aboutsummaryrefslogtreecommitdiffstats
path: root/drivers/block/xen-blkback
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2011-11-04 20:22:14 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2011-11-04 20:22:14 -0400
commit3d0a8d10cfb4cc3d1877c29a866ee7d8a46aa2fa (patch)
tree11a85044d1472f5972ae47ce10a2f446ad981e9f /drivers/block/xen-blkback
parentb4fdcb02f1e39c27058a885905bd0277370ba441 (diff)
parenta0eda62552eba4e1f92d5354bb65c68fb6b45f87 (diff)
Merge branch 'for-3.2/drivers' of git://git.kernel.dk/linux-block
* 'for-3.2/drivers' of git://git.kernel.dk/linux-block: (30 commits) virtio-blk: use ida to allocate disk index hpsa: add small delay when using PCI Power Management to reset for kump cciss: add small delay when using PCI Power Management to reset for kump xen/blkback: Fix two races in the handling of barrier requests. xen/blkback: Check for proper operation. xen/blkback: Fix the inhibition to map pages when discarding sector ranges. xen/blkback: Report VBD_WSECT (wr_sect) properly. xen/blkback: Support 'feature-barrier' aka old-style BARRIER requests. xen-blkfront: plug device number leak in xlblk_init() error path xen-blkfront: If no barrier or flush is supported, use invalid operation. xen-blkback: use kzalloc() in favor of kmalloc()+memset() xen-blkback: fixed indentation and comments xen-blkfront: fix a deadlock while handling discard response xen-blkfront: Handle discard requests. xen-blkback: Implement discard requests ('feature-discard') xen-blkfront: add BLKIF_OP_DISCARD and discard request struct drivers/block/loop.c: remove unnecessary bdev argument from loop_clr_fd() drivers/block/loop.c: emit uevent on auto release drivers/block/cpqarray.c: use pci_dev->revision loop: always allow userspace partitions and optionally support automatic scanning ... Fic up trivial header file includsion conflict in drivers/block/loop.c
Diffstat (limited to 'drivers/block/xen-blkback')
-rw-r--r--drivers/block/xen-blkback/blkback.c130
-rw-r--r--drivers/block/xen-blkback/common.h98
-rw-r--r--drivers/block/xen-blkback/xenbus.c76
3 files changed, 266 insertions, 38 deletions
diff --git a/drivers/block/xen-blkback/blkback.c b/drivers/block/xen-blkback/blkback.c
index 1540792b1e54..15ec4db194d1 100644
--- a/drivers/block/xen-blkback/blkback.c
+++ b/drivers/block/xen-blkback/blkback.c
@@ -39,6 +39,9 @@
39#include <linux/list.h> 39#include <linux/list.h>
40#include <linux/delay.h> 40#include <linux/delay.h>
41#include <linux/freezer.h> 41#include <linux/freezer.h>
42#include <linux/loop.h>
43#include <linux/falloc.h>
44#include <linux/fs.h>
42 45
43#include <xen/events.h> 46#include <xen/events.h>
44#include <xen/page.h> 47#include <xen/page.h>
@@ -258,13 +261,16 @@ irqreturn_t xen_blkif_be_int(int irq, void *dev_id)
258 261
259static void print_stats(struct xen_blkif *blkif) 262static void print_stats(struct xen_blkif *blkif)
260{ 263{
261 pr_info("xen-blkback (%s): oo %3d | rd %4d | wr %4d | f %4d\n", 264 pr_info("xen-blkback (%s): oo %3d | rd %4d | wr %4d | f %4d"
265 " | ds %4d\n",
262 current->comm, blkif->st_oo_req, 266 current->comm, blkif->st_oo_req,
263 blkif->st_rd_req, blkif->st_wr_req, blkif->st_f_req); 267 blkif->st_rd_req, blkif->st_wr_req,
268 blkif->st_f_req, blkif->st_ds_req);
264 blkif->st_print = jiffies + msecs_to_jiffies(10 * 1000); 269 blkif->st_print = jiffies + msecs_to_jiffies(10 * 1000);
