2 * Copyright (C) 2009-2010 Nippon Telegraph and Telephone Corporation.
4 * This program is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU General Public License version
6 * 2 as published by the Free Software Foundation.
8 * You should have received a copy of the GNU General Public License
9 * along with this program. If not, see <http://www.gnu.org/licenses/>.
11 * Contributions after 2012-01-13 are licensed under the terms of the
12 * GNU GPL, version 2 or (at your option) any later version.
15 #include "qemu-common.h"
17 #include "qemu/error-report.h"
18 #include "qemu/sockets.h"
19 #include "block/block_int.h"
20 #include "qemu/bitops.h"
22 #define SD_PROTO_VER 0x01
24 #define SD_DEFAULT_ADDR "localhost"
25 #define SD_DEFAULT_PORT 7000
27 #define SD_OP_CREATE_AND_WRITE_OBJ 0x01
28 #define SD_OP_READ_OBJ 0x02
29 #define SD_OP_WRITE_OBJ 0x03
30 /* 0x04 is used internally by Sheepdog */
31 #define SD_OP_DISCARD_OBJ 0x05
33 #define SD_OP_NEW_VDI 0x11
34 #define SD_OP_LOCK_VDI 0x12
35 #define SD_OP_RELEASE_VDI 0x13
36 #define SD_OP_GET_VDI_INFO 0x14
37 #define SD_OP_READ_VDIS 0x15
38 #define SD_OP_FLUSH_VDI 0x16
39 #define SD_OP_DEL_VDI 0x17
40 #define SD_OP_GET_CLUSTER_DEFAULT 0x18
42 #define SD_FLAG_CMD_WRITE 0x01
43 #define SD_FLAG_CMD_COW 0x02
44 #define SD_FLAG_CMD_CACHE 0x04 /* Writeback mode for cache */
45 #define SD_FLAG_CMD_DIRECT 0x08 /* Don't use cache */
47 #define SD_RES_SUCCESS 0x00 /* Success */
48 #define SD_RES_UNKNOWN 0x01 /* Unknown error */
49 #define SD_RES_NO_OBJ 0x02 /* No object found */
50 #define SD_RES_EIO 0x03 /* I/O error */
51 #define SD_RES_VDI_EXIST 0x04 /* Vdi exists already */
52 #define SD_RES_INVALID_PARMS 0x05 /* Invalid parameters */
53 #define SD_RES_SYSTEM_ERROR 0x06 /* System error */
54 #define SD_RES_VDI_LOCKED 0x07 /* Vdi is locked */
55 #define SD_RES_NO_VDI 0x08 /* No vdi found */
56 #define SD_RES_NO_BASE_VDI 0x09 /* No base vdi found */
57 #define SD_RES_VDI_READ 0x0A /* Cannot read requested vdi */
58 #define SD_RES_VDI_WRITE 0x0B /* Cannot write requested vdi */
59 #define SD_RES_BASE_VDI_READ 0x0C /* Cannot read base vdi */
60 #define SD_RES_BASE_VDI_WRITE 0x0D /* Cannot write base vdi */
61 #define SD_RES_NO_TAG 0x0E /* Requested tag is not found */
62 #define SD_RES_STARTUP 0x0F /* Sheepdog is on starting up */
63 #define SD_RES_VDI_NOT_LOCKED 0x10 /* Vdi is not locked */
64 #define SD_RES_SHUTDOWN 0x11 /* Sheepdog is shutting down */
65 #define SD_RES_NO_MEM 0x12 /* Cannot allocate memory */
66 #define SD_RES_FULL_VDI 0x13 /* we already have the maximum vdis */
67 #define SD_RES_VER_MISMATCH 0x14 /* Protocol version mismatch */
68 #define SD_RES_NO_SPACE 0x15 /* Server has no room for new objects */
69 #define SD_RES_WAIT_FOR_FORMAT 0x16 /* Waiting for a format operation */
70 #define SD_RES_WAIT_FOR_JOIN 0x17 /* Waiting for other nodes joining */
71 #define SD_RES_JOIN_FAILED 0x18 /* Target node had failed to join sheepdog */
72 #define SD_RES_HALT 0x19 /* Sheepdog is stopped serving IO request */
73 #define SD_RES_READONLY 0x1A /* Object is read-only */
78 * 0 - 19 (20 bits): data object space
79 * 20 - 31 (12 bits): reserved data object space
80 * 32 - 55 (24 bits): vdi object space
81 * 56 - 59 ( 4 bits): reserved vdi object space
82 * 60 - 63 ( 4 bits): object type identifier space
85 #define VDI_SPACE_SHIFT 32
86 #define VDI_BIT (UINT64_C(1) << 63)
87 #define VMSTATE_BIT (UINT64_C(1) << 62)
88 #define MAX_DATA_OBJS (UINT64_C(1) << 20)
89 #define MAX_CHILDREN 1024
90 #define SD_MAX_VDI_LEN 256
91 #define SD_MAX_VDI_TAG_LEN 256
92 #define SD_NR_VDIS (1U << 24)
93 #define SD_DATA_OBJ_SIZE (UINT64_C(1) << 22)
94 #define SD_MAX_VDI_SIZE (SD_DATA_OBJ_SIZE * MAX_DATA_OBJS)
95 #define SD_DEFAULT_BLOCK_SIZE_SHIFT 22
97 * For erasure coding, we use at most SD_EC_MAX_STRIP for data strips and
98 * (SD_EC_MAX_STRIP - 1) for parity strips
100 * SD_MAX_COPIES is sum of number of data strips and parity strips.
102 #define SD_EC_MAX_STRIP 16
103 #define SD_MAX_COPIES (SD_EC_MAX_STRIP * 2 - 1)
105 #define SD_INODE_SIZE (sizeof(SheepdogInode))
106 #define CURRENT_VDI_ID 0
108 #define LOCK_TYPE_NORMAL 0
109 #define LOCK_TYPE_SHARED 1 /* for iSCSI multipath */
111 typedef struct SheepdogReq {
117 uint32_t data_length;
118 uint32_t opcode_specific[8];
121 typedef struct SheepdogRsp {
127 uint32_t data_length;
129 uint32_t opcode_specific[7];
132 typedef struct SheepdogObjReq {
138 uint32_t data_length;
147 typedef struct SheepdogObjRsp {
153 uint32_t data_length;
161 typedef struct SheepdogVdiReq {
167 uint32_t data_length;
169 uint32_t base_vdi_id;
172 uint8_t store_policy;
173 uint8_t block_size_shift;
179 typedef struct SheepdogVdiRsp {
185 uint32_t data_length;
192 typedef struct SheepdogClusterRsp {
198 uint32_t data_length;
202 uint8_t block_size_shift;
205 } SheepdogClusterRsp;
207 typedef struct SheepdogInode {
208 char name[SD_MAX_VDI_LEN];
209 char tag[SD_MAX_VDI_TAG_LEN];
212 uint64_t vm_clock_nsec;
214 uint64_t vm_state_size;
215 uint16_t copy_policy;
217 uint8_t block_size_shift;
220 uint32_t parent_vdi_id;
221 uint32_t child_vdi_id[MAX_CHILDREN];
222 uint32_t data_vdi_id[MAX_DATA_OBJS];
225 #define SD_INODE_HEADER_SIZE offsetof(SheepdogInode, data_vdi_id)
228 * 64 bit FNV-1a non-zero initial basis
230 #define FNV1A_64_INIT ((uint64_t)0xcbf29ce484222325ULL)
233 * 64 bit Fowler/Noll/Vo FNV-1a hash code
235 static inline uint64_t fnv_64a_buf(void *buf, size_t len, uint64_t hval)
237 unsigned char *bp = buf;
238 unsigned char *be = bp + len;
240 hval ^= (uint64_t) *bp++;
241 hval += (hval << 1) + (hval << 4) + (hval << 5) +
242 (hval << 7) + (hval << 8) + (hval << 40);
247 static inline bool is_data_obj_writable(SheepdogInode *inode, unsigned int idx)
249 return inode->vdi_id == inode->data_vdi_id[idx];
252 static inline bool is_data_obj(uint64_t oid)
254 return !(VDI_BIT & oid);
257 static inline uint64_t data_oid_to_idx(uint64_t oid)
259 return oid & (MAX_DATA_OBJS - 1);
262 static inline uint32_t oid_to_vid(uint64_t oid)
264 return (oid & ~VDI_BIT) >> VDI_SPACE_SHIFT;
267 static inline uint64_t vid_to_vdi_oid(uint32_t vid)
269 return VDI_BIT | ((uint64_t)vid << VDI_SPACE_SHIFT);
272 static inline uint64_t vid_to_vmstate_oid(uint32_t vid, uint32_t idx)
274 return VMSTATE_BIT | ((uint64_t)vid << VDI_SPACE_SHIFT) | idx;
277 static inline uint64_t vid_to_data_oid(uint32_t vid, uint32_t idx)
279 return ((uint64_t)vid << VDI_SPACE_SHIFT) | idx;
282 static inline bool is_snapshot(struct SheepdogInode *inode)
284 return !!inode->snap_ctime;
289 #define DPRINTF(fmt, args...) \
291 fprintf(stdout, "%s %d: " fmt, __func__, __LINE__, ##args); \
294 #define DPRINTF(fmt, args...)
