aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
-rw-r--r--Documentation/DocBook/kernel-api.tmpl8
-rw-r--r--Documentation/networking/NAPI_HOWTO.txt766
-rw-r--r--Documentation/networking/netdevices.txt12
-rw-r--r--drivers/infiniband/ulp/ipoib/ipoib.h4
-rw-r--r--drivers/infiniband/ulp/ipoib/ipoib_ib.c43
-rw-r--r--drivers/infiniband/ulp/ipoib/ipoib_main.c11
-rw-r--r--drivers/net/8139cp.c41
-rw-r--r--drivers/net/8139too.c48
-rw-r--r--drivers/net/amd8111e.c30
-rw-r--r--drivers/net/amd8111e.h2
-rw-r--r--drivers/net/arm/ep93xx_eth.c72
-rw-r--r--drivers/net/b44.c48
-rw-r--r--drivers/net/b44.h2
-rw-r--r--drivers/net/bnx2.c47
-rw-r--r--drivers/net/bnx2.h2
-rw-r--r--drivers/net/cassini.c40
-rw-r--r--drivers/net/cassini.h2
-rw-r--r--drivers/net/chelsio/common.h1
-rw-r--r--drivers/net/chelsio/cxgb2.c9
-rw-r--r--drivers/net/chelsio/sge.c29
-rw-r--r--drivers/net/chelsio/sge.h2
-rw-r--r--drivers/net/cxgb3/adapter.h22
-rw-r--r--drivers/net/cxgb3/cxgb3_main.c96
-rw-r--r--drivers/net/cxgb3/sge.c170
-rw-r--r--drivers/net/e100.c37
-rw-r--r--drivers/net/e1000/e1000.h1
-rw-r--r--drivers/net/e1000/e1000_main.c45
-rw-r--r--drivers/net/ehea/ehea.h2
-rw-r--r--drivers/net/ehea/ehea_main.c129
-rw-r--r--drivers/net/epic100.c36
-rw-r--r--drivers/net/fec_8xx/fec_8xx.h2
-rw-r--r--drivers/net/fec_8xx/fec_main.c59
-rw-r--r--drivers/net/forcedeth.c69
-rw-r--r--drivers/net/fs_enet/fs_enet-main.c55
-rw-r--r--drivers/net/fs_enet/fs_enet.h1
-rw-r--r--drivers/net/gianfar.c47
-rw-r--r--drivers/net/gianfar.h3
-rw-r--r--drivers/net/ibmveth.c117
-rw-r--r--drivers/net/ibmveth.h1
-rw-r--r--drivers/net/ixgb/ixgb.h1
-rw-r--r--drivers/net/ixgb/ixgb_main.c29
-rw-r--r--drivers/net/ixp2000/ixpdev.c39
-rw-r--r--drivers/net/ixp2000/ixpdev.h2
-rw-r--r--drivers/net/macb.c40
-rw-r--r--drivers/net/macb.h1
-rw-r--r--drivers/net/mv643xx_eth.c48
-rw-r--r--drivers/net/mv643xx_eth.h2
-rw-r--r--drivers/net/myri10ge/myri10ge.c40
-rw-r--r--drivers/net/natsemi.c42
-rw-r--r--drivers/net/netxen/netxen_nic.h1
-rw-r--r--drivers/net/netxen/netxen_nic_main.c39
-rw-r--r--drivers/net/pasemi_mac.c36
-rw-r--r--drivers/net/pasemi_mac.h1
-rw-r--r--drivers/net/pcnet32.c82
-rw-r--r--drivers/net/ps3_gelic_net.c45
-rw-r--r--drivers/net/ps3_gelic_net.h1
-rwxr-xr-xdrivers/net/qla3xxx.c29
-rwxr-xr-xdrivers/net/qla3xxx.h2
-rw-r--r--drivers/net/r8169.c58
-rw-r--r--drivers/net/s2io.c35
-rw-r--r--drivers/net/s2io.h3
-rw-r--r--drivers/net/sb1250-mac.c47
-rw-r--r--drivers/net/sis190.c19
-rw-r--r--drivers/net/skge.c44
-rw-r--r--drivers/net/skge.h1
-rw-r--r--drivers/net/sky2.c94
-rw-r--r--drivers/net/sky2.h1
-rw-r--r--drivers/net/spider_net.c60
-rw-r--r--drivers/net/spider_net.h2
-rw-r--r--drivers/net/starfire.c51
-rw-r--r--drivers/net/sungem.c52
-rw-r--r--drivers/net/sungem.h1
-rw-r--r--drivers/net/tc35815.c49
-rw-r--r--drivers/net/tg3.c61
-rw-r--r--drivers/net/tg3.h1
-rw-r--r--drivers/net/tsi108_eth.c40
-rw-r--r--drivers/net/tulip/interrupt.c54
-rw-r--r--drivers/net/tulip/tulip.h3
-rw-r--r--drivers/net/tulip/tulip_core.c11
-rw-r--r--drivers/net/typhoon.c47
-rw-r--r--drivers/net/ucc_geth.c64
-rw-r--r--drivers/net/ucc_geth.h1
-rw-r--r--drivers/net/via-rhine.c42
-rw-r--r--drivers/net/xen-netfront.c33
-rw-r--r--include/linux/netdevice.h361
-rw-r--r--include/linux/netpoll.h55
-rw-r--r--net/core/dev.c180
-rw-r--r--net/core/net-sysfs.c15
-rw-r--r--net/core/netpoll.c39
-rw-r--r--net/core/rtnetlink.c6
-rw-r--r--net/sched/sch_generic.c12
91 files changed, 1758 insertions, 2375 deletions
diff --git a/Documentation/DocBook/kernel-api.tmpl b/Documentation/DocBook/kernel-api.tmpl
index b886f52a9aac..e5da4f2b7c22 100644
--- a/Documentation/DocBook/kernel-api.tmpl
+++ b/Documentation/DocBook/kernel-api.tmpl
@@ -240,17 +240,23 @@ X!Ilib/string.c
240 <sect1><title>Driver Support</title> 240 <sect1><title>Driver Support</title>
241!Enet/core/dev.c 241!Enet/core/dev.c
242!Enet/ethernet/eth.c 242!Enet/ethernet/eth.c
243!Enet/sched/sch_generic.c
243!Iinclude/linux/etherdevice.h 244!Iinclude/linux/etherdevice.h
245!Iinclude/linux/netdevice.h
246 </sect1>
247 <sect1><title>PHY Support</title>
244!Edrivers/net/phy/phy.c 248!Edrivers/net/phy/phy.c
245!Idrivers/net/phy/phy.c 249!Idrivers/net/phy/phy.c
246!Edrivers/net/phy/phy_device.c 250!Edrivers/net/phy/phy_device.c
247!Idrivers/net/phy/phy_device.c 251!Idrivers/net/phy/phy_device.c
248!Edrivers/net/phy/mdio_bus.c 252!Edrivers/net/phy/mdio_bus.c
249!Idrivers/net/phy/mdio_bus.c 253!Idrivers/net/phy/mdio_bus.c
254 </sect1>
250<!-- FIXME: Removed for now since no structured comments in source 255<!-- FIXME: Removed for now since no structured comments in source
256 <sect1><title>Wireless</title>
251X!Enet/core/wireless.c 257X!Enet/core/wireless.c
252-->
253 </sect1> 258 </sect1>
259-->
254 <sect1><title>Synchronous PPP</title> 260 <sect1><title>Synchronous PPP</title>
255!Edrivers/net/wan/syncppp.c 261!Edrivers/net/wan/syncppp.c
256 </sect1> 262 </sect1>
diff --git a/Documentation/networking/NAPI_HOWTO.txt b/Documentation/networking/NAPI_HOWTO.txt
deleted file mode 100644
index 7907435a661c..000000000000
--- a/Documentation/networking/NAPI_HOWTO.txt
+++ /dev/null
@@ -1,766 +0,0 @@
1HISTORY:
2February 16/2002 -- revision 0.2.1:
3COR typo corrected
4February 10/2002 -- revision 0.2:
5some spell checking ;->
6January 12/2002 -- revision 0.1
7This is still work in progress so may change.
8To keep up to date please watch this space.
9
10Introduction to NAPI
11====================
12
13NAPI is a proven (www.cyberus.ca/~hadi/usenix-paper.tgz) technique
14to improve network performance on Linux. For more details please
15read that paper.
16NAPI provides a "inherent mitigation" which is bound by system capacity
17as can be seen from the following data collected by Robert on Gigabit
18ethernet (e1000):
19
20 Psize Ipps Tput Rxint Txint Done Ndone
21 ---------------------------------------------------------------
22 60 890000 409362 17 27622 7 6823
23 128 758150 464364 21 9301 10 7738
24 256 445632 774646 42 15507 21 12906
25 512 232666 994445 241292 19147 241192 1062
26 1024 119061 1000003 872519 19258 872511 0
27 1440 85193 1000003 946576 19505 946569 0
28
29
30Legend:
31"Ipps" stands for input packets per second.
32"Tput" == packets out of total 1M that made it out.
33"txint" == transmit completion interrupts seen
34"Done" == The number of times that the poll() managed to pull all
35packets out of the rx ring. Note from this that the lower the
36load the more we could clean up the rxring
37"Ndone" == is the converse of "Done". Note again, that the higher
38the load the more times we couldn't clean up the rxring.
39
40Observe that:
41when the NIC receives 890Kpackets/sec only 17 rx interrupts are generated.
42The system cant handle the processing at 1 interrupt/packet at that load level.
43At lower rates on the other hand, rx interrupts go up and therefore the
44interrupt/packet ratio goes up (as observable from that table). So there is
45possibility that under low enough input, you get one poll call for each
46input packet caused by a single interrupt each time. And if the system
47cant handle interrupt per packet ratio of 1, then it will just have to
48chug along ....
49
50
510) Prerequisites:
52==================
53A driver MAY continue using the old 2.4 technique for interfacing
54to the network stack and not benefit from the NAPI changes.
55NAPI additions to the kernel do not break backward compatibility.
56NAPI, however, requires the following features to be available:
57
58A) DMA ring or enough RAM to store packets in software devices.
59
60B) Ability to turn off interrupts or maybe events that send packets up
61the stack.
62
63NAPI processes packet events in what is known as dev->poll() method.
64Typically, only packet receive events are processed in dev->poll().
65The rest of the events MAY be processed by the regular interrupt handler
66to reduce processing latency (justified also because there are not that
67many of them).
68Note, however, NAPI does not enforce that dev->poll() only processes
69receive events.
70Tests with the tulip driver indicated slightly increased latency if
71all of the interrupt handler is moved to dev->poll(). Also MII handling
72gets a little trickier.
73The example used in this document is to move the receive processing only
74to dev->poll(); this is shown with the patch for the tulip driver.
75For an example of code that moves all the interrupt driver to
76dev->poll() look at the ported e1000 code.
77
78There are caveats that might force you to go with moving everything to
79dev->poll(). Different NICs work differently depending on their status/event
80acknowledgement setup.
81There are two types of event register ACK mechanisms.
82 I) what is known as Clear-on-read (COR).
83 when you read the status/event register, it clears everything!
84 The natsemi and sunbmac NICs are known to do this.
85 In this case your only choice is to move all to dev->poll()
86
87 II) Clear-on-write (COW)
88 i) you clear the status by writing a 1 in the bit-location you want.
89 These are the majority of the NICs and work the best with NAPI.
90 Put only receive events in dev->poll(); leave the rest in
91 the old interrupt handler.
92 ii) whatever you write in the status register clears every thing ;->
93 Cant seem to find any supported by Linux which do this. If
94 someone knows such a chip email us please.
95 Move all to dev->poll()
96
97C) Ability to detect new work correctly.
98NAPI works by shutting down event interrupts when there's work and
99turning them on when there's none.
100New packets might show up in the small window while interrupts were being
101re-enabled (refer to appendix 2). A packet might sneak in during the period
102we are enabling interrupts. We only get to know about such a packet when the
103next new packet arrives and generates an interrupt.
104Essentially, there is a small window of opportunity for a race condition
105which for clarity we'll refer to as the "rotting packet".
106
107This is a very important topic and appendix 2 is dedicated for more
108discussion.
109
110Locking rules and environmental guarantees
111==========================================
112
113-Guarantee: Only one CPU at any time can call dev->poll(); this is because
114only one CPU can pick the initial interrupt and hence the initial
115netif_rx_schedule(dev);
116- The core layer invokes devices to send packets in a round robin format.
117This implies receive is totally lockless because of the guarantee that only
118one CPU is executing it.
119- contention can only be the result of some other CPU accessing the rx
120ring. This happens only in close() and suspend() (when these methods
121try to clean the rx ring);
122****guarantee: driver authors need not worry about this; synchronization
123is taken care for them by the top net layer.
124-local interrupts are enabled (if you dont move all to dev->poll()). For
125example link/MII and txcomplete continue functioning just same old way.
126This improves the latency of processing these events. It is also assumed that
127the receive interrupt is the largest cause of noise. Note this might not
128always be true.
129[according to Manfred Spraul, the winbond insists on sending one
130txmitcomplete interrupt for each packet (although this can be mitigated)].
131For these broken drivers, move all to dev->poll().
132
133For the rest of this text, we'll assume that dev->poll() only
134processes receive events.
135
136new methods introduce by NAPI
137=============================
138
139a) netif_rx_schedule(dev)
140Called by an IRQ handler to schedule a poll for device
141
142b) netif_rx_schedule_prep(dev)
143puts the device in a state which allows for it to be added to the
144CPU polling list if it is up and running. You can look at this as
145the first half of netif_rx_schedule(dev) above; the second half
146being c) below.
147
148c) __netif_rx_schedule(dev)
149Add device to the poll list for this CPU; assuming that _prep above
150has already been called and returned 1.
151
152d) netif_rx_reschedule(dev, undo)
153Called to reschedule polling for device specifically for some
154deficient hardware. Read Appendix 2 for more details.
155
156e) netif_rx_complete(dev)
157
158Remove interface from the CPU poll list: it must be in the poll list
159on current cpu. This primitive is called by dev->poll(), when
160it completes its work. The device cannot be out of poll list at this
161call, if it is then clearly it is a BUG(). You'll know ;->
162
163All of the above methods are used below, so keep reading for clarity.
164
165Device driver changes to be made when porting NAPI
166==================================================
167
168Below we describe what kind of changes are required for NAPI to work.
169
1701) introduction of dev->poll() method
171=====================================
172
173This is the method that is invoked by the network core when it requests
174for new packets from the driver. A driver is allowed to send upto
175dev->quota packets by the current CPU before yielding to the network
176subsystem (so other devices can also get opportunity to send to the stack).
177
178dev->poll() prototype looks as follows:
179int my_poll(struct net_device *dev, int *budget)
180
181budget is the remaining number of packets the network subsystem on the
182current CPU can send up the stack before yielding to other system tasks.
183*Each driver is responsible for decrementing budget by the total number of
184packets sent.
185 Total number of packets cannot exceed dev->quota.
186
187dev->poll() method is invoked by the top layer, the driver just sends if it
188can to the stack the packet quantity requested.
189
190more on dev->poll() below after the interrupt changes are explained.
191
1922) registering dev->poll() method
193===================================
194
195dev->poll should be set in the dev->probe() method.
196e.g:
197dev->open = my_open;
198.
199.
200/* two new additions */
201/* first register my poll method */
202dev->poll = my_poll;
203/* next register my weight/quanta; can be overridden in /proc */
204dev->weight = 16;
205.
206.
207dev->stop = my_close;
208
209
210
2113) scheduling dev->poll()
212=============================
213This involves modifying the interrupt handler and the code
214path which takes the packet off the NIC and sends them to the
215stack.
216
217it's important at this point to introduce the classical D Becker
218interrupt processor:
219
220------------------
221static irqreturn_t
222netdevice_interrupt(int irq, void *dev_id, struct pt_regs *regs)
223{
224
225 struct net_device *dev = (struct net_device *)dev_instance;
226 struct my_private *tp = (struct my_private *)dev->priv;
227
228 int work_count = my_work_count;
229 status = read_interrupt_status_reg();
230 if (status == 0)
231 return IRQ_NONE; /* Shared IRQ: not us */
232 if (status == 0xffff)
233 return IRQ_HANDLED; /* Hot unplug */
234 if (status & error)
235 do_some_error_handling()
236
237 do {
238 acknowledge_ints_ASAP();
239
240 if (status & link_interrupt) {
241 spin_lock(&tp->link_lock);
242 do_some_link_stat_stuff();
243 spin_lock(&tp->link_lock);
244 }
245
246 if (status & rx_interrupt) {
247 receive_packets(dev);
248 }
249
250 if (status & rx_nobufs) {
251 make_rx_buffs_avail();
252 }
253
254 if (status & tx_related) {
255 spin_lock(&tp->lock);
256 tx_ring_free(dev);
257 if (tx_died)
258 restart_tx();
259 spin_unlock(&tp->lock);
260 }
261
262 status = read_interrupt_status_reg();
263
264 } while (!(status & error) || more_work_to_be_done);
265 return IRQ_HANDLED;
266}
267
268----------------------------------------------------------------------
269
270We now change this to what is shown below to NAPI-enable it:
271
272----------------------------------------------------------------------
273static irqreturn_t
274netdevice_interrupt(int irq, void *dev_id, struct pt_regs *regs)
275{
276 struct net_device *dev = (struct net_device *)dev_instance;
277 struct my_private *tp = (struct my_private *)dev->priv;
278
279 status = read_interrupt_status_reg();
280 if (status == 0)
281 return IRQ_NONE; /* Shared IRQ: not us */
282 if (status == 0xffff)
283 return IRQ_HANDLED; /* Hot unplug */
284 if (status & error)
285 do_some_error_handling();
286
287 do {
288/************************ start note *********************************/
289 acknowledge_ints_ASAP(); // dont ack rx and rxnobuff here
290/************************ end note *********************************/
291
292 if (status & link_interrupt) {
293 spin_lock(&tp->link_lock);
294 do_some_link_stat_stuff();
295 spin_unlock(&tp->link_lock);
296 }
297/************************ start note *********************************/
298 if (status & rx_interrupt || (status & rx_nobuffs)) {
299 if (netif_rx_schedule_prep(dev)) {
300
301 /* disable interrupts caused
302 * by arriving packets */
303 disable_rx_and_rxnobuff_ints();
304 /* tell system we have work to be done. */
305 __netif_rx_schedule(dev);
306 } else {
307 printk("driver bug! interrupt while in poll\n");
308 /* FIX by disabling interrupts */
309 disable_rx_and_rxnobuff_ints();
310 }
311 }
312/************************ end note note *********************************/
313
314 if (status & tx_related) {
315 spin_lock(&tp->lock);
316 tx_ring_free(dev);
317
318 if (tx_died)
319 restart_tx();
320 spin_unlock(&tp->lock);
321 }
322
323 status = read_interrupt_status_reg();
324
325/************************ start note *********************************/
326 } while (!(status & error) || more_work_to_be_done(status));
327/************************ end note note *********************************/
328 return IRQ_HANDLED;
329}
330
331---------------------------------------------------------------------
332
333
334We note several things from above:
335
336I) Any interrupt source which is caused by arriving packets is now
337turned off when it occurs. Depending on the hardware, there could be
338several reasons that arriving packets would cause interrupts; these are the
339interrupt sources we wish to avoid. The two common ones are a) a packet
340arriving (rxint) b) a packet arriving and finding no DMA buffers available
341(rxnobuff) .
342This means also acknowledge_ints_ASAP() will not clear the status
343register for those two items above; clearing is done in the place where
344proper work is done within NAPI; at the poll() and refill_rx_ring()
345discussed further below.
346netif_rx_schedule_prep() returns 1 if device is in running state and
347gets successfully added to the core poll list. If we get a zero value
348we can _almost_ assume are already added to the list (instead of not running.
349Logic based on the fact that you shouldn't get interrupt if not running)
350We rectify this by disabling rx and rxnobuf interrupts.
351
352II) that receive_packets(dev) and make_rx_buffs_avail() may have disappeared.
353These functionalities are still around actually......
354
355infact, receive_packets(dev) is very close to my_poll() and
356make_rx_buffs_avail() is invoked from my_poll()
357
3584) converting receive_packets() to dev->poll()
359===============================================
360
361We need to convert the classical D Becker receive_packets(dev) to my_poll()
362
363First the typical receive_packets() below:
364-------------------------------------------------------------------
365
366/* this is called by interrupt handler */
367static void receive_packets (struct net_device *dev)
368{
369
370 struct my_private *tp = (struct my_private *)dev->priv;
371 rx_ring = tp->rx_ring;
372 cur_rx = tp->cur_rx;
373 int entry = cur_rx % RX_RING_SIZE;
374 int received = 0;
375 int rx_work_limit = tp->dirty_rx + RX_RING_SIZE - tp->cur_rx;
376
377 while (rx_ring_not_empty) {
378 u32 rx_status;
379 unsigned int rx_size;
380 unsigned int pkt_size;
381 struct sk_buff *skb;
382 /* read size+status of next frame from DMA ring buffer */
383 /* the number 16 and 4 are just examples */
384 rx_status = le32_to_cpu (*(u32 *) (rx_ring + ring_offset));
385 rx_size = rx_status >> 16;
386 pkt_size = rx_size - 4;
387
388 /* process errors */
389 if ((rx_size > (MAX_ETH_FRAME_SIZE+4)) ||
390 (!(rx_status & RxStatusOK))) {
391 netdrv_rx_err (rx_status, dev, tp, ioaddr);
392 return;
393 }
394
395 if (--rx_work_limit < 0)
396 break;
397
398 /* grab a skb */
399 skb = dev_alloc_skb (pkt_size + 2);
400 if (skb) {
401 .
402 .
403 netif_rx (skb);
404 .
405 .
406 } else { /* OOM */
407 /*seems very driver specific ... some just pass
408 whatever is on the ring already. */
409 }
410
411 /* move to the next skb on the ring */
412 entry = (++tp->cur_rx) % RX_RING_SIZE;
413 received++ ;
414
415 }
416
417 /* store current ring pointer state */
418 tp->cur_rx = cur_rx;
419
420 /* Refill the Rx ring buffers if they are needed */
421 refill_rx_ring();
422 .
423 .
424
425}
426-------------------------------------------------------------------
427We change it to a new one below; note the additional parameter in
428the call.
429
430-------------------------------------------------------------------
431
432/* this is called by the network core */
433static int my_poll (struct net_device *dev, int *budget)
434{
435
436 struct my_private *tp = (struct my_private *)dev->priv;
437 rx_ring = tp->rx_ring;
438 cur_rx = tp->cur_rx;
439 int entry = cur_rx % RX_BUF_LEN;
440 /* maximum packets to send to the stack */
441/************************ note note *********************************/
442 int rx_work_limit = dev->quota;
443
444/************************ end note note *********************************/
445 do { // outer beginning loop starts here
446
447 clear_rx_status_register_bit();
448
449 while (rx_ring_not_empty) {
450 u32 rx_status;
451 unsigned int rx_size;
452 unsigned int pkt_size;
453 struct sk_buff *skb;
454 /* read size+status of next frame from DMA ring buffer */
455 /* the number 16 and 4 are just examples */
456 rx_status = le32_to_cpu (*(u32 *) (rx_ring + ring_offset));
457 rx_size = rx_status >> 16;
458 pkt_size = rx_size - 4;
459
460 /* process errors */
461 if ((rx_size > (MAX_ETH_FRAME_SIZE+4)) ||
462 (!(rx_status & RxStatusOK))) {
463 netdrv_rx_err (rx_status, dev, tp, ioaddr);
464 return 1;
465 }
466
467/************************ note note *********************************/
468 if (--rx_work_limit < 0) { /* we got packets, but no quota */
469 /* store current ring pointer state */
470 tp->cur_rx = cur_rx;
471
472 /* Refill the Rx ring buffers if they are needed */
473 refill_rx_ring(dev);
474 goto not_done;
475 }
476/********************** end note **********************************/
477
478 /* grab a skb */
479 skb = dev_alloc_skb (pkt_size + 2);
480 if (skb) {
481 .
482 .
483/************************ note note *********************************/
484 netif_receive_skb (skb);
485/********************** end note **********************************/
486 .
487 .
488 } else { /* OOM */
489 /*seems very driver specific ... common is just pass
490 whatever is on the ring already. */
491 }
492
493 /* move to the next skb on the ring */
494 entry = (++tp->cur_rx) % RX_RING_SIZE;
495 received++ ;
496
497 }
498
499 /* store current ring pointer state */
500 tp->cur_rx = cur_rx;
501
502 /* Refill the Rx ring buffers if they are needed */
503 refill_rx_ring(dev);
504
505 /* no packets on ring; but new ones can arrive since we last
506 checked */
507 status = read_interrupt_status_reg();
508 if (rx status is not set) {
509 /* If something arrives in this narrow window,
510 an interrupt will be generated */
511 goto done;
512 }
513 /* done! at least that's what it looks like ;->
514 if new packets came in after our last check on status bits
515 they'll be caught by the while check and we go back and clear them
516 since we havent exceeded our quota */
517 } while (rx_status_is_set);
518
519done:
520
521/************************ note note *********************************/
522 dev->quota -= received;
523 *budget -= received;
524
525 /* If RX ring is not full we are out of memory. */
526 if (tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
527 goto oom;
528
529 /* we are happy/done, no more packets on ring; put us back
530 to where we can start processing interrupts again */
531 netif_rx_complete(dev);
532 enable_rx_and_rxnobuf_ints();
533
534 /* The last op happens after poll completion. Which means the following:
535 * 1. it can race with disabling irqs in irq handler (which are done to
536 * schedule polls)
537 * 2. it can race with dis/enabling irqs in other poll threads
538 * 3. if an irq raised after the beginning of the outer beginning
539 * loop (marked in the code above), it will be immediately
540 * triggered here.
541 *
542 * Summarizing: the logic may result in some redundant irqs both
543 * due to races in masking and due to too late acking of already
544 * processed irqs. The good news: no events are ever lost.
545 */
546
547 return 0; /* done */
548
549not_done:
550 if (tp->cur_rx - tp->dirty_rx > RX_RING_SIZE/2 ||
551 tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
552 refill_rx_ring(dev);
553
554 if (!received) {
555 printk("received==0\n");
556 received = 1;
557 }
558 dev->quota -= received;
559 *budget -= received;
560 return 1; /* not_done */
561
562oom:
563 /* Start timer, stop polling, but do not enable rx interrupts. */
564 start_poll_timer(dev);
565 return 0; /* we'll take it from here so tell core "done"*/
566
567/************************ End note note *********************************/
568}
569-------------------------------------------------------------------
570
571From above we note that:
5720) rx_work_limit = dev->quota
5731) refill_rx_ring() is in charge of clearing the bit for rxnobuff when
574it does the work.
5752) We have a done and not_done state.
5763) instead of netif_rx() we call netif_receive_skb() to pass the skb.
5774) we have a new way of handling oom condition
5785) A new outer for (;;) loop has been added. This serves the purpose of
579ensuring that if a new packet has come in, after we are all set and done,
580and we have not exceeded our quota that we continue sending packets up.
581
582
583-----------------------------------------------------------
584Poll timer code will need to do the following:
585
586a)
587
588 if (tp->cur_rx - tp->dirty_rx > RX_RING_SIZE/2 ||
589 tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
590 refill_rx_ring(dev);
591
592 /* If RX ring is not full we are still out of memory.
593 Restart the timer again. Else we re-add ourselves
594 to the master poll list.
595 */
596
597 if (tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
598 restart_timer();
599
600 else netif_rx_schedule(dev); /* we are back on the poll list */
601
6025) dev->close() and dev->suspend() issues
603==========================================
604The driver writer needn't worry about this; the top net layer takes
605care of it.
606
6076) Adding new Stats to /proc
608=============================
609In order to debug some of the new features, we introduce new stats
610that need to be collected.
611TODO: Fill this later.
612
613APPENDIX 1: discussion on using ethernet HW FC
614==============================================
615Most chips with FC only send a pause packet when they run out of Rx buffers.
616Since packets are pulled off the DMA ring by a softirq in NAPI,
617if the system is slow in grabbing them and we have a high input
618rate (faster than the system's capacity to remove packets), then theoretically
619there will only be one rx interrupt for all packets during a given packetstorm.
620Under low load, we might have a single interrupt per packet.
621FC should be programmed to apply in the case when the system cant pull out
622packets fast enough i.e send a pause only when you run out of rx buffers.
623Note FC in itself is a good solution but we have found it to not be
624much of a commodity feature (both in NICs and switches) and hence falls
625under the same category as using NIC based mitigation. Also, experiments
626indicate that it's much harder to resolve the resource allocation
627issue (aka lazy receiving that NAPI offers) and hence quantify its usefulness
628proved harder. In any case, FC works even better with NAPI but is not
629necessary.
630
631
632APPENDIX 2: the "rotting packet" race-window avoidance scheme
633=============================================================
634
635There are two types of associations seen here
636
6371) status/int which honors level triggered IRQ
638
639If a status bit for receive or rxnobuff is set and the corresponding
640interrupt-enable bit is not on, then no interrupts will be generated. However,
641as soon as the "interrupt-enable" bit is unmasked, an immediate interrupt is
642generated. [assuming the status bit was not turned off].
643Generally the concept of level triggered IRQs in association with a status and
644interrupt-enable CSR register set is used to avoid the race.
645
646If we take the example of the tulip:
647"pending work" is indicated by the status bit(CSR5 in tulip).
648the corresponding interrupt bit (CSR7 in tulip) might be turned off (but
649the CSR5 will continue to be turned on with new packet arrivals even if
650we clear it the first time)
651Very important is the fact that if we turn on the interrupt bit on when
652status is set that an immediate irq is triggered.
653
654If we cleared the rx ring and proclaimed there was "no more work
655to be done" and then went on to do a few other things; then when we enable
656interrupts, there is a possibility that a new packet might sneak in during
657this phase. It helps to look at the pseudo code for the tulip poll
658routine:
659
660--------------------------
661 do {
662 ACK;
663 while (ring_is_not_empty()) {
664 work-work-work
665 if quota is exceeded: exit, no touching irq status/mask
666 }
667 /* No packets, but new can arrive while we are doing this*/
668 CSR5 := read
669 if (CSR5 is not set) {
670 /* If something arrives in this narrow window here,
671 * where the comments are ;-> irq will be generated */
672 unmask irqs;
673 exit poll;
674 }
675 } while (rx_status_is_set);
676------------------------
677
678CSR5 bit of interest is only the rx status.
679If you look at the last if statement:
680you just finished grabbing all the packets from the rx ring .. you check if
681status bit says there are more packets just in ... it says none; you then
682enable rx interrupts again; if a new packet just came in during this check,
683we are counting that CSR5 will be set in that small window of opportunity
684and that by re-enabling interrupts, we would actually trigger an interrupt
685to register the new packet for processing.
686
687[The above description nay be very verbose, if you have better wording
688that will make this more understandable, please suggest it.]
689
6902) non-capable hardware
691
692These do not generally respect level triggered IRQs. Normally,
693irqs may be lost while being masked and the only way to leave poll is to do
694a double check for new input after netif_rx_complete() is invoked
695and re-enable polling (after seeing this new input).
696
697Sample code:
698
699---------
700 .
701 .
702restart_poll:
703 while (ring_is_not_empty()) {
704 work-work-work
705 if quota is exceeded: exit, not touching irq status/mask
706 }
707 .
708 .
709 .
710 enable_rx_interrupts()
711 netif_rx_complete(dev);
712 if (ring_has_new_packet() && netif_rx_reschedule(dev, received)) {
713 disable_rx_and_rxnobufs()
714 goto restart_poll
715 } while (rx_status_is_set);
716---------
717
718Basically netif_rx_complete() removes us from the poll list, but because a
719new packet which will never be caught due to the possibility of a race
720might come in, we attempt to re-add ourselves to the poll list.
721
722
723
724
725APPENDIX 3: Scheduling issues.
726==============================
727As seen NAPI moves processing to softirq level. Linux uses the ksoftirqd as the
728general solution to schedule softirq's to run before next interrupt and by putting
729them under scheduler control. Also this prevents consecutive softirq's from
730monopolize the CPU. This also have the effect that the priority of ksoftirq needs
731to be considered when running very CPU-intensive applications and networking to
732get the proper balance of softirq/user balance. Increasing ksoftirq priority to 0
733(eventually more) is reported cure problems with low network performance at high
734CPU load.
735
736Most used processes in a GIGE router:
737USER PID %CPU %MEM SIZE RSS TTY STAT START TIME COMMAND
738root 3 0.2 0.0 0 0 ? RWN Aug 15 602:00 (ksoftirqd_CPU0)
739root 232 0.0 7.9 41400 40884 ? S Aug 15 74:12 gated
740
741--------------------------------------------------------------------
742
743relevant sites:
744==================
745ftp://robur.slu.se/pub/Linux/net-development/NAPI/
746
747
748--------------------------------------------------------------------
749TODO: Write net-skeleton.c driver.
750-------------------------------------------------------------
751
752Authors:
753========
754Alexey Kuznetsov <kuznet@ms2.inr.ac.ru>
755Jamal Hadi Salim <hadi@cyberus.ca>
756Robert Olsson <Robert.Olsson@data.slu.se>
757
758Acknowledgements:
759================
760People who made this document better:
761
762Lennert Buytenhek <buytenh@gnu.org>
763Andrew Morton <akpm@zip.com.au>
764Manfred Spraul <manfred@colorfullife.com>
765Donald Becker <becker@scyld.com>
766Jeff Garzik <jgarzik@pobox.com>
diff --git a/Documentation/networking/netdevices.txt b/Documentation/networking/netdevices.txt
index 37869295fc70..9f7be9b7785e 100644
--- a/Documentation/networking/netdevices.txt
+++ b/Documentation/networking/netdevices.txt
@@ -95,9 +95,13 @@ dev->set_multicast_list:
95 Synchronization: netif_tx_lock spinlock. 95 Synchronization: netif_tx_lock spinlock.
96 Context: BHs disabled 96 Context: BHs disabled
97 97
98dev->poll: 98struct napi_struct synchronization rules
99 Synchronization: __LINK_STATE_RX_SCHED bit in dev->state. See 99========================================
100 dev_close code and comments in net/core/dev.c for more info. 100napi->poll:
101 Synchronization: NAPI_STATE_SCHED bit in napi->state. Device
102 driver's dev->close method will invoke napi_disable() on
103 all NAPI instances which will do a sleeping poll on the
104 NAPI_STATE_SCHED napi->state bit, waiting for all pending
105 NAPI activity to cease.
101 Context: softirq 106 Context: softirq
102 will be called with interrupts disabled by netconsole. 107 will be called with interrupts disabled by netconsole.
103
diff --git a/drivers/infiniband/ulp/ipoib/ipoib.h b/drivers/infiniband/ulp/ipoib/ipoib.h
index 285c143115cc..35f3ca42bd60 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib.h
+++ b/drivers/infiniband/ulp/ipoib/ipoib.h
@@ -228,6 +228,8 @@ struct ipoib_dev_priv {
228 228
229 struct net_device *dev; 229 struct net_device *dev;
230 230
231 struct napi_struct napi;
232
231 unsigned long flags; 233 unsigned long flags;
232 234
233 struct mutex mcast_mutex; 235 struct mutex mcast_mutex;
@@ -351,7 +353,7 @@ extern struct workqueue_struct *ipoib_workqueue;
351 353
352/* functions */ 354/* functions */
353 355
354int ipoib_poll(struct net_device *dev, int *budget); 356int ipoib_poll(struct napi_struct *napi, int budget);
355void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr); 357void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr);
356 358
357struct ipoib_ah *ipoib_create_ah(struct net_device *dev, 359struct ipoib_ah *ipoib_create_ah(struct net_device *dev,
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_ib.c b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
index 10944888cffd..481e4b6bd949 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_ib.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_ib.c
@@ -281,63 +281,58 @@ static void ipoib_ib_handle_tx_wc(struct net_device *dev, struct ib_wc *wc)
281 wc->status, wr_id, wc->vendor_err); 281 wc->status, wr_id, wc->vendor_err);
282} 282}
283 283
284int ipoib_poll(struct net_device *dev, int *budget) 284int ipoib_poll(struct napi_struct *napi, int budget)
285{ 285{
286 struct ipoib_dev_priv *priv = netdev_priv(dev); 286 struct ipoib_dev_priv *priv = container_of(napi, struct ipoib_dev_priv, napi);
287 int max = min(*budget, dev->quota); 287 struct net_device *dev = priv->dev;
288 int done; 288 int done;
289 int t; 289 int t;
290 int empty;
291 int n, i; 290 int n, i;
292 291
293 done = 0; 292 done = 0;
294 empty = 0;
295 293
296 while (max) { 294poll_more:
295 while (done < budget) {
296 int max = (budget - done);
297
297 t = min(IPOIB_NUM_WC, max); 298 t = min(IPOIB_NUM_WC, max);
298 n = ib_poll_cq(priv->cq, t, priv->ibwc); 299 n = ib_poll_cq(priv->cq, t, priv->ibwc);
299 300
300 for (i = 0; i < n; ++i) { 301 for (i = 0; i < n; i++) {
301 struct ib_wc *wc = priv->ibwc + i; 302 struct ib_wc *wc = priv->ibwc + i;
302 303
303 if (wc->wr_id & IPOIB_CM_OP_SRQ) { 304 if (wc->wr_id & IPOIB_CM_OP_SRQ) {
304 ++done; 305 ++done;
305 --max;
306 ipoib_cm_handle_rx_wc(dev, wc); 306 ipoib_cm_handle_rx_wc(dev, wc);
307 } else if (wc->wr_id & IPOIB_OP_RECV) { 307 } else if (wc->wr_id & IPOIB_OP_RECV) {
308 ++done; 308 ++done;
309 --max;
310 ipoib_ib_handle_rx_wc(dev, wc); 309 ipoib_ib_handle_rx_wc(dev, wc);
311 } else 310 } else
312 ipoib_ib_handle_tx_wc(dev, wc); 311 ipoib_ib_handle_tx_wc(dev, wc);
313 } 312 }
314 313
315 if (n != t) { 314 if (n != t)
316 empty = 1;
317 break; 315 break;
318 }
319 } 316 }
320 317
321 dev->quota -= done; 318 if (done < budget) {
322 *budget -= done; 319 netif_rx_complete(dev, napi);
323
324 if (empty) {
325 netif_rx_complete(dev);
326 if (unlikely(ib_req_notify_cq(priv->cq, 320 if (unlikely(ib_req_notify_cq(priv->cq,
327 IB_CQ_NEXT_COMP | 321 IB_CQ_NEXT_COMP |
328 IB_CQ_REPORT_MISSED_EVENTS)) && 322 IB_CQ_REPORT_MISSED_EVENTS)) &&
329 netif_rx_reschedule(dev, 0)) 323 netif_rx_reschedule(dev, napi))
330 return 1; 324 goto poll_more;
331
332 return 0;
333 } 325 }
334 326
335 return 1; 327 return done;
336} 328}
337 329
338void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr) 330void ipoib_ib_completion(struct ib_cq *cq, void *dev_ptr)
339{ 331{
340 netif_rx_schedule(dev_ptr); 332 struct net_device *dev = dev_ptr;
333 struct ipoib_dev_priv *priv = netdev_priv(dev);
334
335 netif_rx_schedule(dev, &priv->napi);
341} 336}
342 337
343static inline int post_send(struct ipoib_dev_priv *priv, 338static inline int post_send(struct ipoib_dev_priv *priv,
@@ -577,7 +572,6 @@ int ipoib_ib_dev_stop(struct net_device *dev, int flush)
577 int i; 572 int i;
578 573
579 clear_bit(IPOIB_FLAG_INITIALIZED, &priv->flags); 574 clear_bit(IPOIB_FLAG_INITIALIZED, &priv->flags);
580 netif_poll_disable(dev);
581 575
582 ipoib_cm_dev_stop(dev); 576 ipoib_cm_dev_stop(dev);
583 577
@@ -660,7 +654,6 @@ timeout:
660 msleep(1); 654 msleep(1);
661 } 655 }
662 656
663 netif_poll_enable(dev);
664 ib_req_notify_cq(priv->cq, IB_CQ_NEXT_COMP); 657 ib_req_notify_cq(priv->cq, IB_CQ_NEXT_COMP);
665 658
666 return 0; 659 return 0;
diff --git a/drivers/infiniband/ulp/ipoib/ipoib_main.c b/drivers/infiniband/ulp/ipoib/ipoib_main.c
index 894b1dcdf3eb..a59ff07ec3cd 100644
--- a/drivers/infiniband/ulp/ipoib/ipoib_main.c
+++ b/drivers/infiniband/ulp/ipoib/ipoib_main.c
@@ -98,16 +98,20 @@ int ipoib_open(struct net_device *dev)
98 98
99 ipoib_dbg(priv, "bringing up interface\n"); 99 ipoib_dbg(priv, "bringing up interface\n");
100 100
101 napi_enable(&priv->napi);
101 set_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags); 102 set_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags);
102 103
103 if (ipoib_pkey_dev_delay_open(dev)) 104 if (ipoib_pkey_dev_delay_open(dev))
104 return 0; 105 return 0;
105 106
106 if (ipoib_ib_dev_open(dev)) 107 if (ipoib_ib_dev_open(dev)) {
108 napi_disable(&priv->napi);
107 return -EINVAL; 109 return -EINVAL;
110 }
108 111
109 if (ipoib_ib_dev_up(dev)) { 112 if (ipoib_ib_dev_up(dev)) {
110 ipoib_ib_dev_stop(dev, 1); 113 ipoib_ib_dev_stop(dev, 1);
114 napi_disable(&priv->napi);
111 return -EINVAL; 115 return -EINVAL;
112 } 116 }
113 117
@@ -140,6 +144,7 @@ static int ipoib_stop(struct net_device *dev)
140 ipoib_dbg(priv, "stopping interface\n"); 144 ipoib_dbg(priv, "stopping interface\n");
141 145
142 clear_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags); 146 clear_bit(IPOIB_FLAG_ADMIN_UP, &priv->flags);
147 napi_disable(&priv->napi);
143 148
144 netif_stop_queue(dev); 149 netif_stop_queue(dev);
145 150
@@ -948,8 +953,8 @@ static void ipoib_setup(struct net_device *dev)
948 dev->hard_header = ipoib_hard_header; 953 dev->hard_header = ipoib_hard_header;
949 dev->set_multicast_list = ipoib_set_mcast_list; 954 dev->set_multicast_list = ipoib_set_mcast_list;
950 dev->neigh_setup = ipoib_neigh_setup_dev; 955 dev->neigh_setup = ipoib_neigh_setup_dev;
951 dev->poll = ipoib_poll; 956
952 dev->weight = 100; 957 netif_napi_add(dev, &priv->napi, ipoib_poll, 100);
953 958
954 dev->watchdog_timeo = HZ; 959 dev->watchdog_timeo = HZ;
955 960
diff --git a/drivers/net/8139cp.c b/drivers/net/8139cp.c
index a79f28c7a100..7f18ca23d9f8 100644
--- a/drivers/net/8139cp.c
+++ b/drivers/net/8139cp.c
@@ -334,6 +334,8 @@ struct cp_private {
334 spinlock_t lock; 334 spinlock_t lock;
335 u32 msg_enable; 335 u32 msg_enable;
336 336
337 struct napi_struct napi;
338
337 struct pci_dev *pdev; 339 struct pci_dev *pdev;
338 u32 rx_config; 340 u32 rx_config;
339 u16 cpcmd; 341 u16 cpcmd;
@@ -501,12 +503,12 @@ static inline unsigned int cp_rx_csum_ok (u32 status)
501 return 0; 503 return 0;
502} 504}
503 505
504static int cp_rx_poll (struct net_device *dev, int *budget) 506static int cp_rx_poll(struct napi_struct *napi, int budget)
505{ 507{
506 struct cp_private *cp = netdev_priv(dev); 508 struct cp_private *cp = container_of(napi, struct cp_private, napi);
507 unsigned rx_tail = cp->rx_tail; 509 struct net_device *dev = cp->dev;
508 unsigned rx_work = dev->quota; 510 unsigned int rx_tail = cp->rx_tail;
509 unsigned rx; 511 int rx;
510 512
511rx_status_loop: 513rx_status_loop:
512 rx = 0; 514 rx = 0;
@@ -588,33 +590,28 @@ rx_next:
588 desc->opts1 = cpu_to_le32(DescOwn | cp->rx_buf_sz); 590 desc->opts1 = cpu_to_le32(DescOwn | cp->rx_buf_sz);
589 rx_tail = NEXT_RX(rx_tail); 591 rx_tail = NEXT_RX(rx_tail);
590 592
591 if (!rx_work--) 593 if (rx >= budget)
592 break; 594 break;
593 } 595 }
594 596
595 cp->rx_tail = rx_tail; 597 cp->rx_tail = rx_tail;
596 598
597 dev->quota -= rx;
598 *budget -= rx;
599
600 /* if we did not reach work limit, then we're done with 599 /* if we did not reach work limit, then we're done with
601 * this round of polling 600 * this round of polling
602 */ 601 */
603 if (rx_work) { 602 if (rx < budget) {
604 unsigned long flags; 603 unsigned long flags;
605 604
606 if (cpr16(IntrStatus) & cp_rx_intr_mask) 605 if (cpr16(IntrStatus) & cp_rx_intr_mask)
607 goto rx_status_loop; 606 goto rx_status_loop;
608 607
609 local_irq_save(flags); 608 spin_lock_irqsave(&cp->lock, flags);
610 cpw16_f(IntrMask, cp_intr_mask); 609 cpw16_f(IntrMask, cp_intr_mask);
611 __netif_rx_complete(dev); 610 __netif_rx_complete(dev, napi);
612 local_irq_restore(flags); 611 spin_unlock_irqrestore(&cp->lock, flags);
613
614 return 0; /* done */
615 } 612 }
616 613
617 return 1; /* not done */ 614 return rx;
618} 615}
619 616
620static irqreturn_t cp_interrupt (int irq, void *dev_instance) 617static irqreturn_t cp_interrupt (int irq, void *dev_instance)
@@ -647,9 +644,9 @@ static irqreturn_t cp_interrupt (int irq, void *dev_instance)
647 } 644 }
648 645
649 if (status & (RxOK | RxErr | RxEmpty | RxFIFOOvr)) 646 if (status & (RxOK | RxErr | RxEmpty | RxFIFOOvr))
650 if (netif_rx_schedule_prep(dev)) { 647 if (netif_rx_schedule_prep(dev, &cp->napi)) {
651 cpw16_f(IntrMask, cp_norx_intr_mask); 648 cpw16_f(IntrMask, cp_norx_intr_mask);
652 __netif_rx_schedule(dev); 649 __netif_rx_schedule(dev, &cp->napi);
653 } 650 }
654 651
655 if (status & (TxOK | TxErr | TxEmpty | SWInt)) 652 if (status & (TxOK | TxErr | TxEmpty | SWInt))
@@ -1175,6 +1172,8 @@ static int cp_open (struct net_device *dev)
1175 if (rc) 1172 if (rc)
1176 return rc; 1173 return rc;
1177 1174
1175 napi_enable(&cp->napi);
1176
1178 cp_init_hw(cp); 1177 cp_init_hw(cp);
1179 1178
1180 rc = request_irq(dev->irq, cp_interrupt, IRQF_SHARED, dev->name, dev); 1179 rc = request_irq(dev->irq, cp_interrupt, IRQF_SHARED, dev->name, dev);
@@ -1188,6 +1187,7 @@ static int cp_open (struct net_device *dev)
1188 return 0; 1187 return 0;
1189 1188
1190err_out_hw: 1189err_out_hw:
1190 napi_disable(&cp->napi);
1191 cp_stop_hw(cp); 1191 cp_stop_hw(cp);
1192 cp_free_rings(cp); 1192 cp_free_rings(cp);
1193 return rc; 1193 return rc;
@@ -1198,6 +1198,8 @@ static int cp_close (struct net_device *dev)
1198 struct cp_private *cp = netdev_priv(dev); 1198 struct cp_private *cp = netdev_priv(dev);
1199 unsigned long flags; 1199 unsigned long flags;
1200 1200
1201 napi_disable(&cp->napi);
1202
1201 if (netif_msg_ifdown(cp)) 1203 if (netif_msg_ifdown(cp))
1202 printk(KERN_DEBUG "%s: disabling interface\n", dev->name); 1204 printk(KERN_DEBUG "%s: disabling interface\n", dev->name);
1203 1205
@@ -1933,11 +1935,10 @@ static int cp_init_one (struct pci_dev *pdev, const struct pci_device_id *ent)
1933 dev->hard_start_xmit = cp_start_xmit; 1935 dev->hard_start_xmit = cp_start_xmit;
1934 dev->get_stats = cp_get_stats; 1936 dev->get_stats = cp_get_stats;
1935 dev->do_ioctl = cp_ioctl; 1937 dev->do_ioctl = cp_ioctl;
1936 dev->poll = cp_rx_poll;
1937#ifdef CONFIG_NET_POLL_CONTROLLER 1938#ifdef CONFIG_NET_POLL_CONTROLLER
1938 dev->poll_controller = cp_poll_controller; 1939 dev->poll_controller = cp_poll_controller;
1939#endif 1940#endif
1940 dev->weight = 16; /* arbitrary? from NAPI_HOWTO.txt. */ 1941 netif_napi_add(dev, &cp->napi, cp_rx_poll, 16);
1941#ifdef BROKEN 1942#ifdef BROKEN
1942 dev->change_mtu = cp_change_mtu; 1943 dev->change_mtu = cp_change_mtu;
1943#endif 1944#endif
diff --git a/drivers/net/8139too.c b/drivers/net/8139too.c
index f4e4298d24b9..20af6baecfcb 100644
--- a/drivers/net/8139too.c
+++ b/drivers/net/8139too.c
@@ -573,6 +573,8 @@ struct rtl8139_private {
573 int drv_flags; 573 int drv_flags;
574 struct pci_dev *pci_dev; 574 struct pci_dev *pci_dev;
575 u32 msg_enable; 575 u32 msg_enable;
576 struct napi_struct napi;
577 struct net_device *dev;
576 struct net_device_stats stats; 578 struct net_device_stats stats;
577 unsigned char *rx_ring; 579 unsigned char *rx_ring;
578 unsigned int cur_rx; /* Index into the Rx buffer of next Rx pkt. */ 580 unsigned int cur_rx; /* Index into the Rx buffer of next Rx pkt. */
@@ -625,10 +627,10 @@ static void rtl8139_tx_timeout (struct net_device *dev);
625static void rtl8139_init_ring (struct net_device *dev); 627static void rtl8139_init_ring (struct net_device *dev);
626static int rtl8139_start_xmit (struct sk_buff *skb, 628static int rtl8139_start_xmit (struct sk_buff *skb,
627 struct net_device *dev); 629 struct net_device *dev);
628static int rtl8139_poll(struct net_device *dev, int *budget);
629#ifdef CONFIG_NET_POLL_CONTROLLER 630#ifdef CONFIG_NET_POLL_CONTROLLER
630static void rtl8139_poll_controller(struct net_device *dev); 631static void rtl8139_poll_controller(struct net_device *dev);
631#endif 632#endif
633static int rtl8139_poll(struct napi_struct *napi, int budget);
632static irqreturn_t rtl8139_interrupt (int irq, void *dev_instance); 634static irqreturn_t rtl8139_interrupt (int irq, void *dev_instance);
633static int rtl8139_close (struct net_device *dev); 635static int rtl8139_close (struct net_device *dev);
634static int netdev_ioctl (struct net_device *dev, struct ifreq *rq, int cmd); 636static int netdev_ioctl (struct net_device *dev, struct ifreq *rq, int cmd);
@@ -963,6 +965,7 @@ static int __devinit rtl8139_init_one (struct pci_dev *pdev,
963 965
964 assert (dev != NULL); 966 assert (dev != NULL);
965 tp = netdev_priv(dev); 967 tp = netdev_priv(dev);
968 tp->dev = dev;
966 969
967 ioaddr = tp->mmio_addr; 970 ioaddr = tp->mmio_addr;
968 assert (ioaddr != NULL); 971 assert (ioaddr != NULL);
@@ -976,8 +979,7 @@ static int __devinit rtl8139_init_one (struct pci_dev *pdev,
976 /* The Rtl8139-specific entries in the device structure. */ 979 /* The Rtl8139-specific entries in the device structure. */
977 dev->open = rtl8139_open; 980 dev->open = rtl8139_open;
978 dev->hard_start_xmit = rtl8139_start_xmit; 981 dev->hard_start_xmit = rtl8139_start_xmit;
979 dev->poll = rtl8139_poll; 982 netif_napi_add(dev, &tp->napi, rtl8139_poll, 64);
980 dev->weight = 64;
981 dev->stop = rtl8139_close; 983 dev->stop = rtl8139_close;
982 dev->get_stats = rtl8139_get_stats; 984 dev->get_stats = rtl8139_get_stats;
983 dev->set_multicast_list = rtl8139_set_rx_mode; 985 dev->set_multicast_list = rtl8139_set_rx_mode;
@@ -1332,6 +1334,8 @@ static int rtl8139_open (struct net_device *dev)
1332 1334
1333 } 1335 }
1334 1336
1337 napi_enable(&tp->napi);
1338
1335 tp->mii.full_duplex = tp->mii.force_media; 1339 tp->mii.full_duplex = tp->mii.force_media;
1336 tp->tx_flag = (TX_FIFO_THRESH << 11) & 0x003f0000; 1340 tp->tx_flag = (TX_FIFO_THRESH << 11) & 0x003f0000;
1337 1341
@@ -2103,39 +2107,32 @@ static void rtl8139_weird_interrupt (struct net_device *dev,
2103 } 2107 }
2104} 2108}
2105 2109
2106static int rtl8139_poll(struct net_device *dev, int *budget) 2110static int rtl8139_poll(struct napi_struct *napi, int budget)
2107{ 2111{
2108 struct rtl8139_private *tp = netdev_priv(dev); 2112 struct rtl8139_private *tp = container_of(napi, struct rtl8139_private, napi);
2113 struct net_device *dev = tp->dev;
2109 void __iomem *ioaddr = tp->mmio_addr; 2114 void __iomem *ioaddr = tp->mmio_addr;
2110 int orig_budget = min(*budget, dev->quota); 2115 int work_done;
2111 int done = 1;
2112 2116
2113 spin_lock(&tp->rx_lock); 2117 spin_lock(&tp->rx_lock);
2114 if (likely(RTL_R16(IntrStatus) & RxAckBits)) { 2118 work_done = 0;
2115 int work_done; 2119 if (likely(RTL_R16(IntrStatus) & RxAckBits))
2116 2120 work_done += rtl8139_rx(dev, tp, budget);
2117 work_done = rtl8139_rx(dev, tp, orig_budget);
2118 if (likely(work_done > 0)) {
2119 *budget -= work_done;
2120 dev->quota -= work_done;
2121 done = (work_done < orig_budget);
2122 }
2123 }
2124 2121
2125 if (done) { 2122 if (work_done < budget) {
2126 unsigned long flags; 2123 unsigned long flags;
2127 /* 2124 /*
2128 * Order is important since data can get interrupted 2125 * Order is important since data can get interrupted
2129 * again when we think we are done. 2126 * again when we think we are done.
2130 */ 2127 */
2131 local_irq_save(flags); 2128 spin_lock_irqsave(&tp->lock, flags);
2132 RTL_W16_F(IntrMask, rtl8139_intr_mask); 2129 RTL_W16_F(IntrMask, rtl8139_intr_mask);
2133 __netif_rx_complete(dev); 2130 __netif_rx_complete(dev, napi);
2134 local_irq_restore(flags); 2131 spin_unlock_irqrestore(&tp->lock, flags);
2135 } 2132 }
2136 spin_unlock(&tp->rx_lock); 2133 spin_unlock(&tp->rx_lock);
2137 2134
2138 return !done; 2135 return work_done;
2139} 2136}
2140 2137
2141/* The interrupt handler does all of the Rx thread work and cleans up 2138/* The interrupt handler does all of the Rx thread work and cleans up
@@ -2180,9 +2177,9 @@ static irqreturn_t rtl8139_interrupt (int irq, void *dev_instance)
2180 /* Receive packets are processed by poll routine. 2177 /* Receive packets are processed by poll routine.
2181 If not running start it now. */ 2178 If not running start it now. */
2182 if (status & RxAckBits){ 2179 if (status & RxAckBits){
2183 if (netif_rx_schedule_prep(dev)) { 2180 if (netif_rx_schedule_prep(dev, &tp->napi)) {
2184 RTL_W16_F (IntrMask, rtl8139_norx_intr_mask); 2181 RTL_W16_F (IntrMask, rtl8139_norx_intr_mask);
2185 __netif_rx_schedule (dev); 2182 __netif_rx_schedule(dev, &tp->napi);
2186 } 2183 }
2187 } 2184 }
2188 2185
@@ -2223,7 +2220,8 @@ static int rtl8139_close (struct net_device *dev)
2223 void __iomem *ioaddr = tp->mmio_addr; 2220 void __iomem *ioaddr = tp->mmio_addr;
2224 unsigned long flags; 2221 unsigned long flags;
2225 2222
2226 netif_stop_queue (dev); 2223 netif_stop_queue(dev);
2224 napi_disable(&tp->napi);
2227 2225
2228 if (netif_msg_ifdown(tp)) 2226 if (netif_msg_ifdown(tp))
2229 printk(KERN_DEBUG "%s: Shutting down ethercard, status was 0x%4.4x.\n", 2227 printk(KERN_DEBUG "%s: Shutting down ethercard, status was 0x%4.4x.\n",
diff --git a/drivers/net/amd8111e.c b/drivers/net/amd8111e.c
index a61b2f89fc33..cf06fc067e92 100644
--- a/drivers/net/amd8111e.c
+++ b/drivers/net/amd8111e.c
@@ -723,9 +723,10 @@ static int amd8111e_tx(struct net_device *dev)
723 723
724#ifdef CONFIG_AMD8111E_NAPI 724#ifdef CONFIG_AMD8111E_NAPI
725/* This function handles the driver receive operation in polling mode */ 725/* This function handles the driver receive operation in polling mode */
726static int amd8111e_rx_poll(struct net_device *dev, int * budget) 726static int amd8111e_rx_poll(struct napi_struct *napi, int budget)
727{ 727{
728 struct amd8111e_priv *lp = netdev_priv(dev); 728 struct amd8111e_priv *lp = container_of(napi, struct amd8111e_priv, napi);
729 struct net_device *dev = lp->amd8111e_net_dev;
729 int rx_index = lp->rx_idx & RX_RING_DR_MOD_MASK; 730 int rx_index = lp->rx_idx & RX_RING_DR_MOD_MASK;
730 void __iomem *mmio = lp->mmio; 731 void __iomem *mmio = lp->mmio;
731 struct sk_buff *skb,*new_skb; 732 struct sk_buff *skb,*new_skb;
@@ -737,7 +738,7 @@ static int amd8111e_rx_poll(struct net_device *dev, int * budget)
737#if AMD8111E_VLAN_TAG_USED 738#if AMD8111E_VLAN_TAG_USED
738 short vtag; 739 short vtag;
739#endif 740#endif
740 int rx_pkt_limit = dev->quota; 741 int rx_pkt_limit = budget;
741 unsigned long flags; 742 unsigned long flags;
742 743
743 do{ 744 do{
@@ -838,21 +839,14 @@ static int amd8111e_rx_poll(struct net_device *dev, int * budget)
838 } while(intr0 & RINT0); 839 } while(intr0 & RINT0);
839 840
840 /* Receive descriptor is empty now */ 841 /* Receive descriptor is empty now */
841 dev->quota -= num_rx_pkt;
842 *budget -= num_rx_pkt;
843
844 spin_lock_irqsave(&lp->lock, flags); 842 spin_lock_irqsave(&lp->lock, flags);
845 netif_rx_complete(dev); 843 __netif_rx_complete(dev, napi);
846 writel(VAL0|RINTEN0, mmio + INTEN0); 844 writel(VAL0|RINTEN0, mmio + INTEN0);
847 writel(VAL2 | RDMD0, mmio + CMD0); 845 writel(VAL2 | RDMD0, mmio + CMD0);
848 spin_unlock_irqrestore(&lp->lock, flags); 846 spin_unlock_irqrestore(&lp->lock, flags);
849 return 0;
850 847
851rx_not_empty: 848rx_not_empty:
852 /* Do not call a netif_rx_complete */ 849 return num_rx_pkt;
853 dev->quota -= num_rx_pkt;
854 *budget -= num_rx_pkt;
855 return 1;
856} 850}
857 851
858#else 852#else
@@ -1287,11 +1281,11 @@ static irqreturn_t amd8111e_interrupt(int irq, void *dev_id)
1287 /* Check if Receive Interrupt has occurred. */ 1281 /* Check if Receive Interrupt has occurred. */
1288#ifdef CONFIG_AMD8111E_NAPI 1282#ifdef CONFIG_AMD8111E_NAPI
1289 if(intr0 & RINT0){ 1283 if(intr0 & RINT0){
1290 if(netif_rx_schedule_prep(dev)){ 1284 if(netif_rx_schedule_prep(dev, &lp->napi)){
1291 /* Disable receive interupts */ 1285 /* Disable receive interupts */
1292 writel(RINTEN0, mmio + INTEN0); 1286 writel(RINTEN0, mmio + INTEN0);
1293 /* Schedule a polling routine */ 1287 /* Schedule a polling routine */
1294 __netif_rx_schedule(dev); 1288 __netif_rx_schedule(dev, &lp->napi);
1295 } 1289 }
1296 else if (intren0 & RINTEN0) { 1290 else if (intren0 & RINTEN0) {
1297 printk("************Driver bug! \ 1291 printk("************Driver bug! \
@@ -1345,6 +1339,8 @@ static int amd8111e_close(struct net_device * dev)
1345 struct amd8111e_priv *lp = netdev_priv(dev); 1339 struct amd8111e_priv *lp = netdev_priv(dev);
1346 netif_stop_queue(dev); 1340 netif_stop_queue(dev);
1347 1341
1342 napi_disable(&lp->napi);
1343
1348 spin_lock_irq(&lp->lock); 1344 spin_lock_irq(&lp->lock);
1349 1345
1350 amd8111e_disable_interrupt(lp); 1346 amd8111e_disable_interrupt(lp);
@@ -1375,12 +1371,15 @@ static int amd8111e_open(struct net_device * dev )
1375 dev->name, dev)) 1371 dev->name, dev))
1376 return -EAGAIN; 1372 return -EAGAIN;
1377 1373
1374 napi_enable(&lp->napi);
1375
1378 spin_lock_irq(&lp->lock); 1376 spin_lock_irq(&lp->lock);
1379 1377
1380 amd8111e_init_hw_default(lp); 1378 amd8111e_init_hw_default(lp);
1381 1379
1382 if(amd8111e_restart(dev)){ 1380 if(amd8111e_restart(dev)){
1383 spin_unlock_irq(&lp->lock); 1381 spin_unlock_irq(&lp->lock);
1382 napi_disable(&lp->napi);
1384 if (dev->irq) 1383 if (dev->irq)
1385 free_irq(dev->irq, dev); 1384 free_irq(dev->irq, dev);
1386 return -ENOMEM; 1385 return -ENOMEM;
@@ -2031,8 +2030,7 @@ static int __devinit amd8111e_probe_one(struct pci_dev *pdev,
2031 dev->tx_timeout = amd8111e_tx_timeout; 2030 dev->tx_timeout = amd8111e_tx_timeout;
2032 dev->watchdog_timeo = AMD8111E_TX_TIMEOUT; 2031 dev->watchdog_timeo = AMD8111E_TX_TIMEOUT;
2033#ifdef CONFIG_AMD8111E_NAPI 2032#ifdef CONFIG_AMD8111E_NAPI
2034 dev->poll = amd8111e_rx_poll; 2033 netif_napi_add(dev, &lp->napi, amd8111e_rx_poll, 32);
2035 dev->weight = 32;
2036#endif 2034#endif
2037#ifdef CONFIG_NET_POLL_CONTROLLER 2035#ifdef CONFIG_NET_POLL_CONTROLLER
2038 dev->poll_controller = amd8111e_poll; 2036 dev->poll_controller = amd8111e_poll;
diff --git a/drivers/net/amd8111e.h b/drivers/net/amd8111e.h
index e65080a5994a..612e653610e1 100644
--- a/drivers/net/amd8111e.h
+++ b/drivers/net/amd8111e.h
@@ -763,6 +763,8 @@ struct amd8111e_priv{
763 /* Reg memory mapped address */ 763 /* Reg memory mapped address */
764 void __iomem *mmio; 764 void __iomem *mmio;
765 765
766 struct napi_struct napi;
767
766 spinlock_t lock; /* Guard lock */ 768 spinlock_t lock; /* Guard lock */
767 unsigned long rx_idx, tx_idx; /* The next free ring entry */ 769 unsigned long rx_idx, tx_idx; /* The next free ring entry */
768 unsigned long tx_complete_idx; 770 unsigned long tx_complete_idx;
diff --git a/drivers/net/arm/ep93xx_eth.c b/drivers/net/arm/ep93xx_eth.c
index f6ece1d43f6e..7f016f3d5bf0 100644
--- a/drivers/net/arm/ep93xx_eth.c
+++ b/drivers/net/arm/ep93xx_eth.c
@@ -169,6 +169,9 @@ struct ep93xx_priv
169 spinlock_t tx_pending_lock; 169 spinlock_t tx_pending_lock;
170 unsigned int tx_pending; 170 unsigned int tx_pending;
171 171
172 struct net_device *dev;
173 struct napi_struct napi;
174
172 struct net_device_stats stats; 175 struct net_device_stats stats;
173 176
174 struct mii_if_info mii; 177 struct mii_if_info mii;
@@ -190,15 +193,11 @@ static struct net_device_stats *ep93xx_get_stats(struct net_device *dev)
190 return &(ep->stats); 193 return &(ep->stats);
191} 194}
192 195
193static int ep93xx_rx(struct net_device *dev, int *budget) 196static int ep93xx_rx(struct net_device *dev, int processed, int budget)
194{ 197{
195 struct ep93xx_priv *ep = netdev_priv(dev); 198 struct ep93xx_priv *ep = netdev_priv(dev);
196 int rx_done;
197 int processed;
198 199
199 rx_done = 0; 200 while (processed < budget) {
200 processed = 0;
201 while (*budget > 0) {
202 int entry; 201 int entry;
203 struct ep93xx_rstat *rstat; 202 struct ep93xx_rstat *rstat;
204 u32 rstat0; 203 u32 rstat0;
@@ -211,10 +210,8 @@ static int ep93xx_rx(struct net_device *dev, int *budget)
211 210
212 rstat0 = rstat->rstat0; 211 rstat0 = rstat->rstat0;
213 rstat1 = rstat->rstat1; 212 rstat1 = rstat->rstat1;
214 if (!(rstat0 & RSTAT0_RFP) || !(rstat1 & RSTAT1_RFP)) { 213 if (!(rstat0 & RSTAT0_RFP) || !(rstat1 & RSTAT1_RFP))
215 rx_done = 1;
216 break; 214 break;
217 }
218 215
219 rstat->rstat0 = 0; 216 rstat->rstat0 = 0;
220 rstat->rstat1 = 0; 217 rstat->rstat1 = 0;
@@ -275,8 +272,6 @@ static int ep93xx_rx(struct net_device *dev, int *budget)
275err: 272err:
276 ep->rx_pointer = (entry + 1) & (RX_QUEUE_ENTRIES - 1); 273 ep->rx_pointer = (entry + 1) & (RX_QUEUE_ENTRIES - 1);
277 processed++; 274 processed++;
278 dev->quota--;
279 (*budget)--;
280 } 275 }
281 276
282 if (processed) { 277 if (processed) {
@@ -284,7 +279,7 @@ err:
284 wrw(ep, REG_RXSTSENQ, processed); 279 wrw(ep, REG_RXSTSENQ, processed);
285 } 280 }
286 281
287 return !rx_done; 282 return processed;
288} 283}
289 284
290static int ep93xx_have_more_rx(struct ep93xx_priv *ep) 285static int ep93xx_have_more_rx(struct ep93xx_priv *ep)
@@ -293,36 +288,32 @@ static int ep93xx_have_more_rx(struct ep93xx_priv *ep)
293 return !!((rstat->rstat0 & RSTAT0_RFP) && (rstat->rstat1 & RSTAT1_RFP)); 288 return !!((rstat->rstat0 & RSTAT0_RFP) && (rstat->rstat1 & RSTAT1_RFP));
294} 289}
295 290
296static int ep93xx_poll(struct net_device *dev, int *budget) 291static int ep93xx_poll(struct napi_struct *napi, int budget)
297{ 292{
298 struct ep93xx_priv *ep = netdev_priv(dev); 293 struct ep93xx_priv *ep = container_of(napi, struct ep93xx_priv, napi);
299 294 struct net_device *dev = ep->dev;
300 /* 295 int rx = 0;
301 * @@@ Have to stop polling if device is downed while we
302 * are polling.