265 blkif->st_rd_req = 0; 270 blkif->st_rd_req = 0;
266 blkif->st_wr_req = 0; 271 blkif->st_wr_req = 0;
267 blkif->st_oo_req = 0; 272 blkif->st_oo_req = 0;
273 blkif->st_ds_req = 0;
268} 274}
269 275
270int xen_blkif_schedule(void *arg) 276int xen_blkif_schedule(void *arg)
@@ -410,6 +416,59 @@ static int xen_blkbk_map(struct blkif_request *req,
410 return ret; 416 return ret;
411} 417}
412 418
419static void xen_blk_discard(struct xen_blkif *blkif, struct blkif_request *req)
420{
421 int err = 0;
422 int status = BLKIF_RSP_OKAY;
423 struct block_device *bdev = blkif->vbd.bdev;
424
425 if (blkif->blk_backend_type == BLKIF_BACKEND_PHY)
426 /* just forward the discard request */
427 err = blkdev_issue_discard(bdev,
428 req->u.discard.sector_number,
429 req->u.discard.nr_sectors,
430 GFP_KERNEL, 0);
431 else if (blkif->blk_backend_type == BLKIF_BACKEND_FILE) {
432 /* punch a hole in the backing file */
433 struct loop_device *lo = bdev->bd_disk->private_data;
434 struct file *file = lo->lo_backing_file;
435
436 if (file->f_op->fallocate)
437 err = file->f_op->fallocate(file,
438 FALLOC_FL_KEEP_SIZE | FALLOC_FL_PUNCH_HOLE,
439 req->u.discard.sector_number << 9,
440 req->u.discard.nr_sectors << 9);
441 else
442 err = -EOPNOTSUPP;
443 } else
444 err = -EOPNOTSUPP;
445
446 if (err == -EOPNOTSUPP) {
447 pr_debug(DRV_PFX "discard op failed, not supported\n");
448 status = BLKIF_RSP_EOPNOTSUPP;
449 } else if (err)
450 status = BLKIF_RSP_ERROR;
451
452 make_response(blkif, req->id, req->operation, status);
453}
454
455static void xen_blk_drain_io(struct xen_blkif *blkif)
456{
457 atomic_set(&blkif->drain, 1);
458 do {
459 /* The initial value is one, and one refcnt taken at the
460 * start of the xen_blkif_schedule thread. */
461 if (atomic_read(&blkif->refcnt) <= 2)
462 break;
463 wait_for_completion_interruptible_timeout(
464 &blkif->drain_complete, HZ);
465
466 if (!atomic_read(&blkif->drain))
467 break;
468 } while (!kthread_should_stop());
469 atomic_set(&blkif->drain, 0);
470}
471
413/* 472/*
414 * Completion callback on the bio's. Called as bh->b_end_io() 473 * Completion callback on the bio's. Called as bh->b_end_io()
415 */ 474 */
@@ -422,6 +481,11 @@ static void __end_block_io_op(struct pending_req *pending_req, int error)
422 pr_debug(DRV_PFX "flush diskcache op failed, not supported\n"); 481 pr_debug(DRV_PFX "flush diskcache op failed, not supported\n");
423 xen_blkbk_flush_diskcache(XBT_NIL, pending_req->blkif->be, 0); 482 xen_blkbk_flush_diskcache(XBT_NIL, pending_req->blkif->be, 0);
424 pending_req->status = BLKIF_RSP_EOPNOTSUPP; 483 pending_req->status = BLKIF_RSP_EOPNOTSUPP;
484 } else if ((pending_req->operation == BLKIF_OP_WRITE_BARRIER) &&
485 (error == -EOPNOTSUPP)) {
486 pr_debug(DRV_PFX "write barrier op failed, not supported\n");
487 xen_blkbk_barrier(XBT_NIL, pending_req->blkif->be, 0);
488 pending_req->status = BLKIF_RSP_EOPNOTSUPP;
425 } else if (error) { 489 } else if (error) {
426 pr_debug(DRV_PFX "Buffer not up-to-date at end of operation," 490 pr_debug(DRV_PFX "Buffer not up-to-date at end of operation,"
427 " error=%d\n", error); 491 " error=%d\n", error);
@@ -438,6 +502,10 @@ static void __end_block_io_op(struct pending_req *pending_req, int error)
438 make_response(pending_req->blkif, pending_req->id, 502 make_response(pending_req->blkif, pending_req->id,