297 typedef struct SheepdogAIOCB SheepdogAIOCB;
299 typedef struct AIOReq {
300 SheepdogAIOCB *aiocb;
301 unsigned int iov_offset;
306 unsigned int data_len;
311 QLIST_ENTRY(AIOReq) aio_siblings;
321 #define AIOCBOverwrapping(x, y) \
322 (!(x->max_affect_data_idx < y->min_affect_data_idx \
323 || y->max_affect_data_idx < x->min_affect_data_idx))
325 struct SheepdogAIOCB {
334 enum AIOCBState aiocb_type;
336 Coroutine *coroutine;
337 void (*aio_done_func)(SheepdogAIOCB *);
342 uint32_t min_affect_data_idx;
343 uint32_t max_affect_data_idx;
345 QLIST_ENTRY(SheepdogAIOCB) aiocb_siblings;
348 typedef struct BDRVSheepdogState {
349 BlockDriverState *bs;
350 AioContext *aio_context;
354 uint32_t min_dirty_data_idx;
355 uint32_t max_dirty_data_idx;
357 char name[SD_MAX_VDI_LEN];
359 uint32_t cache_flags;
360 bool discard_supported;
370 uint32_t aioreq_seq_num;
372 /* Every aio request must be linked to either of these queues. */
373 QLIST_HEAD(inflight_aio_head, AIOReq) inflight_aio_head;
374 QLIST_HEAD(failed_aio_head, AIOReq) failed_aio_head;
376 CoQueue overwrapping_queue;
377 QLIST_HEAD(inflight_aiocb_head, SheepdogAIOCB) inflight_aiocb_head;
380 static const char * sd_strerror(int err)
384 static const struct {
388 {SD_RES_SUCCESS, "Success"},
389 {SD_RES_UNKNOWN, "Unknown error"},
390 {SD_RES_NO_OBJ, "No object found"},
391 {SD_RES_EIO, "I/O error"},
392 {SD_RES_VDI_EXIST, "VDI exists already"},
393 {SD_RES_INVALID_PARMS, "Invalid parameters"},
394 {SD_RES_SYSTEM_ERROR, "System error"},
395 {SD_RES_VDI_LOCKED, "VDI is already locked"},
396 {SD_RES_NO_VDI, "No vdi found"},
397 {SD_RES_NO_BASE_VDI, "No base VDI found"},
398 {SD_RES_VDI_READ, "Failed read the requested VDI"},
399 {SD_RES_VDI_WRITE, "Failed to write the requested VDI"},
400 {SD_RES_BASE_VDI_READ, "Failed to read the base VDI"},
401 {SD_RES_BASE_VDI_WRITE, "Failed to write the base VDI"},
402 {SD_RES_NO_TAG, "Failed to find the requested tag"},
403 {SD_RES_STARTUP, "The system is still booting"},
404 {SD_RES_VDI_NOT_LOCKED, "VDI isn't locked"},
405 {SD_RES_SHUTDOWN, "The system is shutting down"},
406 {SD_RES_NO_MEM, "Out of memory on the server"},
407 {SD_RES_FULL_VDI, "We already have the maximum vdis"},
408 {SD_RES_VER_MISMATCH, "Protocol version mismatch"},
409 {SD_RES_NO_SPACE, "Server has no space for new objects"},
410 {SD_RES_WAIT_FOR_FORMAT, "Sheepdog is waiting for a format operation"},
411 {SD_RES_WAIT_FOR_JOIN, "Sheepdog is waiting for other nodes joining"},
412 {SD_RES_JOIN_FAILED, "Target node had failed to join sheepdog"},
413 {SD_RES_HALT, "Sheepdog is stopped serving IO request"},
414 {SD_RES_READONLY, "Object is read-only"},
417 for (i = 0; i < ARRAY_SIZE(errors); ++i) {
418 if (errors[i].err == err) {
419 return errors[i].desc;
423 return "Invalid error code";
427 * Sheepdog I/O handling:
429 * 1. In sd_co_rw_vector, we send the I/O requests to the server and
430 * link the requests to the inflight_list in the
431 * BDRVSheepdogState. The function exits without waiting for
432 * receiving the response.
434 * 2. We receive the response in aio_read_response, the fd handler to
435 * the sheepdog connection. If metadata update is needed, we send
436 * the write request to the vdi object in sd_write_done, the write
437 * completion function. We switch back to sd_co_readv/writev after
438 * all the requests belonging to the AIOCB are finished.
441 static inline AIOReq *alloc_aio_req(BDRVSheepdogState *s, SheepdogAIOCB *acb,
442 uint64_t oid, unsigned int data_len,
443 uint64_t offset, uint8_t flags, bool create,
444 uint64_t base_oid, unsigned int iov_offset)
448 aio_req = g_malloc(sizeof(*aio_req));
449 aio_req->aiocb = acb;
450 aio_req->iov_offset = iov_offset;
452 aio_req->base_oid = base_oid;
453 aio_req->offset = offset;
454 aio_req->data_len = data_len;
455 aio_req->flags = flags;
456 aio_req->id = s->aioreq_seq_num++;
457 aio_req->create = create;
463 static inline void free_aio_req(BDRVSheepdogState *s, AIOReq *aio_req)
465 SheepdogAIOCB *acb = aio_req->aiocb;
467 acb->cancelable = false;
468 QLIST_REMOVE(aio_req, aio_siblings);
474 static void coroutine_fn sd_finish_aiocb(SheepdogAIOCB *acb)
476 qemu_coroutine_enter(acb->coroutine, NULL);
481 * Check whether the specified acb can be canceled
483 * We can cancel aio when any request belonging to the acb is:
484 * - Not processed by the sheepdog server.
485 * - Not linked to the inflight queue.
487 static bool sd_acb_cancelable(const SheepdogAIOCB *acb)
489 BDRVSheepdogState *s = acb->common.bs->opaque;
492 if (!acb->cancelable) {
496 QLIST_FOREACH(aioreq, &s->inflight_aio_head, aio_siblings) {
497 if (aioreq->aiocb == acb) {
505 static void sd_aio_cancel(BlockAIOCB *blockacb)
507 SheepdogAIOCB *acb = (SheepdogAIOCB *)blockacb;
508 BDRVSheepdogState *s = acb->common.bs->opaque;
509 AIOReq *aioreq, *next;
511 if (sd_acb_cancelable(acb)) {
512 /* Remove outstanding requests from failed queue. */
513 QLIST_FOREACH_SAFE(aioreq, &s->failed_aio_head, aio_siblings,
515 if (aioreq->aiocb == acb) {
516 free_aio_req(s, aioreq);
520 assert(acb->nr_pending == 0);
521 if (acb->common.cb) {
522 acb->common.cb(acb->common.opaque, -ECANCELED);
524 sd_finish_aiocb(acb);
528 static const AIOCBInfo sd_aiocb_info = {
529 .aiocb_size = sizeof(SheepdogAIOCB),
530 .cancel_async = sd_aio_cancel,
533 static SheepdogAIOCB *sd_aio_setup(BlockDriverState *bs, QEMUIOVector *qiov,
534 int64_t sector_num, int nb_sectors)
537 uint32_t object_size;
538 BDRVSheepdogState *s = bs->opaque;
540 object_size = (UINT32_C(1) << s->inode.block_size_shift);
542 acb = qemu_aio_get(&sd_aiocb_info, bs, NULL, NULL);
546 acb->sector_num = sector_num;
547 acb->nb_sectors = nb_sectors;
549 acb->aio_done_func = NULL;
550 acb->cancelable = true;
551 acb->coroutine = qemu_coroutine_self();
555 acb->min_affect_data_idx = acb->sector_num * BDRV_SECTOR_SIZE / object_size;
556 acb->max_affect_data_idx = (acb->sector_num * BDRV_SECTOR_SIZE +
557 acb->nb_sectors * BDRV_SECTOR_SIZE) / object_size;
562 /* Return -EIO in case of error, file descriptor on success */
563 static int connect_to_sdog(BDRVSheepdogState *s, Error **errp)
568 fd = unix_connect(s->host_spec, errp);
570 fd = inet_connect(s->host_spec, errp);
573 int ret = socket_set_nodelay(fd);
575 error_report("%s", strerror(errno));
581 qemu_set_nonblock(fd);
589 /* Return 0 on success and -errno in case of error */
590 static coroutine_fn int send_co_req(int sockfd, SheepdogReq *hdr, void *data,
595 ret = qemu_co_send(sockfd, hdr, sizeof(*hdr));
596 if (ret != sizeof(*hdr)) {
597 error_report("failed to send a req, %s", strerror(errno));
598 ret = -socket_error();
602 ret = qemu_co_send(sockfd, data, *wlen);
604 ret = -socket_error();
605 error_report("failed to send a req, %s", strerror(errno));
611 static void restart_co_req(void *opaque)
613 Coroutine *co = opaque;
615 qemu_coroutine_enter(co, NULL);
618 typedef struct SheepdogReqCo {
620 AioContext *aio_context;
629 static coroutine_fn void do_co_req(void *opaque)
633 SheepdogReqCo *srco = opaque;
634 int sockfd = srco->sockfd;
635 SheepdogReq *hdr = srco->hdr;
636 void *data = srco->data;
637 unsigned int *wlen = srco->wlen;
638 unsigned int *rlen = srco->rlen;
640 co = qemu_coroutine_self();
641 aio_set_fd_handler(srco->aio_context, sockfd, NULL, restart_co_req, co);
643 ret = send_co_req(sockfd, hdr, data, wlen);
648 aio_set_fd_handler(srco->aio_context, sockfd, restart_co_req, NULL, co);
650 ret = qemu_co_recv(sockfd, hdr, sizeof(*hdr));
651 if (ret != sizeof(*hdr)) {
652 error_report("failed to get a rsp, %s", strerror(errno));
657 if (*rlen > hdr->data_length) {
658 *rlen = hdr->data_length;
662 ret = qemu_co_recv(sockfd, data, *rlen);
664 error_report("failed to get the data, %s", strerror(errno));
671 /* there is at most one request for this sockfd, so it is safe to
672 * set each handler to NULL. */
673 aio_set_fd_handler(srco->aio_context, sockfd, NULL, NULL, NULL);
676 srco->finished = true;
680 * Send the request to the sheep in a synchronous manner.
682 * Return 0 on success, -errno in case of error.