303 */
304 296
305poll_some_more: 297poll_some_more:
306 if (ep93xx_rx(dev, budget)) 298 rx = ep93xx_rx(dev, rx, budget);
307 return 1; 299 if (rx < budget) {
308 300 int more = 0;
309 netif_rx_complete(dev); 301
310 302 spin_lock_irq(&ep->rx_lock);
311 spin_lock_irq(&ep->rx_lock); 303 __netif_rx_complete(dev, napi);
312 wrl(ep, REG_INTEN, REG_INTEN_TX | REG_INTEN_RX); 304 wrl(ep, REG_INTEN, REG_INTEN_TX | REG_INTEN_RX);
313 if (ep93xx_have_more_rx(ep)) { 305 if (ep93xx_have_more_rx(ep)) {
314 wrl(ep, REG_INTEN, REG_INTEN_TX); 306 wrl(ep, REG_INTEN, REG_INTEN_TX);
315 wrl(ep, REG_INTSTSP, REG_INTSTS_RX); 307 wrl(ep, REG_INTSTSP, REG_INTSTS_RX);
308 more = 1;
309 }
316 spin_unlock_irq(&ep->rx_lock); 310 spin_unlock_irq(&ep->rx_lock);
317 311
318 if (netif_rx_reschedule(dev, 0)) 312 if (more && netif_rx_reschedule(dev, napi))
319 goto poll_some_more; 313 goto poll_some_more;
320
321 return 0;
322 } 314 }
323 spin_unlock_irq(&ep->rx_lock);
324 315
325 return 0; 316 return rx;
326} 317}
327 318
328static int ep93xx_xmit(struct sk_buff *skb, struct net_device *dev) 319static int ep93xx_xmit(struct sk_buff *skb, struct net_device *dev)
@@ -426,9 +417,9 @@ static irqreturn_t ep93xx_irq(int irq, void *dev_id)
426 417
427 if (status & REG_INTSTS_RX) { 418 if (status & REG_INTSTS_RX) {
428 spin_lock(&ep->rx_lock); 419 spin_lock(&ep->rx_lock);
429 if (likely(__netif_rx_schedule_prep(dev))) { 420 if (likely(__netif_rx_schedule_prep(dev, &ep->napi))) {
430 wrl(ep, REG_INTEN, REG_INTEN_TX); 421 wrl(ep, REG_INTEN, REG_INTEN_TX);
431 __netif_rx_schedule(dev); 422 __netif_rx_schedule(dev, &ep->napi);
432 } 423 }
433 spin_unlock(&ep->rx_lock); 424 spin_unlock(&ep->rx_lock);
434 } 425 }
@@ -648,7 +639,10 @@ static int ep93xx_open(struct net_device *dev)
648 dev->dev_addr[4], dev->dev_addr[5]); 639 dev->dev_addr[4], dev->dev_addr[5]);
649 } 640 }
650 641
642 napi_enable(&ep->napi);
643
651 if (ep93xx_start_hw(dev)) { 644 if (ep93xx_start_hw(dev)) {
645 napi_disable(&ep->napi);
652 ep93xx_free_buffers(ep); 646 ep93xx_free_buffers(ep);
653 return -EIO; 647 return -EIO;
654 } 648 }
@@ -662,6 +656,7 @@ static int ep93xx_open(struct net_device *dev)
662 656
663 err = request_irq(ep->irq, ep93xx_irq, IRQF_SHARED, dev->name, dev); 657 err = request_irq(ep->irq, ep93xx_irq, IRQF_SHARED, dev->name, dev);
664 if (err) { 658 if (err) {
659 napi_disable(&ep->napi);
665 ep93xx_stop_hw(dev); 660 ep93xx_stop_hw(dev);
666 ep93xx_free_buffers(ep); 661 ep93xx_free_buffers(ep);
667 return err; 662 return err;
@@ -678,6 +673,7 @@ static int ep93xx_close(struct net_device *dev)
678{ 673{
679 struct ep93xx_priv *ep = netdev_priv(dev); 674 struct ep93xx_priv *ep = netdev_priv(dev);
680 675
676 napi_disable(&ep->napi);
681 netif_stop_queue(dev); 677 netif_stop_queue(dev);
682 678
683 wrl(ep, REG_GIINTMSK, 0); 679 wrl(ep, REG_GIINTMSK, 0);
@@ -788,14 +784,12 @@ struct net_device *ep93xx_dev_alloc(struct ep93xx_eth_data *data)
788 784
789 dev->get_stats = ep93xx_get_stats; 785 dev->get_stats = ep93xx_get_stats;
790 dev->ethtool_ops = &ep93xx_ethtool_ops; 786 dev->ethtool_ops = &ep93xx_ethtool_ops;
791 dev->poll = ep93xx_poll;
792 dev->hard_start_xmit = ep93xx_xmit; 787 dev->hard_start_xmit = ep93xx_xmit;
793 dev->open = ep93xx_open; 788 dev->open = ep93xx_open;
794 dev->stop = ep93xx_close; 789 dev->stop = ep93xx_close;
795 dev->do_ioctl = ep93xx_ioctl; 790 dev->do_ioctl = ep93xx_ioctl;
796 791
797 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; 792 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
798 dev->weight = 64;
799 793
800 return dev; 794 return dev;
801} 795}
@@ -847,6 +841,8 @@ static int ep93xx_eth_probe(struct platform_device *pdev)
847 goto err_out; 841 goto err_out;
848 } 842 }
849 ep = netdev_priv(dev); 843 ep = netdev_priv(dev);
844 ep->dev = dev;
845 netif_napi_add(dev, &ep->napi, ep93xx_poll, 64);
850 846
851 platform_set_drvdata(pdev, dev); 847 platform_set_drvdata(pdev, dev);
852 848
diff --git a/drivers/net/b44.c b/drivers/net/b44.c
index 0795df235492..b92b3e25c42a 100644
--- a/drivers/net/b44.c
+++ b/drivers/net/b44.c
@@ -848,10 +848,11 @@ static int b44_rx(struct b44 *bp, int budget)
848 return received; 848 return received;
849} 849}
850 850
851static int b44_poll(struct net_device *netdev, int *budget) 851static int b44_poll(struct napi_struct *napi, int budget)
852{ 852{
853 struct b44 *bp = netdev_priv(netdev); 853 struct b44 *bp = container_of(napi, struct b44, napi);
854 int done; 854 struct net_device *netdev = bp->dev;
855 int work_done;
855 856
856 spin_lock_irq(&bp->lock); 857 spin_lock_irq(&bp->lock);
857 858
@@ -862,22 +863,9 @@ static int b44_poll(struct net_device *netdev, int *budget)
862 } 863 }
863 spin_unlock_irq(&bp->lock); 864 spin_unlock_irq(&bp->lock);
864 865
865 done = 1; 866 work_done = 0;
866 if (bp->istat & ISTAT_RX) { 867 if (bp->istat & ISTAT_RX)
867 int orig_budget = *budget; 868 work_done += b44_rx(bp, budget);
868 int work_done;
869
870 if (orig_budget > netdev->quota)
871 orig_budget = netdev->quota;
872
873 work_done = b44_rx(bp, orig_budget);
874
875 *budget -= work_done;
876 netdev->quota -= work_done;
877
878 if (work_done >= orig_budget)
879 done = 0;
880 }
881 869
882 if (bp->istat & ISTAT_ERRORS) { 870 if (bp->istat & ISTAT_ERRORS) {
883 unsigned long flags; 871 unsigned long flags;
@@ -888,15 +876,15 @@ static int b44_poll(struct net_device *netdev, int *budget)
888 b44_init_hw(bp, B44_FULL_RESET_SKIP_PHY); 876 b44_init_hw(bp, B44_FULL_RESET_SKIP_PHY);
889 netif_wake_queue(bp->dev); 877 netif_wake_queue(bp->dev);
890 spin_unlock_irqrestore(&bp->lock, flags); 878 spin_unlock_irqrestore(&bp->lock, flags);
891 done = 1; 879 work_done = 0;
892 } 880 }
893 881
894 if (done) { 882 if (work_done < budget) {
895 netif_rx_complete(netdev); 883 netif_rx_complete(netdev, napi);
896 b44_enable_ints(bp); 884 b44_enable_ints(bp);
897 } 885 }
898 886
899 return (done ? 0 : 1); 887 return work_done;
900} 888}
901 889
902static irqreturn_t b44_interrupt(int irq, void *dev_id) 890static irqreturn_t b44_interrupt(int irq, void *dev_id)
@@ -924,13 +912,13 @@ static irqreturn_t b44_interrupt(int irq, void *dev_id)
924 goto irq_ack; 912 goto irq_ack;
925 } 913 }
926 914
927 if (netif_rx_schedule_prep(dev)) { 915 if (netif_rx_schedule_prep(dev, &bp->napi)) {
928 /* NOTE: These writes are posted by the readback of 916 /* NOTE: These writes are posted by the readback of
929 * the ISTAT register below. 917 * the ISTAT register below.
930 */ 918 */
931 bp->istat = istat; 919 bp->istat = istat;
932 __b44_disable_ints(bp); 920 __b44_disable_ints(bp);
933 __netif_rx_schedule(dev); 921 __netif_rx_schedule(dev, &bp->napi);
934 } else { 922 } else {
935 printk(KERN_ERR PFX "%s: Error, poll already scheduled\n", 923 printk(KERN_ERR PFX "%s: Error, poll already scheduled\n",
936 dev->name); 924 dev->name);
@@ -1420,6 +1408,8 @@ static int b44_open(struct net_device *dev)
1420 if (err) 1408 if (err)
1421 goto out; 1409 goto out;
1422 1410
1411 napi_enable(&bp->napi);
1412
1423 b44_init_rings(bp); 1413 b44_init_rings(bp);
1424 b44_init_hw(bp, B44_FULL_RESET); 1414 b44_init_hw(bp, B44_FULL_RESET);
1425 1415
@@ -1427,6 +1417,7 @@ static int b44_open(struct net_device *dev)
1427 1417
1428 err = request_irq(dev->irq, b44_interrupt, IRQF_SHARED, dev->name, dev); 1418 err = request_irq(dev->irq, b44_interrupt, IRQF_SHARED, dev->name, dev);
1429 if (unlikely(err < 0)) { 1419 if (unlikely(err < 0)) {
1420 napi_disable(&bp->napi);
1430 b44_chip_reset(bp); 1421 b44_chip_reset(bp);
1431 b44_free_rings(bp); 1422 b44_free_rings(bp);
1432 b44_free_consistent(bp); 1423 b44_free_consistent(bp);
@@ -1609,7 +1600,7 @@ static int b44_close(struct net_device *dev)
1609 1600
1610 netif_stop_queue(dev); 1601 netif_stop_queue(dev);
1611 1602
1612 netif_poll_disable(dev); 1603 napi_disable(&bp->napi);
1613 1604
1614 del_timer_sync(&bp->timer); 1605 del_timer_sync(&bp->timer);
1615 1606
@@ -1626,8 +1617,6 @@ static int b44_close(struct net_device *dev)
1626 1617
1627 free_irq(dev->irq, dev); 1618 free_irq(dev->irq, dev);
1628 1619
1629 netif_poll_enable(dev);
1630
1631 if (bp->flags & B44_FLAG_WOL_ENABLE) { 1620 if (bp->flags & B44_FLAG_WOL_ENABLE) {
1632 b44_init_hw(bp, B44_PARTIAL_RESET); 1621 b44_init_hw(bp, B44_PARTIAL_RESET);
1633 b44_setup_wol(bp); 1622 b44_setup_wol(bp);
@@ -2194,8 +2183,7 @@ static int __devinit b44_init_one(struct pci_dev *pdev,
2194 dev->set_mac_address = b44_set_mac_addr; 2183 dev->set_mac_address = b44_set_mac_addr;
2195 dev->do_ioctl = b44_ioctl; 2184 dev->do_ioctl = b44_ioctl;
2196 dev->tx_timeout = b44_tx_timeout; 2185 dev->tx_timeout = b44_tx_timeout;
2197 dev->poll = b44_poll; 2186 netif_napi_add(dev, &bp->napi, b44_poll, 64);
2198 dev->weight = 64;
2199 dev->watchdog_timeo = B44_TX_TIMEOUT; 2187 dev->watchdog_timeo = B44_TX_TIMEOUT;
2200#ifdef CONFIG_NET_POLL_CONTROLLER 2188#ifdef CONFIG_NET_POLL_CONTROLLER
2201 dev->poll_controller = b44_poll_controller; 2189 dev->poll_controller = b44_poll_controller;
diff --git a/drivers/net/b44.h b/drivers/net/b44.h
index e537e63f292e..63c55a4ab3cd 100644
--- a/drivers/net/b44.h
+++ b/drivers/net/b44.h
@@ -423,6 +423,8 @@ struct b44 {
423 struct ring_info *rx_buffers; 423 struct ring_info *rx_buffers;
424 struct ring_info *tx_buffers; 424 struct ring_info *tx_buffers;
425 425
426 struct napi_struct napi;
427
426 u32 dma_offset; 428 u32 dma_offset;
427 u32 flags; 429 u32 flags;
428#define B44_FLAG_B0_ANDLATER 0x00000001 430#define B44_FLAG_B0_ANDLATER 0x00000001
diff --git a/drivers/net/bnx2.c b/drivers/net/bnx2.c
index 66eed22cbd21..ab028ad04235 100644
--- a/drivers/net/bnx2.c
+++ b/drivers/net/bnx2.c
@@ -428,7 +428,7 @@ bnx2_netif_stop(struct bnx2 *bp)
428{ 428{
429 bnx2_disable_int_sync(bp); 429 bnx2_disable_int_sync(bp);
430 if (netif_running(bp->dev)) { 430 if (netif_running(bp->dev)) {
431 netif_poll_disable(bp->dev); 431 napi_disable(&bp->napi);
432 netif_tx_disable(bp->dev); 432 netif_tx_disable(bp->dev);
433 bp->dev->trans_start = jiffies; /* prevent tx timeout */ 433 bp->dev->trans_start = jiffies; /* prevent tx timeout */
434 } 434 }
@@ -440,7 +440,7 @@ bnx2_netif_start(struct bnx2 *bp)
440 if (atomic_dec_and_test(&bp->intr_sem)) { 440 if (atomic_dec_and_test(&bp->intr_sem)) {
441 if (netif_running(bp->dev)) { 441 if (netif_running(bp->dev)) {
442 netif_wake_queue(bp->dev); 442 netif_wake_queue(bp->dev);
443 netif_poll_enable(bp->dev); 443 napi_enable(&bp->napi);
444 bnx2_enable_int(bp); 444 bnx2_enable_int(bp);
445 } 445 }
446 } 446 }
@@ -2551,7 +2551,7 @@ bnx2_msi(int irq, void *dev_instance)
2551 if (unlikely(atomic_read(&bp->intr_sem) != 0)) 2551 if (unlikely(atomic_read(&bp->intr_sem) != 0))
2552 return IRQ_HANDLED; 2552 return IRQ_HANDLED;
2553 2553
2554 netif_rx_schedule(dev); 2554 netif_rx_schedule(dev, &bp->napi);
2555 2555
2556 return IRQ_HANDLED; 2556 return IRQ_HANDLED;
2557} 2557}
@@ -2568,7 +2568,7 @@ bnx2_msi_1shot(int irq, void *dev_instance)
2568 if (unlikely(atomic_read(&bp->intr_sem) != 0)) 2568 if (unlikely(atomic_read(&bp->intr_sem) != 0))
2569 return IRQ_HANDLED; 2569 return IRQ_HANDLED;
2570 2570
2571 netif_rx_schedule(dev); 2571 netif_rx_schedule(dev, &bp->napi);
2572 2572
2573 return IRQ_HANDLED; 2573 return IRQ_HANDLED;
2574} 2574}
@@ -2604,9 +2604,9 @@ bnx2_interrupt(int irq, void *dev_instance)
2604 if (unlikely(atomic_read(&bp->intr_sem) != 0)) 2604 if (unlikely(atomic_read(&bp->intr_sem) != 0))
2605 return IRQ_HANDLED; 2605 return IRQ_HANDLED;
2606 2606
2607 if (netif_rx_schedule_prep(dev)) { 2607 if (netif_rx_schedule_prep(dev, &bp->napi)) {
2608 bp->last_status_idx = sblk->status_idx; 2608 bp->last_status_idx = sblk->status_idx;
2609 __netif_rx_schedule(dev); 2609 __netif_rx_schedule(dev, &bp->napi);
2610 } 2610 }
2611 2611
2612 return IRQ_HANDLED; 2612 return IRQ_HANDLED;
@@ -2632,12 +2632,14 @@ bnx2_has_work(struct bnx2 *bp)
2632} 2632}
2633 2633
2634static int 2634static int
2635bnx2_poll(struct net_device *dev, int *budget) 2635bnx2_poll(struct napi_struct *napi, int budget)
2636{ 2636{
2637 struct bnx2 *bp = netdev_priv(dev); 2637 struct bnx2 *bp = container_of(napi, struct bnx2, napi);
2638 struct net_device *dev = bp->dev;
2638 struct status_block *sblk = bp->status_blk; 2639 struct status_block *sblk = bp->status_blk;
2639 u32 status_attn_bits = sblk->status_attn_bits; 2640 u32 status_attn_bits = sblk->status_attn_bits;
2640 u32 status_attn_bits_ack = sblk->status_attn_bits_ack; 2641 u32 status_attn_bits_ack = sblk->status_attn_bits_ack;
2642 int work_done = 0;
2641 2643
2642 if ((status_attn_bits & STATUS_ATTN_EVENTS) != 2644 if ((status_attn_bits & STATUS_ATTN_EVENTS) !=
2643 (status_attn_bits_ack & STATUS_ATTN_EVENTS)) { 2645 (status_attn_bits_ack & STATUS_ATTN_EVENTS)) {
@@ -2655,23 +2657,14 @@ bnx2_poll(struct net_device *dev, int *budget)
2655 if (bp->status_blk->status_tx_quick_consumer_index0 != bp->hw_tx_cons) 2657 if (bp->status_blk->status_tx_quick_consumer_index0 != bp->hw_tx_cons)
2656 bnx2_tx_int(bp); 2658 bnx2_tx_int(bp);
2657 2659
2658 if (bp->status_blk->status_rx_quick_consumer_index0 != bp->hw_rx_cons) { 2660 if (bp->status_blk->status_rx_quick_consumer_index0 != bp->hw_rx_cons)
2659 int orig_budget = *budget; 2661 work_done = bnx2_rx_int(bp, budget);
2660 int work_done;
2661
2662 if (orig_budget > dev->quota)
2663 orig_budget = dev->quota;
2664
2665 work_done = bnx2_rx_int(bp, orig_budget);
2666 *budget -= work_done;
2667 dev->quota -= work_done;
2668 }
2669 2662
2670 bp->last_status_idx = bp->status_blk->status_idx; 2663 bp->last_status_idx = bp->status_blk->status_idx;
2671 rmb(); 2664 rmb();
2672 2665
2673 if (!bnx2_has_work(bp)) { 2666 if (!bnx2_has_work(bp)) {
2674 netif_rx_complete(dev); 2667 netif_rx_complete(dev, napi);
2675 if (likely(bp->flags & USING_MSI_FLAG)) { 2668 if (likely(bp->flags & USING_MSI_FLAG)) {
2676 REG_WR(bp, BNX2_PCICFG_INT_ACK_CMD, 2669 REG_WR(bp, BNX2_PCICFG_INT_ACK_CMD,
2677 BNX2_PCICFG_INT_ACK_CMD_INDEX_VALID | 2670 BNX2_PCICFG_INT_ACK_CMD_INDEX_VALID |
@@ -2686,10 +2679,9 @@ bnx2_poll(struct net_device *dev, int *budget)
2686 REG_WR(bp, BNX2_PCICFG_INT_ACK_CMD, 2679 REG_WR(bp, BNX2_PCICFG_INT_ACK_CMD,
2687 BNX2_PCICFG_INT_ACK_CMD_INDEX_VALID | 2680 BNX2_PCICFG_INT_ACK_CMD_INDEX_VALID |
2688 bp->last_status_idx); 2681 bp->last_status_idx);
2689 return 0;
2690 } 2682 }
2691 2683
2692 return 1; 2684 return work_done;
2693} 2685}
2694 2686
2695/* Called with rtnl_lock from vlan functions and also netif_tx_lock 2687/* Called with rtnl_lock from vlan functions and also netif_tx_lock
@@ -5039,6 +5031,8 @@ bnx2_open(struct net_device *dev)
5039 if (rc) 5031 if (rc)
5040 return rc; 5032 return rc;
5041 5033
5034 napi_enable(&bp->napi);
5035
5042 if ((bp->flags & MSI_CAP_FLAG) && !disable_msi) { 5036 if ((bp->flags & MSI_CAP_FLAG) && !disable_msi) {
5043 if (pci_enable_msi(bp->pdev) == 0) { 5037 if (pci_enable_msi(bp->pdev) == 0) {
5044 bp->flags |= USING_MSI_FLAG; 5038 bp->flags |= USING_MSI_FLAG;
@@ -5049,6 +5043,7 @@ bnx2_open(struct net_device *dev)
5049 rc = bnx2_request_irq(bp); 5043 rc = bnx2_request_irq(bp);
5050 5044
5051 if (rc) { 5045 if (rc) {
5046 napi_disable(&bp->napi);
5052 bnx2_free_mem(bp); 5047 bnx2_free_mem(bp);
5053 return rc; 5048 return rc;
5054 } 5049 }
@@ -5056,6 +5051,7 @@ bnx2_open(struct net_device *dev)
5056 rc = bnx2_init_nic(bp); 5051 rc = bnx2_init_nic(bp);
5057 5052
5058 if (rc) { 5053 if (rc) {
5054 napi_disable(&bp->napi);
5059 bnx2_free_irq(bp); 5055 bnx2_free_irq(bp);
5060 bnx2_free_skbs(bp); 5056 bnx2_free_skbs(bp);
5061 bnx2_free_mem(bp); 5057 bnx2_free_mem(bp);
@@ -5088,6 +5084,7 @@ bnx2_open(struct net_device *dev)
5088 rc = bnx2_request_irq(bp); 5084 rc = bnx2_request_irq(bp);
5089 5085
5090 if (rc) { 5086 if (rc) {
5087 napi_disable(&bp->napi);
5091 bnx2_free_skbs(bp); 5088 bnx2_free_skbs(bp);
5092 bnx2_free_mem(bp); 5089 bnx2_free_mem(bp);
5093 del_timer_sync(&bp->timer); 5090 del_timer_sync(&bp->timer);
@@ -5301,7 +5298,8 @@ bnx2_close(struct net_device *dev)
5301 while (bp->in_reset_task) 5298 while (bp->in_reset_task)
5302 msleep(1); 5299 msleep(1);
5303 5300
5304 bnx2_netif_stop(bp); 5301 bnx2_disable_int_sync(bp);
5302 napi_disable(&bp->napi);
5305 del_timer_sync(&bp->timer); 5303 del_timer_sync(&bp->timer);
5306 if (bp->flags & NO_WOL_FLAG) 5304 if (bp->flags & NO_WOL_FLAG)
5307 reset_code = BNX2_DRV_MSG_CODE_UNLOAD_LNK_DN; 5305 reset_code = BNX2_DRV_MSG_CODE_UNLOAD_LNK_DN;
@@ -6858,11 +6856,10 @@ bnx2_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
6858#ifdef BCM_VLAN 6856#ifdef BCM_VLAN
6859 dev->vlan_rx_register = bnx2_vlan_rx_register; 6857 dev->vlan_rx_register = bnx2_vlan_rx_register;
6860#endif 6858#endif
6861 dev->poll = bnx2_poll;
6862 dev->ethtool_ops = &bnx2_ethtool_ops; 6859 dev->ethtool_ops = &bnx2_ethtool_ops;
6863 dev->weight = 64;
6864 6860
6865 bp = netdev_priv(dev); 6861 bp = netdev_priv(dev);
6862 netif_napi_add(dev, &bp->napi, bnx2_poll, 64);
6866 6863
6867#if defined(HAVE_POLL_CONTROLLER) || defined(CONFIG_NET_POLL_CONTROLLER) 6864#if defined(HAVE_POLL_CONTROLLER) || defined(CONFIG_NET_POLL_CONTROLLER)
6868 dev->poll_controller = poll_bnx2; 6865 dev->poll_controller = poll_bnx2;
diff --git a/drivers/net/bnx2.h b/drivers/net/bnx2.h
index 102adfe1e923..fbae439db647 100644
--- a/drivers/net/bnx2.h
+++ b/drivers/net/bnx2.h
@@ -6473,6 +6473,8 @@ struct bnx2 {
6473 struct net_device *dev; 6473 struct net_device *dev;
6474 struct pci_dev *pdev; 6474 struct pci_dev *pdev;
6475 6475
6476 struct napi_struct napi;
6477
6476 atomic_t intr_sem; 6478 atomic_t intr_sem;
6477 6479
6478 struct status_block *status_blk; 6480 struct status_block *status_blk;
diff --git a/drivers/net/cassini.c b/drivers/net/cassini.c
index f6e4030c73d1..13f14df21e6e 100644
--- a/drivers/net/cassini.c
+++ b/drivers/net/cassini.c
@@ -2485,7 +2485,7 @@ static irqreturn_t cas_interruptN(int irq, void *dev_id)
2485 if (status & INTR_RX_DONE_ALT) { /* handle rx separately */ 2485 if (status & INTR_RX_DONE_ALT) { /* handle rx separately */
2486#ifdef USE_NAPI 2486#ifdef USE_NAPI
2487 cas_mask_intr(cp); 2487 cas_mask_intr(cp);
2488 netif_rx_schedule(dev); 2488 netif_rx_schedule(dev, &cp->napi);
2489#else 2489#else
2490 cas_rx_ringN(cp, ring, 0); 2490 cas_rx_ringN(cp, ring, 0);
2491#endif 2491#endif
@@ -2536,7 +2536,7 @@ static irqreturn_t cas_interrupt1(int irq, void *dev_id)
2536 if (status & INTR_RX_DONE_ALT) { /* handle rx separately */ 2536 if (status & INTR_RX_DONE_ALT) { /* handle rx separately */
2537#ifdef USE_NAPI 2537#ifdef USE_NAPI
2538 cas_mask_intr(cp); 2538 cas_mask_intr(cp);
2539 netif_rx_schedule(dev); 2539 netif_rx_schedule(dev, &cp->napi);
2540#else 2540#else
2541 cas_rx_ringN(cp, 1, 0); 2541 cas_rx_ringN(cp, 1, 0);
2542#endif 2542#endif
@@ -2592,7 +2592,7 @@ static irqreturn_t cas_interrupt(int irq, void *dev_id)
2592 if (status & INTR_RX_DONE) { 2592 if (status & INTR_RX_DONE) {
2593#ifdef USE_NAPI 2593#ifdef USE_NAPI
2594 cas_mask_intr(cp); 2594 cas_mask_intr(cp);
2595 netif_rx_schedule(dev); 2595 netif_rx_schedule(dev, &cp->napi);
2596#else 2596#else
2597 cas_rx_ringN(cp, 0, 0); 2597 cas_rx_ringN(cp, 0, 0);
2598#endif 2598#endif
@@ -2607,9 +2607,10 @@ static irqreturn_t cas_interrupt(int irq, void *dev_id)
2607 2607
2608 2608
2609#ifdef USE_NAPI 2609#ifdef USE_NAPI
2610static int cas_poll(struct net_device *dev, int *budget) 2610static int cas_poll(struct napi_struct *napi, int budget)
2611{ 2611{
2612 struct cas *cp = netdev_priv(dev); 2612 struct cas *cp = container_of(napi, struct cas, napi);
2613 struct net_device *dev = cp->dev;
2613 int i, enable_intr, todo, credits; 2614 int i, enable_intr, todo, credits;
2614 u32 status = readl(cp->regs + REG_INTR_STATUS); 2615 u32 status = readl(cp->regs + REG_INTR_STATUS);
2615 unsigned long flags; 2616 unsigned long flags;
@@ -2620,20 +2621,18 @@ static int cas_poll(struct net_device *dev, int *budget)
2620 2621
2621 /* NAPI rx packets. we spread the credits across all of the 2622 /* NAPI rx packets. we spread the credits across all of the
2622 * rxc rings 2623 * rxc rings
2623 */ 2624 *
2624 todo = min(*budget, dev->quota); 2625 * to make sure we're fair with the work we loop through each
2625
2626 /* to make sure we're fair with the work we loop through each
2627 * ring N_RX_COMP_RING times with a request of 2626 * ring N_RX_COMP_RING times with a request of
2628 * todo / N_RX_COMP_RINGS 2627 * budget / N_RX_COMP_RINGS
2629 */ 2628 */
2630 enable_intr = 1; 2629 enable_intr = 1;
2631 credits = 0; 2630 credits = 0;
2632 for (i = 0; i < N_RX_COMP_RINGS; i++) { 2631 for (i = 0; i < N_RX_COMP_RINGS; i++) {
2633 int j; 2632 int j;
2634 for (j = 0; j < N_RX_COMP_RINGS; j++) { 2633 for (j = 0; j < N_RX_COMP_RINGS; j++) {
2635 credits += cas_rx_ringN(cp, j, todo / N_RX_COMP_RINGS); 2634 credits += cas_rx_ringN(cp, j, budget / N_RX_COMP_RINGS);
2636 if (credits >= todo) { 2635 if (credits >= budget) {
2637 enable_intr = 0; 2636 enable_intr = 0;
2638 goto rx_comp; 2637 goto rx_comp;
2639 } 2638 }
@@ -2641,9 +2640,6 @@ static int cas_poll(struct net_device *dev, int *budget)
2641 } 2640 }
2642 2641
2643rx_comp: 2642rx_comp:
2644 *budget -= credits;
2645 dev->quota -= credits;
2646
2647 /* final rx completion */ 2643 /* final rx completion */
2648 spin_lock_irqsave(&cp->lock, flags); 2644 spin_lock_irqsave(&cp->lock, flags);
2649 if (status) 2645 if (status)
@@ -2674,11 +2670,10 @@ rx_comp:
2674#endif 2670#endif
2675 spin_unlock_irqrestore(&cp->lock, flags); 2671 spin_unlock_irqrestore(&cp->lock, flags);
2676 if (enable_intr) { 2672 if (enable_intr) {
2677 netif_rx_complete(dev); 2673 netif_rx_complete(dev, napi);
2678 cas_unmask_intr(cp); 2674 cas_unmask_intr(cp);
2679 return 0;
2680 } 2675 }
2681 return 1; 2676 return credits;
2682} 2677}
2683#endif 2678#endif
2684 2679
@@ -4351,6 +4346,9 @@ static int cas_open(struct net_device *dev)
4351 goto err_spare; 4346 goto err_spare;
4352 } 4347 }
4353 4348
4349#ifdef USE_NAPI
4350 napi_enable(&cp->napi);
4351#endif
4354 /* init hw */ 4352 /* init hw */
4355 cas_lock_all_save(cp, flags); 4353 cas_lock_all_save(cp, flags);
4356 cas_clean_rings(cp); 4354 cas_clean_rings(cp);
@@ -4376,6 +4374,9 @@ static int cas_close(struct net_device *dev)
4376 unsigned long flags; 4374 unsigned long flags;
4377 struct cas *cp = netdev_priv(dev); 4375 struct cas *cp = netdev_priv(dev);
4378 4376
4377#ifdef USE_NAPI
4378 napi_enable(&cp->napi);
4379#endif
4379 /* Make sure we don't get distracted by suspend/resume */ 4380 /* Make sure we don't get distracted by suspend/resume */
4380 mutex_lock(&cp->pm_mutex); 4381 mutex_lock(&cp->pm_mutex);
4381 4382
@@ -5062,8 +5063,7 @@ static int __devinit cas_init_one(struct pci_dev *pdev,
5062 dev->watchdog_timeo = CAS_TX_TIMEOUT; 5063 dev->watchdog_timeo = CAS_TX_TIMEOUT;
5063 dev->change_mtu = cas_change_mtu; 5064 dev->change_mtu = cas_change_mtu;
5064#ifdef USE_NAPI 5065#ifdef USE_NAPI
5065 dev->poll = cas_poll; 5066 netif_napi_add(dev, &cp->napi, cas_poll, 64);
5066 dev->weight = 64;
5067#endif 5067#endif
5068#ifdef CONFIG_NET_POLL_CONTROLLER 5068#ifdef CONFIG_NET_POLL_CONTROLLER
5069 dev->poll_controller = cas_netpoll; 5069 dev->poll_controller = cas_netpoll;
diff --git a/drivers/net/cassini.h b/drivers/net/cassini.h
index a970804487c7..2f93f83342d2 100644
--- a/drivers/net/cassini.h
+++ b/drivers/net/cassini.h
@@ -4280,6 +4280,8 @@ struct cas {
4280 int rx_cur[N_RX_COMP_RINGS], rx_new[N_RX_COMP_RINGS]; 4280 int rx_cur[N_RX_COMP_RINGS], rx_new[N_RX_COMP_RINGS];
4281 int rx_last[N_RX_DESC_RINGS]; 4281 int rx_last[N_RX_DESC_RINGS];
4282 4282
4283 struct napi_struct napi;
4284
4283 /* Set when chip is actually in operational state 4285 /* Set when chip is actually in operational state
4284 * (ie. not power managed) */ 4286 * (ie. not power managed) */
4285 int hw_running; 4287 int hw_running;
diff --git a/drivers/net/chelsio/common.h b/drivers/net/chelsio/common.h
index 8ba702c8b560..b5de4452cf24 100644
--- a/drivers/net/chelsio/common.h
+++ b/drivers/net/chelsio/common.h
@@ -278,6 +278,7 @@ struct adapter {
278 struct peespi *espi; 278 struct peespi *espi;
279 struct petp *tp; 279 struct petp *tp;
280 280
281 struct napi_struct napi;
281 struct port_info port[MAX_NPORTS]; 282 struct port_info port[MAX_NPORTS];
282 struct delayed_work stats_update_task; 283 struct delayed_work stats_update_task;
283 struct timer_list stats_update_timer; 284 struct timer_list stats_update_timer;
diff --git a/drivers/net/chelsio/cxgb2.c b/drivers/net/chelsio/cxgb2.c
index 231ce43b97cf..593736c7550d 100644
--- a/drivers/net/chelsio/cxgb2.c
+++ b/drivers/net/chelsio/cxgb2.c
@@ -255,8 +255,11 @@ static int cxgb_open(struct net_device *dev)
255 struct adapter *adapter = dev->priv; 255 struct adapter *adapter = dev->priv;
256 int other_ports = adapter->open_device_map & PORT_MASK; 256 int other_ports = adapter->open_device_map & PORT_MASK;
257 257
258 if (!adapter->open_device_map && (err = cxgb_up(adapter)) < 0) 258 napi_enable(&adapter->napi);
259 if (!adapter->open_device_map && (err = cxgb_up(adapter)) < 0) {
260 napi_disable(&adapter->napi);
259 return err; 261 return err;
262 }
260 263
261 __set_bit(dev->if_port, &adapter->open_device_map); 264 __set_bit(dev->if_port, &adapter->open_device_map);
262 link_start(&adapter->port[dev->if_port]); 265 link_start(&adapter->port[dev->if_port]);
@@ -274,6 +277,7 @@ static int cxgb_close(struct net_device *dev)
274 struct cmac *mac = p->mac; 277 struct cmac *mac = p->mac;
275 278
276 netif_stop_queue(dev); 279 netif_stop_queue(dev);
280 napi_disable(&adapter->napi);
277 mac->ops->disable(mac, MAC_DIRECTION_TX | MAC_DIRECTION_RX); 281 mac->ops->disable(mac, MAC_DIRECTION_TX | MAC_DIRECTION_RX);
278 netif_carrier_off(dev); 282 netif_carrier_off(dev);
279 283
@@ -1113,8 +1117,7 @@ static int __devinit init_one(struct pci_dev *pdev,
1113 netdev->poll_controller = t1_netpoll; 1117 netdev->poll_controller = t1_netpoll;
1114#endif 1118#endif
1115#ifdef CONFIG_CHELSIO_T1_NAPI 1119#ifdef CONFIG_CHELSIO_T1_NAPI
1116 netdev->weight = 64; 1120 netif_napi_add(netdev, &adapter->napi, t1_poll, 64);
1117 netdev->poll = t1_poll;
1118#endif 1121#endif
1119 1122
1120 SET_ETHTOOL_OPS(netdev, &t1_ethtool_ops); 1123 SET_ETHTOOL_OPS(netdev, &t1_ethtool_ops);
diff --git a/drivers/net/chelsio/sge.c b/drivers/net/chelsio/sge.c
index e4f874a70fe5..ffa7e649a6ef 100644
--- a/drivers/net/chelsio/sge.c
+++ b/drivers/net/chelsio/sge.c
@@ -1620,23 +1620,20 @@ static int process_pure_responses(struct adapter *adapter)
1620 * or protection from interrupts as data interrupts are off at this point and 1620 * or protection from interrupts as data interrupts are off at this point and
1621 * other adapter interrupts do not interfere. 1621 * other adapter interrupts do not interfere.
1622 */ 1622 */
1623int t1_poll(struct net_device *dev, int *budget) 1623int t1_poll(struct napi_struct *napi, int budget)
1624{ 1624{
1625 struct adapter *adapter = dev->priv; 1625 struct adapter *adapter = container_of(napi, struct adapter, napi);
1626 struct net_device *dev = adapter->port[0].dev;
1626 int work_done; 1627 int work_done;
1627 1628
1628 work_done = process_responses(adapter, min(*budget, dev->quota)); 1629 work_done = process_responses(adapter, budget);
1629 *budget -= work_done;
1630 dev->quota -= work_done;
1631
1632 if (unlikely(responses_pending(adapter)))
1633 return 1;
1634
1635 netif_rx_complete(dev);
1636 writel(adapter->sge->respQ.cidx, adapter->regs + A_SG_SLEEPING);
1637
1638 return 0;
1639 1630
1631 if (likely(!responses_pending(adapter))) {
1632 netif_rx_complete(dev, napi);
1633 writel(adapter->sge->respQ.cidx,
1634 adapter->regs + A_SG_SLEEPING);
1635 }
1636 return work_done;
1640} 1637}
1641 1638
1642/* 1639/*
@@ -1653,13 +1650,13 @@ irqreturn_t t1_interrupt(int irq, void *data)
1653 1650
1654 writel(F_PL_INTR_SGE_DATA, adapter->regs + A_PL_CAUSE); 1651 writel(F_PL_INTR_SGE_DATA, adapter->regs + A_PL_CAUSE);
1655 1652
1656 if (__netif_rx_schedule_prep(dev)) { 1653 if (napi_schedule_prep(&adapter->napi)) {
1657 if (process_pure_responses(adapter)) 1654 if (process_pure_responses(adapter))
1658 __netif_rx_schedule(dev); 1655 __netif_rx_schedule(dev, &adapter->napi);
1659 else { 1656 else {
1660 /* no data, no NAPI needed */ 1657 /* no data, no NAPI needed */
1661 writel(sge->respQ.cidx, adapter->regs + A_SG_SLEEPING); 1658 writel(sge->respQ.cidx, adapter->regs + A_SG_SLEEPING);
1662 netif_poll_enable(dev); /* undo schedule_prep */ 1659 napi_enable(&adapter->napi); /* undo schedule_prep */
1663 } 1660 }
1664 } 1661 }
1665 return IRQ_HANDLED; 1662 return IRQ_HANDLED;
diff --git a/drivers/net/chelsio/sge.h b/drivers/net/chelsio/sge.h
index d132a0ef2a22..713d9c55f24d 100644
--- a/drivers/net/chelsio/sge.h
+++ b/drivers/net/chelsio/sge.h
@@ -77,7 +77,7 @@ int t1_sge_configure(struct sge *, struct sge_params *);
77int t1_sge_set_coalesce_params(struct sge *, struct sge_params *); 77int t1_sge_set_coalesce_params(struct sge *, struct sge_params *);
78void t1_sge_destroy(struct sge *); 78void t1_sge_destroy(struct sge *);
79irqreturn_t t1_interrupt(int irq, void *cookie); 79irqreturn_t t1_interrupt(int irq, void *cookie);
80int t1_poll(struct net_device *, int *); 80int t1_poll(struct napi_struct *, int);
81 81
82int t1_start_xmit(struct sk_buff *skb, struct net_device *dev); 82int t1_start_xmit(struct sk_buff *skb, struct net_device *dev);
83void t1_set_vlan_accel(struct adapter *adapter, int on_off); 83void t1_set_vlan_accel(struct adapter *adapter, int on_off);
diff --git a/drivers/net/cxgb3/adapter.h b/drivers/net/cxgb3/adapter.h
index 20e887de2545..044261703381 100644
--- a/drivers/net/cxgb3/adapter.h
+++ b/drivers/net/cxgb3/adapter.h
@@ -49,11 +49,13 @@
49typedef irqreturn_t(*intr_handler_t) (int, void *); 49typedef irqreturn_t(*intr_handler_t) (int, void *);
50 50
51struct vlan_group; 51struct vlan_group;
52
53struct adapter; 52struct adapter;
53struct sge_qset;
54
54struct port_info { 55struct port_info {
55 struct adapter *adapter; 56 struct adapter *adapter;
56 struct vlan_group *vlan_grp; 57 struct vlan_group *vlan_grp;
58 struct sge_qset *qs;
57 const struct port_type_info *port_type; 59 const struct port_type_info *port_type;
58 u8 port_id; 60 u8 port_id;
59 u8 rx_csum_offload; 61 u8 rx_csum_offload;
@@ -173,10 +175,12 @@ enum { /* per port SGE statistics */
173}; 175};
174 176
175struct sge_qset { /* an SGE queue set */ 177struct sge_qset { /* an SGE queue set */
178 struct adapter *adap;
179 struct napi_struct napi;
176 struct sge_rspq rspq; 180 struct sge_rspq rspq;
177 struct sge_fl fl[SGE_RXQ_PER_SET]; 181 struct sge_fl fl[SGE_RXQ_PER_SET];
178 struct sge_txq txq[SGE_TXQ_PER_SET]; 182 struct sge_txq txq[SGE_TXQ_PER_SET];
179 struct net_device *netdev; /* associated net device */ 183 struct net_device *netdev;
180 unsigned long txq_stopped; /* which Tx queues are stopped */ 184 unsigned long txq_stopped; /* which Tx queues are stopped */
181 struct timer_list tx_reclaim_timer; /* reclaims TX buffers */ 185 struct timer_list tx_reclaim_timer; /* reclaims TX buffers */
182 unsigned long port_stats[SGE_PSTAT_MAX]; 186 unsigned long port_stats[SGE_PSTAT_MAX];
@@ -221,12 +225,6 @@ struct adapter {
221 struct delayed_work adap_check_task; 225 struct delayed_work adap_check_task;
222 struct work_struct ext_intr_handler_task; 226 struct work_struct ext_intr_handler_task;
223 227
224 /*
225 * Dummy netdevices are needed when using multiple receive queues with
226 * NAPI as each netdevice can service only one queue.
227 */
228 struct net_device *dummy_netdev[SGE_QSETS - 1];
229
230 struct dentry *debugfs_root; 228 struct dentry *debugfs_root;
231 229
232 struct mutex mdio_lock; 230 struct mutex mdio_lock;
@@ -253,12 +251,6 @@ static inline struct port_info *adap2pinfo(struct adapter *adap, int idx)
253 return netdev_priv(adap->port[idx]); 251 return netdev_priv(adap->port[idx]);
254} 252}
255 253
256/*
257 * We use the spare atalk_ptr to map a net device to its SGE queue set.
258 * This is a macro so it can be used as l-value.
259 */
260#define dev2qset(netdev) ((netdev)->atalk_ptr)
261
262#define OFFLOAD_DEVMAP_BIT 15 254#define OFFLOAD_DEVMAP_BIT 15
263 255
264#define tdev2adap(d) container_of(d, struct adapter, tdev) 256#define tdev2adap(d) container_of(d, struct adapter, tdev)
@@ -284,7 +276,7 @@ int t3_mgmt_tx(struct adapter *adap, struct sk_buff *skb);
284void t3_update_qset_coalesce(struct sge_qset *qs, const struct qset_params *p); 276void t3_update_qset_coalesce(struct sge_qset *qs, const struct qset_params *p);
285int t3_sge_alloc_qset(struct adapter *adapter, unsigned int id, int nports, 277int t3_sge_alloc_qset(struct adapter *adapter, unsigned int id, int nports,
286 int irq_vec_idx, const struct qset_params *p, 278 int irq_vec_idx, const struct qset_params *p,
287 int ntxq, struct net_device *netdev); 279 int ntxq, struct net_device *dev);
288int t3_get_desc(const struct sge_qset *qs, unsigned int qnum, unsigned int idx, 280int t3_get_desc(const struct sge_qset *qs, unsigned int qnum, unsigned int idx,
289 unsigned char *data); 281 unsigned char *data);
290irqreturn_t t3_sge_intr_msix(int irq, void *cookie); 282irqreturn_t t3_sge_intr_msix(int irq, void *cookie);
diff --git a/drivers/net/cxgb3/cxgb3_main.c b/drivers/net/cxgb3/cxgb3_main.c
index 5ab319cfe5de..5db7d4e27ec0 100644
--- a/drivers/net/cxgb3/cxgb3_main.c
+++ b/drivers/net/cxgb3/cxgb3_main.c
@@ -339,49 +339,17 @@ static void setup_rss(struct adapter *adap)
339 V_RRCPLCPUSIZE(6), cpus, rspq_map); 339 V_RRCPLCPUSIZE(6), cpus, rspq_map);
340} 340}
341 341
342/* 342static void init_napi(struct adapter *adap)
343 * If we have multiple receive queues per port serviced by NAPI we need one
344 * netdevice per queue as NAPI operates on netdevices. We already have one
345 * netdevice, namely the one associated with the interface, so we use dummy
346 * ones for any additional queues. Note that these netdevices exist purely
347 * so that NAPI has something to work with, they do not represent network
348 * ports and are not registered.
349 */
350static int init_dummy_netdevs(struct adapter *adap)
351{ 343{
352 int i, j, dummy_idx = 0; 344 int i;
353 struct net_device *nd;
354
355 for_each_port(adap, i) {
356 struct net_device *dev = adap->port[i];
357 const struct port_info *pi = netdev_priv(dev);
358
359 for (j = 0; j < pi->nqsets - 1; j++) {
360 if (!adap->dummy_netdev[dummy_idx]) {
361 struct port_info *p;
362
363 nd = alloc_netdev(sizeof(*p), "", ether_setup);
364 if (!nd)
365 goto free_all;
366 345
367 p = netdev_priv(nd); 346 for (i = 0; i < SGE_QSETS; i++) {
368 p->adapter = adap; 347 struct sge_qset *qs = &adap->sge.qs[i];
369 nd->weight = 64;
370 set_bit(__LINK_STATE_START, &nd->state);
371 adap->dummy_netdev[dummy_idx] = nd;
372 }
373 strcpy(adap->dummy_netdev[dummy_idx]->name, dev->name);
374 dummy_idx++;
375 }
376 }
377 return 0;
378 348
379free_all: 349 if (qs->adap)
380 while (--dummy_idx >= 0) { 350 netif_napi_add(qs->netdev, &qs->napi, qs->napi.poll,
381 free_netdev(adap->dummy_netdev[dummy_idx]); 351 64);
382 adap->dummy_netdev[dummy_idx] = NULL;
383 } 352 }
384 return -ENOMEM;
385} 353}
386 354
387/* 355/*
@@ -392,20 +360,18 @@ free_all:
392static void quiesce_rx(struct adapter *adap) 360static void quiesce_rx(struct adapter *adap)
393{ 361{
394 int i; 362 int i;
395 struct net_device *dev;
396 363
397 for_each_port(adap, i) { 364 for (i = 0; i < SGE_QSETS; i++)
398 dev = adap->port[i]; 365 if (adap->sge.qs[i].adap)
399 while (test_bit(__LINK_STATE_RX_SCHED, &dev->state)) 366 napi_disable(&adap->sge.qs[i].napi);
400 msleep(1); 367}
401 }
402 368
403 for (i = 0; i < ARRAY_SIZE(adap->dummy_netdev); i++) { 369static void enable_all_napi(struct adapter *adap)
404 dev = adap->dummy_netdev[i]; 370{
405 if (dev) 371 int i;
406 while (test_bit(__LINK_STATE_RX_SCHED, &dev->state)) 372 for (i = 0; i < SGE_QSETS; i++)
407 msleep(1); 373 if (adap->sge.qs[i].adap)
408 } 374 napi_enable(&adap->sge.qs[i].napi);
409} 375}
410 376
411/** 377/**
@@ -418,7 +384,7 @@ static void quiesce_rx(struct adapter *adap)
418 */ 384 */
419static int setup_sge_qsets(struct adapter *adap) 385static int setup_sge_qsets(struct adapter *adap)
420{ 386{
421 int i, j, err, irq_idx = 0, qset_idx = 0, dummy_dev_idx = 0; 387 int i, j, err, irq_idx = 0, qset_idx = 0;
422 unsigned int ntxq = SGE_TXQ_PER_SET; 388 unsigned int ntxq = SGE_TXQ_PER_SET;
423 389
424 if (adap->params.rev > 0 && !(adap->flags & USING_MSI)) 390 if (adap->params.rev > 0 && !(adap->flags & USING_MSI))
@@ -426,15 +392,14 @@ static int setup_sge_qsets(struct adapter *adap)
426 392
427 for_each_port(adap, i) { 393 for_each_port(adap, i) {
428 struct net_device *dev = adap->port[i]; 394 struct net_device *dev = adap->port[i];
429 const struct port_info *pi = netdev_priv(dev); 395 struct port_info *pi = netdev_priv(dev);
430 396
397 pi->qs = &adap->sge.qs[pi->first_qset];
431 for (j = 0; j < pi->nqsets; ++j, ++qset_idx) { 398 for (j = 0; j < pi->nqsets; ++j, ++qset_idx) {
432 err = t3_sge_alloc_qset(adap, qset_idx, 1, 399 err = t3_sge_alloc_qset(adap, qset_idx, 1,
433 (adap->flags & USING_MSIX) ? qset_idx + 1 : 400 (adap->flags & USING_MSIX) ? qset_idx + 1 :
434 irq_idx, 401 irq_idx,
435 &adap->params.sge.qset[qset_idx], ntxq, 402 &adap->params.sge.qset[qset_idx], ntxq, dev);
436 j == 0 ? dev :
437 adap-> dummy_netdev[dummy_dev_idx++]);
438 if (err) { 403 if (err) {
439 t3_free_sge_resources(adap); 404 t3_free_sge_resources(adap);
440 return err; 405 return err;
@@ -845,21 +810,18 @@ static int cxgb_up(struct adapter *adap)
845 goto out; 810 goto out;
846 } 811 }
847 812
848 err = init_dummy_netdevs(adap);
849 if (err)
850 goto out;
851
852 err = t3_init_hw(adap, 0); 813 err = t3_init_hw(adap, 0);
853 if (err) 814 if (err)
854 goto out; 815 goto out;
855 816
856 t3_write_reg(adap, A_ULPRX_TDDP_PSZ, V_HPZ0(PAGE_SHIFT - 12)); 817 t3_write_reg(adap, A_ULPRX_TDDP_PSZ, V_HPZ0(PAGE_SHIFT - 12));
857 818
858 err = setup_sge_qsets(adap); 819 err = setup_sge_qsets(adap);
859 if (err) 820 if (err)
860 goto out; 821 goto out;
861 822
862 setup_rss(adap); 823 setup_rss(adap);
824 init_napi(adap);
863 adap->flags |= FULL_INIT_DONE; 825 adap->flags |= FULL_INIT_DONE;
864 } 826 }
865 827
@@ -886,6 +848,7 @@ static int cxgb_up(struct adapter *adap)
886 adap->name, adap))) 848 adap->name, adap)))
887 goto irq_err; 849 goto irq_err;
888 850
851 enable_all_napi(adap);
889 t3_sge_start(adap); 852 t3_sge_start(adap);
890 t3_intr_enable(adap); 853 t3_intr_enable(adap);
891 854
@@ -1012,8 +975,10 @@ static int cxgb_open(struct net_device *dev)
1012 int other_ports = adapter->open_device_map & PORT_MASK; 975 int other_ports = adapter->open_device_map & PORT_MASK;
1013 int err; 976 int err;
1014 977
1015 if (!adapter->open_device_map && (err = cxgb_up(adapter)) < 0) 978 if (!adapter->open_device_map && (err = cxgb_up(adapter)) < 0) {
979 quiesce_rx(adapter);
1016 return err; 980 return err;
981 }
1017 982
1018 set_bit(pi->port_id, &adapter->open_device_map); 983 set_bit(pi->port_id, &adapter->open_device_map);
1019 if (is_offload(adapter) && !ofld_disable) { 984 if (is_offload(adapter) && !ofld_disable) {
@@ -2524,7 +2489,6 @@ static int __devinit init_one(struct pci_dev *pdev,
2524#ifdef CONFIG_NET_POLL_CONTROLLER 2489#ifdef CONFIG_NET_POLL_CONTROLLER
2525 netdev->poll_controller = cxgb_netpoll; 2490 netdev->poll_controller = cxgb_netpoll;
2526#endif 2491#endif
2527 netdev->weight = 64;
2528 2492
2529 SET_ETHTOOL_OPS(netdev, &cxgb_ethtool_ops); 2493 SET_ETHTOOL_OPS(netdev, &cxgb_ethtool_ops);
2530 } 2494 }
@@ -2625,12 +2589,6 @@ static void __devexit remove_one(struct pci_dev *pdev)
2625 t3_free_sge_resources(adapter); 2589 t3_free_sge_resources(adapter);
2626 cxgb_disable_msi(adapter); 2590 cxgb_disable_msi(adapter);
2627 2591
2628 for (i = 0; i < ARRAY_SIZE(adapter->dummy_netdev); i++)
2629 if (adapter->dummy_netdev[i]) {
2630 free_netdev(adapter->dummy_netdev[i]);
2631 adapter->dummy_netdev[i] = NULL;
2632 }
2633
2634 for_each_port(adapter, i) 2592 for_each_port(adapter, i)
2635 if (adapter->port[i]) 2593 if (adapter->port[i])
2636 free_netdev(adapter->port[i]); 2594 free_netdev(adapter->port[i]);
diff --git a/drivers/net/cxgb3/sge.c b/drivers/net/cxgb3/sge.c
index 58a5f60521ed..069c1aca8a6b 100644
--- a/drivers/net/cxgb3/sge.c
+++ b/drivers/net/cxgb3/sge.c
@@ -591,9 +591,6 @@ void t3_free_qset(struct adapter *adapter, struct sge_qset *q)
591 q->rspq.desc, q->rspq.phys_addr); 591 q->rspq.desc, q->rspq.phys_addr);
592 } 592 }
593 593
594 if (q->netdev)
595 q->netdev->atalk_ptr = NULL;
596
597 memset(q, 0, sizeof(*q)); 594 memset(q, 0, sizeof(*q));
598} 595}
599 596
@@ -1074,7 +1071,7 @@ int t3_eth_xmit(struct sk_buff *skb, struct net_device *dev)
1074 unsigned int ndesc, pidx, credits, gen, compl; 1071 unsigned int ndesc, pidx, credits, gen, compl;
1075 const struct port_info *pi = netdev_priv(dev); 1072 const struct port_info *pi = netdev_priv(dev);
1076 struct adapter *adap = pi->adapter; 1073 struct adapter *adap = pi->adapter;
1077 struct sge_qset *qs = dev2qset(dev); 1074 struct sge_qset *qs = pi->qs;
1078 struct sge_txq *q = &qs->txq[TXQ_ETH]; 1075 struct sge_txq *q = &qs->txq[TXQ_ETH];
1079 1076
1080 /* 1077 /*
@@ -1326,13 +1323,12 @@ static void restart_ctrlq(unsigned long data)
1326 struct sk_buff *skb; 1323 struct sk_buff *skb;
1327 struct sge_qset *qs = (struct sge_qset *)data; 1324 struct sge_qset *qs = (struct sge_qset *)data;
1328 struct sge_txq *q = &qs->txq[TXQ_CTRL]; 1325 struct sge_txq *q = &qs->txq[TXQ_CTRL];
1329 const struct port_info *pi = netdev_priv(qs->netdev);
1330 struct adapter *adap = pi->adapter;
1331 1326
1332 spin_lock(&q->lock); 1327 spin_lock(&q->lock);
1333 again:reclaim_completed_tx_imm(q); 1328 again:reclaim_completed_tx_imm(q);
1334 1329
1335 while (q->in_use < q->size && (skb = __skb_dequeue(&q->sendq)) != NULL) { 1330 while (q->in_use < q->size &&
1331 (skb = __skb_dequeue(&q->sendq)) != NULL) {
1336 1332
1337 write_imm(&q->desc[q->pidx], skb, skb->len, q->gen); 1333 write_imm(&q->desc[q->pidx], skb, skb->len, q->gen);
1338 1334
@@ -1354,7 +1350,7 @@ static void restart_ctrlq(unsigned long data)
1354 } 1350 }
1355 1351
1356 spin_unlock(&q->lock); 1352 spin_unlock(&q->lock);
1357 t3_write_reg(adap, A_SG_KDOORBELL, 1353 t3_write_reg(qs->adap, A_SG_KDOORBELL,
1358 F_SELEGRCNTX | V_EGRCNTX(q->cntxt_id)); 1354 F_SELEGRCNTX | V_EGRCNTX(q->cntxt_id));
1359} 1355}
1360 1356
@@ -1638,8 +1634,7 @@ static inline void offload_enqueue(struct sge_rspq *q, struct sk_buff *skb)
1638 else { 1634 else {
1639 struct sge_qset *qs = rspq_to_qset(q); 1635 struct sge_qset *qs = rspq_to_qset(q);
1640 1636
1641 if (__netif_rx_schedule_prep(qs->netdev)) 1637 napi_schedule(&qs->napi);
1642 __netif_rx_schedule(qs->netdev);
1643 q->rx_head = skb; 1638 q->rx_head = skb;
1644 } 1639 }
1645 q->rx_tail = skb; 1640 q->rx_tail = skb;
@@ -1675,34 +1670,30 @@ static inline void deliver_partial_bundle(struct t3cdev *tdev,
1675 * receive handler. Batches need to be of modest size as we do prefetches 1670 * receive handler. Batches need to be of modest size as we do prefetches
1676 * on the packets in each. 1671 * on the packets in each.
1677 */ 1672 */
1678static int ofld_poll(struct net_device *dev, int *budget) 1673static int ofld_poll(struct napi_struct *napi, int budget)
1679{ 1674{
1680 const struct port_info *pi = netdev_priv(dev); 1675 struct sge_qset *qs = container_of(napi, struct sge_qset, napi);
1681 struct adapter *adapter = pi->adapter;
1682 struct sge_qset *qs = dev2qset(dev);
1683 struct sge_rspq *q = &qs->rspq; 1676 struct sge_rspq *q = &qs->rspq;
1684 int work_done, limit = min(*budget, dev->quota), avail = limit; 1677 struct adapter *adapter = qs->adap;
1678 int work_done = 0;
1685 1679
1686 while (avail) { 1680 while (work_done < budget) {
1687 struct sk_buff *head, *tail, *skbs[RX_BUNDLE_SIZE]; 1681 struct sk_buff *head, *tail, *skbs[RX_BUNDLE_SIZE];
1688 int ngathered; 1682 int ngathered;
1689 1683
1690 spin_lock_irq(&q->lock); 1684 spin_lock_irq(&q->lock);
1691 head = q->rx_head; 1685 head = q->rx_head;
1692 if (!head) { 1686 if (!head) {
1693 work_done = limit - avail; 1687 napi_complete(napi);
1694 *budget -= work_done;
1695 dev->quota -= work_done;
1696 __netif_rx_complete(dev);
1697 spin_unlock_irq(&q->lock); 1688 spin_unlock_irq(&q->lock);
1698 return 0; 1689 return work_done;
1699 } 1690 }
1700 1691
1701 tail = q->rx_tail; 1692 tail = q->rx_tail;
1702 q->rx_head = q->rx_tail = NULL; 1693 q->rx_head = q->rx_tail = NULL;
1703 spin_unlock_irq(&q->lock); 1694 spin_unlock_irq(&q->lock);
1704 1695
1705 for (ngathered = 0; avail && head; avail--) { 1696 for (ngathered = 0; work_done < budget && head; work_done++) {
1706 prefetch(head->data); 1697 prefetch(head->data);
1707 skbs[ngathered] = head; 1698 skbs[ngathered] = head;
1708 head = head->next; 1699 head = head->next;
@@ -1724,10 +1715,8 @@ static int ofld_poll(struct net_device *dev, int *budget)
1724 } 1715 }
1725 deliver_partial_bundle(&adapter->tdev, q, skbs, ngathered); 1716 deliver_partial_bundle(&adapter->tdev, q, skbs, ngathered);
1726 } 1717 }
1727 work_done = limit - avail; 1718
1728 *budget -= work_done; 1719 return work_done;
1729 dev->quota -= work_done;
1730 return 1;
1731} 1720}
1732 1721
1733/** 1722/**
@@ -2071,50 +2060,47 @@ static inline int is_pure_response(const struct rsp_desc *r)
2071 2060
2072/** 2061/**
2073 * napi_rx_handler - the NAPI handler for Rx processing 2062 * napi_rx_handler - the NAPI handler for Rx processing
2074 * @dev: the net device 2063 * @napi: the napi instance
2075 * @budget: how many packets we can process in this round 2064 * @budget: how many packets we can process in this round
2076 * 2065 *
2077 * Handler for new data events when using NAPI. 2066 * Handler for new data events when using NAPI.
2078 */ 2067 */
2079static int napi_rx_handler(struct net_device *dev, int *budget) 2068static int napi_rx_handler(struct napi_struct *napi, int budget)
2080{ 2069{
2081 const struct port_info *pi = netdev_priv(dev); 2070 struct sge_qset *qs = container_of(napi, struct sge_qset, napi);
2082 struct adapter *adap = pi->adapter; 2071 struct adapter *adap = qs->adap;
2083 struct sge_qset *qs = dev2qset(dev); 2072 int work_done = process_responses(adap, qs, budget);
2084 int effective_budget = min(*budget, dev->quota);
2085
2086 int work_done = process_responses(adap, qs, effective_budget);
2087 *budget -= work_done;
2088 dev->quota -= work_done;
2089 2073
2090 if (work_done >= effective_budget) 2074 if (likely(work_done < budget)) {
2091 return 1; 2075 napi_complete(napi);
2092
2093 netif_rx_complete(dev);
2094 2076
2095 /* 2077 /*
2096 * Because we don't atomically flush the following write it is 2078 * Because we don't atomically flush the following
2097 * possible that in very rare cases it can reach the device in a way 2079 * write it is possible that in very rare cases it can
2098 * that races with a new response being written plus an error interrupt 2080 * reach the device in a way that races with a new
2099 * causing the NAPI interrupt handler below to return unhandled status 2081 * response being written plus an error interrupt
2100 * to the OS. To protect against this would require flushing the write 2082 * causing the NAPI interrupt handler below to return
2101 * and doing both the write and the flush with interrupts off. Way too 2083 * unhandled status to the OS. To protect against
2102 * expensive and unjustifiable given the rarity of the race. 2084 * this would require flushing the write and doing
2103 * 2085 * both the write and the flush with interrupts off.
2104 * The race cannot happen at all with MSI-X. 2086 * Way too expensive and unjustifiable given the
2105 */ 2087 * rarity of the race.
2106 t3_write_reg(adap, A_SG_GTS, V_RSPQ(qs->rspq.cntxt_id) | 2088 *
2107 V_NEWTIMER(qs->rspq.next_holdoff) | 2089 * The race cannot happen at all with MSI-X.
2108 V_NEWINDEX(qs->rspq.cidx)); 2090 */
2109 return 0; 2091 t3_write_reg(adap, A_SG_GTS, V_RSPQ(qs->rspq.cntxt_id) |
2092 V_NEWTIMER(qs->rspq.next_holdoff) |
2093 V_NEWINDEX(qs->rspq.cidx));
2094 }
2095 return work_done;
2110} 2096}
2111 2097
2112/* 2098/*
2113 * Returns true if the device is already scheduled for polling. 2099 * Returns true if the device is already scheduled for polling.
