]>
Commit | Line | Data |
---|---|---|
8d6d89cb BR |
1 | /* |
2 | * GlusterFS backend for QEMU | |
3 | * | |
4 | * Copyright (C) 2012 Bharata B Rao <bharata@linux.vnet.ibm.com> | |
5 | * | |
85c09bc0 BR |
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. | |
8d6d89cb | 8 | * |
8d6d89cb | 9 | */ |
80c71a24 | 10 | #include "qemu/osdep.h" |
8d6d89cb | 11 | #include <glusterfs/api/glfs.h> |
737e150e | 12 | #include "block/block_int.h" |
da34e65c | 13 | #include "qapi/error.h" |
1de7afc9 | 14 | #include "qemu/uri.h" |
0552ff24 | 15 | #include "qemu/error-report.h" |
8d6d89cb | 16 | |
f70c50c8 PKK |
17 | #define GLUSTER_OPT_FILENAME "filename" |
18 | #define GLUSTER_OPT_DEBUG "debug" | |
7edac2dd | 19 | #define GLUSTER_DEFAULT_PORT 24007 |
f70c50c8 PKK |
20 | #define GLUSTER_DEBUG_DEFAULT 4 |
21 | #define GLUSTER_DEBUG_MAX 9 | |
22 | ||
23 | ||
8d6d89cb | 24 | typedef struct GlusterAIOCB { |
8d6d89cb BR |
25 | int64_t size; |
26 | int ret; | |
8d6d89cb | 27 | QEMUBH *bh; |
15744b0b | 28 | Coroutine *coroutine; |
6ee50af2 | 29 | AioContext *aio_context; |
8d6d89cb BR |
30 | } GlusterAIOCB; |
31 | ||
32 | typedef struct BDRVGlusterState { | |
33 | struct glfs *glfs; | |
8d6d89cb | 34 | struct glfs_fd *fd; |
947eb203 | 35 | bool supports_seek_data; |
7eac868a | 36 | int debug_level; |
8d6d89cb BR |
37 | } BDRVGlusterState; |
38 | ||
f70c50c8 PKK |
39 | typedef struct BDRVGlusterReopenState { |
40 | struct glfs *glfs; | |
41 | struct glfs_fd *fd; | |
42 | } BDRVGlusterReopenState; | |
43 | ||
f70c50c8 PKK |
44 | |
45 | static QemuOptsList qemu_gluster_create_opts = { | |
46 | .name = "qemu-gluster-create-opts", | |
47 | .head = QTAILQ_HEAD_INITIALIZER(qemu_gluster_create_opts.head), | |
48 | .desc = { | |
49 | { | |
50 | .name = BLOCK_OPT_SIZE, | |
51 | .type = QEMU_OPT_SIZE, | |
52 | .help = "Virtual disk size" | |
53 | }, | |
54 | { | |
55 | .name = BLOCK_OPT_PREALLOC, | |
56 | .type = QEMU_OPT_STRING, | |
57 | .help = "Preallocation mode (allowed values: off, full)" | |
58 | }, | |
59 | { | |
60 | .name = GLUSTER_OPT_DEBUG, | |
61 | .type = QEMU_OPT_NUMBER, | |
62 | .help = "Gluster log level, valid range is 0-9", | |
63 | }, | |
64 | { /* end of list */ } | |
65 | } | |
66 | }; | |
67 | ||
68 | static QemuOptsList runtime_opts = { | |
69 | .name = "gluster", | |
70 | .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head), | |
71 | .desc = { | |
72 | { | |
73 | .name = GLUSTER_OPT_FILENAME, | |
74 | .type = QEMU_OPT_STRING, | |
75 | .help = "URL to the gluster image", | |
76 | }, | |
77 | { | |
78 | .name = GLUSTER_OPT_DEBUG, | |
79 | .type = QEMU_OPT_NUMBER, | |
80 | .help = "Gluster log level, valid range is 0-9", | |
81 | }, | |
82 | { /* end of list */ } | |
83 | }, | |
84 | }; | |
85 | ||
86 | ||
7edac2dd | 87 | static int parse_volume_options(BlockdevOptionsGluster *gconf, char *path) |
8d6d89cb BR |
88 | { |
89 | char *p, *q; | |
90 | ||
91 | if (!path) { | |
92 | return -EINVAL; | |
93 | } | |
94 | ||
95 | /* volume */ | |
96 | p = q = path + strspn(path, "/"); | |
97 | p += strcspn(p, "/"); | |
98 | if (*p == '\0') { | |
99 | return -EINVAL; | |
100 | } | |
d5cf4079 | 101 | gconf->volume = g_strndup(q, p - q); |
8d6d89cb | 102 | |
d5cf4079 | 103 | /* path */ |
8d6d89cb BR |
104 | p += strspn(p, "/"); |
105 | if (*p == '\0') { | |
106 | return -EINVAL; | |
107 | } | |
d5cf4079 | 108 | gconf->path = g_strdup(p); |
8d6d89cb BR |
109 | return 0; |
110 | } | |
111 | ||
112 | /* | |
d5cf4079 | 113 | * file=gluster[+transport]://[host[:port]]/volume/path[?socket=...] |
8d6d89cb BR |
114 | * |
115 | * 'gluster' is the protocol. | |
116 | * | |
117 | * 'transport' specifies the transport type used to connect to gluster | |
118 | * management daemon (glusterd). Valid transport types are | |
0552ff24 | 119 | * tcp or unix. If a transport type isn't specified, then tcp type is assumed. |
8d6d89cb | 120 | * |
d5cf4079 | 121 | * 'host' specifies the host where the volume file specification for |
0552ff24 | 122 | * the given volume resides. This can be either hostname or ipv4 address. |
d5cf4079 | 123 | * If transport type is 'unix', then 'host' field should not be specified. |
8d6d89cb BR |
124 | * The 'socket' field needs to be populated with the path to unix domain |
125 | * socket. | |
126 | * | |
127 | * 'port' is the port number on which glusterd is listening. This is optional | |
128 | * and if not specified, QEMU will send 0 which will make gluster to use the | |
129 | * default port. If the transport type is unix, then 'port' should not be | |
130 | * specified. | |
131 | * | |
d5cf4079 | 132 | * 'volume' is the name of the gluster volume which contains the VM image. |
8d6d89cb | 133 | * |
d5cf4079 | 134 | * 'path' is the path to the actual VM image that resides on gluster volume. |
8d6d89cb BR |
135 | * |
136 | * Examples: | |
137 | * | |
138 | * file=gluster://1.2.3.4/testvol/a.img | |
139 | * file=gluster+tcp://1.2.3.4/testvol/a.img | |
140 | * file=gluster+tcp://1.2.3.4:24007/testvol/dir/a.img | |
d5cf4079 | 141 | * file=gluster+tcp://host.domain.com:24007/testvol/dir/a.img |