684 static int do_req(int sockfd, AioContext *aio_context, SheepdogReq *hdr,
685 void *data, unsigned int *wlen, unsigned int *rlen)
688 SheepdogReqCo srco = {
690 .aio_context = aio_context,
699 if (qemu_in_coroutine()) {
702 co = qemu_coroutine_create(do_co_req);
703 qemu_coroutine_enter(co, &srco);
704 while (!srco.finished) {
705 aio_poll(aio_context, true);
712 static void coroutine_fn add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req,
713 struct iovec *iov, int niov,
714 enum AIOCBState aiocb_type);
715 static void coroutine_fn resend_aioreq(BDRVSheepdogState *s, AIOReq *aio_req);
716 static int reload_inode(BDRVSheepdogState *s, uint32_t snapid, const char *tag);
717 static int get_sheep_fd(BDRVSheepdogState *s, Error **errp);
718 static void co_write_request(void *opaque);
720 static coroutine_fn void reconnect_to_sdog(void *opaque)
722 BDRVSheepdogState *s = opaque;
723 AIOReq *aio_req, *next;
725 aio_set_fd_handler(s->aio_context, s->fd, NULL, NULL, NULL);
729 /* Wait for outstanding write requests to be completed. */
730 while (s->co_send != NULL) {
731 co_write_request(opaque);
734 /* Try to reconnect the sheepdog server every one second. */
736 Error *local_err = NULL;
737 s->fd = get_sheep_fd(s, &local_err);
739 DPRINTF("Wait for connection to be established\n");
740 error_report_err(local_err);
741 co_aio_sleep_ns(bdrv_get_aio_context(s->bs), QEMU_CLOCK_REALTIME,
747 * Now we have to resend all the request in the inflight queue. However,
748 * resend_aioreq() can yield and newly created requests can be added to the
749 * inflight queue before the coroutine is resumed. To avoid mixing them, we
750 * have to move all the inflight requests to the failed queue before
751 * resend_aioreq() is called.
753 QLIST_FOREACH_SAFE(aio_req, &s->inflight_aio_head, aio_siblings, next) {
754 QLIST_REMOVE(aio_req, aio_siblings);
755 QLIST_INSERT_HEAD(&s->failed_aio_head, aio_req, aio_siblings);
758 /* Resend all the failed aio requests. */
759 while (!QLIST_EMPTY(&s->failed_aio_head)) {
760 aio_req = QLIST_FIRST(&s->failed_aio_head);
761 QLIST_REMOVE(aio_req, aio_siblings);
762 QLIST_INSERT_HEAD(&s->inflight_aio_head, aio_req, aio_siblings);
763 resend_aioreq(s, aio_req);
768 * Receive responses of the I/O requests.
770 * This function is registered as a fd handler, and called from the
771 * main loop when s->fd is ready for reading responses.
773 static void coroutine_fn aio_read_response(void *opaque)
776 BDRVSheepdogState *s = opaque;
779 AIOReq *aio_req = NULL;
784 ret = qemu_co_recv(fd, &rsp, sizeof(rsp));
785 if (ret != sizeof(rsp)) {
786 error_report("failed to get the header, %s", strerror(errno));
790 /* find the right aio_req from the inflight aio list */
791 QLIST_FOREACH(aio_req, &s->inflight_aio_head, aio_siblings) {
792 if (aio_req->id == rsp.id) {
797 error_report("cannot find aio_req %x", rsp.id);
801 acb = aio_req->aiocb;
803 switch (acb->aiocb_type) {
804 case AIOCB_WRITE_UDATA:
805 /* this coroutine context is no longer suitable for co_recv
806 * because we may send data to update vdi objects */
808 if (!is_data_obj(aio_req->oid)) {
811 idx = data_oid_to_idx(aio_req->oid);
813 if (aio_req->create) {
815 * If the object is newly created one, we need to update
816 * the vdi object (metadata object). min_dirty_data_idx
817 * and max_dirty_data_idx are changed to include updated
818 * index between them.
820 if (rsp.result == SD_RES_SUCCESS) {
821 s->inode.data_vdi_id[idx] = s->inode.vdi_id;
822 s->max_dirty_data_idx = MAX(idx, s->max_dirty_data_idx);
823 s->min_dirty_data_idx = MIN(idx, s->min_dirty_data_idx);
827 case AIOCB_READ_UDATA:
828 ret = qemu_co_recvv(fd, acb->qiov->iov, acb->qiov->niov,
829 aio_req->iov_offset, rsp.data_length);
830 if (ret != rsp.data_length) {
831 error_report("failed to get the data, %s", strerror(errno));
835 case AIOCB_FLUSH_CACHE:
836 if (rsp.result == SD_RES_INVALID_PARMS) {
837 DPRINTF("disable cache since the server doesn't support it\n");
838 s->cache_flags = SD_FLAG_CMD_DIRECT;
839 rsp.result = SD_RES_SUCCESS;
842 case AIOCB_DISCARD_OBJ:
843 switch (rsp.result) {
844 case SD_RES_INVALID_PARMS:
845 error_report("sheep(%s) doesn't support discard command",
847 rsp.result = SD_RES_SUCCESS;
848 s->discard_supported = false;
851 idx = data_oid_to_idx(aio_req->oid);
852 s->inode.data_vdi_id[idx] = 0;
859 switch (rsp.result) {
862 case SD_RES_READONLY:
863 if (s->inode.vdi_id == oid_to_vid(aio_req->oid)) {
864 ret = reload_inode(s, 0, "");
869 if (is_data_obj(aio_req->oid)) {
870 aio_req->oid = vid_to_data_oid(s->inode.vdi_id,
871 data_oid_to_idx(aio_req->oid));
873 aio_req->oid = vid_to_vdi_oid(s->inode.vdi_id);
875 resend_aioreq(s, aio_req);
879 error_report("%s", sd_strerror(rsp.result));
883 free_aio_req(s, aio_req);
884 if (!acb->nr_pending) {
886 * We've finished all requests which belong to the AIOCB, so
887 * we can switch back to sd_co_readv/writev now.
889 acb->aio_done_func(acb);
896 reconnect_to_sdog(opaque);
899 static void co_read_response(void *opaque)
901 BDRVSheepdogState *s = opaque;
904 s->co_recv = qemu_coroutine_create(aio_read_response);
907 qemu_coroutine_enter(s->co_recv, opaque);
910 static void co_write_request(void *opaque)
912 BDRVSheepdogState *s = opaque;
914 qemu_coroutine_enter(s->co_send, NULL);
918 * Return a socket descriptor to read/write objects.
920 * We cannot use this descriptor for other operations because
921 * the block driver may be on waiting response from the server.
923 static int get_sheep_fd(BDRVSheepdogState *s, Error **errp)
927 fd = connect_to_sdog(s, errp);
932 aio_set_fd_handler(s->aio_context, fd, co_read_response, NULL, s);
936 static int sd_parse_uri(BDRVSheepdogState *s, const char *filename,
937 char *vdi, uint32_t *snapid, char *tag)
940 QueryParams *qp = NULL;
943 uri = uri_parse(filename);
949 if (!strcmp(uri->scheme, "sheepdog")) {
951 } else if (!strcmp(uri->scheme, "sheepdog+tcp")) {
953 } else if (!strcmp(uri->scheme, "sheepdog+unix")) {
960 if (uri->path == NULL || !strcmp(uri->path, "/")) {
964 pstrcpy(vdi, SD_MAX_VDI_LEN, uri->path + 1);
966 qp = query_params_parse(uri->query);
967 if (qp->n > 1 || (s->is_unix && !qp->n) || (!s->is_unix && qp->n)) {
973 /* sheepdog+unix:///vdiname?socket=path */
974 if (uri->server || uri->port || strcmp(qp->p[0].name, "socket")) {
978 s->host_spec = g_strdup(qp->p[0].value);
980 /* sheepdog[+tcp]://[host:port]/vdiname */
981 s->host_spec = g_strdup_printf("%s:%d", uri->server ?: SD_DEFAULT_ADDR,
982 uri->port ?: SD_DEFAULT_PORT);
987 *snapid = strtoul(uri->fragment, NULL, 10);
989 pstrcpy(tag, SD_MAX_VDI_TAG_LEN, uri->fragment);
992 *snapid = CURRENT_VDI_ID; /* search current vdi */
997 query_params_free(qp);
1004 * Parse a filename (old syntax)
1006 * filename must be one of the following formats:
1008 * 2. [vdiname]:[snapid]
1009 * 3. [vdiname]:[tag]
1010 * 4. [hostname]:[port]:[vdiname]
1011 * 5. [hostname]:[port]:[vdiname]:[snapid]
1012 * 6. [hostname]:[port]:[vdiname]:[tag]
1014 * You can boot from the snapshot images by specifying `snapid` or
1017 * You can run VMs outside the Sheepdog cluster by specifying
1018 * `hostname' and `port' (experimental).
1020 static int parse_vdiname(BDRVSheepdogState *s, const char *filename,
1021 char *vdi, uint32_t *snapid, char *tag)
1024 const char *host_spec, *vdi_spec;
1027 strstart(filename, "sheepdog:", (const char **)&filename);
1028 p = q = g_strdup(filename);
1030 /* count the number of separators */
1040 /* use the first two tokens as host_spec. */
1053 p = strchr(vdi_spec, ':');
1058 uri = g_strdup_printf("sheepdog://%s/%s", host_spec, vdi_spec);
1060 ret = sd_parse_uri(s, uri, vdi, snapid, tag);
1068 static int find_vdi_name(BDRVSheepdogState *s, const char *filename,
1069 uint32_t snapid, const char *tag, uint32_t *vid,
1070 bool lock, Error **errp)
1074 SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;
1075 unsigned int wlen, rlen = 0;
1076 char buf[SD_MAX_VDI_LEN + SD_MAX_VDI_TAG_LEN];
1078 fd = connect_to_sdog(s, errp);
1083 /* This pair of strncpy calls ensures that the buffer is zero-filled,
1084 * which is desirable since we'll soon be sending those bytes, and
1085 * don't want the send_req to read uninitialized data.
