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
{
21 typedef struct BDRVGlusterState
{
26 typedef struct GlusterConf
{
34 static void qemu_gluster_gconf_free(GlusterConf
*gconf
)
37 g_free(gconf
->server
);
38 g_free(gconf
->volname
);
40 g_free(gconf
->transport
);
45 static int parse_volume_options(GlusterConf
*gconf
, char *path
)
54 p
= q
= path
+ strspn(path
, "/");
59 gconf
->volname
= g_strndup(q
, p
- q
);
66 gconf
->image
= g_strdup(p
);
71 * file=gluster[+transport]://[server[:port]]/volname/image[?socket=...]
73 * 'gluster' is the protocol.
75 * 'transport' specifies the transport type used to connect to gluster
76 * management daemon (glusterd). Valid transport types are
77 * tcp, unix and rdma. If a transport type isn't specified, then tcp
80 * 'server' specifies the server where the volume file specification for
81 * the given volume resides. This can be either hostname, ipv4 address
82 * or ipv6 address. ipv6 address needs to be within square brackets [ ].
83 * If transport type is 'unix', then 'server' field should not be specified.
84 * The 'socket' field needs to be populated with the path to unix domain
87 * 'port' is the port number on which glusterd is listening. This is optional
88 * and if not specified, QEMU will send 0 which will make gluster to use the
89 * default port. If the transport type is unix, then 'port' should not be
92 * 'volname' is the name of the gluster volume which contains the VM image.
94 * 'image' is the path to the actual VM image that resides on gluster volume.
98 * file=gluster://1.2.3.4/testvol/a.img
99 * file=gluster+tcp://1.2.3.4/testvol/a.img
100 * file=gluster+tcp://1.2.3.4:24007/testvol/dir/a.img
101 * file=gluster+tcp://[1:2:3:4:5:6:7:8]/testvol/dir/a.img
102 * file=gluster+tcp://[1:2:3:4:5:6:7:8]:24007/testvol/dir/a.img
103 * file=gluster+tcp://server.domain.com:24007/testvol/dir/a.img
104 * file=gluster+unix:///testvol/dir/a.img?socket=/tmp/glusterd.socket
105 * file=gluster+rdma://1.2.3.4:24007/testvol/a.img
107 static int qemu_gluster_parseuri(GlusterConf
*gconf
, const char *filename
)
110 QueryParams
*qp
= NULL
;
111 bool is_unix
= false;
114 uri
= uri_parse(filename
);
120 if (!uri
->scheme
|| !strcmp(uri
->scheme
, "gluster")) {
121 gconf
->transport
= g_strdup("tcp");
122 } else if (!strcmp(uri
->scheme
, "gluster+tcp")) {
123 gconf
->transport
= g_strdup("tcp");
124 } else if (!strcmp(uri
->scheme
, "gluster+unix")) {
125 gconf
->transport
= g_strdup("unix");
127 } else if (!strcmp(uri
->scheme
, "gluster+rdma")) {
128 gconf
->transport
= g_strdup("rdma");
134 ret
= parse_volume_options(gconf
, uri
->path
);
139 qp
= query_params_parse(uri
->query
);
140 if (qp
->n
> 1 || (is_unix
&& !qp
->n
) || (!is_unix
&& qp
->n
)) {
146 if (uri
->server
|| uri
->port
) {
150 if (strcmp(qp
->p
[0].name
, "socket")) {
154 gconf
->server
= g_strdup(qp
->p
[0].value
);
156 gconf
->server
= g_strdup(uri
->server
? uri
->server
: "localhost");
157 gconf
->port
= uri
->port
;
162 query_params_free(qp
);
168 static struct glfs
*qemu_gluster_init(GlusterConf
*gconf
, const char *filename
,
171 struct glfs
*glfs
= NULL
;
175 ret
= qemu_gluster_parseuri(gconf
, filename
);
177 error_setg(errp
, "Usage: file=gluster[+transport]://[server[:port]]/"
178 "volname/image[?socket=...]");
183 glfs
= glfs_new(gconf
->volname
);
188 ret
= glfs_set_volfile_server(glfs
, gconf
->transport
, gconf
->server
,
195 * TODO: Use GF_LOG_ERROR instead of hard code value of 4 here when
196 * GlusterFS makes GF_LOG_* macros available to libgfapi users.
