2 * GlusterFS backend for QEMU
4 * Copyright (C) 2012 Bharata B Rao <bharata@linux.vnet.ibm.com>
6 * This work is licensed under the terms of the GNU GPL, version 2 or later.
7 * See the COPYING file in the top-level directory.
10 #include <glusterfs/api/glfs.h>
11 #include "block/block_int.h"
14 typedef struct GlusterAIOCB
{
19 AioContext
*aio_context
;
22 typedef struct BDRVGlusterState
{
27 typedef struct GlusterConf
{
35 static void qemu_gluster_gconf_free(GlusterConf
*gconf
)
38 g_free(gconf
->server
);
39 g_free(gconf
->volname
);
41 g_free(gconf
->transport
);
46 static int parse_volume_options(GlusterConf
*gconf
, char *path
)
55 p
= q
= path
+ strspn(path
, "/");
60 gconf
->volname
= g_strndup(q
, p
- q
);
67 gconf
->image
= g_strdup(p
);
72 * file=gluster[+transport]://[server[:port]]/volname/image[?socket=...]
74 * 'gluster' is the protocol.
76 * 'transport' specifies the transport type used to connect to gluster
77 * management daemon (glusterd). Valid transport types are
78 * tcp, unix and rdma. If a transport type isn't specified, then tcp
81 * 'server' specifies the server where the volume file specification for
82 * the given volume resides. This can be either hostname, ipv4 address
83 * or ipv6 address. ipv6 address needs to be within square brackets [ ].
84 * If transport type is 'unix', then 'server' field should not be specified.
85 * The 'socket' field needs to be populated with the path to unix domain
88 * 'port' is the port number on which glusterd is listening. This is optional
89 * and if not specified, QEMU will send 0 which will make gluster to use the
90 * default port. If the transport type is unix, then 'port' should not be
93 * 'volname' is the name of the gluster volume which contains the VM image.
95 * 'image' is the path to the actual VM image that resides on gluster volume.
99 * file=gluster://1.2.3.4/testvol/a.img
100 * file=gluster+tcp://1.2.3.4/testvol/a.img
101 * file=gluster+tcp://1.2.3.4:24007/testvol/dir/a.img
102 * file=gluster+tcp://[1:2:3:4:5:6:7:8]/testvol/dir/a.img
103 * file=gluster+tcp://[1:2:3:4:5:6:7:8]:24007/testvol/dir/a.img
104 * file=gluster+tcp://server.domain.com:24007/testvol/dir/a.img
105 * file=gluster+unix:///testvol/dir/a.img?socket=/tmp/glusterd.socket
106 * file=gluster+rdma://1.2.3.4:24007/testvol/a.img
108 static int qemu_gluster_parseuri(GlusterConf
*gconf
, const char *filename
)
111 QueryParams
*qp
= NULL
;
112 bool is_unix
= false;
115 uri
= uri_parse(filename
);
121 if (!uri
->scheme
|| !strcmp(uri
->scheme
, "gluster")) {
122 gconf
->transport
= g_strdup("tcp");
123 } else if (!strcmp(uri
->scheme
, "gluster+tcp")) {
124 gconf
->transport
= g_strdup("tcp");
125 } else if (!strcmp(uri
->scheme
, "gluster+unix")) {
126 gconf
->transport
= g_strdup("unix");
128 } else if (!strcmp(uri
->scheme
, "gluster+rdma")) {
129 gconf
->transport
= g_strdup("rdma");
135 ret
= parse_volume_options(gconf
, uri
->path
);
140 qp
= query_params_parse(uri
->query
);
141 if (qp
->n
> 1 || (is_unix
&& !qp
->n
) || (!is_unix
&& qp
->n
)) {
147 if (uri
->server
|| uri
->port
) {
151 if (strcmp(qp
->p
[0].name
, "socket")) {
155 gconf
->server
= g_strdup(qp
->p
[0].value
);
157 gconf
->server
= g_strdup(uri
->server
? uri
->server
: "localhost");
158 gconf
->port
= uri
->port
;
163 query_params_free(qp
);
169 static struct glfs
*qemu_gluster_init(GlusterConf
*gconf
, const char *filename
,
172 struct glfs
*glfs
= NULL
;
176 ret
= qemu_gluster_parseuri(gconf
, filename
);
178 error_setg(errp
, "Usage: file=gluster[+transport]://[server[:port]]/"
179 "volname/image[?socket=...]");
184 glfs
= glfs_new(gconf
->volname
);
189 ret
= glfs_set_volfile_server(glfs
, gconf
->transport
, gconf
->server
,
196 * TODO: Use GF_LOG_ERROR instead of hard code value of 4 here when
197 * GlusterFS makes GF_LOG_* macros available to libgfapi users.