1087 strncpy(buf, filename, SD_MAX_VDI_LEN);
1088 strncpy(buf + SD_MAX_VDI_LEN, tag, SD_MAX_VDI_TAG_LEN);
1090 memset(&hdr, 0, sizeof(hdr));
1092 hdr.opcode = SD_OP_LOCK_VDI;
1093 hdr.type = LOCK_TYPE_NORMAL;
1095 hdr.opcode = SD_OP_GET_VDI_INFO;
1097 wlen = SD_MAX_VDI_LEN + SD_MAX_VDI_TAG_LEN;
1098 hdr.proto_ver = SD_PROTO_VER;
1099 hdr.data_length = wlen;
1100 hdr.snapid = snapid;
1101 hdr.flags = SD_FLAG_CMD_WRITE;
1103 ret = do_req(fd, s->aio_context, (SheepdogReq *)&hdr, buf, &wlen, &rlen);
1105 error_setg_errno(errp, -ret, "cannot get vdi info");
1109 if (rsp->result != SD_RES_SUCCESS) {
1110 error_setg(errp, "cannot get vdi info, %s, %s %" PRIu32 " %s",
1111 sd_strerror(rsp->result), filename, snapid, tag);
1112 if (rsp->result == SD_RES_NO_VDI) {
1114 } else if (rsp->result == SD_RES_VDI_LOCKED) {
1129 static void coroutine_fn add_aio_request(BDRVSheepdogState *s, AIOReq *aio_req,
1130 struct iovec *iov, int niov,
1131 enum AIOCBState aiocb_type)
1133 int nr_copies = s->inode.nr_copies;
1135 unsigned int wlen = 0;
1137 uint64_t oid = aio_req->oid;
1138 unsigned int datalen = aio_req->data_len;
1139 uint64_t offset = aio_req->offset;
1140 uint8_t flags = aio_req->flags;
1141 uint64_t old_oid = aio_req->base_oid;
1142 bool create = aio_req->create;
1145 error_report("bug");
1148 memset(&hdr, 0, sizeof(hdr));
1150 switch (aiocb_type) {
1151 case AIOCB_FLUSH_CACHE:
1152 hdr.opcode = SD_OP_FLUSH_VDI;
1154 case AIOCB_READ_UDATA:
1155 hdr.opcode = SD_OP_READ_OBJ;
1158 case AIOCB_WRITE_UDATA:
1160 hdr.opcode = SD_OP_CREATE_AND_WRITE_OBJ;
1162 hdr.opcode = SD_OP_WRITE_OBJ;
1165 hdr.flags = SD_FLAG_CMD_WRITE | flags;
1167 case AIOCB_DISCARD_OBJ:
1168 hdr.opcode = SD_OP_DISCARD_OBJ;
1172 if (s->cache_flags) {
1173 hdr.flags |= s->cache_flags;
1177 hdr.cow_oid = old_oid;
1178 hdr.copies = s->inode.nr_copies;
1180 hdr.data_length = datalen;
1181 hdr.offset = offset;
1183 hdr.id = aio_req->id;
1185 qemu_co_mutex_lock(&s->lock);
1186 s->co_send = qemu_coroutine_self();
1187 aio_set_fd_handler(s->aio_context, s->fd,
1188 co_read_response, co_write_request, s);
1189 socket_set_cork(s->fd, 1);
1192 ret = qemu_co_send(s->fd, &hdr, sizeof(hdr));
1193 if (ret != sizeof(hdr)) {
1194 error_report("failed to send a req, %s", strerror(errno));
1199 ret = qemu_co_sendv(s->fd, iov, niov, aio_req->iov_offset, wlen);
1201 error_report("failed to send a data, %s", strerror(errno));
1205 socket_set_cork(s->fd, 0);
1206 aio_set_fd_handler(s->aio_context, s->fd, co_read_response, NULL, s);
1208 qemu_co_mutex_unlock(&s->lock);
1211 static int read_write_object(int fd, AioContext *aio_context, char *buf,
1212 uint64_t oid, uint8_t copies,
1213 unsigned int datalen, uint64_t offset,
1214 bool write, bool create, uint32_t cache_flags)
1217 SheepdogObjRsp *rsp = (SheepdogObjRsp *)&hdr;
1218 unsigned int wlen, rlen;
1221 memset(&hdr, 0, sizeof(hdr));
1226 hdr.flags = SD_FLAG_CMD_WRITE;
1228 hdr.opcode = SD_OP_CREATE_AND_WRITE_OBJ;
1230 hdr.opcode = SD_OP_WRITE_OBJ;
1235 hdr.opcode = SD_OP_READ_OBJ;
1238 hdr.flags |= cache_flags;
1241 hdr.data_length = datalen;
1242 hdr.offset = offset;
1243 hdr.copies = copies;
1245 ret = do_req(fd, aio_context, (SheepdogReq *)&hdr, buf, &wlen, &rlen);
1247 error_report("failed to send a request to the sheep");
1251 switch (rsp->result) {
1252 case SD_RES_SUCCESS:
1255 error_report("%s", sd_strerror(rsp->result));
1260 static int read_object(int fd, AioContext *aio_context, char *buf,
1261 uint64_t oid, uint8_t copies,
1262 unsigned int datalen, uint64_t offset,
1263 uint32_t cache_flags)
1265 return read_write_object(fd, aio_context, buf, oid, copies,
1266 datalen, offset, false,
1267 false, cache_flags);
1270 static int write_object(int fd, AioContext *aio_context, char *buf,
1271 uint64_t oid, uint8_t copies,
1272 unsigned int datalen, uint64_t offset, bool create,
1273 uint32_t cache_flags)
1275 return read_write_object(fd, aio_context, buf, oid, copies,
1276 datalen, offset, true,
1277 create, cache_flags);
1280 /* update inode with the latest state */
1281 static int reload_inode(BDRVSheepdogState *s, uint32_t snapid, const char *tag)
1283 Error *local_err = NULL;
1284 SheepdogInode *inode;
1288 fd = connect_to_sdog(s, &local_err);
1290 error_report_err(local_err);
1294 inode = g_malloc(SD_INODE_HEADER_SIZE);
1296 ret = find_vdi_name(s, s->name, snapid, tag, &vid, false, &local_err);
1298 error_report_err(local_err);
1302 ret = read_object(fd, s->aio_context, (char *)inode, vid_to_vdi_oid(vid),
1303 s->inode.nr_copies, SD_INODE_HEADER_SIZE, 0,
1309 if (inode->vdi_id != s->inode.vdi_id) {
1310 memcpy(&s->inode, inode, SD_INODE_HEADER_SIZE);
1320 static void coroutine_fn resend_aioreq(BDRVSheepdogState *s, AIOReq *aio_req)
1322 SheepdogAIOCB *acb = aio_req->aiocb;
1324 aio_req->create = false;
1326 /* check whether this request becomes a CoW one */
1327 if (acb->aiocb_type == AIOCB_WRITE_UDATA && is_data_obj(aio_req->oid)) {
1328 int idx = data_oid_to_idx(aio_req->oid);
1330 if (is_data_obj_writable(&s->inode, idx)) {
1334 if (s->inode.data_vdi_id[idx]) {
1335 aio_req->base_oid = vid_to_data_oid(s->inode.data_vdi_id[idx], idx);
1336 aio_req->flags |= SD_FLAG_CMD_COW;
1338 aio_req->create = true;
1341 if (is_data_obj(aio_req->oid)) {
1342 add_aio_request(s, aio_req, acb->qiov->iov, acb->qiov->niov,
1346 iov.iov_base = &s->inode;
1347 iov.iov_len = sizeof(s->inode);
1348 add_aio_request(s, aio_req, &iov, 1, AIOCB_WRITE_UDATA);
1352 static void sd_detach_aio_context(BlockDriverState *bs)
1354 BDRVSheepdogState *s = bs->opaque;
1356 aio_set_fd_handler(s->aio_context, s->fd, NULL, NULL, NULL);
1359 static void sd_attach_aio_context(BlockDriverState *bs,
1360 AioContext *new_context)
1362 BDRVSheepdogState *s = bs->opaque;
1364 s->aio_context = new_context;
1365 aio_set_fd_handler(new_context, s->fd, co_read_response, NULL, s);
1368 /* TODO Convert to fine grained options */
1369 static QemuOptsList runtime_opts = {
1371 .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head),
1375 .type = QEMU_OPT_STRING,
1376 .help = "URL to the sheepdog image",
1378 { /* end of list */ }
1382 static int sd_open(BlockDriverState *bs, QDict *options, int flags,
1387 BDRVSheepdogState *s = bs->opaque;
1388 char vdi[SD_MAX_VDI_LEN], tag[SD_MAX_VDI_TAG_LEN];
1392 Error *local_err = NULL;
1393 const char *filename;
1396 s->aio_context = bdrv_get_aio_context(bs);
1398 opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort);
1399 qemu_opts_absorb_qdict(opts, options, &local_err);
1401 error_propagate(errp, local_err);
1406 filename = qemu_opt_get(opts, "filename");
1408 QLIST_INIT(&s->inflight_aio_head);
1409 QLIST_INIT(&s->failed_aio_head);
1410 QLIST_INIT(&s->inflight_aiocb_head);
1413 memset(vdi, 0, sizeof(vdi));
1414 memset(tag, 0, sizeof(tag));
1416 if (strstr(filename, "://")) {
1417 ret = sd_parse_uri(s, filename, vdi, &snapid, tag);
1419 ret = parse_vdiname(s, filename, vdi, &snapid, tag);
1422 error_setg(errp, "Can't parse filename");
1425 s->fd = get_sheep_fd(s, errp);
1431 ret = find_vdi_name(s, vdi, snapid, tag, &vid, true, errp);
1437 * QEMU block layer emulates writethrough cache as 'writeback + flush', so
1438 * we always set SD_FLAG_CMD_CACHE (writeback cache) as default.