198 ret
= glfs_set_logging(glfs
, "-", 4);
203 ret
= glfs_init(glfs
);
205 error_setg_errno(errp
, errno
,
206 "Gluster connection failed for server=%s port=%d "
207 "volume=%s image=%s transport=%s", gconf
->server
,
208 gconf
->port
, gconf
->volname
, gconf
->image
,
211 /* glfs_init sometimes doesn't set errno although docs suggest that */
228 static void qemu_gluster_complete_aio(void *opaque
)
230 GlusterAIOCB
*acb
= (GlusterAIOCB
*)opaque
;
232 qemu_bh_delete(acb
->bh
);
234 qemu_coroutine_enter(acb
->coroutine
, NULL
);
238 * AIO callback routine called from GlusterFS thread.
240 static void gluster_finish_aiocb(struct glfs_fd
*fd
, ssize_t ret
, void *arg
)
242 GlusterAIOCB
*acb
= (GlusterAIOCB
*)arg
;
244 if (!ret
|| ret
== acb
->size
) {
245 acb
->ret
= 0; /* Success */
246 } else if (ret
< 0) {
247 acb
->ret
= ret
; /* Read/Write failed */
249 acb
->ret
= -EIO
; /* Partial read/write - fail it */
252 acb
->bh
= qemu_bh_new(qemu_gluster_complete_aio
, acb
);
253 qemu_bh_schedule(acb
->bh
);
256 /* TODO Convert to fine grained options */
257 static QemuOptsList runtime_opts
= {
259 .head
= QTAILQ_HEAD_INITIALIZER(runtime_opts
.head
),
263 .type
= QEMU_OPT_STRING
,
264 .help
= "URL to the gluster image",
266 { /* end of list */ }
270 static void qemu_gluster_parse_flags(int bdrv_flags
, int *open_flags
)
272 assert(open_flags
!= NULL
);
274 *open_flags
|= O_BINARY
;
276 if (bdrv_flags
& BDRV_O_RDWR
) {
277 *open_flags
|= O_RDWR
;
279 *open_flags
|= O_RDONLY
;
282 if ((bdrv_flags
& BDRV_O_NOCACHE
)) {
283 *open_flags
|= O_DIRECT
;
287 static int qemu_gluster_open(BlockDriverState
*bs
, QDict
*options
,
288 int bdrv_flags
, Error
**errp
)
290 BDRVGlusterState
*s
= bs
->opaque
;
293 GlusterConf
*gconf
= g_malloc0(sizeof(GlusterConf
));
295 Error
*local_err
= NULL
;
296 const char *filename
;
298 opts
= qemu_opts_create(&runtime_opts
, NULL
, 0, &error_abort
);
299 qemu_opts_absorb_qdict(opts
, options
, &local_err
);
301 error_propagate(errp
, local_err
);
306 filename
= qemu_opt_get(opts
, "filename");
308 s
->glfs
= qemu_gluster_init(gconf
, filename
, errp
);
314 qemu_gluster_parse_flags(bdrv_flags
, &open_flags
);
316 s
->fd
= glfs_open(s
->glfs
, gconf
->image
, open_flags
);
323 qemu_gluster_gconf_free(gconf
);
336 typedef struct BDRVGlusterReopenState
{
339 } BDRVGlusterReopenState
;
342 static int qemu_gluster_reopen_prepare(BDRVReopenState
*state
,
343 BlockReopenQueue
*queue
, Error
**errp
)
346 BDRVGlusterReopenState
*reop_s
;
347 GlusterConf
*gconf
= NULL
;
350 assert(state
!= NULL
);
351 assert(state
->bs
!= NULL
);
353 state
->opaque
= g_malloc0(sizeof(BDRVGlusterReopenState
));
354 reop_s
= state
->opaque
;
356 qemu_gluster_parse_flags(state
->flags
, &open_flags
);
358 gconf
= g_malloc0(sizeof(GlusterConf
));
360 reop_s
->glfs
= qemu_gluster_init(gconf
, state
->bs
->filename
, errp
);
361 if (reop_s
->glfs
== NULL
) {
366 reop_s
->fd
= glfs_open(reop_s
->glfs
, gconf
->image
, open_flags
);
367 if (reop_s
->fd
== NULL
) {
368 /* reops->glfs will be cleaned up in _abort */
374 /* state->opaque will be freed in either the _abort or _commit */
375 qemu_gluster_gconf_free(gconf
);
379 static void qemu_gluster_reopen_commit(BDRVReopenState
*state