2114 */ 2100 */
2115static inline int napi_is_scheduled(struct net_device *dev) 2101static inline int napi_is_scheduled(struct napi_struct *napi)
2116{ 2102{
2117 return test_bit(__LINK_STATE_RX_SCHED, &dev->state); 2103 return test_bit(NAPI_STATE_SCHED, &napi->state);
2118} 2104}
2119 2105
2120/** 2106/**
@@ -2197,8 +2183,7 @@ static inline int handle_responses(struct adapter *adap, struct sge_rspq *q)
2197 V_NEWTIMER(q->holdoff_tmr) | V_NEWINDEX(q->cidx)); 2183 V_NEWTIMER(q->holdoff_tmr) | V_NEWINDEX(q->cidx));
2198 return 0; 2184 return 0;
2199 } 2185 }
2200 if (likely(__netif_rx_schedule_prep(qs->netdev))) 2186 napi_schedule(&qs->napi);
2201 __netif_rx_schedule(qs->netdev);
2202 return 1; 2187 return 1;
2203} 2188}
2204 2189
@@ -2209,8 +2194,7 @@ static inline int handle_responses(struct adapter *adap, struct sge_rspq *q)
2209irqreturn_t t3_sge_intr_msix(int irq, void *cookie) 2194irqreturn_t t3_sge_intr_msix(int irq, void *cookie)
2210{ 2195{
2211 struct sge_qset *qs = cookie; 2196 struct sge_qset *qs = cookie;
2212 const struct port_info *pi = netdev_priv(qs->netdev); 2197 struct adapter *adap = qs->adap;
2213 struct adapter *adap = pi->adapter;
2214 struct sge_rspq *q = &qs->rspq; 2198 struct sge_rspq *q = &qs->rspq;
2215 2199
2216 spin_lock(&q->lock); 2200 spin_lock(&q->lock);
@@ -2229,13 +2213,11 @@ irqreturn_t t3_sge_intr_msix(int irq, void *cookie)
2229irqreturn_t t3_sge_intr_msix_napi(int irq, void *cookie) 2213irqreturn_t t3_sge_intr_msix_napi(int irq, void *cookie)
2230{ 2214{
2231 struct sge_qset *qs = cookie; 2215 struct sge_qset *qs = cookie;
2232 const struct port_info *pi = netdev_priv(qs->netdev);
2233 struct adapter *adap = pi->adapter;
2234 struct sge_rspq *q = &qs->rspq; 2216 struct sge_rspq *q = &qs->rspq;
2235 2217
2236 spin_lock(&q->lock); 2218 spin_lock(&q->lock);
2237 2219
2238 if (handle_responses(adap, q) < 0) 2220 if (handle_responses(qs->adap, q) < 0)
2239 q->unhandled_irqs++; 2221 q->unhandled_irqs++;
2240 spin_unlock(&q->lock); 2222 spin_unlock(&q->lock);
2241 return IRQ_HANDLED; 2223 return IRQ_HANDLED;
@@ -2278,11 +2260,13 @@ static irqreturn_t t3_intr_msi(int irq, void *cookie)
2278 return IRQ_HANDLED; 2260 return IRQ_HANDLED;
2279} 2261}
2280 2262
2281static int rspq_check_napi(struct net_device *dev, struct sge_rspq *q) 2263static int rspq_check_napi(struct sge_qset *qs)
2282{ 2264{
2283 if (!napi_is_scheduled(dev) && is_new_response(&q->desc[q->cidx], q)) { 2265 struct sge_rspq *q = &qs->rspq;
2284 if (likely(__netif_rx_schedule_prep(dev))) 2266
2285 __netif_rx_schedule(dev); 2267 if (!napi_is_scheduled(&qs->napi) &&
2268 is_new_response(&q->desc[q->cidx], q)) {
2269 napi_schedule(&qs->napi);
2286 return 1; 2270 return 1;
2287 } 2271 }
2288 return 0; 2272 return 0;
@@ -2303,10 +2287,9 @@ irqreturn_t t3_intr_msi_napi(int irq, void *cookie)
2303 2287
2304 spin_lock(&q->lock); 2288 spin_lock(&q->lock);
2305 2289
2306 new_packets = rspq_check_napi(adap->sge.qs[0].netdev, q); 2290 new_packets = rspq_check_napi(&adap->sge.qs[0]);
2307 if (adap->params.nports == 2) 2291 if (adap->params.nports == 2)
2308 new_packets += rspq_check_napi(adap->sge.qs[1].netdev, 2292 new_packets += rspq_check_napi(&adap->sge.qs[1]);
2309 &adap->sge.qs[1].rspq);
2310 if (!new_packets && t3_slow_intr_handler(adap) == 0) 2293 if (!new_packets && t3_slow_intr_handler(adap) == 0)
2311 q->unhandled_irqs++; 2294 q->unhandled_irqs++;
2312 2295
@@ -2409,9 +2392,9 @@ static irqreturn_t t3b_intr(int irq, void *cookie)
2409static irqreturn_t t3b_intr_napi(int irq, void *cookie) 2392static irqreturn_t t3b_intr_napi(int irq, void *cookie)
2410{ 2393{
2411 u32 map; 2394 u32 map;
2412 struct net_device *dev;
2413 struct adapter *adap = cookie; 2395 struct adapter *adap = cookie;
2414 struct sge_rspq *q0 = &adap->sge.qs[0].rspq; 2396 struct sge_qset *qs0 = &adap->sge.qs[0];
2397 struct sge_rspq *q0 = &qs0->rspq;
2415 2398
2416 t3_write_reg(adap, A_PL_CLI, 0); 2399 t3_write_reg(adap, A_PL_CLI, 0);
2417 map = t3_read_reg(adap, A_SG_DATA_INTR); 2400 map = t3_read_reg(adap, A_SG_DATA_INTR);
@@ -2424,18 +2407,11 @@ static irqreturn_t t3b_intr_napi(int irq, void *cookie)
2424 if (unlikely(map & F_ERRINTR)) 2407 if (unlikely(map & F_ERRINTR))
2425 t3_slow_intr_handler(adap); 2408 t3_slow_intr_handler(adap);
2426 2409
2427 if (likely(map & 1)) { 2410 if (likely(map & 1))
2428 dev = adap->sge.qs[0].netdev; 2411 napi_schedule(&qs0->napi);
2429
2430 if (likely(__netif_rx_schedule_prep(dev)))
2431 __netif_rx_schedule(dev);
2432 }
2433 if (map & 2) {
2434 dev = adap->sge.qs[1].netdev;
2435 2412
2436 if (likely(__netif_rx_schedule_prep(dev))) 2413 if (map & 2)
2437 __netif_rx_schedule(dev); 2414 napi_schedule(&adap->sge.qs[1].napi);
2438 }
2439 2415
2440 spin_unlock(&q0->lock); 2416 spin_unlock(&q0->lock);
2441 return IRQ_HANDLED; 2417 return IRQ_HANDLED;
@@ -2514,8 +2490,7 @@ static void sge_timer_cb(unsigned long data)
2514{ 2490{
2515 spinlock_t *lock; 2491 spinlock_t *lock;
2516 struct sge_qset *qs = (struct sge_qset *)data; 2492 struct sge_qset *qs = (struct sge_qset *)data;
2517 const struct port_info *pi = netdev_priv(qs->netdev); 2493 struct adapter *adap = qs->adap;
2518 struct adapter *adap = pi->adapter;
2519 2494
2520 if (spin_trylock(&qs->txq[TXQ_ETH].lock)) { 2495 if (spin_trylock(&qs->txq[TXQ_ETH].lock)) {
2521 reclaim_completed_tx(adap, &qs->txq[TXQ_ETH]); 2496 reclaim_completed_tx(adap, &qs->txq[TXQ_ETH]);
@@ -2526,9 +2501,9 @@ static void sge_timer_cb(unsigned long data)
2526 spin_unlock(&qs->txq[TXQ_OFLD].lock); 2501 spin_unlock(&qs->txq[TXQ_OFLD].lock);
2527 } 2502 }
2528 lock = (adap->flags & USING_MSIX) ? &qs->rspq.lock : 2503 lock = (adap->flags & USING_MSIX) ? &qs->rspq.lock :
2529 &adap->sge.qs[0].rspq.lock; 2504 &adap->sge.qs[0].rspq.lock;
2530 if (spin_trylock_irq(lock)) { 2505 if (spin_trylock_irq(lock)) {
2531 if (!napi_is_scheduled(qs->netdev)) { 2506 if (!napi_is_scheduled(&qs->napi)) {
2532 u32 status = t3_read_reg(adap, A_SG_RSPQ_FL_STATUS); 2507 u32 status = t3_read_reg(adap, A_SG_RSPQ_FL_STATUS);
2533 2508
2534 if (qs->fl[0].credits < qs->fl[0].size) 2509 if (qs->fl[0].credits < qs->fl[0].size)
@@ -2562,12 +2537,9 @@ static void sge_timer_cb(unsigned long data)
2562 */ 2537 */
2563void t3_update_qset_coalesce(struct sge_qset *qs, const struct qset_params *p) 2538void t3_update_qset_coalesce(struct sge_qset *qs, const struct qset_params *p)
2564{ 2539{
2565 if (!qs->netdev)
2566 return;
2567
2568 qs->rspq.holdoff_tmr = max(p->coalesce_usecs * 10, 1U);/* can't be 0 */ 2540 qs->rspq.holdoff_tmr = max(p->coalesce_usecs * 10, 1U);/* can't be 0 */
2569 qs->rspq.polling = p->polling; 2541 qs->rspq.polling = p->polling;
2570 qs->netdev->poll = p->polling ? napi_rx_handler : ofld_poll; 2542 qs->napi.poll = p->polling ? napi_rx_handler : ofld_poll;
2571} 2543}
2572 2544
2573/** 2545/**
@@ -2587,7 +2559,7 @@ void t3_update_qset_coalesce(struct sge_qset *qs, const struct qset_params *p)
2587 */ 2559 */
2588int t3_sge_alloc_qset(struct adapter *adapter, unsigned int id, int nports, 2560int t3_sge_alloc_qset(struct adapter *adapter, unsigned int id, int nports,
2589 int irq_vec_idx, const struct qset_params *p, 2561 int irq_vec_idx, const struct qset_params *p,
2590 int ntxq, struct net_device *netdev) 2562 int ntxq, struct net_device *dev)
2591{ 2563{
2592 int i, ret = -ENOMEM; 2564 int i, ret = -ENOMEM;
2593 struct sge_qset *q = &adapter->sge.qs[id]; 2565 struct sge_qset *q = &adapter->sge.qs[id];
@@ -2708,16 +2680,10 @@ int t3_sge_alloc_qset(struct adapter *adapter, unsigned int id, int nports,
2708 } 2680 }
2709 2681
2710 spin_unlock(&adapter->sge.reg_lock); 2682 spin_unlock(&adapter->sge.reg_lock);
2711 q->netdev = netdev;
2712 t3_update_qset_coalesce(q, p);
2713 2683
2714 /* 2684 q->adap = adapter;
2715 * We use atalk_ptr as a backpointer to a qset. In case a device is 2685 q->netdev = dev;
2716 * associated with multiple queue sets only the first one sets 2686 t3_update_qset_coalesce(q, p);
2717 * atalk_ptr.
2718 */
2719 if (netdev->atalk_ptr == NULL)
2720 netdev->atalk_ptr = q;
2721 2687
2722 refill_fl(adapter, &q->fl[0], q->fl[0].size, GFP_KERNEL); 2688 refill_fl(adapter, &q->fl[0], q->fl[0].size, GFP_KERNEL);
2723 refill_fl(adapter, &q->fl[1], q->fl[1].size, GFP_KERNEL); 2689 refill_fl(adapter, &q->fl[1], q->fl[1].size, GFP_KERNEL);
diff --git a/drivers/net/e100.c b/drivers/net/e100.c
index 280313b9b069..e25f5ec2b279 100644
--- a/drivers/net/e100.c
+++ b/drivers/net/e100.c
@@ -539,6 +539,7 @@ struct nic {
539 struct csr __iomem *csr; 539 struct csr __iomem *csr;
540 enum scb_cmd_lo cuc_cmd; 540 enum scb_cmd_lo cuc_cmd;
541 unsigned int cbs_avail; 541 unsigned int cbs_avail;
542 struct napi_struct napi;
542 struct cb *cbs; 543 struct cb *cbs;
543 struct cb *cb_to_use; 544 struct cb *cb_to_use;
544 struct cb *cb_to_send; 545 struct cb *cb_to_send;
@@ -1974,35 +1975,31 @@ static irqreturn_t e100_intr(int irq, void *dev_id)
1974 if(stat_ack & stat_ack_rnr) 1975 if(stat_ack & stat_ack_rnr)
1975 nic->ru_running = RU_SUSPENDED; 1976 nic->ru_running = RU_SUSPENDED;
1976 1977
1977 if(likely(netif_rx_schedule_prep(netdev))) { 1978 if(likely(netif_rx_schedule_prep(netdev, &nic->napi))) {
1978 e100_disable_irq(nic); 1979 e100_disable_irq(nic);
1979 __netif_rx_schedule(netdev); 1980 __netif_rx_schedule(netdev, &nic->napi);
1980 } 1981 }
1981 1982
1982 return IRQ_HANDLED; 1983 return IRQ_HANDLED;
1983} 1984}
1984 1985
1985static int e100_poll(struct net_device *netdev, int *budget) 1986static int e100_poll(struct napi_struct *napi, int budget)
1986{ 1987{
1987 struct nic *nic = netdev_priv(netdev); 1988 struct nic *nic = container_of(napi, struct nic, napi);
1988 unsigned int work_to_do = min(netdev->quota, *budget); 1989 struct net_device *netdev = nic->netdev;
1989 unsigned int work_done = 0; 1990 int work_done = 0;
1990 int tx_cleaned; 1991 int tx_cleaned;
1991 1992
1992 e100_rx_clean(nic, &work_done, work_to_do); 1993 e100_rx_clean(nic, &work_done, budget);
1993 tx_cleaned = e100_tx_clean(nic); 1994 tx_cleaned = e100_tx_clean(nic);
1994 1995
1995 /* If no Rx and Tx cleanup work was done, exit polling mode. */ 1996 /* If no Rx and Tx cleanup work was done, exit polling mode. */
1996 if((!tx_cleaned && (work_done == 0)) || !netif_running(netdev)) { 1997 if((!tx_cleaned && (work_done == 0)) || !netif_running(netdev)) {
1997 netif_rx_complete(netdev); 1998 netif_rx_complete(netdev, napi);
1998 e100_enable_irq(nic); 1999 e100_enable_irq(nic);
1999 return 0;
2000 } 2000 }
2001 2001
2002 *budget -= work_done; 2002 return work_done;
2003 netdev->quota -= work_done;
2004
2005 return 1;
2006} 2003}
2007 2004
2008#ifdef CONFIG_NET_POLL_CONTROLLER 2005#ifdef CONFIG_NET_POLL_CONTROLLER
@@ -2071,7 +2068,7 @@ static int e100_up(struct nic *nic)
2071 nic->netdev->name, nic->netdev))) 2068 nic->netdev->name, nic->netdev)))
2072 goto err_no_irq; 2069 goto err_no_irq;
2073 netif_wake_queue(nic->netdev); 2070 netif_wake_queue(nic->netdev);
2074 netif_poll_enable(nic->netdev); 2071 napi_enable(&nic->napi);
2075 /* enable ints _after_ enabling poll, preventing a race between 2072 /* enable ints _after_ enabling poll, preventing a race between
2076 * disable ints+schedule */ 2073 * disable ints+schedule */
2077 e100_enable_irq(nic); 2074 e100_enable_irq(nic);
@@ -2089,7 +2086,7 @@ err_rx_clean_list:
2089static void e100_down(struct nic *nic) 2086static void e100_down(struct nic *nic)
2090{ 2087{
2091 /* wait here for poll to complete */ 2088 /* wait here for poll to complete */
2092 netif_poll_disable(nic->netdev); 2089 napi_disable(&nic->napi);
2093 netif_stop_queue(nic->netdev); 2090 netif_stop_queue(nic->netdev);
2094 e100_hw_reset(nic); 2091 e100_hw_reset(nic);
2095 free_irq(nic->pdev->irq, nic->netdev); 2092 free_irq(nic->pdev->irq, nic->netdev);
@@ -2572,14 +2569,13 @@ static int __devinit e100_probe(struct pci_dev *pdev,
2572 SET_ETHTOOL_OPS(netdev, &e100_ethtool_ops); 2569 SET_ETHTOOL_OPS(netdev, &e100_ethtool_ops);
2573 netdev->tx_timeout = e100_tx_timeout; 2570 netdev->tx_timeout = e100_tx_timeout;
2574 netdev->watchdog_timeo = E100_WATCHDOG_PERIOD; 2571 netdev->watchdog_timeo = E100_WATCHDOG_PERIOD;
2575 netdev->poll = e100_poll;
2576 netdev->weight = E100_NAPI_WEIGHT;
2577#ifdef CONFIG_NET_POLL_CONTROLLER 2572#ifdef CONFIG_NET_POLL_CONTROLLER
2578 netdev->poll_controller = e100_netpoll; 2573 netdev->poll_controller = e100_netpoll;
2579#endif 2574#endif
2580 strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1); 2575 strncpy(netdev->name, pci_name(pdev), sizeof(netdev->name) - 1);
2581 2576
2582 nic = netdev_priv(netdev); 2577 nic = netdev_priv(netdev);
2578 netif_napi_add(netdev, &nic->napi, e100_poll, E100_NAPI_WEIGHT);
2583 nic->netdev = netdev; 2579 nic->netdev = netdev;
2584 nic->pdev = pdev; 2580 nic->pdev = pdev;
2585 nic->msg_enable = (1 << debug) - 1; 2581 nic->msg_enable = (1 << debug) - 1;
@@ -2733,7 +2729,7 @@ static int e100_suspend(struct pci_dev *pdev, pm_message_t state)
2733 struct nic *nic = netdev_priv(netdev); 2729 struct nic *nic = netdev_priv(netdev);
2734 2730
2735 if (netif_running(netdev)) 2731 if (netif_running(netdev))
2736 netif_poll_disable(nic->netdev); 2732 napi_disable(&nic->napi);
2737 del_timer_sync(&nic->watchdog); 2733 del_timer_sync(&nic->watchdog);
2738 netif_carrier_off(nic->netdev); 2734 netif_carrier_off(nic->netdev);
2739 netif_device_detach(netdev); 2735 netif_device_detach(netdev);
@@ -2779,7 +2775,7 @@ static void e100_shutdown(struct pci_dev *pdev)
2779 struct nic *nic = netdev_priv(netdev); 2775 struct nic *nic = netdev_priv(netdev);
2780 2776
2781 if (netif_running(netdev)) 2777 if (netif_running(netdev))
2782 netif_poll_disable(nic->netdev); 2778 napi_disable(&nic->napi);
2783 del_timer_sync(&nic->watchdog); 2779 del_timer_sync(&nic->watchdog);
2784 netif_carrier_off(nic->netdev); 2780 netif_carrier_off(nic->netdev);
2785 2781
@@ -2804,12 +2800,13 @@ static void e100_shutdown(struct pci_dev *pdev)
2804static pci_ers_result_t e100_io_error_detected(struct pci_dev *pdev, pci_channel_state_t state) 2800static pci_ers_result_t e100_io_error_detected(struct pci_dev *pdev, pci_channel_state_t state)
2805{ 2801{
2806 struct net_device *netdev = pci_get_drvdata(pdev); 2802 struct net_device *netdev = pci_get_drvdata(pdev);
2803 struct nic *nic = netdev_priv(netdev);
2807 2804
2808 /* Similar to calling e100_down(), but avoids adpater I/O. */ 2805 /* Similar to calling e100_down(), but avoids adpater I/O. */
2809 netdev->stop(netdev); 2806 netdev->stop(netdev);
2810 2807
2811 /* Detach; put netif into state similar to hotplug unplug. */ 2808 /* Detach; put netif into state similar to hotplug unplug. */
2812 netif_poll_enable(netdev); 2809 napi_enable(&nic->napi);
2813 netif_device_detach(netdev); 2810 netif_device_detach(netdev);
2814 pci_disable_device(pdev); 2811 pci_disable_device(pdev);
2815 2812
diff --git a/drivers/net/e1000/e1000.h b/drivers/net/e1000/e1000.h
index 16a6edfeba41..781ed9968489 100644
--- a/drivers/net/e1000/e1000.h
+++ b/drivers/net/e1000/e1000.h
@@ -300,6 +300,7 @@ struct e1000_adapter {
300 int cleaned_count); 300 int cleaned_count);
301 struct e1000_rx_ring *rx_ring; /* One per active queue */ 301 struct e1000_rx_ring *rx_ring; /* One per active queue */
302#ifdef CONFIG_E1000_NAPI 302#ifdef CONFIG_E1000_NAPI
303 struct napi_struct napi;
303 struct net_device *polling_netdev; /* One per active queue */ 304 struct net_device *polling_netdev; /* One per active queue */
304#endif 305#endif
305 int num_tx_queues; 306 int num_tx_queues;
diff --git a/drivers/net/e1000/e1000_main.c b/drivers/net/e1000/e1000_main.c
index e7c8951f47fa..723568d6e44a 100644
--- a/drivers/net/e1000/e1000_main.c
+++ b/drivers/net/e1000/e1000_main.c
@@ -166,7 +166,7 @@ static irqreturn_t e1000_intr_msi(int irq, void *data);
166static boolean_t e1000_clean_tx_irq(struct e1000_adapter *adapter, 166static boolean_t e1000_clean_tx_irq(struct e1000_adapter *adapter,
167 struct e1000_tx_ring *tx_ring); 167 struct e1000_tx_ring *tx_ring);
168#ifdef CONFIG_E1000_NAPI 168#ifdef CONFIG_E1000_NAPI
169static int e1000_clean(struct net_device *poll_dev, int *budget); 169static int e1000_clean(struct napi_struct *napi, int budget);
170static boolean_t e1000_clean_rx_irq(struct e1000_adapter *adapter, 170static boolean_t e1000_clean_rx_irq(struct e1000_adapter *adapter,
171 struct e1000_rx_ring *rx_ring, 171 struct e1000_rx_ring *rx_ring,
172 int *work_done, int work_to_do); 172 int *work_done, int work_to_do);
@@ -545,7 +545,7 @@ int e1000_up(struct e1000_adapter *adapter)
545 clear_bit(__E1000_DOWN, &adapter->flags); 545 clear_bit(__E1000_DOWN, &adapter->flags);
546 546
547#ifdef CONFIG_E1000_NAPI 547#ifdef CONFIG_E1000_NAPI
548 netif_poll_enable(adapter->netdev); 548 napi_enable(&adapter->napi);
549#endif 549#endif
550 e1000_irq_enable(adapter); 550 e1000_irq_enable(adapter);
551 551
@@ -634,7 +634,7 @@ e1000_down(struct e1000_adapter *adapter)
634 set_bit(__E1000_DOWN, &adapter->flags); 634 set_bit(__E1000_DOWN, &adapter->flags);
635 635
636#ifdef CONFIG_E1000_NAPI 636#ifdef CONFIG_E1000_NAPI
637 netif_poll_disable(netdev); 637 napi_disable(&adapter->napi);
638#endif 638#endif
639 e1000_irq_disable(adapter); 639 e1000_irq_disable(adapter);
640 640
@@ -936,8 +936,7 @@ e1000_probe(struct pci_dev *pdev,
936 netdev->tx_timeout = &e1000_tx_timeout; 936 netdev->tx_timeout = &e1000_tx_timeout;
937 netdev->watchdog_timeo = 5 * HZ; 937 netdev->watchdog_timeo = 5 * HZ;
938#ifdef CONFIG_E1000_NAPI 938#ifdef CONFIG_E1000_NAPI
939 netdev->poll = &e1000_clean; 939 netif_napi_add(netdev, &adapter->napi, e1000_clean, 64);
940 netdev->weight = 64;
941#endif 940#endif
942 netdev->vlan_rx_register = e1000_vlan_rx_register; 941 netdev->vlan_rx_register = e1000_vlan_rx_register;
943 netdev->vlan_rx_add_vid = e1000_vlan_rx_add_vid; 942 netdev->vlan_rx_add_vid = e1000_vlan_rx_add_vid;
@@ -1151,9 +1150,6 @@ e1000_probe(struct pci_dev *pdev,
1151 /* tell the stack to leave us alone until e1000_open() is called */ 1150 /* tell the stack to leave us alone until e1000_open() is called */
1152 netif_carrier_off(netdev); 1151 netif_carrier_off(netdev);
1153 netif_stop_queue(netdev); 1152 netif_stop_queue(netdev);
1154#ifdef CONFIG_E1000_NAPI
1155 netif_poll_disable(netdev);
1156#endif
1157 1153
1158 strcpy(netdev->name, "eth%d"); 1154 strcpy(netdev->name, "eth%d");
1159 if ((err = register_netdev(netdev))) 1155 if ((err = register_netdev(netdev)))
@@ -1222,12 +1218,13 @@ e1000_remove(struct pci_dev *pdev)
1222 * would have already happened in close and is redundant. */ 1218 * would have already happened in close and is redundant. */
1223 e1000_release_hw_control(adapter); 1219 e1000_release_hw_control(adapter);
1224 1220
1225 unregister_netdev(netdev);
1226#ifdef CONFIG_E1000_NAPI 1221#ifdef CONFIG_E1000_NAPI
1227 for (i = 0; i < adapter->num_rx_queues; i++) 1222 for (i = 0; i < adapter->num_rx_queues; i++)
1228 dev_put(&adapter->polling_netdev[i]); 1223 dev_put(&adapter->polling_netdev[i]);
1229#endif 1224#endif
1230 1225
1226 unregister_netdev(netdev);
1227
1231 if (!e1000_check_phy_reset_block(&adapter->hw)) 1228 if (!e1000_check_phy_reset_block(&adapter->hw))
1232 e1000_phy_hw_reset(&adapter->hw); 1229 e1000_phy_hw_reset(&adapter->hw);
1233 1230
@@ -1325,8 +1322,6 @@ e1000_sw_init(struct e1000_adapter *adapter)
1325#ifdef CONFIG_E1000_NAPI 1322#ifdef CONFIG_E1000_NAPI
1326 for (i = 0; i < adapter->num_rx_queues; i++) { 1323 for (i = 0; i < adapter->num_rx_queues; i++) {
1327 adapter->polling_netdev[i].priv = adapter; 1324 adapter->polling_netdev[i].priv = adapter;
1328 adapter->polling_netdev[i].poll = &e1000_clean;
1329 adapter->polling_netdev[i].weight = 64;
1330 dev_hold(&adapter->polling_netdev[i]); 1325 dev_hold(&adapter->polling_netdev[i]);
1331 set_bit(__LINK_STATE_START, &adapter->polling_netdev[i].state); 1326 set_bit(__LINK_STATE_START, &adapter->polling_netdev[i].state);
1332 } 1327 }
@@ -1443,7 +1438,7 @@ e1000_open(struct net_device *netdev)
1443 clear_bit(__E1000_DOWN, &adapter->flags); 1438 clear_bit(__E1000_DOWN, &adapter->flags);
1444 1439
1445#ifdef CONFIG_E1000_NAPI 1440#ifdef CONFIG_E1000_NAPI
1446 netif_poll_enable(netdev); 1441 napi_enable(&adapter->napi);
1447#endif 1442#endif
1448 1443
1449 e1000_irq_enable(adapter); 1444 e1000_irq_enable(adapter);
@@ -3786,12 +3781,12 @@ e1000_intr_msi(int irq, void *data)
3786 } 3781 }
3787 3782
3788#ifdef CONFIG_E1000_NAPI 3783#ifdef CONFIG_E1000_NAPI
3789 if (likely(netif_rx_schedule_prep(netdev))) { 3784 if (likely(netif_rx_schedule_prep(netdev, &adapter->napi))) {
3790 adapter->total_tx_bytes = 0; 3785 adapter->total_tx_bytes = 0;
3791 adapter->total_tx_packets = 0; 3786 adapter->total_tx_packets = 0;
3792 adapter->total_rx_bytes = 0; 3787 adapter->total_rx_bytes = 0;
3793 adapter->total_rx_packets = 0; 3788 adapter->total_rx_packets = 0;
3794 __netif_rx_schedule(netdev); 3789 __netif_rx_schedule(netdev, &adapter->napi);
3795 } else 3790 } else
3796 e1000_irq_enable(adapter); 3791 e1000_irq_enable(adapter);
3797#else 3792#else
@@ -3871,12 +3866,12 @@ e1000_intr(int irq, void *data)
3871 E1000_WRITE_REG(hw, IMC, ~0); 3866 E1000_WRITE_REG(hw, IMC, ~0);
3872 E1000_WRITE_FLUSH(hw); 3867 E1000_WRITE_FLUSH(hw);
3873 } 3868 }
3874 if (likely(netif_rx_schedule_prep(netdev))) { 3869 if (likely(netif_rx_schedule_prep(netdev, &adapter->napi))) {
3875 adapter->total_tx_bytes = 0; 3870 adapter->total_tx_bytes = 0;
3876 adapter->total_tx_packets = 0; 3871 adapter->total_tx_packets = 0;
3877 adapter->total_rx_bytes = 0; 3872 adapter->total_rx_bytes = 0;
3878 adapter->total_rx_packets = 0; 3873 adapter->total_rx_packets = 0;
3879 __netif_rx_schedule(netdev); 3874 __netif_rx_schedule(netdev, &adapter->napi);
3880 } else 3875 } else
3881 /* this really should not happen! if it does it is basically a 3876 /* this really should not happen! if it does it is basically a
3882 * bug, but not a hard error, so enable ints and continue */ 3877 * bug, but not a hard error, so enable ints and continue */
@@ -3924,10 +3919,10 @@ e1000_intr(int irq, void *data)
3924 **/ 3919 **/
3925 3920
3926static int 3921static int
3927e1000_clean(struct net_device *poll_dev, int *budget) 3922e1000_clean(struct napi_struct *napi, int budget)
3928{ 3923{
3929 struct e1000_adapter *adapter; 3924 struct e1000_adapter *adapter = container_of(napi, struct e1000_adapter, napi);
3930 int work_to_do = min(*budget, poll_dev->quota); 3925 struct net_device *poll_dev = adapter->netdev;
3931 int tx_cleaned = 0, work_done = 0; 3926 int tx_cleaned = 0, work_done = 0;
3932 3927
3933 /* Must NOT use netdev_priv macro here. */ 3928 /* Must NOT use netdev_priv macro here. */
@@ -3948,23 +3943,19 @@ e1000_clean(struct net_device *poll_dev, int *budget)
3948 } 3943 }
3949 3944
3950 adapter->clean_rx(adapter, &adapter->rx_ring[0], 3945 adapter->clean_rx(adapter, &adapter->rx_ring[0],
3951 &work_done, work_to_do); 3946 &work_done, budget);
3952
3953 *budget -= work_done;
3954 poll_dev->quota -= work_done;
3955 3947
3956 /* If no Tx and not enough Rx work done, exit the polling mode */ 3948 /* If no Tx and not enough Rx work done, exit the polling mode */
3957 if ((!tx_cleaned && (work_done == 0)) || 3949 if ((!tx_cleaned && (work_done < budget)) ||
3958 !netif_running(poll_dev)) { 3950 !netif_running(poll_dev)) {
3959quit_polling: 3951quit_polling:
3960 if (likely(adapter->itr_setting & 3)) 3952 if (likely(adapter->itr_setting & 3))
3961 e1000_set_itr(adapter); 3953 e1000_set_itr(adapter);
3962 netif_rx_complete(poll_dev); 3954 netif_rx_complete(poll_dev, napi);
3963 e1000_irq_enable(adapter); 3955 e1000_irq_enable(adapter);
3964 return 0;
3965 } 3956 }
3966 3957
3967 return 1; 3958 return work_done;
3968} 3959}
3969 3960
3970#endif 3961#endif
diff --git a/drivers/net/ehea/ehea.h b/drivers/net/ehea/ehea.h
index 8d58be56f4e3..a154681165b9 100644
--- a/drivers/net/ehea/ehea.h
+++ b/drivers/net/ehea/ehea.h
@@ -351,6 +351,7 @@ struct ehea_q_skb_arr {
351 * Port resources 351 * Port resources
352 */ 352 */
353struct ehea_port_res { 353struct ehea_port_res {
354 struct napi_struct napi;
354 struct port_stats p_stats; 355 struct port_stats p_stats;
355 struct ehea_mr send_mr; /* send memory region */ 356 struct ehea_mr send_mr; /* send memory region */
356 struct ehea_mr recv_mr; /* receive memory region */ 357 struct ehea_mr recv_mr; /* receive memory region */
@@ -362,7 +363,6 @@ struct ehea_port_res {
362 struct ehea_cq *send_cq; 363 struct ehea_cq *send_cq;
363 struct ehea_cq *recv_cq; 364 struct ehea_cq *recv_cq;
364 struct ehea_eq *eq; 365 struct ehea_eq *eq;
365 struct net_device *d_netdev;
366 struct ehea_q_skb_arr rq1_skba; 366 struct ehea_q_skb_arr rq1_skba;
367 struct ehea_q_skb_arr rq2_skba; 367 struct ehea_q_skb_arr rq2_skba;
368 struct ehea_q_skb_arr rq3_skba; 368 struct ehea_q_skb_arr rq3_skba;
diff --git a/drivers/net/ehea/ehea_main.c b/drivers/net/ehea/ehea_main.c
index 717b12984d10..5ebd545ab04e 100644
--- a/drivers/net/ehea/ehea_main.c
+++ b/drivers/net/ehea/ehea_main.c
@@ -393,9 +393,9 @@ static int ehea_treat_poll_error(struct ehea_port_res *pr, int rq,
393 return 0; 393 return 0;
394} 394}
395 395
396static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev, 396static int ehea_proc_rwqes(struct net_device *dev,
397 struct ehea_port_res *pr, 397 struct ehea_port_res *pr,
398 int *budget) 398 int budget)
399{ 399{
400 struct ehea_port *port = pr->port; 400 struct ehea_port *port = pr->port;
401 struct ehea_qp *qp = pr->qp; 401 struct ehea_qp *qp = pr->qp;
@@ -408,18 +408,16 @@ static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev,
408 int skb_arr_rq2_len = pr->rq2_skba.len; 408 int skb_arr_rq2_len = pr->rq2_skba.len;
409 int skb_arr_rq3_len = pr->rq3_skba.len; 409 int skb_arr_rq3_len = pr->rq3_skba.len;
410 int processed, processed_rq1, processed_rq2, processed_rq3; 410 int processed, processed_rq1, processed_rq2, processed_rq3;
411 int wqe_index, last_wqe_index, rq, my_quota, port_reset; 411 int wqe_index, last_wqe_index, rq, port_reset;
412 412
413 processed = processed_rq1 = processed_rq2 = processed_rq3 = 0; 413 processed = processed_rq1 = processed_rq2 = processed_rq3 = 0;
414 last_wqe_index = 0; 414 last_wqe_index = 0;
415 my_quota = min(*budget, dev->quota);
416 415
417 cqe = ehea_poll_rq1(qp, &wqe_index); 416 cqe = ehea_poll_rq1(qp, &wqe_index);
418 while ((my_quota > 0) && cqe) { 417 while ((processed < budget) && cqe) {
419 ehea_inc_rq1(qp); 418 ehea_inc_rq1(qp);
420 processed_rq1++; 419 processed_rq1++;
421 processed++; 420 processed++;
422 my_quota--;
423 if (netif_msg_rx_status(port)) 421 if (netif_msg_rx_status(port))
424 ehea_dump(cqe, sizeof(*cqe), "CQE"); 422 ehea_dump(cqe, sizeof(*cqe), "CQE");
425 423
@@ -434,14 +432,14 @@ static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev,
434 if (netif_msg_rx_err(port)) 432 if (netif_msg_rx_err(port))
435 ehea_error("LL rq1: skb=NULL"); 433 ehea_error("LL rq1: skb=NULL");
436 434
437 skb = netdev_alloc_skb(port->netdev, 435 skb = netdev_alloc_skb(dev,
438 EHEA_L_PKT_SIZE); 436 EHEA_L_PKT_SIZE);
439 if (!skb) 437 if (!skb)
440 break; 438 break;
441 } 439 }
442 skb_copy_to_linear_data(skb, ((char*)cqe) + 64, 440 skb_copy_to_linear_data(skb, ((char*)cqe) + 64,
443 cqe->num_bytes_transfered - 4); 441 cqe->num_bytes_transfered - 4);
444 ehea_fill_skb(port->netdev, skb, cqe); 442 ehea_fill_skb(dev, skb, cqe);
445 } else if (rq == 2) { /* RQ2 */ 443 } else if (rq == 2) { /* RQ2 */
446 skb = get_skb_by_index(skb_arr_rq2, 444 skb = get_skb_by_index(skb_arr_rq2,
447 skb_arr_rq2_len, cqe); 445 skb_arr_rq2_len, cqe);
@@ -450,7 +448,7 @@ static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev,
450 ehea_error("rq2: skb=NULL"); 448 ehea_error("rq2: skb=NULL");
451 break; 449 break;
452 } 450 }
453 ehea_fill_skb(port->netdev, skb, cqe); 451 ehea_fill_skb(dev, skb, cqe);
454 processed_rq2++; 452 processed_rq2++;
455 } else { /* RQ3 */ 453 } else { /* RQ3 */
456 skb = get_skb_by_index(skb_arr_rq3, 454 skb = get_skb_by_index(skb_arr_rq3,
@@ -460,7 +458,7 @@ static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev,
460 ehea_error("rq3: skb=NULL"); 458 ehea_error("rq3: skb=NULL");
461 break; 459 break;
462 } 460 }
463 ehea_fill_skb(port->netdev, skb, cqe); 461 ehea_fill_skb(dev, skb, cqe);
464 processed_rq3++; 462 processed_rq3++;
465 } 463 }
466 464
@@ -471,7 +469,7 @@ static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev,
471 else 469 else
472 netif_receive_skb(skb); 470 netif_receive_skb(skb);
473 471
474 port->netdev->last_rx = jiffies; 472 dev->last_rx = jiffies;
475 } else { 473 } else {
476 pr->p_stats.poll_receive_errors++; 474 pr->p_stats.poll_receive_errors++;
477 port_reset = ehea_treat_poll_error(pr, rq, cqe, 475 port_reset = ehea_treat_poll_error(pr, rq, cqe,
@@ -484,14 +482,12 @@ static struct ehea_cqe *ehea_proc_rwqes(struct net_device *dev,
484 } 482 }
485 483
486 pr->rx_packets += processed; 484 pr->rx_packets += processed;
487 *budget -= processed;
488 485
489 ehea_refill_rq1(pr, last_wqe_index, processed_rq1); 486 ehea_refill_rq1(pr, last_wqe_index, processed_rq1);
490 ehea_refill_rq2(pr, processed_rq2); 487 ehea_refill_rq2(pr, processed_rq2);
491 ehea_refill_rq3(pr, processed_rq3); 488 ehea_refill_rq3(pr, processed_rq3);
492 489
493 cqe = ehea_poll_rq1(qp, &wqe_index); 490 return processed;
494 return cqe;
495} 491}
496 492
497static struct ehea_cqe *ehea_proc_cqes(struct ehea_port_res *pr, int my_quota) 493static struct ehea_cqe *ehea_proc_cqes(struct ehea_port_res *pr, int my_quota)
@@ -554,22 +550,27 @@ static struct ehea_cqe *ehea_proc_cqes(struct ehea_port_res *pr, int my_quota)
554} 550}
555 551
556#define EHEA_NAPI_POLL_NUM_BEFORE_IRQ 16 552#define EHEA_NAPI_POLL_NUM_BEFORE_IRQ 16
553#define EHEA_POLL_MAX_CQES 65535
557 554
558static int ehea_poll(struct net_device *dev, int *budget) 555static int ehea_poll(struct napi_struct *napi, int budget)
559{ 556{
560 struct ehea_port_res *pr = dev->priv; 557 struct ehea_port_res *pr = container_of(napi, struct ehea_port_res, napi);
558 struct net_device *dev = pr->port->netdev;
561 struct ehea_cqe *cqe; 559 struct ehea_cqe *cqe;
562 struct ehea_cqe *cqe_skb = NULL; 560 struct ehea_cqe *cqe_skb = NULL;
563 int force_irq, wqe_index; 561 int force_irq, wqe_index;
564 562 int rx = 0;
565 cqe = ehea_poll_rq1(pr->qp, &wqe_index);
566 cqe_skb = ehea_poll_cq(pr->send_cq);
567 563
568 force_irq = (pr->poll_counter > EHEA_NAPI_POLL_NUM_BEFORE_IRQ); 564 force_irq = (pr->poll_counter > EHEA_NAPI_POLL_NUM_BEFORE_IRQ);
565 cqe_skb = ehea_proc_cqes(pr, EHEA_POLL_MAX_CQES);
566
567 if (!force_irq)
568 rx += ehea_proc_rwqes(dev, pr, budget - rx);
569 569
570 if ((!cqe && !cqe_skb) || force_irq) { 570 while ((rx != budget) || force_irq) {
571 pr->poll_counter = 0; 571 pr->poll_counter = 0;
572 netif_rx_complete(dev); 572 force_irq = 0;
573 netif_rx_complete(dev, napi);
573 ehea_reset_cq_ep(pr->recv_cq); 574 ehea_reset_cq_ep(pr->recv_cq);
574 ehea_reset_cq_ep(pr->send_cq); 575 ehea_reset_cq_ep(pr->send_cq);
575 ehea_reset_cq_n1(pr->recv_cq); 576 ehea_reset_cq_n1(pr->recv_cq);
@@ -578,43 +579,35 @@ static int ehea_poll(struct net_device *dev, int *budget)
578 cqe_skb = ehea_poll_cq(pr->send_cq); 579 cqe_skb = ehea_poll_cq(pr->send_cq);
579 580
580 if (!cqe && !cqe_skb) 581 if (!cqe && !cqe_skb)
581 return 0; 582 return rx;
582 583
583 if (!netif_rx_reschedule(dev, dev->quota)) 584 if (!netif_rx_reschedule(dev, napi))
584 return 0; 585 return rx;
585 }
586
587 cqe = ehea_proc_rwqes(dev, pr, budget);
588 cqe_skb = ehea_proc_cqes(pr, 300);
589 586
590 if (cqe || cqe_skb) 587 cqe_skb = ehea_proc_cqes(pr, EHEA_POLL_MAX_CQES);
591 pr->poll_counter++; 588 rx += ehea_proc_rwqes(dev, pr, budget - rx);
589 }
592 590
593 return 1; 591 pr->poll_counter++;
592 return rx;
594} 593}
595 594
596#ifdef CONFIG_NET_POLL_CONTROLLER 595#ifdef CONFIG_NET_POLL_CONTROLLER
597static void ehea_netpoll(struct net_device *dev) 596static void ehea_netpoll(struct net_device *dev)
598{ 597{
599 struct ehea_port *port = netdev_priv(dev); 598 struct ehea_port *port = netdev_priv(dev);
599 int i;
600 600
601 netif_rx_schedule(port->port_res[0].d_netdev); 601 for (i = 0; i < port->num_def_qps; i++)
602 netif_rx_schedule(dev, &port->port_res[i].napi);
602} 603}
603#endif 604#endif
604 605
605static int ehea_poll_firstqueue(struct net_device *dev, int *budget)
606{
607 struct ehea_port *port = netdev_priv(dev);
608 struct net_device *d_dev = port->port_res[0].d_netdev;
609
610 return ehea_poll(d_dev, budget);
611}
612
613static irqreturn_t ehea_recv_irq_handler(int irq, void *param) 606static irqreturn_t ehea_recv_irq_handler(int irq, void *param)
614{ 607{
615 struct ehea_port_res *pr = param; 608 struct ehea_port_res *pr = param;
616 609
617 netif_rx_schedule(pr->d_netdev); 610 netif_rx_schedule(pr->port->netdev, &pr->napi);
618 611
619 return IRQ_HANDLED; 612 return IRQ_HANDLED;
620} 613}
@@ -1236,14 +1229,7 @@ static int ehea_init_port_res(struct ehea_port *port, struct ehea_port_res *pr,
1236 1229
1237 kfree(init_attr); 1230 kfree(init_attr);
1238 1231
1239 pr->d_netdev = alloc_netdev(0, "", ether_setup); 1232 netif_napi_add(pr->port->netdev, &pr->napi, ehea_poll, 64);
1240 if (!pr->d_netdev)
1241 goto out_free;
1242 pr->d_netdev->priv = pr;
1243 pr->d_netdev->weight = 64;
1244 pr->d_netdev->poll = ehea_poll;
1245 set_bit(__LINK_STATE_START, &pr->d_netdev->state);
1246 strcpy(pr->d_netdev->name, port->netdev->name);
1247 1233
1248 ret = 0; 1234 ret = 0;
1249 goto out; 1235 goto out;
@@ -1266,8 +1252,6 @@ static int ehea_clean_portres(struct ehea_port *port, struct ehea_port_res *pr)
1266{ 1252{
1267 int ret, i; 1253 int ret, i;
1268 1254
1269 free_netdev(pr->d_netdev);
1270
1271 ret = ehea_destroy_qp(pr->qp); 1255 ret = ehea_destroy_qp(pr->qp);
1272 1256
1273 if (!ret) { 1257 if (!ret) {
@@ -2248,6 +2232,22 @@ out:
2248 return ret; 2232 return ret;
2249} 2233}
2250 2234
2235static void port_napi_disable(struct ehea_port *port)
2236{
2237 int i;
2238
2239 for (i = 0; i < port->num_def_qps; i++)
2240 napi_disable(&port->port_res[i].napi);
2241}
2242
2243static void port_napi_enable(struct ehea_port *port)
2244{
2245 int i;
2246
2247 for (i = 0; i < port->num_def_qps; i++)
2248 napi_enable(&port->port_res[i].napi);
2249}
2250
2251static int ehea_open(struct net_device *dev) 2251static int ehea_open(struct net_device *dev)
2252{ 2252{
2253 int ret; 2253 int ret;
@@ -2259,8 +2259,10 @@ static int ehea_open(struct net_device *dev)
2259 ehea_info("enabling port %s", dev->name); 2259 ehea_info("enabling port %s", dev->name);
2260 2260
2261 ret = ehea_up(dev); 2261 ret = ehea_up(dev);
2262 if (!ret) 2262 if (!ret) {
2263 port_napi_enable(port);
2263 netif_start_queue(dev); 2264 netif_start_queue(dev);
2265 }
2264 2266
2265 up(&port->port_lock); 2267 up(&port->port_lock);
2266 2268
@@ -2269,7 +2271,7 @@ static int ehea_open(struct net_device *dev)
2269 2271
2270static int ehea_down(struct net_device *dev) 2272static int ehea_down(struct net_device *dev)
2271{ 2273{
2272 int ret, i; 2274 int ret;
2273 struct ehea_port *port = netdev_priv(dev); 2275 struct ehea_port *port = netdev_priv(dev);
2274 2276
2275 if (port->state == EHEA_PORT_DOWN) 2277 if (port->state == EHEA_PORT_DOWN)
@@ -2278,10 +2280,7 @@ static int ehea_down(struct net_device *dev)
2278 ehea_drop_multicast_list(dev); 2280 ehea_drop_multicast_list(dev);
2279 ehea_free_interrupts(dev); 2281 ehea_free_interrupts(dev);
2280 2282
2281 for (i = 0; i < port->num_def_qps; i++) 2283 port_napi_disable(port);
2282 while (test_bit(__LINK_STATE_RX_SCHED,
2283 &port->port_res[i].d_netdev->state))
2284 msleep(1);
2285 2284
2286 port->state = EHEA_PORT_DOWN; 2285 port->state = EHEA_PORT_DOWN;
2287 2286
@@ -2319,7 +2318,8 @@ static void ehea_reset_port(struct work_struct *work)
2319 port->resets++; 2318 port->resets++;
2320 down(&port->port_lock); 2319 down(&port->port_lock);
2321 netif_stop_queue(dev); 2320 netif_stop_queue(dev);
2322 netif_poll_disable(dev); 2321
2322 port_napi_disable(port);
2323 2323
2324 ehea_down(dev); 2324 ehea_down(dev);
2325 2325
@@ -2330,7 +2330,8 @@ static void ehea_reset_port(struct work_struct *work)
2330 if (netif_msg_timer(port)) 2330 if (netif_msg_timer(port))
2331 ehea_info("Device %s resetted successfully", dev->name); 2331 ehea_info("Device %s resetted successfully", dev->name);
2332 2332
2333 netif_poll_enable(dev); 2333 port_napi_enable(port);
2334
2334 netif_wake_queue(dev); 2335 netif_wake_queue(dev);
2335out: 2336out:
2336 up(&port->port_lock); 2337 up(&port->port_lock);
@@ -2358,7 +2359,9 @@ static void ehea_rereg_mrs(struct work_struct *work)
2358 dev->name); 2359 dev->name);
2359 down(&port->port_lock); 2360 down(&port->port_lock);
2360 netif_stop_queue(dev); 2361 netif_stop_queue(dev);
2361 netif_poll_disable(dev); 2362
2363 port_napi_disable(port);
2364
2362 ehea_down(dev); 2365 ehea_down(dev);
2363 up(&port->port_lock); 2366 up(&port->port_lock);
2364 } 2367 }
@@ -2406,7 +2409,7 @@ static void ehea_rereg_mrs(struct work_struct *work)
2406 2409
2407 ret = ehea_up(dev); 2410 ret = ehea_up(dev);
2408 if (!ret) { 2411 if (!ret) {
2409 netif_poll_enable(dev); 2412 port_napi_enable(port);
2410 netif_wake_queue(dev); 2413 netif_wake_queue(dev);
2411 } 2414 }
2412 2415
@@ -2644,11 +2647,9 @@ struct ehea_port *ehea_setup_single_port(struct ehea_adapter *adapter,
2644 memcpy(dev->dev_addr, &port->mac_addr, ETH_ALEN); 2647 memcpy(dev->dev_addr, &port->mac_addr, ETH_ALEN);
2645 2648
2646 dev->open = ehea_open; 2649 dev->open = ehea_open;
2647 dev->poll = ehea_poll_firstqueue;
2648#ifdef CONFIG_NET_POLL_CONTROLLER 2650#ifdef CONFIG_NET_POLL_CONTROLLER
2649 dev->poll_controller = ehea_netpoll; 2651 dev->poll_controller = ehea_netpoll;
2650#endif 2652#endif
2651 dev->weight = 64;
2652 dev->stop = ehea_stop; 2653 dev->stop = ehea_stop;
2653 dev->hard_start_xmit = ehea_start_xmit; 2654 dev->hard_start_xmit = ehea_start_xmit;
2654 dev->get_stats = ehea_get_stats; 2655 dev->get_stats = ehea_get_stats;
diff --git a/drivers/net/epic100.c b/drivers/net/epic100.c
index 119778401e48..f8446e373bdd 100644
--- a/drivers/net/epic100.c
+++ b/drivers/net/epic100.c
@@ -262,6 +262,7 @@ struct epic_private {
262 /* Ring pointers. */ 262 /* Ring pointers. */
263 spinlock_t lock; /* Group with Tx control cache line. */ 263 spinlock_t lock; /* Group with Tx control cache line. */
264 spinlock_t napi_lock; 264 spinlock_t napi_lock;
265 struct napi_struct napi;
265 unsigned int reschedule_in_poll; 266 unsigned int reschedule_in_poll;
266 unsigned int cur_tx, dirty_tx; 267 unsigned int cur_tx, dirty_tx;
267 268
@@ -294,7 +295,7 @@ static void epic_tx_timeout(struct net_device *dev);
294static void epic_init_ring(struct net_device *dev); 295static void epic_init_ring(struct net_device *dev);
295static int epic_start_xmit(struct sk_buff *skb, struct net_device *dev); 296static int epic_start_xmit(struct sk_buff *skb, struct net_device *dev);
296static int epic_rx(struct net_device *dev, int budget); 297static int epic_rx(struct net_device *dev, int budget);
297static int epic_poll(struct net_device *dev, int *budget); 298static int epic_poll(struct napi_struct *napi, int budget);
298static irqreturn_t epic_interrupt(int irq, void *dev_instance); 299static irqreturn_t epic_interrupt(int irq, void *dev_instance);
299static int netdev_ioctl(struct net_device *dev, struct ifreq *rq, int cmd); 300static int netdev_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
300static const struct ethtool_ops netdev_ethtool_ops; 301static const struct ethtool_ops netdev_ethtool_ops;
@@ -487,8 +488,7 @@ static int __devinit epic_init_one (struct pci_dev *pdev,
487 dev->ethtool_ops = &netdev_ethtool_ops; 488 dev->ethtool_ops = &netdev_ethtool_ops;
488 dev->watchdog_timeo = TX_TIMEOUT; 489 dev->watchdog_timeo = TX_TIMEOUT;
489 dev->tx_timeout = &epic_tx_timeout; 490 dev->tx_timeout = &epic_tx_timeout;
490 dev->poll = epic_poll; 491 netif_napi_add(dev, &ep->napi, epic_poll, 64);
491 dev->weight = 64;
492 492
493 ret = register_netdev(dev); 493 ret = register_netdev(dev);
494 if (ret < 0) 494 if (ret < 0)
@@ -660,8 +660,11 @@ static int epic_open(struct net_device *dev)
660 /* Soft reset the chip. */ 660 /* Soft reset the chip. */
661 outl(0x4001, ioaddr + GENCTL); 661 outl(0x4001, ioaddr + GENCTL);
662 662
663 if ((retval = request_irq(dev->irq, &epic_interrupt, IRQF_SHARED, dev->name, dev))) 663 napi_enable(&ep->napi);
664 if ((retval = request_irq(dev->irq, &epic_interrupt, IRQF_SHARED, dev->name, dev))) {
665 napi_disable(&ep->napi);
664 return retval; 666 return retval;
667 }
665 668
666 epic_init_ring(dev); 669 epic_init_ring(dev);
667 670
@@ -1103,9 +1106,9 @@ static irqreturn_t epic_interrupt(int irq, void *dev_instance)
1103 1106
1104 if ((status & EpicNapiEvent) && !ep->reschedule_in_poll) { 1107 if ((status & EpicNapiEvent) && !ep->reschedule_in_poll) {
1105 spin_lock(&ep->napi_lock); 1108 spin_lock(&ep->napi_lock);
1106 if (netif_rx_schedule_prep(dev)) { 1109 if (netif_rx_schedule_prep(dev, &ep->napi)) {
1107 epic_napi_irq_off(dev, ep); 1110 epic_napi_irq_off(dev, ep);
1108 __netif_rx_schedule(dev); 1111 __netif_rx_schedule(dev, &ep->napi);
1109 } else 1112 } else
1110 ep->reschedule_in_poll++; 1113 ep->reschedule_in_poll++;
1111 spin_unlock(&ep->napi_lock); 1114 spin_unlock(&ep->napi_lock);
@@ -1257,26 +1260,22 @@ static void epic_rx_err(struct net_device *dev, struct epic_private *ep)
1257 outw(RxQueued, ioaddr + COMMAND); 1260 outw(RxQueued, ioaddr + COMMAND);
1258} 1261}
1259 1262
1260static int epic_poll(struct net_device *dev, int *budget) 1263static int epic_poll(struct napi_struct *napi, int budget)
1261{ 1264{
1262 struct epic_private *ep = dev->priv; 1265 struct epic_private *ep = container_of(napi, struct epic_private, napi);
1263 int work_done = 0, orig_budget; 1266 struct net_device *dev = ep->mii.dev;
1267 int work_done = 0;
1264 long ioaddr = dev->base_addr; 1268 long ioaddr = dev->base_addr;
1265 1269
1266 orig_budget = (*budget > dev->quota) ? dev->quota : *budget;
1267
1268rx_action: 1270rx_action:
1269 1271
1270 epic_tx(dev, ep); 1272 epic_tx(dev, ep);
1271 1273
1272 work_done += epic_rx(dev, *budget); 1274 work_done += epic_rx(dev, budget);
1273 1275
1274 epic_rx_err(dev, ep); 1276 epic_rx_err(dev, ep);
1275 1277
1276 *budget -= work_done; 1278 if (netif_running(dev) && (work_done < budget)) {
1277 dev->quota -= work_done;
1278
1279 if (netif_running(dev) && (work_done < orig_budget)) {
1280 unsigned long flags; 1279 unsigned long flags;
1281 int more; 1280 int more;
1282 1281
@@ -1286,7 +1285,7 @@ rx_action:
1286 1285
1287 more = ep->reschedule_in_poll; 1286 more = ep->reschedule_in_poll;
1288 if (!more) { 1287 if (!more) {
1289 __netif_rx_complete(dev); 1288 __netif_rx_complete(dev, napi);
1290 outl(EpicNapiEvent, ioaddr + INTSTAT); 1289 outl(EpicNapiEvent, ioaddr + INTSTAT);
1291 epic_napi_irq_on(dev, ep); 1290 epic_napi_irq_on(dev, ep);
1292 } else 1291 } else
@@ -1298,7 +1297,7 @@ rx_action:
1298 goto rx_action; 1297 goto rx_action;
1299 } 1298 }
1300 1299
1301 return (work_done >= orig_budget); 1300 return work_done;
1302} 1301}
1303 1302
1304static int epic_close(struct net_device *dev) 1303static int epic_close(struct net_device *dev)
@@ -1309,6 +1308,7 @@ static int epic_close(struct net_device *dev)
1309 int i; 1308 int i;
1310 1309
1311 netif_stop_queue(dev); 1310 netif_stop_queue(dev);
1311 napi_disable(&ep->napi);
1312 1312
1313 if (debug > 1) 1313 if (debug > 1)
1314 printk(KERN_DEBUG "%s: Shutting down ethercard, status was %2.2x.\n", 1314 printk(KERN_DEBUG "%s: Shutting down ethercard, status was %2.2x.\n",
diff --git a/drivers/net/fec_8xx/fec_8xx.h b/drivers/net/fec_8xx/fec_8xx.h
index 5af60b0f9208..f3b1c6fbba8b 100644
--- a/drivers/net/fec_8xx/fec_8xx.h
+++ b/drivers/net/fec_8xx/fec_8xx.h
@@ -105,6 +105,8 @@ struct fec;
105struct fec_enet_private { 105struct fec_enet_private {
106 spinlock_t lock; /* during all ops except TX pckt processing */ 106 spinlock_t lock; /* during all ops except TX pckt processing */
107 spinlock_t tx_lock; /* during fec_start_xmit and fec_tx */ 107 spinlock_t tx_lock; /* during fec_start_xmit and fec_tx */
108 struct net_device *dev;
109 struct napi_struct napi;
108 int fecno; 110 int fecno;
109 struct fec *fecp; 111 struct fec *fecp;
110 const struct fec_platform_info *fpi; 112 const struct fec_platform_info *fpi;
diff --git a/drivers/net/fec_8xx/fec_main.c b/drivers/net/fec_8xx/fec_main.c
index e5502af5b8e2..6348fb93ca9c 100644
--- a/drivers/net/fec_8xx/fec_main.c
+++ b/drivers/net/fec_8xx/fec_main.c
@@ -465,9 +465,9 @@ void fec_stop(struct net_device *dev)
465} 465}
466 466
467/* common receive function */ 467/* common receive function */
468static int fec_enet_rx_common(struct net_device *dev, int *budget) 468static int fec_enet_rx_common(struct fec_enet_private *ep,
469 struct net_device *dev, int budget)
469{ 470{
470 struct fec_enet_private *fep = netdev_priv(dev);
471 fec_t *fecp = fep->fecp; 471 fec_t *fecp = fep->fecp;
472 const struct fec_platform_info *fpi = fep->fpi; 472 const struct fec_platform_info *fpi = fep->fpi;
473 cbd_t *bdp; 473 cbd_t *bdp;
@@ -475,11 +475,8 @@ static int fec_enet_rx_common(struct net_device *dev, int *budget)
475 int received = 0; 475 int received = 0;
476 __u16 pkt_len, sc; 476 __u16 pkt_len, sc;
477 int curidx; 477 int curidx;
478 int rx_work_limit;
479 478
480 if (fpi->use_napi) { 479 if (fpi->use_napi) {
481 rx_work_limit = min(dev->quota, *budget);
482
483 if (!netif_running(dev)) 480 if (!netif_running(dev))
484 return 0; 481 return 0;
485 } 482 }
@@ -530,11 +527,6 @@ static int fec_enet_rx_common(struct net_device *dev, int *budget)
530 BUG_ON(skbn == NULL); 527 BUG_ON(skbn == NULL);
531 528
532 } else { 529 } else {
533
534 /* napi, got packet but no quota */
535 if (fpi->use_napi && --rx_work_limit < 0)
536 break;
537
538 skb = fep->rx_skbuff[curidx]; 530 skb = fep->rx_skbuff[curidx];
539 BUG_ON(skb == NULL); 531 BUG_ON(skb == NULL);
540 532
@@ -599,25 +591,24 @@ static int fec_enet_rx_common(struct net_device *dev, int *budget)
599 * able to keep up at the expense of system resources. 591 * able to keep up at the expense of system resources.
