aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2008-12-30 20:37:25 -0500
committerLinus Torvalds <torvalds@linux-foundation.org>2008-12-30 20:37:25 -0500
commitab70537c32a3245325b48774664da588904e47f2 (patch)
treefdb4447e520bd34dd8696fdd3b976075414d8555
parent14a3c4ab0e58d143c7928c9eb2f2610205e13bf2 (diff)
parentbda53cd510b6777ced652ba279020bb7b414b744 (diff)
Merge git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus
* git://git.kernel.org/pub/scm/linux/kernel/git/rusty/linux-2.6-for-linus: lguest: struct device - replace bus_id with dev_name() lguest: move the initial guest page table creation code to the host kvm-s390: implement config_changed for virtio on s390 virtio_console: support console resizing virtio: add PCI device release() function virtio_blk: fix type warning virtio: block: dynamic maximum segments virtio: set max_segment_size and max_sectors to infinite. virtio: avoid implicit use of Linux page size in balloon interface virtio: hand virtio ring alignment as argument to vring_new_virtqueue virtio: use KVM_S390_VIRTIO_RING_ALIGN instead of relying on pagesize virtio: use LGUEST_VRING_ALIGN instead of relying on pagesize virtio: Don't use PAGE_SIZE for vring alignment in virtio_pci. virtio: rename 'pagesize' arg to vring_init/vring_size virtio: Don't use PAGE_SIZE in virtio_pci.c virtio: struct device - replace bus_id with dev_name(), dev_set_name() virtio-pci queue allocation not page-aligned
-rw-r--r--Documentation/lguest/lguest.c66
-rw-r--r--arch/s390/include/asm/kvm_virtio.h4
-rw-r--r--arch/x86/lguest/i386_head.S15
-rw-r--r--drivers/block/virtio_blk.c41
-rw-r--r--drivers/char/hvc_console.c1
-rw-r--r--drivers/char/virtio_console.c30
-rw-r--r--drivers/lguest/lg.h2
-rw-r--r--drivers/lguest/lguest_device.c8
-rw-r--r--drivers/lguest/lguest_user.c13
-rw-r--r--drivers/lguest/page_tables.c72
-rw-r--r--drivers/s390/kvm/kvm_virtio.c34
-rw-r--r--drivers/virtio/virtio.c2
-rw-r--r--drivers/virtio/virtio_balloon.c13
-rw-r--r--drivers/virtio/virtio_pci.c43
-rw-r--r--drivers/virtio/virtio_ring.c3
-rw-r--r--include/linux/lguest_launcher.h6
-rw-r--r--include/linux/virtio_balloon.h3
-rw-r--r--include/linux/virtio_console.h11
-rw-r--r--include/linux/virtio_pci.h8
-rw-r--r--include/linux/virtio_ring.h13
20 files changed, 253 insertions, 135 deletions
diff --git a/Documentation/lguest/lguest.c b/Documentation/lguest/lguest.c
index 804520633fcf..f2dbbf3bdeab 100644
--- a/Documentation/lguest/lguest.c
+++ b/Documentation/lguest/lguest.c
@@ -481,51 +481,6 @@ static unsigned long load_initrd(const char *name, unsigned long mem)
481 /* We return the initrd size. */ 481 /* We return the initrd size. */
482 return len; 482 return len;
483} 483}
484
485/* Once we know how much memory we have we can construct simple linear page
486 * tables which set virtual == physical which will get the Guest far enough
487 * into the boot to create its own.
488 *
489 * We lay them out of the way, just below the initrd (which is why we need to
490 * know its size here). */
491static unsigned long setup_pagetables(unsigned long mem,
492 unsigned long initrd_size)
493{
494 unsigned long *pgdir, *linear;
495 unsigned int mapped_pages, i, linear_pages;
496 unsigned int ptes_per_page = getpagesize()/sizeof(void *);
497
498 mapped_pages = mem/getpagesize();
499
500 /* Each PTE page can map ptes_per_page pages: how many do we need? */
501 linear_pages = (mapped_pages + ptes_per_page-1)/ptes_per_page;
502
503 /* We put the toplevel page directory page at the top of memory. */
504 pgdir = from_guest_phys(mem) - initrd_size - getpagesize();
505
506 /* Now we use the next linear_pages pages as pte pages */
507 linear = (void *)pgdir - linear_pages*getpagesize();
508
509 /* Linear mapping is easy: put every page's address into the mapping in
510 * order. PAGE_PRESENT contains the flags Present, Writable and
511 * Executable. */
512 for (i = 0; i < mapped_pages; i++)
513 linear[i] = ((i * getpagesize()) | PAGE_PRESENT);
514
515 /* The top level points to the linear page table pages above. */
516 for (i = 0; i < mapped_pages; i += ptes_per_page) {
517 pgdir[i/ptes_per_page]
518 = ((to_guest_phys(linear) + i*sizeof(void *))
519 | PAGE_PRESENT);
520 }
521
522 verbose("Linear mapping of %u pages in %u pte pages at %#lx\n",
523 mapped_pages, linear_pages, to_guest_phys(linear));
524
525 /* We return the top level (guest-physical) address: the kernel needs
526 * to know where it is. */
527 return to_guest_phys(pgdir);
528}
529/*:*/ 484/*:*/
530 485
531/* Simple routine to roll all the commandline arguments together with spaces 486/* Simple routine to roll all the commandline arguments together with spaces
@@ -548,13 +503,13 @@ static void concat(char *dst, char *args[])
548 503
549/*L:185 This is where we actually tell the kernel to initialize the Guest. We 504/*L:185 This is where we actually tell the kernel to initialize the Guest. We
550 * saw the arguments it expects when we looked at initialize() in lguest_user.c: 505 * saw the arguments it expects when we looked at initialize() in lguest_user.c:
551 * the base of Guest "physical" memory, the top physical page to allow, the 506 * the base of Guest "physical" memory, the top physical page to allow and the
552 * top level pagetable and the entry point for the Guest. */ 507 * entry point for the Guest. */
553static int tell_kernel(unsigned long pgdir, unsigned long start) 508static int tell_kernel(unsigned long start)