8d6d89cb | 142 | * file=gluster+unix:///testvol/dir/a.img?socket=/tmp/glusterd.socket |
8d6d89cb | 143 | */ |
7edac2dd PKK |
144 | static int qemu_gluster_parse_uri(BlockdevOptionsGluster *gconf, |
145 | const char *filename) | |
8d6d89cb | 146 | { |
7edac2dd | 147 | GlusterServer *gsconf; |
8d6d89cb BR |
148 | URI *uri; |
149 | QueryParams *qp = NULL; | |
150 | bool is_unix = false; | |
151 | int ret = 0; | |
152 | ||
153 | uri = uri_parse(filename); | |
154 | if (!uri) { | |
155 | return -EINVAL; | |
156 | } | |
157 | ||
7edac2dd PKK |
158 | gconf->server = gsconf = g_new0(GlusterServer, 1); |
159 | ||
8d6d89cb | 160 | /* transport */ |
24897a76 | 161 | if (!uri->scheme || !strcmp(uri->scheme, "gluster")) { |
7edac2dd | 162 | gsconf->type = GLUSTER_TRANSPORT_TCP; |
8d6d89cb | 163 | } else if (!strcmp(uri->scheme, "gluster+tcp")) { |
7edac2dd | 164 | gsconf->type = GLUSTER_TRANSPORT_TCP; |
8d6d89cb | 165 | } else if (!strcmp(uri->scheme, "gluster+unix")) { |
7edac2dd | 166 | gsconf->type = GLUSTER_TRANSPORT_UNIX; |
8d6d89cb BR |
167 | is_unix = true; |
168 | } else if (!strcmp(uri->scheme, "gluster+rdma")) { | |
7edac2dd | 169 | gsconf->type = GLUSTER_TRANSPORT_TCP; |
0552ff24 PKK |
170 | error_report("Warning: rdma feature is not supported, falling " |
171 | "back to tcp"); | |
8d6d89cb BR |
172 | } else { |
173 | ret = -EINVAL; | |
174 | goto out; | |
175 | } | |
176 | ||
177 | ret = parse_volume_options(gconf, uri->path); | |
178 | if (ret < 0) { | |
179 | goto out; | |
180 | } | |
181 | ||
182 | qp = query_params_parse(uri->query); | |
183 | if (qp->n > 1 || (is_unix && !qp->n) || (!is_unix && qp->n)) { | |
184 | ret = -EINVAL; | |
185 | goto out; | |
186 | } | |
187 | ||
188 | if (is_unix) { | |
189 | if (uri->server || uri->port) { | |
190 | ret = -EINVAL; | |
191 | goto out; | |
192 | } | |
193 | if (strcmp(qp->p[0].name, "socket")) { | |
194 | ret = -EINVAL; | |
195 | goto out; | |
196 | } | |
7edac2dd | 197 | gsconf->u.q_unix.path = g_strdup(qp->p[0].value); |
8d6d89cb | 198 | } else { |
7edac2dd PKK |
199 | gsconf->u.tcp.host = g_strdup(uri->server ? uri->server : "localhost"); |
200 | if (uri->port) { | |
201 | gsconf->u.tcp.port = g_strdup_printf("%d", uri->port); | |
202 | } else { | |
203 | gsconf->u.tcp.port = g_strdup_printf("%d", GLUSTER_DEFAULT_PORT); | |
204 | } | |
8d6d89cb BR |
205 | } |
206 | ||
207 | out: | |
208 | if (qp) { | |
209 | query_params_free(qp); | |
210 | } | |
211 | uri_free(uri); | |
212 | return ret; | |
213 | } | |
214 | ||
7edac2dd PKK |
215 | static struct glfs *qemu_gluster_init(BlockdevOptionsGluster *gconf, |
216 | const char *filename, Error **errp) | |
8d6d89cb BR |
217 | { |
218 | struct glfs *glfs = NULL; | |
219 | int ret; | |
220 | int old_errno; | |
221 | ||
7edac2dd | 222 | ret = qemu_gluster_parse_uri(gconf, filename); |
8d6d89cb | 223 | if (ret < 0) { |
7edac2dd PKK |
224 | error_setg(errp, "Invalid URI"); |
225 | error_append_hint(errp, "Usage: file=gluster[+transport]://" | |
226 | "[host[:port]]/volume/path[?socket=...]\n"); | |
8d6d89cb BR |
227 | errno = -ret; |
228 | goto out; | |
229 | } | |
230 | ||
d5cf4079 | 231 | glfs = glfs_new(gconf->volume); |
8d6d89cb BR |
232 | if (!glfs) { |
233 | goto out; | |
234 | } | |
235 | ||
7edac2dd PKK |
236 | if (gconf->server->type == GLUSTER_TRANSPORT_UNIX) { |
237 | ret = glfs_set_volfile_server(glfs, | |
238 | GlusterTransport_lookup[gconf->server->type], | |
239 | gconf->server->u.q_unix.path, 0); | |
240 | } else { | |
241 | ret = glfs_set_volfile_server(glfs, | |
242 | GlusterTransport_lookup[gconf->server->type], | |
243 | gconf->server->u.tcp.host, | |
244 | atoi(gconf->server->u.tcp.port)); | |
245 | } | |
8d6d89cb BR |
246 | if (ret < 0) { |
247 | goto out; | |
248 | } | |
249 | ||
7eac868a | 250 | ret = glfs_set_logging(glfs, "-", gconf->debug_level); |
8d6d89cb BR |
251 | if (ret < 0) { |
252 | goto out; | |
253 | } | |
254 | ||
255 | ret = glfs_init(glfs); | |
256 | if (ret) { | |
7edac2dd PKK |
257 | if (gconf->server->type == GLUSTER_TRANSPORT_UNIX) { |
258 | error_setg(errp, | |
259 | "Gluster connection for volume %s, path %s failed on " | |
260 | "socket %s ", gconf->volume, gconf->path, | |
261 | gconf->server->u.q_unix.path); | |
262 | } else { | |
263 | error_setg(errp, | |
264 | "Gluster connection for volume %s, path %s failed on " | |
265 | "host %s and port %s ", gconf->volume, gconf->path, | |
266 | gconf->server->u.tcp.host, gconf->server->u.tcp.port); | |
267 | } | |
4557117d PK |
268 | |
269 | /* glfs_init sometimes doesn't set errno although docs suggest that */ | |
7edac2dd | 270 | if (errno == 0) { |
4557117d | 271 | errno = EINVAL; |
7edac2dd | 272 | } |
4557117d | 273 | |
8d6d89cb BR |
274 | goto out; |
275 | } | |
276 | return glfs; | |
277 | ||
278 | out: | |
279 | if (glfs) { | |
280 | old_errno = errno; | |
281 | glfs_fini(glfs); | |
282 | errno = old_errno; | |
283 | } | |
284 | return NULL; | |
285 | } | |
286 | ||
15744b0b | 287 | static void qemu_gluster_complete_aio(void *opaque) |
8d6d89cb | 288 | { |
15744b0b | 289 | GlusterAIOCB *acb = (GlusterAIOCB *)opaque; |
8d6d89cb | 290 | |
15744b0b BR |