1440 s->cache_flags = SD_FLAG_CMD_CACHE;
1441 if (flags & BDRV_O_NOCACHE) {
1442 s->cache_flags = SD_FLAG_CMD_DIRECT;
1444 s->discard_supported = true;
1446 if (snapid || tag[0] != '\0') {
1447 DPRINTF("%" PRIx32 " snapshot inode was open.\n", vid);
1448 s->is_snapshot = true;
1451 fd = connect_to_sdog(s, errp);
1457 buf = g_malloc(SD_INODE_SIZE);
1458 ret = read_object(fd, s->aio_context, buf, vid_to_vdi_oid(vid),
1459 0, SD_INODE_SIZE, 0, s->cache_flags);
1464 error_setg(errp, "Can't read snapshot inode");
1468 memcpy(&s->inode, buf, sizeof(s->inode));
1469 s->min_dirty_data_idx = UINT32_MAX;
1470 s->max_dirty_data_idx = 0;
1472 bs->total_sectors = s->inode.vdi_size / BDRV_SECTOR_SIZE;
1473 pstrcpy(s->name, sizeof(s->name), vdi);
1474 qemu_co_mutex_init(&s->lock);
1475 qemu_co_queue_init(&s->overwrapping_queue);
1476 qemu_opts_del(opts);
1480 aio_set_fd_handler(bdrv_get_aio_context(bs), s->fd, NULL, NULL, NULL);
1484 qemu_opts_del(opts);
1489 static int do_sd_create(BDRVSheepdogState *s, uint32_t *vdi_id, int snapshot,
1493 SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;
1495 unsigned int wlen, rlen = 0;
1496 char buf[SD_MAX_VDI_LEN];
1498 fd = connect_to_sdog(s, errp);
1503 /* FIXME: would it be better to fail (e.g., return -EIO) when filename
1504 * does not fit in buf? For now, just truncate and avoid buffer overrun.
1506 memset(buf, 0, sizeof(buf));
1507 pstrcpy(buf, sizeof(buf), s->name);
1509 memset(&hdr, 0, sizeof(hdr));
1510 hdr.opcode = SD_OP_NEW_VDI;
1511 hdr.base_vdi_id = s->inode.vdi_id;
1513 wlen = SD_MAX_VDI_LEN;
1515 hdr.flags = SD_FLAG_CMD_WRITE;
1516 hdr.snapid = snapshot;
1518 hdr.data_length = wlen;
1519 hdr.vdi_size = s->inode.vdi_size;
1520 hdr.copy_policy = s->inode.copy_policy;
1521 hdr.copies = s->inode.nr_copies;
1522 hdr.block_size_shift = s->inode.block_size_shift;
1524 ret = do_req(fd, s->aio_context, (SheepdogReq *)&hdr, buf, &wlen, &rlen);
1529 error_setg_errno(errp, -ret, "create failed");
1533 if (rsp->result != SD_RES_SUCCESS) {
1534 error_setg(errp, "%s, %s", sd_strerror(rsp->result), s->inode.name);
1539 *vdi_id = rsp->vdi_id;
1545 static int sd_prealloc(const char *filename, Error **errp)
1547 BlockDriverState *bs = NULL;
1548 BDRVSheepdogState *base = NULL;
1549 unsigned long buf_size;
1550 uint32_t idx, max_idx;
1551 uint32_t object_size;
1556 ret = bdrv_open(&bs, filename, NULL, NULL, BDRV_O_RDWR | BDRV_O_PROTOCOL,
1559 goto out_with_err_set;
1562 vdi_size = bdrv_getlength(bs);
1569 object_size = (UINT32_C(1) << base->inode.block_size_shift);
1570 buf_size = MIN(object_size, SD_DATA_OBJ_SIZE);
1571 buf = g_malloc0(buf_size);
1573 max_idx = DIV_ROUND_UP(vdi_size, buf_size);
1575 for (idx = 0; idx < max_idx; idx++) {
1577 * The created image can be a cloned image, so we need to read
1578 * a data from the source image.
1580 ret = bdrv_pread(bs, idx * buf_size, buf, buf_size);
1584 ret = bdrv_pwrite(bs, idx * buf_size, buf, buf_size);
1592 error_setg_errno(errp, -ret, "Can't pre-allocate");
1604 * Sheepdog support two kinds of redundancy, full replication and erasure
1607 * # create a fully replicated vdi with x copies
1608 * -o redundancy=x (1 <= x <= SD_MAX_COPIES)
1610 * # create a erasure coded vdi with x data strips and y parity strips
1611 * -o redundancy=x:y (x must be one of {2,4,8,16} and 1 <= y < SD_EC_MAX_STRIP)
1613 static int parse_redundancy(BDRVSheepdogState *s, const char *opt)
1615 struct SheepdogInode *inode = &s->inode;
1616 const char *n1, *n2;
1620 pstrcpy(p, sizeof(p), opt);
1621 n1 = strtok(p, ":");
1622 n2 = strtok(NULL, ":");
1628 copy = strtol(n1, NULL, 10);
1629 if (copy > SD_MAX_COPIES || copy < 1) {
1633 inode->copy_policy = 0;
1634 inode->nr_copies = copy;
1638 if (copy != 2 && copy != 4 && copy != 8 && copy != 16) {
1642 parity = strtol(n2, NULL, 10);
1643 if (parity >= SD_EC_MAX_STRIP || parity < 1) {
1648 * 4 bits for parity and 4 bits for data.
1649 * We have to compress upper data bits because it can't represent 16
1651 inode->copy_policy = ((copy / 2) << 4) + parity;
1652 inode->nr_copies = copy + parity;
1657 static int parse_block_size_shift(BDRVSheepdogState *s, QemuOpts *opt)
1659 struct SheepdogInode *inode = &s->inode;
1660 uint64_t object_size;
1663 object_size = qemu_opt_get_size_del(opt, BLOCK_OPT_OBJECT_SIZE, 0);
1665 if ((object_size - 1) & object_size) { /* not a power of 2? */
1668 obj_order = ctz32(object_size);
1669 if (obj_order < 20 || obj_order > 31) {
1672 inode->block_size_shift = (uint8_t)obj_order;
1678 static int sd_create(const char *filename, QemuOpts *opts,
1683 char *backing_file = NULL;
1685 BDRVSheepdogState *s;
1686 char tag[SD_MAX_VDI_TAG_LEN];
1688 uint64_t max_vdi_size;
1689 bool prealloc = false;
1691 s = g_new0(BDRVSheepdogState, 1);
1693 memset(tag, 0, sizeof(tag));
1694 if (strstr(filename, "://")) {
1695 ret = sd_parse_uri(s, filename, s->name, &snapid, tag);
1697 ret = parse_vdiname(s, filename, s->name, &snapid, tag);
1700 error_setg(errp, "Can't parse filename");
1704 s->inode.vdi_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0),
1706 backing_file = qemu_opt_get_del(opts, BLOCK_OPT_BACKING_FILE);
1707 buf = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC);
1708 if (!buf || !strcmp(buf, "off")) {
1710 } else if (!strcmp(buf, "full")) {
1713 error_setg(errp, "Invalid preallocation mode: '%s'", buf);
1719 buf = qemu_opt_get_del(opts, BLOCK_OPT_REDUNDANCY);
1721 ret = parse_redundancy(s, buf);
1723 error_setg(errp, "Invalid redundancy mode: '%s'", buf);
1727 ret = parse_block_size_shift(s, opts);
1729 error_setg(errp, "Invalid object_size."
1730 " obect_size needs to be power of 2"
1731 " and be limited from 2^20 to 2^31");
1736 BlockDriverState *bs;
1737 BDRVSheepdogState *base;
1740 /* Currently, only Sheepdog backing image is supported. */
1741 drv = bdrv_find_protocol(backing_file, true, NULL);
1742 if (!drv || strcmp(drv->protocol_name, "sheepdog") != 0) {
1743 error_setg(errp, "backing_file must be a sheepdog image");
1749 ret = bdrv_open(&bs, backing_file, NULL, NULL, BDRV_O_PROTOCOL, NULL,
1757 if (!is_snapshot(&base->inode)) {
1758 error_setg(errp, "cannot clone from a non snapshot vdi");
1763 s->inode.vdi_id = base->inode.vdi_id;
1767 s->aio_context = qemu_get_aio_context();
1769 /* if block_size_shift is not specified, get cluster default value */
1770 if (s->inode.block_size_shift == 0) {
1772 SheepdogClusterRsp *rsp = (SheepdogClusterRsp *)&hdr;
1773 Error *local_err = NULL;
1775 unsigned int wlen = 0, rlen = 0;
1777 fd = connect_to_sdog(s, &local_err);
1779 error_report("%s", error_get_pretty(local_err));
1780 error_free(local_err);
1785 memset(&hdr, 0, sizeof(hdr));
1786 hdr.opcode = SD_OP_GET_CLUSTER_DEFAULT;
1787 hdr.proto_ver = SD_PROTO_VER;
1789 ret = do_req(fd, s->aio_context, (SheepdogReq *)&hdr,
1790 NULL, &wlen, &rlen);
1793 error_setg_errno(errp, -ret, "failed to get cluster default");
1796 if (rsp->result == SD_RES_SUCCESS) {
1797 s->inode.block_size_shift = rsp->block_size_shift;
1799 s->inode.block_size_shift = SD_DEFAULT_BLOCK_SIZE_SHIFT;
1803 max_vdi_size = (UINT64_C(1) << s->inode.block_size_shift) * MAX_DATA_OBJS;
1805 if (s->inode.vdi_size > max_vdi_size) {
1806 error_setg(errp, "An image is too large."