554{ 509{
555 unsigned long args[] = { LHREQ_INITIALIZE, 510 unsigned long args[] = { LHREQ_INITIALIZE,
556 (unsigned long)guest_base, 511 (unsigned long)guest_base,
557 guest_limit / getpagesize(), pgdir, start }; 512 guest_limit / getpagesize(), start };
558 int fd; 513 int fd;
559 514
560 verbose("Guest: %p - %p (%#lx)\n", 515 verbose("Guest: %p - %p (%#lx)\n",
@@ -1030,7 +985,7 @@ static void update_device_status(struct device *dev)
1030 /* Zero out the virtqueues. */ 985 /* Zero out the virtqueues. */
1031 for (vq = dev->vq; vq; vq = vq->next) { 986 for (vq = dev->vq; vq; vq = vq->next) {
1032 memset(vq->vring.desc, 0, 987 memset(vq->vring.desc, 0,
1033 vring_size(vq->config.num, getpagesize())); 988 vring_size(vq->config.num, LGUEST_VRING_ALIGN));
1034 lg_last_avail(vq) = 0; 989 lg_last_avail(vq) = 0;
1035 } 990 }
1036 } else if (dev->desc->status & VIRTIO_CONFIG_S_FAILED) { 991 } else if (dev->desc->status & VIRTIO_CONFIG_S_FAILED) {
@@ -1211,7 +1166,7 @@ static void add_virtqueue(struct device *dev, unsigned int num_descs,
1211 void *p; 1166 void *p;
1212 1167
1213 /* First we need some memory for this virtqueue. */ 1168 /* First we need some memory for this virtqueue. */
1214 pages = (vring_size(num_descs, getpagesize()) + getpagesize() - 1) 1169 pages = (vring_size(num_descs, LGUEST_VRING_ALIGN) + getpagesize() - 1)
1215 / getpagesize(); 1170 / getpagesize();
1216 p = get_pages(pages); 1171 p = get_pages(pages);
1217 1172
@@ -1228,7 +1183,7 @@ static void add_virtqueue(struct device *dev, unsigned int num_descs,
1228 vq->config.pfn = to_guest_phys(p) / getpagesize(); 1183 vq->config.pfn = to_guest_phys(p) / getpagesize();
1229 1184
1230 /* Initialize the vring. */ 1185 /* Initialize the vring. */
1231 vring_init(&vq->vring, num_descs, p, getpagesize()); 1186 vring_init(&vq->vring, num_descs, p, LGUEST_VRING_ALIGN);
1232 1187
1233 /* Append virtqueue to this device's descriptor. We use 1188 /* Append virtqueue to this device's descriptor. We use
1234 * device_config() to get the end of the device's current virtqueues; 1189 * device_config() to get the end of the device's current virtqueues;
@@ -1941,7 +1896,7 @@ int main(int argc, char *argv[])
1941{ 1896{
1942 /* Memory, top-level pagetable, code startpoint and size of the 1897 /* Memory, top-level pagetable, code startpoint and size of the
1943 * (optional) initrd. */ 1898 * (optional) initrd. */
1944 unsigned long mem = 0, pgdir, start, initrd_size = 0; 1899 unsigned long mem = 0, start, initrd_size = 0;
1945 /* Two temporaries and the /dev/lguest file descriptor. */ 1900 /* Two temporaries and the /dev/lguest file descriptor. */
1946 int i, c, lguest_fd; 1901 int i, c, lguest_fd;
1947 /* The boot information for the Guest. */ 1902 /* The boot information for the Guest. */
@@ -2040,9 +1995,6 @@ int main(int argc, char *argv[])
2040 boot->hdr.type_of_loader = 0xFF; 1995 boot->hdr.type_of_loader = 0xFF;
2041 } 1996 }
2042 1997
2043 /* Set up the initial linear pagetables, starting below the initrd. */
2044 pgdir = setup_pagetables(mem, initrd_size);
2045
2046 /* The Linux boot header contains an "E820" memory map: ours is a 1998 /* The Linux boot header contains an "E820" memory map: ours is a
2047 * simple, single region. */ 1999 * simple, single region. */
2048 boot->e820_entries = 1; 2000 boot->e820_entries = 1;
@@ -2064,7 +2016,7 @@ int main(int argc, char *argv[])
2064 2016
2065 /* We tell the kernel to initialize the Guest: this returns the open 2017 /* We tell the kernel to initialize the Guest: this returns the open
2066 * /dev/lguest file descriptor. */ 2018 * /dev/lguest file descriptor. */
2067 lguest_fd = tell_kernel(pgdir, start); 2019 lguest_fd = tell_kernel(start);
2068 2020
2069 /* We clone off a thread, which wakes the Launcher whenever one of the 2021 /* We clone off a thread, which wakes the Launcher whenever one of the
2070 * input file descriptors needs attention. We call this the Waker, and 2022 * input file descriptors needs attention. We call this the Waker, and
diff --git a/arch/s390/include/asm/kvm_virtio.h b/arch/s390/include/asm/kvm_virtio.h
index c13568b9351c..0503936f101f 100644
--- a/arch/s390/include/asm/kvm_virtio.h
+++ b/arch/s390/include/asm/kvm_virtio.h
@@ -50,6 +50,10 @@ struct kvm_vqconfig {
50#define KVM_S390_VIRTIO_RESET 1 50#define KVM_S390_VIRTIO_RESET 1
51#define KVM_S390_VIRTIO_SET_STATUS 2 51#define KVM_S390_VIRTIO_SET_STATUS 2
52 52
53/* The alignment to use between consumer and producer parts of vring.
54 * This is pagesize for historical reasons. */
55#define KVM_S390_VIRTIO_RING_ALIGN 4096
56
53#ifdef __KERNEL__ 57#ifdef __KERNEL__
54/* early virtio console setup */ 58/* early virtio console setup */
55#ifdef CONFIG_S390_GUEST 59#ifdef CONFIG_S390_GUEST
diff --git a/arch/x86/lguest/i386_head.S b/arch/x86/lguest/i386_head.S
index 5c7cef34c9e7..10b9bd35a8ff 100644
--- a/arch/x86/lguest/i386_head.S
+++ b/arch/x86/lguest/i386_head.S
@@ -30,21 +30,6 @@ ENTRY(lguest_entry)
30 movl $lguest_data - __PAGE_OFFSET, %edx 30 movl $lguest_data - __PAGE_OFFSET, %edx
31 int $LGUEST_TRAP_ENTRY 31 int $LGUEST_TRAP_ENTRY
32 32
33 /* The Host put the toplevel pagetable in lguest_data.pgdir. The movsl
34 * instruction uses %esi implicitly as the source for the copy we're
35 * about to do. */
36 movl lguest_data - __PAGE_OFFSET + LGUEST_DATA_pgdir, %esi
37
38 /* Copy first 32 entries of page directory to __PAGE_OFFSET entries.