291 | qemu_bh_delete(acb->bh); |
292 | acb->bh = NULL; | |
0b8b8753 | 293 | qemu_coroutine_enter(acb->coroutine); |
8d6d89cb BR |
294 | } |
295 | ||
7c815372 BR |
296 | /* |
297 | * AIO callback routine called from GlusterFS thread. | |
298 | */ | |
299 | static void gluster_finish_aiocb(struct glfs_fd *fd, ssize_t ret, void *arg) | |
300 | { | |
301 | GlusterAIOCB *acb = (GlusterAIOCB *)arg; | |
302 | ||
303 | if (!ret || ret == acb->size) { | |
304 | acb->ret = 0; /* Success */ | |
305 | } else if (ret < 0) { | |
a8827453 | 306 | acb->ret = -errno; /* Read/Write failed */ |
7c815372 BR |
307 | } else { |
308 | acb->ret = -EIO; /* Partial read/write - fail it */ | |
309 | } | |
310 | ||
6ee50af2 | 311 | acb->bh = aio_bh_new(acb->aio_context, qemu_gluster_complete_aio, acb); |
7c815372 BR |
312 | qemu_bh_schedule(acb->bh); |
313 | } | |
314 | ||
1b37b344 JC |
315 | static void qemu_gluster_parse_flags(int bdrv_flags, int *open_flags) |
316 | { | |
317 | assert(open_flags != NULL); | |
318 | ||
319 | *open_flags |= O_BINARY; | |
320 | ||
321 | if (bdrv_flags & BDRV_O_RDWR) { | |
322 | *open_flags |= O_RDWR; | |
323 | } else { | |
324 | *open_flags |= O_RDONLY; | |
325 | } | |
326 | ||
327 | if ((bdrv_flags & BDRV_O_NOCACHE)) { | |
328 | *open_flags |= O_DIRECT; | |
329 | } | |
330 | } | |
331 | ||
947eb203 NV |
332 | /* |
333 | * Do SEEK_DATA/HOLE to detect if it is functional. Older broken versions of | |
334 | * gfapi incorrectly return the current offset when SEEK_DATA/HOLE is used. | |
335 | * - Corrected versions return -1 and set errno to EINVAL. | |
336 | * - Versions that support SEEK_DATA/HOLE correctly, will return -1 and set | |
337 | * errno to ENXIO when SEEK_DATA is called with a position of EOF. | |
338 | */ | |
339 | static bool qemu_gluster_test_seek(struct glfs_fd *fd) | |
340 | { | |
341 | off_t ret, eof; | |
342 | ||
343 | eof = glfs_lseek(fd, 0, SEEK_END); | |
344 | if (eof < 0) { | |
345 | /* this should never occur */ | |
346 | return false; | |
347 | } | |
348 | ||
349 | /* this should always fail with ENXIO if SEEK_DATA is supported */ | |
350 | ret = glfs_lseek(fd, eof, SEEK_DATA); | |
351 | return (ret < 0) && (errno == ENXIO); | |
352 | } | |
353 | ||
56d1b4d2 | 354 | static int qemu_gluster_open(BlockDriverState *bs, QDict *options, |
015a1036 | 355 | int bdrv_flags, Error **errp) |
8d6d89cb BR |
356 | { |
357 | BDRVGlusterState *s = bs->opaque; | |
1b37b344 | 358 | int open_flags = 0; |
8d6d89cb | 359 | int ret = 0; |
7edac2dd | 360 | BlockdevOptionsGluster *gconf = NULL; |
b4894776 KW |
361 | QemuOpts *opts; |
362 | Error *local_err = NULL; | |
363 | const char *filename; | |
364 | ||
87ea75d5 | 365 | opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort); |
b4894776 | 366 | qemu_opts_absorb_qdict(opts, options, &local_err); |
84d18f06 | 367 | if (local_err) { |
a7451cb8 | 368 | error_propagate(errp, local_err); |
b4894776 KW |
369 | ret = -EINVAL; |
370 | goto out; | |
371 | } | |
372 | ||
7eac868a JC |
373 | filename = qemu_opt_get(opts, GLUSTER_OPT_FILENAME); |
374 | ||
375 | s->debug_level = qemu_opt_get_number(opts, GLUSTER_OPT_DEBUG, | |
376 | GLUSTER_DEBUG_DEFAULT); | |
377 | if (s->debug_level < 0) { | |
378 | s->debug_level = 0; | |
379 | } else if (s->debug_level > GLUSTER_DEBUG_MAX) { | |
380 | s->debug_level = GLUSTER_DEBUG_MAX; | |
381 | } | |
b4894776 | 382 | |
7edac2dd | 383 | gconf = g_new0(BlockdevOptionsGluster, 1); |
7eac868a | 384 | gconf->debug_level = s->debug_level; |
7edac2dd | 385 | gconf->has_debug_level = true; |
a7451cb8 | 386 | s->glfs = qemu_gluster_init(gconf, filename, errp); |
8d6d89cb BR |
387 | if (!s->glfs) { |
388 | ret = -errno; | |
389 | goto out; | |
390 | } | |
391 | ||
d85fa9eb JC |
392 | #ifdef CONFIG_GLUSTERFS_XLATOR_OPT |
393 | /* Without this, if fsync fails for a recoverable reason (for instance, | |
394 | * ENOSPC), gluster will dump its cache, preventing retries. This means | |
395 | * almost certain data loss. Not all gluster versions support the | |
396 | * 'resync-failed-syncs-after-fsync' key value, but there is no way to | |
397 | * discover during runtime if it is supported (this api returns success for | |
398 | * unknown key/value pairs) */ | |
399 | ret = glfs_set_xlator_option(s->glfs, "*-write-behind", | |
400 | "resync-failed-syncs-after-fsync", | |
401 | "on"); | |
402 | if (ret < 0) { | |
403 | error_setg_errno(errp, errno, "Unable to set xlator key/value pair"); | |
404 | ret = -errno; | |
405 | goto out; | |
406 | } | |
407 | #endif | |
408 | ||
1b37b344 | 409 | qemu_gluster_parse_flags(bdrv_flags, &open_flags); |
8d6d89cb | 410 | |
d5cf4079 | 411 | s->fd = glfs_open(s->glfs, gconf->path, open_flags); |
8d6d89cb BR |
412 | if (!s->fd) { |
413 | ret = -errno; | |
8d6d89cb | 414 | } |
8d6d89cb | 415 | |
947eb203 NV |
416 | s->supports_seek_data = qemu_gluster_test_seek(s->fd); |
417 | ||
8d6d89cb | 418 | out: |
b4894776 | 419 | qemu_opts_del(opts); |
7edac2dd | 420 | qapi_free_BlockdevOptionsGluster(gconf); |
8d6d89cb BR |
421 | if (!ret) { |
422 | return ret; | |
423 | } | |
424 | if (s->fd) { | |
425 | glfs_close(s->fd); | |
426 | } | |