199 ret
= glfs_set_logging(glfs
, "-", 4);
204 ret
= glfs_init(glfs
);
206 error_setg_errno(errp
, errno
,
207 "Gluster connection failed for server=%s port=%d "
208 "volume=%s image=%s transport=%s", gconf
->server
,
209 gconf
->port
, gconf
->volname
, gconf
->image
,
212 /* glfs_init sometimes doesn't set errno although docs suggest that */
229 static void qemu_gluster_complete_aio(void *opaque
)
231 GlusterAIOCB
*acb
= (GlusterAIOCB
*)opaque
;
233 qemu_bh_delete(acb
->bh
);
235 qemu_coroutine_enter(acb
->coroutine
, NULL
);
239 * AIO callback routine called from GlusterFS thread.
241 static void gluster_finish_aiocb(struct glfs_fd
*fd
, ssize_t ret
, void *arg
)
243 GlusterAIOCB
*acb
= (GlusterAIOCB
*)arg
;
245 if (!ret
|| ret
== acb
->size
) {
246 acb
->ret
= 0; /* Success */
247 } else if (ret
< 0) {
248 acb
->ret
= ret
; /* Read/Write failed */
250 acb
->ret
= -EIO
; /* Partial read/write - fail it */
253 acb
->bh
= aio_bh_new(acb
->aio_context
, qemu_gluster_complete_aio
, acb
);
254 qemu_bh_schedule(acb
->bh
);
257 /* TODO Convert to fine grained options */
258 static QemuOptsList runtime_opts
= {
260 .head
= QTAILQ_HEAD_INITIALIZER(runtime_opts
.head
),
264 .type
= QEMU_OPT_STRING
,
265 .help
= "URL to the gluster image",
267 { /* end of list */ }
271 static void qemu_gluster_parse_flags(int bdrv_flags
, int *open_flags
)
273 assert(open_flags
!= NULL
);
275 *open_flags
|= O_BINARY
;
277 if (bdrv_flags
& BDRV_O_RDWR
) {
278 *open_flags
|= O_RDWR
;
280 *open_flags
|= O_RDONLY
;
283 if ((bdrv_flags
& BDRV_O_NOCACHE
)) {
284 *open_flags
|= O_DIRECT
;
288 static int qemu_gluster_open(BlockDriverState
*bs
, QDict
*options
,
289 int bdrv_flags
, Error
**errp
)
291 BDRVGlusterState
*s
= bs
->opaque
;
294 GlusterConf
*gconf
= g_malloc0(sizeof(GlusterConf
));
296 Error
*local_err
= NULL
;
297 const char *filename
;
299 opts
= qemu_opts_create(&runtime_opts
, NULL
, 0, &error_abort
);
300 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
302 error_propagate(errp
, local_err
);
307 filename
= qemu_opt_get(opts
, "filename");
309 s
->glfs
= qemu_gluster_init(gconf
, filename
, errp
);
315 qemu_gluster_parse_flags(bdrv_flags
, &open_flags
);
317 s
->fd
= glfs_open(s
->glfs
, gconf
->image
, open_flags
);
324 qemu_gluster_gconf_free(gconf
);
337 typedef struct BDRVGlusterReopenState
{
340 } BDRVGlusterReopenState
;
343 static int qemu_gluster_reopen_prepare(BDRVReopenState
*state
,
344 BlockReopenQueue
*queue
, Error
**errp
)
347 BDRVGlusterReopenState
*reop_s
;
348 GlusterConf
*gconf
= NULL
;
351 assert(state
!= NULL
);
352 assert(state
->bs
!= NULL
);
354 state
->opaque
= g_malloc0(sizeof(BDRVGlusterReopenState
));
355 reop_s
= state
->opaque
;
357 qemu_gluster_parse_flags(state
->flags
, &open_flags
);
359 gconf
= g_malloc0(sizeof(GlusterConf
));
361 reop_s
->glfs
= qemu_gluster_init(gconf
, state
->bs
->filename
, errp
);
362 if (reop_s
->glfs
== NULL
) {
367 reop_s
->fd
= glfs_open(reop_s
->glfs
, gconf
->image
, open_flags
);
368 if (reop_s
->fd
== NULL
) {
369 /* reops->glfs will be cleaned up in _abort */
375 /* state->opaque will be freed in either the _abort or _commit */
376 qemu_gluster_gconf_free(gconf
);
380 static void qemu_gluster_reopen_commit(BDRVReopenState
*state
)
382 BDRVGlusterReopenState
*reop_s
= state
->opaque
;
383 BDRVGlusterState
*s
= state
->bs
->opaque
;