39 * This means the first 128M of kernel memory will be mapped at
40 * PAGE_OFFSET where the kernel expects to run. This will get it far
41 * enough through boot to switch to its own pagetables. */
42 movl $32, %ecx
43 movl %esi, %edi
44 addl $((__PAGE_OFFSET >> 22) * 4), %edi
45 rep
46 movsl
47
48 /* Set up the initial stack so we can run C code. */ 33 /* Set up the initial stack so we can run C code. */
49 movl $(init_thread_union+THREAD_SIZE),%esp 34 movl $(init_thread_union+THREAD_SIZE),%esp
50 35
diff --git a/drivers/block/virtio_blk.c b/drivers/block/virtio_blk.c
index f151592ecf73..5d34764c8a87 100644
--- a/drivers/block/virtio_blk.c
+++ b/drivers/block/virtio_blk.c
@@ -6,7 +6,6 @@
6#include <linux/virtio_blk.h> 6#include <linux/virtio_blk.h>
7#include <linux/scatterlist.h> 7#include <linux/scatterlist.h>
8 8
9#define VIRTIO_MAX_SG (3+MAX_PHYS_SEGMENTS)
10#define PART_BITS 4 9#define PART_BITS 4
11 10
12static int major, index; 11static int major, index;
@@ -26,8 +25,11 @@ struct virtio_blk
26 25
27 mempool_t *pool; 26 mempool_t *pool;
28 27
28 /* What host tells us, plus 2 for header & tailer. */
29 unsigned int sg_elems;
30
29 /* Scatterlist: can be too big for stack. */ 31 /* Scatterlist: can be too big for stack. */
30 struct scatterlist sg[VIRTIO_MAX_SG]; 32 struct scatterlist sg[/*sg_elems*/];
31}; 33};
32 34
33struct virtblk_req 35struct virtblk_req
@@ -97,8 +99,6 @@ static bool do_req(struct request_queue *q, struct virtio_blk *vblk,
97 if (blk_barrier_rq(vbr->req)) 99 if (blk_barrier_rq(vbr->req))
98 vbr->out_hdr.type |= VIRTIO_BLK_T_BARRIER; 100 vbr->out_hdr.type |= VIRTIO_BLK_T_BARRIER;
99 101
100 /* This init could be done at vblk creation time */
101 sg_init_table(vblk->sg, VIRTIO_MAX_SG);
102 sg_set_buf(&vblk->sg[0], &vbr->out_hdr, sizeof(vbr->out_hdr)); 102 sg_set_buf(&vblk->sg[0], &vbr->out_hdr, sizeof(vbr->out_hdr));
103 num = blk_rq_map_sg(q, vbr->req, vblk->sg+1); 103 num = blk_rq_map_sg(q, vbr->req, vblk->sg+1);
104 sg_set_buf(&vblk->sg[num+1], &vbr->status, sizeof(vbr->status)); 104 sg_set_buf(&vblk->sg[num+1], &vbr->status, sizeof(vbr->status));
@@ -130,7 +130,7 @@ static void do_virtblk_request(struct request_queue *q)
130 130
131 while ((req = elv_next_request(q)) != NULL) { 131 while ((req = elv_next_request(q)) != NULL) {
132 vblk = req->rq_disk->private_data; 132 vblk = req->rq_disk->private_data;
133 BUG_ON(req->nr_phys_segments > ARRAY_SIZE(vblk->sg)); 133 BUG_ON(req->nr_phys_segments + 2 > vblk->sg_elems);
134 134
135 /* If this request fails, stop queue and wait for something to 135 /* If this request fails, stop queue and wait for something to
136 finish to restart it. */ 136 finish to restart it. */
@@ -196,12 +196,22 @@ static int virtblk_probe(struct virtio_device *vdev)
196 int err; 196 int err;
197 u64 cap; 197 u64 cap;
198 u32 v; 198 u32 v;
199 u32 blk_size; 199 u32 blk_size, sg_elems;
200 200
201 if (index_to_minor(index) >= 1 << MINORBITS) 201 if (index_to_minor(index) >= 1 << MINORBITS)
202 return -ENOSPC; 202 return -ENOSPC;
203 203
204 vdev->priv = vblk = kmalloc(sizeof(*vblk), GFP_KERNEL); 204 /* We need to know how many segments before we allocate. */
205 err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX,
206 offsetof(struct virtio_blk_config, seg_max),
207 &sg_elems);
208 if (err)
209 sg_elems = 1;
210
211 /* We need an extra sg elements at head and tail. */
212 sg_elems += 2;
213 vdev->priv = vblk = kmalloc(sizeof(*vblk) +
214 sizeof(vblk->sg[0]) * sg_elems, GFP_KERNEL);
205 if (!vblk) { 215 if (!vblk) {
206 err = -ENOMEM; 216 err = -ENOMEM;
207 goto out; 217 goto out;
@@ -210,6 +220,8 @@ static int virtblk_probe(struct virtio_device *vdev)
210 INIT_LIST_HEAD(&vblk->reqs); 220 INIT_LIST_HEAD(&vblk->reqs);
211 spin_lock_init(&vblk->lock); 221 spin_lock_init(&vblk->lock);
212 vblk->vdev = vdev; 222 vblk->vdev = vdev;
223 vblk->sg_elems = sg_elems;
224 sg_init_table(vblk->sg, vblk->sg_elems);
213 225
214 /* We expect one virtqueue, for output. */ 226 /* We expect one virtqueue, for output. */
215 vblk->vq = vdev->config->find_vq(vdev, 0, blk_done); 227 vblk->vq = vdev->config->find_vq(vdev, 0, blk_done);
@@ -279,6 +291,13 @@ static int virtblk_probe(struct virtio_device *vdev)
279 } 291 }
280 set_capacity(vblk->disk, cap); 292 set_capacity(vblk->disk, cap);
281 293
294 /* We can handle whatever the host told us to handle. */
295 blk_queue_max_phys_segments(vblk->disk->queue, vblk->sg_elems-2);
296 blk_queue_max_hw_segments(vblk->disk->queue, vblk->sg_elems-2);
297
298 /* No real sector limit. */
299 blk_queue_max_sectors(vblk->disk->queue, -1U);
300
282 /* Host can optionally specify maximum segment size and number of 301 /* Host can optionally specify maximum segment size and number of
283 * segments. */ 302 * segments. */
284 err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX, 303 err = virtio_config_val(vdev, VIRTIO_BLK_F_SIZE_MAX,
@@ -286,12 +305,8 @@ static int virtblk_probe(struct virtio_device *vdev)
286 &v); 305 &v);
287 if (!err) 306 if (!err)
288 blk_queue_max_segment_size(vblk->disk->queue, v); 307 blk_queue_max_segment_size(vblk->disk->queue, v);
289 308 else
290 err = virtio_config_val(vdev, VIRTIO_BLK_F_SEG_MAX, 309 blk_queue_max_segment_size(vblk->disk->queue, -1U);
291 offsetof(struct virtio_blk_config, seg_max),
292 &v);
293 if (!err)
294 blk_queue_max_hw_segments(vblk->disk->queue, v);
295 310
296 /* Host can optionally specify the block size of the device */ 311 /* Host can optionally specify the block size of the device */
297 err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE, 312 err = virtio_config_val(vdev, VIRTIO_BLK_F_BLK_SIZE,
diff --git a/drivers/char/hvc_console.c b/drivers/char/hvc_console.c
index fb57f67bb427..0587b66d6fc7 100644
--- a/drivers/char/hvc_console.c
+++ b/drivers/char/hvc_console.c
@@ -695,6 +695,7 @@ void hvc_resize(struct hvc_struct *hp, struct winsize ws)
695 hp->ws = ws; 695 hp->ws = ws;
696 schedule_work(&hp->tty_resize); 696 schedule_work(&hp->tty_resize);
697} 697}
698EXPORT_SYMBOL_GPL(hvc_resize);
698 699
699/* 700/*
700 * This kthread is either polling or interrupt driven. This is determined by 701 * This kthread is either polling or interrupt driven. This is determined by
diff --git a/drivers/char/virtio_console.c b/drivers/char/virtio_console.c
index 3fb0d2c88ba5..ff6f5a4b58fb 100644