427 | if (s->glfs) { | |
428 | glfs_fini(s->glfs); | |
429 | } | |
430 | return ret; | |
431 | } | |
432 | ||
adccfbcd JC |
433 | static int qemu_gluster_reopen_prepare(BDRVReopenState *state, |
434 | BlockReopenQueue *queue, Error **errp) | |
435 | { | |
436 | int ret = 0; | |
7eac868a | 437 | BDRVGlusterState *s; |
adccfbcd | 438 | BDRVGlusterReopenState *reop_s; |
7edac2dd | 439 | BlockdevOptionsGluster *gconf; |
adccfbcd JC |
440 | int open_flags = 0; |
441 | ||
442 | assert(state != NULL); | |
443 | assert(state->bs != NULL); | |
444 | ||
7eac868a JC |
445 | s = state->bs->opaque; |
446 | ||
5839e53b | 447 | state->opaque = g_new0(BDRVGlusterReopenState, 1); |
adccfbcd JC |
448 | reop_s = state->opaque; |
449 | ||
450 | qemu_gluster_parse_flags(state->flags, &open_flags); | |
451 | ||
7edac2dd | 452 | gconf = g_new0(BlockdevOptionsGluster, 1); |
7eac868a | 453 | gconf->debug_level = s->debug_level; |
7edac2dd | 454 | gconf->has_debug_level = true; |
f55ea629 | 455 | reop_s->glfs = qemu_gluster_init(gconf, state->bs->filename, errp); |
adccfbcd JC |
456 | if (reop_s->glfs == NULL) { |
457 | ret = -errno; | |
458 | goto exit; | |
459 | } | |
460 | ||
d85fa9eb JC |
461 | #ifdef CONFIG_GLUSTERFS_XLATOR_OPT |
462 | ret = glfs_set_xlator_option(reop_s->glfs, "*-write-behind", | |
463 | "resync-failed-syncs-after-fsync", "on"); | |
464 | if (ret < 0) { | |
465 | error_setg_errno(errp, errno, "Unable to set xlator key/value pair"); | |
466 | ret = -errno; | |
467 | goto exit; | |
468 | } | |
469 | #endif | |
470 | ||
d5cf4079 | 471 | reop_s->fd = glfs_open(reop_s->glfs, gconf->path, open_flags); |
adccfbcd JC |
472 | if (reop_s->fd == NULL) { |
473 | /* reops->glfs will be cleaned up in _abort */ | |
474 | ret = -errno; | |
475 | goto exit; | |
476 | } | |
477 | ||
478 | exit: | |
479 | /* state->opaque will be freed in either the _abort or _commit */ | |
7edac2dd | 480 | qapi_free_BlockdevOptionsGluster(gconf); |
adccfbcd JC |
481 | return ret; |
482 | } | |
483 | ||
484 | static void qemu_gluster_reopen_commit(BDRVReopenState *state) | |
485 | { | |
486 | BDRVGlusterReopenState *reop_s = state->opaque; | |
487 | BDRVGlusterState *s = state->bs->opaque; | |
488 | ||
489 | ||
490 | /* close the old */ | |
491 | if (s->fd) { | |
492 | glfs_close(s->fd); | |
493 | } | |
494 | if (s->glfs) { | |
495 | glfs_fini(s->glfs); | |
496 | } | |
497 | ||
498 | /* use the newly opened image / connection */ | |
499 | s->fd = reop_s->fd; | |
500 | s->glfs = reop_s->glfs; | |
501 | ||
502 | g_free(state->opaque); | |
503 | state->opaque = NULL; | |
504 | ||
505 | return; | |
506 | } | |
507 | ||
508 | ||
509 | static void qemu_gluster_reopen_abort(BDRVReopenState *state) | |
510 | { | |
511 | BDRVGlusterReopenState *reop_s = state->opaque; | |
512 | ||
513 | if (reop_s == NULL) { | |
514 | return; | |
515 | } | |
516 | ||
517 | if (reop_s->fd) { | |
518 | glfs_close(reop_s->fd); | |
519 | } | |
520 | ||
521 | if (reop_s->glfs) { | |
522 | glfs_fini(reop_s->glfs); | |
523 | } | |
524 | ||
525 | g_free(state->opaque); | |
526 | state->opaque = NULL; | |
527 | ||
528 | return; | |
529 | } | |
530 | ||
7c815372 | 531 | #ifdef CONFIG_GLUSTERFS_ZEROFILL |
e88a36eb | 532 | static coroutine_fn int qemu_gluster_co_pwrite_zeroes(BlockDriverState *bs, |
f70c50c8 PKK |
533 | int64_t offset, |
534 | int size, | |
535 | BdrvRequestFlags flags) | |
7c815372 BR |
536 | { |
537 | int ret; | |
c833d1e8 | 538 | GlusterAIOCB acb; |
7c815372 | 539 | BDRVGlusterState *s = bs->opaque; |
7c815372 | 540 | |
c833d1e8 PB |
541 | acb.size = size; |
542 | acb.ret = 0; | |
543 | acb.coroutine = qemu_coroutine_self(); | |
544 | acb.aio_context = bdrv_get_aio_context(bs); | |
7c815372 | 545 | |
c833d1e8 | 546 | ret = glfs_zerofill_async(s->fd, offset, size, gluster_finish_aiocb, &acb); |
7c815372 | 547 | if (ret < 0) { |
c833d1e8 | 548 | return -errno; |
7c815372 BR |
549 | } |
550 | ||
551 | qemu_coroutine_yield(); | |
c833d1e8 | 552 | return acb.ret; |
7c815372 | 553 | } |
cf7f616b BR |
554 | |
555 | static inline bool gluster_supports_zerofill(void) | |
556 | { | |
557 | return 1; | |
558 | } | |
559 | ||
560 | static inline int qemu_gluster_zerofill(struct glfs_fd *fd, int64_t offset, | |
f70c50c8 | 561 | int64_t size) |
cf7f616b BR |
562 | { |
563 | return glfs_zerofill(fd, offset, size); | |
564 | } | |
565 | ||
566 | #else | |
567 | static inline bool gluster_supports_zerofill(void) | |
568 | { | |
569 | return 0; | |
570 | } | |
571 | ||
572 | static inline int qemu_gluster_zerofill(struct glfs_fd *fd, int64_t offset, | |
f70c50c8 | 573 | int64_t size) |
cf7f616b BR |
574 | { |
575 | return 0; | |
576 | } | |
7c815372 BR |
577 | #endif |
578 | ||
8d6d89cb | 579 | static int qemu_gluster_create(const char *filename, |
90c772de | 580 | QemuOpts *opts, Error **errp) |
8d6d89cb | 581 | { |
7edac2dd | 582 | BlockdevOptionsGluster *gconf; |
8d6d89cb BR |
583 | struct glfs *glfs; |
584 | struct glfs_fd *fd; | |
585 | int ret = 0; | |
cf7f616b | 586 | int prealloc = 0; |
8d6d89cb | 587 | int64_t total_size = 0; |
90c772de | 588 | char *tmp = NULL; |
8d6d89cb | 589 | |