439 pending_req->operation, pending_req->status); 503 pending_req->operation, pending_req->status);
440 xen_blkif_put(pending_req->blkif); 504 xen_blkif_put(pending_req->blkif);
505 if (atomic_read(&pending_req->blkif->refcnt) <= 2) {
506 if (atomic_read(&pending_req->blkif->drain))
507 complete(&pending_req->blkif->drain_complete);
508 }
441 free_req(pending_req); 509 free_req(pending_req);
442 } 510 }
443} 511}
@@ -532,7 +600,6 @@ do_block_io_op(struct xen_blkif *blkif)
532 600
533 return more_to_do; 601 return more_to_do;
534} 602}
535
536/* 603/*
537 * Transmutation of the 'struct blkif_request' to a proper 'struct bio' 604 * Transmutation of the 'struct blkif_request' to a proper 'struct bio'
538 * and call the 'submit_bio' to pass it to the underlying storage. 605 * and call the 'submit_bio' to pass it to the underlying storage.
@@ -549,6 +616,7 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
549 int i, nbio = 0; 616 int i, nbio = 0;
550 int operation; 617 int operation;
551 struct blk_plug plug; 618 struct blk_plug plug;
619 bool drain = false;
552 620
553 switch (req->operation) { 621 switch (req->operation) {
554 case BLKIF_OP_READ: 622 case BLKIF_OP_READ:
@@ -559,11 +627,16 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
559 blkif->st_wr_req++; 627 blkif->st_wr_req++;
560 operation = WRITE_ODIRECT; 628 operation = WRITE_ODIRECT;
561 break; 629 break;
630 case BLKIF_OP_WRITE_BARRIER:
631 drain = true;
562 case BLKIF_OP_FLUSH_DISKCACHE: 632 case BLKIF_OP_FLUSH_DISKCACHE:
563 blkif->st_f_req++; 633 blkif->st_f_req++;
564 operation = WRITE_FLUSH; 634 operation = WRITE_FLUSH;
565 break; 635 break;
566 case BLKIF_OP_WRITE_BARRIER: 636 case BLKIF_OP_DISCARD:
637 blkif->st_ds_req++;
638 operation = REQ_DISCARD;
639 break;
567 default: 640 default:
568 operation = 0; /* make gcc happy */ 641 operation = 0; /* make gcc happy */
569 goto fail_response; 642 goto fail_response;
@@ -572,7 +645,8 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
572 645
573 /* Check that the number of segments is sane. */ 646 /* Check that the number of segments is sane. */
574 nseg = req->nr_segments; 647 nseg = req->nr_segments;
575 if (unlikely(nseg == 0 && operation != WRITE_FLUSH) || 648 if (unlikely(nseg == 0 && operation != WRITE_FLUSH &&
649 operation != REQ_DISCARD) ||
576 unlikely(nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) { 650 unlikely(nseg > BLKIF_MAX_SEGMENTS_PER_REQUEST)) {
577 pr_debug(DRV_PFX "Bad number of segments in request (%d)\n", 651 pr_debug(DRV_PFX "Bad number of segments in request (%d)\n",
578 nseg); 652 nseg);
@@ -621,16 +695,25 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
621 } 695 }
622 } 696 }
623 697
698 /* Wait on all outstanding I/O's and once that has been completed
699 * issue the WRITE_FLUSH.
700 */
701 if (drain)
702 xen_blk_drain_io(pending_req->blkif);
703
624 /* 704 /*
625 * If we have failed at this point, we need to undo the M2P override, 705 * If we have failed at this point, we need to undo the M2P override,
626 * set gnttab_set_unmap_op on all of the grant references and perform 706 * set gnttab_set_unmap_op on all of the grant references and perform
627 * the hypercall to unmap the grants - that is all done in 707 * the hypercall to unmap the grants - that is all done in
628 * xen_blkbk_unmap. 708 * xen_blkbk_unmap.