600 */ 592 */
601 FW(fecp, r_des_active, 0x01000000); 593 FW(fecp, r_des_active, 0x01000000);
594
595 if (received >= budget)
596 break;
597
602 } 598 }
603 599
604 fep->cur_rx = bdp; 600 fep->cur_rx = bdp;
605 601
606 if (fpi->use_napi) { 602 if (fpi->use_napi) {
607 dev->quota -= received; 603 if (received < budget) {
608 *budget -= received; 604 netif_rx_complete(dev, &fep->napi);
609
610 if (rx_work_limit < 0)
611 return 1; /* not done */
612 605
613 /* done */ 606 /* enable RX interrupt bits */
614 netif_rx_complete(dev); 607 FS(fecp, imask, FEC_ENET_RXF | FEC_ENET_RXB);
615 608 }
616 /* enable RX interrupt bits */
617 FS(fecp, imask, FEC_ENET_RXF | FEC_ENET_RXB);
618 } 609 }
619 610
620 return 0; 611 return received;
621} 612}
622 613
623static void fec_enet_tx(struct net_device *dev) 614static void fec_enet_tx(struct net_device *dev)
@@ -743,12 +734,12 @@ fec_enet_interrupt(int irq, void *dev_id)
743 734
744 if ((int_events & FEC_ENET_RXF) != 0) { 735 if ((int_events & FEC_ENET_RXF) != 0) {
745 if (!fpi->use_napi) 736 if (!fpi->use_napi)
746 fec_enet_rx_common(dev, NULL); 737 fec_enet_rx_common(fep, dev, ~0);
747 else { 738 else {
748 if (netif_rx_schedule_prep(dev)) { 739 if (netif_rx_schedule_prep(dev, &fep->napi)) {
749 /* disable rx interrupts */ 740 /* disable rx interrupts */
750 FC(fecp, imask, FEC_ENET_RXF | FEC_ENET_RXB); 741 FC(fecp, imask, FEC_ENET_RXF | FEC_ENET_RXB);
751 __netif_rx_schedule(dev); 742 __netif_rx_schedule(dev, &fep->napi);
752 } else { 743 } else {
753 printk(KERN_ERR DRV_MODULE_NAME 744 printk(KERN_ERR DRV_MODULE_NAME
754 ": %s driver bug! interrupt while in poll!\n", 745 ": %s driver bug! interrupt while in poll!\n",
@@ -893,10 +884,13 @@ static int fec_enet_open(struct net_device *dev)
893 const struct fec_platform_info *fpi = fep->fpi; 884 const struct fec_platform_info *fpi = fep->fpi;
894 unsigned long flags; 885 unsigned long flags;
895 886
887 napi_enable(&fep->napi);
888
896 /* Install our interrupt handler. */ 889 /* Install our interrupt handler. */
897 if (request_irq(fpi->fec_irq, fec_enet_interrupt, 0, "fec", dev) != 0) { 890 if (request_irq(fpi->fec_irq, fec_enet_interrupt, 0, "fec", dev) != 0) {
898 printk(KERN_ERR DRV_MODULE_NAME 891 printk(KERN_ERR DRV_MODULE_NAME
899 ": %s Could not allocate FEC IRQ!", dev->name); 892 ": %s Could not allocate FEC IRQ!", dev->name);
893 napi_disable(&fep->napi);
900 return -EINVAL; 894 return -EINVAL;
901 } 895 }
902 896
@@ -907,6 +901,7 @@ static int fec_enet_open(struct net_device *dev)
907 printk(KERN_ERR DRV_MODULE_NAME 901 printk(KERN_ERR DRV_MODULE_NAME
908 ": %s Could not allocate PHY IRQ!", dev->name); 902 ": %s Could not allocate PHY IRQ!", dev->name);
909 free_irq(fpi->fec_irq, dev); 903 free_irq(fpi->fec_irq, dev);
904 napi_disable(&fep->napi);
910 return -EINVAL; 905 return -EINVAL;
911 } 906 }
912 907
@@ -932,6 +927,7 @@ static int fec_enet_close(struct net_device *dev)
932 unsigned long flags; 927 unsigned long flags;
933 928
934 netif_stop_queue(dev); 929 netif_stop_queue(dev);
930 napi_disable(&fep->napi);
935 netif_carrier_off(dev); 931 netif_carrier_off(dev);
936 932
937 if (fpi->use_mdio) 933 if (fpi->use_mdio)
@@ -955,9 +951,12 @@ static struct net_device_stats *fec_enet_get_stats(struct net_device *dev)
955 return &fep->stats; 951 return &fep->stats;
956} 952}
957 953
958static int fec_enet_poll(struct net_device *dev, int *budget) 954static int fec_enet_poll(struct napi_struct *napi, int budget)
959{ 955{
960 return fec_enet_rx_common(dev, budget); 956 struct fec_enet_private *fep = container_of(napi, struct fec_enet_private, napi);
957 struct net_device *dev = fep->dev;
958
959 return fec_enet_rx_common(fep, dev, budget);
961} 960}
962 961
963/*************************************************************************/ 962/*************************************************************************/
@@ -1107,6 +1106,7 @@ int fec_8xx_init_one(const struct fec_platform_info *fpi,
1107 SET_MODULE_OWNER(dev); 1106 SET_MODULE_OWNER(dev);
1108 1107
1109 fep = netdev_priv(dev); 1108 fep = netdev_priv(dev);
1109 fep->dev = dev;
1110 1110
1111 /* partial reset of FEC */ 1111 /* partial reset of FEC */
1112 fec_whack_reset(fecp); 1112 fec_whack_reset(fecp);
@@ -1172,10 +1172,9 @@ int fec_8xx_init_one(const struct fec_platform_info *fpi,
1172 dev->get_stats = fec_enet_get_stats; 1172 dev->get_stats = fec_enet_get_stats;
1173 dev->set_multicast_list = fec_set_multicast_list; 1173 dev->set_multicast_list = fec_set_multicast_list;
1174 dev->set_mac_address = fec_set_mac_address; 1174 dev->set_mac_address = fec_set_mac_address;
1175 if (fpi->use_napi) { 1175 netif_napi_add(dev, &fec->napi,
1176 dev->poll = fec_enet_poll; 1176 fec_enet_poll, fpi->napi_weight);
1177 dev->weight = fpi->napi_weight; 1177
1178 }
1179 dev->ethtool_ops = &fec_ethtool_ops; 1178 dev->ethtool_ops = &fec_ethtool_ops;
1180 dev->do_ioctl = fec_ioctl; 1179 dev->do_ioctl = fec_ioctl;
1181 1180
diff --git a/drivers/net/forcedeth.c b/drivers/net/forcedeth.c
index 1938d6dfc863..24c1294614f2 100644
--- a/drivers/net/forcedeth.c
+++ b/drivers/net/forcedeth.c
@@ -159,6 +159,8 @@
159#define dprintk(x...) do { } while (0) 159#define dprintk(x...) do { } while (0)
160#endif 160#endif
161 161
162#define TX_WORK_PER_LOOP 64
163#define RX_WORK_PER_LOOP 64
162 164
163/* 165/*
164 * Hardware access: 166 * Hardware access:
@@ -745,6 +747,9 @@ struct nv_skb_map {
745struct fe_priv { 747struct fe_priv {
746 spinlock_t lock; 748 spinlock_t lock;
747 749
750 struct net_device *dev;
751 struct napi_struct napi;
752
748 /* General data: 753 /* General data:
749 * Locking: spin_lock(&np->lock); */ 754 * Locking: spin_lock(&np->lock); */
750 struct net_device_stats stats; 755 struct net_device_stats stats;
@@ -1586,9 +1591,10 @@ static int nv_alloc_rx_optimized(struct net_device *dev)
1586static void nv_do_rx_refill(unsigned long data) 1591static void nv_do_rx_refill(unsigned long data)
1587{ 1592{
1588 struct net_device *dev = (struct net_device *) data; 1593 struct net_device *dev = (struct net_device *) data;
1594 struct fe_priv *np = netdev_priv(dev);
1589 1595
1590 /* Just reschedule NAPI rx processing */ 1596 /* Just reschedule NAPI rx processing */
1591 netif_rx_schedule(dev); 1597 netif_rx_schedule(dev, &np->napi);
1592} 1598}
1593#else 1599#else
1594static void nv_do_rx_refill(unsigned long data) 1600static void nv_do_rx_refill(unsigned long data)
@@ -2997,7 +3003,7 @@ static irqreturn_t nv_nic_irq(int foo, void *data)
2997 3003
2998#ifdef CONFIG_FORCEDETH_NAPI 3004#ifdef CONFIG_FORCEDETH_NAPI
2999 if (events & NVREG_IRQ_RX_ALL) { 3005 if (events & NVREG_IRQ_RX_ALL) {
3000 netif_rx_schedule(dev); 3006 netif_rx_schedule(dev, &np->napi);
3001 3007
3002 /* Disable furthur receive irq's */ 3008 /* Disable furthur receive irq's */
3003 spin_lock(&np->lock); 3009 spin_lock(&np->lock);
@@ -3010,7 +3016,7 @@ static irqreturn_t nv_nic_irq(int foo, void *data)
3010 spin_unlock(&np->lock); 3016 spin_unlock(&np->lock);
3011 } 3017 }
3012#else 3018#else
3013 if (nv_rx_process(dev, dev->weight)) { 3019 if (nv_rx_process(dev, RX_WORK_PER_LOOP)) {
3014 if (unlikely(nv_alloc_rx(dev))) { 3020 if (unlikely(nv_alloc_rx(dev))) {
3015 spin_lock(&np->lock); 3021 spin_lock(&np->lock);
3016 if (!np->in_shutdown) 3022 if (!np->in_shutdown)
@@ -3079,8 +3085,6 @@ static irqreturn_t nv_nic_irq(int foo, void *data)
3079 return IRQ_RETVAL(i); 3085 return IRQ_RETVAL(i);
3080} 3086}
3081 3087
3082#define TX_WORK_PER_LOOP 64
3083#define RX_WORK_PER_LOOP 64
3084/** 3088/**
3085 * All _optimized functions are used to help increase performance 3089 * All _optimized functions are used to help increase performance
3086 * (reduce CPU and increase throughput). They use descripter version 3, 3090 * (reduce CPU and increase throughput). They use descripter version 3,
@@ -3114,7 +3118,7 @@ static irqreturn_t nv_nic_irq_optimized(int foo, void *data)
3114 3118
3115#ifdef CONFIG_FORCEDETH_NAPI 3119#ifdef CONFIG_FORCEDETH_NAPI
3116 if (events & NVREG_IRQ_RX_ALL) { 3120 if (events & NVREG_IRQ_RX_ALL) {
3117 netif_rx_schedule(dev); 3121 netif_rx_schedule(dev, &np->napi);
3118 3122
3119 /* Disable furthur receive irq's */ 3123 /* Disable furthur receive irq's */
3120 spin_lock(&np->lock); 3124 spin_lock(&np->lock);
@@ -3127,7 +3131,7 @@ static irqreturn_t nv_nic_irq_optimized(int foo, void *data)
3127 spin_unlock(&np->lock); 3131 spin_unlock(&np->lock);
3128 } 3132 }
3129#else 3133#else
3130 if (nv_rx_process_optimized(dev, dev->weight)) { 3134 if (nv_rx_process_optimized(dev, RX_WORK_PER_LOOP)) {
3131 if (unlikely(nv_alloc_rx_optimized(dev))) { 3135 if (unlikely(nv_alloc_rx_optimized(dev))) {
3132 spin_lock(&np->lock); 3136 spin_lock(&np->lock);
3133 if (!np->in_shutdown) 3137 if (!np->in_shutdown)
@@ -3245,19 +3249,19 @@ static irqreturn_t nv_nic_irq_tx(int foo, void *data)
3245} 3249}
3246 3250
3247#ifdef CONFIG_FORCEDETH_NAPI 3251#ifdef CONFIG_FORCEDETH_NAPI
3248static int nv_napi_poll(struct net_device *dev, int *budget) 3252static int nv_napi_poll(struct napi_struct *napi, int budget)
3249{ 3253{
3250 int pkts, limit = min(*budget, dev->quota); 3254 struct fe_priv *np = container_of(napi, struct fe_priv, napi);
3251 struct fe_priv *np = netdev_priv(dev); 3255 struct net_device *dev = np->dev;
3252 u8 __iomem *base = get_hwbase(dev); 3256 u8 __iomem *base = get_hwbase(dev);
3253 unsigned long flags; 3257 unsigned long flags;
3254 int retcode; 3258 int pkts, retcode;
3255 3259
3256 if (np->desc_ver == DESC_VER_1 || np->desc_ver == DESC_VER_2) { 3260 if (np->desc_ver == DESC_VER_1 || np->desc_ver == DESC_VER_2) {
3257 pkts = nv_rx_process(dev, limit); 3261 pkts = nv_rx_process(dev, budget);
3258 retcode = nv_alloc_rx(dev); 3262 retcode = nv_alloc_rx(dev);
3259 } else { 3263 } else {
3260 pkts = nv_rx_process_optimized(dev, limit); 3264 pkts = nv_rx_process_optimized(dev, budget);
3261 retcode = nv_alloc_rx_optimized(dev); 3265 retcode = nv_alloc_rx_optimized(dev);
3262 } 3266 }
3263 3267
@@ -3268,13 +3272,12 @@ static int nv_napi_poll(struct net_device *dev, int *budget)
3268 spin_unlock_irqrestore(&np->lock, flags); 3272 spin_unlock_irqrestore(&np->lock, flags);
3269 } 3273 }
3270 3274
3271 if (pkts < limit) { 3275 if (pkts < budget) {
3272 /* all done, no more packets present */
3273 netif_rx_complete(dev);
3274
3275 /* re-enable receive interrupts */ 3276 /* re-enable receive interrupts */
3276 spin_lock_irqsave(&np->lock, flags); 3277 spin_lock_irqsave(&np->lock, flags);
3277 3278
3279 __netif_rx_complete(dev, napi);
3280
3278 np->irqmask |= NVREG_IRQ_RX_ALL; 3281 np->irqmask |= NVREG_IRQ_RX_ALL;
3279 if (np->msi_flags & NV_MSI_X_ENABLED) 3282 if (np->msi_flags & NV_MSI_X_ENABLED)
3280 writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask); 3283 writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask);
@@ -3282,13 +3285,8 @@ static int nv_napi_poll(struct net_device *dev, int *budget)
3282 writel(np->irqmask, base + NvRegIrqMask); 3285 writel(np->irqmask, base + NvRegIrqMask);
3283 3286
3284 spin_unlock_irqrestore(&np->lock, flags); 3287 spin_unlock_irqrestore(&np->lock, flags);
3285 return 0;
3286 } else {
3287 /* used up our quantum, so reschedule */
3288 dev->quota -= pkts;
3289 *budget -= pkts;
3290 return 1;
3291 } 3288 }
3289 return pkts;
3292} 3290}
3293#endif 3291#endif
3294 3292
@@ -3296,6 +3294,7 @@ static int nv_napi_poll(struct net_device *dev, int *budget)
3296static irqreturn_t nv_nic_irq_rx(int foo, void *data) 3294static irqreturn_t nv_nic_irq_rx(int foo, void *data)
3297{ 3295{
3298 struct net_device *dev = (struct net_device *) data; 3296 struct net_device *dev = (struct net_device *) data;
3297 struct fe_priv *np = netdev_priv(dev);
3299 u8 __iomem *base = get_hwbase(dev); 3298 u8 __iomem *base = get_hwbase(dev);
3300 u32 events; 3299 u32 events;
3301 3300
@@ -3303,7 +3302,7 @@ static irqreturn_t nv_nic_irq_rx(int foo, void *data)
3303 writel(NVREG_IRQ_RX_ALL, base + NvRegMSIXIrqStatus); 3302 writel(NVREG_IRQ_RX_ALL, base + NvRegMSIXIrqStatus);
3304 3303
3305 if (events) { 3304 if (events) {
3306 netif_rx_schedule(dev); 3305 netif_rx_schedule(dev, &np->napi);
3307 /* disable receive interrupts on the nic */ 3306 /* disable receive interrupts on the nic */
3308 writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask); 3307 writel(NVREG_IRQ_RX_ALL, base + NvRegIrqMask);
3309 pci_push(base); 3308 pci_push(base);
@@ -3329,7 +3328,7 @@ static irqreturn_t nv_nic_irq_rx(int foo, void *data)
3329 if (!(events & np->irqmask)) 3328 if (!(events & np->irqmask))
3330 break; 3329 break;
3331 3330
3332 if (nv_rx_process_optimized(dev, dev->weight)) { 3331 if (nv_rx_process_optimized(dev, RX_WORK_PER_LOOP)) {
3333 if (unlikely(nv_alloc_rx_optimized(dev))) { 3332 if (unlikely(nv_alloc_rx_optimized(dev))) {
3334 spin_lock_irqsave(&np->lock, flags); 3333 spin_lock_irqsave(&np->lock, flags);
3335 if (!np->in_shutdown) 3334 if (!np->in_shutdown)
@@ -4620,7 +4619,9 @@ static void nv_self_test(struct net_device *dev, struct ethtool_test *test, u64
4620 if (test->flags & ETH_TEST_FL_OFFLINE) { 4619 if (test->flags & ETH_TEST_FL_OFFLINE) {
4621 if (netif_running(dev)) { 4620 if (netif_running(dev)) {
4622 netif_stop_queue(dev); 4621 netif_stop_queue(dev);
4623 netif_poll_disable(dev); 4622#ifdef CONFIG_FORCEDETH_NAPI
4623 napi_disable(&np->napi);
4624#endif
4624 netif_tx_lock_bh(dev); 4625 netif_tx_lock_bh(dev);
4625 spin_lock_irq(&np->lock); 4626 spin_lock_irq(&np->lock);
4626 nv_disable_hw_interrupts(dev, np->irqmask); 4627 nv_disable_hw_interrupts(dev, np->irqmask);
@@ -4679,7 +4680,9 @@ static void nv_self_test(struct net_device *dev, struct ethtool_test *test, u64
4679 nv_start_rx(dev); 4680 nv_start_rx(dev);
4680 nv_start_tx(dev); 4681 nv_start_tx(dev);
4681 netif_start_queue(dev); 4682 netif_start_queue(dev);
4682 netif_poll_enable(dev); 4683#ifdef CONFIG_FORCEDETH_NAPI
4684 napi_enable(&np->napi);
4685#endif
4683 nv_enable_hw_interrupts(dev, np->irqmask); 4686 nv_enable_hw_interrupts(dev, np->irqmask);
4684 } 4687 }
4685 } 4688 }
@@ -4911,7 +4914,9 @@ static int nv_open(struct net_device *dev)
4911 nv_start_rx(dev); 4914 nv_start_rx(dev);
4912 nv_start_tx(dev); 4915 nv_start_tx(dev);
4913 netif_start_queue(dev); 4916 netif_start_queue(dev);
4914 netif_poll_enable(dev); 4917#ifdef CONFIG_FORCEDETH_NAPI
4918 napi_enable(&np->napi);
4919#endif
4915 4920
4916 if (ret) { 4921 if (ret) {
4917 netif_carrier_on(dev); 4922 netif_carrier_on(dev);
@@ -4942,7 +4947,9 @@ static int nv_close(struct net_device *dev)
4942 spin_lock_irq(&np->lock); 4947 spin_lock_irq(&np->lock);
4943 np->in_shutdown = 1; 4948 np->in_shutdown = 1;
4944 spin_unlock_irq(&np->lock); 4949 spin_unlock_irq(&np->lock);
4945 netif_poll_disable(dev); 4950#ifdef CONFIG_FORCEDETH_NAPI
4951 napi_disable(&np->napi);
4952#endif
4946 synchronize_irq(dev->irq); 4953 synchronize_irq(dev->irq);
4947 4954
4948 del_timer_sync(&np->oom_kick); 4955 del_timer_sync(&np->oom_kick);
@@ -4994,6 +5001,7 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i
4994 goto out; 5001 goto out;
4995 5002
4996 np = netdev_priv(dev); 5003 np = netdev_priv(dev);
5004 np->dev = dev;
4997 np->pci_dev = pci_dev; 5005 np->pci_dev = pci_dev;
4998 spin_lock_init(&np->lock); 5006 spin_lock_init(&np->lock);
4999 SET_MODULE_OWNER(dev); 5007 SET_MODULE_OWNER(dev);
@@ -5155,9 +5163,8 @@ static int __devinit nv_probe(struct pci_dev *pci_dev, const struct pci_device_i
5155#ifdef CONFIG_NET_POLL_CONTROLLER 5163#ifdef CONFIG_NET_POLL_CONTROLLER
5156 dev->poll_controller = nv_poll_controller; 5164 dev->poll_controller = nv_poll_controller;
5157#endif 5165#endif
5158 dev->weight = RX_WORK_PER_LOOP;
5159#ifdef CONFIG_FORCEDETH_NAPI 5166#ifdef CONFIG_FORCEDETH_NAPI
5160 dev->poll = nv_napi_poll; 5167 netif_napi_add(dev, &np->napi, nv_napi_poll, RX_WORK_PER_LOOP);
5161#endif 5168#endif
5162 SET_ETHTOOL_OPS(dev, &ops); 5169 SET_ETHTOOL_OPS(dev, &ops);
5163 dev->tx_timeout = nv_tx_timeout; 5170 dev->tx_timeout = nv_tx_timeout;
diff --git a/drivers/net/fs_enet/fs_enet-main.c b/drivers/net/fs_enet/fs_enet-main.c
index a4a2a0ea43d3..c509cb13222d 100644
--- a/drivers/net/fs_enet/fs_enet-main.c
+++ b/drivers/net/fs_enet/fs_enet-main.c
@@ -70,18 +70,16 @@ static void fs_set_multicast_list(struct net_device *dev)
70} 70}
71 71
72/* NAPI receive function */ 72/* NAPI receive function */
73static int fs_enet_rx_napi(struct net_device *dev, int *budget) 73static int fs_enet_rx_napi(struct napi_struct *napi, int budget)
74{ 74{
75 struct fs_enet_private *fep = netdev_priv(dev); 75 struct fs_enet_private *fep = container_of(napi, struct fs_enet_private, napi);
76 struct net_device *dev = to_net_dev(fep->dev);
76 const struct fs_platform_info *fpi = fep->fpi; 77 const struct fs_platform_info *fpi = fep->fpi;
77 cbd_t *bdp; 78 cbd_t *bdp;
78 struct sk_buff *skb, *skbn, *skbt; 79 struct sk_buff *skb, *skbn, *skbt;
79 int received = 0; 80 int received = 0;
80 u16 pkt_len, sc; 81 u16 pkt_len, sc;
81 int curidx; 82 int curidx;
82 int rx_work_limit = 0; /* pacify gcc */
83
84 rx_work_limit = min(dev->quota, *budget);
85 83
86 if (!netif_running(dev)) 84 if (!netif_running(dev))
87 return 0; 85 return 0;
@@ -96,7 +94,6 @@ static int fs_enet_rx_napi(struct net_device *dev, int *budget)
96 (*fep->ops->napi_clear_rx_event)(dev); 94 (*fep->ops->napi_clear_rx_event)(dev);
97 95
98 while (((sc = CBDR_SC(bdp)) & BD_ENET_RX_EMPTY) == 0) { 96 while (((sc = CBDR_SC(bdp)) & BD_ENET_RX_EMPTY) == 0) {
99
100 curidx = bdp - fep->rx_bd_base; 97 curidx = bdp - fep->rx_bd_base;
101 98
102 /* 99 /*
@@ -136,11 +133,6 @@ static int fs_enet_rx_napi(struct net_device *dev, int *budget)
136 skbn = skb; 133 skbn = skb;
137 134
138 } else { 135 } else {
139
140 /* napi, got packet but no quota */
141 if (--rx_work_limit < 0)
142 break;
143
144 skb = fep->rx_skbuff[curidx]; 136 skb = fep->rx_skbuff[curidx];
145 137
146 dma_unmap_single(fep->dev, CBDR_BUFADDR(bdp), 138 dma_unmap_single(fep->dev, CBDR_BUFADDR(bdp),
@@ -199,22 +191,19 @@ static int fs_enet_rx_napi(struct net_device *dev, int *budget)
199 bdp = fep->rx_bd_base; 191 bdp = fep->rx_bd_base;
200 192
201 (*fep->ops->rx_bd_done)(dev); 193 (*fep->ops->rx_bd_done)(dev);
194
195 if (received >= budget)
196 break;
202 } 197 }
203 198
204 fep->cur_rx = bdp; 199 fep->cur_rx = bdp;
205 200
206 dev->quota -= received; 201 if (received >= budget) {
207 *budget -= received; 202 /* done */
208 203 netif_rx_complete(dev, napi);
209 if (rx_work_limit < 0) 204 (*fep->ops->napi_enable_rx)(dev);
210 return 1; /* not done */ 205 }
211 206 return received;
212 /* done */
213 netif_rx_complete(dev);
214
215 (*fep->ops->napi_enable_rx)(dev);
216
217 return 0;
218} 207}
219 208
220/* non NAPI receive function */ 209/* non NAPI receive function */
@@ -470,7 +459,7 @@ fs_enet_interrupt(int irq, void *dev_id)
470 if (!fpi->use_napi) 459 if (!fpi->use_napi)
471 fs_enet_rx_non_napi(dev); 460 fs_enet_rx_non_napi(dev);
472 else { 461 else {
473 napi_ok = netif_rx_schedule_prep(dev); 462 napi_ok = napi_schedule_prep(&fep->napi);
474 463
475 (*fep->ops->napi_disable_rx)(dev); 464 (*fep->ops->napi_disable_rx)(dev);
476 (*fep->ops->clear_int_events)(dev, fep->ev_napi_rx); 465 (*fep->ops->clear_int_events)(dev, fep->ev_napi_rx);
@@ -478,7 +467,7 @@ fs_enet_interrupt(int irq, void *dev_id)
478 /* NOTE: it is possible for FCCs in NAPI mode */ 467 /* NOTE: it is possible for FCCs in NAPI mode */
479 /* to submit a spurious interrupt while in poll */ 468 /* to submit a spurious interrupt while in poll */
480 if (napi_ok) 469 if (napi_ok)
481 __netif_rx_schedule(dev); 470 __netif_rx_schedule(dev, &fep->napi);
482 } 471 }
483 } 472 }
484 473
@@ -799,18 +788,22 @@ static int fs_enet_open(struct net_device *dev)
799 int r; 788 int r;
800 int err; 789 int err;
801 790
791 napi_enable(&fep->napi);
792
802 /* Install our interrupt handler. */ 793 /* Install our interrupt handler. */
803 r = fs_request_irq(dev, fep->interrupt, "fs_enet-mac", fs_enet_interrupt); 794 r = fs_request_irq(dev, fep->interrupt, "fs_enet-mac", fs_enet_interrupt);
804 if (r != 0) { 795 if (r != 0) {
805 printk(KERN_ERR DRV_MODULE_NAME 796 printk(KERN_ERR DRV_MODULE_NAME
806 ": %s Could not allocate FS_ENET IRQ!", dev->name); 797 ": %s Could not allocate FS_ENET IRQ!", dev->name);
798 napi_disable(&fep->napi);
807 return -EINVAL; 799 return -EINVAL;
808 } 800 }
809 801
810 err = fs_init_phy(dev); 802 err = fs_init_phy(dev);
811 if(err) 803 if(err) {
804 napi_disable(&fep->napi);
812 return err; 805 return err;
813 806 }
814 phy_start(fep->phydev); 807 phy_start(fep->phydev);
815 808
816 return 0; 809 return 0;
@@ -823,6 +816,7 @@ static int fs_enet_close(struct net_device *dev)
823 816
824 netif_stop_queue(dev); 817 netif_stop_queue(dev);
825 netif_carrier_off(dev); 818 netif_carrier_off(dev);
819 napi_disable(&fep->napi);
826 phy_stop(fep->phydev); 820 phy_stop(fep->phydev);
827 821
828 spin_lock_irqsave(&fep->lock, flags); 822 spin_lock_irqsave(&fep->lock, flags);
@@ -1047,10 +1041,9 @@ static struct net_device *fs_init_instance(struct device *dev,
1047 ndev->stop = fs_enet_close; 1041 ndev->stop = fs_enet_close;
1048 ndev->get_stats = fs_enet_get_stats; 1042 ndev->get_stats = fs_enet_get_stats;
1049 ndev->set_multicast_list = fs_set_multicast_list; 1043 ndev->set_multicast_list = fs_set_multicast_list;
1050 if (fpi->use_napi) { 1044 netif_napi_add(ndev, &fep->napi,
1051 ndev->poll = fs_enet_rx_napi; 1045 fs_enet_rx_napi, fpi->napi_weight);
1052 ndev->weight = fpi->napi_weight; 1046
1053 }
1054 ndev->ethtool_ops = &fs_ethtool_ops; 1047 ndev->ethtool_ops = &fs_ethtool_ops;
1055 ndev->do_ioctl = fs_ioctl; 1048 ndev->do_ioctl = fs_ioctl;
1056 1049
diff --git a/drivers/net/fs_enet/fs_enet.h b/drivers/net/fs_enet/fs_enet.h
index 569be225cd05..46d0606b1439 100644
--- a/drivers/net/fs_enet/fs_enet.h
+++ b/drivers/net/fs_enet/fs_enet.h
@@ -121,6 +121,7 @@ struct fs_enet_mii_bus {
121}; 121};
122 122
123struct fs_enet_private { 123struct fs_enet_private {
124 struct napi_struct napi;
124 struct device *dev; /* pointer back to the device (must be initialized first) */ 125 struct device *dev; /* pointer back to the device (must be initialized first) */
125 spinlock_t lock; /* during all ops except TX pckt processing */ 126 spinlock_t lock; /* during all ops except TX pckt processing */
126 spinlock_t tx_lock; /* during fs_start_xmit and fs_tx */ 127 spinlock_t tx_lock; /* during fs_start_xmit and fs_tx */
diff --git a/drivers/net/gianfar.c b/drivers/net/gianfar.c
index f92690555dd9..bd2de325bbdd 100644
--- a/drivers/net/gianfar.c
+++ b/drivers/net/gianfar.c
@@ -134,7 +134,7 @@ static void gfar_configure_serdes(struct net_device *dev);
134extern int gfar_local_mdio_write(struct gfar_mii *regs, int mii_id, int regnum, u16 value); 134extern int gfar_local_mdio_write(struct gfar_mii *regs, int mii_id, int regnum, u16 value);
135extern int gfar_local_mdio_read(struct gfar_mii *regs, int mii_id, int regnum); 135extern int gfar_local_mdio_read(struct gfar_mii *regs, int mii_id, int regnum);
136#ifdef CONFIG_GFAR_NAPI 136#ifdef CONFIG_GFAR_NAPI
137static int gfar_poll(struct net_device *dev, int *budget); 137static int gfar_poll(struct napi_struct *napi, int budget);
138#endif 138#endif
139#ifdef CONFIG_NET_POLL_CONTROLLER 139#ifdef CONFIG_NET_POLL_CONTROLLER
140static void gfar_netpoll(struct net_device *dev); 140static void gfar_netpoll(struct net_device *dev);
@@ -188,6 +188,7 @@ static int gfar_probe(struct platform_device *pdev)
188 return -ENOMEM; 188 return -ENOMEM;
189 189
190 priv = netdev_priv(dev); 190 priv = netdev_priv(dev);
191 priv->dev = dev;
191 192
192 /* Set the info in the priv to the current info */ 193 /* Set the info in the priv to the current info */
193 priv->einfo = einfo; 194 priv->einfo = einfo;
@@ -261,10 +262,7 @@ static int gfar_probe(struct platform_device *pdev)
261 dev->hard_start_xmit = gfar_start_xmit; 262 dev->hard_start_xmit = gfar_start_xmit;
262 dev->tx_timeout = gfar_timeout; 263 dev->tx_timeout = gfar_timeout;
263 dev->watchdog_timeo = TX_TIMEOUT; 264 dev->watchdog_timeo = TX_TIMEOUT;
264#ifdef CONFIG_GFAR_NAPI 265 netif_napi_add(dev, &priv->napi, gfar_poll, GFAR_DEV_WEIGHT);
265 dev->poll = gfar_poll;
266 dev->weight = GFAR_DEV_WEIGHT;
267#endif
268#ifdef CONFIG_NET_POLL_CONTROLLER 266#ifdef CONFIG_NET_POLL_CONTROLLER
269 dev->poll_controller = gfar_netpoll; 267 dev->poll_controller = gfar_netpoll;
270#endif 268#endif
@@ -939,6 +937,8 @@ static int gfar_enet_open(struct net_device *dev)
939{ 937{
940 int err; 938 int err;
941 939
940 napi_enable(&priv->napi);
941
942 /* Initialize a bunch of registers */ 942 /* Initialize a bunch of registers */
943 init_registers(dev); 943 init_registers(dev);
944 944
@@ -946,10 +946,14 @@ static int gfar_enet_open(struct net_device *dev)
946 946
947 err = init_phy(dev); 947 err = init_phy(dev);
948 948
949 if(err) 949 if(err) {
950 napi_disable(&priv->napi);
950 return err; 951 return err;
952 }
951 953
952 err = startup_gfar(dev); 954 err = startup_gfar(dev);
955 if (err)
956 napi_disable(&priv->napi);
953 957
954 netif_start_queue(dev); 958 netif_start_queue(dev);
955 959
@@ -1102,6 +1106,9 @@ static int gfar_start_xmit(struct sk_buff *skb, struct net_device *dev)
1102static int gfar_close(struct net_device *dev) 1106static int gfar_close(struct net_device *dev)
1103{ 1107{
1104 struct gfar_private *priv = netdev_priv(dev); 1108 struct gfar_private *priv = netdev_priv(dev);
1109
1110 napi_disable(&priv->napi);
1111
1105 stop_gfar(dev); 1112 stop_gfar(dev);
1106 1113
1107 /* Disconnect from the PHY */ 1114 /* Disconnect from the PHY */
@@ -1318,7 +1325,7 @@ struct sk_buff * gfar_new_skb(struct net_device *dev, struct rxbd8 *bdp)
1318 return NULL; 1325 return NULL;
1319 1326
1320 alignamount = RXBUF_ALIGNMENT - 1327 alignamount = RXBUF_ALIGNMENT -
1321 (((unsigned) skb->data) & (RXBUF_ALIGNMENT - 1)); 1328 (((unsigned long) skb->data) & (RXBUF_ALIGNMENT - 1));
1322 1329
1323 /* We need the data buffer to be aligned properly. We will reserve 1330 /* We need the data buffer to be aligned properly. We will reserve
1324 * as many bytes as needed to align the data properly 1331 * as many bytes as needed to align the data properly
@@ -1390,12 +1397,12 @@ irqreturn_t gfar_receive(int irq, void *dev_id)
1390 1397
1391 /* support NAPI */ 1398 /* support NAPI */
1392#ifdef CONFIG_GFAR_NAPI 1399#ifdef CONFIG_GFAR_NAPI
1393 if (netif_rx_schedule_prep(dev)) { 1400 if (netif_rx_schedule_prep(dev, &priv->napi)) {
1394 tempval = gfar_read(&priv->regs->imask); 1401 tempval = gfar_read(&priv->regs->imask);
1395 tempval &= IMASK_RX_DISABLED; 1402 tempval &= IMASK_RX_DISABLED;
1396 gfar_write(&priv->regs->imask, tempval); 1403 gfar_write(&priv->regs->imask, tempval);
1397 1404
1398 __netif_rx_schedule(dev); 1405 __netif_rx_schedule(dev, &priv->napi);
1399 } else { 1406 } else {
1400 if (netif_msg_rx_err(priv)) 1407 if (netif_msg_rx_err(priv))
1401 printk(KERN_DEBUG "%s: receive called twice (%x)[%x]\n", 1408 printk(KERN_DEBUG "%s: receive called twice (%x)[%x]\n",
@@ -1569,23 +1576,16 @@ int gfar_clean_rx_ring(struct net_device *dev, int rx_work_limit)
1569} 1576}
1570 1577
1571#ifdef CONFIG_GFAR_NAPI 1578#ifdef CONFIG_GFAR_NAPI
1572static int gfar_poll(struct net_device *dev, int *budget) 1579static int gfar_poll(struct napi_struct *napi, int budget)
1573{ 1580{
1581 struct gfar_private *priv = container_of(napi, struct gfar_private, napi);
1582 struct net_device *dev = priv->dev;
1574 int howmany; 1583 int howmany;
1575 struct gfar_private *priv = netdev_priv(dev);
1576 int rx_work_limit = *budget;
1577
1578 if (rx_work_limit > dev->quota)
1579 rx_work_limit = dev->quota;
1580 1584
1581 howmany = gfar_clean_rx_ring(dev, rx_work_limit); 1585 howmany = gfar_clean_rx_ring(dev, budget);
1582 1586
1583 dev->quota -= howmany; 1587 if (howmany < budget) {
1584 rx_work_limit -= howmany; 1588 netif_rx_complete(dev, napi);
1585 *budget -= howmany;
1586
1587 if (rx_work_limit > 0) {
1588 netif_rx_complete(dev);
1589 1589
1590 /* Clear the halt bit in RSTAT */ 1590 /* Clear the halt bit in RSTAT */
1591 gfar_write(&priv->regs->rstat, RSTAT_CLEAR_RHALT); 1591 gfar_write(&priv->regs->rstat, RSTAT_CLEAR_RHALT);
@@ -1601,8 +1601,7 @@ static int gfar_poll(struct net_device *dev, int *budget)
1601 gfar_write(&priv->regs->rxic, 0); 1601 gfar_write(&priv->regs->rxic, 0);
1602 } 1602 }
1603 1603
1604 /* Return 1 if there's more work to do */ 1604 return howmany;
1605 return (rx_work_limit > 0) ? 0 : 1;
1606} 1605}
1607#endif 1606#endif
1608 1607
diff --git a/drivers/net/gianfar.h b/drivers/net/gianfar.h
index d8e779c102fa..b8714e00482d 100644
--- a/drivers/net/gianfar.h
+++ b/drivers/net/gianfar.h
@@ -691,6 +691,9 @@ struct gfar_private {
691 /* RX Locked fields */ 691 /* RX Locked fields */
692 spinlock_t rxlock; 692 spinlock_t rxlock;
693 693
694 struct net_device *dev;
695 struct napi_struct napi;
696
694 /* skb array and index */ 697 /* skb array and index */
695 struct sk_buff ** rx_skbuff; 698 struct sk_buff ** rx_skbuff;
696 u16 skb_currx; 699 u16 skb_currx;
diff --git a/drivers/net/ibmveth.c b/drivers/net/ibmveth.c
index acba90f1638e..78e28ada1e21 100644
--- a/drivers/net/ibmveth.c
+++ b/drivers/net/ibmveth.c
@@ -83,7 +83,7 @@
83static int ibmveth_open(struct net_device *dev); 83static int ibmveth_open(struct net_device *dev);
84static int ibmveth_close(struct net_device *dev); 84static int ibmveth_close(struct net_device *dev);
85static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd); 85static int ibmveth_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd);
86static int ibmveth_poll(struct net_device *dev, int *budget); 86static int ibmveth_poll(struct napi_struct *napi, int budget);
87static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *dev); 87static int ibmveth_start_xmit(struct sk_buff *skb, struct net_device *dev);
88static struct net_device_stats *ibmveth_get_stats(struct net_device *dev); 88static struct net_device_stats *ibmveth_get_stats(struct net_device *dev);
89static void ibmveth_set_multicast_list(struct net_device *dev); 89static void ibmveth_set_multicast_list(struct net_device *dev);
@@ -480,6 +480,8 @@ static int ibmveth_open(struct net_device *netdev)
480 480
481 ibmveth_debug_printk("open starting\n"); 481 ibmveth_debug_printk("open starting\n");
482 482
483 napi_enable(&adapter->napi);
484
483 for(i = 0; i<IbmVethNumBufferPools; i++) 485 for(i = 0; i<IbmVethNumBufferPools; i++)
484 rxq_entries += adapter->rx_buff_pool[i].size; 486 rxq_entries += adapter->rx_buff_pool[i].size;
485 487
@@ -489,6 +491,7 @@ static int ibmveth_open(struct net_device *netdev)
489 if(!adapter->buffer_list_addr || !adapter->filter_list_addr) { 491 if(!adapter->buffer_list_addr || !adapter->filter_list_addr) {
490 ibmveth_error_printk("unable to allocate filter or buffer list pages\n"); 492 ibmveth_error_printk("unable to allocate filter or buffer list pages\n");
491 ibmveth_cleanup(adapter); 493 ibmveth_cleanup(adapter);
494 napi_disable(&adapter->napi);
492 return -ENOMEM; 495 return -ENOMEM;
493 } 496 }
494 497
@@ -498,6 +501,7 @@ static int ibmveth_open(struct net_device *netdev)
498 if(!adapter->rx_queue.queue_addr) { 501 if(!adapter->rx_queue.queue_addr) {
499 ibmveth_error_printk("unable to allocate rx queue pages\n"); 502 ibmveth_error_printk("unable to allocate rx queue pages\n");
500 ibmveth_cleanup(adapter); 503 ibmveth_cleanup(adapter);
504 napi_disable(&adapter->napi);
501 return -ENOMEM; 505 return -ENOMEM;
502 } 506 }
503 507
@@ -514,6 +518,7 @@ static int ibmveth_open(struct net_device *netdev)
514 (dma_mapping_error(adapter->rx_queue.queue_dma))) { 518 (dma_mapping_error(adapter->rx_queue.queue_dma))) {
515 ibmveth_error_printk("unable to map filter or buffer list pages\n"); 519 ibmveth_error_printk("unable to map filter or buffer list pages\n");
516 ibmveth_cleanup(adapter); 520 ibmveth_cleanup(adapter);
521 napi_disable(&adapter->napi);
517 return -ENOMEM; 522 return -ENOMEM;
518 } 523 }
519 524
@@ -545,6 +550,7 @@ static int ibmveth_open(struct net_device *netdev)
545 rxq_desc.desc, 550 rxq_desc.desc,
546 mac_address); 551 mac_address);
547 ibmveth_cleanup(adapter); 552 ibmveth_cleanup(adapter);
553 napi_disable(&adapter->napi);
548 return -ENONET; 554 return -ENONET;
549 } 555 }
550 556
@@ -555,6 +561,7 @@ static int ibmveth_open(struct net_device *netdev)
555 ibmveth_error_printk("unable to alloc pool\n"); 561 ibmveth_error_printk("unable to alloc pool\n");
556 adapter->rx_buff_pool[i].active = 0; 562 adapter->rx_buff_pool[i].active = 0;
557 ibmveth_cleanup(adapter); 563 ibmveth_cleanup(adapter);
564 napi_disable(&adapter->napi);
558 return -ENOMEM ; 565 return -ENOMEM ;
559 } 566 }
560 } 567 }
@@ -567,6 +574,7 @@ static int ibmveth_open(struct net_device *netdev)
567 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY)); 574 } while (H_IS_LONG_BUSY(rc) || (rc == H_BUSY));
568 575
569 ibmveth_cleanup(adapter); 576 ibmveth_cleanup(adapter);
577 napi_disable(&adapter->napi);
570 return rc; 578 return rc;
571 } 579 }
572 580
@@ -587,6 +595,8 @@ static int ibmveth_close(struct net_device *netdev)
587 595
588 ibmveth_debug_printk("close starting\n"); 596 ibmveth_debug_printk("close starting\n");
589 597
598 napi_disable(&adapter->napi);
599
590 if (!adapter->pool_config) 600 if (!adapter->pool_config)
591 netif_stop_queue(netdev); 601 netif_stop_queue(netdev);
592 602
@@ -767,80 +777,68 @@ out: spin_lock_irqsave(&adapter->stats_lock, flags);
767 return 0; 777 return 0;
768} 778}
769 779
770static int ibmveth_poll(struct net_device *netdev, int *budget) 780static int ibmveth_poll(struct napi_struct *napi, int budget)
771{ 781{
772 struct ibmveth_adapter *adapter = netdev->priv; 782 struct ibmveth_adapter *adapter = container_of(napi, struct ibmveth_adapter, napi);
773 int max_frames_to_process = netdev->quota; 783 struct net_device *netdev = adapter->netdev;
774 int frames_processed = 0; 784 int frames_processed = 0;
775 int more_work = 1;
776 unsigned long lpar_rc; 785 unsigned long lpar_rc;
777 786
778 restart_poll: 787 restart_poll:
779 do { 788 do {
780 struct net_device *netdev = adapter->netdev; 789 struct sk_buff *skb;
781
782 if(ibmveth_rxq_pending_buffer(adapter)) {
783 struct sk_buff *skb;
784 790
785 rmb(); 791 if (!ibmveth_rxq_pending_buffer(adapter))
792 break;
786 793
787 if(!ibmveth_rxq_buffer_valid(adapter)) { 794 rmb();
788 wmb(); /* suggested by larson1 */ 795 if (!ibmveth_rxq_buffer_valid(adapter)) {
789 adapter->rx_invalid_buffer++; 796 wmb(); /* suggested by larson1 */
790 ibmveth_debug_printk("recycling invalid buffer\n"); 797 adapter->rx_invalid_buffer++;
791 ibmveth_rxq_recycle_buffer(adapter); 798 ibmveth_debug_printk("recycling invalid buffer\n");
792 } else { 799 ibmveth_rxq_recycle_buffer(adapter);
793 int length = ibmveth_rxq_frame_length(adapter); 800 } else {
794 int offset = ibmveth_rxq_frame_offset(adapter); 801 int length = ibmveth_rxq_frame_length(adapter);
795 skb = ibmveth_rxq_get_buffer(adapter); 802 int offset = ibmveth_rxq_frame_offset(adapter);
803 skb = ibmveth_rxq_get_buffer(adapter);
796 804
797 ibmveth_rxq_harvest_buffer(adapter); 805 ibmveth_rxq_harvest_buffer(adapter);
798 806
799 skb_reserve(skb, offset); 807 skb_reserve(skb, offset);
800 skb_put(skb, length); 808 skb_put(skb, length);
801 skb->protocol = eth_type_trans(skb, netdev); 809 skb->protocol = eth_type_trans(skb, netdev);
802 810
803 netif_receive_skb(skb); /* send it up */ 811 netif_receive_skb(skb); /* send it up */
804 812
805 adapter->stats.rx_packets++; 813 adapter->stats.rx_packets++;
806 adapter->stats.rx_bytes += length; 814 adapter->stats.rx_bytes += length;
807 frames_processed++; 815 frames_processed++;
808 netdev->last_rx = jiffies; 816 netdev->last_rx = jiffies;
809 }
810 } else {
811 more_work = 0;
812 } 817 }
813 } while(more_work && (frames_processed < max_frames_to_process)); 818 } while (frames_processed < budget);
814 819
815 ibmveth_replenish_task(adapter); 820 ibmveth_replenish_task(adapter);
816 821
817 if(more_work) { 822 if (frames_processed < budget) {
818 /* more work to do - return that we are not done yet */ 823 /* We think we are done - reenable interrupts,
819 netdev->quota -= frames_processed; 824 * then check once more to make sure we are done.
820 *budget -= frames_processed; 825 */
821 return 1; 826 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
822 } 827 VIO_IRQ_ENABLE);
823
824 /* we think we are done - reenable interrupts, then check once more to make sure we are done */
825 lpar_rc = h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_ENABLE);
826 828
827 ibmveth_assert(lpar_rc == H_SUCCESS); 829 ibmveth_assert(lpar_rc == H_SUCCESS);
828 830
829 netif_rx_complete(netdev); 831 netif_rx_complete(netdev, napi);
830 832
831 if(ibmveth_rxq_pending_buffer(adapter) && netif_rx_reschedule(netdev, frames_processed)) 833 if (ibmveth_rxq_pending_buffer(adapter) &&
832 { 834 netif_rx_reschedule(netdev, napi)) {
833 lpar_rc = h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_DISABLE); 835 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
834 ibmveth_assert(lpar_rc == H_SUCCESS); 836 VIO_IRQ_DISABLE);
835 more_work = 1; 837 goto restart_poll;
836 goto restart_poll; 838 }
837 } 839 }
838 840
839 netdev->quota -= frames_processed; 841 return frames_processed;
840 *budget -= frames_processed;
841
842 /* we really are done */
843 return 0;
844} 842}
845 843
846static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance) 844static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance)
@@ -849,10 +847,11 @@ static irqreturn_t ibmveth_interrupt(int irq, void *dev_instance)
849 struct ibmveth_adapter *adapter = netdev->priv; 847 struct ibmveth_adapter *adapter = netdev->priv;
850 unsigned long lpar_rc; 848 unsigned long lpar_rc;
851 849
852 if(netif_rx_schedule_prep(netdev)) { 850 if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
853 lpar_rc = h_vio_signal(adapter->vdev->unit_address, VIO_IRQ_DISABLE); 851 lpar_rc = h_vio_signal(adapter->vdev->unit_address,
852 VIO_IRQ_DISABLE);
854 ibmveth_assert(lpar_rc == H_SUCCESS); 853 ibmveth_assert(lpar_rc == H_SUCCESS);
855 __netif_rx_schedule(netdev); 854 __netif_rx_schedule(netdev, &adapter->napi);
856 } 855 }
857 return IRQ_HANDLED; 856 return IRQ_HANDLED;
858} 857}
@@ -1004,6 +1003,8 @@ static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_
1004 adapter->mcastFilterSize= *mcastFilterSize_p; 1003 adapter->mcastFilterSize= *mcastFilterSize_p;
1005 adapter->pool_config = 0; 1004 adapter->pool_config = 0;
1006 1005
1006 netif_napi_add(netdev, &adapter->napi, ibmveth_poll, 16);
1007
1007 /* Some older boxes running PHYP non-natively have an OF that 1008 /* Some older boxes running PHYP non-natively have an OF that
1008 returns a 8-byte local-mac-address field (and the first 1009 returns a 8-byte local-mac-address field (and the first
1009 2 bytes have to be ignored) while newer boxes' OF return 1010 2 bytes have to be ignored) while newer boxes' OF return
@@ -1020,8 +1021,6 @@ static int __devinit ibmveth_probe(struct vio_dev *dev, const struct vio_device_
1020 1021
1021 netdev->irq = dev->irq; 1022 netdev->irq = dev->irq;
1022 netdev->open = ibmveth_open; 1023 netdev->open = ibmveth_open;
1023 netdev->poll = ibmveth_poll;
1024 netdev->weight = 16;
1025 netdev->stop = ibmveth_close; 1024 netdev->stop = ibmveth_close;
1026 netdev->hard_start_xmit = ibmveth_start_xmit; 1025 netdev->hard_start_xmit = ibmveth_start_xmit;
1027 netdev->get_stats = ibmveth_get_stats; 1026 netdev->get_stats = ibmveth_get_stats;
diff --git a/drivers/net/ibmveth.h b/drivers/net/ibmveth.h
index 72cc15a6cab7..e05694126f85 100644
--- a/drivers/net/ibmveth.h
+++ b/drivers/net/ibmveth.h
@@ -112,6 +112,7 @@ struct ibmveth_rx_q {
112struct ibmveth_adapter { 112struct ibmveth_adapter {
113 struct vio_dev *vdev; 113 struct vio_dev *vdev;
114 struct net_device *netdev; 114 struct net_device *netdev;
115 struct napi_struct napi;
115 struct net_device_stats stats; 116 struct net_device_stats stats;
116 unsigned int mcastFilterSize; 117 unsigned int mcastFilterSize;
117 unsigned long mac_addr; 118 unsigned long mac_addr;
diff --git a/drivers/net/ixgb/ixgb.h b/drivers/net/ixgb/ixgb.h
index 3569d5b03388..1eee8894c732 100644
--- a/drivers/net/ixgb/ixgb.h
+++ b/drivers/net/ixgb/ixgb.h
@@ -184,6 +184,7 @@ struct ixgb_adapter {
184 boolean_t rx_csum; 184 boolean_t rx_csum;
185 185
186 /* OS defined structs */ 186 /* OS defined structs */
187 struct napi_struct napi;
187 struct net_device *netdev; 188 struct net_device *netdev;
188 struct pci_dev *pdev; 189 struct pci_dev *pdev;
189 struct net_device_stats net_stats; 190 struct net_device_stats net_stats;
diff --git a/drivers/net/ixgb/ixgb_main.c b/drivers/net/ixgb/ixgb_main.c
index 991c8833e23c..e3f27c67fb28 100644
--- a/drivers/net/ixgb/ixgb_main.c
+++ b/drivers/net/ixgb/ixgb_main.c
@@ -97,7 +97,7 @@ static irqreturn_t ixgb_intr(int irq, void *data);
97static boolean_t ixgb_clean_tx_irq(struct ixgb_adapter *adapter); 97static boolean_t ixgb_clean_tx_irq(struct ixgb_adapter *adapter);
98 98
99#ifdef CONFIG_IXGB_NAPI 99#ifdef CONFIG_IXGB_NAPI
100static int ixgb_clean(struct net_device *netdev, int *budget); 100static int ixgb_clean(struct napi_struct *napi, int budget);
101static boolean_t ixgb_clean_rx_irq(struct ixgb_adapter *adapter, 101static boolean_t ixgb_clean_rx_irq(struct ixgb_adapter *adapter,
102 int *work_done, int work_to_do); 102 int *work_done, int work_to_do);
103#else 103#else
@@ -288,7 +288,7 @@ ixgb_up(struct ixgb_adapter *adapter)
288 mod_timer(&adapter->watchdog_timer, jiffies); 288 mod_timer(&adapter->watchdog_timer, jiffies);
289 289
290#ifdef CONFIG_IXGB_NAPI 290#ifdef CONFIG_IXGB_NAPI
291 netif_poll_enable(netdev); 291 napi_enable(&adapter->napi);
292#endif 292#endif
293 ixgb_irq_enable(adapter); 293 ixgb_irq_enable(adapter);
294 294
@@ -309,7 +309,7 @@ ixgb_down(struct ixgb_adapter *adapter, boolean_t kill_watchdog)
309 if(kill_watchdog) 309 if(kill_watchdog)
310 del_timer_sync(&adapter->watchdog_timer); 310 del_timer_sync(&adapter->watchdog_timer);
311#ifdef CONFIG_IXGB_NAPI 311#ifdef CONFIG_IXGB_NAPI
312 netif_poll_disable(netdev); 312 napi_disable(&adapter->napi);
313#endif 313#endif
314 adapter->link_speed = 0; 314 adapter->link_speed = 0;
315 adapter->link_duplex = 0; 315 adapter->link_duplex = 0;
@@ -421,8 +421,7 @@ ixgb_probe(struct pci_dev *pdev,
421 netdev->tx_timeout = &ixgb_tx_timeout; 421 netdev->tx_timeout = &ixgb_tx_timeout;
422 netdev->watchdog_timeo = 5 * HZ; 422 netdev->watchdog_timeo = 5 * HZ;
423#ifdef CONFIG_IXGB_NAPI 423#ifdef CONFIG_IXGB_NAPI
424 netdev->poll = &ixgb_clean; 424 netif_napi_add(netdev, &adapter->napi, ixgb_clean, 64);
425 netdev->weight = 64;
426#endif 425#endif
427 netdev->vlan_rx_register = ixgb_vlan_rx_register; 426 netdev->vlan_rx_register = ixgb_vlan_rx_register;
428 netdev->vlan_rx_add_vid = ixgb_vlan_rx_add_vid; 427 netdev->vlan_rx_add_vid = ixgb_vlan_rx_add_vid;
@@ -1746,7 +1745,7 @@ ixgb_intr(int irq, void *data)
1746 } 1745 }
1747 1746
1748#ifdef CONFIG_IXGB_NAPI 1747#ifdef CONFIG_IXGB_NAPI
1749 if(netif_rx_schedule_prep(netdev)) { 1748 if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
1750 1749
1751 /* Disable interrupts and register for poll. The flush 1750 /* Disable interrupts and register for poll. The flush
1752 of the posted write is intentionally left out. 1751 of the posted write is intentionally left out.
@@ -1754,7 +1753,7 @@ ixgb_intr(int irq, void *data)
1754 1753
1755 atomic_inc(&adapter->irq_sem); 1754 atomic_inc(&adapter->irq_sem);
1756 IXGB_WRITE_REG(&adapter->hw, IMC, ~0); 1755 IXGB_WRITE_REG(&adapter->hw, IMC, ~0);
1757 __netif_rx_schedule(netdev); 1756 __netif_rx_schedule(netdev, &adapter->napi);
1758 } 1757 }
1759#else 1758#else
1760 /* yes, that is actually a & and it is meant to make sure that 1759 /* yes, that is actually a & and it is meant to make sure that
@@ -1776,27 +1775,23 @@ ixgb_intr(int irq, void *data)
1776 **/ 1775 **/
1777 1776
1778static int 1777static int
1779ixgb_clean(struct net_device *netdev, int *budget) 1778ixgb_clean(struct napi_struct *napi, int budget)
1780{ 1779{
1781 struct ixgb_adapter *adapter = netdev_priv(netdev); 1780 struct ixgb_adapter *adapter = container_of(napi, struct ixgb_adapter, napi);
1782 int work_to_do = min(*budget, netdev->quota); 1781 struct net_device *netdev = adapter->netdev;
1783 int tx_cleaned; 1782 int tx_cleaned;
1784 int work_done = 0; 1783 int work_done = 0;
1785 1784
1786 tx_cleaned = ixgb_clean_tx_irq(adapter); 1785 tx_cleaned = ixgb_clean_tx_irq(adapter);
1787 ixgb_clean_rx_irq(adapter, &work_done, work_to_do); 1786 ixgb_clean_rx_irq(adapter, &work_done, budget);
1788
1789 *budget -= work_done;
1790 netdev->quota -= work_done;
1791 1787
1792 /* if no Tx and not enough Rx work done, exit the polling mode */ 1788 /* if no Tx and not enough Rx work done, exit the polling mode */
1793 if((!tx_cleaned && (work_done == 0)) || !netif_running(netdev)) { 1789 if((!tx_cleaned && (work_done == 0)) || !netif_running(netdev)) {
1794 netif_rx_complete(netdev); 1790 netif_rx_complete(netdev, napi);
1795 ixgb_irq_enable(adapter); 1791 ixgb_irq_enable(adapter);
1796 return 0;
1797 } 1792 }
1798 1793
1799 return 1; 1794 return work_done;
1800} 1795}
1801#endif 1796#endif
1802 1797
diff --git a/drivers/net/ixp2000/ixpdev.c b/drivers/net/ixp2000/ixpdev.c
index d9ce1aef148a..6c0dd49149d0 100644
--- a/drivers/net/ixp2000/ixpdev.c
+++ b/drivers/net/ixp2000/ixpdev.c
@@ -74,9 +74,9 @@ static int ixpdev_xmit(struct sk_buff *skb, struct net_device *dev)
74} 74}
75 75
76 76
77static int ixpdev_rx(struct net_device *dev, int *budget) 77static int ixpdev_rx(struct net_device *dev, int processed, int budget)
78{ 78{
79 while (*budget > 0) { 79 while (processed < budget) {
80 struct ixpdev_rx_desc *desc; 80 struct ixpdev_rx_desc *desc;
81 struct sk_buff *skb; 81 struct sk_buff *skb;
82 void *buf; 82 void *buf;
@@ -122,29 +122,34 @@ static int ixpdev_rx(struct net_device *dev, int *budget)
122 122
123err: 123err:
124 ixp2000_reg_write(RING_RX_PENDING, _desc); 124 ixp2000_reg_write(RING_RX_PENDING, _desc);
125 dev->quota--; 125 processed++;
126 (*budget)--;
127 } 126 }
128 127
129 return 1; 128 return processed;
130} 129}
131 130
132/* dev always points to nds[0]. */ 131/* dev always points to nds[0]. */
133static int ixpdev_poll(struct net_device *dev, int *budget) 132static int ixpdev_poll(struct napi_struct *napi, int budget)
134{ 133{
134 struct ixpdev_priv *ip = container_of(napi, struct ixpdev_priv, napi);
135 struct net_device *dev = ip->dev;
136 int rx;
137
135 /* @@@ Have to stop polling when nds[0] is administratively 138 /* @@@ Have to stop polling when nds[0] is administratively
136 * downed while we are polling. */ 139 * downed while we are polling. */
140 rx = 0;
137 do { 141 do {
138 ixp2000_reg_write(IXP2000_IRQ_THD_RAW_STATUS_A_0, 0x00ff); 142 ixp2000_reg_write(IXP2000_IRQ_THD_RAW_STATUS_A_0, 0x00ff);
139 143
140 if (ixpdev_rx(dev, budget)) 144 rx = ixpdev_rx(dev, rx, budget);
141 return 1; 145 if (rx >= budget)
146 break;
142 } while (ixp2000_reg_read(IXP2000_IRQ_THD_RAW_STATUS_A_0) & 0x00ff); 147 } while (ixp2000_reg_read(IXP2000_IRQ_THD_RAW_STATUS_A_0) & 0x00ff);
143 148
144 netif_rx_complete(dev); 149 netif_rx_complete(dev, napi);
145 ixp2000_reg_write(IXP2000_IRQ_THD_ENABLE_SET_A_0, 0x00ff); 150 ixp2000_reg_write(IXP2000_IRQ_THD_ENABLE_SET_A_0, 0x00ff);
146 151
147 return 0; 152 return rx;
148} 153}
149 154
150static void ixpdev_tx_complete(void) 155static void ixpdev_tx_complete(void)
@@ -199,9 +204,12 @@ static irqreturn_t ixpdev_interrupt(int irq, void *dev_id)
199 * Any of the eight receive units signaled RX? 204 * Any of the eight receive units signaled RX?
200 */ 205 */
201 if (status & 0x00ff) { 206 if (status & 0x00ff) {
207 struct net_device *dev = nds[0];
208 struct ixpdev_priv *ip = netdev_priv(dev);
209
202 ixp2000_reg_wrb(IXP2000_IRQ_THD_ENABLE_CLEAR_A_0, 0x00ff); 210 ixp2000_reg_wrb(IXP2000_IRQ_THD_ENABLE_CLEAR_A_0, 0x00ff);
203 if (likely(__netif_rx_schedule_prep(nds[0]))) { 211 if (likely(napi_schedule_prep(&ip->napi))) {
204 __netif_rx_schedule(nds[0]); 212 __netif_rx_schedule(dev, &ip->napi);
205 } else { 213 } else {
206 printk(KERN_CRIT "ixp2000: irq while polling!!\n"); 214 printk(KERN_CRIT "ixp2000: irq while polling!!\n");
207 } 215 }
@@ -232,11 +240,13 @@ static int ixpdev_open(struct net_device *dev)
232 struct ixpdev_priv *ip = netdev_priv(dev); 240 struct ixpdev_priv *ip = netdev_priv(dev);
233 int err; 241 int err;
234 242
243 napi_enable(&ip->napi);
235 if (!nds_open++) { 244 if (!nds_open++) {
236 err = request_irq(IRQ_IXP2000_THDA0, ixpdev_interrupt, 245 err = request_irq(IRQ_IXP2000_THDA0, ixpdev_interrupt,
237 IRQF_SHARED, "ixp2000_eth", nds); 246 IRQF_SHARED, "ixp2000_eth", nds);
238 if (err) { 247 if (err) {
239 nds_open--; 248 nds_open--;
249 napi_disable(&ip->napi);
240 return err; 250 return err;
241 } 251 }
242 252
@@ -254,6 +264,7 @@ static int ixpdev_close(struct net_device *dev)
254 struct ixpdev_priv *ip = netdev_priv(dev); 264 struct ixpdev_priv *ip = netdev_priv(dev);
255 265
256 netif_stop_queue(dev); 266 netif_stop_queue(dev);
267 napi_disable(&ip->napi);
257 set_port_admin_status(ip->channel, 0); 268 set_port_admin_status(ip->channel, 0);
258 269
259 if (!--nds_open) { 270 if (!--nds_open) {
@@ -274,7 +285,6 @@ struct net_device *ixpdev_alloc(int channel, int sizeof_priv)
274 return NULL; 285 return NULL;
275 286
276 dev->hard_start_xmit = ixpdev_xmit; 287 dev->hard_start_xmit = ixpdev_xmit;
277 dev->poll = ixpdev_poll;
278 dev->open = ixpdev_open; 288 dev->open = ixpdev_open;
279 dev->stop = ixpdev_close; 289 dev->stop = ixpdev_close;
280#ifdef CONFIG_NET_POLL_CONTROLLER 290#ifdef CONFIG_NET_POLL_CONTROLLER
@@ -282,9 +292,10 @@ struct net_device *ixpdev_alloc(int channel, int sizeof_priv)
282#endif 292#endif
283 293
284 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM; 294 dev->features |= NETIF_F_SG | NETIF_F_HW_CSUM;
285 dev->weight = 64;
286 295
287 ip = netdev_priv(dev); 296 ip = netdev_priv(dev);
297 ip->dev = dev;
298 netif_napi_add(dev, &ip->napi, ixpdev_poll, 64);
288 ip->channel = channel; 299 ip->channel = channel;
289 ip->tx_queue_entries = 0; 300 ip->tx_queue_entries = 0;
290 301
diff --git a/drivers/net/ixp2000/ixpdev.h b/drivers/net/ixp2000/ixpdev.h
index bd686cb63058..391ece623243 100644
--- a/drivers/net/ixp2000/ixpdev.h
+++ b/drivers/net/ixp2000/ixpdev.h
@@ -14,6 +14,8 @@
14 14
15struct ixpdev_priv 15struct ixpdev_priv
16{ 16{
17 struct net_device *dev;
18 struct napi_struct napi;
17 int channel; 19 int channel;
18 int tx_queue_entries; 20 int tx_queue_entries;
19}; 21};
diff --git a/drivers/net/macb.c b/drivers/net/macb.c
index a4bb0264180a..74c3f7a7ae4a 100644
--- a/drivers/net/macb.c
+++ b/drivers/net/macb.c
@@ -470,47 +470,41 @@ static int macb_rx(struct macb *bp, int budget)
470 return received; 470 return received;
471} 471}
472 472
473static int macb_poll(struct net_device *dev, int *budget) 473static int macb_poll(struct napi_struct *napi, int budget)
474{ 474{
475 struct macb *bp = netdev_priv(dev); 475 struct macb *bp = container_of(napi, struct macb, napi);
476 int orig_budget, work_done, retval = 0; 476 struct net_device *dev = bp->dev;
477 int work_done;
477 u32 status; 478 u32 status;
478 479
479 status = macb_readl(bp, RSR); 480 status = macb_readl(bp, RSR);
480 macb_writel(bp, RSR, status); 481 macb_writel(bp, RSR, status);
481 482
483 work_done = 0;
482 if (!status) { 484 if (!status) {
483 /* 485 /*
484 * This may happen if an interrupt was pending before 486 * This may happen if an interrupt was pending before
485 * this function was called last time, and no packets 487 * this function was called last time, and no packets
486 * have been received since. 488 * have been received since.
487 */ 489 */
488 netif_rx_complete(dev); 490 netif_rx_complete(dev, napi);
489 goto out; 491 goto out;
490 } 492 }
491 493
492 dev_dbg(&bp->pdev->dev, "poll: status = %08lx, budget = %d\n", 494 dev_dbg(&bp->pdev->dev, "poll: status = %08lx, budget = %d\n",
493 (unsigned long)status, *budget); 495 (unsigned long)status, budget);
494 496
495 if (!(status & MACB_BIT(REC))) { 497 if (!(status & MACB_BIT(REC))) {
496 dev_warn(&bp->pdev->dev, 498 dev_warn(&bp->pdev->dev,
497 "No RX buffers complete, status = %02lx\n", 499 "No RX buffers complete, status = %02lx\n",
498 (unsigned long)status); 500 (unsigned long)status);
499 netif_rx_complete(dev); 501 netif_rx_complete(dev, napi);
500 goto out; 502 goto out;
501 } 503 }
502 504
503 orig_budget = *budget; 505 work_done = macb_rx(bp, budget);
504 if (orig_budget > dev->quota) 506 if (work_done < budget)
505 orig_budget = dev->quota; 507 netif_rx_complete(dev, napi);
506
507 work_done = macb_rx(bp, orig_budget);
508 if (work_done < orig_budget) {
509 netif_rx_complete(dev);
510 retval = 0;
511 } else {
512 retval = 1;
513 }
514 508
515 /* 509 /*
516 * We've done what we can to clean the buffers. Make sure we 510 * We've done what we can to clean the buffers. Make sure we
@@ -521,7 +515,7 @@ out:
521 515
522 /* TODO: Handle errors */ 516 /* TODO: Handle errors */
523 517
524 return retval; 518 return work_done;
525} 519}
526 520
527static irqreturn_t macb_interrupt(int irq, void *dev_id) 521static irqreturn_t macb_interrupt(int irq, void *dev_id)
@@ -545,7 +539,7 @@ static irqreturn_t macb_interrupt(int irq, void *dev_id)
545 } 539 }
546 540
547 if (status & MACB_RX_INT_FLAGS) { 541 if (status & MACB_RX_INT_FLAGS) {
548 if (netif_rx_schedule_prep(dev)) { 542 if (netif_rx_schedule_prep(dev, &bp->napi)) {
549 /* 543 /*
550 * There's no point taking any more interrupts 544 * There's no point taking any more interrupts
551 * until we have processed the buffers 545 * until we have processed the buffers
@@ -553,7 +547,7 @@ static irqreturn_t macb_interrupt(int irq, void *dev_id)
553 macb_writel(bp, IDR, MACB_RX_INT_FLAGS); 547 macb_writel(bp, IDR, MACB_RX_INT_FLAGS);
554 dev_dbg(&bp->pdev->dev, 548 dev_dbg(&bp->pdev->dev,
555 "scheduling RX softirq\n"); 549 "scheduling RX softirq\n");
556 __netif_rx_schedule(dev); 550 __netif_rx_schedule(dev, &bp->napi);
557 } 551 }
558 } 552 }
559 553
@@ -937,6 +931,8 @@ static int macb_open(struct net_device *dev)
937 return err; 931 return err;
938 } 932 }
939 933
934 napi_enable(&bp->napi);
935
940 macb_init_rings(bp); 936 macb_init_rings(bp);
941 macb_init_hw(bp); 937 macb_init_hw(bp);
942 938
@@ -954,6 +950,7 @@ static int macb_close(struct net_device *dev)
954 unsigned long flags; 950 unsigned long flags;
955 951
956 netif_stop_queue(dev); 952 netif_stop_queue(dev);
953 napi_disable(&bp->napi);
957 954
958 if (bp->phy_dev) 955 if (bp->phy_dev)
959 phy_stop(bp->phy_dev); 956 phy_stop(bp->phy_dev);
@@ -1146,8 +1143,7 @@ static int __devinit macb_probe(struct platform_device *pdev)
1146 dev->get_stats = macb_get_stats; 1143 dev->get_stats = macb_get_stats;
1147 dev->set_multicast_list = macb_set_rx_mode; 1144 dev->set_multicast_list = macb_set_rx_mode;
1148 dev->do_ioctl = macb_ioctl; 1145 dev->do_ioctl = macb_ioctl;
1149 dev->poll = macb_poll; 1146 netif_napi_add(dev, &bp->napi, macb_poll, 64);
1150 dev->weight = 64;
1151 dev->ethtool_ops = &macb_ethtool_ops; 1147 dev->ethtool_ops = &macb_ethtool_ops;
1152 1148
1153 dev->base_addr = regs->start; 1149 dev->base_addr = regs->start;
diff --git a/drivers/net/macb.h b/drivers/net/macb.h
index 4e3283ebd97c..57b85acf0d16 100644
--- a/drivers/net/macb.h
+++ b/drivers/net/macb.h
@@ -374,6 +374,7 @@ struct macb {
374 struct clk *pclk; 374 struct clk *pclk;
375 struct clk *hclk; 375 struct clk *hclk;
376 struct net_device *dev; 376 struct net_device *dev;
377 struct napi_struct napi;
377 struct net_device_stats stats; 378 struct net_device_stats stats;
378 struct macb_stats hw_stats; 379 struct macb_stats hw_stats;
379 380
diff --git a/drivers/net/mv643xx_eth.c b/drivers/net/mv643xx_eth.c
index 315335671f0f..702eba549161 100644
--- a/drivers/net/mv643xx_eth.c
+++ b/drivers/net/mv643xx_eth.c
@@ -66,7 +66,7 @@ static int mv643xx_eth_change_mtu(struct net_device *, int);
66static struct net_device_stats *mv643xx_eth_get_stats(struct net_device *); 66static struct net_device_stats *mv643xx_eth_get_stats(struct net_device *);
67static void eth_port_init_mac_tables(unsigned int eth_port_num); 67static void eth_port_init_mac_tables(unsigned int eth_port_num);
68#ifdef MV643XX_NAPI 68#ifdef MV643XX_NAPI
69static int mv643xx_poll(struct net_device *dev, int *budget); 69static int mv643xx_poll(struct napi_struct *napi, int budget);
70#endif 70#endif
71static int ethernet_phy_get(unsigned int eth_port_num); 71static int ethernet_phy_get(unsigned int eth_port_num);
72static void ethernet_phy_set(unsigned int eth_port_num, int phy_addr); 72static void ethernet_phy_set(unsigned int eth_port_num, int phy_addr);
@@ -562,7 +562,7 @@ static irqreturn_t mv643xx_eth_int_handler(int irq, void *dev_id)
562 /* wait for previous write to complete */ 562 /* wait for previous write to complete */
563 mv_read(MV643XX_ETH_INTERRUPT_MASK_REG(port_num)); 563 mv_read(MV643XX_ETH_INTERRUPT_MASK_REG(port_num));
564 564
565 netif_rx_schedule(dev); 565 netif_rx_schedule(dev, &mp->napi);
566 } 566 }
567#else 567#else
568 if (eth_int_cause & ETH_INT_CAUSE_RX) 568 if (eth_int_cause & ETH_INT_CAUSE_RX)
@@ -880,6 +880,10 @@ static int mv643xx_eth_open(struct net_device *dev)
880 880
881 mv643xx_eth_rx_refill_descs(dev); /* Fill RX ring with skb's */ 881 mv643xx_eth_rx_refill_descs(dev); /* Fill RX ring with skb's */
882 882
883#ifdef MV643XX_NAPI
884 napi_enable(&mp->napi);
885#endif
886
883 eth_port_start(dev); 887 eth_port_start(dev);
884 888
885 /* Interrupt Coalescing */ 889 /* Interrupt Coalescing */
@@ -982,7 +986,7 @@ static int mv643xx_eth_stop(struct net_device *dev)
982 mv_read(MV643XX_ETH_INTERRUPT_MASK_REG(port_num)); 986 mv_read(MV643XX_ETH_INTERRUPT_MASK_REG(port_num));
983 987
984#ifdef MV643XX_NAPI 988#ifdef MV643XX_NAPI
985 netif_poll_disable(dev); 989 napi_disable(&mp->napi);
986#endif 990#endif
987 netif_carrier_off(dev); 991 netif_carrier_off(dev);
988 netif_stop_queue(dev); 992 netif_stop_queue(dev);
@@ -992,10 +996,6 @@ static int mv643xx_eth_stop(struct net_device *dev)
992 mv643xx_eth_free_tx_rings(dev); 996 mv643xx_eth_free_tx_rings(dev);
993 mv643xx_eth_free_rx_rings(dev); 997 mv643xx_eth_free_rx_rings(dev);
994 998
995#ifdef MV643XX_NAPI
996 netif_poll_enable(dev);
997#endif
998
999 free_irq(dev->irq, dev); 999 free_irq(dev->irq, dev);
1000 1000
1001 return 0; 1001 return 0;
@@ -1007,11 +1007,12 @@ static int mv643xx_eth_stop(struct net_device *dev)
1007 * 1007 *
1008 * This function is used in case of NAPI 1008 * This function is used in case of NAPI
1009 */ 1009 */
1010static int mv643xx_poll(struct net_device *dev, int *budget) 1010static int mv643xx_poll(struct napi_struct *napi, int budget)
1011{ 1011{
1012 struct mv643xx_private *mp = netdev_priv(dev); 1012 struct mv643xx_private *mp = container_of(napi, struct mv643xx_private, napi);
1013 int done = 1, orig_budget, work_done; 1013 struct net_device *dev = mp->dev;
1014 unsigned int port_num = mp->port_num; 1014 unsigned int port_num = mp->port_num;
1015 int work_done;
1015 1016
1016#ifdef MV643XX_TX_FAST_REFILL 1017#ifdef MV643XX_TX_FAST_REFILL
1017 if (++mp->tx_clean_threshold > 5) { 1018 if (++mp->tx_clean_threshold > 5) {
@@ -1020,27 +1021,20 @@ static int mv643xx_poll(struct net_device *dev, int *budget)
1020 } 1021 }
1021#endif 1022#endif
1022 1023
1024 work_done = 0;
1023 if ((mv_read(MV643XX_ETH_RX_CURRENT_QUEUE_DESC_PTR_0(port_num))) 1025 if ((mv_read(MV643XX_ETH_RX_CURRENT_QUEUE_DESC_PTR_0(port_num)))
1024 != (u32) mp->rx_used_desc_q) { 1026 != (u32) mp->rx_used_desc_q)
1025 orig_budget = *budget; 1027 work_done = mv643xx_eth_receive_queue(dev, budget);
1026 if (orig_budget > dev->quota)
1027 orig_budget = dev->quota;
1028 work_done = mv643xx_eth_receive_queue(dev, orig_budget);
1029 *budget -= work_done;
1030 dev->quota -= work_done;
1031 if (work_done >= orig_budget)
1032 done = 0;
1033 }
1034 1028
1035 if (done) { 1029 if (work_done < budget) {
1036 netif_rx_complete(dev); 1030 netif_rx_complete(dev, napi);
1037 mv_write(MV643XX_ETH_INTERRUPT_CAUSE_REG(port_num), 0); 1031 mv_write(MV643XX_ETH_INTERRUPT_CAUSE_REG(port_num), 0);
1038 mv_write(MV643XX_ETH_INTERRUPT_CAUSE_EXTEND_REG(port_num), 0); 1032 mv_write(MV643XX_ETH_INTERRUPT_CAUSE_EXTEND_REG(port_num), 0);
1039 mv_write(MV643XX_ETH_INTERRUPT_MASK_REG(port_num), 1033 mv_write(MV643XX_ETH_INTERRUPT_MASK_REG(port_num),
1040 ETH_INT_UNMASK_ALL); 1034 ETH_INT_UNMASK_ALL);
1041 } 1035 }
1042 1036
1043 return done ? 0 : 1; 1037 return work_done;
1044} 1038}
1045#endif 1039#endif
1046 1040
@@ -1333,6 +1327,10 @@ static int mv643xx_eth_probe(struct platform_device *pdev)
1333 platform_set_drvdata(pdev, dev); 1327 platform_set_drvdata(pdev, dev);
1334 1328
1335 mp = netdev_priv(dev); 1329 mp = netdev_priv(dev);
1330 mp->dev = dev;
1331#ifdef MV643XX_NAPI
1332 netif_napi_add(dev, &mp->napi, mv643xx_poll, 64);
1333#endif
1336 1334
1337 res = platform_get_resource(pdev, IORESOURCE_IRQ, 0); 1335 res = platform_get_resource(pdev, IORESOURCE_IRQ, 0);
1338 BUG_ON(!res); 1336 BUG_ON(!res);
@@ -1347,10 +1345,6 @@ static int mv643xx_eth_probe(struct platform_device *pdev)
1347 1345
1348 /* No need to Tx Timeout */ 1346 /* No need to Tx Timeout */
1349 dev->tx_timeout = mv643xx_eth_tx_timeout; 1347 dev->tx_timeout = mv643xx_eth_tx_timeout;
1350#ifdef MV643XX_NAPI
1351 dev->poll = mv643xx_poll;
1352 dev->weight = 64;
1353#endif
1354 1348
1355#ifdef CONFIG_NET_POLL_CONTROLLER 1349#ifdef CONFIG_NET_POLL_CONTROLLER
1356 dev->poll_controller = mv643xx_netpoll; 1350 dev->poll_controller = mv643xx_netpoll;
diff --git a/drivers/net/mv643xx_eth.h b/drivers/net/mv643xx_eth.h
index 565b96696aca..be669eb23788 100644
--- a/drivers/net/mv643xx_eth.h
+++ b/drivers/net/mv643xx_eth.h
@@ -320,6 +320,8 @@ struct mv643xx_private {
320 320
321 struct work_struct tx_timeout_task; 321 struct work_struct tx_timeout_task;
322 322
323 struct net_device *dev;
324 struct napi_struct napi;
323 struct net_device_stats stats; 325 struct net_device_stats stats;
324 struct mv643xx_mib_counters mib_counters; 326 struct mv643xx_mib_counters mib_counters;
325 spinlock_t lock; 327 spinlock_t lock;
diff --git a/drivers/net/myri10ge/myri10ge.c b/drivers/net/myri10ge/myri10ge.c
index 556962f9612d..a30146ea51f0 100644
--- a/drivers/net/myri10ge/myri10ge.c
+++ b/drivers/net/myri10ge/myri10ge.c
@@ -163,6 +163,7 @@ struct myri10ge_priv {
163 int small_bytes; 163 int small_bytes;
164 int big_bytes; 164 int big_bytes;
165 struct net_device *dev; 165 struct net_device *dev;
166 struct napi_struct napi;
166 struct net_device_stats stats; 167 struct net_device_stats stats;
167 u8 __iomem *sram; 168 u8 __iomem *sram;
168 int sram_size; 169 int sram_size;
@@ -1100,7 +1101,7 @@ static inline void myri10ge_tx_done(struct myri10ge_priv *mgp, int mcp_index)
1100 } 1101 }
1101} 1102}
1102 1103
1103static inline void myri10ge_clean_rx_done(struct myri10ge_priv *mgp, int *limit) 1104static inline int myri10ge_clean_rx_done(struct myri10ge_priv *mgp, int budget)
1104{ 1105{
1105 struct myri10ge_rx_done *rx_done = &mgp->rx_done; 1106 struct myri10ge_rx_done *rx_done = &mgp->rx_done;
1106 unsigned long rx_bytes = 0; 1107 unsigned long rx_bytes = 0;
@@ -1109,10 +1110,11 @@ static inline void myri10ge_clean_rx_done(struct myri10ge_priv *mgp, int *limit)
1109 1110
1110 int idx = rx_done->idx; 1111 int idx = rx_done->idx;
1111 int cnt = rx_done->cnt; 1112 int cnt = rx_done->cnt;
1113 int work_done = 0;
1112 u16 length; 1114 u16 length;
1113 __wsum checksum; 1115 __wsum checksum;
1114 1116
1115 while (rx_done->entry[idx].length != 0 && *limit != 0) { 1117 while (rx_done->entry[idx].length != 0 && work_done++ < budget) {
1116 length = ntohs(rx_done->entry[idx].length); 1118 length = ntohs(rx_done->entry[idx].length);
1117 rx_done->entry[idx].length = 0; 1119 rx_done->entry[idx].length = 0;
1118 checksum = csum_unfold(rx_done->entry[idx].checksum); 1120 checksum = csum_unfold(rx_done->entry[idx].checksum);
@@ -1128,10 +1130,6 @@ static inline void myri10ge_clean_rx_done(struct myri10ge_priv *mgp, int *limit)
1128 rx_bytes += rx_ok * (unsigned long)length; 1130 rx_bytes += rx_ok * (unsigned long)length;
1129 cnt++; 1131 cnt++;
1130 idx = cnt & (myri10ge_max_intr_slots - 1); 1132 idx = cnt & (myri10ge_max_intr_slots - 1);
1131
1132 /* limit potential for livelock by only handling a
1133 * limited number of frames. */
1134 (*limit)--;
1135 } 1133 }
1136 rx_done->idx = idx; 1134 rx_done->idx = idx;
1137 rx_done->cnt = cnt; 1135 rx_done->cnt = cnt;
@@ -1145,6 +1143,7 @@ static inline void myri10ge_clean_rx_done(struct myri10ge_priv *mgp, int *limit)
1145 if (mgp->rx_big.fill_cnt - mgp->rx_big.cnt < myri10ge_fill_thresh) 1143 if (mgp->rx_big.fill_cnt - mgp->rx_big.cnt < myri10ge_fill_thresh)
1146 myri10ge_alloc_rx_pages(mgp, &mgp->rx_big, mgp->big_bytes, 0); 1144 myri10ge_alloc_rx_pages(mgp, &mgp->rx_big, mgp->big_bytes, 0);
1147 1145
1146 return work_done;
1148} 1147}
1149 1148
1150static inline void myri10ge_check_statblock(struct myri10ge_priv *mgp) 1149static inline void myri10ge_check_statblock(struct myri10ge_priv *mgp)
@@ -1189,26 +1188,21 @@ static inline void myri10ge_check_statblock(struct myri10ge_priv *mgp)
1189 } 1188 }
1190} 1189}
1191 1190
1192static int myri10ge_poll(struct net_device *netdev, int *budget) 1191static int myri10ge_poll(struct napi_struct *napi, int budget)
1193{ 1192{
1194 struct myri10ge_priv *mgp = netdev_priv(netdev); 1193 struct myri10ge_priv *mgp = container_of(napi, struct myri10ge_priv, napi);
1194 struct net_device *netdev = mgp->dev;
1195 struct myri10ge_rx_done *rx_done = &mgp->rx_done; 1195 struct myri10ge_rx_done *rx_done = &mgp->rx_done;
1196 int limit, orig_limit, work_done; 1196 int work_done;
1197 1197
1198 /* process as many rx events as NAPI will allow */ 1198 /* process as many rx events as NAPI will allow */
1199 limit = min(*budget, netdev->quota); 1199 work_done = myri10ge_clean_rx_done(mgp, budget);
1200 orig_limit = limit;
1201 myri10ge_clean_rx_done(mgp, &limit);
1202 work_done = orig_limit - limit;
1203 *budget -= work_done;
1204 netdev->quota -= work_done;
1205 1200
1206 if (rx_done->entry[rx_done->idx].length == 0 || !netif_running(netdev)) { 1201 if (rx_done->entry[rx_done->idx].length == 0 || !netif_running(netdev)) {
1207 netif_rx_complete(netdev); 1202 netif_rx_complete(netdev, napi);
1208 put_be32(htonl(3), mgp->irq_claim); 1203 put_be32(htonl(3), mgp->irq_claim);
1209 return 0;
1210 } 1204 }
1211 return 1; 1205 return work_done;
1212} 1206}
1213 1207
1214static irqreturn_t myri10ge_intr(int irq, void *arg) 1208static irqreturn_t myri10ge_intr(int irq, void *arg)
@@ -1226,7 +1220,7 @@ static irqreturn_t myri10ge_intr(int irq, void *arg)
1226 /* low bit indicates receives are present, so schedule 1220 /* low bit indicates receives are present, so schedule
1227 * napi poll handler */ 1221 * napi poll handler */
1228 if (stats->valid & 1) 1222 if (stats->valid & 1)
1229 netif_rx_schedule(mgp->dev); 1223 netif_rx_schedule(mgp->dev, &mgp->napi);
1230 1224
1231 if (!mgp->msi_enabled) { 1225 if (!mgp->msi_enabled) {
1232 put_be32(0, mgp->irq_deassert); 1226 put_be32(0, mgp->irq_deassert);
@@ -1853,7 +1847,7 @@ static int myri10ge_open(struct net_device *dev)
1853 mgp->link_state = htonl(~0U); 1847 mgp->link_state = htonl(~0U);
1854 mgp->rdma_tags_available = 15; 1848 mgp->rdma_tags_available = 15;
1855 1849
1856 netif_poll_enable(mgp->dev); /* must happen prior to any irq */ 1850 napi_enable(&mgp->napi); /* must happen prior to any irq */
1857 1851
1858 status = myri10ge_send_cmd(mgp, MXGEFW_CMD_ETHERNET_UP, &cmd, 0); 1852 status = myri10ge_send_cmd(mgp, MXGEFW_CMD_ETHERNET_UP, &cmd, 0);
1859 if (status) { 1853 if (status) {
@@ -1897,7 +1891,7 @@ static int myri10ge_close(struct net_device *dev)
1897 1891
1898 del_timer_sync(&mgp->watchdog_timer); 1892 del_timer_sync(&mgp->watchdog_timer);
1899 mgp->running = MYRI10GE_ETH_STOPPING; 1893 mgp->running = MYRI10GE_ETH_STOPPING;
1900 netif_poll_disable(mgp->dev); 1894 napi_disable(&mgp->napi);
1901 netif_carrier_off(dev); 1895 netif_carrier_off(dev);
1902 netif_stop_queue(dev); 1896 netif_stop_queue(dev);
1903 old_down_cnt = mgp->down_cnt; 1897 old_down_cnt = mgp->down_cnt;
@@ -2857,6 +2851,8 @@ static int myri10ge_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2857 mgp = netdev_priv(netdev); 2851 mgp = netdev_priv(netdev);
2858 memset(mgp, 0, sizeof(*mgp)); 2852 memset(mgp, 0, sizeof(*mgp));
2859 mgp->dev = netdev; 2853 mgp->dev = netdev;
2854 netif_napi_add(netdev, &mgp->napi,
2855 myri10ge_poll, myri10ge_napi_weight);
2860 mgp->pdev = pdev; 2856 mgp->pdev = pdev;
2861 mgp->csum_flag = MXGEFW_FLAGS_CKSUM; 2857 mgp->csum_flag = MXGEFW_FLAGS_CKSUM;
2862 mgp->pause = myri10ge_flow_control; 2858 mgp->pause = myri10ge_flow_control;
@@ -2981,8 +2977,6 @@ static int myri10ge_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
2981 netdev->features = NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_TSO; 2977 netdev->features = NETIF_F_SG | NETIF_F_HW_CSUM | NETIF_F_TSO;
2982 if (dac_enabled) 2978 if (dac_enabled)
2983 netdev->features |= NETIF_F_HIGHDMA; 2979 netdev->features |= NETIF_F_HIGHDMA;
2984 netdev->poll = myri10ge_poll;
2985 netdev->weight = myri10ge_napi_weight;
2986 2980
2987 /* make sure we can get an irq, and that MSI can be 2981 /* make sure we can get an irq, and that MSI can be
2988 * setup (if available). Also ensure netdev->irq 2982 * setup (if available). Also ensure netdev->irq
diff --git a/drivers/net/natsemi.c b/drivers/net/natsemi.c
index b47a12d684f9..43cfa4b3e294 100644
--- a/drivers/net/natsemi.c
+++ b/drivers/net/natsemi.c
@@ -560,6 +560,8 @@ struct netdev_private {
560 /* address of a sent-in-place packet/buffer, for later free() */ 560 /* address of a sent-in-place packet/buffer, for later free() */
561 struct sk_buff *tx_skbuff[TX_RING_SIZE]; 561 struct sk_buff *tx_skbuff[TX_RING_SIZE];
562 dma_addr_t tx_dma[TX_RING_SIZE]; 562 dma_addr_t tx_dma[TX_RING_SIZE];
563 struct net_device *dev;
564 struct napi_struct napi;
563 struct net_device_stats stats; 565 struct net_device_stats stats;
564 /* Media monitoring timer */ 566 /* Media monitoring timer */
565 struct timer_list timer; 567 struct timer_list timer;
@@ -636,7 +638,7 @@ static void init_registers(struct net_device *dev);
636static int start_tx(struct sk_buff *skb, struct net_device *dev); 638static int start_tx(struct sk_buff *skb, struct net_device *dev);
637static irqreturn_t intr_handler(int irq, void *dev_instance); 639static irqreturn_t intr_handler(int irq, void *dev_instance);
638static void netdev_error(struct net_device *dev, int intr_status); 640static void netdev_error(struct net_device *dev, int intr_status);
639static int natsemi_poll(struct net_device *dev, int *budget); 641static int natsemi_poll(struct napi_struct *napi, int budget);
640static void netdev_rx(struct net_device *dev, int *work_done, int work_to_do); 642static void netdev_rx(struct net_device *dev, int *work_done, int work_to_do);
641static void netdev_tx_done(struct net_device *dev); 643static void netdev_tx_done(struct net_device *dev);
642static int natsemi_change_mtu(struct net_device *dev, int new_mtu); 644static int natsemi_change_mtu(struct net_device *dev, int new_mtu);
@@ -861,6 +863,7 @@ static int __devinit natsemi_probe1 (struct pci_dev *pdev,
861 dev->irq = irq; 863 dev->irq = irq;
862 864
863 np = netdev_priv(dev); 865 np = netdev_priv(dev);
866 netif_napi_add(dev, &np->napi, natsemi_poll, 64);
864 867
865 np->pci_dev = pdev; 868 np->pci_dev = pdev;
866 pci_set_drvdata(pdev, dev); 869 pci_set_drvdata(pdev, dev);
@@ -931,8 +934,6 @@ static int __devinit natsemi_probe1 (struct pci_dev *pdev,
931 dev->do_ioctl = &netdev_ioctl; 934 dev->do_ioctl = &netdev_ioctl;
932 dev->tx_timeout = &tx_timeout; 935 dev->tx_timeout = &tx_timeout;
933 dev->watchdog_timeo = TX_TIMEOUT; 936 dev->watchdog_timeo = TX_TIMEOUT;
934 dev->poll = natsemi_poll;
935 dev->weight = 64;
936 937
937#ifdef CONFIG_NET_POLL_CONTROLLER 938#ifdef CONFIG_NET_POLL_CONTROLLER
938 dev->poll_controller = &natsemi_poll_controller; 939 dev->poll_controller = &natsemi_poll_controller;
@@ -1554,6 +1555,8 @@ static int netdev_open(struct net_device *dev)
1554 free_irq(dev->irq, dev); 1555 free_irq(dev->irq, dev);
1555 return i; 1556 return i;
1556 } 1557 }
1558 napi_enable(&np->napi);
1559
1557 init_ring(dev); 1560 init_ring(dev);
1558 spin_lock_irq(&np->lock); 1561 spin_lock_irq(&np->lock);
1559 init_registers(dev); 1562 init_registers(dev);
@@ -2200,10 +2203,10 @@ static irqreturn_t intr_handler(int irq, void *dev_instance)
2200 2203
2201 prefetch(&np->rx_skbuff[np->cur_rx % RX_RING_SIZE]); 2204 prefetch(&np->rx_skbuff[np->cur_rx % RX_RING_SIZE]);
2202 2205
2203 if (netif_rx_schedule_prep(dev)) { 2206 if (netif_rx_schedule_prep(dev, &np->napi)) {
2204 /* Disable interrupts and register for poll */ 2207 /* Disable interrupts and register for poll */
2205 natsemi_irq_disable(dev); 2208 natsemi_irq_disable(dev);
2206 __netif_rx_schedule(dev); 2209 __netif_rx_schedule(dev, &np->napi);
2207 } else 2210 } else
2208 printk(KERN_WARNING 2211 printk(KERN_WARNING
2209 "%s: Ignoring interrupt, status %#08x, mask %#08x.\n", 2212 "%s: Ignoring interrupt, status %#08x, mask %#08x.\n",
@@ -2216,12 +2219,11 @@ static irqreturn_t intr_handler(int irq, void *dev_instance)
2216/* This is the NAPI poll routine. As well as the standard RX handling 2219/* This is the NAPI poll routine. As well as the standard RX handling
2217 * it also handles all other interrupts that the chip might raise. 2220 * it also handles all other interrupts that the chip might raise.