1807 " The maximum image size is %"PRIu64 "GB",
1808 max_vdi_size / 1024 / 1024 / 1024);
1813 ret = do_sd_create(s, &vid, 0, errp);
1819 ret = sd_prealloc(filename, errp);
1822 g_free(backing_file);
1828 static void sd_close(BlockDriverState *bs)
1830 Error *local_err = NULL;
1831 BDRVSheepdogState *s = bs->opaque;
1833 SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;
1834 unsigned int wlen, rlen = 0;
1837 DPRINTF("%s\n", s->name);
1839 fd = connect_to_sdog(s, &local_err);
1841 error_report_err(local_err);
1845 memset(&hdr, 0, sizeof(hdr));
1847 hdr.opcode = SD_OP_RELEASE_VDI;
1848 hdr.type = LOCK_TYPE_NORMAL;
1849 hdr.base_vdi_id = s->inode.vdi_id;
1850 wlen = strlen(s->name) + 1;
1851 hdr.data_length = wlen;
1852 hdr.flags = SD_FLAG_CMD_WRITE;
1854 ret = do_req(fd, s->aio_context, (SheepdogReq *)&hdr,
1855 s->name, &wlen, &rlen);
1859 if (!ret && rsp->result != SD_RES_SUCCESS &&
1860 rsp->result != SD_RES_VDI_NOT_LOCKED) {
1861 error_report("%s, %s", sd_strerror(rsp->result), s->name);
1864 aio_set_fd_handler(bdrv_get_aio_context(bs), s->fd, NULL, NULL, NULL);
1866 g_free(s->host_spec);
1869 static int64_t sd_getlength(BlockDriverState *bs)
1871 BDRVSheepdogState *s = bs->opaque;
1873 return s->inode.vdi_size;
1876 static int sd_truncate(BlockDriverState *bs, int64_t offset)
1878 Error *local_err = NULL;
1879 BDRVSheepdogState *s = bs->opaque;
1881 unsigned int datalen;
1882 uint64_t max_vdi_size;
1884 max_vdi_size = (UINT64_C(1) << s->inode.block_size_shift) * MAX_DATA_OBJS;
1885 if (offset < s->inode.vdi_size) {
1886 error_report("shrinking is not supported");
1888 } else if (offset > max_vdi_size) {
1889 error_report("too big image size");
1893 fd = connect_to_sdog(s, &local_err);
1895 error_report_err(local_err);
1899 /* we don't need to update entire object */
1900 datalen = SD_INODE_SIZE - sizeof(s->inode.data_vdi_id);
1901 s->inode.vdi_size = offset;
1902 ret = write_object(fd, s->aio_context, (char *)&s->inode,
1903 vid_to_vdi_oid(s->inode.vdi_id), s->inode.nr_copies,
1904 datalen, 0, false, s->cache_flags);
1908 error_report("failed to update an inode.");
1915 * This function is called after writing data objects. If we need to
1916 * update metadata, this sends a write request to the vdi object.
1917 * Otherwise, this switches back to sd_co_readv/writev.
1919 static void coroutine_fn sd_write_done(SheepdogAIOCB *acb)
1921 BDRVSheepdogState *s = acb->common.bs->opaque;
1924 uint32_t offset, data_len, mn, mx;
1926 mn = s->min_dirty_data_idx;
1927 mx = s->max_dirty_data_idx;
1929 /* we need to update the vdi object. */
1930 offset = sizeof(s->inode) - sizeof(s->inode.data_vdi_id) +
1931 mn * sizeof(s->inode.data_vdi_id[0]);
1932 data_len = (mx - mn + 1) * sizeof(s->inode.data_vdi_id[0]);
1934 s->min_dirty_data_idx = UINT32_MAX;
1935 s->max_dirty_data_idx = 0;
1937 iov.iov_base = &s->inode;
1938 iov.iov_len = sizeof(s->inode);
1939 aio_req = alloc_aio_req(s, acb, vid_to_vdi_oid(s->inode.vdi_id),
1940 data_len, offset, 0, false, 0, offset);
1941 QLIST_INSERT_HEAD(&s->inflight_aio_head, aio_req, aio_siblings);
1942 add_aio_request(s, aio_req, &iov, 1, AIOCB_WRITE_UDATA);
1944 acb->aio_done_func = sd_finish_aiocb;
1945 acb->aiocb_type = AIOCB_WRITE_UDATA;
1949 sd_finish_aiocb(acb);
1952 /* Delete current working VDI on the snapshot chain */
1953 static bool sd_delete(BDRVSheepdogState *s)
1955 Error *local_err = NULL;
1956 unsigned int wlen = SD_MAX_VDI_LEN, rlen = 0;
1957 SheepdogVdiReq hdr = {
1958 .opcode = SD_OP_DEL_VDI,
1959 .base_vdi_id = s->inode.vdi_id,
1960 .data_length = wlen,
1961 .flags = SD_FLAG_CMD_WRITE,
1963 SheepdogVdiRsp *rsp = (SheepdogVdiRsp *)&hdr;
1966 fd = connect_to_sdog(s, &local_err);
1968 error_report_err(local_err);
1972 ret = do_req(fd, s->aio_context, (SheepdogReq *)&hdr,
1973 s->name, &wlen, &rlen);
1978 switch (rsp->result) {
1980 error_report("%s was already deleted", s->name);
1982 case SD_RES_SUCCESS:
1985 error_report("%s, %s", sd_strerror(rsp->result), s->name);
1993 * Create a writable VDI from a snapshot
1995 static int sd_create_branch(BDRVSheepdogState *s)
1997 Error *local_err = NULL;
2003 DPRINTF("%" PRIx32 " is snapshot.\n", s->inode.vdi_id);
2005 buf = g_malloc(SD_INODE_SIZE);
2008 * Even If deletion fails, we will just create extra snapshot based on
2009 * the working VDI which was supposed to be deleted. So no need to
2012 deleted = sd_delete(s);
2013 ret = do_sd_create(s, &vid, !deleted, &local_err);
2015 error_report_err(local_err);
2019 DPRINTF("%" PRIx32 " is created.\n", vid);
2021 fd = connect_to_sdog(s, &local_err);
2023 error_report_err(local_err);
2028 ret = read_object(fd, s->aio_context, buf, vid_to_vdi_oid(vid),
2029 s->inode.nr_copies, SD_INODE_SIZE, 0, s->cache_flags);
2037 memcpy(&s->inode, buf, sizeof(s->inode));
2039 s->is_snapshot = false;
2041 DPRINTF("%" PRIx32 " was newly created.\n", s->inode.vdi_id);
2050 * Send I/O requests to the server.
2052 * This function sends requests to the server, links the requests to
2053 * the inflight_list in BDRVSheepdogState, and exits without
2054 * waiting the response. The responses are received in the
2055 * `aio_read_response' function which is called from the main loop as
2058 * Returns 1 when we need to wait a response, 0 when there is no sent
2059 * request and -errno in error cases.
2061 static int coroutine_fn sd_co_rw_vector(void *p)
2063 SheepdogAIOCB *acb = p;
2065 unsigned long len, done = 0, total = acb->nb_sectors * BDRV_SECTOR_SIZE;
2067 uint32_t object_size;
2070 BDRVSheepdogState *s = acb->common.bs->opaque;
2071 SheepdogInode *inode = &s->inode;
2074 if (acb->aiocb_type == AIOCB_WRITE_UDATA && s->is_snapshot) {
2076 * In the case we open the snapshot VDI, Sheepdog creates the
2077 * writable VDI when we do a write operation first.
2079 ret = sd_create_branch(s);
2086 object_size = (UINT32_C(1) << inode->block_size_shift);
2087 idx = acb->sector_num * BDRV_SECTOR_SIZE / object_size;
2088 offset = (acb->sector_num * BDRV_SECTOR_SIZE) % object_size;
2091 * Make sure we don't free the aiocb before we are done with all requests.
2092 * This additional reference is dropped at the end of this function.
2096 while (done != total) {
2098 uint64_t old_oid = 0;
2099 bool create = false;
2101 oid = vid_to_data_oid(inode->data_vdi_id[idx], idx);
2103 len = MIN(total - done, object_size - offset);
2105 switch (acb->aiocb_type) {
2106 case AIOCB_READ_UDATA:
2107 if (!inode->data_vdi_id[idx]) {
2108 qemu_iovec_memset(acb->qiov, done, 0, len);
2112 case AIOCB_WRITE_UDATA:
2113 if (!inode->data_vdi_id[idx]) {
2115 } else if (!is_data_obj_writable(inode, idx)) {
2119 flags = SD_FLAG_CMD_COW;
2122 case AIOCB_DISCARD_OBJ:
2124 * We discard the object only when the whole object is
2125 * 1) allocated 2) trimmed. Otherwise, simply skip it.