394 /* use the newly opened image / connection */
396 s
->glfs
= reop_s
->glfs
;
398 g_free(state
->opaque
);
399 state
->opaque
= NULL
;
405 static void qemu_gluster_reopen_abort(BDRVReopenState
*state
)
407 BDRVGlusterReopenState
*reop_s
= state
->opaque
;
409 if (reop_s
== NULL
) {
414 glfs_close(reop_s
->fd
);
418 glfs_fini(reop_s
->glfs
);
421 g_free(state
->opaque
);
422 state
->opaque
= NULL
;
427 #ifdef CONFIG_GLUSTERFS_ZEROFILL
428 static coroutine_fn
int qemu_gluster_co_write_zeroes(BlockDriverState
*bs
,
429 int64_t sector_num
, int nb_sectors
, BdrvRequestFlags flags
)
432 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
433 BDRVGlusterState
*s
= bs
->opaque
;
434 off_t size
= nb_sectors
* BDRV_SECTOR_SIZE
;
435 off_t offset
= sector_num
* BDRV_SECTOR_SIZE
;
439 acb
->coroutine
= qemu_coroutine_self();
440 acb
->aio_context
= bdrv_get_aio_context(bs
);
442 ret
= glfs_zerofill_async(s
->fd
, offset
, size
, &gluster_finish_aiocb
, acb
);
448 qemu_coroutine_yield();
452 g_slice_free(GlusterAIOCB
, acb
);
456 static inline bool gluster_supports_zerofill(void)
461 static inline int qemu_gluster_zerofill(struct glfs_fd
*fd
, int64_t offset
,
464 return glfs_zerofill(fd
, offset
, size
);
468 static inline bool gluster_supports_zerofill(void)
473 static inline int qemu_gluster_zerofill(struct glfs_fd
*fd
, int64_t offset
,
480 static int qemu_gluster_create(const char *filename
,
481 QemuOpts
*opts
, Error
**errp
)
487 int64_t total_size
= 0;
489 GlusterConf
*gconf
= g_malloc0(sizeof(GlusterConf
));
491 glfs
= qemu_gluster_init(gconf
, filename
, errp
);
498 qemu_opt_get_size_del(opts
, BLOCK_OPT_SIZE
, 0) / BDRV_SECTOR_SIZE
;
500 tmp
= qemu_opt_get_del(opts
, BLOCK_OPT_PREALLOC
);
501 if (!tmp
|| !strcmp(tmp
, "off")) {
503 } else if (!strcmp(tmp
, "full") &&
504 gluster_supports_zerofill()) {
507 error_setg(errp
, "Invalid preallocation mode: '%s'"
508 " or GlusterFS doesn't support zerofill API",
514 fd
= glfs_creat(glfs
, gconf
->image
,
515 O_WRONLY
| O_CREAT
| O_TRUNC
| O_BINARY
, S_IRUSR
| S_IWUSR
);
519 if (!glfs_ftruncate(fd
, total_size
* BDRV_SECTOR_SIZE
)) {
520 if (prealloc
&& qemu_gluster_zerofill(fd
, 0,
521 total_size
* BDRV_SECTOR_SIZE
)) {
528 if (glfs_close(fd
) != 0) {
534 qemu_gluster_gconf_free(gconf
);
541 static coroutine_fn
int qemu_gluster_co_rw(BlockDriverState
*bs
,
542 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
, int write
)
545 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
546 BDRVGlusterState
*s
= bs
->opaque
;
547 size_t size
= nb_sectors
* BDRV_SECTOR_SIZE
;
548 off_t offset
= sector_num
* BDRV_SECTOR_SIZE
;
552 acb
->coroutine
= qemu_coroutine_self();
553 acb
->aio_context
= bdrv_get_aio_context(bs
);
556 ret
= glfs_pwritev_async(s
->fd
, qiov
->iov
, qiov
->niov
, offset
, 0,
557 &gluster_finish_aiocb
, acb
);
559 ret
= glfs_preadv_async(s
->fd
, qiov
->iov
, qiov
->niov
, offset
, 0,
560 &gluster_finish_aiocb
, acb
);
568 qemu_coroutine_yield();
572 g_slice_free(GlusterAIOCB
, acb
);
576 static int qemu_gluster_truncate(BlockDriverState
*bs
, int64_t offset
)
579 BDRVGlusterState
*s
= bs
->opaque
;
581 ret
= glfs_ftruncate(s
->fd
, offset
);
589 static coroutine_fn
int qemu_gluster_co_readv(BlockDriverState
*bs
,
590 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
592 return qemu_gluster_co_rw(bs
, sector_num
, nb_sectors