7edac2dd | 590 | gconf = g_new0(BlockdevOptionsGluster, 1); |
7eac868a JC |
591 | gconf->debug_level = qemu_opt_get_number_del(opts, GLUSTER_OPT_DEBUG, |
592 | GLUSTER_DEBUG_DEFAULT); | |
593 | if (gconf->debug_level < 0) { | |
594 | gconf->debug_level = 0; | |
595 | } else if (gconf->debug_level > GLUSTER_DEBUG_MAX) { | |
596 | gconf->debug_level = GLUSTER_DEBUG_MAX; | |
597 | } | |
7edac2dd | 598 | gconf->has_debug_level = true; |
7eac868a | 599 | |
a7451cb8 | 600 | glfs = qemu_gluster_init(gconf, filename, errp); |
8d6d89cb | 601 | if (!glfs) { |
4557117d | 602 | ret = -errno; |
8d6d89cb BR |
603 | goto out; |
604 | } | |
605 | ||
180e9526 HT |
606 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
607 | BDRV_SECTOR_SIZE); | |
90c772de CL |
608 | |
609 | tmp = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
610 | if (!tmp || !strcmp(tmp, "off")) { | |
611 | prealloc = 0; | |
f70c50c8 | 612 | } else if (!strcmp(tmp, "full") && gluster_supports_zerofill()) { |
90c772de CL |
613 | prealloc = 1; |
614 | } else { | |
615 | error_setg(errp, "Invalid preallocation mode: '%s'" | |
f70c50c8 | 616 | " or GlusterFS doesn't support zerofill API", tmp); |
90c772de CL |
617 | ret = -EINVAL; |
618 | goto out; | |
8d6d89cb BR |
619 | } |
620 | ||
d5cf4079 | 621 | fd = glfs_creat(glfs, gconf->path, |
f70c50c8 | 622 | O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, S_IRUSR | S_IWUSR); |
8d6d89cb BR |
623 | if (!fd) { |
624 | ret = -errno; | |
625 | } else { | |
180e9526 HT |
626 | if (!glfs_ftruncate(fd, total_size)) { |
627 | if (prealloc && qemu_gluster_zerofill(fd, 0, total_size)) { | |
cf7f616b BR |
628 | ret = -errno; |
629 | } | |
630 | } else { | |
8d6d89cb BR |
631 | ret = -errno; |
632 | } | |
cf7f616b | 633 | |
8d6d89cb BR |
634 | if (glfs_close(fd) != 0) { |
635 | ret = -errno; | |
636 | } | |
637 | } | |
638 | out: | |
90c772de | 639 | g_free(tmp); |
7edac2dd | 640 | qapi_free_BlockdevOptionsGluster(gconf); |
8d6d89cb BR |
641 | if (glfs) { |
642 | glfs_fini(glfs); | |
643 | } | |
644 | return ret; | |
645 | } | |
646 | ||
15744b0b | 647 | static coroutine_fn int qemu_gluster_co_rw(BlockDriverState *bs, |
f70c50c8 PKK |
648 | int64_t sector_num, int nb_sectors, |
649 | QEMUIOVector *qiov, int write) | |
8d6d89cb BR |
650 | { |
651 | int ret; | |
c833d1e8 | 652 | GlusterAIOCB acb; |
8d6d89cb | 653 | BDRVGlusterState *s = bs->opaque; |
15744b0b BR |
654 | size_t size = nb_sectors * BDRV_SECTOR_SIZE; |
655 | off_t offset = sector_num * BDRV_SECTOR_SIZE; | |
8d6d89cb | 656 | |
c833d1e8 PB |
657 | acb.size = size; |
658 | acb.ret = 0; | |
659 | acb.coroutine = qemu_coroutine_self(); | |
660 | acb.aio_context = bdrv_get_aio_context(bs); | |
8d6d89cb BR |
661 | |
662 | if (write) { | |
663 | ret = glfs_pwritev_async(s->fd, qiov->iov, qiov->niov, offset, 0, | |
f70c50c8 | 664 | gluster_finish_aiocb, &acb); |
8d6d89cb BR |
665 | } else { |
666 | ret = glfs_preadv_async(s->fd, qiov->iov, qiov->niov, offset, 0, | |
f70c50c8 | 667 | gluster_finish_aiocb, &acb); |
8d6d89cb BR |
668 | } |
669 | ||
670 | if (ret < 0) { | |
c833d1e8 | 671 | return -errno; |
8d6d89cb | 672 | } |
15744b0b BR |
673 | |
674 | qemu_coroutine_yield(); | |
c833d1e8 | 675 | return acb.ret; |
8d6d89cb BR |
676 | } |
677 | ||
42ec24e2 PB |
678 | static int qemu_gluster_truncate(BlockDriverState *bs, int64_t offset) |
679 | { | |
680 | int ret; | |
681 | BDRVGlusterState *s = bs->opaque; | |
682 | ||
683 | ret = glfs_ftruncate(s->fd, offset); | |
684 | if (ret < 0) { | |
685 | return -errno; | |
686 | } | |
687 | ||
688 | return 0; | |
689 | } | |
690 | ||
15744b0b | 691 | static coroutine_fn int qemu_gluster_co_readv(BlockDriverState *bs, |
f70c50c8 PKK |
692 | int64_t sector_num, |
693 | int nb_sectors, | |
694 | QEMUIOVector *qiov) | |
8d6d89cb | 695 | { |
15744b0b | 696 | return qemu_gluster_co_rw(bs, sector_num, nb_sectors, qiov, 0); |
8d6d89cb BR |
697 | } |
698 | ||
15744b0b | 699 | static coroutine_fn int qemu_gluster_co_writev(BlockDriverState *bs, |
f70c50c8 PKK |
700 | int64_t sector_num, |
701 | int nb_sectors, | |
702 | QEMUIOVector *qiov) | |
8d6d89cb | 703 | { |
15744b0b | 704 | return qemu_gluster_co_rw(bs, sector_num, nb_sectors, qiov, 1); |
8d6d89cb BR |
705 | } |
706 | ||
5d4343e6 JC |
707 | static void qemu_gluster_close(BlockDriverState *bs) |
708 | { | |
709 | BDRVGlusterState *s = bs->opaque; | |
710 | ||
711 | if (s->fd) { | |
712 | glfs_close(s->fd); | |
713 | s->fd = NULL; | |
714 | } | |
715 | glfs_fini(s->glfs); | |
716 | } | |
717 | ||
15744b0b | 718 | static coroutine_fn int qemu_gluster_co_flush_to_disk(BlockDriverState *bs) |
8d6d89cb BR |
719 | { |
720 | int ret; | |
c833d1e8 | 721 | GlusterAIOCB acb; |
8d6d89cb BR |
722 | BDRVGlusterState *s = bs->opaque; |
723 | ||
c833d1e8 PB |
724 | acb.size = 0; |
725 | acb.ret = 0; | |
726 | acb.coroutine = qemu_coroutine_self(); | |
727 | acb.aio_context = bdrv_get_aio_context(bs); | |
8d6d89cb | 728 | |
c833d1e8 | 729 | ret = glfs_fsync_async(s->fd, gluster_finish_aiocb, &acb); |
8d6d89cb | 730 | if (ret < 0) { |
d85fa9eb JC |
731 | ret = -errno; |
732 | goto error; | |