--- a/drivers/char/virtio_console.c
+++ b/drivers/char/virtio_console.c
@@ -138,12 +138,33 @@ int __init virtio_cons_early_init(int (*put_chars)(u32, const char *, int))
138} 138}
139 139
140/* 140/*
141 * virtio console configuration. This supports:
142 * - console resize
143 */
144static void virtcons_apply_config(struct virtio_device *dev)
145{
146 struct winsize ws;
147
148 if (virtio_has_feature(dev, VIRTIO_CONSOLE_F_SIZE)) {
149 dev->config->get(dev,
150 offsetof(struct virtio_console_config, cols),
151 &ws.ws_col, sizeof(u16));
152 dev->config->get(dev,
153 offsetof(struct virtio_console_config, rows),
154 &ws.ws_row, sizeof(u16));
155 hvc_resize(hvc, ws);
156 }
157}
158
159/*
141 * we support only one console, the hvc struct is a global var 160 * we support only one console, the hvc struct is a global var
142 * There is no need to do anything 161 * We set the configuration at this point, since we now have a tty
143 */ 162 */
144static int notifier_add_vio(struct hvc_struct *hp, int data) 163static int notifier_add_vio(struct hvc_struct *hp, int data)
145{ 164{
146 hp->irq_requested = 1; 165 hp->irq_requested = 1;
166 virtcons_apply_config(vdev);
167
147 return 0; 168 return 0;
148} 169}
149 170
@@ -234,11 +255,18 @@ static struct virtio_device_id id_table[] = {
234 { 0 }, 255 { 0 },
235}; 256};
236 257
258static unsigned int features[] = {
259 VIRTIO_CONSOLE_F_SIZE,
260};
261
237static struct virtio_driver virtio_console = { 262static struct virtio_driver virtio_console = {
263 .feature_table = features,
264 .feature_table_size = ARRAY_SIZE(features),
238 .driver.name = KBUILD_MODNAME, 265 .driver.name = KBUILD_MODNAME,
239 .driver.owner = THIS_MODULE, 266 .driver.owner = THIS_MODULE,
240 .id_table = id_table, 267 .id_table = id_table,
241 .probe = virtcons_probe, 268 .probe = virtcons_probe,
269 .config_changed = virtcons_apply_config,
242}; 270};
243 271
244static int __init init(void) 272static int __init init(void)
diff --git a/drivers/lguest/lg.h b/drivers/lguest/lg.h
index 5faefeaf6790..f2c641e0bdde 100644
--- a/drivers/lguest/lg.h
+++ b/drivers/lguest/lg.h
@@ -164,7 +164,7 @@ void copy_gdt(const struct lg_cpu *cpu, struct desc_struct *gdt);
164void copy_gdt_tls(const struct lg_cpu *cpu, struct desc_struct *gdt); 164void copy_gdt_tls(const struct lg_cpu *cpu, struct desc_struct *gdt);
165 165
166/* page_tables.c: */ 166/* page_tables.c: */
167int init_guest_pagetable(struct lguest *lg, unsigned long pgtable); 167int init_guest_pagetable(struct lguest *lg);
168void free_guest_pagetable(struct lguest *lg); 168void free_guest_pagetable(struct lguest *lg);
169void guest_new_pagetable(struct lg_cpu *cpu, unsigned long pgtable); 169void guest_new_pagetable(struct lg_cpu *cpu, unsigned long pgtable);
170void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 i); 170void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 i);
diff --git a/drivers/lguest/lguest_device.c b/drivers/lguest/lguest_device.c
index a661bbdae3d6..915da6b8c924 100644
--- a/drivers/lguest/lguest_device.c
+++ b/drivers/lguest/lguest_device.c
@@ -250,7 +250,7 @@ static struct virtqueue *lg_find_vq(struct virtio_device *vdev,
250 /* Figure out how many pages the ring will take, and map that memory */ 250 /* Figure out how many pages the ring will take, and map that memory */
251 lvq->pages = lguest_map((unsigned long)lvq->config.pfn << PAGE_SHIFT, 251 lvq->pages = lguest_map((unsigned long)lvq->config.pfn << PAGE_SHIFT,
252 DIV_ROUND_UP(vring_size(lvq->config.num, 252 DIV_ROUND_UP(vring_size(lvq->config.num,
253 PAGE_SIZE), 253 LGUEST_VRING_ALIGN),
254 PAGE_SIZE)); 254 PAGE_SIZE));
255 if (!lvq->pages) { 255 if (!lvq->pages) {
256 err = -ENOMEM; 256 err = -ENOMEM;
@@ -259,8 +259,8 @@ static struct virtqueue *lg_find_vq(struct virtio_device *vdev,
259 259
260 /* OK, tell virtio_ring.c to set up a virtqueue now we know its size 260 /* OK, tell virtio_ring.c to set up a virtqueue now we know its size
261 * and we've got a pointer to its pages. */ 261 * and we've got a pointer to its pages. */
262 vq = vring_new_virtqueue(lvq->config.num, vdev, lvq->pages, 262 vq = vring_new_virtqueue(lvq->config.num, LGUEST_VRING_ALIGN,
263 lg_notify, callback); 263 vdev, lvq->pages, lg_notify, callback);
264 if (!vq) { 264 if (!vq) {
265 err = -ENOMEM; 265 err = -ENOMEM;
266 goto unmap; 266 goto unmap;
@@ -272,7 +272,7 @@ static struct virtqueue *lg_find_vq(struct virtio_device *vdev,
272 * the interrupt as a source of randomness: it'd be nice to have that 272 * the interrupt as a source of randomness: it'd be nice to have that
273 * back.. */ 273 * back.. */
274 err = request_irq(lvq->config.irq, vring_interrupt, IRQF_SHARED, 274 err = request_irq(lvq->config.irq, vring_interrupt, IRQF_SHARED,
275 vdev->dev.bus_id, vq); 275 dev_name(&vdev->dev), vq);
276 if (err) 276 if (err)
277 goto destroy_vring; 277 goto destroy_vring;
278 278
diff --git a/drivers/lguest/lguest_user.c b/drivers/lguest/lguest_user.c
index e73a000473cc..34bc017b8b3c 100644
--- a/drivers/lguest/lguest_user.c
+++ b/drivers/lguest/lguest_user.c
@@ -146,7 +146,7 @@ static int lg_cpu_start(struct lg_cpu *cpu, unsigned id, unsigned long start_ip)
146 return 0; 146 return 0;
147} 147}
148 148
149/*L:020 The initialization write supplies 4 pointer sized (32 or 64 bit) 149/*L:020 The initialization write supplies 3 pointer sized (32 or 64 bit)
150 * values (in addition to the LHREQ_INITIALIZE value). These are: 150 * values (in addition to the LHREQ_INITIALIZE value). These are:
151 * 151 *
152 * base: The start of the Guest-physical memory inside the Launcher memory. 152 * base: The start of the Guest-physical memory inside the Launcher memory.
@@ -155,9 +155,6 @@ static int lg_cpu_start(struct lg_cpu *cpu, unsigned id, unsigned long start_ip)
155 * allowed to access. The Guest memory lives inside the Launcher, so it sets 155 * allowed to access. The Guest memory lives inside the Launcher, so it sets
156 * this to ensure the Guest can only reach its own memory. 156 * this to ensure the Guest can only reach its own memory.
157 * 157 *
158 * pgdir: The (Guest-physical) address of the top of the initial Guest
159 * pagetables (which are set up by the Launcher).
160 *
161 * start: The first instruction to execute ("eip" in x86-speak). 158 * start: The first instruction to execute ("eip" in x86-speak).