2218 */ 2221 */
2219static int natsemi_poll(struct net_device *dev, int *budget) 2222static int natsemi_poll(struct napi_struct *napi, int budget)
2220{ 2223{
2221 struct netdev_private *np = netdev_priv(dev); 2224 struct netdev_private *np = container_of(napi, struct netdev_private, napi);
2225 struct net_device *dev = np->dev;
2222 void __iomem * ioaddr = ns_ioaddr(dev); 2226 void __iomem * ioaddr = ns_ioaddr(dev);
2223
2224 int work_to_do = min(*budget, dev->quota);
2225 int work_done = 0; 2227 int work_done = 0;
2226 2228
2227 do { 2229 do {
@@ -2236,7 +2238,7 @@ static int natsemi_poll(struct net_device *dev, int *budget)
2236 if (np->intr_status & 2238 if (np->intr_status &
2237 (IntrRxDone | IntrRxIntr | RxStatusFIFOOver | 2239 (IntrRxDone | IntrRxIntr | RxStatusFIFOOver |
2238 IntrRxErr | IntrRxOverrun)) { 2240 IntrRxErr | IntrRxOverrun)) {
2239 netdev_rx(dev, &work_done, work_to_do); 2241 netdev_rx(dev, &work_done, budget);
2240 } 2242 }
2241 2243
2242 if (np->intr_status & 2244 if (np->intr_status &
@@ -2250,16 +2252,13 @@ static int natsemi_poll(struct net_device *dev, int *budget)
2250 if (np->intr_status & IntrAbnormalSummary) 2252 if (np->intr_status & IntrAbnormalSummary)
2251 netdev_error(dev, np->intr_status); 2253 netdev_error(dev, np->intr_status);
2252 2254
2253 *budget -= work_done; 2255 if (work_done >= budget)
2254 dev->quota -= work_done; 2256 return work_done;
2255
2256 if (work_done >= work_to_do)
2257 return 1;
2258 2257
2259 np->intr_status = readl(ioaddr + IntrStatus); 2258 np->intr_status = readl(ioaddr + IntrStatus);
2260 } while (np->intr_status); 2259 } while (np->intr_status);
2261 2260
2262 netif_rx_complete(dev); 2261 netif_rx_complete(dev, napi);
2263 2262
2264 /* Reenable interrupts providing nothing is trying to shut 2263 /* Reenable interrupts providing nothing is trying to shut
2265 * the chip down. */ 2264 * the chip down. */
@@ -2268,7 +2267,7 @@ static int natsemi_poll(struct net_device *dev, int *budget)
2268 natsemi_irq_enable(dev); 2267 natsemi_irq_enable(dev);
2269 spin_unlock(&np->lock); 2268 spin_unlock(&np->lock);
2270 2269
2271 return 0; 2270 return work_done;
2272} 2271}
2273 2272
2274/* This routine is logically part of the interrupt handler, but separated 2273/* This routine is logically part of the interrupt handler, but separated
@@ -3158,6 +3157,8 @@ static int netdev_close(struct net_device *dev)
3158 dev->name, np->cur_tx, np->dirty_tx, 3157 dev->name, np->cur_tx, np->dirty_tx,
3159 np->cur_rx, np->dirty_rx); 3158 np->cur_rx, np->dirty_rx);
3160 3159
3160 napi_disable(&np->napi);
3161
3161 /* 3162 /*
3162 * FIXME: what if someone tries to close a device 3163 * FIXME: what if someone tries to close a device
3163 * that is suspended? 3164 * that is suspended?
@@ -3253,7 +3254,7 @@ static void __devexit natsemi_remove1 (struct pci_dev *pdev)
3253 * disable_irq() to enforce synchronization. 3254 * disable_irq() to enforce synchronization.
3254 * * natsemi_poll: checks before reenabling interrupts. suspend 3255 * * natsemi_poll: checks before reenabling interrupts. suspend
3255 * sets hands_off, disables interrupts and then waits with 3256 * sets hands_off, disables interrupts and then waits with
3256 * netif_poll_disable(). 3257 * napi_disable().
3257 * 3258 *
3258 * Interrupts must be disabled, otherwise hands_off can cause irq storms. 3259 * Interrupts must be disabled, otherwise hands_off can cause irq storms.
3259 */ 3260 */
@@ -3279,7 +3280,7 @@ static int natsemi_suspend (struct pci_dev *pdev, pm_message_t state)
3279 spin_unlock_irq(&np->lock); 3280 spin_unlock_irq(&np->lock);
3280 enable_irq(dev->irq); 3281 enable_irq(dev->irq);
3281 3282
3282 netif_poll_disable(dev); 3283 napi_disable(&np->napi);
3283 3284
3284 /* Update the error counts. */ 3285 /* Update the error counts. */
3285 __get_stats(dev); 3286 __get_stats(dev);
@@ -3320,6 +3321,8 @@ static int natsemi_resume (struct pci_dev *pdev)
3320 pci_enable_device(pdev); 3321 pci_enable_device(pdev);
3321 /* pci_power_on(pdev); */ 3322 /* pci_power_on(pdev); */
3322 3323
3324 napi_enable(&np->napi);
3325
3323 natsemi_reset(dev); 3326 natsemi_reset(dev);
3324 init_ring(dev); 3327 init_ring(dev);
3325 disable_irq(dev->irq); 3328 disable_irq(dev->irq);
@@ -3333,7 +3336,6 @@ static int natsemi_resume (struct pci_dev *pdev)
3333 mod_timer(&np->timer, jiffies + 1*HZ); 3336 mod_timer(&np->timer, jiffies + 1*HZ);
3334 } 3337 }
3335 netif_device_attach(dev); 3338 netif_device_attach(dev);
3336 netif_poll_enable(dev);
3337out: 3339out:
3338 rtnl_unlock(); 3340 rtnl_unlock();
3339 return 0; 3341 return 0;
diff --git a/drivers/net/netxen/netxen_nic.h b/drivers/net/netxen/netxen_nic.h
index d4c92cc879d4..aaa34939485b 100644
--- a/drivers/net/netxen/netxen_nic.h
+++ b/drivers/net/netxen/netxen_nic.h
@@ -880,6 +880,7 @@ struct netxen_adapter {
880 struct netxen_adapter *master; 880 struct netxen_adapter *master;
881 struct net_device *netdev; 881 struct net_device *netdev;
882 struct pci_dev *pdev; 882 struct pci_dev *pdev;
883 struct napi_struct napi;
883 struct net_device_stats net_stats; 884 struct net_device_stats net_stats;
884 unsigned char mac_addr[ETH_ALEN]; 885 unsigned char mac_addr[ETH_ALEN];
885 int mtu; 886 int mtu;
diff --git a/drivers/net/netxen/netxen_nic_main.c b/drivers/net/netxen/netxen_nic_main.c
index 3122d0101638..a10bbefbdadd 100644
--- a/drivers/net/netxen/netxen_nic_main.c
+++ b/drivers/net/netxen/netxen_nic_main.c
@@ -68,7 +68,7 @@ static void netxen_tx_timeout(struct net_device *netdev);
68static void netxen_tx_timeout_task(struct work_struct *work); 68static void netxen_tx_timeout_task(struct work_struct *work);
69static void netxen_watchdog(unsigned long); 69static void netxen_watchdog(unsigned long);
70static int netxen_handle_int(struct netxen_adapter *, struct net_device *); 70static int netxen_handle_int(struct netxen_adapter *, struct net_device *);
71static int netxen_nic_poll(struct net_device *dev, int *budget); 71static int netxen_nic_poll(struct napi_struct *napi, int budget);
72#ifdef CONFIG_NET_POLL_CONTROLLER 72#ifdef CONFIG_NET_POLL_CONTROLLER
73static void netxen_nic_poll_controller(struct net_device *netdev); 73static void netxen_nic_poll_controller(struct net_device *netdev);
74#endif 74#endif
@@ -402,6 +402,9 @@ netxen_nic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
402 adapter->netdev = netdev; 402 adapter->netdev = netdev;
403 adapter->pdev = pdev; 403 adapter->pdev = pdev;
404 404
405 netif_napi_add(netdev, &adapter->napi,
406 netxen_nic_poll, NETXEN_NETDEV_WEIGHT);
407
405 /* this will be read from FW later */ 408 /* this will be read from FW later */
406 adapter->intr_scheme = -1; 409 adapter->intr_scheme = -1;
407 410
@@ -422,8 +425,6 @@ netxen_nic_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
422 netxen_nic_change_mtu(netdev, netdev->mtu); 425 netxen_nic_change_mtu(netdev, netdev->mtu);
423 426
424 SET_ETHTOOL_OPS(netdev, &netxen_nic_ethtool_ops); 427 SET_ETHTOOL_OPS(netdev, &netxen_nic_ethtool_ops);
425 netdev->poll = netxen_nic_poll;
426 netdev->weight = NETXEN_NETDEV_WEIGHT;
427#ifdef CONFIG_NET_POLL_CONTROLLER 428#ifdef CONFIG_NET_POLL_CONTROLLER
428 netdev->poll_controller = netxen_nic_poll_controller; 429 netdev->poll_controller = netxen_nic_poll_controller;
429#endif 430#endif
@@ -885,6 +886,8 @@ static int netxen_nic_open(struct net_device *netdev)
885 if (!adapter->driver_mismatch) 886 if (!adapter->driver_mismatch)
886 mod_timer(&adapter->watchdog_timer, jiffies); 887 mod_timer(&adapter->watchdog_timer, jiffies);
887 888
889 napi_enable(&adapter->napi);
890
888 netxen_nic_enable_int(adapter); 891 netxen_nic_enable_int(adapter);
889 892
890 /* Done here again so that even if phantom sw overwrote it, 893 /* Done here again so that even if phantom sw overwrote it,
@@ -894,6 +897,7 @@ static int netxen_nic_open(struct net_device *netdev)
894 del_timer_sync(&adapter->watchdog_timer); 897 del_timer_sync(&adapter->watchdog_timer);
895 printk(KERN_ERR "%s: Failed to initialize port %d\n", 898 printk(KERN_ERR "%s: Failed to initialize port %d\n",
896 netxen_nic_driver_name, adapter->portnum); 899 netxen_nic_driver_name, adapter->portnum);
900 napi_disable(&adapter->napi);
897 return -EIO; 901 return -EIO;
898 } 902 }
899 if (adapter->macaddr_set) 903 if (adapter->macaddr_set)
@@ -923,6 +927,7 @@ static int netxen_nic_close(struct net_device *netdev)
923 927
924 netif_carrier_off(netdev); 928 netif_carrier_off(netdev);
925 netif_stop_queue(netdev); 929 netif_stop_queue(netdev);
930 napi_disable(&adapter->napi);
926 931
927 netxen_nic_disable_int(adapter); 932 netxen_nic_disable_int(adapter);
928 933
@@ -1243,11 +1248,11 @@ netxen_handle_int(struct netxen_adapter *adapter, struct net_device *netdev)
1243 netxen_nic_disable_int(adapter); 1248 netxen_nic_disable_int(adapter);
1244 1249
1245 if (netxen_nic_rx_has_work(adapter) || netxen_nic_tx_has_work(adapter)) { 1250 if (netxen_nic_rx_has_work(adapter) || netxen_nic_tx_has_work(adapter)) {
1246 if (netif_rx_schedule_prep(netdev)) { 1251 if (netif_rx_schedule_prep(netdev, &adapter->napi)) {
1247 /* 1252 /*
1248 * Interrupts are already disabled. 1253 * Interrupts are already disabled.
1249 */ 1254 */
1250 __netif_rx_schedule(netdev); 1255 __netif_rx_schedule(netdev, &adapter->napi);
1251 } else { 1256 } else {
1252 static unsigned int intcount = 0; 1257 static unsigned int intcount = 0;
1253 if ((++intcount & 0xfff) == 0xfff) 1258 if ((++intcount & 0xfff) == 0xfff)
@@ -1305,14 +1310,13 @@ irqreturn_t netxen_intr(int irq, void *data)
1305 return IRQ_HANDLED; 1310 return IRQ_HANDLED;
1306} 1311}
1307 1312
1308static int netxen_nic_poll(struct net_device *netdev, int *budget) 1313static int netxen_nic_poll(struct napi_struct *napi, int budget)
1309{ 1314{
1310 struct netxen_adapter *adapter = netdev_priv(netdev); 1315 struct netxen_adapter *adapter = container_of(napi, struct netxen_adapter, napi);
1311 int work_to_do = min(*budget, netdev->quota); 1316 struct net_device *netdev = adapter->netdev;
1312 int done = 1; 1317 int done = 1;
1313 int ctx; 1318 int ctx;
1314 int this_work_done; 1319 int work_done;
1315 int work_done = 0;
1316 1320
1317 DPRINTK(INFO, "polling for %d descriptors\n", *budget); 1321 DPRINTK(INFO, "polling for %d descriptors\n", *budget);
1318 1322
@@ -1330,16 +1334,11 @@ static int netxen_nic_poll(struct net_device *netdev, int *budget)
1330 * packets are on one context, it gets only half of the quota, 1334 * packets are on one context, it gets only half of the quota,
1331 * and ends up not processing it. 1335 * and ends up not processing it.
1332 */ 1336 */
1333 this_work_done = netxen_process_rcv_ring(adapter, ctx, 1337 work_done += netxen_process_rcv_ring(adapter, ctx,
1334 work_to_do / 1338 budget / MAX_RCV_CTX);
1335 MAX_RCV_CTX);
1336 work_done += this_work_done;
1337 } 1339 }
1338 1340
1339 netdev->quota -= work_done; 1341 if (work_done >= budget && netxen_nic_rx_has_work(adapter) != 0)
1340 *budget -= work_done;
1341
1342 if (work_done >= work_to_do && netxen_nic_rx_has_work(adapter) != 0)
1343 done = 0; 1342 done = 0;
1344 1343
1345 if (netxen_process_cmd_ring((unsigned long)adapter) == 0) 1344 if (netxen_process_cmd_ring((unsigned long)adapter) == 0)
@@ -1348,11 +1347,11 @@ static int netxen_nic_poll(struct net_device *netdev, int *budget)
1348 DPRINTK(INFO, "new work_done: %d work_to_do: %d\n", 1347 DPRINTK(INFO, "new work_done: %d work_to_do: %d\n",
1349 work_done, work_to_do); 1348 work_done, work_to_do);
1350 if (done) { 1349 if (done) {
1351 netif_rx_complete(netdev); 1350 netif_rx_complete(netdev, napi);
1352 netxen_nic_enable_int(adapter); 1351 netxen_nic_enable_int(adapter);
1353 } 1352 }
1354 1353
1355 return !done; 1354 return work_done;
1356} 1355}
1357 1356
1358#ifdef CONFIG_NET_POLL_CONTROLLER 1357#ifdef CONFIG_NET_POLL_CONTROLLER
diff --git a/drivers/net/pasemi_mac.c b/drivers/net/pasemi_mac.c
index 0b3066a6fe40..e63cc335a4ba 100644
--- a/drivers/net/pasemi_mac.c
+++ b/drivers/net/pasemi_mac.c
@@ -584,7 +584,7 @@ static irqreturn_t pasemi_mac_rx_intr(int irq, void *data)
584 if (*mac->rx_status & PAS_STATUS_TIMER) 584 if (*mac->rx_status & PAS_STATUS_TIMER)
585 reg |= PAS_IOB_DMA_RXCH_RESET_TINTC; 585 reg |= PAS_IOB_DMA_RXCH_RESET_TINTC;
586 586
587 netif_rx_schedule(dev); 587 netif_rx_schedule(dev, &mac->napi);
588 588
589 pci_write_config_dword(mac->iob_pdev, 589 pci_write_config_dword(mac->iob_pdev,
590 PAS_IOB_DMA_RXCH_RESET(mac->dma_rxch), reg); 590 PAS_IOB_DMA_RXCH_RESET(mac->dma_rxch), reg);
@@ -808,7 +808,7 @@ static int pasemi_mac_open(struct net_device *dev)
808 dev_warn(&mac->pdev->dev, "phy init failed: %d\n", ret); 808 dev_warn(&mac->pdev->dev, "phy init failed: %d\n", ret);
809 809
810 netif_start_queue(dev); 810 netif_start_queue(dev);
811 netif_poll_enable(dev); 811 napi_enable(&mac->napi);
812 812
813 /* Interrupts are a bit different for our DMA controller: While 813 /* Interrupts are a bit different for our DMA controller: While
814 * it's got one a regular PCI device header, the interrupt there 814 * it's got one a regular PCI device header, the interrupt there
@@ -845,7 +845,7 @@ static int pasemi_mac_open(struct net_device *dev)
845out_rx_int: 845out_rx_int:
846 free_irq(mac->tx_irq, dev); 846 free_irq(mac->tx_irq, dev);
847out_tx_int: 847out_tx_int:
848 netif_poll_disable(dev); 848 napi_disable(&mac->napi);
849 netif_stop_queue(dev); 849 netif_stop_queue(dev);
850 pasemi_mac_free_tx_resources(dev); 850 pasemi_mac_free_tx_resources(dev);
851out_tx_resources: 851out_tx_resources:
@@ -869,6 +869,7 @@ static int pasemi_mac_close(struct net_device *dev)
869 } 869 }
870 870
871 netif_stop_queue(dev); 871 netif_stop_queue(dev);
872 napi_disable(&mac->napi);
872 873
873 /* Clean out any pending buffers */ 874 /* Clean out any pending buffers */
874 pasemi_mac_clean_tx(mac); 875 pasemi_mac_clean_tx(mac);
@@ -1047,26 +1048,20 @@ static void pasemi_mac_set_rx_mode(struct net_device *dev)
1047} 1048}
1048 1049
1049 1050
1050static int pasemi_mac_poll(struct net_device *dev, int *budget) 1051static int pasemi_mac_poll(struct napi_struct *napi, int budget)
1051{ 1052{
1052 int pkts, limit = min(*budget, dev->quota); 1053 struct pasemi_mac *mac = container_of(napi, struct pasemi_mac, napi);
1053 struct pasemi_mac *mac = netdev_priv(dev); 1054 struct net_device *dev = mac->netdev;
1054 1055 int pkts;
1055 pkts = pasemi_mac_clean_rx(mac, limit);
1056 1056
1057 dev->quota -= pkts; 1057 pkts = pasemi_mac_clean_rx(mac, budget);
1058 *budget -= pkts; 1058 if (pkts < budget) {
1059
1060 if (pkts < limit) {
1061 /* all done, no more packets present */ 1059 /* all done, no more packets present */
1062 netif_rx_complete(dev); 1060 netif_rx_complete(dev, napi);
1063 1061
1064 pasemi_mac_restart_rx_intr(mac); 1062 pasemi_mac_restart_rx_intr(mac);
1065 return 0;
1066 } else {
1067 /* used up our quantum, so reschedule */
1068 return 1;
1069 } 1063 }
1064 return pkts;
1070} 1065}
1071 1066
1072static int __devinit 1067static int __devinit
@@ -1099,6 +1094,10 @@ pasemi_mac_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
1099 mac->netdev = dev; 1094 mac->netdev = dev;
1100 mac->dma_pdev = pci_get_device(PCI_VENDOR_ID_PASEMI, 0xa007, NULL); 1095 mac->dma_pdev = pci_get_device(PCI_VENDOR_ID_PASEMI, 0xa007, NULL);
1101 1096
1097 netif_napi_add(dev, &mac->napi, pasemi_mac_poll, 64);
1098
1099 dev->features = NETIF_F_HW_CSUM;
1100
1102 if (!mac->dma_pdev) { 1101 if (!mac->dma_pdev) {
1103 dev_err(&pdev->dev, "Can't find DMA Controller\n"); 1102 dev_err(&pdev->dev, "Can't find DMA Controller\n");
1104 err = -ENODEV; 1103 err = -ENODEV;
@@ -1150,9 +1149,6 @@ pasemi_mac_probe(struct pci_dev *pdev, const struct pci_device_id *ent)
1150 dev->hard_start_xmit = pasemi_mac_start_tx; 1149 dev->hard_start_xmit = pasemi_mac_start_tx;
1151 dev->get_stats = pasemi_mac_get_stats; 1150 dev->get_stats = pasemi_mac_get_stats;
1152 dev->set_multicast_list = pasemi_mac_set_rx_mode; 1151 dev->set_multicast_list = pasemi_mac_set_rx_mode;
1153 dev->weight = 64;
1154 dev->poll = pasemi_mac_poll;
1155 dev->features = NETIF_F_HW_CSUM;
1156 1152
1157 /* The dma status structure is located in the I/O bridge, and 1153 /* The dma status structure is located in the I/O bridge, and
1158 * is cache coherent. 1154 * is cache coherent.
diff --git a/drivers/net/pasemi_mac.h b/drivers/net/pasemi_mac.h
index c29ee159c33d..85d3b7856e5f 100644
--- a/drivers/net/pasemi_mac.h
+++ b/drivers/net/pasemi_mac.h
@@ -56,6 +56,7 @@ struct pasemi_mac {
56 struct pci_dev *dma_pdev; 56 struct pci_dev *dma_pdev;
57 struct pci_dev *iob_pdev; 57 struct pci_dev *iob_pdev;
58 struct phy_device *phydev; 58 struct phy_device *phydev;
59 struct napi_struct napi;
59 struct net_device_stats stats; 60 struct net_device_stats stats;
60 61
61 /* Pointer to the cacheable per-channel status registers */ 62 /* Pointer to the cacheable per-channel status registers */
diff --git a/drivers/net/pcnet32.c b/drivers/net/pcnet32.c
index e6a67531de99..a9973490dba9 100644
--- a/drivers/net/pcnet32.c
+++ b/drivers/net/pcnet32.c
@@ -280,6 +280,8 @@ struct pcnet32_private {
280 unsigned int dirty_rx, /* ring entries to be freed. */ 280 unsigned int dirty_rx, /* ring entries to be freed. */
281 dirty_tx; 281 dirty_tx;
282 282
283 struct net_device *dev;
284 struct napi_struct napi;
283 struct net_device_stats stats; 285 struct net_device_stats stats;
284 char tx_full; 286 char tx_full;
285 char phycount; /* number of phys found */ 287 char phycount; /* number of phys found */
@@ -440,15 +442,21 @@ static struct pcnet32_access pcnet32_dwio = {
440 442
441static void pcnet32_netif_stop(struct net_device *dev) 443static void pcnet32_netif_stop(struct net_device *dev)
442{ 444{
445 struct pcnet32_private *lp = netdev_priv(dev);
443 dev->trans_start = jiffies; 446 dev->trans_start = jiffies;
444 netif_poll_disable(dev); 447#ifdef CONFIG_PCNET32_NAPI
448 napi_disable(&lp->napi);
449#endif
445 netif_tx_disable(dev); 450 netif_tx_disable(dev);
446} 451}
447 452
448static void pcnet32_netif_start(struct net_device *dev) 453static void pcnet32_netif_start(struct net_device *dev)
449{ 454{
455 struct pcnet32_private *lp = netdev_priv(dev);
450 netif_wake_queue(dev); 456 netif_wake_queue(dev);
451 netif_poll_enable(dev); 457#ifdef CONFIG_PCNET32_NAPI
458 napi_enable(&lp->napi);
459#endif
452} 460}
453 461
454/* 462/*
@@ -816,7 +824,7 @@ static int pcnet32_set_ringparam(struct net_device *dev,
816 if ((1 << i) != lp->rx_ring_size) 824 if ((1 << i) != lp->rx_ring_size)
817 pcnet32_realloc_rx_ring(dev, lp, i); 825 pcnet32_realloc_rx_ring(dev, lp, i);
818 826
819 dev->weight = lp->rx_ring_size / 2; 827 lp->napi.weight = lp->rx_ring_size / 2;
820 828
821 if (netif_running(dev)) { 829 if (netif_running(dev)) {
822 pcnet32_netif_start(dev); 830 pcnet32_netif_start(dev);
@@ -1255,7 +1263,7 @@ static void pcnet32_rx_entry(struct net_device *dev,
1255 return; 1263 return;
1256} 1264}
1257 1265
1258static int pcnet32_rx(struct net_device *dev, int quota) 1266static int pcnet32_rx(struct net_device *dev, int budget)
1259{ 1267{
1260 struct pcnet32_private *lp = netdev_priv(dev); 1268 struct pcnet32_private *lp = netdev_priv(dev);
1261 int entry = lp->cur_rx & lp->rx_mod_mask; 1269 int entry = lp->cur_rx & lp->rx_mod_mask;
@@ -1263,7 +1271,7 @@ static int pcnet32_rx(struct net_device *dev, int quota)
1263 int npackets = 0; 1271 int npackets = 0;
1264 1272
1265 /* If we own the next entry, it's a new packet. Send it up. */ 1273 /* If we own the next entry, it's a new packet. Send it up. */
1266 while (quota > npackets && (short)le16_to_cpu(rxp->status) >= 0) { 1274 while (npackets < budget && (short)le16_to_cpu(rxp->status) >= 0) {
1267 pcnet32_rx_entry(dev, lp, rxp, entry); 1275 pcnet32_rx_entry(dev, lp, rxp, entry);
1268 npackets += 1; 1276 npackets += 1;
1269 /* 1277 /*
@@ -1379,15 +1387,16 @@ static int pcnet32_tx(struct net_device *dev)
1379} 1387}
1380 1388
1381#ifdef CONFIG_PCNET32_NAPI 1389#ifdef CONFIG_PCNET32_NAPI
1382static int pcnet32_poll(struct net_device *dev, int *budget) 1390static int pcnet32_poll(struct napi_struct *napi, int budget)
1383{ 1391{
1384 struct pcnet32_private *lp = netdev_priv(dev); 1392 struct pcnet32_private *lp = container_of(napi, struct pcnet32_private, napi);
1385 int quota = min(dev->quota, *budget); 1393 struct net_device *dev = lp->dev;
1386 unsigned long ioaddr = dev->base_addr; 1394 unsigned long ioaddr = dev->base_addr;
1387 unsigned long flags; 1395 unsigned long flags;
1396 int work_done;
1388 u16 val; 1397 u16 val;
1389 1398
1390 quota = pcnet32_rx(dev, quota); 1399 work_done = pcnet32_rx(dev, budget);
1391 1400
1392 spin_lock_irqsave(&lp->lock, flags); 1401 spin_lock_irqsave(&lp->lock, flags);
1393 if (pcnet32_tx(dev)) { 1402 if (pcnet32_tx(dev)) {
@@ -1399,28 +1408,22 @@ static int pcnet32_poll(struct net_device *dev, int *budget)
1399 } 1408 }
1400 spin_unlock_irqrestore(&lp->lock, flags); 1409 spin_unlock_irqrestore(&lp->lock, flags);
1401 1410
1402 *budget -= quota; 1411 if (work_done < budget) {
1403 dev->quota -= quota; 1412 spin_lock_irqsave(&lp->lock, flags);
1404
1405 if (dev->quota == 0) {
1406 return 1;
1407 }
1408
1409 netif_rx_complete(dev);
1410
1411 spin_lock_irqsave(&lp->lock, flags);
1412 1413
1413 /* clear interrupt masks */ 1414 __netif_rx_complete(dev, napi);
1414 val = lp->a.read_csr(ioaddr, CSR3);
1415 val &= 0x00ff;
1416 lp->a.write_csr(ioaddr, CSR3, val);
1417 1415
1418 /* Set interrupt enable. */ 1416 /* clear interrupt masks */
1419 lp->a.write_csr(ioaddr, CSR0, CSR0_INTEN); 1417 val = lp->a.read_csr(ioaddr, CSR3);
1420 mmiowb(); 1418 val &= 0x00ff;
1421 spin_unlock_irqrestore(&lp->lock, flags); 1419 lp->a.write_csr(ioaddr, CSR3, val);
1422 1420
1423 return 0; 1421 /* Set interrupt enable. */
1422 lp->a.write_csr(ioaddr, CSR0, CSR0_INTEN);
1423 mmiowb();
1424 spin_unlock_irqrestore(&lp->lock, flags);
1425 }
1426 return work_done;
1424} 1427}
1425#endif 1428#endif
1426 1429
@@ -1815,6 +1818,8 @@ pcnet32_probe1(unsigned long ioaddr, int shared, struct pci_dev *pdev)
1815 } 1818 }
1816 lp->pci_dev = pdev; 1819 lp->pci_dev = pdev;
1817 1820
1821 lp->dev = dev;
1822
1818 spin_lock_init(&lp->lock); 1823 spin_lock_init(&lp->lock);
1819 1824
1820 SET_MODULE_OWNER(dev); 1825 SET_MODULE_OWNER(dev);
@@ -1843,6 +1848,10 @@ pcnet32_probe1(unsigned long ioaddr, int shared, struct pci_dev *pdev)
1843 lp->mii_if.mdio_read = mdio_read; 1848 lp->mii_if.mdio_read = mdio_read;
1844 lp->mii_if.mdio_write = mdio_write; 1849 lp->mii_if.mdio_write = mdio_write;
1845 1850
1851#ifdef CONFIG_PCNET32_NAPI
1852 netif_napi_add(dev, &lp->napi, pcnet32_poll, lp->rx_ring_size / 2);
1853#endif
1854
1846 if (fdx && !(lp->options & PCNET32_PORT_ASEL) && 1855 if (fdx && !(lp->options & PCNET32_PORT_ASEL) &&
1847 ((cards_found >= MAX_UNITS) || full_duplex[cards_found])) 1856 ((cards_found >= MAX_UNITS) || full_duplex[cards_found]))
1848 lp->options |= PCNET32_PORT_FD; 1857 lp->options |= PCNET32_PORT_FD;
@@ -1953,10 +1962,6 @@ pcnet32_probe1(unsigned long ioaddr, int shared, struct pci_dev *pdev)
1953 dev->ethtool_ops = &pcnet32_ethtool_ops; 1962 dev->ethtool_ops = &pcnet32_ethtool_ops;
1954 dev->tx_timeout = pcnet32_tx_timeout; 1963 dev->tx_timeout = pcnet32_tx_timeout;
1955 dev->watchdog_timeo = (5 * HZ); 1964 dev->watchdog_timeo = (5 * HZ);
1956 dev->weight = lp->rx_ring_size / 2;
1957#ifdef CONFIG_PCNET32_NAPI
1958 dev->poll = pcnet32_poll;
1959#endif
1960 1965
1961#ifdef CONFIG_NET_POLL_CONTROLLER 1966#ifdef CONFIG_NET_POLL_CONTROLLER
1962 dev->poll_controller = pcnet32_poll_controller; 1967 dev->poll_controller = pcnet32_poll_controller;
@@ -2276,6 +2281,10 @@ static int pcnet32_open(struct net_device *dev)
2276 goto err_free_ring; 2281 goto err_free_ring;
2277 } 2282 }
2278 2283
2284#ifdef CONFIG_PCNET32_NAPI
2285 napi_enable(&lp->napi);
2286#endif
2287
2279 /* Re-initialize the PCNET32, and start it when done. */ 2288 /* Re-initialize the PCNET32, and start it when done. */
2280 lp->a.write_csr(ioaddr, 1, (lp->init_dma_addr & 0xffff)); 2289 lp->a.write_csr(ioaddr, 1, (lp->init_dma_addr & 0xffff));
2281 lp->a.write_csr(ioaddr, 2, (lp->init_dma_addr >> 16)); 2290 lp->a.write_csr(ioaddr, 2, (lp->init_dma_addr >> 16));
@@ -2599,18 +2608,18 @@ pcnet32_interrupt(int irq, void *dev_id)
2599 /* unlike for the lance, there is no restart needed */ 2608 /* unlike for the lance, there is no restart needed */
2600 } 2609 }
2601#ifdef CONFIG_PCNET32_NAPI 2610#ifdef CONFIG_PCNET32_NAPI
2602 if (netif_rx_schedule_prep(dev)) { 2611 if (netif_rx_schedule_prep(dev, &lp->napi)) {
2603 u16 val; 2612 u16 val;
2604 /* set interrupt masks */ 2613 /* set interrupt masks */
2605 val = lp->a.read_csr(ioaddr, CSR3); 2614 val = lp->a.read_csr(ioaddr, CSR3);
2606 val |= 0x5f00; 2615 val |= 0x5f00;
2607 lp->a.write_csr(ioaddr, CSR3, val); 2616 lp->a.write_csr(ioaddr, CSR3, val);
2608 mmiowb(); 2617 mmiowb();
2609 __netif_rx_schedule(dev); 2618 __netif_rx_schedule(dev, &lp->napi);
2610 break; 2619 break;
2611 } 2620 }
2612#else 2621#else
2613 pcnet32_rx(dev, dev->weight); 2622 pcnet32_rx(dev, lp->napi.weight);
2614 if (pcnet32_tx(dev)) { 2623 if (pcnet32_tx(dev)) {
2615 /* reset the chip to clear the error condition, then restart */ 2624 /* reset the chip to clear the error condition, then restart */
2616 lp->a.reset(ioaddr); 2625 lp->a.reset(ioaddr);
@@ -2645,6 +2654,9 @@ static int pcnet32_close(struct net_device *dev)
2645 del_timer_sync(&lp->watchdog_timer); 2654 del_timer_sync(&lp->watchdog_timer);
2646 2655
2647 netif_stop_queue(dev); 2656 netif_stop_queue(dev);
2657#ifdef CONFIG_PCNET32_NAPI
2658 napi_disable(&lp->napi);
2659#endif
2648 2660
2649 spin_lock_irqsave(&lp->lock, flags); 2661 spin_lock_irqsave(&lp->lock, flags);
2650 2662
diff --git a/drivers/net/ps3_gelic_net.c b/drivers/net/ps3_gelic_net.c
index e56503918436..92561c0450bc 100644
--- a/drivers/net/ps3_gelic_net.c
+++ b/drivers/net/ps3_gelic_net.c
@@ -556,6 +556,7 @@ static int gelic_net_stop(struct net_device *netdev)
556{ 556{
557 struct gelic_net_card *card = netdev_priv(netdev); 557 struct gelic_net_card *card = netdev_priv(netdev);
558 558
559 napi_disable(&card->napi);
559 netif_stop_queue(netdev); 560 netif_stop_queue(netdev);
560 561
561 /* turn off DMA, force end */ 562 /* turn off DMA, force end */
@@ -987,32 +988,24 @@ refill:
987 * if the quota is exceeded, but the driver has still packets. 988 * if the quota is exceeded, but the driver has still packets.
988 * 989 *
989 */ 990 */
990static int gelic_net_poll(struct net_device *netdev, int *budget) 991static int gelic_net_poll(struct napi_struct *napi, int budget)
991{ 992{
992 struct gelic_net_card *card = netdev_priv(netdev); 993 struct gelic_net_card *card = container_of(napi, struct gelic_net_card, napi);
993 int packets_to_do, packets_done = 0; 994 struct net_device *netdev = card->netdev;
994 int no_more_packets = 0; 995 int packets_done = 0;
995
996 packets_to_do = min(*budget, netdev->quota);
997 996
998 while (packets_to_do) { 997 while (packets_done < budget) {
999 if (gelic_net_decode_one_descr(card)) { 998 if (!gelic_net_decode_one_descr(card))
1000 packets_done++;
1001 packets_to_do--;
1002 } else {
1003 /* no more packets for the stack */
1004 no_more_packets = 1;
1005 break; 999 break;
1006 } 1000
1001 packets_done++;
1007 } 1002 }
1008 netdev->quota -= packets_done; 1003
1009 *budget -= packets_done; 1004 if (packets_done < budget) {
1010 if (no_more_packets) { 1005 netif_rx_complete(netdev, napi);
1011 netif_rx_complete(netdev);
1012 gelic_net_rx_irq_on(card); 1006 gelic_net_rx_irq_on(card);
1013 return 0; 1007 }
1014 } else 1008 return packets_done;
1015 return 1;
1016} 1009}
1017/** 1010/**
1018 * gelic_net_change_mtu - changes the MTU of an interface 1011 * gelic_net_change_mtu - changes the MTU of an interface
@@ -1055,7 +1048,7 @@ static irqreturn_t gelic_net_interrupt(int irq, void *ptr)
1055 1048
1056 if (status & GELIC_NET_RXINT) { 1049 if (status & GELIC_NET_RXINT) {
1057 gelic_net_rx_irq_off(card); 1050 gelic_net_rx_irq_off(card);
1058 netif_rx_schedule(netdev); 1051 netif_rx_schedule(netdev, &card->napi);
1059 } 1052 }
1060 1053
1061 if (status & GELIC_NET_TXINT) { 1054 if (status & GELIC_NET_TXINT) {
@@ -1159,6 +1152,8 @@ static int gelic_net_open(struct net_device *netdev)
1159 if (gelic_net_alloc_rx_skbs(card)) 1152 if (gelic_net_alloc_rx_skbs(card))
1160 goto alloc_skbs_failed; 1153 goto alloc_skbs_failed;
1161 1154
1155 napi_enable(&card->napi);
1156
1162 card->tx_dma_progress = 0; 1157 card->tx_dma_progress = 0;
1163 card->ghiintmask = GELIC_NET_RXINT | GELIC_NET_TXINT; 1158 card->ghiintmask = GELIC_NET_RXINT | GELIC_NET_TXINT;
1164 1159
@@ -1360,9 +1355,6 @@ static void gelic_net_setup_netdev_ops(struct net_device *netdev)
1360 /* tx watchdog */ 1355 /* tx watchdog */
1361 netdev->tx_timeout = &gelic_net_tx_timeout; 1356 netdev->tx_timeout = &gelic_net_tx_timeout;
1362 netdev->watchdog_timeo = GELIC_NET_WATCHDOG_TIMEOUT; 1357 netdev->watchdog_timeo = GELIC_NET_WATCHDOG_TIMEOUT;
1363 /* NAPI */
1364 netdev->poll = &gelic_net_poll;
1365 netdev->weight = GELIC_NET_NAPI_WEIGHT;
1366 netdev->ethtool_ops = &gelic_net_ethtool_ops; 1358 netdev->ethtool_ops = &gelic_net_ethtool_ops;
1367} 1359}
1368 1360
@@ -1390,6 +1382,9 @@ static int gelic_net_setup_netdev(struct gelic_net_card *card)
1390 1382
1391 gelic_net_setup_netdev_ops(netdev); 1383 gelic_net_setup_netdev_ops(netdev);
1392 1384
1385 netif_napi_add(netdev, &card->napi,
1386 gelic_net_poll, GELIC_NET_NAPI_WEIGHT);
1387
1393 netdev->features = NETIF_F_IP_CSUM; 1388 netdev->features = NETIF_F_IP_CSUM;
1394 1389
1395 status = lv1_net_control(bus_id(card), dev_id(card), 1390 status = lv1_net_control(bus_id(card), dev_id(card),
diff --git a/drivers/net/ps3_gelic_net.h b/drivers/net/ps3_gelic_net.h
index a9c4c4fc2547..968560269a3b 100644
--- a/drivers/net/ps3_gelic_net.h
+++ b/drivers/net/ps3_gelic_net.h
@@ -194,6 +194,7 @@ struct gelic_net_descr_chain {
194 194
195struct gelic_net_card { 195struct gelic_net_card {
196 struct net_device *netdev; 196 struct net_device *netdev;
197 struct napi_struct napi;
197 /* 198 /*
198 * hypervisor requires irq_status should be 199 * hypervisor requires irq_status should be
199 * 8 bytes aligned, but u64 member is 200 * 8 bytes aligned, but u64 member is
diff --git a/drivers/net/qla3xxx.c b/drivers/net/qla3xxx.c
index ea151315050c..bf9f8f64ba67 100755
--- a/drivers/net/qla3xxx.c
+++ b/drivers/net/qla3xxx.c
@@ -2310,10 +2310,10 @@ static int ql_tx_rx_clean(struct ql3_adapter *qdev,
2310 return work_done; 2310 return work_done;
2311} 2311}
2312 2312
2313static int ql_poll(struct net_device *ndev, int *budget) 2313static int ql_poll(struct napi_struct *napi, int budget)
2314{ 2314{
2315 struct ql3_adapter *qdev = netdev_priv(ndev); 2315 struct ql3_adapter *qdev = container_of(napi, struct ql3_adapter, napi);
2316 int work_to_do = min(*budget, ndev->quota); 2316 struct net_device *ndev = qdev->ndev;
2317 int rx_cleaned = 0, tx_cleaned = 0; 2317 int rx_cleaned = 0, tx_cleaned = 0;
2318 unsigned long hw_flags; 2318 unsigned long hw_flags;
2319 struct ql3xxx_port_registers __iomem *port_regs = qdev->mem_map_registers; 2319 struct ql3xxx_port_registers __iomem *port_regs = qdev->mem_map_registers;
@@ -2321,16 +2321,13 @@ static int ql_poll(struct net_device *ndev, int *budget)
2321 if (!netif_carrier_ok(ndev)) 2321 if (!netif_carrier_ok(ndev))
2322 goto quit_polling; 2322 goto quit_polling;
2323 2323
2324 ql_tx_rx_clean(qdev, &tx_cleaned, &rx_cleaned, work_to_do); 2324 ql_tx_rx_clean(qdev, &tx_cleaned, &rx_cleaned, budget);
2325 *budget -= rx_cleaned;
2326 ndev->quota -= rx_cleaned;
2327 2325
2328 if( tx_cleaned + rx_cleaned != work_to_do || 2326 if (tx_cleaned + rx_cleaned != budget ||
2329 !netif_running(ndev)) { 2327 !netif_running(ndev)) {
2330quit_polling: 2328quit_polling:
2331 netif_rx_complete(ndev);
2332
2333 spin_lock_irqsave(&qdev->hw_lock, hw_flags); 2329 spin_lock_irqsave(&qdev->hw_lock, hw_flags);
2330 __netif_rx_complete(ndev, napi);
2334 ql_update_small_bufq_prod_index(qdev); 2331 ql_update_small_bufq_prod_index(qdev);
2335 ql_update_lrg_bufq_prod_index(qdev); 2332 ql_update_lrg_bufq_prod_index(qdev);
2336 writel(qdev->rsp_consumer_index, 2333 writel(qdev->rsp_consumer_index,
@@ -2338,9 +2335,8 @@ quit_polling:
2338 spin_unlock_irqrestore(&qdev->hw_lock, hw_flags); 2335 spin_unlock_irqrestore(&qdev->hw_lock, hw_flags);
2339 2336
2340 ql_enable_interrupts(qdev); 2337 ql_enable_interrupts(qdev);
2341 return 0;
2342 } 2338 }
2343 return 1; 2339 return tx_cleaned + rx_cleaned;
2344} 2340}
2345 2341
2346static irqreturn_t ql3xxx_isr(int irq, void *dev_id) 2342static irqreturn_t ql3xxx_isr(int irq, void *dev_id)
@@ -2390,8 +2386,8 @@ static irqreturn_t ql3xxx_isr(int irq, void *dev_id)
2390 spin_unlock(&qdev->adapter_lock); 2386 spin_unlock(&qdev->adapter_lock);
2391 } else if (value & ISP_IMR_DISABLE_CMPL_INT) { 2387 } else if (value & ISP_IMR_DISABLE_CMPL_INT) {
2392 ql_disable_interrupts(qdev); 2388 ql_disable_interrupts(qdev);
2393 if (likely(netif_rx_schedule_prep(ndev))) { 2389 if (likely(netif_rx_schedule_prep(ndev, &qdev->napi))) {
2394 __netif_rx_schedule(ndev); 2390 __netif_rx_schedule(ndev, &qdev->napi);
2395 } 2391 }
2396 } else { 2392 } else {
2397 return IRQ_NONE; 2393 return IRQ_NONE;
@@ -3617,7 +3613,7 @@ static int ql_adapter_down(struct ql3_adapter *qdev, int do_reset)
3617 3613
3618 del_timer_sync(&qdev->adapter_timer); 3614 del_timer_sync(&qdev->adapter_timer);
3619 3615
3620 netif_poll_disable(ndev); 3616 napi_disable(&qdev->napi);
3621 3617
3622 if (do_reset) { 3618 if (do_reset) {
3623 int soft_reset; 3619 int soft_reset;
@@ -3705,7 +3701,7 @@ static int ql_adapter_up(struct ql3_adapter *qdev)
3705 3701
3706 mod_timer(&qdev->adapter_timer, jiffies + HZ * 1); 3702 mod_timer(&qdev->adapter_timer, jiffies + HZ * 1);
3707 3703
3708 netif_poll_enable(ndev); 3704 napi_enable(&qdev->napi);
3709 ql_enable_interrupts(qdev); 3705 ql_enable_interrupts(qdev);
3710 return 0; 3706 return 0;
3711 3707
@@ -4061,8 +4057,7 @@ static int __devinit ql3xxx_probe(struct pci_dev *pdev,
4061 ndev->tx_timeout = ql3xxx_tx_timeout; 4057 ndev->tx_timeout = ql3xxx_tx_timeout;
4062 ndev->watchdog_timeo = 5 * HZ; 4058 ndev->watchdog_timeo = 5 * HZ;
4063 4059
4064 ndev->poll = &ql_poll; 4060 netif_napi_add(ndev, &qdev->napi, ql_poll, 64);
4065 ndev->weight = 64;
4066 4061
4067 ndev->irq = pdev->irq; 4062 ndev->irq = pdev->irq;
4068 4063
diff --git a/drivers/net/qla3xxx.h b/drivers/net/qla3xxx.h
index 4a832c46c274..aa2216f0d7b8 100755
--- a/drivers/net/qla3xxx.h
+++ b/drivers/net/qla3xxx.h
@@ -1175,6 +1175,8 @@ struct ql3_adapter {
1175 struct pci_dev *pdev; 1175 struct pci_dev *pdev;
1176 struct net_device *ndev; /* Parent NET device */ 1176 struct net_device *ndev; /* Parent NET device */
1177 1177
1178 struct napi_struct napi;
1179
1178 /* Hardware information */ 1180 /* Hardware information */
1179 u8 chip_rev_id; 1181 u8 chip_rev_id;
1180 u8 pci_slot; 1182 u8 pci_slot;
diff --git a/drivers/net/r8169.c b/drivers/net/r8169.c
index c76dd29c8e9a..3f2306e3f517 100644
--- a/drivers/net/r8169.c
+++ b/drivers/net/r8169.c
@@ -384,6 +384,7 @@ struct rtl8169_private {
384 void __iomem *mmio_addr; /* memory map physical address */ 384 void __iomem *mmio_addr; /* memory map physical address */
385 struct pci_dev *pci_dev; /* Index of PCI device */ 385 struct pci_dev *pci_dev; /* Index of PCI device */
386 struct net_device *dev; 386 struct net_device *dev;
387 struct napi_struct napi;
387 struct net_device_stats stats; /* statistics of net device */ 388 struct net_device_stats stats; /* statistics of net device */
388 spinlock_t lock; /* spin lock flag */ 389 spinlock_t lock; /* spin lock flag */
389 u32 msg_enable; 390 u32 msg_enable;
@@ -443,13 +444,13 @@ static void rtl_set_rx_mode(struct net_device *dev);
443static void rtl8169_tx_timeout(struct net_device *dev); 444static void rtl8169_tx_timeout(struct net_device *dev);
444static struct net_device_stats *rtl8169_get_stats(struct net_device *dev); 445static struct net_device_stats *rtl8169_get_stats(struct net_device *dev);
445static int rtl8169_rx_interrupt(struct net_device *, struct rtl8169_private *, 446static int rtl8169_rx_interrupt(struct net_device *, struct rtl8169_private *,
446 void __iomem *); 447 void __iomem *, u32 budget);
447static int rtl8169_change_mtu(struct net_device *dev, int new_mtu); 448static int rtl8169_change_mtu(struct net_device *dev, int new_mtu);
448static void rtl8169_down(struct net_device *dev); 449static void rtl8169_down(struct net_device *dev);
449static void rtl8169_rx_clear(struct rtl8169_private *tp); 450static void rtl8169_rx_clear(struct rtl8169_private *tp);
450 451
451#ifdef CONFIG_R8169_NAPI 452#ifdef CONFIG_R8169_NAPI
452static int rtl8169_poll(struct net_device *dev, int *budget); 453static int rtl8169_poll(struct napi_struct *napi, int budget);
453#endif 454#endif
454 455
455static const unsigned int rtl8169_rx_config = 456static const unsigned int rtl8169_rx_config =
@@ -1656,8 +1657,7 @@ rtl8169_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
1656 dev->set_mac_address = rtl_set_mac_address; 1657 dev->set_mac_address = rtl_set_mac_address;
1657 1658
1658#ifdef CONFIG_R8169_NAPI 1659#ifdef CONFIG_R8169_NAPI
1659 dev->poll = rtl8169_poll; 1660 netif_napi_add(dev, &tp->napi, rtl8169_poll, R8169_NAPI_WEIGHT);
1660 dev->weight = R8169_NAPI_WEIGHT;
1661#endif 1661#endif
1662 1662
1663#ifdef CONFIG_R8169_VLAN 1663#ifdef CONFIG_R8169_VLAN
@@ -1777,6 +1777,10 @@ static int rtl8169_open(struct net_device *dev)
1777 if (retval < 0) 1777 if (retval < 0)
1778 goto err_release_ring_2; 1778 goto err_release_ring_2;
1779 1779
1780#ifdef CONFIG_R8169_NAPI
1781 napi_enable(&tp->napi);
1782#endif
1783
1780 rtl_hw_start(dev); 1784 rtl_hw_start(dev);
1781 1785
1782 rtl8169_request_timer(dev); 1786 rtl8169_request_timer(dev);
@@ -2082,7 +2086,9 @@ static int rtl8169_change_mtu(struct net_device *dev, int new_mtu)
2082 if (ret < 0) 2086 if (ret < 0)
2083 goto out; 2087 goto out;
2084 2088
2085 netif_poll_enable(dev); 2089#ifdef CONFIG_R8169_NAPI
2090 napi_enable(&tp->napi);
2091#endif
2086 2092
2087 rtl_hw_start(dev); 2093 rtl_hw_start(dev);
2088 2094
@@ -2274,11 +2280,15 @@ static void rtl8169_wait_for_quiescence(struct net_device *dev)
2274 synchronize_irq(dev->irq); 2280 synchronize_irq(dev->irq);
2275 2281
2276 /* Wait for any pending NAPI task to complete */ 2282 /* Wait for any pending NAPI task to complete */
2277 netif_poll_disable(dev); 2283#ifdef CONFIG_R8169_NAPI
2284 napi_disable(&tp->napi);
2285#endif
2278 2286
2279 rtl8169_irq_mask_and_ack(ioaddr); 2287 rtl8169_irq_mask_and_ack(ioaddr);
2280 2288
2281 netif_poll_enable(dev); 2289#ifdef CONFIG_R8169_NAPI
2290 napi_enable(&tp->napi);
2291#endif
2282} 2292}
2283 2293
2284static void rtl8169_reinit_task(struct work_struct *work) 2294static void rtl8169_reinit_task(struct work_struct *work)
@@ -2322,7 +2332,7 @@ static void rtl8169_reset_task(struct work_struct *work)
2322 2332
2323 rtl8169_wait_for_quiescence(dev); 2333 rtl8169_wait_for_quiescence(dev);
2324 2334
2325 rtl8169_rx_interrupt(dev, tp, tp->mmio_addr); 2335 rtl8169_rx_interrupt(dev, tp, tp->mmio_addr, ~(u32)0);
2326 rtl8169_tx_clear(tp); 2336 rtl8169_tx_clear(tp);
2327 2337
2328 if (tp->dirty_rx == tp->cur_rx) { 2338 if (tp->dirty_rx == tp->cur_rx) {
@@ -2636,14 +2646,14 @@ out:
2636 2646
2637static int rtl8169_rx_interrupt(struct net_device *dev, 2647static int rtl8169_rx_interrupt(struct net_device *dev,
2638 struct rtl8169_private *tp, 2648 struct rtl8169_private *tp,
2639 void __iomem *ioaddr) 2649 void __iomem *ioaddr, u32 budget)
2640{ 2650{
2641 unsigned int cur_rx, rx_left; 2651 unsigned int cur_rx, rx_left;
2642 unsigned int delta, count; 2652 unsigned int delta, count;
2643 2653
2644 cur_rx = tp->cur_rx; 2654 cur_rx = tp->cur_rx;
2645 rx_left = NUM_RX_DESC + tp->dirty_rx - cur_rx; 2655 rx_left = NUM_RX_DESC + tp->dirty_rx - cur_rx;
2646 rx_left = rtl8169_rx_quota(rx_left, (u32) dev->quota); 2656 rx_left = rtl8169_rx_quota(rx_left, budget);
2647 2657
2648 for (; rx_left > 0; rx_left--, cur_rx++) { 2658 for (; rx_left > 0; rx_left--, cur_rx++) {
2649 unsigned int entry = cur_rx % NUM_RX_DESC; 2659 unsigned int entry = cur_rx % NUM_RX_DESC;
@@ -2792,8 +2802,8 @@ static irqreturn_t rtl8169_interrupt(int irq, void *dev_instance)
2792 RTL_W16(IntrMask, tp->intr_event & ~tp->napi_event); 2802 RTL_W16(IntrMask, tp->intr_event & ~tp->napi_event);
2793 tp->intr_mask = ~tp->napi_event; 2803 tp->intr_mask = ~tp->napi_event;
2794 2804
2795 if (likely(netif_rx_schedule_prep(dev))) 2805 if (likely(netif_rx_schedule_prep(dev, &tp->napi)))
2796 __netif_rx_schedule(dev); 2806 __netif_rx_schedule(dev, &tp->napi);
2797 else if (netif_msg_intr(tp)) { 2807 else if (netif_msg_intr(tp)) {
2798 printk(KERN_INFO "%s: interrupt %04x in poll\n", 2808 printk(KERN_INFO "%s: interrupt %04x in poll\n",
2799 dev->name, status); 2809 dev->name, status);
@@ -2803,7 +2813,7 @@ static irqreturn_t rtl8169_interrupt(int irq, void *dev_instance)
2803#else 2813#else
2804 /* Rx interrupt */ 2814 /* Rx interrupt */
2805 if (status & (RxOK | RxOverflow | RxFIFOOver)) 2815 if (status & (RxOK | RxOverflow | RxFIFOOver))
2806 rtl8169_rx_interrupt(dev, tp, ioaddr); 2816 rtl8169_rx_interrupt(dev, tp, ioaddr, ~(u32)0);
2807 2817
2808 /* Tx interrupt */ 2818 /* Tx interrupt */
2809 if (status & (TxOK | TxErr)) 2819 if (status & (TxOK | TxErr))
@@ -2826,20 +2836,18 @@ out:
2826} 2836}
2827 2837
2828#ifdef CONFIG_R8169_NAPI 2838#ifdef CONFIG_R8169_NAPI
2829static int rtl8169_poll(struct net_device *dev, int *budget) 2839static int rtl8169_poll(struct napi_struct *napi, int budget)
2830{ 2840{
2831 unsigned int work_done, work_to_do = min(*budget, dev->quota); 2841 struct rtl8169_private *tp = container_of(napi, struct rtl8169_private, napi);
2832 struct rtl8169_private *tp = netdev_priv(dev); 2842 struct net_device *dev = tp->dev;
2833 void __iomem *ioaddr = tp->mmio_addr; 2843 void __iomem *ioaddr = tp->mmio_addr;
2844 int work_done;
2834 2845
2835 work_done = rtl8169_rx_interrupt(dev, tp, ioaddr); 2846 work_done = rtl8169_rx_interrupt(dev, tp, ioaddr, (u32) budget);
2836 rtl8169_tx_interrupt(dev, tp, ioaddr); 2847 rtl8169_tx_interrupt(dev, tp, ioaddr);
2837 2848
2838 *budget -= work_done; 2849 if (work_done < budget) {
2839 dev->quota -= work_done; 2850 netif_rx_complete(dev, napi);
2840
2841 if (work_done < work_to_do) {
2842 netif_rx_complete(dev);
2843 tp->intr_mask = 0xffff; 2851 tp->intr_mask = 0xffff;
2844 /* 2852 /*
2845 * 20040426: the barrier is not strictly required but the 2853 * 20040426: the barrier is not strictly required but the
@@ -2851,7 +2859,7 @@ static int rtl8169_poll(struct net_device *dev, int *budget)
2851 RTL_W16(IntrMask, tp->intr_event); 2859 RTL_W16(IntrMask, tp->intr_event);
2852 } 2860 }
2853 2861
2854 return (work_done >= work_to_do); 2862 return work_done;
2855} 2863}
2856#endif 2864#endif
2857 2865
@@ -2880,7 +2888,7 @@ core_down:
2880 synchronize_irq(dev->irq); 2888 synchronize_irq(dev->irq);
2881 2889
2882 if (!poll_locked) { 2890 if (!poll_locked) {
2883 netif_poll_disable(dev); 2891 napi_disable(&tp->napi);
2884 poll_locked++; 2892 poll_locked++;
2885 } 2893 }
2886 2894
@@ -2918,8 +2926,6 @@ static int rtl8169_close(struct net_device *dev)
2918 2926
2919 free_irq(dev->irq, dev); 2927 free_irq(dev->irq, dev);
2920 2928
2921 netif_poll_enable(dev);
2922
2923 pci_free_consistent(pdev, R8169_RX_RING_BYTES, tp->RxDescArray, 2929 pci_free_consistent(pdev, R8169_RX_RING_BYTES, tp->RxDescArray,
2924 tp->RxPhyAddr); 2930 tp->RxPhyAddr);
2925 pci_free_consistent(pdev, R8169_TX_RING_BYTES, tp->TxDescArray, 2931 pci_free_consistent(pdev, R8169_TX_RING_BYTES, tp->TxDescArray,
diff --git a/drivers/net/s2io.c b/drivers/net/s2io.c
index 24feb00600ee..dd012322cdbe 100644
--- a/drivers/net/s2io.c
+++ b/drivers/net/s2io.c
@@ -2568,7 +2568,7 @@ static void free_rx_buffers(struct s2io_nic *sp)
2568 2568
2569/** 2569/**
2570 * s2io_poll - Rx interrupt handler for NAPI support 2570 * s2io_poll - Rx interrupt handler for NAPI support
2571 * @dev : pointer to the device structure. 2571 * @napi : pointer to the napi structure.