8d6d89cb | 733 | } |
15744b0b BR |
734 | |
735 | qemu_coroutine_yield(); | |
d85fa9eb JC |
736 | if (acb.ret < 0) { |
737 | ret = acb.ret; | |
738 | goto error; | |
739 | } | |
740 | ||
c833d1e8 | 741 | return acb.ret; |
d85fa9eb JC |
742 | |
743 | error: | |
744 | /* Some versions of Gluster (3.5.6 -> 3.5.8?) will not retain its cache | |
745 | * after a fsync failure, so we have no way of allowing the guest to safely | |
746 | * continue. Gluster versions prior to 3.5.6 don't retain the cache | |
747 | * either, but will invalidate the fd on error, so this is again our only | |
748 | * option. | |
749 | * | |
750 | * The 'resync-failed-syncs-after-fsync' xlator option for the | |
751 | * write-behind cache will cause later gluster versions to retain its | |
752 | * cache after error, so long as the fd remains open. However, we | |
753 | * currently have no way of knowing if this option is supported. | |
754 | * | |
755 | * TODO: Once gluster provides a way for us to determine if the option | |
756 | * is supported, bypass the closure and setting drv to NULL. */ | |
757 | qemu_gluster_close(bs); | |
758 | bs->drv = NULL; | |
759 | return ret; | |
8d6d89cb BR |
760 | } |
761 | ||
0c14fb47 | 762 | #ifdef CONFIG_GLUSTERFS_DISCARD |
15744b0b | 763 | static coroutine_fn int qemu_gluster_co_discard(BlockDriverState *bs, |
f70c50c8 PKK |
764 | int64_t sector_num, |
765 | int nb_sectors) | |
0c14fb47 BR |
766 | { |
767 | int ret; | |
c833d1e8 | 768 | GlusterAIOCB acb; |
0c14fb47 | 769 | BDRVGlusterState *s = bs->opaque; |
15744b0b BR |
770 | size_t size = nb_sectors * BDRV_SECTOR_SIZE; |
771 | off_t offset = sector_num * BDRV_SECTOR_SIZE; | |
0c14fb47 | 772 | |
c833d1e8 PB |
773 | acb.size = 0; |
774 | acb.ret = 0; | |
775 | acb.coroutine = qemu_coroutine_self(); | |
776 | acb.aio_context = bdrv_get_aio_context(bs); | |
0c14fb47 | 777 | |
c833d1e8 | 778 | ret = glfs_discard_async(s->fd, offset, size, gluster_finish_aiocb, &acb); |
0c14fb47 | 779 | if (ret < 0) { |
c833d1e8 | 780 | return -errno; |
0c14fb47 | 781 | } |
15744b0b BR |
782 | |
783 | qemu_coroutine_yield(); | |
c833d1e8 | 784 | return acb.ret; |
0c14fb47 BR |
785 | } |
786 | #endif | |
787 | ||
8d6d89cb BR |
788 | static int64_t qemu_gluster_getlength(BlockDriverState *bs) |
789 | { | |
790 | BDRVGlusterState *s = bs->opaque; | |
791 | int64_t ret; | |
792 | ||
793 | ret = glfs_lseek(s->fd, 0, SEEK_END); | |
794 | if (ret < 0) { | |
795 | return -errno; | |
796 | } else { | |
797 | return ret; | |
798 | } | |
799 | } | |
800 | ||
801 | static int64_t qemu_gluster_allocated_file_size(BlockDriverState *bs) | |
802 | { | |
803 | BDRVGlusterState *s = bs->opaque; | |
804 | struct stat st; | |
805 | int ret; | |
806 | ||
807 | ret = glfs_fstat(s->fd, &st); | |
808 | if (ret < 0) { | |
809 | return -errno; | |
810 | } else { | |
811 | return st.st_blocks * 512; | |
812 | } | |
813 | } | |
814 | ||
8ab6feec KW |
815 | static int qemu_gluster_has_zero_init(BlockDriverState *bs) |
816 | { | |
817 | /* GlusterFS volume could be backed by a block device */ | |
818 | return 0; | |
819 | } | |
820 | ||
947eb203 NV |
821 | /* |
822 | * Find allocation range in @bs around offset @start. | |
823 | * May change underlying file descriptor's file offset. | |
824 | * If @start is not in a hole, store @start in @data, and the | |
825 | * beginning of the next hole in @hole, and return 0. | |
826 | * If @start is in a non-trailing hole, store @start in @hole and the | |
827 | * beginning of the next non-hole in @data, and return 0. | |
828 | * If @start is in a trailing hole or beyond EOF, return -ENXIO. | |
829 | * If we can't find out, return a negative errno other than -ENXIO. | |
830 | * | |
831 | * (Shamefully copied from raw-posix.c, only miniscule adaptions.) | |
832 | */ | |
833 | static int find_allocation(BlockDriverState *bs, off_t start, | |
834 | off_t *data, off_t *hole) | |
835 | { | |
836 | BDRVGlusterState *s = bs->opaque; | |
837 | off_t offs; | |
838 | ||
839 | if (!s->supports_seek_data) { | |
840 | return -ENOTSUP; | |
841 | } | |
842 | ||
843 | /* | |
844 | * SEEK_DATA cases: | |
845 | * D1. offs == start: start is in data | |
846 | * D2. offs > start: start is in a hole, next data at offs | |
847 | * D3. offs < 0, errno = ENXIO: either start is in a trailing hole | |
848 | * or start is beyond EOF | |
849 | * If the latter happens, the file has been truncated behind | |
850 | * our back since we opened it. All bets are off then. | |
851 | * Treating like a trailing hole is simplest. | |
852 | * D4. offs < 0, errno != ENXIO: we learned nothing | |
853 | */ | |
854 | offs = glfs_lseek(s->fd, start, SEEK_DATA); | |
855 | if (offs < 0) { | |
856 | return -errno; /* D3 or D4 */ | |
857 | } | |
858 | assert(offs >= start); | |
859 | ||
860 | if (offs > start) { | |
861 | /* D2: in hole, next data at offs */ | |
862 | *hole = start; | |
863 | *data = offs; | |
864 | return 0; | |
865 | } | |
866 | ||
867 | /* D1: in data, end not yet known */ | |
868 | ||
869 | /* | |
870 | * SEEK_HOLE cases: | |
871 | * H1. offs == start: start is in a hole | |