162 */ 159 */
163static int initialize(struct file *file, const unsigned long __user *input) 160static int initialize(struct file *file, const unsigned long __user *input)
@@ -166,7 +163,7 @@ static int initialize(struct file *file, const unsigned long __user *input)
166 * Guest. */ 163 * Guest. */
167 struct lguest *lg; 164 struct lguest *lg;
168 int err; 165 int err;
169 unsigned long args[4]; 166 unsigned long args[3];
170 167
171 /* We grab the Big Lguest lock, which protects against multiple 168 /* We grab the Big Lguest lock, which protects against multiple
172 * simultaneous initializations. */ 169 * simultaneous initializations. */
@@ -192,14 +189,14 @@ static int initialize(struct file *file, const unsigned long __user *input)
192 lg->mem_base = (void __user *)args[0]; 189 lg->mem_base = (void __user *)args[0];
193 lg->pfn_limit = args[1]; 190 lg->pfn_limit = args[1];
194 191
195 /* This is the first cpu (cpu 0) and it will start booting at args[3] */ 192 /* This is the first cpu (cpu 0) and it will start booting at args[2] */
196 err = lg_cpu_start(&lg->cpus[0], 0, args[3]); 193 err = lg_cpu_start(&lg->cpus[0], 0, args[2]);
197 if (err) 194 if (err)
198 goto release_guest; 195 goto release_guest;
199 196
200 /* Initialize the Guest's shadow page tables, using the toplevel 197 /* Initialize the Guest's shadow page tables, using the toplevel
201 * address the Launcher gave us. This allocates memory, so can fail. */ 198 * address the Launcher gave us. This allocates memory, so can fail. */
202 err = init_guest_pagetable(lg, args[2]); 199 err = init_guest_pagetable(lg);
203 if (err) 200 if (err)
204 goto free_regs; 201 goto free_regs;
205 202
diff --git a/drivers/lguest/page_tables.c b/drivers/lguest/page_tables.c
index 81d0c6053447..576a8318221c 100644
--- a/drivers/lguest/page_tables.c
+++ b/drivers/lguest/page_tables.c
@@ -14,6 +14,7 @@
14#include <linux/percpu.h> 14#include <linux/percpu.h>
15#include <asm/tlbflush.h> 15#include <asm/tlbflush.h>
16#include <asm/uaccess.h> 16#include <asm/uaccess.h>
17#include <asm/bootparam.h>
17#include "lg.h" 18#include "lg.h"
18 19
19/*M:008 We hold reference to pages, which prevents them from being swapped. 20/*M:008 We hold reference to pages, which prevents them from being swapped.
@@ -581,15 +582,82 @@ void guest_set_pmd(struct lguest *lg, unsigned long gpgdir, u32 idx)
581 release_pgd(lg, lg->pgdirs[pgdir].pgdir + idx); 582 release_pgd(lg, lg->pgdirs[pgdir].pgdir + idx);
582} 583}
583 584
585/* Once we know how much memory we have we can construct simple identity
586 * (which set virtual == physical) and linear mappings
587 * which will get the Guest far enough into the boot to create its own.
588 *
589 * We lay them out of the way, just below the initrd (which is why we need to
590 * know its size here). */
591static unsigned long setup_pagetables(struct lguest *lg,
592 unsigned long mem,
593 unsigned long initrd_size)
594{
595 pgd_t __user *pgdir;
596 pte_t __user *linear;
597 unsigned int mapped_pages, i, linear_pages, phys_linear;
598 unsigned long mem_base = (unsigned long)lg->mem_base;
599
600 /* We have mapped_pages frames to map, so we need
601 * linear_pages page tables to map them. */
602 mapped_pages = mem / PAGE_SIZE;
603 linear_pages = (mapped_pages + PTRS_PER_PTE - 1) / PTRS_PER_PTE;
604
605 /* We put the toplevel page directory page at the top of memory. */
606 pgdir = (pgd_t *)(mem + mem_base - initrd_size - PAGE_SIZE);
607
608 /* Now we use the next linear_pages pages as pte pages */
609 linear = (void *)pgdir - linear_pages * PAGE_SIZE;
610
611 /* Linear mapping is easy: put every page's address into the
612 * mapping in order. */
613 for (i = 0; i < mapped_pages; i++) {
614 pte_t pte;
615 pte = pfn_pte(i, __pgprot(_PAGE_PRESENT|_PAGE_RW|_PAGE_USER));
616 if (copy_to_user(&linear[i], &pte, sizeof(pte)) != 0)
617 return -EFAULT;
618 }
619
620 /* The top level points to the linear page table pages above.
621 * We setup the identity and linear mappings here. */
622 phys_linear = (unsigned long)linear - mem_base;
623 for (i = 0; i < mapped_pages; i += PTRS_PER_PTE) {
624 pgd_t pgd;
625 pgd = __pgd((phys_linear + i * sizeof(pte_t)) |
626 (_PAGE_PRESENT | _PAGE_RW | _PAGE_USER));
627
628 if (copy_to_user(&pgdir[i / PTRS_PER_PTE], &pgd, sizeof(pgd))
629 || copy_to_user(&pgdir[pgd_index(PAGE_OFFSET)
630 + i / PTRS_PER_PTE],
631 &pgd, sizeof(pgd)))
632 return -EFAULT;
633 }
634
635 /* We return the top level (guest-physical) address: remember where
636 * this is. */
637 return (unsigned long)pgdir - mem_base;
638}
639
584/*H:500 (vii) Setting up the page tables initially. 640/*H:500 (vii) Setting up the page tables initially.
585 * 641 *
586 * When a Guest is first created, the Launcher tells us where the toplevel of 642 * When a Guest is first created, the Launcher tells us where the toplevel of
587 * its first page table is. We set some things up here: */ 643 * its first page table is. We set some things up here: */
588int init_guest_pagetable(struct lguest *lg, unsigned long pgtable) 644int init_guest_pagetable(struct lguest *lg)
589{ 645{
646 u64 mem;
647 u32 initrd_size;
648 struct boot_params __user *boot = (struct boot_params *)lg->mem_base;
649
650 /* Get the Guest memory size and the ramdisk size from the boot header
651 * located at lg->mem_base (Guest address 0). */
652 if (copy_from_user(&mem, &boot->e820_map[0].size, sizeof(mem))
653 || get_user(initrd_size, &boot->hdr.ramdisk_size))
654 return -EFAULT;
655
590 /* We start on the first shadow page table, and give it a blank PGD 656 /* We start on the first shadow page table, and give it a blank PGD
591 * page. */ 657 * page. */
592 lg->pgdirs[0].gpgdir = pgtable; 658 lg->pgdirs[0].gpgdir = setup_pagetables(lg, mem, initrd_size);
659 if (IS_ERR_VALUE(lg->pgdirs[0].gpgdir))
660 return lg->pgdirs[0].gpgdir;
593 lg->pgdirs[0].pgdir = (pgd_t *)get_zeroed_page(GFP_KERNEL); 661 lg->pgdirs[0].pgdir = (pgd_t *)get_zeroed_page(GFP_KERNEL);
594 if (!lg->pgdirs[0].pgdir) 662 if (!lg->pgdirs[0].pgdir)
595 return -ENOMEM; 663 return -ENOMEM;
diff --git a/drivers/s390/kvm/kvm_virtio.c b/drivers/s390/kvm/kvm_virtio.c
index 3d442444c618..28c90b89f2b4 100644
--- a/drivers/s390/kvm/kvm_virtio.c
+++ b/drivers/s390/kvm/kvm_virtio.c
@@ -188,11 +188,13 @@ static struct virtqueue *kvm_find_vq(struct virtio_device *vdev,
188 config = kvm_vq_config(kdev->desc)+index; 188 config = kvm_vq_config(kdev->desc)+index;
189 189
190 err = vmem_add_mapping(config->address, 190 err = vmem_add_mapping(config->address,
191 vring_size(config->num, PAGE_SIZE)); 191 vring_size(config->num,