629 */ 709 */
630 if (xen_blkbk_map(req, pending_req, seg)) 710 if (operation != REQ_DISCARD && xen_blkbk_map(req, pending_req, seg))
631 goto fail_flush; 711 goto fail_flush;
632 712
633 /* This corresponding xen_blkif_put is done in __end_block_io_op */ 713 /*
714 * This corresponding xen_blkif_put is done in __end_block_io_op, or
715 * below (in "!bio") if we are handling a BLKIF_OP_DISCARD.
716 */
634 xen_blkif_get(blkif); 717 xen_blkif_get(blkif);
635 718
636 for (i = 0; i < nseg; i++) { 719 for (i = 0; i < nseg; i++) {
@@ -654,18 +737,25 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
654 preq.sector_number += seg[i].nsec; 737 preq.sector_number += seg[i].nsec;
655 } 738 }
656 739
657 /* This will be hit if the operation was a flush. */ 740 /* This will be hit if the operation was a flush or discard. */
658 if (!bio) { 741 if (!bio) {
659 BUG_ON(operation != WRITE_FLUSH); 742 BUG_ON(operation != WRITE_FLUSH && operation != REQ_DISCARD);
660 743
661 bio = bio_alloc(GFP_KERNEL, 0); 744 if (operation == WRITE_FLUSH) {
662 if (unlikely(bio == NULL)) 745 bio = bio_alloc(GFP_KERNEL, 0);
663 goto fail_put_bio; 746 if (unlikely(bio == NULL))
747 goto fail_put_bio;
664 748
665 biolist[nbio++] = bio; 749 biolist[nbio++] = bio;
666 bio->bi_bdev = preq.bdev; 750 bio->bi_bdev = preq.bdev;
667 bio->bi_private = pending_req; 751 bio->bi_private = pending_req;
668 bio->bi_end_io = end_block_io_op; 752 bio->bi_end_io = end_block_io_op;
753 } else if (operation == REQ_DISCARD) {
754 xen_blk_discard(blkif, req);
755 xen_blkif_put(blkif);
756 free_req(pending_req);
757 return 0;
758 }
669 } 759 }
670 760
671 /* 761 /*
@@ -685,7 +775,7 @@ static int dispatch_rw_block_io(struct xen_blkif *blkif,
685 775
686 if (operation == READ) 776 if (operation == READ)
687 blkif->st_rd_sect += preq.nr_sects; 777 blkif->st_rd_sect += preq.nr_sects;
688 else if (operation == WRITE || operation == WRITE_FLUSH) 778 else if (operation & WRITE)
689 blkif->st_wr_sect += preq.nr_sects; 779 blkif->st_wr_sect += preq.nr_sects;
690 780
691 return 0; 781 return 0;
@@ -765,9 +855,9 @@ static int __init xen_blkif_init(void)
765 855
766 mmap_pages = xen_blkif_reqs * BLKIF_MAX_SEGMENTS_PER_REQUEST; 856 mmap_pages = xen_blkif_reqs * BLKIF_MAX_SEGMENTS_PER_REQUEST;
767 857
768 blkbk->pending_reqs = kmalloc(sizeof(blkbk->pending_reqs[0]) * 858 blkbk->pending_reqs = kzalloc(sizeof(blkbk->pending_reqs[0]) *
769 xen_blkif_reqs, GFP_KERNEL); 859 xen_blkif_reqs, GFP_KERNEL);
770 blkbk->pending_grant_handles = kzalloc(sizeof(blkbk->pending_grant_handles[0]) * 860 blkbk->pending_grant_handles = kmalloc(sizeof(blkbk->pending_grant_handles[0]) *
771 mmap_pages, GFP_KERNEL); 861 mmap_pages, GFP_KERNEL);
772 blkbk->pending_pages = kzalloc(sizeof(blkbk->pending_pages[0]) * 862 blkbk->pending_pages = kzalloc(sizeof(blkbk->pending_pages[0]) *
773 mmap_pages, GFP_KERNEL); 863 mmap_pages, GFP_KERNEL);
@@ -790,8 +880,6 @@ static int __init xen_blkif_init(void)
790 if (rc) 880 if (rc)
791 goto failed_init; 881 goto failed_init;
792 882
793 memset(blkbk->pending_reqs, 0, sizeof(blkbk->pending_reqs));
794