872 | * If this happens here, a hole has been dug behind our back | |
873 | * since the previous lseek(). | |
874 | * H2. offs > start: either start is in data, next hole at offs, | |
875 | * or start is in trailing hole, EOF at offs | |
876 | * Linux treats trailing holes like any other hole: offs == | |
877 | * start. Solaris seeks to EOF instead: offs > start (blech). | |
878 | * If that happens here, a hole has been dug behind our back | |
879 | * since the previous lseek(). | |
880 | * H3. offs < 0, errno = ENXIO: start is beyond EOF | |
881 | * If this happens, the file has been truncated behind our | |
882 | * back since we opened it. Treat it like a trailing hole. | |
883 | * H4. offs < 0, errno != ENXIO: we learned nothing | |
884 | * Pretend we know nothing at all, i.e. "forget" about D1. | |
885 | */ | |
886 | offs = glfs_lseek(s->fd, start, SEEK_HOLE); | |
887 | if (offs < 0) { | |
888 | return -errno; /* D1 and (H3 or H4) */ | |
889 | } | |
890 | assert(offs >= start); | |
891 | ||
892 | if (offs > start) { | |
893 | /* | |
894 | * D1 and H2: either in data, next hole at offs, or it was in | |
895 | * data but is now in a trailing hole. In the latter case, | |
896 | * all bets are off. Treating it as if it there was data all | |
897 | * the way to EOF is safe, so simply do that. | |
898 | */ | |
899 | *data = start; | |
900 | *hole = offs; | |
901 | return 0; | |
902 | } | |
903 | ||
904 | /* D1 and H1 */ | |
905 | return -EBUSY; | |
906 | } | |
907 | ||
908 | /* | |
909 | * Returns the allocation status of the specified sectors. | |
910 | * | |
911 | * If 'sector_num' is beyond the end of the disk image the return value is 0 | |
912 | * and 'pnum' is set to 0. | |
913 | * | |
914 | * 'pnum' is set to the number of sectors (including and immediately following | |
915 | * the specified sector) that are known to be in the same | |
916 | * allocated/unallocated state. | |
917 | * | |
918 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes | |
919 | * beyond the end of the disk image it will be clamped. | |
920 | * | |
921 | * (Based on raw_co_get_block_status() from raw-posix.c.) | |
922 | */ | |
923 | static int64_t coroutine_fn qemu_gluster_co_get_block_status( | |
924 | BlockDriverState *bs, int64_t sector_num, int nb_sectors, int *pnum, | |
925 | BlockDriverState **file) | |
926 | { | |
927 | BDRVGlusterState *s = bs->opaque; | |
928 | off_t start, data = 0, hole = 0; | |
929 | int64_t total_size; | |
930 | int ret = -EINVAL; | |
931 | ||
932 | if (!s->fd) { | |
933 | return ret; | |
934 | } | |
935 | ||
936 | start = sector_num * BDRV_SECTOR_SIZE; | |
937 | total_size = bdrv_getlength(bs); | |
938 | if (total_size < 0) { | |
939 | return total_size; | |
940 | } else if (start >= total_size) { | |
941 | *pnum = 0; | |
942 | return 0; | |
943 | } else if (start + nb_sectors * BDRV_SECTOR_SIZE > total_size) { | |
944 | nb_sectors = DIV_ROUND_UP(total_size - start, BDRV_SECTOR_SIZE); | |
945 | } | |
946 | ||
947 | ret = find_allocation(bs, start, &data, &hole); | |
948 | if (ret == -ENXIO) { | |
949 | /* Trailing hole */ | |
950 | *pnum = nb_sectors; | |
951 | ret = BDRV_BLOCK_ZERO; | |
952 | } else if (ret < 0) { | |
953 | /* No info available, so pretend there are no holes */ | |
954 | *pnum = nb_sectors; | |
955 | ret = BDRV_BLOCK_DATA; | |
956 | } else if (data == start) { | |
957 | /* On a data extent, compute sectors to the end of the extent, | |
958 | * possibly including a partial sector at EOF. */ | |
959 | *pnum = MIN(nb_sectors, DIV_ROUND_UP(hole - start, BDRV_SECTOR_SIZE)); | |
960 | ret = BDRV_BLOCK_DATA; | |
961 | } else { | |
962 | /* On a hole, compute sectors to the beginning of the next extent. */ | |
963 | assert(hole == start); | |
964 | *pnum = MIN(nb_sectors, (data - start) / BDRV_SECTOR_SIZE); | |
965 | ret = BDRV_BLOCK_ZERO; | |
966 | } | |
967 | ||
968 | *file = bs; | |
969 | ||
970 | return ret | BDRV_BLOCK_OFFSET_VALID | start; | |
971 | } | |
972 | ||
973 | ||
8d6d89cb BR |
974 | static BlockDriver bdrv_gluster = { |
975 | .format_name = "gluster", | |
976 | .protocol_name = "gluster", | |
977 | .instance_size = sizeof(BDRVGlusterState), | |
030be321 | 978 | .bdrv_needs_filename = true, |
8d6d89cb | 979 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
980 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
981 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
982 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 983 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 984 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
985 | .bdrv_getlength = qemu_gluster_getlength, |
986 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 987 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
988 | .bdrv_co_readv = qemu_gluster_co_readv, |
989 | .bdrv_co_writev = qemu_gluster_co_writev, | |
990 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 991 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 992 | #ifdef CONFIG_GLUSTERFS_DISCARD |
15744b0b | 993 | .bdrv_co_discard = qemu_gluster_co_discard, |
7c815372 BR |
994 | #endif |