)
381 BDRVGlusterReopenState
*reop_s
= state
->opaque
;
382 BDRVGlusterState
*s
= state
->bs
->opaque
;
393 /* use the newly opened image / connection */
395 s
->glfs
= reop_s
->glfs
;
397 g_free(state
->opaque
);
398 state
->opaque
= NULL
;
404 static void qemu_gluster_reopen_abort(BDRVReopenState
*state
)
406 BDRVGlusterReopenState
*reop_s
= state
->opaque
;
408 if (reop_s
== NULL
) {
413 glfs_close(reop_s
->fd
);
417 glfs_fini(reop_s
->glfs
);
420 g_free(state
->opaque
);
421 state
->opaque
= NULL
;
426 #ifdef CONFIG_GLUSTERFS_ZEROFILL
427 static coroutine_fn
int qemu_gluster_co_write_zeroes(BlockDriverState
*bs
,
428 int64_t sector_num
, int nb_sectors
, BdrvRequestFlags flags
)
431 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
432 BDRVGlusterState
*s
= bs
->opaque
;
433 off_t size
= nb_sectors
* BDRV_SECTOR_SIZE
;
434 off_t offset
= sector_num
* BDRV_SECTOR_SIZE
;
438 acb
->coroutine
= qemu_coroutine_self();
440 ret
= glfs_zerofill_async(s
->fd
, offset
, size
, &gluster_finish_aiocb
, acb
);
446 qemu_coroutine_yield();
450 g_slice_free(GlusterAIOCB
, acb
);
454 static inline bool gluster_supports_zerofill(void)
459 static inline int qemu_gluster_zerofill(struct glfs_fd
*fd
, int64_t offset
,
462 return glfs_zerofill(fd
, offset
, size
);
466 static inline bool gluster_supports_zerofill(void)
471 static inline int qemu_gluster_zerofill(struct glfs_fd
*fd
, int64_t offset
,
478 static int qemu_gluster_create(const char *filename
,
479 QEMUOptionParameter
*options
, Error
**errp
)
485 int64_t total_size
= 0;
486 GlusterConf
*gconf
= g_malloc0(sizeof(GlusterConf
));
488 glfs
= qemu_gluster_init(gconf
, filename
, errp
);
494 while (options
&& options
->name
) {
495 if (!strcmp(options
->name
, BLOCK_OPT_SIZE
)) {
496 total_size
= options
->value
.n
/ BDRV_SECTOR_SIZE
;
497 } else if (!strcmp(options
->name
, BLOCK_OPT_PREALLOC
)) {
498 if (!options
->value
.s
|| !strcmp(options
->value
.s
, "off")) {
500 } else if (!strcmp(options
->value
.s
, "full") &&
501 gluster_supports_zerofill()) {
504 error_setg(errp
, "Invalid preallocation mode: '%s'"
505 " 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) {
533 qemu_gluster_gconf_free(gconf
);
540 static coroutine_fn
int qemu_gluster_co_rw(BlockDriverState
*bs
,
541 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
, int write
)
544 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
545 BDRVGlusterState
*s
= bs
->opaque
;
546 size_t size
= nb_sectors
* BDRV_SECTOR_SIZE
;
547 off_t offset
= sector_num
* BDRV_SECTOR_SIZE
;
551 acb
->coroutine
= qemu_coroutine_self();
554 ret
= glfs_pwritev_async(s
->fd
, qiov
->iov
, qiov
->niov
, offset
, 0,
555 &gluster_finish_aiocb
, acb
);
557 ret
= glfs_preadv_async(s
->fd
, qiov
->iov
, qiov
->niov
, offset
, 0,
558 &gluster_finish_aiocb
, acb
);
566 qemu_coroutine_yield();
570 g_slice_free(GlusterAIOCB
, acb
);
574 static int qemu_gluster_truncate(BlockDriverState
*bs
, int64_t offset
)
577 BDRVGlusterState
*s
= bs
->opaque
;
579 ret
= glfs_ftruncate(s
->fd
, offset
);
587 static coroutine_fn
int qemu_gluster_co_readv(BlockDriverState
*bs
,
588 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
590 return qemu_gluster_co_rw(bs
, sector_num
, nb_sectors