192 KVM_S390_VIRTIO_RING_ALIGN));
192 if (err) 193 if (err)
193 goto out; 194 goto out;
194 195
195 vq = vring_new_virtqueue(config->num, vdev, (void *) config->address, 196 vq = vring_new_virtqueue(config->num, KVM_S390_VIRTIO_RING_ALIGN,
197 vdev, (void *) config->address,
196 kvm_notify, callback); 198 kvm_notify, callback);
197 if (!vq) { 199 if (!vq) {
198 err = -ENOMEM; 200 err = -ENOMEM;
@@ -209,7 +211,8 @@ static struct virtqueue *kvm_find_vq(struct virtio_device *vdev,
209 return vq; 211 return vq;
210unmap: 212unmap:
211 vmem_remove_mapping(config->address, 213 vmem_remove_mapping(config->address,
212 vring_size(config->num, PAGE_SIZE)); 214 vring_size(config->num,
215 KVM_S390_VIRTIO_RING_ALIGN));
213out: 216out:
214 return ERR_PTR(err); 217 return ERR_PTR(err);
215} 218}
@@ -220,7 +223,8 @@ static void kvm_del_vq(struct virtqueue *vq)
220 223
221 vring_del_virtqueue(vq); 224 vring_del_virtqueue(vq);
222 vmem_remove_mapping(config->address, 225 vmem_remove_mapping(config->address,
223 vring_size(config->num, PAGE_SIZE)); 226 vring_size(config->num,
227 KVM_S390_VIRTIO_RING_ALIGN));
224} 228}
225 229
226/* 230/*
@@ -295,13 +299,29 @@ static void scan_devices(void)
295 */ 299 */
296static void kvm_extint_handler(u16 code) 300static void kvm_extint_handler(u16 code)
297{ 301{
298 void *data = (void *) *(long *) __LC_PFAULT_INTPARM; 302 struct virtqueue *vq;
299 u16 subcode = S390_lowcore.cpu_addr; 303 u16 subcode;
304 int config_changed;
300 305
306 subcode = S390_lowcore.cpu_addr;
301 if ((subcode & 0xff00) != VIRTIO_SUBCODE_64) 307 if ((subcode & 0xff00) != VIRTIO_SUBCODE_64)
302 return; 308 return;
303 309
304 vring_interrupt(0, data); 310 /* The LSB might be overloaded, we have to mask it */
311 vq = (struct virtqueue *) ((*(long *) __LC_PFAULT_INTPARM) & ~1UL);
312
313 /* We use the LSB of extparam, to decide, if this interrupt is a config
314 * change or a "standard" interrupt */
315 config_changed = (*(int *) __LC_EXT_PARAMS & 1);
316
317 if (config_changed) {
318 struct virtio_driver *drv;
319 drv = container_of(vq->vdev->dev.driver,
320 struct virtio_driver, driver);
321 if (drv->config_changed)
322 drv->config_changed(vq->vdev);
323 } else
324 vring_interrupt(0, vq);
305} 325}
306 326
307/* 327/*
diff --git a/drivers/virtio/virtio.c b/drivers/virtio/virtio.c
index 5b78fd0aff0a..018c070a357f 100644
--- a/drivers/virtio/virtio.c
+++ b/drivers/virtio/virtio.c
@@ -176,7 +176,7 @@ int register_virtio_device(struct virtio_device *dev)
176 176
177 /* Assign a unique device index and hence name. */ 177 /* Assign a unique device index and hence name. */
178 dev->index = dev_index++; 178 dev->index = dev_index++;
179 sprintf(dev->dev.bus_id, "virtio%u", dev->index); 179 dev_set_name(&dev->dev, "virtio%u", dev->index);
180 180
181 /* We always start by resetting the device, in case a previous 181 /* We always start by resetting the device, in case a previous
182 * driver messed it up. This also tests that code path a little. */ 182 * driver messed it up. This also tests that code path a little. */
diff --git a/drivers/virtio/virtio_balloon.c b/drivers/virtio/virtio_balloon.c
index 62eab43152d2..59268266b79a 100644
--- a/drivers/virtio/virtio_balloon.c
+++ b/drivers/virtio/virtio_balloon.c
@@ -56,6 +56,15 @@ static struct virtio_device_id id_table[] = {
56 { 0 }, 56 { 0 },
57}; 57};
58 58
59static u32 page_to_balloon_pfn(struct page *page)
60{
61 unsigned long pfn = page_to_pfn(page);
62
63 BUILD_BUG_ON(PAGE_SHIFT < VIRTIO_BALLOON_PFN_SHIFT);
64 /* Convert pfn from Linux page size to balloon page size. */
65 return pfn >> (PAGE_SHIFT - VIRTIO_BALLOON_PFN_SHIFT);
66}
67
59static void balloon_ack(struct virtqueue *vq) 68static void balloon_ack(struct virtqueue *vq)
60{ 69{
61 struct virtio_balloon *vb; 70 struct virtio_balloon *vb;
@@ -99,7 +108,7 @@ static void fill_balloon(struct virtio_balloon *vb, size_t num)
99 msleep(200); 108 msleep(200);
100 break; 109 break;
101 } 110 }
102 vb->pfns[vb->num_pfns] = page_to_pfn(page); 111 vb->pfns[vb->num_pfns] = page_to_balloon_pfn(page);
103 totalram_pages--; 112 totalram_pages--;
104 vb->num_pages++; 113 vb->num_pages++;
105 list_add(&page->lru, &vb->pages); 114 list_add(&page->lru, &vb->pages);
@@ -132,7 +141,7 @@ static void leak_balloon(struct virtio_balloon *vb, size_t num)
132 for (vb->num_pfns = 0; vb->num_pfns < num; vb->num_pfns++) { 141 for (vb->num_pfns = 0; vb->num_pfns < num; vb->num_pfns++) {
133 page = list_first_entry(&vb->pages, struct page, lru); 142 page = list_first_entry(&vb->pages, struct page, lru);
134 list_del(&page->lru); 143 list_del(&page->lru);
135 vb->pfns[vb->num_pfns] = page_to_pfn(page); 144 vb->pfns[vb->num_pfns] = page_to_balloon_pfn(page);
136 vb->num_pages--; 145 vb->num_pages--;
137 } 146 }
138 147
diff --git a/drivers/virtio/virtio_pci.c b/drivers/virtio/virtio_pci.c
index c7dc37c7cce9..265fdf2d1276 100644
--- a/drivers/virtio/virtio_pci.c
+++ b/drivers/virtio/virtio_pci.c
@@ -75,7 +75,7 @@ MODULE_DEVICE_TABLE(pci, virtio_pci_id_table);
75 * would make more sense for virtio to not insist on having it's own device. */ 75 * would make more sense for virtio to not insist on having it's own device. */
76static struct device virtio_pci_root = { 76static struct device virtio_pci_root = {
77 .parent = NULL, 77 .parent = NULL,
78 .bus_id = "virtio-pci", 78 .init_name = "virtio-pci",
79}; 79};
80 80
81/* Convert a generic virtio device to our structure */ 81/* Convert a generic virtio device to our structure */
@@ -216,7 +216,7 @@ static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
216 struct virtio_pci_device *vp_dev = to_vp_device(vdev); 216 struct virtio_pci_device *vp_dev = to_vp_device(vdev);
217 struct virtio_pci_vq_info *info; 217 struct virtio_pci_vq_info *info;
218 struct virtqueue *vq; 218 struct virtqueue *vq;
219 unsigned long flags; 219 unsigned long flags, size;
220 u16 num; 220 u16 num;
221 int err; 221 int err;
222 222
@@ -237,19 +237,20 @@ static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
237 info->queue_index = index; 237 info->queue_index = index;
238 info->num = num; 238 info->num = num;
239 239
240 info->queue = kzalloc(PAGE_ALIGN(vring_size(num,PAGE_SIZE)), GFP_KERNEL); 240 size = PAGE_ALIGN(vring_size(num, VIRTIO_PCI_VRING_ALIGN));
241 info->queue = alloc_pages_exact(size, GFP_KERNEL|__GFP_ZERO);
241 if (info->queue == NULL) { 242 if (info->queue == NULL) {
242 err = -ENOMEM; 243 err = -ENOMEM;
243 goto out_info; 244 goto out_info;
244 } 245 }
245 246
246 /* activate the queue */ 247 /* activate the queue */