795 INIT_LIST_HEAD(&blkbk->pending_free); 883 INIT_LIST_HEAD(&blkbk->pending_free);
796 spin_lock_init(&blkbk->pending_free_lock); 884 spin_lock_init(&blkbk->pending_free_lock);
797 init_waitqueue_head(&blkbk->pending_free_wq); 885 init_waitqueue_head(&blkbk->pending_free_wq);
diff --git a/drivers/block/xen-blkback/common.h b/drivers/block/xen-blkback/common.h
index c4bd34063ecc..de09f525d6c1 100644
--- a/drivers/block/xen-blkback/common.h
+++ b/drivers/block/xen-blkback/common.h
@@ -62,13 +62,26 @@ struct blkif_common_response {
62 62
63/* i386 protocol version */ 63/* i386 protocol version */
64#pragma pack(push, 4) 64#pragma pack(push, 4)
65
66struct blkif_x86_32_request_rw {
67 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
68 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
69};
70
71struct blkif_x86_32_request_discard {
72 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
73 uint64_t nr_sectors;
74};
75
65struct blkif_x86_32_request { 76struct blkif_x86_32_request {
66 uint8_t operation; /* BLKIF_OP_??? */ 77 uint8_t operation; /* BLKIF_OP_??? */
67 uint8_t nr_segments; /* number of segments */ 78 uint8_t nr_segments; /* number of segments */
68 blkif_vdev_t handle; /* only for read/write requests */ 79 blkif_vdev_t handle; /* only for read/write requests */
69 uint64_t id; /* private guest value, echoed in resp */ 80 uint64_t id; /* private guest value, echoed in resp */
70 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 81 union {
71 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 82 struct blkif_x86_32_request_rw rw;
83 struct blkif_x86_32_request_discard discard;
84 } u;
72}; 85};
73struct blkif_x86_32_response { 86struct blkif_x86_32_response {
74 uint64_t id; /* copied from request */ 87 uint64_t id; /* copied from request */
@@ -78,13 +91,26 @@ struct blkif_x86_32_response {
78#pragma pack(pop) 91#pragma pack(pop)
79 92
80/* x86_64 protocol version */ 93/* x86_64 protocol version */
94
95struct blkif_x86_64_request_rw {
96 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
97 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST];
98};
99
100struct blkif_x86_64_request_discard {
101 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */
102 uint64_t nr_sectors;
103};
104
81struct blkif_x86_64_request { 105struct blkif_x86_64_request {
82 uint8_t operation; /* BLKIF_OP_??? */ 106 uint8_t operation; /* BLKIF_OP_??? */
83 uint8_t nr_segments; /* number of segments */ 107 uint8_t nr_segments; /* number of segments */
84 blkif_vdev_t handle; /* only for read/write requests */ 108 blkif_vdev_t handle; /* only for read/write requests */
85 uint64_t __attribute__((__aligned__(8))) id; 109 uint64_t __attribute__((__aligned__(8))) id;
86 blkif_sector_t sector_number;/* start sector idx on disk (r/w only) */ 110 union {
87 struct blkif_request_segment seg[BLKIF_MAX_SEGMENTS_PER_REQUEST]; 111 struct blkif_x86_64_request_rw rw;
112 struct blkif_x86_64_request_discard discard;
113 } u;
88}; 114};
89struct blkif_x86_64_response { 115struct blkif_x86_64_response {
90 uint64_t __attribute__((__aligned__(8))) id; 116 uint64_t __attribute__((__aligned__(8))) id;