, qiov
, 0);
593 static coroutine_fn
int qemu_gluster_co_writev(BlockDriverState
*bs
,
594 int64_t sector_num
, int nb_sectors
, QEMUIOVector
*qiov
)
596 return qemu_gluster_co_rw(bs
, sector_num
, nb_sectors
, qiov
, 1);
599 static coroutine_fn
int qemu_gluster_co_flush_to_disk(BlockDriverState
*bs
)
602 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
603 BDRVGlusterState
*s
= bs
->opaque
;
607 acb
->coroutine
= qemu_coroutine_self();
609 ret
= glfs_fsync_async(s
->fd
, &gluster_finish_aiocb
, acb
);
615 qemu_coroutine_yield();
619 g_slice_free(GlusterAIOCB
, acb
);
623 #ifdef CONFIG_GLUSTERFS_DISCARD
624 static coroutine_fn
int qemu_gluster_co_discard(BlockDriverState
*bs
,
625 int64_t sector_num
, int nb_sectors
)
628 GlusterAIOCB
*acb
= g_slice_new(GlusterAIOCB
);
629 BDRVGlusterState
*s
= bs
->opaque
;
630 size_t size
= nb_sectors
* BDRV_SECTOR_SIZE
;
631 off_t offset
= sector_num
* BDRV_SECTOR_SIZE
;
635 acb
->coroutine
= qemu_coroutine_self();
637 ret
= glfs_discard_async(s
->fd
, offset
, size
, &gluster_finish_aiocb
, acb
);
643 qemu_coroutine_yield();
647 g_slice_free(GlusterAIOCB
, acb
);
652 static int64_t qemu_gluster_getlength(BlockDriverState
*bs
)
654 BDRVGlusterState
*s
= bs
->opaque
;
657 ret
= glfs_lseek(s
->fd
, 0, SEEK_END
);
665 static int64_t qemu_gluster_allocated_file_size(BlockDriverState
*bs
)
667 BDRVGlusterState
*s
= bs
->opaque
;
671 ret
= glfs_fstat(s
->fd
, &st
);
675 return st
.st_blocks
* 512;
679 static void qemu_gluster_close(BlockDriverState
*bs
)
681 BDRVGlusterState
*s
= bs
->opaque
;
690 static int qemu_gluster_has_zero_init(BlockDriverState
*bs
)
692 /* GlusterFS volume could be backed by a block device */
696 static QEMUOptionParameter qemu_gluster_create_options
[] = {
698 .name
= BLOCK_OPT_SIZE
,
700 .help
= "Virtual disk size"
703 .name
= BLOCK_OPT_PREALLOC
,
705 .help
= "Preallocation mode (allowed values: off, full)"
710 static BlockDriver bdrv_gluster
= {
711 .format_name
= "gluster",
712 .protocol_name
= "gluster",
713 .instance_size
= sizeof(BDRVGlusterState
),
714 .bdrv_needs_filename
= true,
715 .bdrv_file_open
= qemu_gluster_open
,
716 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
717 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
718 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
719 .bdrv_close
= qemu_gluster_close
,
720 .bdrv_create
= qemu_gluster_create
,
721 .bdrv_getlength
= qemu_gluster_getlength
,
722 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
723 .bdrv_truncate
= qemu_gluster_truncate
,
724 .bdrv_co_readv
= qemu_gluster_co_readv
,
725 .bdrv_co_writev
= qemu_gluster_co_writev
,
726 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
727 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
728 #ifdef CONFIG_GLUSTERFS_DISCARD
729 .bdrv_co_discard
= qemu_gluster_co_discard
,
731 #ifdef CONFIG_GLUSTERFS_ZEROFILL
732 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
734 .create_options
= qemu_gluster_create_options
,
737 static BlockDriver bdrv_gluster_tcp
= {
738 .format_name
= "gluster",
739 .protocol_name
= "gluster+tcp",
740 .instance_size
= sizeof(BDRVGlusterState
),
741 .bdrv_needs_filename
= true,
742 .bdrv_file_open
= qemu_gluster_open
,