, qiov
, 0);
595 static coroutine_fn
int qemu_gluster_co_writev(BlockDriverState
*bs
,
596 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
598 return qemu_gluster_co_rw(bs
, sector_num
, nb_sectors
, qiov
, 1);
601 static coroutine_fn
int qemu_gluster_co_flush_to_disk(BlockDriverState
*bs
)
604 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
605 BDRVGlusterState
*s
= bs
->opaque
;
609 acb
->coroutine
= qemu_coroutine_self();
610 acb
->aio_context
= bdrv_get_aio_context(bs
);
612 ret
= glfs_fsync_async(s
->fd
, &gluster_finish_aiocb
, acb
);
618 qemu_coroutine_yield();
622 g_slice_free(GlusterAIOCB
, acb
);
626 #ifdef CONFIG_GLUSTERFS_DISCARD
627 static coroutine_fn
int qemu_gluster_co_discard(BlockDriverState
*bs
,
628 int64_t sector_num
, int nb_sectors
)
631 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
632 BDRVGlusterState
*s
= bs
->opaque
;
633 size_t size
= nb_sectors
* BDRV_SECTOR_SIZE
;
634 off_t offset
= sector_num
* BDRV_SECTOR_SIZE
;
638 acb
->coroutine
= qemu_coroutine_self();
639 acb
->aio_context
= bdrv_get_aio_context(bs
);
641 ret
= glfs_discard_async(s
->fd
, offset
, size
, &gluster_finish_aiocb
, acb
);
647 qemu_coroutine_yield();
651 g_slice_free(GlusterAIOCB
, acb
);
656 static int64_t qemu_gluster_getlength(BlockDriverState
*bs
)
658 BDRVGlusterState
*s
= bs
->opaque
;
661 ret
= glfs_lseek(s
->fd
, 0, SEEK_END
);
669 static int64_t qemu_gluster_allocated_file_size(BlockDriverState
*bs
)
671 BDRVGlusterState
*s
= bs
->opaque
;
675 ret
= glfs_fstat(s
->fd
, &st
);
679 return st
.st_blocks
* 512;
683 static void qemu_gluster_close(BlockDriverState
*bs
)
685 BDRVGlusterState
*s
= bs
->opaque
;
694 static int qemu_gluster_has_zero_init(BlockDriverState
*bs
)
696 /* GlusterFS volume could be backed by a block device */
700 static QemuOptsList qemu_gluster_create_opts
= {
701 .name
= "qemu-gluster-create-opts",
702 .head
= QTAILQ_HEAD_INITIALIZER(qemu_gluster_create_opts
.head
),
705 .name
= BLOCK_OPT_SIZE
,
706 .type
= QEMU_OPT_SIZE
,
707 .help
= "Virtual disk size"
710 .name
= BLOCK_OPT_PREALLOC
,
711 .type
= QEMU_OPT_STRING
,
712 .help
= "Preallocation mode (allowed values: off, full)"
714 { /* end of list */ }
718 static BlockDriver bdrv_gluster
= {
719 .format_name
= "gluster",
720 .protocol_name
= "gluster",
721 .instance_size
= sizeof(BDRVGlusterState
),
722 .bdrv_needs_filename
= true,
723 .bdrv_file_open
= qemu_gluster_open
,
724 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
725 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
726 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
727 .bdrv_close
= qemu_gluster_close
,
728 .bdrv_create
= qemu_gluster_create
,
729 .bdrv_getlength
= qemu_gluster_getlength
,
730 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
731 .bdrv_truncate
= qemu_gluster_truncate
,
732 .bdrv_co_readv
= qemu_gluster_co_readv
,
733 .bdrv_co_writev
= qemu_gluster_co_writev
,
734 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
735 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
736 #ifdef CONFIG_GLUSTERFS_DISCARD
737 .bdrv_co_discard
= qemu_gluster_co_discard
,
739 #ifdef CONFIG_GLUSTERFS_ZEROFILL
740 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
742 .create_opts
= &qemu_gluster_create_opts
,
745 static BlockDriver bdrv_gluster_tcp
= {
746 .format_name
= "gluster",
747 .protocol_name