@@ -112,6 +138,11 @@ enum blkif_protocol {
112 BLKIF_PROTOCOL_X86_64 = 3, 138 BLKIF_PROTOCOL_X86_64 = 3,
113}; 139};
114 140
141enum blkif_backend_type {
142 BLKIF_BACKEND_PHY = 1,
143 BLKIF_BACKEND_FILE = 2,
144};
145
115struct xen_vbd { 146struct xen_vbd {
116 /* What the domain refers to this vbd as. */ 147 /* What the domain refers to this vbd as. */
117 blkif_vdev_t handle; 148 blkif_vdev_t handle;
@@ -137,6 +168,7 @@ struct xen_blkif {
137 unsigned int irq; 168 unsigned int irq;
138 /* Comms information. */ 169 /* Comms information. */
139 enum blkif_protocol blk_protocol; 170 enum blkif_protocol blk_protocol;
171 enum blkif_backend_type blk_backend_type;
140 union blkif_back_rings blk_rings; 172 union blkif_back_rings blk_rings;
141 struct vm_struct *blk_ring_area; 173 struct vm_struct *blk_ring_area;
142 /* The VBD attached to this interface. */ 174 /* The VBD attached to this interface. */
@@ -148,6 +180,9 @@ struct xen_blkif {
148 atomic_t refcnt; 180 atomic_t refcnt;
149 181
150 wait_queue_head_t wq; 182 wait_queue_head_t wq;
183 /* for barrier (drain) requests */
184 struct completion drain_complete;
185 atomic_t drain;
151 /* One thread per one blkif. */ 186 /* One thread per one blkif. */
152 struct task_struct *xenblkd; 187 struct task_struct *xenblkd;
153 unsigned int waiting_reqs; 188 unsigned int waiting_reqs;
@@ -158,6 +193,7 @@ struct xen_blkif {
158 int st_wr_req; 193 int st_wr_req;
159 int st_oo_req; 194 int st_oo_req;
160 int st_f_req; 195 int st_f_req;
196 int st_ds_req;
161 int st_rd_sect; 197 int st_rd_sect;
162 int st_wr_sect; 198 int st_wr_sect;
163 199
@@ -181,7 +217,7 @@ struct xen_blkif {
181 217
182struct phys_req { 218struct phys_req {
183 unsigned short dev; 219 unsigned short dev;
184 unsigned short nr_sects; 220 blkif_sector_t nr_sects;
185 struct block_device *bdev; 221 struct block_device *bdev;
186 blkif_sector_t sector_number; 222 blkif_sector_t sector_number;
187}; 223};
@@ -195,6 +231,8 @@ int xen_blkif_schedule(void *arg);
195int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt, 231int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
196 struct backend_info *be, int state); 232 struct backend_info *be, int state);
197 233
234int xen_blkbk_barrier(struct xenbus_transaction xbt,
235 struct backend_info *be, int state);
198struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be); 236struct xenbus_device *xen_blkbk_xenbus(struct backend_info *be);
199 237
200static inline void blkif_get_x86_32_req(struct blkif_request *dst, 238static inline void blkif_get_x86_32_req(struct blkif_request *dst,
@@ -205,12 +243,25 @@ static inline void blkif_get_x86_32_req(struct blkif_request *dst,
205 dst->nr_segments = src->nr_segments; 243 dst->nr_segments = src->nr_segments;
206 dst->handle = src->handle; 244 dst->handle = src->handle;
207 dst->id = src->id; 245 dst->id = src->id;
208 dst->u.rw.sector_number = src->sector_number; 246 switch (src->operation) {
209 barrier(); 247 case BLKIF_OP_READ:
210 if (n > dst->nr_segments) 248 case BLKIF_OP_WRITE:
211 n = dst->nr_segments; 249 case BLKIF_OP_WRITE_BARRIER:
212 for (i = 0; i < n; i++) 250 case BLKIF_OP_FLUSH_DISKCACHE:
213 dst->u.rw.seg[i] = src->seg[i]; 251 dst->u.rw.sector_number = src->u.rw.sector_number;
252 barrier();
253 if (n > dst->nr_segments)
254 n = dst->nr_segments;
255 for (i = 0; i < n; i++)
256 dst->u.rw.seg[i] = src->u.rw.seg[i];
257 break;
258 case BLKIF_OP_DISCARD:
259 dst->u.discard.sector_number = src->u.discard.sector_number;
260 dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
261 break;
262 default:
263 break;
264 }
214} 265}
215 266
216static inline void blkif_get_x86_64_req(struct blkif_request *dst, 267static inline void blkif_get_x86_64_req(struct blkif_request *dst,
@@ -221,12 +272,25 @@ static inline void blkif_get_x86_64_req(struct blkif_request *dst,
221 dst->nr_segments = src->nr_segments; 272 dst->nr_segments = src->nr_segments;
222 dst->handle = src->handle; 273 dst->handle = src->handle;
223 dst->id = src->id; 274 dst->id = src->id;
224 dst->u.rw.sector_number = src->sector_number; 275 switch (src->operation) {
225 barrier(); 276 case BLKIF_OP_READ:
226 if (n > dst->nr_segments) 277 case BLKIF_OP_WRITE:
227 n = dst->nr_segments; 278 case BLKIF_OP_WRITE_BARRIER:
228 for (i = 0; i < n; i++) 279 case BLKIF_OP_FLUSH_DISKCACHE:
229 dst->u.rw.seg[i] = src->seg[i]; 280 dst->u.rw.sector_number = src->u.rw.sector_number;
281 barrier();
282 if (n > dst->nr_segments)
283 n = dst->nr_segments;
284 for (i = 0; i < n; i++)
285 dst->u.rw.seg[i] = src->u.rw.seg[i];
286 break;
287 case BLKIF_OP_DISCARD:
288 dst->u.discard.sector_number = src->u.discard.sector_number;
289 dst->u.discard.nr_sectors = src->u.discard.nr_sectors;
290 break;
291 default:
292 break;
293 }
230} 294}
231 295
232#endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */ 296#endif /* __XEN_BLKIF__BACKEND__COMMON_H__ */
diff --git a/drivers/block/xen-blkback/xenbus.c b/drivers/block/xen-blkback/xenbus.c
index 5fd2010f7d2b..2c008afe63d9 100644
--- a/drivers/block/xen-blkback/xenbus.c
+++ b/drivers/block/xen-blkback/xenbus.c
@@ -114,6 +114,8 @@ static struct xen_blkif *xen_blkif_alloc(domid_t domid)
114 spin_lock_init(&blkif->blk_ring_lock); 114 spin_lock_init(&blkif->blk_ring_lock);
115 atomic_set(&blkif->refcnt, 1); 115 atomic_set(&blkif->refcnt, 1);
116 init_waitqueue_head(&blkif->wq); 116 init_waitqueue_head(&blkif->wq);
117 init_completion(&blkif->drain_complete);
118 atomic_set(&blkif->drain, 0);
117 blkif->st_print = jiffies; 119 blkif->st_print = jiffies;
118 init_waitqueue_head(&blkif->waiting_to_free); 120 init_waitqueue_head(&blkif->waiting_to_free);
119 121
@@ -272,6 +274,7 @@ VBD_SHOW(oo_req, "%d\n", be->blkif->st_oo_req);
272VBD_SHOW(rd_req, "%d\n", be->blkif->st_rd_req); 274VBD_SHOW(rd_req, "%d\n", be->blkif->st_rd_req);
273VBD_SHOW(wr_req, "%d\n", be->blkif->st_wr_req); 275VBD_SHOW(wr_req, "%d\n", be->blkif->st_wr_req);
274VBD_SHOW(f_req, "%d\n", be->blkif->st_f_req); 276VBD_SHOW(f_req, "%d\n", be->blkif->st_f_req);
277VBD_SHOW(ds_req, "%d\n", be->blkif->st_ds_req);
275VBD_SHOW(rd_sect, "%d\n", be->blkif->st_rd_sect); 278VBD_SHOW(rd_sect, "%d\n", be->blkif->st_rd_sect);
276VBD_SHOW(wr_sect, "%d\n", be->blkif->st_wr_sect); 279VBD_SHOW(wr_sect, "%d\n", be->blkif->st_wr_sect);