2572 * @budget : The number of packets that were budgeted to be processed 2572 * @budget : The number of packets that were budgeted to be processed
2573 * during one pass through the 'Poll" function. 2573 * during one pass through the 'Poll" function.
2574 * Description: 2574 * Description:
@@ -2579,9 +2579,10 @@ static void free_rx_buffers(struct s2io_nic *sp)
2579 * 0 on success and 1 if there are No Rx packets to be processed. 2579 * 0 on success and 1 if there are No Rx packets to be processed.
2580 */ 2580 */
2581 2581
2582static int s2io_poll(struct net_device *dev, int *budget) 2582static int s2io_poll(struct napi_struct *napi, int budget)
2583{ 2583{
2584 struct s2io_nic *nic = dev->priv; 2584 struct s2io_nic *nic = container_of(napi, struct s2io_nic, napi);
2585 struct net_device *dev = nic->dev;
2585 int pkt_cnt = 0, org_pkts_to_process; 2586 int pkt_cnt = 0, org_pkts_to_process;
2586 struct mac_info *mac_control; 2587 struct mac_info *mac_control;
2587 struct config_param *config; 2588 struct config_param *config;
@@ -2592,9 +2593,7 @@ static int s2io_poll(struct net_device *dev, int *budget)
2592 mac_control = &nic->mac_control; 2593 mac_control = &nic->mac_control;
2593 config = &nic->config; 2594 config = &nic->config;
2594 2595
2595 nic->pkts_to_process = *budget; 2596 nic->pkts_to_process = budget;
2596 if (nic->pkts_to_process > dev->quota)
2597 nic->pkts_to_process = dev->quota;
2598 org_pkts_to_process = nic->pkts_to_process; 2597 org_pkts_to_process = nic->pkts_to_process;
2599 2598
2600 writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_int); 2599 writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_int);
@@ -2608,12 +2607,8 @@ static int s2io_poll(struct net_device *dev, int *budget)
2608 goto no_rx; 2607 goto no_rx;
2609 } 2608 }
2610 } 2609 }
2611 if (!pkt_cnt)
2612 pkt_cnt = 1;
2613 2610
2614 dev->quota -= pkt_cnt; 2611 netif_rx_complete(dev, napi);
2615 *budget -= pkt_cnt;
2616 netif_rx_complete(dev);
2617 2612
2618 for (i = 0; i < config->rx_ring_num; i++) { 2613 for (i = 0; i < config->rx_ring_num; i++) {
2619 if (fill_rx_buffers(nic, i) == -ENOMEM) { 2614 if (fill_rx_buffers(nic, i) == -ENOMEM) {
@@ -2626,12 +2621,9 @@ static int s2io_poll(struct net_device *dev, int *budget)
2626 writeq(0x0, &bar0->rx_traffic_mask); 2621 writeq(0x0, &bar0->rx_traffic_mask);
2627 readl(&bar0->rx_traffic_mask); 2622 readl(&bar0->rx_traffic_mask);
2628 atomic_dec(&nic->isr_cnt); 2623 atomic_dec(&nic->isr_cnt);
2629 return 0; 2624 return pkt_cnt;
2630 2625
2631no_rx: 2626no_rx:
2632 dev->quota -= pkt_cnt;
2633 *budget -= pkt_cnt;
2634
2635 for (i = 0; i < config->rx_ring_num; i++) { 2627 for (i = 0; i < config->rx_ring_num; i++) {
2636 if (fill_rx_buffers(nic, i) == -ENOMEM) { 2628 if (fill_rx_buffers(nic, i) == -ENOMEM) {
2637 DBG_PRINT(INFO_DBG, "%s:Out of memory", dev->name); 2629 DBG_PRINT(INFO_DBG, "%s:Out of memory", dev->name);
@@ -2640,7 +2632,7 @@ no_rx:
2640 } 2632 }
2641 } 2633 }
2642 atomic_dec(&nic->isr_cnt); 2634 atomic_dec(&nic->isr_cnt);
2643 return 1; 2635 return pkt_cnt;
2644} 2636}
2645 2637
2646#ifdef CONFIG_NET_POLL_CONTROLLER 2638#ifdef CONFIG_NET_POLL_CONTROLLER
@@ -3809,6 +3801,8 @@ static int s2io_open(struct net_device *dev)
3809 netif_carrier_off(dev); 3801 netif_carrier_off(dev);
3810 sp->last_link_state = 0; 3802 sp->last_link_state = 0;
3811 3803
3804 napi_enable(&sp->napi);
3805
3812 /* Initialize H/W and enable interrupts */ 3806 /* Initialize H/W and enable interrupts */
3813 err = s2io_card_up(sp); 3807 err = s2io_card_up(sp);
3814 if (err) { 3808 if (err) {
@@ -3828,6 +3822,7 @@ static int s2io_open(struct net_device *dev)
3828 return 0; 3822 return 0;
3829 3823
3830hw_init_failed: 3824hw_init_failed:
3825 napi_disable(&sp->napi);
3831 if (sp->intr_type == MSI_X) { 3826 if (sp->intr_type == MSI_X) {
3832 if (sp->entries) { 3827 if (sp->entries) {
3833 kfree(sp->entries); 3828 kfree(sp->entries);
@@ -3861,6 +3856,7 @@ static int s2io_close(struct net_device *dev)
3861 struct s2io_nic *sp = dev->priv; 3856 struct s2io_nic *sp = dev->priv;
3862 3857
3863 netif_stop_queue(dev); 3858 netif_stop_queue(dev);
3859 napi_disable(&sp->napi);
3864 /* Reset card, kill tasklet and free Tx and Rx buffers. */ 3860 /* Reset card, kill tasklet and free Tx and Rx buffers. */
3865 s2io_card_down(sp); 3861 s2io_card_down(sp);
3866 3862
@@ -4232,8 +4228,8 @@ static irqreturn_t s2io_isr(int irq, void *dev_id)
4232 4228
4233 if (napi) { 4229 if (napi) {
4234 if (reason & GEN_INTR_RXTRAFFIC) { 4230 if (reason & GEN_INTR_RXTRAFFIC) {
4235 if ( likely ( netif_rx_schedule_prep(dev)) ) { 4231 if (likely (netif_rx_schedule_prep(dev, &sp->napi))) {
4236 __netif_rx_schedule(dev); 4232 __netif_rx_schedule(dev, &sp->napi);
4237 writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_mask); 4233 writeq(S2IO_MINUS_ONE, &bar0->rx_traffic_mask);
4238 } 4234 }
4239 else 4235 else
@@ -7215,8 +7211,7 @@ s2io_init_nic(struct pci_dev *pdev, const struct pci_device_id *pre)
7215 * will use eth_mac_addr() for dev->set_mac_address 7211 * will use eth_mac_addr() for dev->set_mac_address
7216 * mac address will be set every time dev->open() is called 7212 * mac address will be set every time dev->open() is called
7217 */ 7213 */
7218 dev->poll = s2io_poll; 7214 netif_napi_add(dev, &sp->napi, s2io_poll, 32);
7219 dev->weight = 32;
7220 7215
7221#ifdef CONFIG_NET_POLL_CONTROLLER 7216#ifdef CONFIG_NET_POLL_CONTROLLER
7222 dev->poll_controller = s2io_netpoll; 7217 dev->poll_controller = s2io_netpoll;
diff --git a/drivers/net/s2io.h b/drivers/net/s2io.h
index 92983ee7df8c..420fefb99188 100644
--- a/drivers/net/s2io.h
+++ b/drivers/net/s2io.h
@@ -786,6 +786,7 @@ struct s2io_nic {
786 */ 786 */
787 int pkts_to_process; 787 int pkts_to_process;
788 struct net_device *dev; 788 struct net_device *dev;
789 struct napi_struct napi;
789 struct mac_info mac_control; 790 struct mac_info mac_control;
790 struct config_param config; 791 struct config_param config;
791 struct pci_dev *pdev; 792 struct pci_dev *pdev;
@@ -1019,7 +1020,7 @@ static void s2io_set_multicast(struct net_device *dev);
1019static int rx_osm_handler(struct ring_info *ring_data, struct RxD_t * rxdp); 1020static int rx_osm_handler(struct ring_info *ring_data, struct RxD_t * rxdp);
1020static void s2io_link(struct s2io_nic * sp, int link); 1021static void s2io_link(struct s2io_nic * sp, int link);
1021static void s2io_reset(struct s2io_nic * sp); 1022static void s2io_reset(struct s2io_nic * sp);
1022static int s2io_poll(struct net_device *dev, int *budget); 1023static int s2io_poll(struct napi_struct *napi, int budget);
1023static void s2io_init_pci(struct s2io_nic * sp); 1024static void s2io_init_pci(struct s2io_nic * sp);
1024static int s2io_set_mac_addr(struct net_device *dev, u8 * addr); 1025static int s2io_set_mac_addr(struct net_device *dev, u8 * addr);
1025static void s2io_alarm_handle(unsigned long data); 1026static void s2io_alarm_handle(unsigned long data);
diff --git a/drivers/net/sb1250-mac.c b/drivers/net/sb1250-mac.c
index e7fdcf15b5a7..53845ebb649f 100644
--- a/drivers/net/sb1250-mac.c
+++ b/drivers/net/sb1250-mac.c
@@ -238,6 +238,7 @@ struct sbmac_softc {
238 */ 238 */
239 239
240 struct net_device *sbm_dev; /* pointer to linux device */ 240 struct net_device *sbm_dev; /* pointer to linux device */
241 struct napi_struct napi;
241 spinlock_t sbm_lock; /* spin lock */ 242 spinlock_t sbm_lock; /* spin lock */
242 struct timer_list sbm_timer; /* for monitoring MII */ 243 struct timer_list sbm_timer; /* for monitoring MII */
243 struct net_device_stats sbm_stats; 244 struct net_device_stats sbm_stats;
@@ -320,7 +321,7 @@ static struct net_device_stats *sbmac_get_stats(struct net_device *dev);
320static void sbmac_set_rx_mode(struct net_device *dev); 321static void sbmac_set_rx_mode(struct net_device *dev);
321static int sbmac_mii_ioctl(struct net_device *dev, struct ifreq *rq, int cmd); 322static int sbmac_mii_ioctl(struct net_device *dev, struct ifreq *rq, int cmd);
322static int sbmac_close(struct net_device *dev); 323static int sbmac_close(struct net_device *dev);
323static int sbmac_poll(struct net_device *poll_dev, int *budget); 324static int sbmac_poll(struct napi_struct *napi, int budget);
324 325
325static int sbmac_mii_poll(struct sbmac_softc *s,int noisy); 326static int sbmac_mii_poll(struct sbmac_softc *s,int noisy);
326static int sbmac_mii_probe(struct net_device *dev); 327static int sbmac_mii_probe(struct net_device *dev);
@@ -2154,20 +2155,13 @@ static irqreturn_t sbmac_intr(int irq,void *dev_instance)
2154 * Transmits on channel 0 2155 * Transmits on channel 0
2155 */ 2156 */
2156 2157
2157 if (isr & (M_MAC_INT_CHANNEL << S_MAC_TX_CH0)) { 2158 if (isr & (M_MAC_INT_CHANNEL << S_MAC_TX_CH0))
2158 sbdma_tx_process(sc,&(sc->sbm_txdma), 0); 2159 sbdma_tx_process(sc,&(sc->sbm_txdma), 0);
2159#ifdef CONFIG_NETPOLL_TRAP
2160 if (netpoll_trap()) {
2161 if (test_and_clear_bit(__LINK_STATE_XOFF, &dev->state))
2162 __netif_schedule(dev);
2163 }
2164#endif
2165 }
2166 2160
2167 if (isr & (M_MAC_INT_CHANNEL << S_MAC_RX_CH0)) { 2161 if (isr & (M_MAC_INT_CHANNEL << S_MAC_RX_CH0)) {
2168 if (netif_rx_schedule_prep(dev)) { 2162 if (netif_rx_schedule_prep(dev, &sc->napi)) {
2169 __raw_writeq(0, sc->sbm_imr); 2163 __raw_writeq(0, sc->sbm_imr);
2170 __netif_rx_schedule(dev); 2164 __netif_rx_schedule(dev, &sc->napi);
2171 /* Depend on the exit from poll to reenable intr */ 2165 /* Depend on the exit from poll to reenable intr */
2172 } 2166 }
2173 else { 2167 else {
@@ -2470,8 +2464,8 @@ static int sbmac_init(struct net_device *dev, int idx)
2470 dev->do_ioctl = sbmac_mii_ioctl; 2464 dev->do_ioctl = sbmac_mii_ioctl;
2471 dev->tx_timeout = sbmac_tx_timeout; 2465 dev->tx_timeout = sbmac_tx_timeout;
2472 dev->watchdog_timeo = TX_TIMEOUT; 2466 dev->watchdog_timeo = TX_TIMEOUT;
2473 dev->poll = sbmac_poll; 2467
2474 dev->weight = 16; 2468 netif_napi_add(dev, &sc->napi, sbmac_poll, 16);
2475 2469
2476 dev->change_mtu = sb1250_change_mtu; 2470 dev->change_mtu = sb1250_change_mtu;
2477#ifdef CONFIG_NET_POLL_CONTROLLER 2471#ifdef CONFIG_NET_POLL_CONTROLLER
@@ -2537,6 +2531,8 @@ static int sbmac_open(struct net_device *dev)
2537 return -EINVAL; 2531 return -EINVAL;
2538 } 2532 }
2539 2533
2534 napi_enable(&sc->napi);
2535
2540 /* 2536 /*
2541 * Configure default speed 2537 * Configure default speed
2542 */ 2538 */
@@ -2850,6 +2846,8 @@ static int sbmac_close(struct net_device *dev)
2850 unsigned long flags; 2846 unsigned long flags;
2851 int irq; 2847 int irq;
2852 2848
2849 napi_disable(&sc->napi);
2850
2853 sbmac_set_channel_state(sc,sbmac_state_off); 2851 sbmac_set_channel_state(sc,sbmac_state_off);
2854 2852
2855 del_timer_sync(&sc->sbm_timer); 2853 del_timer_sync(&sc->sbm_timer);
@@ -2874,26 +2872,17 @@ static int sbmac_close(struct net_device *dev)
2874 return 0; 2872 return 0;
2875} 2873}
2876 2874
2877static int sbmac_poll(struct net_device *dev, int *budget) 2875static int sbmac_poll(struct napi_struct *napi, int budget)
2878{ 2876{
2879 int work_to_do; 2877 struct sbmac_softc *sc = container_of(napi, struct sbmac_softc, napi);
2878 struct net_device *dev = sc->sbm_dev;
2880 int work_done; 2879 int work_done;
2881 struct sbmac_softc *sc = netdev_priv(dev);
2882
2883 work_to_do = min(*budget, dev->quota);
2884 work_done = sbdma_rx_process(sc, &(sc->sbm_rxdma), work_to_do, 1);
2885
2886 if (work_done > work_to_do)
2887 printk(KERN_ERR "%s exceeded work_to_do budget=%d quota=%d work-done=%d\n",
2888 sc->sbm_dev->name, *budget, dev->quota, work_done);
2889 2880
2881 work_done = sbdma_rx_process(sc, &(sc->sbm_rxdma), budget, 1);
2890 sbdma_tx_process(sc, &(sc->sbm_txdma), 1); 2882 sbdma_tx_process(sc, &(sc->sbm_txdma), 1);
2891 2883
2892 *budget -= work_done; 2884 if (work_done < budget) {
2893 dev->quota -= work_done; 2885 netif_rx_complete(dev, napi);
2894
2895 if (work_done < work_to_do) {
2896 netif_rx_complete(dev);
2897 2886
2898#ifdef CONFIG_SBMAC_COALESCE 2887#ifdef CONFIG_SBMAC_COALESCE
2899 __raw_writeq(((M_MAC_INT_EOP_COUNT | M_MAC_INT_EOP_TIMER) << S_MAC_TX_CH0) | 2888 __raw_writeq(((M_MAC_INT_EOP_COUNT | M_MAC_INT_EOP_TIMER) << S_MAC_TX_CH0) |
@@ -2905,7 +2894,7 @@ static int sbmac_poll(struct net_device *dev, int *budget)
2905#endif 2894#endif
2906 } 2895 }
2907 2896
2908 return (work_done >= work_to_do); 2897 return work_done;
2909} 2898}
2910 2899
2911#if defined(SBMAC_ETH0_HWADDR) || defined(SBMAC_ETH1_HWADDR) || defined(SBMAC_ETH2_HWADDR) || defined(SBMAC_ETH3_HWADDR) 2900#if defined(SBMAC_ETH0_HWADDR) || defined(SBMAC_ETH1_HWADDR) || defined(SBMAC_ETH2_HWADDR) || defined(SBMAC_ETH3_HWADDR)
diff --git a/drivers/net/sis190.c b/drivers/net/sis190.c
index d470b19c0810..038ccfbafdd1 100644
--- a/drivers/net/sis190.c
+++ b/drivers/net/sis190.c
@@ -47,24 +47,13 @@
47#define PHY_ID_ANY 0x1f 47#define PHY_ID_ANY 0x1f
48#define MII_REG_ANY 0x1f 48#define MII_REG_ANY 0x1f
49 49
50#ifdef CONFIG_SIS190_NAPI 50#define DRV_VERSION "1.2"
51#define NAPI_SUFFIX "-NAPI"
52#else
53#define NAPI_SUFFIX ""
54#endif
55
56#define DRV_VERSION "1.2" NAPI_SUFFIX
57#define DRV_NAME "sis190" 51#define DRV_NAME "sis190"
58#define SIS190_DRIVER_NAME DRV_NAME " Gigabit Ethernet driver " DRV_VERSION 52#define SIS190_DRIVER_NAME DRV_NAME " Gigabit Ethernet driver " DRV_VERSION
59#define PFX DRV_NAME ": " 53#define PFX DRV_NAME ": "
60 54
61#ifdef CONFIG_SIS190_NAPI
62#define sis190_rx_skb netif_receive_skb
63#define sis190_rx_quota(count, quota) min(count, quota)
64#else
65#define sis190_rx_skb netif_rx 55#define sis190_rx_skb netif_rx
66#define sis190_rx_quota(count, quota) count 56#define sis190_rx_quota(count, quota) count
67#endif
68 57
69#define MAC_ADDR_LEN 6 58#define MAC_ADDR_LEN 6
70 59
@@ -1115,10 +1104,8 @@ static void sis190_down(struct net_device *dev)
1115 1104
1116 synchronize_irq(dev->irq); 1105 synchronize_irq(dev->irq);
1117 1106
1118 if (!poll_locked) { 1107 if (!poll_locked)
1119 netif_poll_disable(dev);
1120 poll_locked++; 1108 poll_locked++;
1121 }
1122 1109
1123 synchronize_sched(); 1110 synchronize_sched();
1124 1111
@@ -1137,8 +1124,6 @@ static int sis190_close(struct net_device *dev)
1137 1124
1138 free_irq(dev->irq, dev); 1125 free_irq(dev->irq, dev);
1139 1126
1140 netif_poll_enable(dev);
1141
1142 pci_free_consistent(pdev, TX_RING_BYTES, tp->TxDescRing, tp->tx_dma); 1127 pci_free_consistent(pdev, TX_RING_BYTES, tp->TxDescRing, tp->tx_dma);
1143 pci_free_consistent(pdev, RX_RING_BYTES, tp->RxDescRing, tp->rx_dma); 1128 pci_free_consistent(pdev, RX_RING_BYTES, tp->RxDescRing, tp->rx_dma);
1144 1129
diff --git a/drivers/net/skge.c b/drivers/net/skge.c
index e3d8520209b8..0bf46ed4e684 100644
--- a/drivers/net/skge.c
+++ b/drivers/net/skge.c
@@ -2528,7 +2528,7 @@ static int skge_up(struct net_device *dev)
2528 skge_write32(hw, B0_IMSK, hw->intr_mask); 2528 skge_write32(hw, B0_IMSK, hw->intr_mask);
2529 spin_unlock_irq(&hw->hw_lock); 2529 spin_unlock_irq(&hw->hw_lock);
2530 2530
2531 netif_poll_enable(dev); 2531 napi_enable(&skge->napi);
2532 return 0; 2532 return 0;
2533 2533
2534 free_rx_ring: 2534 free_rx_ring:
@@ -2558,7 +2558,7 @@ static int skge_down(struct net_device *dev)
2558 if (hw->chip_id == CHIP_ID_GENESIS && hw->phy_type == SK_PHY_XMAC) 2558 if (hw->chip_id == CHIP_ID_GENESIS && hw->phy_type == SK_PHY_XMAC)
2559 del_timer_sync(&skge->link_timer); 2559 del_timer_sync(&skge->link_timer);
2560 2560
2561 netif_poll_disable(dev); 2561 napi_disable(&skge->napi);
2562 netif_carrier_off(dev); 2562 netif_carrier_off(dev);
2563 2563
2564 spin_lock_irq(&hw->hw_lock); 2564 spin_lock_irq(&hw->hw_lock);
@@ -3044,14 +3044,13 @@ static void skge_tx_done(struct net_device *dev)
3044 } 3044 }
3045} 3045}
3046 3046
3047static int skge_poll(struct net_device *dev, int *budget) 3047static int skge_poll(struct napi_struct *napi, int to_do)
3048{ 3048{
3049 struct skge_port *skge = netdev_priv(dev); 3049 struct skge_port *skge = container_of(napi, struct skge_port, napi);
3050 struct net_device *dev = skge->netdev;
3050 struct skge_hw *hw = skge->hw; 3051 struct skge_hw *hw = skge->hw;
3051 struct skge_ring *ring = &skge->rx_ring; 3052 struct skge_ring *ring = &skge->rx_ring;
3052 struct skge_element *e; 3053 struct skge_element *e;
3053 unsigned long flags;
3054 int to_do = min(dev->quota, *budget);
3055 int work_done = 0; 3054 int work_done = 0;
3056 3055
3057 skge_tx_done(dev); 3056 skge_tx_done(dev);
@@ -3082,20 +3081,16 @@ static int skge_poll(struct net_device *dev, int *budget)
3082 wmb(); 3081 wmb();
3083 skge_write8(hw, Q_ADDR(rxqaddr[skge->port], Q_CSR), CSR_START); 3082 skge_write8(hw, Q_ADDR(rxqaddr[skge->port], Q_CSR), CSR_START);
3084 3083
3085 *budget -= work_done; 3084 if (work_done < to_do) {
3086 dev->quota -= work_done; 3085 spin_lock_irq(&hw->hw_lock);
3087 3086 __netif_rx_complete(dev, napi);
3088 if (work_done >= to_do) 3087 hw->intr_mask |= napimask[skge->port];
3089 return 1; /* not done */ 3088 skge_write32(hw, B0_IMSK, hw->intr_mask);
3090 3089 skge_read32(hw, B0_IMSK);
3091 spin_lock_irqsave(&hw->hw_lock, flags); 3090 spin_unlock_irq(&hw->hw_lock);
3092 __netif_rx_complete(dev); 3091 }
3093 hw->intr_mask |= napimask[skge->port];
3094 skge_write32(hw, B0_IMSK, hw->intr_mask);
3095 skge_read32(hw, B0_IMSK);
3096 spin_unlock_irqrestore(&hw->hw_lock, flags);
3097 3092
3098 return 0; 3093 return work_done;
3099} 3094}
3100 3095
3101/* Parity errors seem to happen when Genesis is connected to a switch 3096/* Parity errors seem to happen when Genesis is connected to a switch
@@ -3252,8 +3247,9 @@ static irqreturn_t skge_intr(int irq, void *dev_id)
3252 } 3247 }
3253 3248
3254 if (status & (IS_XA1_F|IS_R1_F)) { 3249 if (status & (IS_XA1_F|IS_R1_F)) {
3250 struct skge_port *skge = netdev_priv(hw->dev[0]);
3255 hw->intr_mask &= ~(IS_XA1_F|IS_R1_F); 3251 hw->intr_mask &= ~(IS_XA1_F|IS_R1_F);
3256 netif_rx_schedule(hw->dev[0]); 3252 netif_rx_schedule(hw->dev[0], &skge->napi);
3257 } 3253 }
3258 3254
3259 if (status & IS_PA_TO_TX1) 3255 if (status & IS_PA_TO_TX1)
@@ -3271,13 +3267,14 @@ static irqreturn_t skge_intr(int irq, void *dev_id)
3271 skge_mac_intr(hw, 0); 3267 skge_mac_intr(hw, 0);
3272 3268
3273 if (hw->dev[1]) { 3269 if (hw->dev[1]) {
3270 struct skge_port *skge = netdev_priv(hw->dev[1]);
3271
3274 if (status & (IS_XA2_F|IS_R2_F)) { 3272 if (status & (IS_XA2_F|IS_R2_F)) {
3275 hw->intr_mask &= ~(IS_XA2_F|IS_R2_F); 3273 hw->intr_mask &= ~(IS_XA2_F|IS_R2_F);
3276 netif_rx_schedule(hw->dev[1]); 3274 netif_rx_schedule(hw->dev[1], &skge->napi);
3277 } 3275 }
3278 3276
3279 if (status & IS_PA_TO_RX2) { 3277 if (status & IS_PA_TO_RX2) {
3280 struct skge_port *skge = netdev_priv(hw->dev[1]);
3281 ++skge->net_stats.rx_over_errors; 3278 ++skge->net_stats.rx_over_errors;
3282 skge_write16(hw, B3_PA_CTRL, PA_CLR_TO_RX2); 3279 skge_write16(hw, B3_PA_CTRL, PA_CLR_TO_RX2);
3283 } 3280 }
@@ -3569,8 +3566,6 @@ static struct net_device *skge_devinit(struct skge_hw *hw, int port,
3569 SET_ETHTOOL_OPS(dev, &skge_ethtool_ops); 3566 SET_ETHTOOL_OPS(dev, &skge_ethtool_ops);
3570 dev->tx_timeout = skge_tx_timeout; 3567 dev->tx_timeout = skge_tx_timeout;
3571 dev->watchdog_timeo = TX_WATCHDOG; 3568 dev->watchdog_timeo = TX_WATCHDOG;
3572 dev->poll = skge_poll;
3573 dev->weight = NAPI_WEIGHT;
3574#ifdef CONFIG_NET_POLL_CONTROLLER 3569#ifdef CONFIG_NET_POLL_CONTROLLER
3575 dev->poll_controller = skge_netpoll; 3570 dev->poll_controller = skge_netpoll;
3576#endif 3571#endif
@@ -3580,6 +3575,7 @@ static struct net_device *skge_devinit(struct skge_hw *hw, int port,
3580 dev->features |= NETIF_F_HIGHDMA; 3575 dev->features |= NETIF_F_HIGHDMA;
3581 3576
3582 skge = netdev_priv(dev); 3577 skge = netdev_priv(dev);
3578 netif_napi_add(dev, &skge->napi, skge_poll, NAPI_WEIGHT);
3583 skge->netdev = dev; 3579 skge->netdev = dev;
3584 skge->hw = hw; 3580 skge->hw = hw;
3585 skge->msg_enable = netif_msg_init(debug, default_msg); 3581 skge->msg_enable = netif_msg_init(debug, default_msg);
diff --git a/drivers/net/skge.h b/drivers/net/skge.h
index edd71468220c..dd0fd45c7155 100644
--- a/drivers/net/skge.h
+++ b/drivers/net/skge.h
@@ -2448,6 +2448,7 @@ enum pause_status {
2448struct skge_port { 2448struct skge_port {
2449 struct skge_hw *hw; 2449 struct skge_hw *hw;
2450 struct net_device *netdev; 2450 struct net_device *netdev;
2451 struct napi_struct napi;
2451 int port; 2452 int port;
2452 u32 msg_enable; 2453 u32 msg_enable;
2453 2454
diff --git a/drivers/net/sky2.c b/drivers/net/sky2.c
index ea117fc3d5e3..a0d75b0f3798 100644
--- a/drivers/net/sky2.c
+++ b/drivers/net/sky2.c
@@ -1130,7 +1130,7 @@ static void sky2_vlan_rx_register(struct net_device *dev, struct vlan_group *grp
1130 u16 port = sky2->port; 1130 u16 port = sky2->port;
1131 1131
1132 netif_tx_lock_bh(dev); 1132 netif_tx_lock_bh(dev);
1133 netif_poll_disable(sky2->hw->dev[0]); 1133 napi_disable(&hw->napi);
1134 1134
1135 sky2->vlgrp = grp; 1135 sky2->vlgrp = grp;
1136 if (grp) { 1136 if (grp) {
@@ -1145,7 +1145,7 @@ static void sky2_vlan_rx_register(struct net_device *dev, struct vlan_group *grp
1145 TX_VLAN_TAG_OFF); 1145 TX_VLAN_TAG_OFF);
1146 } 1146 }
1147 1147
1148 netif_poll_enable(sky2->hw->dev[0]); 1148 napi_enable(&hw->napi);
1149 netif_tx_unlock_bh(dev); 1149 netif_tx_unlock_bh(dev);
1150} 1150}
1151#endif 1151#endif
@@ -1385,9 +1385,13 @@ static int sky2_up(struct net_device *dev)
1385 sky2_prefetch_init(hw, txqaddr[port], sky2->tx_le_map, 1385 sky2_prefetch_init(hw, txqaddr[port], sky2->tx_le_map,
1386 TX_RING_SIZE - 1); 1386 TX_RING_SIZE - 1);
1387 1387
1388 napi_enable(&hw->napi);
1389
1388 err = sky2_rx_start(sky2); 1390 err = sky2_rx_start(sky2);
1389 if (err) 1391 if (err) {
1392 napi_disable(&hw->napi);
1390 goto err_out; 1393 goto err_out;
1394 }
1391 1395
1392 /* Enable interrupts from phy/mac for port */ 1396 /* Enable interrupts from phy/mac for port */
1393 imask = sky2_read32(hw, B0_IMSK); 1397 imask = sky2_read32(hw, B0_IMSK);
@@ -1676,6 +1680,8 @@ static int sky2_down(struct net_device *dev)
1676 /* Stop more packets from being queued */ 1680 /* Stop more packets from being queued */
1677 netif_stop_queue(dev); 1681 netif_stop_queue(dev);
1678 1682
1683 napi_disable(&hw->napi);
1684
1679 /* Disable port IRQ */ 1685 /* Disable port IRQ */
1680 imask = sky2_read32(hw, B0_IMSK); 1686 imask = sky2_read32(hw, B0_IMSK);
1681 imask &= ~portirq_msk[port]; 1687 imask &= ~portirq_msk[port];
@@ -2016,7 +2022,7 @@ static int sky2_change_mtu(struct net_device *dev, int new_mtu)
2016 2022
2017 dev->trans_start = jiffies; /* prevent tx timeout */ 2023 dev->trans_start = jiffies; /* prevent tx timeout */
2018 netif_stop_queue(dev); 2024 netif_stop_queue(dev);
2019 netif_poll_disable(hw->dev[0]); 2025 napi_disable(&hw->napi);
2020 2026
2021 synchronize_irq(hw->pdev->irq); 2027 synchronize_irq(hw->pdev->irq);
2022 2028
@@ -2043,12 +2049,16 @@ static int sky2_change_mtu(struct net_device *dev, int new_mtu)
2043 err = sky2_rx_start(sky2); 2049 err = sky2_rx_start(sky2);
2044 sky2_write32(hw, B0_IMSK, imask); 2050 sky2_write32(hw, B0_IMSK, imask);
2045 2051
2052 /* Unconditionally re-enable NAPI because even if we
2053 * call dev_close() that will do a napi_disable().
2054 */
2055 napi_enable(&hw->napi);
2056
2046 if (err) 2057 if (err)
2047 dev_close(dev); 2058 dev_close(dev);
2048 else { 2059 else {
2049 gma_write16(hw, port, GM_GP_CTRL, ctl); 2060 gma_write16(hw, port, GM_GP_CTRL, ctl);
2050 2061
2051 netif_poll_enable(hw->dev[0]);
2052 netif_wake_queue(dev); 2062 netif_wake_queue(dev);
2053 } 2063 }
2054 2064
@@ -2544,18 +2554,15 @@ static int sky2_rx_hung(struct net_device *dev)
2544static void sky2_watchdog(unsigned long arg) 2554static void sky2_watchdog(unsigned long arg)
2545{ 2555{
2546 struct sky2_hw *hw = (struct sky2_hw *) arg; 2556 struct sky2_hw *hw = (struct sky2_hw *) arg;
2547 struct net_device *dev;
2548 2557
2549 /* Check for lost IRQ once a second */ 2558 /* Check for lost IRQ once a second */
2550 if (sky2_read32(hw, B0_ISRC)) { 2559 if (sky2_read32(hw, B0_ISRC)) {
2551 dev = hw->dev[0]; 2560 napi_schedule(&hw->napi);
2552 if (__netif_rx_schedule_prep(dev))
2553 __netif_rx_schedule(dev);
2554 } else { 2561 } else {
2555 int i, active = 0; 2562 int i, active = 0;
2556 2563
2557 for (i = 0; i < hw->ports; i++) { 2564 for (i = 0; i < hw->ports; i++) {
2558 dev = hw->dev[i]; 2565 struct net_device *dev = hw->dev[i];
2559 if (!netif_running(dev)) 2566 if (!netif_running(dev))
2560 continue; 2567 continue;
2561 ++active; 2568 ++active;
@@ -2605,11 +2612,11 @@ static void sky2_err_intr(struct sky2_hw *hw, u32 status)
2605 sky2_le_error(hw, 1, Q_XA2, TX_RING_SIZE); 2612 sky2_le_error(hw, 1, Q_XA2, TX_RING_SIZE);
2606} 2613}
2607 2614
2608static int sky2_poll(struct net_device *dev0, int *budget) 2615static int sky2_poll(struct napi_struct *napi, int work_limit)
2609{ 2616{
2610 struct sky2_hw *hw = ((struct sky2_port *) netdev_priv(dev0))->hw; 2617 struct sky2_hw *hw = container_of(napi, struct sky2_hw, napi);
2611 int work_done;
2612 u32 status = sky2_read32(hw, B0_Y2_SP_EISR); 2618 u32 status = sky2_read32(hw, B0_Y2_SP_EISR);
2619 int work_done;
2613 2620
2614 if (unlikely(status & Y2_IS_ERROR)) 2621 if (unlikely(status & Y2_IS_ERROR))
2615 sky2_err_intr(hw, status); 2622 sky2_err_intr(hw, status);
@@ -2620,31 +2627,27 @@ static int sky2_poll(struct net_device *dev0, int *budget)
2620 if (status & Y2_IS_IRQ_PHY2) 2627 if (status & Y2_IS_IRQ_PHY2)
2621 sky2_phy_intr(hw, 1); 2628 sky2_phy_intr(hw, 1);
2622 2629
2623 work_done = sky2_status_intr(hw, min(dev0->quota, *budget)); 2630 work_done = sky2_status_intr(hw, work_limit);
2624 *budget -= work_done;
2625 dev0->quota -= work_done;
2626 2631
2627 /* More work? */ 2632 /* More work? */
2628 if (hw->st_idx != sky2_read16(hw, STAT_PUT_IDX)) 2633 if (hw->st_idx == sky2_read16(hw, STAT_PUT_IDX)) {
2629 return 1; 2634 /* Bug/Errata workaround?
2635 * Need to kick the TX irq moderation timer.
2636 */
2637 if (sky2_read8(hw, STAT_TX_TIMER_CTRL) == TIM_START) {
2638 sky2_write8(hw, STAT_TX_TIMER_CTRL, TIM_STOP);
2639 sky2_write8(hw, STAT_TX_TIMER_CTRL, TIM_START);
2640 }
2630 2641
2631 /* Bug/Errata workaround? 2642 napi_complete(napi);
2632 * Need to kick the TX irq moderation timer. 2643 sky2_read32(hw, B0_Y2_SP_LISR);
2633 */
2634 if (sky2_read8(hw, STAT_TX_TIMER_CTRL) == TIM_START) {
2635 sky2_write8(hw, STAT_TX_TIMER_CTRL, TIM_STOP);
2636 sky2_write8(hw, STAT_TX_TIMER_CTRL, TIM_START);
2637 } 2644 }
2638 netif_rx_complete(dev0); 2645 return work_done;
2639
2640 sky2_read32(hw, B0_Y2_SP_LISR);
2641 return 0;
2642} 2646}
2643 2647
2644static irqreturn_t sky2_intr(int irq, void *dev_id) 2648static irqreturn_t sky2_intr(int irq, void *dev_id)
2645{ 2649{
2646 struct sky2_hw *hw = dev_id; 2650 struct sky2_hw *hw = dev_id;
2647 struct net_device *dev0 = hw->dev[0];
2648 u32 status; 2651 u32 status;
2649 2652
2650 /* Reading this mask interrupts as side effect */ 2653 /* Reading this mask interrupts as side effect */
@@ -2653,8 +2656,8 @@ static irqreturn_t sky2_intr(int irq, void *dev_id)
2653 return IRQ_NONE; 2656 return IRQ_NONE;
2654 2657
2655 prefetch(&hw->st_le[hw->st_idx]); 2658 prefetch(&hw->st_le[hw->st_idx]);
2656 if (likely(__netif_rx_schedule_prep(dev0))) 2659
2657 __netif_rx_schedule(dev0); 2660 napi_schedule(&hw->napi);
2658 2661
2659 return IRQ_HANDLED; 2662 return IRQ_HANDLED;
2660} 2663}
@@ -2663,10 +2666,8 @@ static irqreturn_t sky2_intr(int irq, void *dev_id)
2663static void sky2_netpoll(struct net_device *dev) 2666static void sky2_netpoll(struct net_device *dev)
2664{ 2667{
2665 struct sky2_port *sky2 = netdev_priv(dev); 2668 struct sky2_port *sky2 = netdev_priv(dev);
2666 struct net_device *dev0 = sky2->hw->dev[0];
2667 2669
2668 if (netif_running(dev) && __netif_rx_schedule_prep(dev0)) 2670 napi_schedule(&sky2->hw->napi);
2669 __netif_rx_schedule(dev0);
2670} 2671}
2671#endif 2672#endif
2672 2673
@@ -2914,8 +2915,6 @@ static void sky2_restart(struct work_struct *work)
2914 sky2_write32(hw, B0_IMSK, 0); 2915 sky2_write32(hw, B0_IMSK, 0);
2915 sky2_read32(hw, B0_IMSK); 2916 sky2_read32(hw, B0_IMSK);
2916 2917
2917 netif_poll_disable(hw->dev[0]);
2918
2919 for (i = 0; i < hw->ports; i++) { 2918 for (i = 0; i < hw->ports; i++) {
2920 dev = hw->dev[i]; 2919 dev = hw->dev[i];
2921 if (netif_running(dev)) 2920 if (netif_running(dev))
@@ -2924,7 +2923,6 @@ static void sky2_restart(struct work_struct *work)
2924 2923
2925 sky2_reset(hw); 2924 sky2_reset(hw);
2926 sky2_write32(hw, B0_IMSK, Y2_IS_BASE); 2925 sky2_write32(hw, B0_IMSK, Y2_IS_BASE);
2927 netif_poll_enable(hw->dev[0]);
2928 2926
2929 for (i = 0; i < hw->ports; i++) { 2927 for (i = 0; i < hw->ports; i++) {
2930 dev = hw->dev[i]; 2928 dev = hw->dev[i];
@@ -3735,7 +3733,7 @@ static int sky2_debug_show(struct seq_file *seq, void *v)
3735{ 3733{
3736 struct net_device *dev = seq->private; 3734 struct net_device *dev = seq->private;
3737 const struct sky2_port *sky2 = netdev_priv(dev); 3735 const struct sky2_port *sky2 = netdev_priv(dev);
3738 const struct sky2_hw *hw = sky2->hw; 3736 struct sky2_hw *hw = sky2->hw;
3739 unsigned port = sky2->port; 3737 unsigned port = sky2->port;
3740 unsigned idx, last; 3738 unsigned idx, last;
3741 int sop; 3739 int sop;
@@ -3748,7 +3746,7 @@ static int sky2_debug_show(struct seq_file *seq, void *v)
3748 sky2_read32(hw, B0_IMSK), 3746 sky2_read32(hw, B0_IMSK),
3749 sky2_read32(hw, B0_Y2_SP_ICR)); 3747 sky2_read32(hw, B0_Y2_SP_ICR));
3750 3748
3751 netif_poll_disable(hw->dev[0]); 3749 napi_disable(&hw->napi);
3752 last = sky2_read16(hw, STAT_PUT_IDX); 3750 last = sky2_read16(hw, STAT_PUT_IDX);
3753 3751
3754 if (hw->st_idx == last) 3752 if (hw->st_idx == last)
@@ -3818,7 +3816,7 @@ static int sky2_debug_show(struct seq_file *seq, void *v)
3818 last = sky2_read16(hw, Y2_QADDR(rxqaddr[port], PREF_UNIT_PUT_IDX)), 3816 last = sky2_read16(hw, Y2_QADDR(rxqaddr[port], PREF_UNIT_PUT_IDX)),
3819 sky2_read16(hw, Y2_QADDR(rxqaddr[port], PREF_UNIT_LAST_IDX))); 3817 sky2_read16(hw, Y2_QADDR(rxqaddr[port], PREF_UNIT_LAST_IDX)));
3820 3818
3821 netif_poll_enable(hw->dev[0]); 3819 napi_enable(&hw->napi);
3822 return 0; 3820 return 0;
3823} 3821}
3824 3822
@@ -3943,15 +3941,8 @@ static __devinit struct net_device *sky2_init_netdev(struct sky2_hw *hw,
3943 SET_ETHTOOL_OPS(dev, &sky2_ethtool_ops); 3941 SET_ETHTOOL_OPS(dev, &sky2_ethtool_ops);
3944 dev->tx_timeout = sky2_tx_timeout; 3942 dev->tx_timeout = sky2_tx_timeout;
3945 dev->watchdog_timeo = TX_WATCHDOG; 3943 dev->watchdog_timeo = TX_WATCHDOG;
3946 if (port == 0)
3947 dev->poll = sky2_poll;
3948 dev->weight = NAPI_WEIGHT;
3949#ifdef CONFIG_NET_POLL_CONTROLLER 3944#ifdef CONFIG_NET_POLL_CONTROLLER
3950 /* Network console (only works on port 0) 3945 dev->poll_controller = sky2_netpoll;
3951 * because netpoll makes assumptions about NAPI
3952 */
3953 if (port == 0)
3954 dev->poll_controller = sky2_netpoll;
3955#endif 3946#endif
3956 3947
3957 sky2 = netdev_priv(dev); 3948 sky2 = netdev_priv(dev);
@@ -4166,6 +4157,7 @@ static int __devinit sky2_probe(struct pci_dev *pdev,
4166 err = -ENOMEM; 4157 err = -ENOMEM;
4167 goto err_out_free_pci; 4158 goto err_out_free_pci;
4168 } 4159 }
4160 netif_napi_add(dev, &hw->napi, sky2_poll, NAPI_WEIGHT);
4169 4161
4170 if (!disable_msi && pci_enable_msi(pdev) == 0) { 4162 if (!disable_msi && pci_enable_msi(pdev) == 0) {
4171 err = sky2_test_msi(hw); 4163 err = sky2_test_msi(hw);
@@ -4288,8 +4280,6 @@ static int sky2_suspend(struct pci_dev *pdev, pm_message_t state)
4288 if (!hw) 4280 if (!hw)
4289 return 0; 4281 return 0;
4290 4282
4291 netif_poll_disable(hw->dev[0]);
4292
4293 for (i = 0; i < hw->ports; i++) { 4283 for (i = 0; i < hw->ports; i++) {
4294 struct net_device *dev = hw->dev[i]; 4284 struct net_device *dev = hw->dev[i];
4295 struct sky2_port *sky2 = netdev_priv(dev); 4285 struct sky2_port *sky2 = netdev_priv(dev);
@@ -4356,8 +4346,6 @@ static int sky2_resume(struct pci_dev *pdev)
4356 } 4346 }
4357 } 4347 }
4358 4348
4359 netif_poll_enable(hw->dev[0]);
4360
4361 return 0; 4349 return 0;
4362out: 4350out:
4363 dev_err(&pdev->dev, "resume failed (%d)\n", err); 4351 dev_err(&pdev->dev, "resume failed (%d)\n", err);
@@ -4374,7 +4362,7 @@ static void sky2_shutdown(struct pci_dev *pdev)
4374 if (!hw) 4362 if (!hw)
4375 return; 4363 return;
4376 4364
4377 netif_poll_disable(hw->dev[0]); 4365 napi_disable(&hw->napi);
4378 4366
4379 for (i = 0; i < hw->ports; i++) { 4367 for (i = 0; i < hw->ports; i++) {
4380 struct net_device *dev = hw->dev[i]; 4368 struct net_device *dev = hw->dev[i];
diff --git a/drivers/net/sky2.h b/drivers/net/sky2.h
index 8bc5c54e3efa..f18f8752118e 100644
--- a/drivers/net/sky2.h
+++ b/drivers/net/sky2.h
@@ -2057,6 +2057,7 @@ struct sky2_port {
2057struct sky2_hw { 2057struct sky2_hw {
2058 void __iomem *regs; 2058 void __iomem *regs;
2059 struct pci_dev *pdev; 2059 struct pci_dev *pdev;
2060 struct napi_struct napi;
2060 struct net_device *dev[2]; 2061 struct net_device *dev[2];
2061 unsigned long flags; 2062 unsigned long flags;
2062#define SKY2_HW_USE_MSI 0x00000001 2063#define SKY2_HW_USE_MSI 0x00000001
diff --git a/drivers/net/spider_net.c b/drivers/net/spider_net.c
index 82d837ab4db9..6d8f2bb7e0f9 100644
--- a/drivers/net/spider_net.c
+++ b/drivers/net/spider_net.c
@@ -1278,34 +1278,26 @@ bad_desc:
1278 * (using netif_receive_skb). If all/enough packets are up, the driver 1278 * (using netif_receive_skb). If all/enough packets are up, the driver
1279 * reenables interrupts and returns 0. If not, 1 is returned. 1279 * reenables interrupts and returns 0. If not, 1 is returned.
1280 */ 1280 */
1281static int 1281static int spider_net_poll(struct napi_struct *napi, int budget)
1282spider_net_poll(struct net_device *netdev, int *budget)
1283{ 1282{
1284 struct spider_net_card *card = netdev_priv(netdev); 1283 struct spider_net_card *card = container_of(napi, struct spider_net_card, napi);
1285 int packets_to_do, packets_done = 0; 1284 struct net_device *netdev = card->netdev;
1286 int no_more_packets = 0; 1285 int packets_done = 0;
1287 1286
1288 packets_to_do = min(*budget, netdev->quota); 1287 while (packets_done < budget) {
1289 1288 if (!spider_net_decode_one_descr(card))
1290 while (packets_to_do) {
1291 if (spider_net_decode_one_descr(card)) {
1292 packets_done++;
1293 packets_to_do--;
1294 } else {
1295 /* no more packets for the stack */
1296 no_more_packets = 1;
1297 break; 1289 break;
1298 } 1290
1291 packets_done++;
1299 } 1292 }
1300 1293
1301 if ((packets_done == 0) && (card->num_rx_ints != 0)) { 1294 if ((packets_done == 0) && (card->num_rx_ints != 0)) {
1302 no_more_packets = spider_net_resync_tail_ptr(card); 1295 if (!spider_net_resync_tail_ptr(card))
1296 packets_done = budget;
1303 spider_net_resync_head_ptr(card); 1297 spider_net_resync_head_ptr(card);
1304 } 1298 }
1305 card->num_rx_ints = 0; 1299 card->num_rx_ints = 0;
1306 1300
1307 netdev->quota -= packets_done;
1308 *budget -= packets_done;
1309 spider_net_refill_rx_chain(card); 1301 spider_net_refill_rx_chain(card);
1310 spider_net_enable_rxdmac(card); 1302 spider_net_enable_rxdmac(card);
1311 1303
@@ -1313,14 +1305,13 @@ spider_net_poll(struct net_device *netdev, int *budget)
1313 1305
1314 /* if all packets are in the stack, enable interrupts and return 0 */ 1306 /* if all packets are in the stack, enable interrupts and return 0 */
1315 /* if not, return 1 */ 1307 /* if not, return 1 */
1316 if (no_more_packets) { 1308 if (packets_done < budget) {
1317 netif_rx_complete(netdev); 1309 netif_rx_complete(netdev, napi);
1318 spider_net_rx_irq_on(card); 1310 spider_net_rx_irq_on(card);
1319 card->ignore_rx_ramfull = 0; 1311 card->ignore_rx_ramfull = 0;
1320 return 0;
1321 } 1312 }
1322 1313
1323 return 1; 1314 return packets_done;
1324} 1315}
1325 1316
1326/** 1317/**
@@ -1560,7 +1551,8 @@ spider_net_handle_error_irq(struct spider_net_card *card, u32 status_reg)
1560 spider_net_refill_rx_chain(card); 1551 spider_net_refill_rx_chain(card);
1561 spider_net_enable_rxdmac(card); 1552 spider_net_enable_rxdmac(card);
1562 card->num_rx_ints ++; 1553 card->num_rx_ints ++;
1563 netif_rx_schedule(card->netdev); 1554 netif_rx_schedule(card->netdev,
1555 &card->napi);
1564 } 1556 }
1565 show_error = 0; 1557 show_error = 0;
1566 break; 1558 break;
@@ -1580,7 +1572,8 @@ spider_net_handle_error_irq(struct spider_net_card *card, u32 status_reg)
1580 spider_net_refill_rx_chain(card); 1572 spider_net_refill_rx_chain(card);
1581 spider_net_enable_rxdmac(card); 1573 spider_net_enable_rxdmac(card);
1582 card->num_rx_ints ++; 1574 card->num_rx_ints ++;
1583 netif_rx_schedule(card->netdev); 1575 netif_rx_schedule(card->netdev,
1576 &card->napi);
1584 show_error = 0; 1577 show_error = 0;
1585 break; 1578 break;
1586 1579
@@ -1594,7 +1587,8 @@ spider_net_handle_error_irq(struct spider_net_card *card, u32 status_reg)
1594 spider_net_refill_rx_chain(card); 1587 spider_net_refill_rx_chain(card);
1595 spider_net_enable_rxdmac(card); 1588 spider_net_enable_rxdmac(card);
1596 card->num_rx_ints ++; 1589 card->num_rx_ints ++;
1597 netif_rx_schedule(card->netdev); 1590 netif_rx_schedule(card->netdev,
1591 &card->napi);
1598 show_error = 0; 1592 show_error = 0;
1599 break; 1593 break;
1600 1594
@@ -1686,11 +1680,11 @@ spider_net_interrupt(int irq, void *ptr)
1686 1680
1687 if (status_reg & SPIDER_NET_RXINT ) { 1681 if (status_reg & SPIDER_NET_RXINT ) {
1688 spider_net_rx_irq_off(card); 1682 spider_net_rx_irq_off(card);
1689 netif_rx_schedule(netdev); 1683 netif_rx_schedule(netdev, &card->napi);
1690 card->num_rx_ints ++; 1684 card->num_rx_ints ++;
1691 } 1685 }
1692 if (status_reg & SPIDER_NET_TXINT) 1686 if (status_reg & SPIDER_NET_TXINT)
1693 netif_rx_schedule(netdev); 1687 netif_rx_schedule(netdev, &card->napi);
1694 1688
1695 if (status_reg & SPIDER_NET_LINKINT) 1689 if (status_reg & SPIDER_NET_LINKINT)
1696 spider_net_link_reset(netdev); 1690 spider_net_link_reset(netdev);
@@ -2034,7 +2028,7 @@ spider_net_open(struct net_device *netdev)
2034 2028
2035 netif_start_queue(netdev); 2029 netif_start_queue(netdev);
2036 netif_carrier_on(netdev); 2030 netif_carrier_on(netdev);
2037 netif_poll_enable(netdev); 2031 napi_enable(&card->napi);
2038 2032
2039 spider_net_enable_interrupts(card); 2033 spider_net_enable_interrupts(card);
2040 2034
@@ -2204,7 +2198,7 @@ spider_net_stop(struct net_device *netdev)
2204{ 2198{
2205 struct spider_net_card *card = netdev_priv(netdev); 2199 struct spider_net_card *card = netdev_priv(netdev);
2206 2200
2207 netif_poll_disable(netdev); 2201 napi_disable(&card->napi);
2208 netif_carrier_off(netdev); 2202 netif_carrier_off(netdev);
2209 netif_stop_queue(netdev); 2203 netif_stop_queue(netdev);
2210 del_timer_sync(&card->tx_timer); 2204 del_timer_sync(&card->tx_timer);
@@ -2304,9 +2298,6 @@ spider_net_setup_netdev_ops(struct net_device *netdev)
2304 /* tx watchdog */ 2298 /* tx watchdog */
2305 netdev->tx_timeout = &spider_net_tx_timeout; 2299 netdev->tx_timeout = &spider_net_tx_timeout;
2306 netdev->watchdog_timeo = SPIDER_NET_WATCHDOG_TIMEOUT; 2300 netdev->watchdog_timeo = SPIDER_NET_WATCHDOG_TIMEOUT;
2307 /* NAPI */
2308 netdev->poll = &spider_net_poll;
2309 netdev->weight = SPIDER_NET_NAPI_WEIGHT;
2310 /* HW VLAN */ 2301 /* HW VLAN */
2311#ifdef CONFIG_NET_POLL_CONTROLLER 2302#ifdef CONFIG_NET_POLL_CONTROLLER
2312 /* poll controller */ 2303 /* poll controller */
@@ -2351,6 +2342,9 @@ spider_net_setup_netdev(struct spider_net_card *card)
2351 2342
2352 card->options.rx_csum = SPIDER_NET_RX_CSUM_DEFAULT; 2343 card->options.rx_csum = SPIDER_NET_RX_CSUM_DEFAULT;
2353 2344
2345 netif_napi_add(netdev, &card->napi,
2346 spider_net_poll, SPIDER_NET_NAPI_WEIGHT);
2347
2354 spider_net_setup_netdev_ops(netdev); 2348 spider_net_setup_netdev_ops(netdev);
2355 2349
2356 netdev->features = NETIF_F_IP_CSUM | NETIF_F_LLTX; 2350 netdev->features = NETIF_F_IP_CSUM | NETIF_F_LLTX;
diff --git a/drivers/net/spider_net.h b/drivers/net/spider_net.h
index dbbdb8cee3c6..a2fcdebc3790 100644
--- a/drivers/net/spider_net.h
+++ b/drivers/net/spider_net.h
@@ -466,6 +466,8 @@ struct spider_net_card {
466 struct pci_dev *pdev; 466 struct pci_dev *pdev;
467 struct mii_phy phy; 467 struct mii_phy phy;
468 468
469 struct napi_struct napi;
470
469 int medium; 471 int medium;
470 472
471 void __iomem *regs; 473 void __iomem *regs;
diff --git a/drivers/net/starfire.c b/drivers/net/starfire.c
index 8b6478663a56..3b9336c34206 100644
--- a/drivers/net/starfire.c
+++ b/drivers/net/starfire.c
@@ -178,16 +178,13 @@ static int full_duplex[MAX_UNITS] = {0, };
178#define skb_num_frags(skb) (skb_shinfo(skb)->nr_frags + 1) 178#define skb_num_frags(skb) (skb_shinfo(skb)->nr_frags + 1)
179 179
180#ifdef HAVE_NETDEV_POLL 180#ifdef HAVE_NETDEV_POLL
181#define init_poll(dev) \ 181#define init_poll(dev, np) \
182do { \ 182 netif_napi_add(dev, &np->napi, netdev_poll, max_interrupt_work)
183 dev->poll = &netdev_poll; \ 183#define netdev_rx(dev, np, ioaddr) \
184 dev->weight = max_interrupt_work; \
185} while (0)
186#define netdev_rx(dev, ioaddr) \
187do { \ 184do { \
188 u32 intr_enable; \ 185 u32 intr_enable; \
189 if (netif_rx_schedule_prep(dev)) { \ 186 if (netif_rx_schedule_prep(dev, &np->napi)) { \
190 __netif_rx_schedule(dev); \ 187 __netif_rx_schedule(dev, &np->napi); \
191 intr_enable = readl(ioaddr + IntrEnable); \ 188 intr_enable = readl(ioaddr + IntrEnable); \
192 intr_enable &= ~(IntrRxDone | IntrRxEmpty); \ 189 intr_enable &= ~(IntrRxDone | IntrRxEmpty); \
193 writel(intr_enable, ioaddr + IntrEnable); \ 190 writel(intr_enable, ioaddr + IntrEnable); \
@@ -204,12 +201,12 @@ do { \
204} while (0) 201} while (0)
205#define netdev_receive_skb(skb) netif_receive_skb(skb) 202#define netdev_receive_skb(skb) netif_receive_skb(skb)
206#define vlan_netdev_receive_skb(skb, vlgrp, vlid) vlan_hwaccel_receive_skb(skb, vlgrp, vlid) 203#define vlan_netdev_receive_skb(skb, vlgrp, vlid) vlan_hwaccel_receive_skb(skb, vlgrp, vlid)
207static int netdev_poll(struct net_device *dev, int *budget); 204static int netdev_poll(struct napi_struct *napi, int budget);
208#else /* not HAVE_NETDEV_POLL */ 205#else /* not HAVE_NETDEV_POLL */
209#define init_poll(dev) 206#define init_poll(dev, np)
210#define netdev_receive_skb(skb) netif_rx(skb) 207#define netdev_receive_skb(skb) netif_rx(skb)
211#define vlan_netdev_receive_skb(skb, vlgrp, vlid) vlan_hwaccel_rx(skb, vlgrp, vlid) 208#define vlan_netdev_receive_skb(skb, vlgrp, vlid) vlan_hwaccel_rx(skb, vlgrp, vlid)
212#define netdev_rx(dev, ioaddr) \ 209#define netdev_rx(dev, np, ioaddr) \
213do { \ 210do { \
214 int quota = np->dirty_rx + RX_RING_SIZE - np->cur_rx; \ 211 int quota = np->dirty_rx + RX_RING_SIZE - np->cur_rx; \
215 __netdev_rx(dev, &quota);\ 212 __netdev_rx(dev, &quota);\
@@ -599,6 +596,8 @@ struct netdev_private {
599 struct tx_done_desc *tx_done_q; 596 struct tx_done_desc *tx_done_q;
600 dma_addr_t tx_done_q_dma; 597 dma_addr_t tx_done_q_dma;
601 unsigned int tx_done; 598 unsigned int tx_done;
599 struct napi_struct napi;
600 struct net_device *dev;
602 struct net_device_stats stats; 601 struct net_device_stats stats;
603 struct pci_dev *pci_dev; 602 struct pci_dev *pci_dev;
604#ifdef VLAN_SUPPORT 603#ifdef VLAN_SUPPORT
@@ -791,6 +790,7 @@ static int __devinit starfire_init_one(struct pci_dev *pdev,
791 dev->irq = irq; 790 dev->irq = irq;
792 791
793 np = netdev_priv(dev); 792 np = netdev_priv(dev);
793 np->dev = dev;
794 np->base = base; 794 np->base = base;
795 spin_lock_init(&np->lock); 795 spin_lock_init(&np->lock);
796 pci_set_drvdata(pdev, dev); 796 pci_set_drvdata(pdev, dev);
@@ -851,7 +851,7 @@ static int __devinit starfire_init_one(struct pci_dev *pdev,
851 dev->hard_start_xmit = &start_tx; 851 dev->hard_start_xmit = &start_tx;
852 dev->tx_timeout = tx_timeout; 852 dev->tx_timeout = tx_timeout;
853 dev->watchdog_timeo = TX_TIMEOUT; 853 dev->watchdog_timeo = TX_TIMEOUT;
854 init_poll(dev); 854 init_poll(dev, np);
855 dev->stop = &netdev_close; 855 dev->stop = &netdev_close;
856 dev->get_stats = &get_stats; 856 dev->get_stats = &get_stats;
857 dev->set_multicast_list = &set_rx_mode; 857 dev->set_multicast_list = &set_rx_mode;
@@ -1056,6 +1056,9 @@ static int netdev_open(struct net_device *dev)
1056 1056
1057 writel(np->intr_timer_ctrl, ioaddr + IntrTimerCtrl); 1057 writel(np->intr_timer_ctrl, ioaddr + IntrTimerCtrl);
1058 1058
1059#ifdef HAVE_NETDEV_POLL
1060 napi_enable(&np->napi);
1061#endif
1059 netif_start_queue(dev); 1062 netif_start_queue(dev);
1060 1063
1061 if (debug > 1) 1064 if (debug > 1)
@@ -1330,7 +1333,7 @@ static irqreturn_t intr_handler(int irq, void *dev_instance)
1330 handled = 1; 1333 handled = 1;
1331 1334
1332 if (intr_status & (IntrRxDone | IntrRxEmpty)) 1335 if (intr_status & (IntrRxDone | IntrRxEmpty))
1333 netdev_rx(dev, ioaddr); 1336 netdev_rx(dev, np, ioaddr);
1334 1337
1335 /* Scavenge the skbuff list based on the Tx-done queue. 1338 /* Scavenge the skbuff list based on the Tx-done queue.
1336 There are redundant checks here that may be cleaned up 1339 There are redundant checks here that may be cleaned up
@@ -1531,36 +1534,35 @@ static int __netdev_rx(struct net_device *dev, int *quota)
1531 1534
1532 1535
1533#ifdef HAVE_NETDEV_POLL 1536#ifdef HAVE_NETDEV_POLL
1534static int netdev_poll(struct net_device *dev, int *budget) 1537static int netdev_poll(struct napi_struct *napi, int budget)
1535{ 1538{
1539 struct netdev_private *np = container_of(napi, struct netdev_private, napi);
1540 struct net_device *dev = np->dev;
1536 u32 intr_status; 1541 u32 intr_status;
1537 struct netdev_private *np = netdev_priv(dev);
1538 void __iomem *ioaddr = np->base; 1542 void __iomem *ioaddr = np->base;
1539 int retcode = 0, quota = dev->quota; 1543 int quota = budget;
1540 1544
1541 do { 1545 do {
1542 writel(IntrRxDone | IntrRxEmpty, ioaddr + IntrClear); 1546 writel(IntrRxDone | IntrRxEmpty, ioaddr + IntrClear);
1543 1547
1544 retcode = __netdev_rx(dev, &quota); 1548 if (__netdev_rx(dev, &quota))
1545 *budget -= (dev->quota - quota);
1546 dev->quota = quota;
1547 if (retcode)
1548 goto out; 1549 goto out;
1549 1550
1550 intr_status = readl(ioaddr + IntrStatus); 1551 intr_status = readl(ioaddr + IntrStatus);
1551 } while (intr_status & (IntrRxDone | IntrRxEmpty)); 1552 } while (intr_status & (IntrRxDone | IntrRxEmpty));
1552 1553
1553 netif_rx_complete(dev); 1554 netif_rx_complete(dev, napi);
1554 intr_status = readl(ioaddr + IntrEnable); 1555 intr_status = readl(ioaddr + IntrEnable);
1555 intr_status |= IntrRxDone | IntrRxEmpty; 1556 intr_status |= IntrRxDone | IntrRxEmpty;
1556 writel(intr_status, ioaddr + IntrEnable); 1557 writel(intr_status, ioaddr + IntrEnable);
1557 1558
1558 out: 1559 out:
1559 if (debug > 5) 1560 if (debug > 5)
1560 printk(KERN_DEBUG " exiting netdev_poll(): %d.\n", retcode); 1561 printk(KERN_DEBUG " exiting netdev_poll(): %d.\n",
1562 budget - quota);
1561 1563
1562 /* Restart Rx engine if stopped. */ 1564 /* Restart Rx engine if stopped. */
1563 return retcode; 1565 return budget - quota;
1564} 1566}
1565#endif /* HAVE_NETDEV_POLL */ 1567#endif /* HAVE_NETDEV_POLL */
1566 1568
@@ -1904,6 +1906,9 @@ static int netdev_close(struct net_device *dev)
1904 int i; 1906 int i;
1905 1907
1906 netif_stop_queue(dev); 1908 netif_stop_queue(dev);
1909#ifdef HAVE_NETDEV_POLL
1910 napi_disable(&np->napi);
1911#endif
1907 1912
1908 if (debug > 1) { 1913 if (debug > 1) {
1909 printk(KERN_DEBUG "%s: Shutting down ethercard, Intr status %#8.8x.\n", 1914 printk(KERN_DEBUG "%s: Shutting down ethercard, Intr status %#8.8x.\n",
diff --git a/drivers/net/sungem.c b/drivers/net/sungem.c
index 432803855034..bf821e96f7b2 100644
--- a/drivers/net/sungem.c
+++ b/drivers/net/sungem.c
@@ -19,7 +19,7 @@
19 * 19 *
20 * gem_change_mtu() and gem_set_multicast() are called with a read_lock() 20 * gem_change_mtu() and gem_set_multicast() are called with a read_lock()
21 * help by net/core/dev.c, thus they can't schedule. That means they can't 21 * help by net/core/dev.c, thus they can't schedule. That means they can't
22 * call netif_poll_disable() neither, thus force gem_poll() to keep a spinlock 22 * call napi_disable() neither, thus force gem_poll() to keep a spinlock
23 * where it could have been dropped. change_mtu especially would love also to 23 * where it could have been dropped. change_mtu especially would love also to
24 * be able to msleep instead of horrid locked delays when resetting the HW, 24 * be able to msleep instead of horrid locked delays when resetting the HW,
25 * but that read_lock() makes it impossible, unless I defer it's action to 25 * but that read_lock() makes it impossible, unless I defer it's action to
@@ -878,19 +878,20 @@ static int gem_rx(struct gem *gp, int work_to_do)
878 return work_done; 878 return work_done;
879} 879}
880 880
881static int gem_poll(struct net_device *dev, int *budget) 881static int gem_poll(struct napi_struct *napi, int budget)
882{ 882{
883 struct gem *gp = dev->priv; 883 struct gem *gp = container_of(napi, struct gem, napi);
884 struct net_device *dev = gp->dev;
884 unsigned long flags; 885 unsigned long flags;
886 int work_done;
885 887
886 /* 888 /*
887 * NAPI locking nightmare: See comment at head of driver 889 * NAPI locking nightmare: See comment at head of driver
888 */ 890 */
889 spin_lock_irqsave(&gp->lock, flags); 891 spin_lock_irqsave(&gp->lock, flags);
890 892
893 work_done = 0;
891 do { 894 do {
892 int work_to_do, work_done;
893
894 /* Handle anomalies */ 895 /* Handle anomalies */
895 if (gp->status & GREG_STAT_ABNORMAL) { 896 if (gp->status & GREG_STAT_ABNORMAL) {
896 if (gem_abnormal_irq(dev, gp, gp->status)) 897 if (gem_abnormal_irq(dev, gp, gp->status))
@@ -906,29 +907,25 @@ static int gem_poll(struct net_device *dev, int *budget)
906 907
907 /* Run RX thread. We don't use any locking here, 908 /* Run RX thread. We don't use any locking here,
908 * code willing to do bad things - like cleaning the 909 * code willing to do bad things - like cleaning the
909 * rx ring - must call netif_poll_disable(), which 910 * rx ring - must call napi_disable(), which
910 * schedule_timeout()'s if polling is already disabled. 911 * schedule_timeout()'s if polling is already disabled.