247 iowrite32(virt_to_phys(info->queue) >> PAGE_SHIFT, 248 iowrite32(virt_to_phys(info->queue) >> VIRTIO_PCI_QUEUE_ADDR_SHIFT,
248 vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN); 249 vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
249 250
250 /* create the vring */ 251 /* create the vring */
251 vq = vring_new_virtqueue(info->num, vdev, info->queue, 252 vq = vring_new_virtqueue(info->num, VIRTIO_PCI_VRING_ALIGN,
252 vp_notify, callback); 253 vdev, info->queue, vp_notify, callback);
253 if (!vq) { 254 if (!vq) {
254 err = -ENOMEM; 255 err = -ENOMEM;
255 goto out_activate_queue; 256 goto out_activate_queue;
@@ -266,7 +267,7 @@ static struct virtqueue *vp_find_vq(struct virtio_device *vdev, unsigned index,
266 267
267out_activate_queue: 268out_activate_queue:
268 iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN); 269 iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
269 kfree(info->queue); 270 free_pages_exact(info->queue, size);
270out_info: 271out_info:
271 kfree(info); 272 kfree(info);
272 return ERR_PTR(err); 273 return ERR_PTR(err);
@@ -277,7 +278,7 @@ static void vp_del_vq(struct virtqueue *vq)
277{ 278{
278 struct virtio_pci_device *vp_dev = to_vp_device(vq->vdev); 279 struct virtio_pci_device *vp_dev = to_vp_device(vq->vdev);
279 struct virtio_pci_vq_info *info = vq->priv; 280 struct virtio_pci_vq_info *info = vq->priv;
280 unsigned long flags; 281 unsigned long flags, size;
281 282
282 spin_lock_irqsave(&vp_dev->lock, flags); 283 spin_lock_irqsave(&vp_dev->lock, flags);
283 list_del(&info->node); 284 list_del(&info->node);
@@ -289,7 +290,8 @@ static void vp_del_vq(struct virtqueue *vq)
289 iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL); 290 iowrite16(info->queue_index, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_SEL);
290 iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN); 291 iowrite32(0, vp_dev->ioaddr + VIRTIO_PCI_QUEUE_PFN);
291 292
292 kfree(info->queue); 293 size = PAGE_ALIGN(vring_size(info->num, VIRTIO_PCI_VRING_ALIGN));
294 free_pages_exact(info->queue, size);
293 kfree(info); 295 kfree(info);
294} 296}
295 297
@@ -305,6 +307,20 @@ static struct virtio_config_ops virtio_pci_config_ops = {
305 .finalize_features = vp_finalize_features, 307 .finalize_features = vp_finalize_features,
306}; 308};
307 309
310static void virtio_pci_release_dev(struct device *_d)
311{
312 struct virtio_device *dev = container_of(_d, struct virtio_device, dev);
313 struct virtio_pci_device *vp_dev = to_vp_device(dev);
314 struct pci_dev *pci_dev = vp_dev->pci_dev;
315
316 free_irq(pci_dev->irq, vp_dev);
317 pci_set_drvdata(pci_dev, NULL);
318 pci_iounmap(pci_dev, vp_dev->ioaddr);
319 pci_release_regions(pci_dev);
320 pci_disable_device(pci_dev);
321 kfree(vp_dev);
322}
323
308/* the PCI probing function */ 324/* the PCI probing function */
309static int __devinit virtio_pci_probe(struct pci_dev *pci_dev, 325static int __devinit virtio_pci_probe(struct pci_dev *pci_dev,
310 const struct pci_device_id *id) 326 const struct pci_device_id *id)
@@ -328,6 +344,7 @@ static int __devinit virtio_pci_probe(struct pci_dev *pci_dev,
328 return -ENOMEM; 344 return -ENOMEM;
329 345
330 vp_dev->vdev.dev.parent = &virtio_pci_root; 346 vp_dev->vdev.dev.parent = &virtio_pci_root;
347 vp_dev->vdev.dev.release = virtio_pci_release_dev;
331 vp_dev->vdev.config = &virtio_pci_config_ops; 348 vp_dev->vdev.config = &virtio_pci_config_ops;
332 vp_dev->pci_dev = pci_dev; 349 vp_dev->pci_dev = pci_dev;
333 INIT_LIST_HEAD(&vp_dev->virtqueues); 350 INIT_LIST_HEAD(&vp_dev->virtqueues);
@@ -357,7 +374,7 @@ static int __devinit virtio_pci_probe(struct pci_dev *pci_dev,
357 374
358 /* register a handler for the queue with the PCI device's interrupt */ 375 /* register a handler for the queue with the PCI device's interrupt */
359 err = request_irq(vp_dev->pci_dev->irq, vp_interrupt, IRQF_SHARED, 376 err = request_irq(vp_dev->pci_dev->irq, vp_interrupt, IRQF_SHARED,
360 vp_dev->vdev.dev.bus_id, vp_dev); 377 dev_name(&vp_dev->vdev.dev), vp_dev);
361 if (err) 378 if (err)
362 goto out_set_drvdata; 379 goto out_set_drvdata;
363 380
@@ -387,12 +404,6 @@ static void __devexit virtio_pci_remove(struct pci_dev *pci_dev)
387 struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev); 404 struct virtio_pci_device *vp_dev = pci_get_drvdata(pci_dev);
388 405
389 unregister_virtio_device(&vp_dev->vdev); 406 unregister_virtio_device(&vp_dev->vdev);
390 free_irq(pci_dev->irq, vp_dev);
391 pci_set_drvdata(pci_dev, NULL);
392 pci_iounmap(pci_dev, vp_dev->ioaddr);
393 pci_release_regions(pci_dev);
394 pci_disable_device(pci_dev);
395 kfree(vp_dev);
396} 407}
397 408
398#ifdef CONFIG_PM 409#ifdef CONFIG_PM
diff --git a/drivers/virtio/virtio_ring.c b/drivers/virtio/virtio_ring.c
index 6eb5303fed11..5777196bf6c9 100644
--- a/drivers/virtio/virtio_ring.c
+++ b/drivers/virtio/virtio_ring.c
@@ -274,6 +274,7 @@ static struct virtqueue_ops vring_vq_ops = {
274}; 274};
275 275
276struct virtqueue *vring_new_virtqueue(unsigned int num, 276struct virtqueue *vring_new_virtqueue(unsigned int num,
277 unsigned int vring_align,
277 struct virtio_device *vdev, 278 struct virtio_device *vdev,
278 void *pages, 279 void *pages,
279 void (*notify)(struct virtqueue *), 280 void (*notify)(struct virtqueue *),
@@ -292,7 +293,7 @@ struct virtqueue *vring_new_virtqueue(unsigned int num,
292 if (!vq) 293 if (!vq)
293 return NULL; 294 return NULL;
294 295
295 vring_init(&vq->vring, num, pages, PAGE_SIZE); 296 vring_init(&vq->vring, num, pages, vring_align);
296 vq->vq.callback = callback; 297 vq->vq.callback = callback;
297 vq->vq.vdev = vdev; 298 vq->vq.vdev = vdev;
298 vq->vq.vq_ops = &vring_vq_ops; 299 vq->vq.vq_ops = &vring_vq_ops;
diff --git a/include/linux/lguest_launcher.h b/include/linux/lguest_launcher.h
index e7217dc58f39..a53407a4165c 100644
--- a/include/linux/lguest_launcher.h
+++ b/include/linux/lguest_launcher.h
@@ -54,9 +54,13 @@ struct lguest_vqconfig {
54/* Write command first word is a request. */ 54/* Write command first word is a request. */
55enum lguest_req 55enum lguest_req
56{ 56{
57 LHREQ_INITIALIZE, /* + base, pfnlimit, pgdir, start */ 57 LHREQ_INITIALIZE, /* + base, pfnlimit, start */
58 LHREQ_GETDMA, /* No longer used */ 58 LHREQ_GETDMA, /* No longer used */
59 LHREQ_IRQ, /* + irq */ 59 LHREQ_IRQ, /* + irq */
60 LHREQ_BREAK, /* + on/off flag (on blocks until someone does off) */ 60 LHREQ_BREAK, /* + on/off flag (on blocks until someone does off) */
61}; 61};
62
63/* The alignment to use between consumer and producer parts of vring.