2127 if (len != object_size || inode->data_vdi_id[idx] == 0) {
2136 DPRINTF("update ino (%" PRIu32 ") %" PRIu64 " %" PRIu64 " %ld\n",
2138 vid_to_data_oid(inode->data_vdi_id[idx], idx), idx);
2139 oid = vid_to_data_oid(inode->vdi_id, idx);
2140 DPRINTF("new oid %" PRIx64 "\n", oid);
2143 aio_req = alloc_aio_req(s, acb, oid, len, offset, flags, create,
2145 QLIST_INSERT_HEAD(&s->inflight_aio_head, aio_req, aio_siblings);
2147 add_aio_request(s, aio_req, acb->qiov->iov, acb->qiov->niov,
2155 if (!--acb->nr_pending) {
2161 static bool check_overwrapping_aiocb(BDRVSheepdogState *s, SheepdogAIOCB *aiocb)
2165 QLIST_FOREACH(cb, &s->inflight_aiocb_head, aiocb_siblings) {
2166 if (AIOCBOverwrapping(aiocb, cb)) {
2171 QLIST_INSERT_HEAD(&s->inflight_aiocb_head, aiocb, aiocb_siblings);
2175 static coroutine_fn int sd_co_writev(BlockDriverState *bs, int64_t sector_num,
2176 int nb_sectors, QEMUIOVector *qiov)
2180 int64_t offset = (sector_num + nb_sectors) * BDRV_SECTOR_SIZE;
2181 BDRVSheepdogState *s = bs->opaque;
2183 if (offset > s->inode.vdi_size) {
2184 ret = sd_truncate(bs, offset);
2190 acb = sd_aio_setup(bs, qiov, sector_num, nb_sectors);
2191 acb->aio_done_func = sd_write_done;
2192 acb->aiocb_type = AIOCB_WRITE_UDATA;
2195 if (check_overwrapping_aiocb(s, acb)) {
2196 qemu_co_queue_wait(&s->overwrapping_queue);
2200 ret = sd_co_rw_vector(acb);
2202 QLIST_REMOVE(acb, aiocb_siblings);
2203 qemu_co_queue_restart_all(&s->overwrapping_queue);
2204 qemu_aio_unref(acb);
2208 qemu_coroutine_yield();
2210 QLIST_REMOVE(acb, aiocb_siblings);
2211 qemu_co_queue_restart_all(&s->overwrapping_queue);
2216 static coroutine_fn int sd_co_readv(BlockDriverState *bs, int64_t sector_num,
2217 int nb_sectors, QEMUIOVector *qiov)
2221 BDRVSheepdogState *s = bs->opaque;
2223 acb = sd_aio_setup(bs, qiov, sector_num, nb_sectors);
2224 acb->aiocb_type = AIOCB_READ_UDATA;
2225 acb->aio_done_func = sd_finish_aiocb;
2228 if (check_overwrapping_aiocb(s, acb)) {
2229 qemu_co_queue_wait(&s->overwrapping_queue);
2233 ret = sd_co_rw_vector(acb);
2235 QLIST_REMOVE(acb, aiocb_siblings);
2236 qemu_co_queue_restart_all(&s->overwrapping_queue);
2237 qemu_aio_unref(acb);
2241 qemu_coroutine_yield();
2243 QLIST_REMOVE(acb, aiocb_siblings);
2244 qemu_co_queue_restart_all(&s->overwrapping_queue);
2248 static int coroutine_fn sd_co_flush_to_disk(BlockDriverState *bs)
2250 BDRVSheepdogState *s = bs->opaque;
2254 if (s->cache_flags != SD_FLAG_CMD_CACHE) {
2258 acb = sd_aio_setup(bs, NULL, 0, 0);
2259 acb->aiocb_type = AIOCB_FLUSH_CACHE;
2260 acb->aio_done_func = sd_finish_aiocb;
2262 aio_req = alloc_aio_req(s, acb, vid_to_vdi_oid(s->inode.vdi_id),
2263 0, 0, 0, false, 0, 0);
2264 QLIST_INSERT_HEAD(&s->inflight_aio_head, aio_req, aio_siblings);
2265 add_aio_request(s, aio_req, NULL, 0, acb->aiocb_type);
2267 qemu_coroutine_yield();
2271 static int sd_snapshot_create(BlockDriverState *bs, QEMUSnapshotInfo *sn_info)
2273 Error *local_err = NULL;
2274 BDRVSheepdogState *s = bs->opaque;
2277 SheepdogInode *inode;
2278 unsigned int datalen;
2280 DPRINTF("sn_info: name %s id_str %s s: name %s vm_state_size %" PRId64 " "
2281 "is_snapshot %d\n", sn_info->name, sn_info->id_str,
2282 s->name, sn_info->vm_state_size, s->is_snapshot);
2284 if (s->is_snapshot) {
2285 error_report("You can't create a snapshot of a snapshot VDI, "
2286 "%s (%" PRIu32 ").", s->name, s->inode.vdi_id);
2291 DPRINTF("%s %s\n", sn_info->name, sn_info->id_str);
2293 s->inode.vm_state_size = sn_info->vm_state_size;
2294 s->inode.vm_clock_nsec = sn_info->vm_clock_nsec;
2295 /* It appears that inode.tag does not require a NUL terminator,
2296 * which means this use of strncpy is ok.
2298 strncpy(s->inode.tag, sn_info->name, sizeof(s->inode.tag));
2299 /* we don't need to update entire object */
2300 datalen = SD_INODE_SIZE - sizeof(s->inode.data_vdi_id);
2301 inode = g_malloc(datalen);
2303 /* refresh inode. */
2304 fd = connect_to_sdog(s, &local_err);
2306 error_report_err(local_err);
2311 ret = write_object(fd, s->aio_context, (char *)&s->inode,
2312 vid_to_vdi_oid(s->inode.vdi_id), s->inode.nr_copies,
2313 datalen, 0, false, s->cache_flags);
2315 error_report("failed to write snapshot's inode.");
2319 ret = do_sd_create(s, &new_vid, 1, &local_err);
2321 error_report("failed to create inode for snapshot: %s",
2322 error_get_pretty(local_err));
2323 error_free(local_err);
2327 ret = read_object(fd, s->aio_context, (char *)inode,
2328 vid_to_vdi_oid(new_vid), s->inode.nr_copies, datalen, 0,
2332 error_report("failed to read new inode info. %s", strerror(errno));
2336 memcpy(&s->inode, inode, datalen);
2337 DPRINTF("s->inode: name %s snap_id %x oid %x\n",
2338 s->inode.name, s->inode.snap_id, s->inode.vdi_id);
2347 * We implement rollback(loadvm) operation to the specified snapshot by
2348 * 1) switch to the snapshot
2349 * 2) rely on sd_create_branch to delete working VDI and
2350 * 3) create a new working VDI based on the specified snapshot
2352 static int sd_snapshot_goto(BlockDriverState *bs, const char *snapshot_id)
2354 BDRVSheepdogState *s = bs->opaque;
2355 BDRVSheepdogState *old_s;
2356 char tag[SD_MAX_VDI_TAG_LEN];
2357 uint32_t snapid = 0;
2360 old_s = g_new(BDRVSheepdogState, 1);
2362 memcpy(old_s, s, sizeof(BDRVSheepdogState));
2364 snapid = strtoul(snapshot_id, NULL, 10);
2368 pstrcpy(tag, sizeof(tag), snapshot_id);
2371 ret = reload_inode(s, snapid, tag);
2376 ret = sd_create_branch(s);
2385 /* recover bdrv_sd_state */
2386 memcpy(s, old_s, sizeof(BDRVSheepdogState));
2389 error_report("failed to open. recover old bdrv_sd_state.");
2394 static int sd_snapshot_delete(BlockDriverState *bs,
2395 const char *snapshot_id,
2399 /* FIXME: Delete specified snapshot id. */
2403 static int sd_snapshot_list(BlockDriverState *bs, QEMUSnapshotInfo **psn_tab)
2405 Error *local_err = NULL;
2406 BDRVSheepdogState *s = bs->opaque;
2408 int fd, nr = 1024, ret, max = BITS_TO_LONGS(SD_NR_VDIS) * sizeof(long);
2409 QEMUSnapshotInfo *sn_tab = NULL;
2410 unsigned wlen, rlen;
2412 static SheepdogInode inode;
2413 unsigned long *vdi_inuse;
2414 unsigned int start_nr;
2418 vdi_inuse = g_malloc(max);
2420 fd = connect_to_sdog(s, &local_err);
2422 error_report_err(local_err);
2430 memset(&req, 0, sizeof(req));
2432 req.opcode = SD_OP_READ_VDIS;
2433 req.data_length = max;
2435 ret = do_req(fd, s->aio_context, (SheepdogReq *)&req,
2436 vdi_inuse, &wlen, &rlen);
2443 sn_tab = g_new0(QEMUSnapshotInfo, nr);
2445 /* calculate a vdi id with hash function */
2446 hval = fnv_64a_buf(s->name, strlen(s->name), FNV1A_64_INIT);
2447 start_nr = hval & (SD_NR_VDIS - 1);
2449 fd = connect_to_sdog(s, &local_err);
2451 error_report_err(local_err);
2456 for (vid = start_nr; found < nr; vid = (vid + 1) % SD_NR_VDIS) {
2457 if (!test_bit(vid, vdi_inuse)) {
2461 /* we don't need to read entire object */
2462 ret = read_object(fd, s->aio_context, (char *)&inode,
2463 vid_to_vdi_oid(vid),
2464 0, SD_INODE_SIZE - sizeof(inode.data_vdi_id), 0,
2471 if (!strcmp(inode.name, s->name) && is_snapshot(&inode)) {
2472 sn_tab[found].date_sec = inode.snap_ctime >> 32;
2473 sn_tab[found].date_nsec = inode.snap_ctime & 0xffffffff;
2474 sn_tab[found].vm_state_size = inode.vm_state_size;
2475 sn_tab[found].vm_clock_nsec = inode.vm_clock_nsec;
2477 snprintf(sn_tab[found].id_str, sizeof(sn_tab[found].id_str),
2478 "%" PRIu32, inode.snap_id);
2479 pstrcpy(sn_tab[found].name,
2480 MIN(sizeof(sn_tab[found].name), sizeof(inode.tag)),
2499 static int do_load_save_vmstate(BDRVSheepdogState *s, uint8_t *data,
2500 int64_t pos, int size, int load)