911 */ 912 */
912 work_to_do = min(*budget, dev->quota); 913 work_done += gem_rx(gp, budget);
913
914 work_done = gem_rx(gp, work_to_do);
915
916 *budget -= work_done;
917 dev->quota -= work_done;
918 914
919 if (work_done >= work_to_do) 915 if (work_done >= budget)
920 return 1; 916 return work_done;
921 917
922 spin_lock_irqsave(&gp->lock, flags); 918 spin_lock_irqsave(&gp->lock, flags);
923 919
924 gp->status = readl(gp->regs + GREG_STAT); 920 gp->status = readl(gp->regs + GREG_STAT);
925 } while (gp->status & GREG_STAT_NAPI); 921 } while (gp->status & GREG_STAT_NAPI);
926 922
927 __netif_rx_complete(dev); 923 __netif_rx_complete(dev, napi);
928 gem_enable_ints(gp); 924 gem_enable_ints(gp);
929 925
930 spin_unlock_irqrestore(&gp->lock, flags); 926 spin_unlock_irqrestore(&gp->lock, flags);
931 return 0; 927
928 return work_done;
932} 929}
933 930
934static irqreturn_t gem_interrupt(int irq, void *dev_id) 931static irqreturn_t gem_interrupt(int irq, void *dev_id)
@@ -946,17 +943,17 @@ static irqreturn_t gem_interrupt(int irq, void *dev_id)
946 943
947 spin_lock_irqsave(&gp->lock, flags); 944 spin_lock_irqsave(&gp->lock, flags);
948 945
949 if (netif_rx_schedule_prep(dev)) { 946 if (netif_rx_schedule_prep(dev, &gp->napi)) {
950 u32 gem_status = readl(gp->regs + GREG_STAT); 947 u32 gem_status = readl(gp->regs + GREG_STAT);
951 948
952 if (gem_status == 0) { 949 if (gem_status == 0) {
953 netif_poll_enable(dev); 950 napi_enable(&gp->napi);
954 spin_unlock_irqrestore(&gp->lock, flags); 951 spin_unlock_irqrestore(&gp->lock, flags);
955 return IRQ_NONE; 952 return IRQ_NONE;
956 } 953 }
957 gp->status = gem_status; 954 gp->status = gem_status;
958 gem_disable_ints(gp); 955 gem_disable_ints(gp);
959 __netif_rx_schedule(dev); 956 __netif_rx_schedule(dev, &gp->napi);
960 } 957 }
961 958
962 spin_unlock_irqrestore(&gp->lock, flags); 959 spin_unlock_irqrestore(&gp->lock, flags);
@@ -2284,7 +2281,7 @@ static void gem_reset_task(struct work_struct *work)
2284 2281
2285 mutex_lock(&gp->pm_mutex); 2282 mutex_lock(&gp->pm_mutex);
2286 2283
2287 netif_poll_disable(gp->dev); 2284 napi_disable(&gp->napi);
2288 2285
2289 spin_lock_irq(&gp->lock); 2286 spin_lock_irq(&gp->lock);
2290 spin_lock(&gp->tx_lock); 2287 spin_lock(&gp->tx_lock);
@@ -2307,7 +2304,7 @@ static void gem_reset_task(struct work_struct *work)
2307 spin_unlock(&gp->tx_lock); 2304 spin_unlock(&gp->tx_lock);
2308 spin_unlock_irq(&gp->lock); 2305 spin_unlock_irq(&gp->lock);
2309 2306
2310 netif_poll_enable(gp->dev); 2307 napi_enable(&gp->napi);
2311 2308
2312 mutex_unlock(&gp->pm_mutex); 2309 mutex_unlock(&gp->pm_mutex);
2313} 2310}
@@ -2324,6 +2321,8 @@ static int gem_open(struct net_device *dev)
2324 if (!gp->asleep) 2321 if (!gp->asleep)
2325 rc = gem_do_start(dev); 2322 rc = gem_do_start(dev);
2326 gp->opened = (rc == 0); 2323 gp->opened = (rc == 0);
2324 if (gp->opened)
2325 napi_enable(&gp->napi);
2327 2326
2328 mutex_unlock(&gp->pm_mutex); 2327 mutex_unlock(&gp->pm_mutex);
2329 2328
@@ -2334,9 +2333,7 @@ static int gem_close(struct net_device *dev)
2334{ 2333{
2335 struct gem *gp = dev->priv; 2334 struct gem *gp = dev->priv;
2336 2335
2337 /* Note: we don't need to call netif_poll_disable() here because 2336 napi_disable(&gp->napi);
2338 * our caller (dev_close) already did it for us
2339 */
2340 2337
2341 mutex_lock(&gp->pm_mutex); 2338 mutex_lock(&gp->pm_mutex);
2342 2339
@@ -2358,7 +2355,7 @@ static int gem_suspend(struct pci_dev *pdev, pm_message_t state)
2358 2355
2359 mutex_lock(&gp->pm_mutex); 2356 mutex_lock(&gp->pm_mutex);
2360 2357
2361 netif_poll_disable(dev); 2358 napi_disable(&gp->napi);
2362 2359
2363 printk(KERN_INFO "%s: suspending, WakeOnLan %s\n", 2360 printk(KERN_INFO "%s: suspending, WakeOnLan %s\n",
2364 dev->name, 2361 dev->name,
@@ -2482,7 +2479,7 @@ static int gem_resume(struct pci_dev *pdev)
2482 spin_unlock(&gp->tx_lock); 2479 spin_unlock(&gp->tx_lock);
2483 spin_unlock_irqrestore(&gp->lock, flags); 2480 spin_unlock_irqrestore(&gp->lock, flags);
2484 2481
2485 netif_poll_enable(dev); 2482 napi_enable(&gp->napi);
2486 2483
2487 mutex_unlock(&gp->pm_mutex); 2484 mutex_unlock(&gp->pm_mutex);
2488 2485
@@ -3121,8 +3118,7 @@ static int __devinit gem_init_one(struct pci_dev *pdev,
3121 dev->get_stats = gem_get_stats; 3118 dev->get_stats = gem_get_stats;
3122 dev->set_multicast_list = gem_set_multicast; 3119 dev->set_multicast_list = gem_set_multicast;
3123 dev->do_ioctl = gem_ioctl; 3120 dev->do_ioctl = gem_ioctl;
3124 dev->poll = gem_poll; 3121 netif_napi_add(dev, &gp->napi, gem_poll, 64);
3125 dev->weight = 64;
3126 dev->ethtool_ops = &gem_ethtool_ops; 3122 dev->ethtool_ops = &gem_ethtool_ops;
3127 dev->tx_timeout = gem_tx_timeout; 3123 dev->tx_timeout = gem_tx_timeout;
3128 dev->watchdog_timeo = 5 * HZ; 3124 dev->watchdog_timeo = 5 * HZ;
diff --git a/drivers/net/sungem.h b/drivers/net/sungem.h
index 58cf87c5751e..76d760acc9e2 100644
--- a/drivers/net/sungem.h
+++ b/drivers/net/sungem.h
@@ -993,6 +993,7 @@ struct gem {
993 u32 msg_enable; 993 u32 msg_enable;
994 u32 status; 994 u32 status;
995 995
996 struct napi_struct napi;
996 struct net_device_stats net_stats; 997 struct net_device_stats net_stats;
997 998
998 int tx_fifo_sz; 999 int tx_fifo_sz;
diff --git a/drivers/net/tc35815.c b/drivers/net/tc35815.c
index ec41469eee82..b5e0dff67230 100644
--- a/drivers/net/tc35815.c
+++ b/drivers/net/tc35815.c
@@ -414,6 +414,9 @@ enum tc35815_timer_state {
414struct tc35815_local { 414struct tc35815_local {
415 struct pci_dev *pci_dev; 415 struct pci_dev *pci_dev;
416 416
417 struct net_device *dev;
418 struct napi_struct napi;
419
417 /* statistics */ 420 /* statistics */
418 struct net_device_stats stats; 421 struct net_device_stats stats;
419 struct { 422 struct {
@@ -566,7 +569,7 @@ static int tc35815_send_packet(struct sk_buff *skb, struct net_device *dev);
566static irqreturn_t tc35815_interrupt(int irq, void *dev_id); 569static irqreturn_t tc35815_interrupt(int irq, void *dev_id);
567#ifdef TC35815_NAPI 570#ifdef TC35815_NAPI
568static int tc35815_rx(struct net_device *dev, int limit); 571static int tc35815_rx(struct net_device *dev, int limit);
569static int tc35815_poll(struct net_device *dev, int *budget); 572static int tc35815_poll(struct napi_struct *napi, int budget);
570#else 573#else
571static void tc35815_rx(struct net_device *dev); 574static void tc35815_rx(struct net_device *dev);
572#endif 575#endif
@@ -685,6 +688,7 @@ static int __devinit tc35815_init_one (struct pci_dev *pdev,
685 SET_MODULE_OWNER(dev); 688 SET_MODULE_OWNER(dev);
686 SET_NETDEV_DEV(dev, &pdev->dev); 689 SET_NETDEV_DEV(dev, &pdev->dev);
687 lp = dev->priv; 690 lp = dev->priv;
691 lp->dev = dev;
688 692
689 /* enable device (incl. PCI PM wakeup), and bus-mastering */ 693 /* enable device (incl. PCI PM wakeup), and bus-mastering */
690 rc = pci_enable_device (pdev); 694 rc = pci_enable_device (pdev);
@@ -738,8 +742,7 @@ static int __devinit tc35815_init_one (struct pci_dev *pdev,
738 dev->tx_timeout = tc35815_tx_timeout; 742 dev->tx_timeout = tc35815_tx_timeout;
739 dev->watchdog_timeo = TC35815_TX_TIMEOUT; 743 dev->watchdog_timeo = TC35815_TX_TIMEOUT;
740#ifdef TC35815_NAPI 744#ifdef TC35815_NAPI
741 dev->poll = tc35815_poll; 745 netif_napi_add(dev, &lp->napi, tc35815_poll, NAPI_WEIGHT);
742 dev->weight = NAPI_WEIGHT;
743#endif 746#endif
744#ifdef CONFIG_NET_POLL_CONTROLLER 747#ifdef CONFIG_NET_POLL_CONTROLLER
745 dev->poll_controller = tc35815_poll_controller; 748 dev->poll_controller = tc35815_poll_controller;
@@ -748,8 +751,6 @@ static int __devinit tc35815_init_one (struct pci_dev *pdev,
748 dev->irq = pdev->irq; 751 dev->irq = pdev->irq;
749 dev->base_addr = (unsigned long) ioaddr; 752 dev->base_addr = (unsigned long) ioaddr;
750 753
751 /* dev->priv/lp zeroed and aligned in alloc_etherdev */
752 lp = dev->priv;
753 spin_lock_init(&lp->lock); 754 spin_lock_init(&lp->lock);
754 lp->pci_dev = pdev; 755 lp->pci_dev = pdev;
755 lp->boardtype = ent->driver_data; 756 lp->boardtype = ent->driver_data;
@@ -1237,6 +1238,10 @@ tc35815_open(struct net_device *dev)
1237 return -EAGAIN; 1238 return -EAGAIN;
1238 } 1239 }
1239 1240
1241#ifdef TC35815_NAPI
1242 napi_enable(&lp->napi);
1243#endif
1244
1240 /* Reset the hardware here. Don't forget to set the station address. */ 1245 /* Reset the hardware here. Don't forget to set the station address. */
1241 spin_lock_irq(&lp->lock); 1246 spin_lock_irq(&lp->lock);
1242 tc35815_chip_init(dev); 1247 tc35815_chip_init(dev);
@@ -1436,6 +1441,7 @@ static int tc35815_do_interrupt(struct net_device *dev, u32 status)
1436static irqreturn_t tc35815_interrupt(int irq, void *dev_id) 1441static irqreturn_t tc35815_interrupt(int irq, void *dev_id)
1437{ 1442{
1438 struct net_device *dev = dev_id; 1443 struct net_device *dev = dev_id;
1444 struct tc35815_local *lp = netdev_priv(dev);
1439 struct tc35815_regs __iomem *tr = 1445 struct tc35815_regs __iomem *tr =
1440 (struct tc35815_regs __iomem *)dev->base_addr; 1446 (struct tc35815_regs __iomem *)dev->base_addr;
1441#ifdef TC35815_NAPI 1447#ifdef TC35815_NAPI
@@ -1444,8 +1450,8 @@ static irqreturn_t tc35815_interrupt(int irq, void *dev_id)
1444 if (!(dmactl & DMA_IntMask)) { 1450 if (!(dmactl & DMA_IntMask)) {
1445 /* disable interrupts */ 1451 /* disable interrupts */
1446 tc_writel(dmactl | DMA_IntMask, &tr->DMA_Ctl); 1452 tc_writel(dmactl | DMA_IntMask, &tr->DMA_Ctl);
1447 if (netif_rx_schedule_prep(dev)) 1453 if (netif_rx_schedule_prep(dev, &lp->napi))
1448 __netif_rx_schedule(dev); 1454 __netif_rx_schedule(dev, &lp->napi);
1449 else { 1455 else {
1450 printk(KERN_ERR "%s: interrupt taken in poll\n", 1456 printk(KERN_ERR "%s: interrupt taken in poll\n",
1451 dev->name); 1457 dev->name);
@@ -1726,13 +1732,12 @@ tc35815_rx(struct net_device *dev)
1726} 1732}
1727 1733
1728#ifdef TC35815_NAPI 1734#ifdef TC35815_NAPI
1729static int 1735static int tc35815_poll(struct napi_struct *napi, int budget)
1730tc35815_poll(struct net_device *dev, int *budget)
1731{ 1736{
1732 struct tc35815_local *lp = dev->priv; 1737 struct tc35815_local *lp = container_of(napi, struct tc35815_local, napi);
1738 struct net_device *dev = lp->dev;
1733 struct tc35815_regs __iomem *tr = 1739 struct tc35815_regs __iomem *tr =
1734 (struct tc35815_regs __iomem *)dev->base_addr; 1740 (struct tc35815_regs __iomem *)dev->base_addr;
1735 int limit = min(*budget, dev->quota);
1736 int received = 0, handled; 1741 int received = 0, handled;
1737 u32 status; 1742 u32 status;
1738 1743
@@ -1744,23 +1749,19 @@ tc35815_poll(struct net_device *dev, int *budget)
1744 handled = tc35815_do_interrupt(dev, status, limit); 1749 handled = tc35815_do_interrupt(dev, status, limit);
1745 if (handled >= 0) { 1750 if (handled >= 0) {
1746 received += handled; 1751 received += handled;
1747 limit -= handled; 1752 if (received >= budget)
1748 if (limit <= 0)
1749 break; 1753 break;
1750 } 1754 }
1751 status = tc_readl(&tr->Int_Src); 1755 status = tc_readl(&tr->Int_Src);
1752 } while (status); 1756 } while (status);
1753 spin_unlock(&lp->lock); 1757 spin_unlock(&lp->lock);
1754 1758
1755 dev->quota -= received; 1759 if (received < budget) {
1756 *budget -= received; 1760 netif_rx_complete(dev, napi);
1757 if (limit <= 0) 1761 /* enable interrupts */
1758 return 1; 1762 tc_writel(tc_readl(&tr->DMA_Ctl) & ~DMA_IntMask, &tr->DMA_Ctl);
1759 1763 }
1760 netif_rx_complete(dev); 1764 return received;
1761 /* enable interrupts */
1762 tc_writel(tc_readl(&tr->DMA_Ctl) & ~DMA_IntMask, &tr->DMA_Ctl);
1763 return 0;
1764} 1765}
1765#endif 1766#endif
1766 1767
@@ -1949,7 +1950,11 @@ static int
1949tc35815_close(struct net_device *dev) 1950tc35815_close(struct net_device *dev)
1950{ 1951{
1951 struct tc35815_local *lp = dev->priv; 1952 struct tc35815_local *lp = dev->priv;
1953
1952 netif_stop_queue(dev); 1954 netif_stop_queue(dev);
1955#ifdef TC35815_NAPI
1956 napi_disable(&lp->napi);
1957#endif
1953 1958
1954 /* Flush the Tx and disable Rx here. */ 1959 /* Flush the Tx and disable Rx here. */
1955 1960
diff --git a/drivers/net/tg3.c b/drivers/net/tg3.c
index 9034a05734ef..ef1e3d1173c4 100644
--- a/drivers/net/tg3.c
+++ b/drivers/net/tg3.c
@@ -574,7 +574,7 @@ static void tg3_restart_ints(struct tg3 *tp)
574static inline void tg3_netif_stop(struct tg3 *tp) 574static inline void tg3_netif_stop(struct tg3 *tp)
575{ 575{
576 tp->dev->trans_start = jiffies; /* prevent tx timeout */ 576 tp->dev->trans_start = jiffies; /* prevent tx timeout */
577 netif_poll_disable(tp->dev); 577 napi_disable(&tp->napi);
578 netif_tx_disable(tp->dev); 578 netif_tx_disable(tp->dev);
579} 579}
580 580
@@ -585,7 +585,7 @@ static inline void tg3_netif_start(struct tg3 *tp)
585 * so long as all callers are assured to have free tx slots 585 * so long as all callers are assured to have free tx slots
586 * (such as after tg3_init_hw) 586 * (such as after tg3_init_hw)
587 */ 587 */
588 netif_poll_enable(tp->dev); 588 napi_enable(&tp->napi);
589 tp->hw_status->status |= SD_STATUS_UPDATED; 589 tp->hw_status->status |= SD_STATUS_UPDATED;
590 tg3_enable_ints(tp); 590 tg3_enable_ints(tp);
591} 591}
@@ -3471,11 +3471,12 @@ next_pkt_nopost:
3471 return received; 3471 return received;
3472} 3472}
3473 3473
3474static int tg3_poll(struct net_device *netdev, int *budget) 3474static int tg3_poll(struct napi_struct *napi, int budget)
3475{ 3475{
3476 struct tg3 *tp = netdev_priv(netdev); 3476 struct tg3 *tp = container_of(napi, struct tg3, napi);
3477 struct net_device *netdev = tp->dev;
3477 struct tg3_hw_status *sblk = tp->hw_status; 3478 struct tg3_hw_status *sblk = tp->hw_status;
3478 int done; 3479 int work_done = 0;
3479 3480
3480 /* handle link change and other phy events */ 3481 /* handle link change and other phy events */
3481 if (!(tp->tg3_flags & 3482 if (!(tp->tg3_flags &
@@ -3494,7 +3495,7 @@ static int tg3_poll(struct net_device *netdev, int *budget)
3494 if (sblk->idx[0].tx_consumer != tp->tx_cons) { 3495 if (sblk->idx[0].tx_consumer != tp->tx_cons) {
3495 tg3_tx(tp); 3496 tg3_tx(tp);
3496 if (unlikely(tp->tg3_flags & TG3_FLAG_TX_RECOVERY_PENDING)) { 3497 if (unlikely(tp->tg3_flags & TG3_FLAG_TX_RECOVERY_PENDING)) {
3497 netif_rx_complete(netdev); 3498 netif_rx_complete(netdev, napi);
3498 schedule_work(&tp->reset_task); 3499 schedule_work(&tp->reset_task);
3499 return 0; 3500 return 0;
3500 } 3501 }
@@ -3502,20 +3503,10 @@ static int tg3_poll(struct net_device *netdev, int *budget)
3502 3503
3503 /* run RX thread, within the bounds set by NAPI. 3504 /* run RX thread, within the bounds set by NAPI.
3504 * All RX "locking" is done by ensuring outside 3505 * All RX "locking" is done by ensuring outside
3505 * code synchronizes with dev->poll() 3506 * code synchronizes with tg3->napi.poll()
3506 */ 3507 */
3507 if (sblk->idx[0].rx_producer != tp->rx_rcb_ptr) { 3508 if (sblk->idx[0].rx_producer != tp->rx_rcb_ptr)
3508 int orig_budget = *budget; 3509 work_done = tg3_rx(tp, budget);
3509 int work_done;
3510
3511 if (orig_budget > netdev->quota)
3512 orig_budget = netdev->quota;
3513
3514 work_done = tg3_rx(tp, orig_budget);
3515
3516 *budget -= work_done;
3517 netdev->quota -= work_done;
3518 }
3519 3510
3520 if (tp->tg3_flags & TG3_FLAG_TAGGED_STATUS) { 3511 if (tp->tg3_flags & TG3_FLAG_TAGGED_STATUS) {
3521 tp->last_tag = sblk->status_tag; 3512 tp->last_tag = sblk->status_tag;
@@ -3524,13 +3515,12 @@ static int tg3_poll(struct net_device *netdev, int *budget)
3524 sblk->status &= ~SD_STATUS_UPDATED; 3515 sblk->status &= ~SD_STATUS_UPDATED;
3525 3516
3526 /* if no more work, tell net stack and NIC we're done */ 3517 /* if no more work, tell net stack and NIC we're done */
3527 done = !tg3_has_work(tp); 3518 if (!tg3_has_work(tp)) {
3528 if (done) { 3519 netif_rx_complete(netdev, napi);
3529 netif_rx_complete(netdev);
3530 tg3_restart_ints(tp); 3520 tg3_restart_ints(tp);
3531 } 3521 }
3532 3522
3533 return (done ? 0 : 1); 3523 return work_done;
3534} 3524}
3535 3525
3536static void tg3_irq_quiesce(struct tg3 *tp) 3526static void tg3_irq_quiesce(struct tg3 *tp)
@@ -3577,7 +3567,7 @@ static irqreturn_t tg3_msi_1shot(int irq, void *dev_id)
3577 prefetch(&tp->rx_rcb[tp->rx_rcb_ptr]); 3567 prefetch(&tp->rx_rcb[tp->rx_rcb_ptr]);
3578 3568
3579 if (likely(!tg3_irq_sync(tp))) 3569 if (likely(!tg3_irq_sync(tp)))
3580 netif_rx_schedule(dev); /* schedule NAPI poll */ 3570 netif_rx_schedule(dev, &tp->napi);
3581 3571
3582 return IRQ_HANDLED; 3572 return IRQ_HANDLED;
3583} 3573}
@@ -3602,7 +3592,7 @@ static irqreturn_t tg3_msi(int irq, void *dev_id)
3602 */ 3592 */
3603 tw32_mailbox(MAILBOX_INTERRUPT_0 + TG3_64BIT_REG_LOW, 0x00000001); 3593 tw32_mailbox(MAILBOX_INTERRUPT_0 + TG3_64BIT_REG_LOW, 0x00000001);
3604 if (likely(!tg3_irq_sync(tp))) 3594 if (likely(!tg3_irq_sync(tp)))
3605 netif_rx_schedule(dev); /* schedule NAPI poll */ 3595 netif_rx_schedule(dev, &tp->napi);
3606 3596
3607 return IRQ_RETVAL(1); 3597 return IRQ_RETVAL(1);
3608} 3598}
@@ -3644,7 +3634,7 @@ static irqreturn_t tg3_interrupt(int irq, void *dev_id)
3644 sblk->status &= ~SD_STATUS_UPDATED; 3634 sblk->status &= ~SD_STATUS_UPDATED;
3645 if (likely(tg3_has_work(tp))) { 3635 if (likely(tg3_has_work(tp))) {
3646 prefetch(&tp->rx_rcb[tp->rx_rcb_ptr]); 3636 prefetch(&tp->rx_rcb[tp->rx_rcb_ptr]);
3647 netif_rx_schedule(dev); /* schedule NAPI poll */ 3637 netif_rx_schedule(dev, &tp->napi);
3648 } else { 3638 } else {
3649 /* No work, shared interrupt perhaps? re-enable 3639 /* No work, shared interrupt perhaps? re-enable
3650 * interrupts, and flush that PCI write 3640 * interrupts, and flush that PCI write
@@ -3690,7 +3680,7 @@ static irqreturn_t tg3_interrupt_tagged(int irq, void *dev_id)
3690 tw32_mailbox_f(MAILBOX_INTERRUPT_0 + TG3_64BIT_REG_LOW, 0x00000001); 3680 tw32_mailbox_f(MAILBOX_INTERRUPT_0 + TG3_64BIT_REG_LOW, 0x00000001);
3691 if (tg3_irq_sync(tp)) 3681 if (tg3_irq_sync(tp))
3692 goto out; 3682 goto out;
3693 if (netif_rx_schedule_prep(dev)) { 3683 if (netif_rx_schedule_prep(dev, &tp->napi)) {
3694 prefetch(&tp->rx_rcb[tp->rx_rcb_ptr]); 3684 prefetch(&tp->rx_rcb[tp->rx_rcb_ptr]);
3695 /* Update last_tag to mark that this status has been 3685 /* Update last_tag to mark that this status has been
3696 * seen. Because interrupt may be shared, we may be 3686 * seen. Because interrupt may be shared, we may be
@@ -3698,7 +3688,7 @@ static irqreturn_t tg3_interrupt_tagged(int irq, void *dev_id)
3698 * if tg3_poll() is not scheduled. 3688 * if tg3_poll() is not scheduled.
3699 */ 3689 */
3700 tp->last_tag = sblk->status_tag; 3690 tp->last_tag = sblk->status_tag;
3701 __netif_rx_schedule(dev); 3691 __netif_rx_schedule(dev, &tp->napi);
3702 } 3692 }
3703out: 3693out:
3704 return IRQ_RETVAL(handled); 3694 return IRQ_RETVAL(handled);
@@ -3737,7 +3727,7 @@ static int tg3_restart_hw(struct tg3 *tp, int reset_phy)
3737 tg3_full_unlock(tp); 3727 tg3_full_unlock(tp);
3738 del_timer_sync(&tp->timer); 3728 del_timer_sync(&tp->timer);
3739 tp->irq_sync = 0; 3729 tp->irq_sync = 0;
3740 netif_poll_enable(tp->dev); 3730 napi_enable(&tp->napi);
3741 dev_close(tp->dev); 3731 dev_close(tp->dev);
3742 tg3_full_lock(tp, 0); 3732 tg3_full_lock(tp, 0);
3743 } 3733 }
@@ -3932,7 +3922,7 @@ static int tg3_start_xmit(struct sk_buff *skb, struct net_device *dev)
3932 len = skb_headlen(skb); 3922 len = skb_headlen(skb);
3933 3923
3934 /* We are running in BH disabled context with netif_tx_lock 3924 /* We are running in BH disabled context with netif_tx_lock
3935 * and TX reclaim runs via tp->poll inside of a software 3925 * and TX reclaim runs via tp->napi.poll inside of a software
3936 * interrupt. Furthermore, IRQ processing runs lockless so we have 3926 * interrupt. Furthermore, IRQ processing runs lockless so we have
3937 * no IRQ context deadlocks to worry about either. Rejoice! 3927 * no IRQ context deadlocks to worry about either. Rejoice!
3938 */ 3928 */
@@ -4087,7 +4077,7 @@ static int tg3_start_xmit_dma_bug(struct sk_buff *skb, struct net_device *dev)
4087 len = skb_headlen(skb); 4077 len = skb_headlen(skb);
4088 4078
4089 /* We are running in BH disabled context with netif_tx_lock 4079 /* We are running in BH disabled context with netif_tx_lock
4090 * and TX reclaim runs via tp->poll inside of a software 4080 * and TX reclaim runs via tp->napi.poll inside of a software
4091 * interrupt. Furthermore, IRQ processing runs lockless so we have 4081 * interrupt. Furthermore, IRQ processing runs lockless so we have
4092 * no IRQ context deadlocks to worry about either. Rejoice! 4082 * no IRQ context deadlocks to worry about either. Rejoice!
4093 */ 4083 */
@@ -7147,6 +7137,8 @@ static int tg3_open(struct net_device *dev)
7147 return err; 7137 return err;
7148 } 7138 }
7149 7139
7140 napi_enable(&tp->napi);
7141
7150 tg3_full_lock(tp, 0); 7142 tg3_full_lock(tp, 0);
7151 7143
7152 err = tg3_init_hw(tp, 1); 7144 err = tg3_init_hw(tp, 1);
@@ -7174,6 +7166,7 @@ static int tg3_open(struct net_device *dev)
7174 tg3_full_unlock(tp); 7166 tg3_full_unlock(tp);
7175 7167
7176 if (err) { 7168 if (err) {
7169 napi_disable(&tp->napi);
7177 free_irq(tp->pdev->irq, dev); 7170 free_irq(tp->pdev->irq, dev);
7178 if (tp->tg3_flags2 & TG3_FLG2_USING_MSI) { 7171 if (tp->tg3_flags2 & TG3_FLG2_USING_MSI) {
7179 pci_disable_msi(tp->pdev); 7172 pci_disable_msi(tp->pdev);
@@ -7199,6 +7192,8 @@ static int tg3_open(struct net_device *dev)
7199 7192
7200 tg3_full_unlock(tp); 7193 tg3_full_unlock(tp);
7201 7194
7195 napi_disable(&tp->napi);
7196
7202 return err; 7197 return err;
7203 } 7198 }
7204 7199
@@ -7460,6 +7455,7 @@ static int tg3_close(struct net_device *dev)
7460{ 7455{
7461 struct tg3 *tp = netdev_priv(dev); 7456 struct tg3 *tp = netdev_priv(dev);
7462 7457
7458 napi_disable(&tp->napi);
7463 cancel_work_sync(&tp->reset_task); 7459 cancel_work_sync(&tp->reset_task);
7464 7460
7465 netif_stop_queue(dev); 7461 netif_stop_queue(dev);
@@ -11900,9 +11896,8 @@ static int __devinit tg3_init_one(struct pci_dev *pdev,
11900 dev->set_mac_address = tg3_set_mac_addr; 11896 dev->set_mac_address = tg3_set_mac_addr;
11901 dev->do_ioctl = tg3_ioctl; 11897 dev->do_ioctl = tg3_ioctl;
11902 dev->tx_timeout = tg3_tx_timeout; 11898 dev->tx_timeout = tg3_tx_timeout;
11903 dev->poll = tg3_poll; 11899 netif_napi_add(dev, &tp->napi, tg3_poll, 64);
11904 dev->ethtool_ops = &tg3_ethtool_ops; 11900 dev->ethtool_ops = &tg3_ethtool_ops;
11905 dev->weight = 64;
11906 dev->watchdog_timeo = TG3_TX_TIMEOUT; 11901 dev->watchdog_timeo = TG3_TX_TIMEOUT;
11907 dev->change_mtu = tg3_change_mtu; 11902 dev->change_mtu = tg3_change_mtu;
11908 dev->irq = pdev->irq; 11903 dev->irq = pdev->irq;
diff --git a/drivers/net/tg3.h b/drivers/net/tg3.h
index 5c21f49026c9..a6a23bbcdfee 100644
--- a/drivers/net/tg3.h
+++ b/drivers/net/tg3.h
@@ -2176,6 +2176,7 @@ struct tg3 {
2176 dma_addr_t tx_desc_mapping; 2176 dma_addr_t tx_desc_mapping;
2177 2177
2178 /* begin "rx thread" cacheline section */ 2178 /* begin "rx thread" cacheline section */
2179 struct napi_struct napi;
2179 void (*write32_rx_mbox) (struct tg3 *, u32, 2180 void (*write32_rx_mbox) (struct tg3 *, u32,
2180 u32); 2181 u32);
2181 u32 rx_rcb_ptr; 2182 u32 rx_rcb_ptr;
diff --git a/drivers/net/tsi108_eth.c b/drivers/net/tsi108_eth.c
index 1aabc91f6458..b3069ee34bd2 100644
--- a/drivers/net/tsi108_eth.c
+++ b/drivers/net/tsi108_eth.c
@@ -79,6 +79,9 @@ struct tsi108_prv_data {
79 void __iomem *regs; /* Base of normal regs */ 79 void __iomem *regs; /* Base of normal regs */
80 void __iomem *phyregs; /* Base of register bank used for PHY access */ 80 void __iomem *phyregs; /* Base of register bank used for PHY access */
81 81
82 struct net_device *dev;
83 struct napi_struct napi;
84
82 unsigned int phy; /* Index of PHY for this interface */ 85 unsigned int phy; /* Index of PHY for this interface */
83 unsigned int irq_num; 86 unsigned int irq_num;
84 unsigned int id; 87 unsigned int id;
@@ -837,13 +840,13 @@ static int tsi108_refill_rx(struct net_device *dev, int budget)
837 return done; 840 return done;
838} 841}
839 842
840static int tsi108_poll(struct net_device *dev, int *budget) 843static int tsi108_poll(struct napi_struct *napi, int budget)
841{ 844{
842 struct tsi108_prv_data *data = netdev_priv(dev); 845 struct tsi108_prv_data *data = container_of(napi, struct tsi108_prv_data, napi);
846 struct net_device *dev = data->dev;
843 u32 estat = TSI_READ(TSI108_EC_RXESTAT); 847 u32 estat = TSI_READ(TSI108_EC_RXESTAT);
844 u32 intstat = TSI_READ(TSI108_EC_INTSTAT); 848 u32 intstat = TSI_READ(TSI108_EC_INTSTAT);
845 int total_budget = min(*budget, dev->quota); 849 int num_received = 0, num_filled = 0;
846 int num_received = 0, num_filled = 0, budget_used;
847 850
848 intstat &= TSI108_INT_RXQUEUE0 | TSI108_INT_RXTHRESH | 851 intstat &= TSI108_INT_RXQUEUE0 | TSI108_INT_RXTHRESH |
849 TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR | TSI108_INT_RXWAIT; 852 TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR | TSI108_INT_RXWAIT;
@@ -852,7 +855,7 @@ static int tsi108_poll(struct net_device *dev, int *budget)
852 TSI_WRITE(TSI108_EC_INTSTAT, intstat); 855 TSI_WRITE(TSI108_EC_INTSTAT, intstat);
853 856
854 if (data->rxpending || (estat & TSI108_EC_RXESTAT_Q0_DESCINT)) 857 if (data->rxpending || (estat & TSI108_EC_RXESTAT_Q0_DESCINT))
855 num_received = tsi108_complete_rx(dev, total_budget); 858 num_received = tsi108_complete_rx(dev, budget);
856 859
857 /* This should normally fill no more slots than the number of 860 /* This should normally fill no more slots than the number of
858 * packets received in tsi108_complete_rx(). The exception 861 * packets received in tsi108_complete_rx(). The exception
@@ -867,7 +870,7 @@ static int tsi108_poll(struct net_device *dev, int *budget)
867 */ 870 */
868 871
869 if (data->rxfree < TSI108_RXRING_LEN) 872 if (data->rxfree < TSI108_RXRING_LEN)
870 num_filled = tsi108_refill_rx(dev, total_budget * 2); 873 num_filled = tsi108_refill_rx(dev, budget * 2);
871 874
872 if (intstat & TSI108_INT_RXERROR) { 875 if (intstat & TSI108_INT_RXERROR) {
873 u32 err = TSI_READ(TSI108_EC_RXERR); 876 u32 err = TSI_READ(TSI108_EC_RXERR);
@@ -890,14 +893,9 @@ static int tsi108_poll(struct net_device *dev, int *budget)
890 spin_unlock_irq(&data->misclock); 893 spin_unlock_irq(&data->misclock);
891 } 894 }
892 895
893 budget_used = max(num_received, num_filled / 2); 896 if (num_received < budget) {
894
895 *budget -= budget_used;
896 dev->quota -= budget_used;
897
898 if (budget_used != total_budget) {
899 data->rxpending = 0; 897 data->rxpending = 0;
900 netif_rx_complete(dev); 898 netif_rx_complete(dev, napi);
901 899
902 TSI_WRITE(TSI108_EC_INTMASK, 900 TSI_WRITE(TSI108_EC_INTMASK,
903 TSI_READ(TSI108_EC_INTMASK) 901 TSI_READ(TSI108_EC_INTMASK)
@@ -906,14 +904,11 @@ static int tsi108_poll(struct net_device *dev, int *budget)
906 TSI108_INT_RXOVERRUN | 904 TSI108_INT_RXOVERRUN |
907 TSI108_INT_RXERROR | 905 TSI108_INT_RXERROR |
908 TSI108_INT_RXWAIT)); 906 TSI108_INT_RXWAIT));
909
910 /* IRQs are level-triggered, so no need to re-check */
911 return 0;
912 } else { 907 } else {
913 data->rxpending = 1; 908 data->rxpending = 1;
914 } 909 }
915 910
916 return 1; 911 return num_received;
917} 912}
918 913
919static void tsi108_rx_int(struct net_device *dev) 914static void tsi108_rx_int(struct net_device *dev)
@@ -931,7 +926,7 @@ static void tsi108_rx_int(struct net_device *dev)
931 * from tsi108_check_rxring(). 926 * from tsi108_check_rxring().
932 */ 927 */
933 928
934 if (netif_rx_schedule_prep(dev)) { 929 if (netif_rx_schedule_prep(dev, &data->napi)) {
935 /* Mask, rather than ack, the receive interrupts. The ack 930 /* Mask, rather than ack, the receive interrupts. The ack
936 * will happen in tsi108_poll(). 931 * will happen in tsi108_poll().
937 */ 932 */
@@ -942,7 +937,7 @@ static void tsi108_rx_int(struct net_device *dev)
942 | TSI108_INT_RXTHRESH | 937 | TSI108_INT_RXTHRESH |
943 TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR | 938 TSI108_INT_RXOVERRUN | TSI108_INT_RXERROR |
944 TSI108_INT_RXWAIT); 939 TSI108_INT_RXWAIT);
945 __netif_rx_schedule(dev); 940 __netif_rx_schedule(dev, &data->napi);
946 } else { 941 } else {
947 if (!netif_running(dev)) { 942 if (!netif_running(dev)) {
948 /* This can happen if an interrupt occurs while the 943 /* This can happen if an interrupt occurs while the
@@ -1401,6 +1396,8 @@ static int tsi108_open(struct net_device *dev)
1401 TSI_WRITE(TSI108_EC_TXQ_PTRLOW, data->txdma); 1396 TSI_WRITE(TSI108_EC_TXQ_PTRLOW, data->txdma);
1402 tsi108_init_phy(dev); 1397 tsi108_init_phy(dev);
1403 1398
1399 napi_enable(&data->napi);
1400
1404 setup_timer(&data->timer, tsi108_timed_checker, (unsigned long)dev); 1401 setup_timer(&data->timer, tsi108_timed_checker, (unsigned long)dev);
1405 mod_timer(&data->timer, jiffies + 1); 1402 mod_timer(&data->timer, jiffies + 1);
1406 1403
@@ -1425,6 +1422,7 @@ static int tsi108_close(struct net_device *dev)
1425 struct tsi108_prv_data *data = netdev_priv(dev); 1422 struct tsi108_prv_data *data = netdev_priv(dev);
1426 1423
1427 netif_stop_queue(dev); 1424 netif_stop_queue(dev);
1425 napi_disable(&data->napi);
1428 1426
1429 del_timer_sync(&data->timer); 1427 del_timer_sync(&data->timer);
1430 1428
@@ -1562,6 +1560,7 @@ tsi108_init_one(struct platform_device *pdev)
1562 1560
1563 printk("tsi108_eth%d: probe...\n", pdev->id); 1561 printk("tsi108_eth%d: probe...\n", pdev->id);
1564 data = netdev_priv(dev); 1562 data = netdev_priv(dev);
1563 data->dev = dev;
1565 1564
1566 pr_debug("tsi108_eth%d:regs:phyresgs:phy:irq_num=0x%x:0x%x:0x%x:0x%x\n", 1565 pr_debug("tsi108_eth%d:regs:phyresgs:phy:irq_num=0x%x:0x%x:0x%x:0x%x\n",
1567 pdev->id, einfo->regs, einfo->phyregs, 1566 pdev->id, einfo->regs, einfo->phyregs,
@@ -1597,9 +1596,8 @@ tsi108_init_one(struct platform_device *pdev)
1597 dev->set_mac_address = tsi108_set_mac; 1596 dev->set_mac_address = tsi108_set_mac;
1598 dev->set_multicast_list = tsi108_set_rx_mode; 1597 dev->set_multicast_list = tsi108_set_rx_mode;
1599 dev->get_stats = tsi108_get_stats; 1598 dev->get_stats = tsi108_get_stats;
1600 dev->poll = tsi108_poll; 1599 netif_napi_add(dev, &data->napi, tsi108_poll, 64);
1601 dev->do_ioctl = tsi108_do_ioctl; 1600 dev->do_ioctl = tsi108_do_ioctl;
1602 dev->weight = 64; /* 64 is more suitable for GigE interface - klai */
1603 1601
1604 /* Apparently, the Linux networking code won't use scatter-gather 1602 /* Apparently, the Linux networking code won't use scatter-gather
1605 * if the hardware doesn't do checksums. However, it's faster 1603 * if the hardware doesn't do checksums. However, it's faster
diff --git a/drivers/net/tulip/interrupt.c b/drivers/net/tulip/interrupt.c
index 53efd6694e75..365331446387 100644
--- a/drivers/net/tulip/interrupt.c
+++ b/drivers/net/tulip/interrupt.c
@@ -103,28 +103,29 @@ int tulip_refill_rx(struct net_device *dev)
103void oom_timer(unsigned long data) 103void oom_timer(unsigned long data)
104{ 104{
105 struct net_device *dev = (struct net_device *)data; 105 struct net_device *dev = (struct net_device *)data;
106 netif_rx_schedule(dev); 106 struct tulip_private *tp = netdev_priv(dev);
107 netif_rx_schedule(dev, &tp->napi);
107} 108}
108 109
109int tulip_poll(struct net_device *dev, int *budget) 110int tulip_poll(struct napi_struct *napi, int budget)
110{ 111{
111 struct tulip_private *tp = netdev_priv(dev); 112 struct tulip_private *tp = container_of(napi, struct tulip_private, napi);
113 struct net_device *dev = tp->dev;
112 int entry = tp->cur_rx % RX_RING_SIZE; 114 int entry = tp->cur_rx % RX_RING_SIZE;
113 int rx_work_limit = *budget; 115 int work_done = 0;
116#ifdef CONFIG_TULIP_NAPI_HW_MITIGATION
114 int received = 0; 117 int received = 0;
118#endif
115 119
116 if (!netif_running(dev)) 120 if (!netif_running(dev))
117 goto done; 121 goto done;
118 122
119 if (rx_work_limit > dev->quota)
120 rx_work_limit = dev->quota;
121
122#ifdef CONFIG_TULIP_NAPI_HW_MITIGATION 123#ifdef CONFIG_TULIP_NAPI_HW_MITIGATION
123 124
124/* that one buffer is needed for mit activation; or might be a 125/* that one buffer is needed for mit activation; or might be a
125 bug in the ring buffer code; check later -- JHS*/ 126 bug in the ring buffer code; check later -- JHS*/
126 127
127 if (rx_work_limit >=RX_RING_SIZE) rx_work_limit--; 128 if (budget >=RX_RING_SIZE) budget--;
128#endif 129#endif
129 130
130 if (tulip_debug > 4) 131 if (tulip_debug > 4)
@@ -144,14 +145,13 @@ int tulip_poll(struct net_device *dev, int *budget)
144 while ( ! (tp->rx_ring[entry].status & cpu_to_le32(DescOwned))) { 145 while ( ! (tp->rx_ring[entry].status & cpu_to_le32(DescOwned))) {
145 s32 status = le32_to_cpu(tp->rx_ring[entry].status); 146 s32 status = le32_to_cpu(tp->rx_ring[entry].status);
146 147
147
148 if (tp->dirty_rx + RX_RING_SIZE == tp->cur_rx) 148 if (tp->dirty_rx + RX_RING_SIZE == tp->cur_rx)
149 break; 149 break;
150 150
151 if (tulip_debug > 5) 151 if (tulip_debug > 5)
152 printk(KERN_DEBUG "%s: In tulip_rx(), entry %d %8.8x.\n", 152 printk(KERN_DEBUG "%s: In tulip_rx(), entry %d %8.8x.\n",
153 dev->name, entry, status); 153 dev->name, entry, status);
154 if (--rx_work_limit < 0) 154 if (work_done++ >= budget)
155 goto not_done; 155 goto not_done;
156 156
157 if ((status & 0x38008300) != 0x0300) { 157 if ((status & 0x38008300) != 0x0300) {
@@ -238,7 +238,9 @@ int tulip_poll(struct net_device *dev, int *budget)
238 tp->stats.rx_packets++; 238 tp->stats.rx_packets++;
239 tp->stats.rx_bytes += pkt_len; 239 tp->stats.rx_bytes += pkt_len;
240 } 240 }
241 received++; 241#ifdef CONFIG_TULIP_NAPI_HW_MITIGATION
242 received++;
243#endif
242 244
243 entry = (++tp->cur_rx) % RX_RING_SIZE; 245 entry = (++tp->cur_rx) % RX_RING_SIZE;
244 if (tp->cur_rx - tp->dirty_rx > RX_RING_SIZE/4) 246 if (tp->cur_rx - tp->dirty_rx > RX_RING_SIZE/4)
@@ -296,17 +298,15 @@ done:
296 298
297#endif /* CONFIG_TULIP_NAPI_HW_MITIGATION */ 299#endif /* CONFIG_TULIP_NAPI_HW_MITIGATION */
298 300
299 dev->quota -= received;
300 *budget -= received;
301
302 tulip_refill_rx(dev); 301 tulip_refill_rx(dev);
303 302
304 /* If RX ring is not full we are out of memory. */ 303 /* If RX ring is not full we are out of memory. */
305 if (tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL) goto oom; 304 if (tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
305 goto oom;
306 306
307 /* Remove us from polling list and enable RX intr. */ 307 /* Remove us from polling list and enable RX intr. */
308 308
309 netif_rx_complete(dev); 309 netif_rx_complete(dev, napi);
310 iowrite32(tulip_tbl[tp->chip_id].valid_intrs, tp->base_addr+CSR7); 310 iowrite32(tulip_tbl[tp->chip_id].valid_intrs, tp->base_addr+CSR7);
311 311
312 /* The last op happens after poll completion. Which means the following: 312 /* The last op happens after poll completion. Which means the following:
@@ -320,28 +320,20 @@ done:
320 * processed irqs. But it must not result in losing events. 320 * processed irqs. But it must not result in losing events.
321 */ 321 */
322 322
323 return 0; 323 return work_done;
324 324
325 not_done: 325 not_done:
326 if (!received) {
327
328 received = dev->quota; /* Not to happen */
329 }
330 dev->quota -= received;
331 *budget -= received;
332
333 if (tp->cur_rx - tp->dirty_rx > RX_RING_SIZE/2 || 326 if (tp->cur_rx - tp->dirty_rx > RX_RING_SIZE/2 ||
334 tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL) 327 tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
335 tulip_refill_rx(dev); 328 tulip_refill_rx(dev);
336 329
337 if (tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL) goto oom; 330 if (tp->rx_buffers[tp->dirty_rx % RX_RING_SIZE].skb == NULL)
338 331 goto oom;
339 return 1;
340 332
333 return work_done;
341 334
342 oom: /* Executed with RX ints disabled */ 335 oom: /* Executed with RX ints disabled */
343 336
344
345 /* Start timer, stop polling, but do not enable rx interrupts. */ 337 /* Start timer, stop polling, but do not enable rx interrupts. */
346 mod_timer(&tp->oom_timer, jiffies+1); 338 mod_timer(&tp->oom_timer, jiffies+1);
347 339
@@ -350,9 +342,9 @@ done:
350 * before we did netif_rx_complete(). See? We would lose it. */ 342 * before we did netif_rx_complete(). See? We would lose it. */
351 343
352 /* remove ourselves from the polling list */ 344 /* remove ourselves from the polling list */
353 netif_rx_complete(dev); 345 netif_rx_complete(dev, napi);
354 346
355 return 0; 347 return work_done;
356} 348}
357 349
358#else /* CONFIG_TULIP_NAPI */ 350#else /* CONFIG_TULIP_NAPI */
@@ -534,7 +526,7 @@ irqreturn_t tulip_interrupt(int irq, void *dev_instance)
534 rxd++; 526 rxd++;
535 /* Mask RX intrs and add the device to poll list. */ 527 /* Mask RX intrs and add the device to poll list. */
536 iowrite32(tulip_tbl[tp->chip_id].valid_intrs&~RxPollInt, ioaddr + CSR7); 528 iowrite32(tulip_tbl[tp->chip_id].valid_intrs&~RxPollInt, ioaddr + CSR7);
537 netif_rx_schedule(dev); 529 netif_rx_schedule(dev, &tp->napi);
538 530
539 if (!(csr5&~(AbnormalIntr|NormalIntr|RxPollInt|TPLnkPass))) 531 if (!(csr5&~(AbnormalIntr|NormalIntr|RxPollInt|TPLnkPass)))
540 break; 532 break;
diff --git a/drivers/net/tulip/tulip.h b/drivers/net/tulip/tulip.h
index 16f26a8364f0..5a4d7270973e 100644
--- a/drivers/net/tulip/tulip.h
+++ b/drivers/net/tulip/tulip.h
@@ -353,6 +353,7 @@ struct tulip_private {
353 int chip_id; 353 int chip_id;
354 int revision; 354 int revision;
355 int flags; 355 int flags;
356 struct napi_struct napi;
356 struct net_device_stats stats; 357 struct net_device_stats stats;
357 struct timer_list timer; /* Media selection timer. */ 358 struct timer_list timer; /* Media selection timer. */
358 struct timer_list oom_timer; /* Out of memory timer. */ 359 struct timer_list oom_timer; /* Out of memory timer. */
@@ -429,7 +430,7 @@ extern int tulip_rx_copybreak;
429irqreturn_t tulip_interrupt(int irq, void *dev_instance); 430irqreturn_t tulip_interrupt(int irq, void *dev_instance);
430int tulip_refill_rx(struct net_device *dev); 431int tulip_refill_rx(struct net_device *dev);
431#ifdef CONFIG_TULIP_NAPI 432#ifdef CONFIG_TULIP_NAPI
432int tulip_poll(struct net_device *dev, int *budget); 433int tulip_poll(struct napi_struct *napi, int budget);
433#endif 434#endif
434 435
435 436
diff --git a/drivers/net/tulip/tulip_core.c b/drivers/net/tulip/tulip_core.c
index eca984f89bbf..7040a59fa3c9 100644
--- a/drivers/net/tulip/tulip_core.c
+++ b/drivers/net/tulip/tulip_core.c
@@ -294,6 +294,10 @@ static void tulip_up(struct net_device *dev)
294 int next_tick = 3*HZ; 294 int next_tick = 3*HZ;
295 int i; 295 int i;
296 296
297#ifdef CONFIG_TULIP_NAPI
298 napi_enable(&tp->napi);
299#endif
300
297 /* Wake the chip from sleep/snooze mode. */ 301 /* Wake the chip from sleep/snooze mode. */
298 tulip_set_power_state (tp, 0, 0); 302 tulip_set_power_state (tp, 0, 0);
299 303
@@ -728,6 +732,10 @@ static void tulip_down (struct net_device *dev)
728 732
729 flush_scheduled_work(); 733 flush_scheduled_work();
730 734
735#ifdef CONFIG_TULIP_NAPI
736 napi_disable(&tp->napi);
737#endif
738
731 del_timer_sync (&tp->timer); 739 del_timer_sync (&tp->timer);
732#ifdef CONFIG_TULIP_NAPI 740#ifdef CONFIG_TULIP_NAPI
733 del_timer_sync (&tp->oom_timer); 741 del_timer_sync (&tp->oom_timer);
@@ -1606,8 +1614,7 @@ static int __devinit tulip_init_one (struct pci_dev *pdev,
1606 dev->tx_timeout = tulip_tx_timeout; 1614 dev->tx_timeout = tulip_tx_timeout;
1607 dev->watchdog_timeo = TX_TIMEOUT; 1615 dev->watchdog_timeo = TX_TIMEOUT;
1608#ifdef CONFIG_TULIP_NAPI 1616#ifdef CONFIG_TULIP_NAPI
1609 dev->poll = tulip_poll; 1617 netif_napi_add(dev, &tp->napi, tulip_poll, 16);
1610 dev->weight = 16;
1611#endif 1618#endif
1612 dev->stop = tulip_close; 1619 dev->stop = tulip_close;
1613 dev->get_stats = tulip_get_stats; 1620 dev->get_stats = tulip_get_stats;
diff --git a/drivers/net/typhoon.c b/drivers/net/typhoon.c
index 03587205546e..0377b8b64c78 100644
--- a/drivers/net/typhoon.c
+++ b/drivers/net/typhoon.c
@@ -284,6 +284,7 @@ struct typhoon {
284 struct basic_ring rxLoRing; 284 struct basic_ring rxLoRing;
285 struct pci_dev * pdev; 285 struct pci_dev * pdev;
286 struct net_device * dev; 286 struct net_device * dev;
287 struct napi_struct napi;
287 spinlock_t state_lock; 288 spinlock_t state_lock;
288 struct vlan_group * vlgrp; 289 struct vlan_group * vlgrp;
289 struct basic_ring rxHiRing; 290 struct basic_ring rxHiRing;
@@ -1759,12 +1760,12 @@ typhoon_fill_free_ring(struct typhoon *tp)
1759} 1760}
1760 1761
1761static int 1762static int
1762typhoon_poll(struct net_device *dev, int *total_budget) 1763typhoon_poll(struct napi_struct *napi, int budget)
1763{ 1764{
1764 struct typhoon *tp = netdev_priv(dev); 1765 struct typhoon *tp = container_of(napi, struct typhoon, napi);
1766 struct net_device *dev = tp->dev;
1765 struct typhoon_indexes *indexes = tp->indexes; 1767 struct typhoon_indexes *indexes = tp->indexes;
1766 int orig_budget = *total_budget; 1768 int work_done;
1767 int budget, work_done, done;
1768 1769
1769 rmb(); 1770 rmb();
1770 if(!tp->awaiting_resp && indexes->respReady != indexes->respCleared) 1771 if(!tp->awaiting_resp && indexes->respReady != indexes->respCleared)
@@ -1773,30 +1774,16 @@ typhoon_poll(struct net_device *dev, int *total_budget)
1773 if(le32_to_cpu(indexes->txLoCleared) != tp->txLoRing.lastRead) 1774 if(le32_to_cpu(indexes->txLoCleared) != tp->txLoRing.lastRead)
1774 typhoon_tx_complete(tp, &tp->txLoRing, &indexes->txLoCleared); 1775 typhoon_tx_complete(tp, &tp->txLoRing, &indexes->txLoCleared);
1775 1776
1776 if(orig_budget > dev->quota)
1777 orig_budget = dev->quota;
1778
1779 budget = orig_budget;
1780 work_done = 0; 1777 work_done = 0;
1781 done = 1;
1782 1778
1783 if(indexes->rxHiCleared != indexes->rxHiReady) { 1779 if(indexes->rxHiCleared != indexes->rxHiReady) {
1784 work_done = typhoon_rx(tp, &tp->rxHiRing, &indexes->rxHiReady, 1780 work_done += typhoon_rx(tp, &tp->rxHiRing, &indexes->rxHiReady,
1785 &indexes->rxHiCleared, budget); 1781 &indexes->rxHiCleared, budget);
1786 budget -= work_done;
1787 } 1782 }
1788 1783
1789 if(indexes->rxLoCleared != indexes->rxLoReady) { 1784 if(indexes->rxLoCleared != indexes->rxLoReady) {
1790 work_done += typhoon_rx(tp, &tp->rxLoRing, &indexes->rxLoReady, 1785 work_done += typhoon_rx(tp, &tp->rxLoRing, &indexes->rxLoReady,
1791 &indexes->rxLoCleared, budget); 1786 &indexes->rxLoCleared, budget - work_done);
1792 }
1793
1794 if(work_done) {
1795 *total_budget -= work_done;
1796 dev->quota -= work_done;
1797
1798 if(work_done >= orig_budget)
1799 done = 0;
1800 } 1787 }
1801 1788
1802 if(le32_to_cpu(indexes->rxBuffCleared) == tp->rxBuffRing.lastWrite) { 1789 if(le32_to_cpu(indexes->rxBuffCleared) == tp->rxBuffRing.lastWrite) {
@@ -1804,14 +1791,14 @@ typhoon_poll(struct net_device *dev, int *total_budget)
1804 typhoon_fill_free_ring(tp); 1791 typhoon_fill_free_ring(tp);
1805 } 1792 }
1806 1793
1807 if(done) { 1794 if (work_done < budget) {
1808 netif_rx_complete(dev); 1795 netif_rx_complete(dev, napi);
1809 iowrite32(TYPHOON_INTR_NONE, 1796 iowrite32(TYPHOON_INTR_NONE,
1810 tp->ioaddr + TYPHOON_REG_INTR_MASK); 1797 tp->ioaddr + TYPHOON_REG_INTR_MASK);
1811 typhoon_post_pci_writes(tp->ioaddr); 1798 typhoon_post_pci_writes(tp->ioaddr);
1812 } 1799 }
1813 1800
1814 return (done ? 0 : 1); 1801 return work_done;
1815} 1802}
1816 1803
1817static irqreturn_t 1804static irqreturn_t
@@ -1828,10 +1815,10 @@ typhoon_interrupt(int irq, void *dev_instance)
1828 1815
1829 iowrite32(intr_status, ioaddr + TYPHOON_REG_INTR_STATUS); 1816 iowrite32(intr_status, ioaddr + TYPHOON_REG_INTR_STATUS);
1830 1817
1831 if(netif_rx_schedule_prep(dev)) { 1818 if (netif_rx_schedule_prep(dev, &tp->napi)) {
1832 iowrite32(TYPHOON_INTR_ALL, ioaddr + TYPHOON_REG_INTR_MASK); 1819 iowrite32(TYPHOON_INTR_ALL, ioaddr + TYPHOON_REG_INTR_MASK);
1833 typhoon_post_pci_writes(ioaddr); 1820 typhoon_post_pci_writes(ioaddr);
1834 __netif_rx_schedule(dev); 1821 __netif_rx_schedule(dev, &tp->napi);
1835 } else { 1822 } else {
1836 printk(KERN_ERR "%s: Error, poll already scheduled\n", 1823 printk(KERN_ERR "%s: Error, poll already scheduled\n",
1837 dev->name); 1824 dev->name);
@@ -2119,9 +2106,13 @@ typhoon_open(struct net_device *dev)
2119 if(err < 0) 2106 if(err < 0)
2120 goto out_sleep; 2107 goto out_sleep;
2121 2108
2109 napi_enable(&tp->napi);
2110
2122 err = typhoon_start_runtime(tp); 2111 err = typhoon_start_runtime(tp);
2123 if(err < 0) 2112 if(err < 0) {
2113 napi_disable(&tp->napi);
2124 goto out_irq; 2114 goto out_irq;
2115 }
2125 2116
2126 netif_start_queue(dev); 2117 netif_start_queue(dev);
2127 return 0; 2118 return 0;
@@ -2150,6 +2141,7 @@ typhoon_close(struct net_device *dev)
2150 struct typhoon *tp = netdev_priv(dev); 2141 struct typhoon *tp = netdev_priv(dev);
2151 2142
2152 netif_stop_queue(dev); 2143 netif_stop_queue(dev);
2144 napi_disable(&tp->napi);
2153 2145
2154 if(typhoon_stop_runtime(tp, WaitSleep) < 0) 2146 if(typhoon_stop_runtime(tp, WaitSleep) < 0)
2155 printk(KERN_ERR "%s: unable to stop runtime\n", dev->name); 2147 printk(KERN_ERR "%s: unable to stop runtime\n", dev->name);
@@ -2521,8 +2513,7 @@ typhoon_init_one(struct pci_dev *pdev, const struct pci_device_id *ent)
2521 dev->stop = typhoon_close; 2513 dev->stop = typhoon_close;
2522 dev->set_multicast_list = typhoon_set_rx_mode; 2514 dev->set_multicast_list = typhoon_set_rx_mode;
2523 dev->tx_timeout = typhoon_tx_timeout; 2515 dev->tx_timeout = typhoon_tx_timeout;
2524 dev->poll = typhoon_poll; 2516 netif_napi_add(dev, &tp->napi, typhoon_poll, 16);
2525 dev->weight = 16;
2526 dev->watchdog_timeo = TX_TIMEOUT; 2517 dev->watchdog_timeo = TX_TIMEOUT;
2527 dev->get_stats = typhoon_get_stats; 2518 dev->get_stats = typhoon_get_stats;
2528 dev->set_mac_address = typhoon_set_mac_address; 2519 dev->set_mac_address = typhoon_set_mac_address;
diff --git a/drivers/net/ucc_geth.c b/drivers/net/ucc_geth.c
index 9a38dfe45f8f..72f617bf2520 100644
--- a/drivers/net/ucc_geth.c
+++ b/drivers/net/ucc_geth.c
@@ -3582,41 +3582,31 @@ static int ucc_geth_tx(struct net_device *dev, u8 txQ)
3582} 3582}
3583 3583
3584#ifdef CONFIG_UGETH_NAPI 3584#ifdef CONFIG_UGETH_NAPI
3585static int ucc_geth_poll(struct net_device *dev, int *budget) 3585static int ucc_geth_poll(struct napi_struct *napi, int budget)
3586{ 3586{
3587 struct ucc_geth_private *ugeth = netdev_priv(dev); 3587 struct ucc_geth_private *ugeth = container_of(napi, struct ucc_geth_private, napi);
3588 struct net_device *dev = ugeth->dev;
3588 struct ucc_geth_info *ug_info; 3589 struct ucc_geth_info *ug_info;
3589 struct ucc_fast_private *uccf; 3590 int howmany, i;
3590 int howmany;
3591 u8 i;
3592 int rx_work_limit;
3593 register u32 uccm;
3594 3591
3595 ug_info = ugeth->ug_info; 3592 ug_info = ugeth->ug_info;
3596 3593
3597 rx_work_limit = *budget;
3598 if (rx_work_limit > dev->quota)
3599 rx_work_limit = dev->quota;
3600
3601 howmany = 0; 3594 howmany = 0;
3595 for (i = 0; i < ug_info->numQueuesRx; i++)
3596 howmany += ucc_geth_rx(ugeth, i, budget - howmany);
3602 3597
3603 for (i = 0; i < ug_info->numQueuesRx; i++) { 3598 if (howmany < budget) {
3604 howmany += ucc_geth_rx(ugeth, i, rx_work_limit); 3599 struct ucc_fast_private *uccf;
3605 } 3600 u32 uccm;
3606
3607 dev->quota -= howmany;
3608 rx_work_limit -= howmany;
3609 *budget -= howmany;
3610 3601
3611 if (rx_work_limit > 0) { 3602 netif_rx_complete(dev, napi);
3612 netif_rx_complete(dev);
3613 uccf = ugeth->uccf; 3603 uccf = ugeth->uccf;
3614 uccm = in_be32(uccf->p_uccm); 3604 uccm = in_be32(uccf->p_uccm);
3615 uccm |= UCCE_RX_EVENTS; 3605 uccm |= UCCE_RX_EVENTS;
3616 out_be32(uccf->p_uccm, uccm); 3606 out_be32(uccf->p_uccm, uccm);
3617 } 3607 }
3618 3608
3619 return (rx_work_limit > 0) ? 0 : 1; 3609 return howmany;
3620} 3610}
3621#endif /* CONFIG_UGETH_NAPI */ 3611#endif /* CONFIG_UGETH_NAPI */
3622 3612
@@ -3651,10 +3641,10 @@ static irqreturn_t ucc_geth_irq_handler(int irq, void *info)
3651 /* check for receive events that require processing */ 3641 /* check for receive events that require processing */
3652 if (ucce & UCCE_RX_EVENTS) { 3642 if (ucce & UCCE_RX_EVENTS) {
3653#ifdef CONFIG_UGETH_NAPI 3643#ifdef CONFIG_UGETH_NAPI
3654 if (netif_rx_schedule_prep(dev)) { 3644 if (netif_rx_schedule_prep(dev, &ugeth->napi)) {
3655 uccm &= ~UCCE_RX_EVENTS; 3645 uccm &= ~UCCE_RX_EVENTS;
3656 out_be32(uccf->p_uccm, uccm); 3646 out_be32(uccf->p_uccm, uccm);
3657 __netif_rx_schedule(dev); 3647 __netif_rx_schedule(dev, &ugeth->napi);
3658 } 3648 }
3659#else 3649#else
3660 rx_mask = UCCE_RXBF_SINGLE_MASK; 3650 rx_mask = UCCE_RXBF_SINGLE_MASK;
@@ -3717,12 +3707,15 @@ static int ucc_geth_open(struct net_device *dev)
3717 return err; 3707 return err;
3718 } 3708 }
3719 3709
3710#ifdef CONFIG_UGETH_NAPI
3711 napi_enable(&ugeth->napi);
3712#endif
3720 err = ucc_geth_startup(ugeth); 3713 err = ucc_geth_startup(ugeth);
3721 if (err) { 3714 if (err) {
3722 if (netif_msg_ifup(ugeth)) 3715 if (netif_msg_ifup(ugeth))
3723 ugeth_err("%s: Cannot configure net device, aborting.", 3716 ugeth_err("%s: Cannot configure net device, aborting.",
3724 dev->name); 3717 dev->name);
3725 return err; 3718 goto out_err;
3726 } 3719 }
3727 3720
3728 err = adjust_enet_interface(ugeth); 3721 err = adjust_enet_interface(ugeth);
@@ -3730,7 +3723,7 @@ static int ucc_geth_open(struct net_device *dev)
3730 if (netif_msg_ifup(ugeth)) 3723 if (netif_msg_ifup(ugeth))
3731 ugeth_err("%s: Cannot configure net device, aborting.", 3724 ugeth_err("%s: Cannot configure net device, aborting.",
3732 dev->name); 3725 dev->name);
3733 return err; 3726 goto out_err;
3734 } 3727 }
3735 3728
3736 /* Set MACSTNADDR1, MACSTNADDR2 */ 3729 /* Set MACSTNADDR1, MACSTNADDR2 */
@@ -3748,7 +3741,7 @@ static int ucc_geth_open(struct net_device *dev)
3748 if (err) { 3741 if (err) {
3749 if (netif_msg_ifup(ugeth)) 3742 if (netif_msg_ifup(ugeth))
3750 ugeth_err("%s: Cannot initialize PHY, aborting.", dev->name); 3743 ugeth_err("%s: Cannot initialize PHY, aborting.", dev->name);
3751 return err; 3744 goto out_err;
3752 } 3745 }
3753 3746
3754 phy_start(ugeth->phydev); 3747 phy_start(ugeth->phydev);
@@ -3761,7 +3754,7 @@ static int ucc_geth_open(struct net_device *dev)
3761 ugeth_err("%s: Cannot get IRQ for net device, aborting.", 3754 ugeth_err("%s: Cannot get IRQ for net device, aborting.",
3762 dev->name); 3755 dev->name);
3763 ucc_geth_stop(ugeth); 3756 ucc_geth_stop(ugeth);
3764 return err; 3757 goto out_err;
3765 } 3758 }
3766 3759
3767 err = ugeth_enable(ugeth, COMM_DIR_RX_AND_TX); 3760 err = ugeth_enable(ugeth, COMM_DIR_RX_AND_TX);
@@ -3769,12 +3762,18 @@ static int ucc_geth_open(struct net_device *dev)
3769 if (netif_msg_ifup(ugeth)) 3762 if (netif_msg_ifup(ugeth))
3770 ugeth_err("%s: Cannot enable net device, aborting.", dev->name); 3763 ugeth_err("%s: Cannot enable net device, aborting.", dev->name);
3771 ucc_geth_stop(ugeth); 3764 ucc_geth_stop(ugeth);
3772 return err; 3765 goto out_err;
3773 } 3766 }
3774 3767
3775 netif_start_queue(dev); 3768 netif_start_queue(dev);
3776 3769
3777 return err; 3770 return err;
3771
3772out_err:
3773#ifdef CONFIG_UGETH_NAPI
3774 napi_disable(&ugeth->napi);
3775#endif
3776 return err;
3778} 3777}
3779 3778
3780/* Stops the kernel queue, and halts the controller */ 3779/* Stops the kernel queue, and halts the controller */
@@ -3784,6 +3783,10 @@ static int ucc_geth_close(struct net_device *dev)
3784 3783
3785 ugeth_vdbg("%s: IN", __FUNCTION__); 3784 ugeth_vdbg("%s: IN", __FUNCTION__);
3786 3785
3786#ifdef CONFIG_UGETH_NAPI
3787 napi_disable(&ugeth->napi);
3788#endif
3789
3787 ucc_geth_stop(ugeth); 3790 ucc_geth_stop(ugeth);
3788 3791
3789 phy_disconnect(ugeth->phydev); 3792 phy_disconnect(ugeth->phydev);
@@ -3964,8 +3967,7 @@ static int ucc_geth_probe(struct of_device* ofdev, const struct of_device_id *ma
3964 dev->tx_timeout = ucc_geth_timeout; 3967 dev->tx_timeout = ucc_geth_timeout;
3965 dev->watchdog_timeo = TX_TIMEOUT; 3968 dev->watchdog_timeo = TX_TIMEOUT;
3966#ifdef CONFIG_UGETH_NAPI 3969#ifdef CONFIG_UGETH_NAPI
3967 dev->poll = ucc_geth_poll; 3970 netif_napi_add(dev, &ugeth->napi, ucc_geth_poll, UCC_GETH_DEV_WEIGHT);
3968 dev->weight = UCC_GETH_DEV_WEIGHT;
3969#endif /* CONFIG_UGETH_NAPI */ 3971#endif /* CONFIG_UGETH_NAPI */
3970 dev->stop = ucc_geth_close; 3972 dev->stop = ucc_geth_close;
3971 dev->get_stats = ucc_geth_get_stats; 3973 dev->get_stats = ucc_geth_get_stats;
diff --git a/drivers/net/ucc_geth.h b/drivers/net/ucc_geth.h
index bb4dac8c0c65..0579ba081aa5 100644
--- a/drivers/net/ucc_geth.h
+++ b/drivers/net/ucc_geth.h
@@ -1184,6 +1184,7 @@ struct ucc_geth_private {
1184 struct ucc_geth_info *ug_info; 1184 struct ucc_geth_info *ug_info;
1185 struct ucc_fast_private *uccf; 1185 struct ucc_fast_private *uccf;
1186 struct net_device *dev; 1186 struct net_device *dev;
1187 struct napi_struct napi;
1187 struct net_device_stats stats; /* linux network statistics */ 1188 struct net_device_stats stats; /* linux network statistics */
1188 struct ucc_geth *ug_regs; 1189 struct ucc_geth *ug_regs;
1189 struct ucc_geth_init_pram *p_init_enet_param_shadow; 1190 struct ucc_geth_init_pram *p_init_enet_param_shadow;
diff --git a/drivers/net/via-rhine.c b/drivers/net/via-rhine.c
index b56dff26772d..7a5899059c44 100644
--- a/drivers/net/via-rhine.c
+++ b/drivers/net/via-rhine.c
@@ -389,6 +389,8 @@ struct rhine_private {
389 389
390 struct pci_dev *pdev; 390 struct pci_dev *pdev;
391 long pioaddr; 391 long pioaddr;
392 struct net_device *dev;
393 struct napi_struct napi;
392 struct net_device_stats stats; 394 struct net_device_stats stats;
393 spinlock_t lock; 395 spinlock_t lock;
394 396
@@ -582,28 +584,25 @@ static void rhine_poll(struct net_device *dev)
582#endif 584#endif
583 585
584#ifdef CONFIG_VIA_RHINE_NAPI 586#ifdef CONFIG_VIA_RHINE_NAPI
585static int rhine_napipoll(struct net_device *dev, int *budget) 587static int rhine_napipoll(struct napi_struct *napi, int budget)
586{ 588{
587 struct rhine_private *rp = netdev_priv(dev); 589 struct rhine_private *rp = container_of(napi, struct rhine_private, napi);
590 struct net_device *dev = rp->dev;
588 void __iomem *ioaddr = rp->base; 591 void __iomem *ioaddr = rp->base;
589 int done, limit = min(dev->quota, *budget); 592 int work_done;
590 593
591 done = rhine_rx(dev, limit); 594 work_done = rhine_rx(dev, budget);
592 *budget -= done;
593 dev->quota -= done;
594 595
595 if (done < limit) { 596 if (work_done < budget) {
596 netif_rx_complete(dev); 597 netif_rx_complete(dev, napi);
597 598
598 iowrite16(IntrRxDone | IntrRxErr | IntrRxEmpty| IntrRxOverflow | 599 iowrite16(IntrRxDone | IntrRxErr | IntrRxEmpty| IntrRxOverflow |
599 IntrRxDropped | IntrRxNoBuf | IntrTxAborted | 600 IntrRxDropped | IntrRxNoBuf | IntrTxAborted |
600 IntrTxDone | IntrTxError | IntrTxUnderrun | 601 IntrTxDone | IntrTxError | IntrTxUnderrun |
601 IntrPCIErr | IntrStatsMax | IntrLinkChange, 602 IntrPCIErr | IntrStatsMax | IntrLinkChange,
602 ioaddr + IntrEnable); 603 ioaddr + IntrEnable);
603 return 0;
604 } 604 }
605 else 605 return work_done;
606 return 1;
607} 606}
608#endif 607#endif
609 608
@@ -707,6 +706,7 @@ static int __devinit rhine_init_one(struct pci_dev *pdev,
707 SET_NETDEV_DEV(dev, &pdev->dev); 706 SET_NETDEV_DEV(dev, &pdev->dev);
708 707
709 rp = netdev_priv(dev); 708 rp = netdev_priv(dev);
709 rp->dev = dev;
710 rp->quirks = quirks; 710 rp->quirks = quirks;
711 rp->pioaddr = pioaddr; 711 rp->pioaddr = pioaddr;
712 rp->pdev = pdev; 712 rp->pdev = pdev;
@@ -785,8 +785,7 @@ static int __devinit rhine_init_one(struct pci_dev *pdev,
785 dev->poll_controller = rhine_poll; 785 dev->poll_controller = rhine_poll;
786#endif 786#endif
787#ifdef CONFIG_VIA_RHINE_NAPI 787#ifdef CONFIG_VIA_RHINE_NAPI
788 dev->poll = rhine_napipoll; 788 netif_napi_add(dev, &rp->napi, rhine_napipoll, 64);
789 dev->weight = 64;
790#endif 789#endif
791 if (rp->quirks & rqRhineI) 790 if (rp->quirks & rqRhineI)
792 dev->features |= NETIF_F_SG|NETIF_F_HW_CSUM; 791 dev->features |= NETIF_F_SG|NETIF_F_HW_CSUM;
@@ -1061,7 +1060,9 @@ static void init_registers(struct net_device *dev)
1061 1060
1062 rhine_set_rx_mode(dev); 1061 rhine_set_rx_mode(dev);
1063 1062
1064 netif_poll_enable(dev); 1063#ifdef CONFIG_VIA_RHINE_NAPI
1064 napi_enable(&rp->napi);
1065#endif
1065 1066
1066 /* Enable interrupts by setting the interrupt mask. */ 1067 /* Enable interrupts by setting the interrupt mask. */
1067 iowrite16(IntrRxDone | IntrRxErr | IntrRxEmpty| IntrRxOverflow | 1068 iowrite16(IntrRxDone | IntrRxErr | IntrRxEmpty| IntrRxOverflow |
@@ -1196,6 +1197,10 @@ static void rhine_tx_timeout(struct net_device *dev)
1196 /* protect against concurrent rx interrupts */ 1197 /* protect against concurrent rx interrupts */
1197 disable_irq(rp->pdev->irq); 1198 disable_irq(rp->pdev->irq);
1198 1199
1200#ifdef CONFIG_VIA_RHINE_NAPI
1201 napi_disable(&rp->napi);
1202#endif
1203
1199 spin_lock(&rp->lock); 1204 spin_lock(&rp->lock);
1200 1205
1201 /* clear all descriptors */ 1206 /* clear all descriptors */
@@ -1324,7 +1329,7 @@ static irqreturn_t rhine_interrupt(int irq, void *dev_instance)
1324 IntrPCIErr | IntrStatsMax | IntrLinkChange, 1329 IntrPCIErr | IntrStatsMax | IntrLinkChange,
1325 ioaddr + IntrEnable); 1330 ioaddr + IntrEnable);
1326 1331
1327 netif_rx_schedule(dev); 1332 netif_rx_schedule(dev, &rp->napi);
1328#else 1333#else
1329 rhine_rx(dev, RX_RING_SIZE); 1334 rhine_rx(dev, RX_RING_SIZE);
1330#endif 1335#endif
@@ -1837,7 +1842,9 @@ static int rhine_close(struct net_device *dev)
1837 spin_lock_irq(&rp->lock); 1842 spin_lock_irq(&rp->lock);
1838 1843
1839 netif_stop_queue(dev); 1844 netif_stop_queue(dev);
1840 netif_poll_disable(dev); 1845#ifdef CONFIG_VIA_RHINE_NAPI
1846 napi_disable(&rp->napi);
1847#endif
1841 1848
1842 if (debug > 1) 1849 if (debug > 1)
1843 printk(KERN_DEBUG "%s: Shutting down ethercard, " 1850 printk(KERN_DEBUG "%s: Shutting down ethercard, "
@@ -1936,6 +1943,9 @@ static int rhine_suspend(struct pci_dev *pdev, pm_message_t state)
1936 if (!netif_running(dev)) 1943 if (!netif_running(dev))
1937 return 0; 1944 return 0;
1938 1945
1946#ifdef CONFIG_VIA_RHINE_NAPI
1947 napi_disable(&rp->napi);
1948#endif
1939 netif_device_detach(dev); 1949 netif_device_detach(dev);
1940 pci_save_state(pdev); 1950 pci_save_state(pdev);
1941 1951
diff --git a/drivers/net/xen-netfront.c b/drivers/net/xen-netfront.c
index 4445810335a8..70e551c19e3a 100644
--- a/drivers/net/xen-netfront.c
+++ b/drivers/net/xen-netfront.c
@@ -72,6 +72,7 @@ struct netfront_info {
72 struct list_head list; 72 struct list_head list;
73 struct net_device *netdev; 73 struct net_device *netdev;
74 74
75 struct napi_struct napi;
75 struct net_device_stats stats; 76 struct net_device_stats stats;
76 77
77 struct xen_netif_tx_front_ring tx; 78 struct xen_netif_tx_front_ring tx;
@@ -185,7 +186,8 @@ static int xennet_can_sg(struct net_device *dev)
185static void rx_refill_timeout(unsigned long data) 186static void rx_refill_timeout(unsigned long data)
186{ 187{
187 struct net_device *dev = (struct net_device *)data; 188 struct net_device *dev = (struct net_device *)data;
188 netif_rx_schedule(dev); 189 struct netfront_info *np = netdev_priv(dev);
190 netif_rx_schedule(dev, &np->napi);
189} 191}
190 192
191static int netfront_tx_slot_available(struct netfront_info *np) 193static int netfront_tx_slot_available(struct netfront_info *np)
@@ -342,12 +344,14 @@ static int xennet_open(struct net_device *dev)
342 344
343 memset(&np->stats, 0, sizeof(np->stats)); 345 memset(&np->stats, 0, sizeof(np->stats));
344 346
347 napi_enable(&np->napi);
348
345 spin_lock_bh(&np->rx_lock); 349 spin_lock_bh(&np->rx_lock);
346 if (netif_carrier_ok(dev)) { 350 if (netif_carrier_ok(dev)) {
347 xennet_alloc_rx_buffers(dev); 351 xennet_alloc_rx_buffers(dev);
348 np->rx.sring->rsp_event = np->rx.rsp_cons + 1; 352 np->rx.sring->rsp_event = np->rx.rsp_cons + 1;
349 if (RING_HAS_UNCONSUMED_RESPONSES(&np->rx)) 353 if (RING_HAS_UNCONSUMED_RESPONSES(&np->rx))
350 netif_rx_schedule(dev); 354 netif_rx_schedule(dev, &np->napi);
351 } 355 }
352 spin_unlock_bh(&np->rx_lock); 356 spin_unlock_bh(&np->rx_lock);
353 357
@@ -589,6 +593,7 @@ static int xennet_close(struct net_device *dev)
589{ 593{
590 struct netfront_info *np = netdev_priv(dev); 594 struct netfront_info *np = netdev_priv(dev);
591 netif_stop_queue(np->netdev); 595 netif_stop_queue(np->netdev);
596 napi_disable(&np->napi);
592 return 0; 597 return 0;
593} 598}
594 599
@@ -872,15 +877,16 @@ static int handle_incoming_queue(struct net_device *dev,
872 return packets_dropped; 877 return packets_dropped;
873} 878}
874 879
875static int xennet_poll(struct net_device *dev, int *pbudget) 880static int xennet_poll(struct napi_struct *napi, int budget)
876{ 881{
877 struct netfront_info *np = netdev_priv(dev); 882 struct netfront_info *np = container_of(napi, struct netfront_info, napi);
883 struct net_device *dev = np->netdev;
878 struct sk_buff *skb; 884 struct sk_buff *skb;
879 struct netfront_rx_info rinfo; 885 struct netfront_rx_info rinfo;
880 struct xen_netif_rx_response *rx = &rinfo.rx; 886 struct xen_netif_rx_response *rx = &rinfo.rx;
881 struct xen_netif_extra_info *extras = rinfo.extras; 887 struct xen_netif_extra_info *extras = rinfo.extras;
882 RING_IDX i, rp; 888 RING_IDX i, rp;
883 int work_done, budget, more_to_do = 1; 889 int work_done;
884 struct sk_buff_head rxq; 890 struct sk_buff_head rxq;
885 struct sk_buff_head errq; 891 struct sk_buff_head errq;
886 struct sk_buff_head tmpq; 892 struct sk_buff_head tmpq;
@@ -899,9 +905,6 @@ static int xennet_poll(struct net_device *dev, int *pbudget)
899 skb_queue_head_init(&errq); 905 skb_queue_head_init(&errq);
900 skb_queue_head_init(&tmpq); 906 skb_queue_head_init(&tmpq);
901 907
902 budget = *pbudget;
903 if (budget > dev->quota)
904 budget = dev->quota;
905 rp = np->rx.sring->rsp_prod; 908 rp = np->rx.sring->rsp_prod;
906 rmb(); /* Ensure we see queued responses up to 'rp'. */ 909 rmb(); /* Ensure we see queued responses up to 'rp'. */
907 910
@@ -1006,22 +1009,21 @@ err:
1006 1009
1007 xennet_alloc_rx_buffers(dev); 1010 xennet_alloc_rx_buffers(dev);
1008 1011
1009 *pbudget -= work_done;
1010 dev->quota -= work_done;
1011
1012 if (work_done < budget) { 1012 if (work_done < budget) {
1013 int more_to_do = 0;
1014
1013 local_irq_save(flags); 1015 local_irq_save(flags);
1014 1016
1015 RING_FINAL_CHECK_FOR_RESPONSES(&np->rx, more_to_do); 1017 RING_FINAL_CHECK_FOR_RESPONSES(&np->rx, more_to_do);
1016 if (!more_to_do) 1018 if (!more_to_do)
1017 __netif_rx_complete(dev); 1019 __netif_rx_complete(dev, napi);
1018 1020
1019 local_irq_restore(flags); 1021 local_irq_restore(flags);
1020 } 1022 }
1021 1023
1022 spin_unlock(&np->rx_lock); 1024 spin_unlock(&np->rx_lock);
1023 1025
1024 return more_to_do; 1026 return work_done;
1025} 1027}
1026 1028
1027static int xennet_change_mtu(struct net_device *dev, int mtu) 1029static int xennet_change_mtu(struct net_device *dev, int mtu)
@@ -1201,10 +1203,9 @@ static struct net_device * __devinit xennet_create_dev(struct xenbus_device *dev
1201 netdev->hard_start_xmit = xennet_start_xmit; 1203 netdev->hard_start_xmit = xennet_start_xmit;
1202 netdev->stop = xennet_close; 1204 netdev->stop = xennet_close;
1203 netdev->get_stats = xennet_get_stats; 1205 netdev->get_stats = xennet_get_stats;
1204 netdev->poll = xennet_poll; 1206 netif_napi_add(netdev, &np->napi, xennet_poll, 64);
1205 netdev->uninit = xennet_uninit; 1207 netdev->uninit = xennet_uninit;
1206 netdev->change_mtu = xennet_change_mtu; 1208 netdev->change_mtu = xennet_change_mtu;
1207 netdev->weight = 64;
1208 netdev->features = NETIF_F_IP_CSUM; 1209 netdev->features = NETIF_F_IP_CSUM;
1209 1210
1210 SET_ETHTOOL_OPS(netdev, &xennet_ethtool_ops); 1211 SET_ETHTOOL_OPS(netdev, &xennet_ethtool_ops);
@@ -1349,7 +1350,7 @@ static irqreturn_t xennet_interrupt(int irq, void *dev_id)
1349 xennet_tx_buf_gc(dev); 1350 xennet_tx_buf_gc(dev);
1350 /* Under tx_lock: protects access to rx shared-ring indexes. */ 1351 /* Under tx_lock: protects access to rx shared-ring indexes. */
1351 if (RING_HAS_UNCONSUMED_RESPONSES(&np->rx)) 1352 if (RING_HAS_UNCONSUMED_RESPONSES(&np->rx))
1352 netif_rx_schedule(dev); 1353 netif_rx_schedule(dev, &np->napi);
1353 } 1354 }
1354 1355
1355 spin_unlock_irqrestore(&np->tx_lock, flags); 1356 spin_unlock_irqrestore(&np->tx_lock, flags);
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h
index e679b2751665..b93575db8cce 100644
--- a/include/linux/netdevice.h
+++ b/include/linux/netdevice.h
@@ -31,6 +31,7 @@
31 31
32#ifdef __KERNEL__ 32#ifdef __KERNEL__
33#include <linux/timer.h> 33#include <linux/timer.h>
34#include <linux/delay.h>
34#include <asm/atomic.h> 35#include <asm/atomic.h>
35#include <asm/cache.h> 36#include <asm/cache.h>
36#include <asm/byteorder.h> 37#include <asm/byteorder.h>
@@ -38,6 +39,7 @@
38#include <linux/device.h> 39#include <linux/device.h>
39#include <linux/percpu.h> 40#include <linux/percpu.h>
40#include <linux/dmaengine.h> 41#include <linux/dmaengine.h>
42#include <linux/workqueue.h>
41 43
42struct vlan_group; 44struct vlan_group;
43struct ethtool_ops; 45struct ethtool_ops;
@@ -258,7 +260,6 @@ enum netdev_state_t
258 __LINK_STATE_PRESENT, 260 __LINK_STATE_PRESENT,
259 __LINK_STATE_SCHED, 261 __LINK_STATE_SCHED,
260 __LINK_STATE_NOCARRIER, 262 __LINK_STATE_NOCARRIER,
261 __LINK_STATE_RX_SCHED,
262 __LINK_STATE_LINKWATCH_PENDING, 263 __LINK_STATE_LINKWATCH_PENDING,
263 __LINK_STATE_DORMANT, 264 __LINK_STATE_DORMANT,
264 __LINK_STATE_QDISC_RUNNING, 265 __LINK_STATE_QDISC_RUNNING,
@@ -278,6 +279,110 @@ struct netdev_boot_setup {
278extern int __init netdev_boot_setup(char *str); 279extern int __init netdev_boot_setup(char *str);
279 280
280/* 281/*
282 * Structure for NAPI scheduling similar to tasklet but with weighting
283 */
284struct napi_struct {
285 /* The poll_list must only be managed by the entity which
286 * changes the state of the NAPI_STATE_SCHED bit. This means
287 * whoever atomically sets that bit can add this napi_struct
288 * to the per-cpu poll_list, and whoever clears that bit
289 * can remove from the list right before clearing the bit.