995 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 996 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 997 | #endif |
947eb203 | 998 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 999 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1000 | }; |
1001 | ||
1002 | static BlockDriver bdrv_gluster_tcp = { | |
1003 | .format_name = "gluster", | |
1004 | .protocol_name = "gluster+tcp", | |
1005 | .instance_size = sizeof(BDRVGlusterState), | |
030be321 | 1006 | .bdrv_needs_filename = true, |
8d6d89cb | 1007 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1008 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1009 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1010 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1011 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1012 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1013 | .bdrv_getlength = qemu_gluster_getlength, |
1014 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1015 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1016 | .bdrv_co_readv = qemu_gluster_co_readv, |
1017 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1018 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1019 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1020 | #ifdef CONFIG_GLUSTERFS_DISCARD |
15744b0b | 1021 | .bdrv_co_discard = qemu_gluster_co_discard, |
7c815372 BR |
1022 | #endif |
1023 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1024 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1025 | #endif |
947eb203 | 1026 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1027 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1028 | }; |
1029 | ||
1030 | static BlockDriver bdrv_gluster_unix = { | |
1031 | .format_name = "gluster", | |
1032 | .protocol_name = "gluster+unix", | |
1033 | .instance_size = sizeof(BDRVGlusterState), | |
030be321 | 1034 | .bdrv_needs_filename = true, |
8d6d89cb | 1035 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1036 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1037 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1038 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1039 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1040 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1041 | .bdrv_getlength = qemu_gluster_getlength, |
1042 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1043 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1044 | .bdrv_co_readv = qemu_gluster_co_readv, |
1045 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1046 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1047 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1048 | #ifdef CONFIG_GLUSTERFS_DISCARD |
15744b0b | 1049 | .bdrv_co_discard = qemu_gluster_co_discard, |
7c815372 BR |
1050 | #endif |
1051 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1052 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1053 | #endif |
947eb203 | 1054 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1055 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1056 | }; |
1057 | ||
0552ff24 PKK |
1058 | /* rdma is deprecated (actually never supported for volfile fetch). |
1059 | * Let's maintain it for the protocol compatibility, to make sure things | |
1060 | * won't break immediately. For now, gluster+rdma will fall back to gluster+tcp | |
1061 | * protocol with a warning. | |
1062 | * TODO: remove gluster+rdma interface support | |
1063 | */ | |
8d6d89cb BR |
1064 | static BlockDriver bdrv_gluster_rdma = { |
1065 | .format_name = "gluster", | |
1066 | .protocol_name = "gluster+rdma", | |
1067 | .instance_size = sizeof(BDRVGlusterState), | |
030be321 | 1068 | .bdrv_needs_filename = true, |
8d6d89cb | 1069 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1070 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1071 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1072 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1073 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1074 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1075 | .bdrv_getlength = qemu_gluster_getlength, |
1076 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1077 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1078 | .bdrv_co_readv = qemu_gluster_co_readv, |
1079 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1080 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1081 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1082 | #ifdef CONFIG_GLUSTERFS_DISCARD |
15744b0b | 1083 | .bdrv_co_discard = qemu_gluster_co_discard, |
7c815372 BR |
1084 | #endif |
1085 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1086 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1087 | #endif |
947eb203 | 1088 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1089 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1090 | }; |
1091 | ||
1092 | static void bdrv_gluster_init(void) | |
1093 | { | |
1094 | bdrv_register(&bdrv_gluster_rdma); | |
1095 | bdrv_register(&bdrv_gluster_unix); | |
1096 | bdrv_register(&bdrv_gluster_tcp); | |
1097 | bdrv_register(&bdrv_gluster); | |
1098 | } | |
1099 | ||
1100 | block_init(bdrv_gluster_init); |