2502 Error *local_err = NULL;
2504 int fd, ret = 0, remaining = size;
2505 unsigned int data_len;
2506 uint64_t vmstate_oid;
2509 uint32_t vdi_id = load ? s->inode.parent_vdi_id : s->inode.vdi_id;
2510 uint32_t object_size = (UINT32_C(1) << s->inode.block_size_shift);
2512 fd = connect_to_sdog(s, &local_err);
2514 error_report_err(local_err);
2519 vdi_index = pos / object_size;
2520 offset = pos % object_size;
2522 data_len = MIN(remaining, object_size - offset);
2524 vmstate_oid = vid_to_vmstate_oid(vdi_id, vdi_index);
2526 create = (offset == 0);
2528 ret = read_object(fd, s->aio_context, (char *)data, vmstate_oid,
2529 s->inode.nr_copies, data_len, offset,
2532 ret = write_object(fd, s->aio_context, (char *)data, vmstate_oid,
2533 s->inode.nr_copies, data_len, offset, create,
2538 error_report("failed to save vmstate %s", strerror(errno));
2544 remaining -= data_len;
2552 static int sd_save_vmstate(BlockDriverState *bs, QEMUIOVector *qiov,
2555 BDRVSheepdogState *s = bs->opaque;
2559 buf = qemu_blockalign(bs, qiov->size);
2560 qemu_iovec_to_buf(qiov, 0, buf, qiov->size);
2561 ret = do_load_save_vmstate(s, (uint8_t *) buf, pos, qiov->size, 0);
2567 static int sd_load_vmstate(BlockDriverState *bs, uint8_t *data,
2568 int64_t pos, int size)
2570 BDRVSheepdogState *s = bs->opaque;
2572 return do_load_save_vmstate(s, data, pos, size, 1);
2576 static coroutine_fn int sd_co_discard(BlockDriverState *bs, int64_t sector_num,
2581 BDRVSheepdogState *s = bs->opaque;
2584 if (!s->discard_supported) {
2588 acb = sd_aio_setup(bs, &dummy, sector_num, nb_sectors);
2589 acb->aiocb_type = AIOCB_DISCARD_OBJ;
2590 acb->aio_done_func = sd_finish_aiocb;
2593 if (check_overwrapping_aiocb(s, acb)) {
2594 qemu_co_queue_wait(&s->overwrapping_queue);
2598 ret = sd_co_rw_vector(acb);
2600 QLIST_REMOVE(acb, aiocb_siblings);
2601 qemu_co_queue_restart_all(&s->overwrapping_queue);
2602 qemu_aio_unref(acb);
2606 qemu_coroutine_yield();
2608 QLIST_REMOVE(acb, aiocb_siblings);
2609 qemu_co_queue_restart_all(&s->overwrapping_queue);
2614 static coroutine_fn int64_t
2615 sd_co_get_block_status(BlockDriverState *bs, int64_t sector_num, int nb_sectors,
2618 BDRVSheepdogState *s = bs->opaque;
2619 SheepdogInode *inode = &s->inode;
2620 uint32_t object_size = (UINT32_C(1) << inode->block_size_shift);
2621 uint64_t offset = sector_num * BDRV_SECTOR_SIZE;
2622 unsigned long start = offset / object_size,
2623 end = DIV_ROUND_UP((sector_num + nb_sectors) *
2624 BDRV_SECTOR_SIZE, object_size);
2626 int64_t ret = BDRV_BLOCK_DATA | BDRV_BLOCK_OFFSET_VALID | offset;
2628 for (idx = start; idx < end; idx++) {
2629 if (inode->data_vdi_id[idx] == 0) {
2634 /* Get the longest length of unallocated sectors */
2636 for (idx = start + 1; idx < end; idx++) {
2637 if (inode->data_vdi_id[idx] != 0) {
2643 *pnum = (idx - start) * object_size / BDRV_SECTOR_SIZE;
2644 if (*pnum > nb_sectors) {
2650 static int64_t sd_get_allocated_file_size(BlockDriverState *bs)
2652 BDRVSheepdogState *s = bs->opaque;
2653 SheepdogInode *inode = &s->inode;
2654 uint32_t object_size = (UINT32_C(1) << inode->block_size_shift);
2655 unsigned long i, last = DIV_ROUND_UP(inode->vdi_size, object_size);
2658 for (i = 0; i < last; i++) {
2659 if (inode->data_vdi_id[i] == 0) {
2662 size += object_size;
2667 static QemuOptsList sd_create_opts = {
2668 .name = "sheepdog-create-opts",
2669 .head = QTAILQ_HEAD_INITIALIZER(sd_create_opts.head),
2672 .name = BLOCK_OPT_SIZE,
2673 .type = QEMU_OPT_SIZE,
2674 .help = "Virtual disk size"
2677 .name = BLOCK_OPT_BACKING_FILE,
2678 .type = QEMU_OPT_STRING,
2679 .help = "File name of a base image"
2682 .name = BLOCK_OPT_PREALLOC,
2683 .type = QEMU_OPT_STRING,
2684 .help = "Preallocation mode (allowed values: off, full)"
2687 .name = BLOCK_OPT_REDUNDANCY,
2688 .type = QEMU_OPT_STRING,
2689 .help = "Redundancy of the image"
2692 .name = BLOCK_OPT_OBJECT_SIZE,
2693 .type = QEMU_OPT_SIZE,
2694 .help = "Object size of the image"
2696 { /* end of list */ }
2700 static BlockDriver bdrv_sheepdog = {
2701 .format_name = "sheepdog",
2702 .protocol_name = "sheepdog",
2703 .instance_size = sizeof(BDRVSheepdogState),
2704 .bdrv_needs_filename = true,
2705 .bdrv_file_open = sd_open,
2706 .bdrv_close = sd_close,
2707 .bdrv_create = sd_create,
2708 .bdrv_has_zero_init = bdrv_has_zero_init_1,
2709 .bdrv_getlength = sd_getlength,
2710 .bdrv_get_allocated_file_size = sd_get_allocated_file_size,
2711 .bdrv_truncate = sd_truncate,
2713 .bdrv_co_readv = sd_co_readv,
2714 .bdrv_co_writev = sd_co_writev,
2715 .bdrv_co_flush_to_disk = sd_co_flush_to_disk,
2716 .bdrv_co_discard = sd_co_discard,
2717 .bdrv_co_get_block_status = sd_co_get_block_status,
2719 .bdrv_snapshot_create = sd_snapshot_create,
2720 .bdrv_snapshot_goto = sd_snapshot_goto,
2721 .bdrv_snapshot_delete = sd_snapshot_delete,
2722 .bdrv_snapshot_list = sd_snapshot_list,
2724 .bdrv_save_vmstate = sd_save_vmstate,
2725 .bdrv_load_vmstate = sd_load_vmstate,
2727 .bdrv_detach_aio_context = sd_detach_aio_context,
2728 .bdrv_attach_aio_context = sd_attach_aio_context,
2730 .create_opts = &sd_create_opts,
2733 static BlockDriver bdrv_sheepdog_tcp = {
2734 .format_name = "sheepdog",
2735 .protocol_name = "sheepdog+tcp",
2736 .instance_size = sizeof(BDRVSheepdogState),
2737 .bdrv_needs_filename = true,
2738 .bdrv_file_open = sd_open,
2739 .bdrv_close = sd_close,
2740 .bdrv_create = sd_create,
2741 .bdrv_has_zero_init = bdrv_has_zero_init_1,
2742 .bdrv_getlength = sd_getlength,
2743 .bdrv_get_allocated_file_size = sd_get_allocated_file_size,
2744 .bdrv_truncate = sd_truncate,
2746 .bdrv_co_readv = sd_co_readv,
2747 .bdrv_co_writev = sd_co_writev,
2748 .bdrv_co_flush_to_disk = sd_co_flush_to_disk,
2749 .bdrv_co_discard = sd_co_discard,
2750 .bdrv_co_get_block_status = sd_co_get_block_status,
2752 .bdrv_snapshot_create = sd_snapshot_create,
2753 .bdrv_snapshot_goto = sd_snapshot_goto,
2754 .bdrv_snapshot_delete = sd_snapshot_delete,
2755 .bdrv_snapshot_list = sd_snapshot_list,
2757 .bdrv_save_vmstate = sd_save_vmstate,
2758 .bdrv_load_vmstate = sd_load_vmstate,
2760 .bdrv_detach_aio_context = sd_detach_aio_context,
2761 .bdrv_attach_aio_context = sd_attach_aio_context,
2763 .create_opts = &sd_create_opts,
2766 static BlockDriver bdrv_sheepdog_unix = {
2767 .format_name = "sheepdog",
2768 .protocol_name = "sheepdog+unix",
2769 .instance_size = sizeof(BDRVSheepdogState),
2770 .bdrv_needs_filename = true,
2771 .bdrv_file_open = sd_open,
2772 .bdrv_close = sd_close,
2773 .bdrv_create = sd_create,
2774 .bdrv_has_zero_init = bdrv_has_zero_init_1,
2775 .bdrv_getlength = sd_getlength,
2776 .bdrv_get_allocated_file_size = sd_get_allocated_file_size,
2777 .bdrv_truncate = sd_truncate,
2779 .bdrv_co_readv = sd_co_readv,
2780 .bdrv_co_writev = sd_co_writev,
2781 .bdrv_co_flush_to_disk = sd_co_flush_to_disk,
2782 .bdrv_co_discard = sd_co_discard,
2783 .bdrv_co_get_block_status = sd_co_get_block_status,
2785 .bdrv_snapshot_create = sd_snapshot_create,
2786 .bdrv_snapshot_goto = sd_snapshot_goto,
2787 .bdrv_snapshot_delete = sd_snapshot_delete,
2788 .bdrv_snapshot_list = sd_snapshot_list,
2790 .bdrv_save_vmstate = sd_save_vmstate,
2791 .bdrv_load_vmstate = sd_load_vmstate,
2793 .bdrv_detach_aio_context = sd_detach_aio_context,
2794 .bdrv_attach_aio_context = sd_attach_aio_context,
2796 .create_opts = &sd_create_opts,
2799 static void bdrv_sheepdog_init(void)
2801 bdrv_register(&bdrv_sheepdog);
2802 bdrv_register(&bdrv_sheepdog_tcp);
2803 bdrv_register(&bdrv_sheepdog_unix);
2805 block_init(bdrv_sheepdog_init);