290 */
291 struct list_head poll_list;
292
293 unsigned long state;
294 int weight;
295 int (*poll)(struct napi_struct *, int);
296#ifdef CONFIG_NETPOLL
297 spinlock_t poll_lock;
298 int poll_owner;
299 struct net_device *dev;
300 struct list_head dev_list;
301#endif
302};
303
304enum
305{
306 NAPI_STATE_SCHED, /* Poll is scheduled */
307};
308
309extern void FASTCALL(__napi_schedule(struct napi_struct *n));
310
311/**
312 * napi_schedule_prep - check if napi can be scheduled
313 * @n: napi context
314 *
315 * Test if NAPI routine is already running, and if not mark
316 * it as running. This is used as a condition variable
317 * insure only one NAPI poll instance runs
318 */
319static inline int napi_schedule_prep(struct napi_struct *n)
320{
321 return !test_and_set_bit(NAPI_STATE_SCHED, &n->state);
322}
323
324/**
325 * napi_schedule - schedule NAPI poll
326 * @n: napi context
327 *
328 * Schedule NAPI poll routine to be called if it is not already
329 * running.
330 */
331static inline void napi_schedule(struct napi_struct *n)
332{
333 if (napi_schedule_prep(n))
334 __napi_schedule(n);
335}
336
337/**
338 * napi_complete - NAPI processing complete
339 * @n: napi context
340 *
341 * Mark NAPI processing as complete.
342 */
343static inline void __napi_complete(struct napi_struct *n)
344{
345 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
346 list_del(&n->poll_list);
347 smp_mb__before_clear_bit();
348 clear_bit(NAPI_STATE_SCHED, &n->state);
349}
350
351static inline void napi_complete(struct napi_struct *n)
352{
353 local_irq_disable();
354 __napi_complete(n);
355 local_irq_enable();
356}
357
358/**
359 * napi_disable - prevent NAPI from scheduling
360 * @n: napi context
361 *
362 * Stop NAPI from being scheduled on this context.
363 * Waits till any outstanding processing completes.
364 */
365static inline void napi_disable(struct napi_struct *n)
366{
367 while (test_and_set_bit(NAPI_STATE_SCHED, &n->state))
368 msleep_interruptible(1);
369}
370
371/**
372 * napi_enable - enable NAPI scheduling
373 * @n: napi context
374 *
375 * Resume NAPI from being scheduled on this context.
376 * Must be paired with napi_disable.
377 */
378static inline void napi_enable(struct napi_struct *n)
379{
380 BUG_ON(!test_bit(NAPI_STATE_SCHED, &n->state));
381 smp_mb__before_clear_bit();
382 clear_bit(NAPI_STATE_SCHED, &n->state);
383}
384
385/*
281 * The DEVICE structure. 386 * The DEVICE structure.
282 * Actually, this whole structure is a big mistake. It mixes I/O 387 * Actually, this whole structure is a big mistake. It mixes I/O
283 * data with strictly "high-level" data, and it has to know about 388 * data with strictly "high-level" data, and it has to know about
@@ -319,6 +424,9 @@ struct net_device
319 unsigned long state; 424 unsigned long state;
320 425
321 struct list_head dev_list; 426 struct list_head dev_list;
427#ifdef CONFIG_NETPOLL
428 struct list_head napi_list;
429#endif
322 430
323 /* The device initialization function. Called only once. */ 431 /* The device initialization function. Called only once. */
324 int (*init)(struct net_device *dev); 432 int (*init)(struct net_device *dev);
@@ -430,12 +538,6 @@ struct net_device
430/* 538/*
431 * Cache line mostly used on receive path (including eth_type_trans()) 539 * Cache line mostly used on receive path (including eth_type_trans())
432 */ 540 */
433 struct list_head poll_list ____cacheline_aligned_in_smp;
434 /* Link to poll list */
435
436 int (*poll) (struct net_device *dev, int *quota);
437 int quota;
438 int weight;
439 unsigned long last_rx; /* Time of last Rx */ 541 unsigned long last_rx; /* Time of last Rx */
440 /* Interface address info used in eth_type_trans() */ 542 /* Interface address info used in eth_type_trans() */
441 unsigned char dev_addr[MAX_ADDR_LEN]; /* hw address, (before bcast 543 unsigned char dev_addr[MAX_ADDR_LEN]; /* hw address, (before bcast
@@ -582,6 +684,12 @@ struct net_device
582#define NETDEV_ALIGN 32 684#define NETDEV_ALIGN 32
583#define NETDEV_ALIGN_CONST (NETDEV_ALIGN - 1) 685#define NETDEV_ALIGN_CONST (NETDEV_ALIGN - 1)
584 686
687/**
688 * netdev_priv - access network device private data
689 * @dev: network device
690 *
691 * Get network device private data
692 */
585static inline void *netdev_priv(const struct net_device *dev) 693static inline void *netdev_priv(const struct net_device *dev)
586{ 694{
587 return dev->priv; 695 return dev->priv;
@@ -593,6 +701,23 @@ static inline void *netdev_priv(const struct net_device *dev)
593 */ 701 */
594#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev)) 702#define SET_NETDEV_DEV(net, pdev) ((net)->dev.parent = (pdev))
595 703
704static inline void netif_napi_add(struct net_device *dev,
705 struct napi_struct *napi,
706 int (*poll)(struct napi_struct *, int),
707 int weight)
708{
709 INIT_LIST_HEAD(&napi->poll_list);
710 napi->poll = poll;
711 napi->weight = weight;
712#ifdef CONFIG_NETPOLL
713 napi->dev = dev;
714 list_add(&napi->dev_list, &dev->napi_list);
715 spin_lock_init(&napi->poll_lock);
716 napi->poll_owner = -1;
717#endif
718 set_bit(NAPI_STATE_SCHED, &napi->state);
719}
720
596struct packet_type { 721struct packet_type {
597 __be16 type; /* This is really htons(ether_type). */ 722 __be16 type; /* This is really htons(ether_type). */
598 struct net_device *dev; /* NULL is wildcarded here */ 723 struct net_device *dev; /* NULL is wildcarded here */
@@ -678,7 +803,6 @@ static inline int unregister_gifconf(unsigned int family)
678 * Incoming packets are placed on per-cpu queues so that 803 * Incoming packets are placed on per-cpu queues so that
679 * no locking is needed. 804 * no locking is needed.
680 */ 805 */
681
682struct softnet_data 806struct softnet_data
683{ 807{
684 struct net_device *output_queue; 808 struct net_device *output_queue;
@@ -686,7 +810,7 @@ struct softnet_data
686 struct list_head poll_list; 810 struct list_head poll_list;
687 struct sk_buff *completion_queue; 811 struct sk_buff *completion_queue;
688 812
689 struct net_device backlog_dev; /* Sorry. 8) */ 813 struct napi_struct backlog;
690#ifdef CONFIG_NET_DMA 814#ifdef CONFIG_NET_DMA
691 struct dma_chan *net_dma; 815 struct dma_chan *net_dma;
692#endif 816#endif
@@ -704,11 +828,24 @@ static inline void netif_schedule(struct net_device *dev)
704 __netif_schedule(dev); 828 __netif_schedule(dev);
705} 829}
706 830
831/**
832 * netif_start_queue - allow transmit
833 * @dev: network device
834 *
835 * Allow upper layers to call the device hard_start_xmit routine.
836 */
707static inline void netif_start_queue(struct net_device *dev) 837static inline void netif_start_queue(struct net_device *dev)
708{ 838{
709 clear_bit(__LINK_STATE_XOFF, &dev->state); 839 clear_bit(__LINK_STATE_XOFF, &dev->state);
710} 840}
711 841
842/**
843 * netif_wake_queue - restart transmit
844 * @dev: network device
845 *
846 * Allow upper layers to call the device hard_start_xmit routine.
847 * Used for flow control when transmit resources are available.
848 */
712static inline void netif_wake_queue(struct net_device *dev) 849static inline void netif_wake_queue(struct net_device *dev)
713{ 850{
714#ifdef CONFIG_NETPOLL_TRAP 851#ifdef CONFIG_NETPOLL_TRAP
@@ -721,16 +858,35 @@ static inline void netif_wake_queue(struct net_device *dev)
721 __netif_schedule(dev); 858 __netif_schedule(dev);
722} 859}
723 860
861/**
862 * netif_stop_queue - stop transmitted packets
863 * @dev: network device
864 *
865 * Stop upper layers calling the device hard_start_xmit routine.
866 * Used for flow control when transmit resources are unavailable.
867 */
724static inline void netif_stop_queue(struct net_device *dev) 868static inline void netif_stop_queue(struct net_device *dev)
725{ 869{
726 set_bit(__LINK_STATE_XOFF, &dev->state); 870 set_bit(__LINK_STATE_XOFF, &dev->state);
727} 871}
728 872
873/**
874 * netif_queue_stopped - test if transmit queue is flowblocked
875 * @dev: network device
876 *
877 * Test if transmit queue on device is currently unable to send.
878 */
729static inline int netif_queue_stopped(const struct net_device *dev) 879static inline int netif_queue_stopped(const struct net_device *dev)
730{ 880{
731 return test_bit(__LINK_STATE_XOFF, &dev->state); 881 return test_bit(__LINK_STATE_XOFF, &dev->state);
732} 882}
733 883
884/**
885 * netif_running - test if up
886 * @dev: network device
887 *
888 * Test if the device has been brought up.
889 */
734static inline int netif_running(const struct net_device *dev) 890static inline int netif_running(const struct net_device *dev)
735{ 891{
736 return test_bit(__LINK_STATE_START, &dev->state); 892 return test_bit(__LINK_STATE_START, &dev->state);
@@ -742,6 +898,14 @@ static inline int netif_running(const struct net_device *dev)
742 * done at the overall netdevice level. 898 * done at the overall netdevice level.
743 * Also test the device if we're multiqueue. 899 * Also test the device if we're multiqueue.
744 */ 900 */
901
902/**
903 * netif_start_subqueue - allow sending packets on subqueue
904 * @dev: network device
905 * @queue_index: sub queue index
906 *
907 * Start individual transmit queue of a device with multiple transmit queues.
908 */
745static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index) 909static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
746{ 910{
747#ifdef CONFIG_NETDEVICES_MULTIQUEUE 911#ifdef CONFIG_NETDEVICES_MULTIQUEUE
@@ -749,6 +913,13 @@ static inline void netif_start_subqueue(struct net_device *dev, u16 queue_index)
749#endif 913#endif
750} 914}
751 915
916/**
917 * netif_stop_subqueue - stop sending packets on subqueue
918 * @dev: network device
919 * @queue_index: sub queue index
920 *
921 * Stop individual transmit queue of a device with multiple transmit queues.
922 */
752static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index) 923static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
753{ 924{
754#ifdef CONFIG_NETDEVICES_MULTIQUEUE 925#ifdef CONFIG_NETDEVICES_MULTIQUEUE
@@ -760,6 +931,13 @@ static inline void netif_stop_subqueue(struct net_device *dev, u16 queue_index)
760#endif 931#endif
761} 932}
762 933
934/**
935 * netif_subqueue_stopped - test status of subqueue
936 * @dev: network device
937 * @queue_index: sub queue index
938 *
939 * Check individual transmit queue of a device with multiple transmit queues.
940 */
763static inline int netif_subqueue_stopped(const struct net_device *dev, 941static inline int netif_subqueue_stopped(const struct net_device *dev,
764 u16 queue_index) 942 u16 queue_index)
765{ 943{
@@ -771,6 +949,14 @@ static inline int netif_subqueue_stopped(const struct net_device *dev,
771#endif 949#endif
772} 950}
773 951
952
953/**
954 * netif_wake_subqueue - allow sending packets on subqueue
955 * @dev: network device
956 * @queue_index: sub queue index
957 *
958 * Resume individual transmit queue of a device with multiple transmit queues.
959 */
774static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index) 960static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
775{ 961{
776#ifdef CONFIG_NETDEVICES_MULTIQUEUE 962#ifdef CONFIG_NETDEVICES_MULTIQUEUE
@@ -784,6 +970,13 @@ static inline void netif_wake_subqueue(struct net_device *dev, u16 queue_index)
784#endif 970#endif
785} 971}
786 972
973/**
974 * netif_is_multiqueue - test if device has multiple transmit queues
975 * @dev: network device
976 *
977 * Check if device has multiple transmit queues
978 * Always falls if NETDEVICE_MULTIQUEUE is not configured
979 */
787static inline int netif_is_multiqueue(const struct net_device *dev) 980static inline int netif_is_multiqueue(const struct net_device *dev)
788{ 981{
789#ifdef CONFIG_NETDEVICES_MULTIQUEUE 982#ifdef CONFIG_NETDEVICES_MULTIQUEUE
@@ -796,20 +989,7 @@ static inline int netif_is_multiqueue(const struct net_device *dev)
796/* Use this variant when it is known for sure that it 989/* Use this variant when it is known for sure that it
797 * is executing from interrupt context. 990 * is executing from interrupt context.
798 */ 991 */
799static inline void dev_kfree_skb_irq(struct sk_buff *skb) 992extern void dev_kfree_skb_irq(struct sk_buff *skb);
800{
801 if (atomic_dec_and_test(&skb->users)) {
802 struct softnet_data *sd;
803 unsigned long flags;
804
805 local_irq_save(flags);
806 sd = &__get_cpu_var(softnet_data);
807 skb->next = sd->completion_queue;
808 sd->completion_queue = skb;
809 raise_softirq_irqoff(NET_TX_SOFTIRQ);
810 local_irq_restore(flags);
811 }
812}
813 993
814/* Use this variant in places where it could be invoked 994/* Use this variant in places where it could be invoked
815 * either from interrupt or non-interrupt context. 995 * either from interrupt or non-interrupt context.
@@ -833,18 +1013,28 @@ extern int dev_set_mac_address(struct net_device *,
833extern int dev_hard_start_xmit(struct sk_buff *skb, 1013extern int dev_hard_start_xmit(struct sk_buff *skb,
834 struct net_device *dev); 1014 struct net_device *dev);
835 1015
836extern void dev_init(void);
837
838extern int netdev_budget; 1016extern int netdev_budget;
839 1017
840/* Called by rtnetlink.c:rtnl_unlock() */ 1018/* Called by rtnetlink.c:rtnl_unlock() */
841extern void netdev_run_todo(void); 1019extern void netdev_run_todo(void);
842 1020
1021/**
1022 * dev_put - release reference to device
1023 * @dev: network device
1024 *
1025 * Hold reference to device to keep it from being freed.
1026 */
843static inline void dev_put(struct net_device *dev) 1027static inline void dev_put(struct net_device *dev)
844{ 1028{
845 atomic_dec(&dev->refcnt); 1029 atomic_dec(&dev->refcnt);
846} 1030}
847 1031
1032/**
1033 * dev_hold - get reference to device
1034 * @dev: network device
1035 *
1036 * Release reference to device to allow it to be freed.
1037 */
848static inline void dev_hold(struct net_device *dev) 1038static inline void dev_hold(struct net_device *dev)
849{ 1039{
850 atomic_inc(&dev->refcnt); 1040 atomic_inc(&dev->refcnt);
@@ -861,6 +1051,12 @@ static inline void dev_hold(struct net_device *dev)
861 1051
862extern void linkwatch_fire_event(struct net_device *dev); 1052extern void linkwatch_fire_event(struct net_device *dev);
863 1053
1054/**
1055 * netif_carrier_ok - test if carrier present
1056 * @dev: network device
1057 *
1058 * Check if carrier is present on device
1059 */
864static inline int netif_carrier_ok(const struct net_device *dev) 1060static inline int netif_carrier_ok(const struct net_device *dev)
865{ 1061{
866 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state); 1062 return !test_bit(__LINK_STATE_NOCARRIER, &dev->state);
@@ -872,30 +1068,66 @@ extern void netif_carrier_on(struct net_device *dev);
872 1068
873extern void netif_carrier_off(struct net_device *dev); 1069extern void netif_carrier_off(struct net_device *dev);
874 1070
1071/**
1072 * netif_dormant_on - mark device as dormant.
1073 * @dev: network device
1074 *
1075 * Mark device as dormant (as per RFC2863).
1076 *
1077 * The dormant state indicates that the relevant interface is not
1078 * actually in a condition to pass packets (i.e., it is not 'up') but is
1079 * in a "pending" state, waiting for some external event. For "on-
1080 * demand" interfaces, this new state identifies the situation where the
1081 * interface is waiting for events to place it in the up state.
1082 *
1083 */
875static inline void netif_dormant_on(struct net_device *dev) 1084static inline void netif_dormant_on(struct net_device *dev)
876{ 1085{
877 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state)) 1086 if (!test_and_set_bit(__LINK_STATE_DORMANT, &dev->state))
878 linkwatch_fire_event(dev); 1087 linkwatch_fire_event(dev);
879} 1088}
880 1089
1090/**
1091 * netif_dormant_off - set device as not dormant.
1092 * @dev: network device
1093 *
1094 * Device is not in dormant state.
1095 */
881static inline void netif_dormant_off(struct net_device *dev) 1096static inline void netif_dormant_off(struct net_device *dev)
882{ 1097{
883 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state)) 1098 if (test_and_clear_bit(__LINK_STATE_DORMANT, &dev->state))
884 linkwatch_fire_event(dev); 1099 linkwatch_fire_event(dev);
885} 1100}
886 1101
1102/**
1103 * netif_dormant - test if carrier present
1104 * @dev: network device
1105 *
1106 * Check if carrier is present on device
1107 */
887static inline int netif_dormant(const struct net_device *dev) 1108static inline int netif_dormant(const struct net_device *dev)
888{ 1109{
889 return test_bit(__LINK_STATE_DORMANT, &dev->state); 1110 return test_bit(__LINK_STATE_DORMANT, &dev->state);
890} 1111}
891 1112
892 1113
1114/**
1115 * netif_oper_up - test if device is operational
1116 * @dev: network device
1117 *
1118 * Check if carrier is operational
1119 */
893static inline int netif_oper_up(const struct net_device *dev) { 1120static inline int netif_oper_up(const struct net_device *dev) {
894 return (dev->operstate == IF_OPER_UP || 1121 return (dev->operstate == IF_OPER_UP ||
895 dev->operstate == IF_OPER_UNKNOWN /* backward compat */); 1122 dev->operstate == IF_OPER_UNKNOWN /* backward compat */);
896} 1123}
897 1124
898/* Hot-plugging. */ 1125/**
1126 * netif_device_present - is device available or removed
1127 * @dev: network device
1128 *
1129 * Check if device has not been removed from system.
1130 */
899static inline int netif_device_present(struct net_device *dev) 1131static inline int netif_device_present(struct net_device *dev)
900{ 1132{
901 return test_bit(__LINK_STATE_PRESENT, &dev->state); 1133 return test_bit(__LINK_STATE_PRESENT, &dev->state);
@@ -955,46 +1187,38 @@ static inline u32 netif_msg_init(int debug_value, int default_msg_enable_bits)
955 return (1 << debug_value) - 1; 1187 return (1 << debug_value) - 1;
956} 1188}
957 1189
958/* Test if receive needs to be scheduled */
959static inline int __netif_rx_schedule_prep(struct net_device *dev)
960{
961 return !test_and_set_bit(__LINK_STATE_RX_SCHED, &dev->state);
962}
963
964/* Test if receive needs to be scheduled but only if up */ 1190/* Test if receive needs to be scheduled but only if up */
965static inline int netif_rx_schedule_prep(struct net_device *dev) 1191static inline int netif_rx_schedule_prep(struct net_device *dev,
1192 struct napi_struct *napi)
966{ 1193{
967 return netif_running(dev) && __netif_rx_schedule_prep(dev); 1194 return netif_running(dev) && napi_schedule_prep(napi);
968} 1195}
969 1196
970/* Add interface to tail of rx poll list. This assumes that _prep has 1197/* Add interface to tail of rx poll list. This assumes that _prep has
971 * already been called and returned 1. 1198 * already been called and returned 1.
972 */ 1199 */
973 1200static inline void __netif_rx_schedule(struct net_device *dev,
974extern void __netif_rx_schedule(struct net_device *dev); 1201 struct napi_struct *napi)
1202{
1203 dev_hold(dev);
1204 __napi_schedule(napi);
1205}
975 1206
976/* Try to reschedule poll. Called by irq handler. */ 1207/* Try to reschedule poll. Called by irq handler. */
977 1208
978static inline void netif_rx_schedule(struct net_device *dev) 1209static inline void netif_rx_schedule(struct net_device *dev,
1210 struct napi_struct *napi)
979{ 1211{
980 if (netif_rx_schedule_prep(dev)) 1212 if (netif_rx_schedule_prep(dev, napi))
981 __netif_rx_schedule(dev); 1213 __netif_rx_schedule(dev, napi);
982} 1214}
983 1215
984/* Try to reschedule poll. Called by dev->poll() after netif_rx_complete(). 1216/* Try to reschedule poll. Called by dev->poll() after netif_rx_complete(). */
985 * Do not inline this? 1217static inline int netif_rx_reschedule(struct net_device *dev,
986 */ 1218 struct napi_struct *napi)
987static inline int netif_rx_reschedule(struct net_device *dev, int undo)
988{ 1219{
989 if (netif_rx_schedule_prep(dev)) { 1220 if (napi_schedule_prep(napi)) {
990 unsigned long flags; 1221 __netif_rx_schedule(dev, napi);
991
992 dev->quota += undo;
993
994 local_irq_save(flags);
995 list_add_tail(&dev->poll_list, &__get_cpu_var(softnet_data).poll_list);
996 __raise_softirq_irqoff(NET_RX_SOFTIRQ);
997 local_irq_restore(flags);
998 return 1; 1222 return 1;
999 } 1223 }
1000 return 0; 1224 return 0;
@@ -1003,12 +1227,11 @@ static inline int netif_rx_reschedule(struct net_device *dev, int undo)
1003/* same as netif_rx_complete, except that local_irq_save(flags) 1227/* same as netif_rx_complete, except that local_irq_save(flags)
1004 * has already been issued 1228 * has already been issued
1005 */ 1229 */
1006static inline void __netif_rx_complete(struct net_device *dev) 1230static inline void __netif_rx_complete(struct net_device *dev,
1231 struct napi_struct *napi)
1007{ 1232{
1008 BUG_ON(!test_bit(__LINK_STATE_RX_SCHED, &dev->state)); 1233 __napi_complete(napi);
1009 list_del(&dev->poll_list); 1234 dev_put(dev);
1010 smp_mb__before_clear_bit();
1011 clear_bit(__LINK_STATE_RX_SCHED, &dev->state);
1012} 1235}
1013 1236
1014/* Remove interface from poll list: it must be in the poll list 1237/* Remove interface from poll list: it must be in the poll list
@@ -1016,28 +1239,22 @@ static inline void __netif_rx_complete(struct net_device *dev)
1016 * it completes the work. The device cannot be out of poll list at this 1239 * it completes the work. The device cannot be out of poll list at this
1017 * moment, it is BUG(). 1240 * moment, it is BUG().
1018 */ 1241 */
1019static inline void netif_rx_complete(struct net_device *dev) 1242static inline void netif_rx_complete(struct net_device *dev,
1243 struct napi_struct *napi)
1020{ 1244{
1021 unsigned long flags; 1245 unsigned long flags;
1022 1246
1023 local_irq_save(flags); 1247 local_irq_save(flags);
1024 __netif_rx_complete(dev); 1248 __netif_rx_complete(dev, napi);
1025 local_irq_restore(flags); 1249 local_irq_restore(flags);
1026} 1250}
1027 1251
1028static inline void netif_poll_disable(struct net_device *dev) 1252/**
1029{ 1253 * netif_tx_lock - grab network device transmit lock
1030 while (test_and_set_bit(__LINK_STATE_RX_SCHED, &dev->state)) 1254 * @dev: network device
1031 /* No hurry. */ 1255 *
1032 schedule_timeout_interruptible(1); 1256 * Get network device transmit lock
1033} 1257 */
1034
1035static inline void netif_poll_enable(struct net_device *dev)
1036{
1037 smp_mb__before_clear_bit();
1038 clear_bit(__LINK_STATE_RX_SCHED, &dev->state);
1039}
1040
1041static inline void netif_tx_lock(struct net_device *dev) 1258static inline void netif_tx_lock(struct net_device *dev)
1042{ 1259{
1043 spin_lock(&dev->_xmit_lock); 1260 spin_lock(&dev->_xmit_lock);
diff --git a/include/linux/netpoll.h b/include/linux/netpoll.h
index 29930b71a9aa..08dcc39ec18d 100644
--- a/include/linux/netpoll.h
+++ b/include/linux/netpoll.h
@@ -25,8 +25,6 @@ struct netpoll {
25 25
26struct netpoll_info { 26struct netpoll_info {
27 atomic_t refcnt; 27 atomic_t refcnt;
28 spinlock_t poll_lock;
29 int poll_owner;
30 int rx_flags; 28 int rx_flags;
31 spinlock_t rx_lock; 29 spinlock_t rx_lock;
32 struct netpoll *rx_np; /* netpoll that registered an rx_hook */ 30 struct netpoll *rx_np; /* netpoll that registered an rx_hook */
@@ -64,32 +62,61 @@ static inline int netpoll_rx(struct sk_buff *skb)
64 return ret; 62 return ret;
65} 63}
66 64
67static inline void *netpoll_poll_lock(struct net_device *dev) 65static inline int netpoll_receive_skb(struct sk_buff *skb)
68{ 66{
67 if (!list_empty(&skb->dev->napi_list))
68 return netpoll_rx(skb);
69 return 0;
70}
71
72static inline void *netpoll_poll_lock(struct napi_struct *napi)
73{
74 struct net_device *dev = napi->dev;
75
69 rcu_read_lock(); /* deal with race on ->npinfo */ 76 rcu_read_lock(); /* deal with race on ->npinfo */
70 if (dev->npinfo) { 77 if (dev && dev->npinfo) {
71 spin_lock(&dev->npinfo->poll_lock); 78 spin_lock(&napi->poll_lock);
72 dev->npinfo->poll_owner = smp_processor_id(); 79 napi->poll_owner = smp_processor_id();
73 return dev->npinfo; 80 return napi;
74 } 81 }
75 return NULL; 82 return NULL;
76} 83}
77 84
78static inline void netpoll_poll_unlock(void *have) 85static inline void netpoll_poll_unlock(void *have)
79{ 86{
80 struct netpoll_info *npi = have; 87 struct napi_struct *napi = have;
81 88
82 if (npi) { 89 if (napi) {
83 npi->poll_owner = -1; 90 napi->poll_owner = -1;
84 spin_unlock(&npi->poll_lock); 91 spin_unlock(&napi->poll_lock);
85 } 92 }
86 rcu_read_unlock(); 93 rcu_read_unlock();
87} 94}
88 95
96static inline void netpoll_netdev_init(struct net_device *dev)
97{
98 INIT_LIST_HEAD(&dev->napi_list);
99}
100
89#else 101#else
90#define netpoll_rx(a) 0 102static inline int netpoll_rx(struct sk_buff *skb)
91#define netpoll_poll_lock(a) NULL 103{
92#define netpoll_poll_unlock(a) 104 return 0;
105}
106static inline int netpoll_receive_skb(struct sk_buff *skb)
107{
108 return 0;
109}
110static inline void *netpoll_poll_lock(struct napi_struct *napi)
111{
112 return NULL;
113}
114static inline void netpoll_poll_unlock(void *have)
115{
116}
117static inline void netpoll_netdev_init(struct net_device *dev)
118{
119}
93#endif 120#endif
94 121
95#endif 122#endif
diff --git a/net/core/dev.c b/net/core/dev.c
index a76021c71207..29cf00c5d865 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -220,7 +220,8 @@ static RAW_NOTIFIER_HEAD(netdev_chain);
220 * Device drivers call our routines to queue packets here. We empty the 220 * Device drivers call our routines to queue packets here. We empty the
221 * queue in the local softnet handler. 221 * queue in the local softnet handler.
222 */ 222 */
223DEFINE_PER_CPU(struct softnet_data, softnet_data) = { NULL }; 223
224DEFINE_PER_CPU(struct softnet_data, softnet_data);
224 225
225#ifdef CONFIG_SYSFS 226#ifdef CONFIG_SYSFS
226extern int netdev_sysfs_init(void); 227extern int netdev_sysfs_init(void);
@@ -1018,16 +1019,12 @@ int dev_close(struct net_device *dev)
1018 clear_bit(__LINK_STATE_START, &dev->state); 1019 clear_bit(__LINK_STATE_START, &dev->state);
1019 1020
1020 /* Synchronize to scheduled poll. We cannot touch poll list, 1021 /* Synchronize to scheduled poll. We cannot touch poll list,
1021 * it can be even on different cpu. So just clear netif_running(), 1022 * it can be even on different cpu. So just clear netif_running().
1022 * and wait when poll really will happen. Actually, the best place 1023 *
1023 * for this is inside dev->stop() after device stopped its irq 1024 * dev->stop() will invoke napi_disable() on all of it's
1024 * engine, but this requires more changes in devices. */ 1025 * napi_struct instances on this device.
1025 1026 */
1026 smp_mb__after_clear_bit(); /* Commit netif_running(). */ 1027 smp_mb__after_clear_bit(); /* Commit netif_running(). */
1027 while (test_bit(__LINK_STATE_RX_SCHED, &dev->state)) {
1028 /* No hurry. */
1029 msleep(1);
1030 }
1031 1028
1032 /* 1029 /*
1033 * Call the device specific close. This cannot fail. 1030 * Call the device specific close. This cannot fail.
@@ -1233,21 +1230,21 @@ void __netif_schedule(struct net_device *dev)
1233} 1230}
1234EXPORT_SYMBOL(__netif_schedule); 1231EXPORT_SYMBOL(__netif_schedule);
1235 1232
1236void __netif_rx_schedule(struct net_device *dev) 1233void dev_kfree_skb_irq(struct sk_buff *skb)
1237{ 1234{
1238 unsigned long flags; 1235 if (atomic_dec_and_test(&skb->users)) {
1236 struct softnet_data *sd;
1237 unsigned long flags;
1239 1238
1240 local_irq_save(flags); 1239 local_irq_save(flags);
1241 dev_hold(dev); 1240 sd = &__get_cpu_var(softnet_data);
1242 list_add_tail(&dev->poll_list, &__get_cpu_var(softnet_data).poll_list); 1241 skb->next = sd->completion_queue;
1243 if (dev->quota < 0) 1242 sd->completion_queue = skb;
1244 dev->quota += dev->weight; 1243 raise_softirq_irqoff(NET_TX_SOFTIRQ);
1245 else 1244 local_irq_restore(flags);
1246 dev->quota = dev->weight; 1245 }
1247 __raise_softirq_irqoff(NET_RX_SOFTIRQ);
1248 local_irq_restore(flags);
1249} 1246}
1250EXPORT_SYMBOL(__netif_rx_schedule); 1247EXPORT_SYMBOL(dev_kfree_skb_irq);
1251 1248
1252void dev_kfree_skb_any(struct sk_buff *skb) 1249void dev_kfree_skb_any(struct sk_buff *skb)
1253{ 1250{
@@ -1259,7 +1256,12 @@ void dev_kfree_skb_any(struct sk_buff *skb)
1259EXPORT_SYMBOL(dev_kfree_skb_any); 1256EXPORT_SYMBOL(dev_kfree_skb_any);
1260 1257
1261 1258
1262/* Hot-plugging. */ 1259/**
1260 * netif_device_detach - mark device as removed
1261 * @dev: network device
1262 *
1263 * Mark device as removed from system and therefore no longer available.
1264 */
1263void netif_device_detach(struct net_device *dev) 1265void netif_device_detach(struct net_device *dev)
1264{ 1266{
1265 if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) && 1267 if (test_and_clear_bit(__LINK_STATE_PRESENT, &dev->state) &&
@@ -1269,6 +1271,12 @@ void netif_device_detach(struct net_device *dev)
1269} 1271}
1270EXPORT_SYMBOL(netif_device_detach); 1272EXPORT_SYMBOL(netif_device_detach);
1271 1273
1274/**
1275 * netif_device_attach - mark device as attached
1276 * @dev: network device
1277 *
1278 * Mark device as attached from system and restart if needed.
1279 */
1272void netif_device_attach(struct net_device *dev) 1280void netif_device_attach(struct net_device *dev)
1273{ 1281{
1274 if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) && 1282 if (!test_and_set_bit(__LINK_STATE_PRESENT, &dev->state) &&
@@ -1730,7 +1738,7 @@ enqueue:
1730 return NET_RX_SUCCESS; 1738 return NET_RX_SUCCESS;
1731 } 1739 }
1732 1740
1733 netif_rx_schedule(&queue->backlog_dev); 1741 napi_schedule(&queue->backlog);
1734 goto enqueue; 1742 goto enqueue;
1735 } 1743 }
1736 1744
@@ -1771,6 +1779,7 @@ static inline struct net_device *skb_bond(struct sk_buff *skb)
1771 return dev; 1779 return dev;
1772} 1780}
1773 1781
1782
1774static void net_tx_action(struct softirq_action *h) 1783static void net_tx_action(struct softirq_action *h)
1775{ 1784{
1776 struct softnet_data *sd = &__get_cpu_var(softnet_data); 1785 struct softnet_data *sd = &__get_cpu_var(softnet_data);
@@ -1927,7 +1936,7 @@ int netif_receive_skb(struct sk_buff *skb)
1927 __be16 type; 1936 __be16 type;
1928 1937
1929 /* if we've gotten here through NAPI, check netpoll */ 1938 /* if we've gotten here through NAPI, check netpoll */
1930 if (skb->dev->poll && netpoll_rx(skb)) 1939 if (netpoll_receive_skb(skb))
1931 return NET_RX_DROP; 1940 return NET_RX_DROP;
1932 1941
1933 if (!skb->tstamp.tv64) 1942 if (!skb->tstamp.tv64)
@@ -2017,22 +2026,25 @@ out:
2017 return ret; 2026 return ret;
2018} 2027}
2019 2028
2020static int process_backlog(struct net_device *backlog_dev, int *budget) 2029static int process_backlog(struct napi_struct *napi, int quota)
2021{ 2030{
2022 int work = 0; 2031 int work = 0;
2023 int quota = min(backlog_dev->quota, *budget);
2024 struct softnet_data *queue = &__get_cpu_var(softnet_data); 2032 struct softnet_data *queue = &__get_cpu_var(softnet_data);
2025 unsigned long start_time = jiffies; 2033 unsigned long start_time = jiffies;
2026 2034
2027 backlog_dev->weight = weight_p; 2035 napi->weight = weight_p;
2028 for (;;) { 2036 do {
2029 struct sk_buff *skb; 2037 struct sk_buff *skb;
2030 struct net_device *dev; 2038 struct net_device *dev;
2031 2039
2032 local_irq_disable(); 2040 local_irq_disable();
2033 skb = __skb_dequeue(&queue->input_pkt_queue); 2041 skb = __skb_dequeue(&queue->input_pkt_queue);
2034 if (!skb) 2042 if (!skb) {
2035 goto job_done; 2043 __napi_complete(napi);
2044 local_irq_enable();
2045 break;
2046 }
2047
2036 local_irq_enable(); 2048 local_irq_enable();
2037 2049
2038 dev = skb->dev; 2050 dev = skb->dev;
@@ -2040,67 +2052,86 @@ static int process_backlog(struct net_device *backlog_dev, int *budget)
2040 netif_receive_skb(skb); 2052 netif_receive_skb(skb);
2041 2053
2042 dev_put(dev); 2054 dev_put(dev);
2055 } while (++work < quota && jiffies == start_time);
2043 2056
2044 work++; 2057 return work;
2045 2058}
2046 if (work >= quota || jiffies - start_time > 1)
2047 break;
2048
2049 }
2050
2051 backlog_dev->quota -= work;
2052 *budget -= work;
2053 return -1;
2054
2055job_done:
2056 backlog_dev->quota -= work;
2057 *budget -= work;
2058 2059
2059 list_del(&backlog_dev->poll_list); 2060/**
2060 smp_mb__before_clear_bit(); 2061 * __napi_schedule - schedule for receive
2061 netif_poll_enable(backlog_dev); 2062 * @napi: entry to schedule
2063 *
2064 * The entry's receive function will be scheduled to run
2065 */
2066void fastcall __napi_schedule(struct napi_struct *n)
2067{
2068 unsigned long flags;
2062 2069
2063 local_irq_enable(); 2070 local_irq_save(flags);
2064 return 0; 2071 list_add_tail(&n->poll_list, &__get_cpu_var(softnet_data).poll_list);
2072 __raise_softirq_irqoff(NET_RX_SOFTIRQ);
2073 local_irq_restore(flags);
2065} 2074}
2075EXPORT_SYMBOL(__napi_schedule);
2076
2066 2077
2067static void net_rx_action(struct softirq_action *h) 2078static void net_rx_action(struct softirq_action *h)
2068{ 2079{
2069 struct softnet_data *queue = &__get_cpu_var(softnet_data); 2080 struct list_head *list = &__get_cpu_var(softnet_data).poll_list;
2070 unsigned long start_time = jiffies; 2081 unsigned long start_time = jiffies;
2071 int budget = netdev_budget; 2082 int budget = netdev_budget;
2072 void *have; 2083 void *have;
2073 2084
2074 local_irq_disable(); 2085 local_irq_disable();
2075 2086
2076 while (!list_empty(&queue->poll_list)) { 2087 while (!list_empty(list)) {
2077 struct net_device *dev; 2088 struct napi_struct *n;
2089 int work, weight;
2078 2090
2079 if (budget <= 0 || jiffies - start_time > 1) 2091 /* If softirq window is exhuasted then punt.
2092 *
2093 * Note that this is a slight policy change from the
2094 * previous NAPI code, which would allow up to 2
2095 * jiffies to pass before breaking out. The test
2096 * used to be "jiffies - start_time > 1".
2097 */
2098 if (unlikely(budget <= 0 || jiffies != start_time))
2080 goto softnet_break; 2099 goto softnet_break;
2081 2100
2082 local_irq_enable(); 2101 local_irq_enable();
2083 2102
2084 dev = list_entry(queue->poll_list.next, 2103 /* Even though interrupts have been re-enabled, this
2085 struct net_device, poll_list); 2104 * access is safe because interrupts can only add new
2086 have = netpoll_poll_lock(dev); 2105 * entries to the tail of this list, and only ->poll()
2106 * calls can remove this head entry from the list.
2107 */
2108 n = list_entry(list->next, struct napi_struct, poll_list);
2087 2109
2088 if (dev->quota <= 0 || dev->poll(dev, &budget)) { 2110 have = netpoll_poll_lock(n);
2089 netpoll_poll_unlock(have); 2111
2090 local_irq_disable(); 2112 weight = n->weight;
2091 list_move_tail(&dev->poll_list, &queue->poll_list); 2113
2092 if (dev->quota < 0) 2114 work = n->poll(n, weight);
2093 dev->quota += dev->weight; 2115
2094 else 2116 WARN_ON_ONCE(work > weight);
2095 dev->quota = dev->weight; 2117
2096 } else { 2118 budget -= work;
2097 netpoll_poll_unlock(have); 2119
2098 dev_put(dev); 2120 local_irq_disable();
2099 local_irq_disable(); 2121
2100 } 2122 /* Drivers must not modify the NAPI state if they
2123 * consume the entire weight. In such cases this code
2124 * still "owns" the NAPI instance and therefore can
2125 * move the instance around on the list at-will.
2126 */
2127 if (unlikely(work == weight))
2128 list_move_tail(&n->poll_list, list);
2129
2130 netpoll_poll_unlock(have);
2101 } 2131 }
2102out: 2132out:
2103 local_irq_enable(); 2133 local_irq_enable();
2134
2104#ifdef CONFIG_NET_DMA 2135#ifdef CONFIG_NET_DMA
2105 /* 2136 /*
2106 * There may not be any more sk_buffs coming right now, so push 2137 * There may not be any more sk_buffs coming right now, so push
@@ -2115,6 +2146,7 @@ out:
2115 } 2146 }
2116 } 2147 }
2117#endif 2148#endif
2149
2118 return; 2150 return;
2119 2151
2120softnet_break: 2152softnet_break:
@@ -3704,6 +3736,7 @@ struct net_device *alloc_netdev_mq(int sizeof_priv, const char *name,
3704 dev->egress_subqueue_count = queue_count; 3736 dev->egress_subqueue_count = queue_count;
3705 3737
3706 dev->get_stats = internal_stats; 3738 dev->get_stats = internal_stats;
3739 netpoll_netdev_init(dev);
3707 setup(dev); 3740 setup(dev);
3708 strcpy(dev->name, name); 3741 strcpy(dev->name, name);
3709 return dev; 3742 return dev;
@@ -4076,10 +4109,9 @@ static int __init net_dev_init(void)
4076 skb_queue_head_init(&queue->input_pkt_queue); 4109 skb_queue_head_init(&queue->input_pkt_queue);
4077 queue->completion_queue = NULL; 4110 queue->completion_queue = NULL;
4078 INIT_LIST_HEAD(&queue->poll_list); 4111 INIT_LIST_HEAD(&queue->poll_list);
4079 set_bit(__LINK_STATE_START, &queue->backlog_dev.state); 4112
4080 queue->backlog_dev.weight = weight_p; 4113 queue->backlog.poll = process_backlog;
4081 queue->backlog_dev.poll = process_backlog; 4114 queue->backlog.weight = weight_p;
4082 atomic_set(&queue->backlog_dev.refcnt, 1);
4083 } 4115 }
4084 4116
4085 netdev_dma_register(); 4117 netdev_dma_register();
diff --git a/net/core/net-sysfs.c b/net/core/net-sysfs.c
index 5c19b0646d7a..79159db6acb9 100644
--- a/net/core/net-sysfs.c
+++ b/net/core/net-sysfs.c
@@ -216,20 +216,6 @@ static ssize_t store_tx_queue_len(struct device *dev,
216 return netdev_store(dev, attr, buf, len, change_tx_queue_len); 216 return netdev_store(dev, attr, buf, len, change_tx_queue_len);
217} 217}
218 218
219NETDEVICE_SHOW(weight, fmt_dec);
220
221static int change_weight(struct net_device *net, unsigned long new_weight)
222{
223 net->weight = new_weight;
224 return 0;
225}
226
227static ssize_t store_weight(struct device *dev, struct device_attribute *attr,
228 const char *buf, size_t len)
229{
230 return netdev_store(dev, attr, buf, len, change_weight);
231}
232
233static struct device_attribute net_class_attributes[] = { 219static struct device_attribute net_class_attributes[] = {
234 __ATTR(addr_len, S_IRUGO, show_addr_len, NULL), 220 __ATTR(addr_len, S_IRUGO, show_addr_len, NULL),
235 __ATTR(iflink, S_IRUGO, show_iflink, NULL), 221 __ATTR(iflink, S_IRUGO, show_iflink, NULL),
@@ -246,7 +232,6 @@ static struct device_attribute net_class_attributes[] = {
246 __ATTR(flags, S_IRUGO | S_IWUSR, show_flags, store_flags), 232 __ATTR(flags, S_IRUGO | S_IWUSR, show_flags, store_flags),
247 __ATTR(tx_queue_len, S_IRUGO | S_IWUSR, show_tx_queue_len, 233 __ATTR(tx_queue_len, S_IRUGO | S_IWUSR, show_tx_queue_len,
248 store_tx_queue_len), 234 store_tx_queue_len),
249 __ATTR(weight, S_IRUGO | S_IWUSR, show_weight, store_weight),
250 {} 235 {}
251}; 236};
252 237
diff --git a/net/core/netpoll.c b/net/core/netpoll.c
index de1b26aa5720..abe6e3a4cc44 100644
--- a/net/core/netpoll.c
+++ b/net/core/netpoll.c
@@ -119,19 +119,22 @@ static __sum16 checksum_udp(struct sk_buff *skb, struct udphdr *uh,
119static void poll_napi(struct netpoll *np) 119static void poll_napi(struct netpoll *np)
120{ 120{
121 struct netpoll_info *npinfo = np->dev->npinfo; 121 struct netpoll_info *npinfo = np->dev->npinfo;
122 struct napi_struct *napi;
122 int budget = 16; 123 int budget = 16;
123 124
124 if (test_bit(__LINK_STATE_RX_SCHED, &np->dev->state) && 125 list_for_each_entry(napi, &np->dev->napi_list, dev_list) {
125 npinfo->poll_owner != smp_processor_id() && 126 if (test_bit(NAPI_STATE_SCHED, &napi->state) &&
126 spin_trylock(&npinfo->poll_lock)) { 127 napi->poll_owner != smp_processor_id() &&
127 npinfo->rx_flags |= NETPOLL_RX_DROP; 128 spin_trylock(&napi->poll_lock)) {
128 atomic_inc(&trapped); 129 npinfo->rx_flags |= NETPOLL_RX_DROP;
130 atomic_inc(&trapped);
129 131
130 np->dev->poll(np->dev, &budget); 132 napi->poll(napi, budget);
131 133
132 atomic_dec(&trapped); 134 atomic_dec(&trapped);
133 npinfo->rx_flags &= ~NETPOLL_RX_DROP; 135 npinfo->rx_flags &= ~NETPOLL_RX_DROP;
134 spin_unlock(&npinfo->poll_lock); 136 spin_unlock(&napi->poll_lock);
137 }
135 } 138 }
136} 139}
137 140
@@ -157,7 +160,7 @@ void netpoll_poll(struct netpoll *np)
157 160
158 /* Process pending work on NIC */ 161 /* Process pending work on NIC */
159 np->dev->poll_controller(np->dev); 162 np->dev->poll_controller(np->dev);
160 if (np->dev->poll) 163 if (!list_empty(&np->dev->napi_list))
161 poll_napi(np); 164 poll_napi(np);
162 165
163 service_arp_queue(np->dev->npinfo); 166 service_arp_queue(np->dev->npinfo);
@@ -233,6 +236,17 @@ repeat:
233 return skb; 236 return skb;
234} 237}
235 238
239static int netpoll_owner_active(struct net_device *dev)
240{
241 struct napi_struct *napi;
242
243 list_for_each_entry(napi, &dev->napi_list, dev_list) {
244 if (napi->poll_owner == smp_processor_id())
245 return 1;
246 }
247 return 0;
248}
249
236static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb) 250static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
237{ 251{
238 int status = NETDEV_TX_BUSY; 252 int status = NETDEV_TX_BUSY;
@@ -246,8 +260,7 @@ static void netpoll_send_skb(struct netpoll *np, struct sk_buff *skb)
246 } 260 }
247 261
248 /* don't get messages out of order, and no recursion */ 262 /* don't get messages out of order, and no recursion */
249 if (skb_queue_len(&npinfo->txq) == 0 && 263 if (skb_queue_len(&npinfo->txq) == 0 && !netpoll_owner_active(dev)) {
250 npinfo->poll_owner != smp_processor_id()) {
251 unsigned long flags; 264 unsigned long flags;
252 265
253 local_irq_save(flags); 266 local_irq_save(flags);
@@ -652,8 +665,6 @@ int netpoll_setup(struct netpoll *np)
652 665
653 npinfo->rx_flags = 0; 666 npinfo->rx_flags = 0;
654 npinfo->rx_np = NULL; 667 npinfo->rx_np = NULL;
655 spin_lock_init(&npinfo->poll_lock);
656 npinfo->poll_owner = -1;
657 668
658 spin_lock_init(&npinfo->rx_lock); 669 spin_lock_init(&npinfo->rx_lock);
659 skb_queue_head_init(&npinfo->arp_tx); 670 skb_queue_head_init(&npinfo->arp_tx);
diff --git a/net/core/rtnetlink.c b/net/core/rtnetlink.c
index 4756d5857abf..2b0b6fac6cef 100644
--- a/net/core/rtnetlink.c
+++ b/net/core/rtnetlink.c
@@ -634,7 +634,6 @@ static int rtnl_fill_ifinfo(struct sk_buff *skb, struct net_device *dev,
634 634
635 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name); 635 NLA_PUT_STRING(skb, IFLA_IFNAME, dev->name);
636 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len); 636 NLA_PUT_U32(skb, IFLA_TXQLEN, dev->tx_queue_len);
637 NLA_PUT_U32(skb, IFLA_WEIGHT, dev->weight);
638 NLA_PUT_U8(skb, IFLA_OPERSTATE, 637 NLA_PUT_U8(skb, IFLA_OPERSTATE,
639 netif_running(dev) ? dev->operstate : IF_OPER_DOWN); 638 netif_running(dev) ? dev->operstate : IF_OPER_DOWN);
640 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode); 639 NLA_PUT_U8(skb, IFLA_LINKMODE, dev->link_mode);
@@ -834,9 +833,6 @@ static int do_setlink(struct net_device *dev, struct ifinfomsg *ifm,
834 if (tb[IFLA_TXQLEN]) 833 if (tb[IFLA_TXQLEN])
835 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 834 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
836 835
837 if (tb[IFLA_WEIGHT])
838 dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
839
840 if (tb[IFLA_OPERSTATE]) 836 if (tb[IFLA_OPERSTATE])
841 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 837 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
842 838
@@ -1074,8 +1070,6 @@ replay:
1074 nla_len(tb[IFLA_BROADCAST])); 1070 nla_len(tb[IFLA_BROADCAST]));
1075 if (tb[IFLA_TXQLEN]) 1071 if (tb[IFLA_TXQLEN])
1076 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]); 1072 dev->tx_queue_len = nla_get_u32(tb[IFLA_TXQLEN]);
1077 if (tb[IFLA_WEIGHT])
1078 dev->weight = nla_get_u32(tb[IFLA_WEIGHT]);
1079 if (tb[IFLA_OPERSTATE]) 1073 if (tb[IFLA_OPERSTATE])
1080 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE])); 1074 set_operstate(dev, nla_get_u8(tb[IFLA_OPERSTATE]));
1081 if (tb[IFLA_LINKMODE]) 1075 if (tb[IFLA_LINKMODE])
diff --git a/net/sched/sch_generic.c b/net/sched/sch_generic.c
index c81649cf0b9e..e970e8e75720 100644
--- a/net/sched/sch_generic.c
+++ b/net/sched/sch_generic.c
@@ -256,6 +256,12 @@ static void dev_watchdog_down(struct net_device *dev)
256 netif_tx_unlock_bh(dev); 256 netif_tx_unlock_bh(dev);
257} 257}
258 258
259/**
260 * netif_carrier_on - set carrier
261 * @dev: network device
262 *
263 * Device has detected that carrier.
264 */
259void netif_carrier_on(struct net_device *dev) 265void netif_carrier_on(struct net_device *dev)
260{ 266{
261 if (test_and_clear_bit(__LINK_STATE_NOCARRIER, &dev->state)) 267 if (test_and_clear_bit(__LINK_STATE_NOCARRIER, &dev->state))
@@ -264,6 +270,12 @@ void netif_carrier_on(struct net_device *dev)
264 __netdev_watchdog_up(dev); 270 __netdev_watchdog_up(dev);
265} 271}
266 272
273/**
274 * netif_carrier_off - clear carrier
275 * @dev: network device
276 *
277 * Device has detected loss of carrier.
278 */
267void netif_carrier_off(struct net_device *dev) 279void netif_carrier_off(struct net_device *dev)
268{ 280{
269 if (!test_and_set_bit(__LINK_STATE_NOCARRIER, &dev->state)) 281 if (!test_and_set_bit(__LINK_STATE_NOCARRIER, &dev->state))