743 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
744 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
745 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
746 .bdrv_close
= qemu_gluster_close
,
747 .bdrv_create
= qemu_gluster_create
,
748 .bdrv_getlength
= qemu_gluster_getlength
,
749 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
750 .bdrv_truncate
= qemu_gluster_truncate
,
751 .bdrv_co_readv
= qemu_gluster_co_readv
,
752 .bdrv_co_writev
= qemu_gluster_co_writev
,
753 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
754 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
755 #ifdef CONFIG_GLUSTERFS_DISCARD
756 .bdrv_co_discard
= qemu_gluster_co_discard
,
758 #ifdef CONFIG_GLUSTERFS_ZEROFILL
759 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
761 .create_options
= qemu_gluster_create_options
,
764 static BlockDriver bdrv_gluster_unix
= {
765 .format_name
= "gluster",
766 .protocol_name
= "gluster+unix",
767 .instance_size
= sizeof(BDRVGlusterState
),
768 .bdrv_needs_filename
= true,
769 .bdrv_file_open
= qemu_gluster_open
,
770 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
771 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
772 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
773 .bdrv_close
= qemu_gluster_close
,
774 .bdrv_create
= qemu_gluster_create
,
775 .bdrv_getlength
= qemu_gluster_getlength
,
776 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
777 .bdrv_truncate
= qemu_gluster_truncate
,
778 .bdrv_co_readv
= qemu_gluster_co_readv
,
779 .bdrv_co_writev
= qemu_gluster_co_writev
,
780 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
781 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
782 #ifdef CONFIG_GLUSTERFS_DISCARD
783 .bdrv_co_discard
= qemu_gluster_co_discard
,
785 #ifdef CONFIG_GLUSTERFS_ZEROFILL
786 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
788 .create_options
= qemu_gluster_create_options
,
791 static BlockDriver bdrv_gluster_rdma
= {
792 .format_name
= "gluster",
793 .protocol_name
= "gluster+rdma",
794 .instance_size
= sizeof(BDRVGlusterState
),
795 .bdrv_needs_filename
= true,
796 .bdrv_file_open
= qemu_gluster_open
,
797 .bdrv_reopen_prepare
= qemu_gluster_reopen_prepare
,
798 .bdrv_reopen_commit
= qemu_gluster_reopen_commit
,
799 .bdrv_reopen_abort
= qemu_gluster_reopen_abort
,
800 .bdrv_close
= qemu_gluster_close
,
801 .bdrv_create
= qemu_gluster_create
,
802 .bdrv_getlength
= qemu_gluster_getlength
,
803 .bdrv_get_allocated_file_size
= qemu_gluster_allocated_file_size
,
804 .bdrv_truncate
= qemu_gluster_truncate
,
805 .bdrv_co_readv
= qemu_gluster_co_readv
,
806 .bdrv_co_writev
= qemu_gluster_co_writev
,
807 .bdrv_co_flush_to_disk
= qemu_gluster_co_flush_to_disk
,
808 .bdrv_has_zero_init
= qemu_gluster_has_zero_init
,
809 #ifdef CONFIG_GLUSTERFS_DISCARD
810 .bdrv_co_discard
= qemu_gluster_co_discard
,
812 #ifdef CONFIG_GLUSTERFS_ZEROFILL
813 .bdrv_co_write_zeroes
= qemu_gluster_co_write_zeroes
,
815 .create_options
= qemu_gluster_create_options
,
818 static void bdrv_gluster_init(void)
820 bdrv_register(&bdrv_gluster_rdma
);
821 bdrv_register(&bdrv_gluster_unix
);
822 bdrv_register(&bdrv_gluster_tcp
);
823 bdrv_register(&bdrv_gluster
);
826 block_init(bdrv_gluster_init
);