= "gluster+tcp",
748 .instance_size
= sizeof(BDRVGlusterState
),
749 .bdrv_needs_filename
= true,
750 .bdrv_file_open
= qemu_gluster_open
,
751 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
752 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
753 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
754 .bdrv_close
= qemu_gluster_close
,
755 .bdrv_create
= qemu_gluster_create
,
756 .bdrv_getlength
= qemu_gluster_getlength
,
757 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
758 .bdrv_truncate
= qemu_gluster_truncate
,
759 .bdrv_co_readv
= qemu_gluster_co_readv
,
760 .bdrv_co_writev
= qemu_gluster_co_writev
,
761 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
762 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
763 #ifdef CONFIG_GLUSTERFS_DISCARD
764 .bdrv_co_discard
= qemu_gluster_co_discard
,
766 #ifdef CONFIG_GLUSTERFS_ZEROFILL
767 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
769 .create_opts
= &qemu_gluster_create_opts
,
772 static BlockDriver bdrv_gluster_unix
= {
773 .format_name
= "gluster",
774 .protocol_name
= "gluster+unix",
775 .instance_size
= sizeof(BDRVGlusterState
),
776 .bdrv_needs_filename
= true,
777 .bdrv_file_open
= qemu_gluster_open
,
778 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
779 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
780 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
781 .bdrv_close
= qemu_gluster_close
,
782 .bdrv_create
= qemu_gluster_create
,
783 .bdrv_getlength
= qemu_gluster_getlength
,
784 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
785 .bdrv_truncate
= qemu_gluster_truncate
,
786 .bdrv_co_readv
= qemu_gluster_co_readv
,
787 .bdrv_co_writev
= qemu_gluster_co_writev
,
788 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
789 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
790 #ifdef CONFIG_GLUSTERFS_DISCARD
791 .bdrv_co_discard
= qemu_gluster_co_discard
,
793 #ifdef CONFIG_GLUSTERFS_ZEROFILL
794 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
796 .create_opts
= &qemu_gluster_create_opts
,
799 static BlockDriver bdrv_gluster_rdma
= {
800 .format_name
= "gluster",
801 .protocol_name
= "gluster+rdma",
802 .instance_size
= sizeof(BDRVGlusterState
),
803 .bdrv_needs_filename
= true,
804 .bdrv_file_open
= qemu_gluster_open
,
805 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
806 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
807 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
808 .bdrv_close
= qemu_gluster_close
,
809 .bdrv_create
= qemu_gluster_create
,
810 .bdrv_getlength
= qemu_gluster_getlength
,
811 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
812 .bdrv_truncate
= qemu_gluster_truncate
,
813 .bdrv_co_readv
= qemu_gluster_co_readv
,
814 .bdrv_co_writev
= qemu_gluster_co_writev
,
815 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
816 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
817 #ifdef CONFIG_GLUSTERFS_DISCARD
818 .bdrv_co_discard
= qemu_gluster_co_discard
,
820 #ifdef CONFIG_GLUSTERFS_ZEROFILL
821 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
823 .create_opts
= &qemu_gluster_create_opts
,
826 static void bdrv_gluster_init(void)
828 bdrv_register(&bdrv_gluster_rdma
);
829 bdrv_register(&bdrv_gluster_unix
);
830 bdrv_register(&bdrv_gluster_tcp
);
831 bdrv_register(&bdrv_gluster
);
834 block_init(bdrv_gluster_init
);