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