277 280
@@ -280,6 +283,7 @@ static struct attribute *xen_vbdstat_attrs[] = {
280 &dev_attr_rd_req.attr, 283 &dev_attr_rd_req.attr,
281 &dev_attr_wr_req.attr, 284 &dev_attr_wr_req.attr,
282 &dev_attr_f_req.attr, 285 &dev_attr_f_req.attr,
286 &dev_attr_ds_req.attr,
283 &dev_attr_rd_sect.attr, 287 &dev_attr_rd_sect.attr,
284 &dev_attr_wr_sect.attr, 288 &dev_attr_wr_sect.attr,
285 NULL 289 NULL
@@ -419,6 +423,73 @@ int xen_blkbk_flush_diskcache(struct xenbus_transaction xbt,
419 return err; 423 return err;
420} 424}
421 425
426int xen_blkbk_discard(struct xenbus_transaction xbt, struct backend_info *be)
427{
428 struct xenbus_device *dev = be->dev;
429 struct xen_blkif *blkif = be->blkif;
430 char *type;
431 int err;
432 int state = 0;
433
434 type = xenbus_read(XBT_NIL, dev->nodename, "type", NULL);
435 if (!IS_ERR(type)) {
436 if (strncmp(type, "file", 4) == 0) {
437 state = 1;
438 blkif->blk_backend_type = BLKIF_BACKEND_FILE;
439 }
440 if (strncmp(type, "phy", 3) == 0) {
441 struct block_device *bdev = be->blkif->vbd.bdev;
442 struct request_queue *q = bdev_get_queue(bdev);
443 if (blk_queue_discard(q)) {
444 err = xenbus_printf(xbt, dev->nodename,
445 "discard-granularity", "%u",
446 q->limits.discard_granularity);
447 if (err) {
448 xenbus_dev_fatal(dev, err,
449 "writing discard-granularity");
450 goto kfree;
451 }
452 err = xenbus_printf(xbt, dev->nodename,
453 "discard-alignment", "%u",
454 q->limits.discard_alignment);
455 if (err) {
456 xenbus_dev_fatal(dev, err,
457 "writing discard-alignment");
458 goto kfree;
459 }
460 state = 1;
461 blkif->blk_backend_type = BLKIF_BACKEND_PHY;
462 }
463 }
464 } else {
465 err = PTR_ERR(type);
466 xenbus_dev_fatal(dev, err, "reading type");
467 goto out;
468 }
469
470 err = xenbus_printf(xbt, dev->nodename, "feature-discard",
471 "%d", state);
472 if (err)
473 xenbus_dev_fatal(dev, err, "writing feature-discard");
474kfree:
475 kfree(type);
476out:
477 return err;
478}
479int xen_blkbk_barrier(struct xenbus_transaction xbt,
480 struct backend_info *be, int state)
481{
482 struct xenbus_device *dev = be->dev;
483 int err;
484
485 err = xenbus_printf(xbt, dev->nodename, "feature-barrier",
486 "%d", state);
487 if (err)
488 xenbus_dev_fatal(dev, err, "writing feature-barrier");
489
490 return err;
491}
492
422/* 493/*
423 * Entry point to this code when a new device is created. Allocate the basic 494 * Entry point to this code when a new device is created. Allocate the basic
424 * structures, and watch the store waiting for the hotplug scripts to tell us 495 * structures, and watch the store waiting for the hotplug scripts to tell us
@@ -650,6 +721,11 @@ again:
650 if (err) 721 if (err)
651 goto abort; 722 goto abort;
652 723
724 err = xen_blkbk_discard(xbt, be);
725
726 /* If we can't advertise it is OK. */
727 err = xen_blkbk_barrier(xbt, be, be->blkif->vbd.flush_support);
728
653 err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu", 729 err = xenbus_printf(xbt, dev->nodename, "sectors", "%llu",
654 (unsigned long long)vbd_sz(&be->blkif->vbd)); 730 (unsigned long long)vbd_sz(&be->blkif->vbd));
655 if (err) { 731 if (err) {