64 * x86 pagesize for historical reasons. */
65#define LGUEST_VRING_ALIGN 4096
62#endif /* _LINUX_LGUEST_LAUNCHER */ 66#endif /* _LINUX_LGUEST_LAUNCHER */
diff --git a/include/linux/virtio_balloon.h b/include/linux/virtio_balloon.h
index c30c7bfbf39b..8726ff77763e 100644
--- a/include/linux/virtio_balloon.h
+++ b/include/linux/virtio_balloon.h
@@ -10,6 +10,9 @@
10/* The feature bitmap for virtio balloon */ 10/* The feature bitmap for virtio balloon */
11#define VIRTIO_BALLOON_F_MUST_TELL_HOST 0 /* Tell before reclaiming pages */ 11#define VIRTIO_BALLOON_F_MUST_TELL_HOST 0 /* Tell before reclaiming pages */
12 12
13/* Size of a PFN in the balloon interface. */
14#define VIRTIO_BALLOON_PFN_SHIFT 12
15
13struct virtio_balloon_config 16struct virtio_balloon_config
14{ 17{
15 /* Number of pages host wants Guest to give up. */ 18 /* Number of pages host wants Guest to give up. */
diff --git a/include/linux/virtio_console.h b/include/linux/virtio_console.h
index 19a0da0dba41..7615ffcdd555 100644
--- a/include/linux/virtio_console.h
+++ b/include/linux/virtio_console.h
@@ -7,6 +7,17 @@
7/* The ID for virtio console */ 7/* The ID for virtio console */
8#define VIRTIO_ID_CONSOLE 3 8#define VIRTIO_ID_CONSOLE 3
9 9
10/* Feature bits */
11#define VIRTIO_CONSOLE_F_SIZE 0 /* Does host provide console size? */
12
13struct virtio_console_config {
14 /* colums of the screens */
15 __u16 cols;
16 /* rows of the screens */
17 __u16 rows;
18} __attribute__((packed));
19
20
10#ifdef __KERNEL__ 21#ifdef __KERNEL__
11int __init virtio_cons_early_init(int (*put_chars)(u32, const char *, int)); 22int __init virtio_cons_early_init(int (*put_chars)(u32, const char *, int));
12#endif /* __KERNEL__ */ 23#endif /* __KERNEL__ */
diff --git a/include/linux/virtio_pci.h b/include/linux/virtio_pci.h
index cdef35742932..cd0fd5d181a6 100644
--- a/include/linux/virtio_pci.h
+++ b/include/linux/virtio_pci.h
@@ -53,4 +53,12 @@
53 53
54/* Virtio ABI version, this must match exactly */ 54/* Virtio ABI version, this must match exactly */
55#define VIRTIO_PCI_ABI_VERSION 0 55#define VIRTIO_PCI_ABI_VERSION 0
56
57/* How many bits to shift physical queue address written to QUEUE_PFN.
58 * 12 is historical, and due to x86 page size. */
59#define VIRTIO_PCI_QUEUE_ADDR_SHIFT 12
60
61/* The alignment to use between consumer and producer parts of vring.
62 * x86 pagesize again. */
63#define VIRTIO_PCI_VRING_ALIGN 4096
56#endif 64#endif
diff --git a/include/linux/virtio_ring.h b/include/linux/virtio_ring.h
index c4a598fb3826..71e03722fb59 100644
--- a/include/linux/virtio_ring.h
+++ b/include/linux/virtio_ring.h
@@ -83,7 +83,7 @@ struct vring {
83 * __u16 avail_idx; 83 * __u16 avail_idx;
84 * __u16 available[num]; 84 * __u16 available[num];
85 * 85 *
86 * // Padding to the next page boundary. 86 * // Padding to the next align boundary.
87 * char pad[]; 87 * char pad[];
88 * 88 *
89 * // A ring of used descriptor heads with free-running index. 89 * // A ring of used descriptor heads with free-running index.
@@ -93,19 +93,19 @@ struct vring {
93 * }; 93 * };
94 */ 94 */
95static inline void vring_init(struct vring *vr, unsigned int num, void *p, 95static inline void vring_init(struct vring *vr, unsigned int num, void *p,
96 unsigned long pagesize) 96 unsigned long align)
97{ 97{
98 vr->num = num; 98 vr->num = num;
99 vr->desc = p; 99 vr->desc = p;
100 vr->avail = p + num*sizeof(struct vring_desc); 100 vr->avail = p + num*sizeof(struct vring_desc);
101 vr->used = (void *)(((unsigned long)&vr->avail->ring[num] + pagesize-1) 101 vr->used = (void *)(((unsigned long)&vr->avail->ring[num] + align-1)
102 & ~(pagesize - 1)); 102 & ~(align - 1));
103} 103}
104 104
105static inline unsigned vring_size(unsigned int num, unsigned long pagesize) 105static inline unsigned vring_size(unsigned int num, unsigned long align)
106{ 106{
107 return ((sizeof(struct vring_desc) * num + sizeof(__u16) * (2 + num) 107 return ((sizeof(struct vring_desc) * num + sizeof(__u16) * (2 + num)
108 + pagesize - 1) & ~(pagesize - 1)) 108 + align - 1) & ~(align - 1))
109 + sizeof(__u16) * 2 + sizeof(struct vring_used_elem) * num; 109 + sizeof(__u16) * 2 + sizeof(struct vring_used_elem) * num;
110} 110}
111 111
@@ -115,6 +115,7 @@ struct virtio_device;
115struct virtqueue; 115struct virtqueue;
116 116
117struct virtqueue *vring_new_virtqueue(unsigned int num, 117struct virtqueue *vring_new_virtqueue(unsigned int num,
118 unsigned int vring_align,
118 struct virtio_device *vdev, 119 struct virtio_device *vdev,
119 void *pages, 120 void *pages,
120 void (*notify)(struct virtqueue *vq), 121 void (*notify)(struct virtqueue *vq),