]>
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" |
6c7189bb | 14 | #include "qapi/qmp/qerror.h" |
e152ef7a | 15 | #include "qapi/util.h" |
1de7afc9 | 16 | #include "qemu/uri.h" |
0552ff24 | 17 | #include "qemu/error-report.h" |
c56ac33b | 18 | #include "qemu/cutils.h" |
8d6d89cb | 19 | |
f70c50c8 | 20 | #define GLUSTER_OPT_FILENAME "filename" |
6c7189bb PKK |
21 | #define GLUSTER_OPT_VOLUME "volume" |
22 | #define GLUSTER_OPT_PATH "path" | |
23 | #define GLUSTER_OPT_TYPE "type" | |
24 | #define GLUSTER_OPT_SERVER_PATTERN "server." | |
25 | #define GLUSTER_OPT_HOST "host" | |
26 | #define GLUSTER_OPT_PORT "port" | |
27 | #define GLUSTER_OPT_TO "to" | |
28 | #define GLUSTER_OPT_IPV4 "ipv4" | |
29 | #define GLUSTER_OPT_IPV6 "ipv6" | |
30 | #define GLUSTER_OPT_SOCKET "socket" | |
f70c50c8 | 31 | #define GLUSTER_OPT_DEBUG "debug" |
7edac2dd | 32 | #define GLUSTER_DEFAULT_PORT 24007 |
f70c50c8 PKK |
33 | #define GLUSTER_DEBUG_DEFAULT 4 |
34 | #define GLUSTER_DEBUG_MAX 9 | |
e9db8ff3 PKK |
35 | #define GLUSTER_OPT_LOGFILE "logfile" |
36 | #define GLUSTER_LOGFILE_DEFAULT "-" /* handled in libgfapi as /dev/stderr */ | |
f70c50c8 | 37 | |
6c7189bb | 38 | #define GERR_INDEX_HINT "hint: check in 'server' array index '%d'\n" |
f70c50c8 | 39 | |
8d6d89cb | 40 | typedef struct GlusterAIOCB { |
8d6d89cb BR |
41 | int64_t size; |
42 | int ret; | |
15744b0b | 43 | Coroutine *coroutine; |
6ee50af2 | 44 | AioContext *aio_context; |
8d6d89cb BR |
45 | } GlusterAIOCB; |
46 | ||
47 | typedef struct BDRVGlusterState { | |
48 | struct glfs *glfs; | |
8d6d89cb | 49 | struct glfs_fd *fd; |
e9db8ff3 | 50 | char *logfile; |
947eb203 | 51 | bool supports_seek_data; |
1a417e46 | 52 | int debug; |
8d6d89cb BR |
53 | } BDRVGlusterState; |
54 | ||
f70c50c8 PKK |
55 | typedef struct BDRVGlusterReopenState { |
56 | struct glfs *glfs; | |
57 | struct glfs_fd *fd; | |
58 | } BDRVGlusterReopenState; | |
59 | ||
f70c50c8 | 60 | |
6349c154 PKK |
61 | typedef struct GlfsPreopened { |
62 | char *volume; | |
63 | glfs_t *fs; | |
64 | int ref; | |
65 | } GlfsPreopened; | |
66 | ||
67 | typedef struct ListElement { | |
68 | QLIST_ENTRY(ListElement) list; | |
69 | GlfsPreopened saved; | |
70 | } ListElement; | |
71 | ||
72 | static QLIST_HEAD(glfs_list, ListElement) glfs_list; | |
73 | ||
f70c50c8 PKK |
74 | static QemuOptsList qemu_gluster_create_opts = { |
75 | .name = "qemu-gluster-create-opts", | |
76 | .head = QTAILQ_HEAD_INITIALIZER(qemu_gluster_create_opts.head), | |
77 | .desc = { | |
78 | { | |
79 | .name = BLOCK_OPT_SIZE, | |
80 | .type = QEMU_OPT_SIZE, | |
81 | .help = "Virtual disk size" | |
82 | }, | |
83 | { | |
84 | .name = BLOCK_OPT_PREALLOC, | |
85 | .type = QEMU_OPT_STRING, | |
86 | .help = "Preallocation mode (allowed values: off, full)" | |
87 | }, | |
88 | { | |
89 | .name = GLUSTER_OPT_DEBUG, | |
90 | .type = QEMU_OPT_NUMBER, | |
91 | .help = "Gluster log level, valid range is 0-9", | |
92 | }, | |
e9db8ff3 PKK |
93 | { |
94 | .name = GLUSTER_OPT_LOGFILE, | |
95 | .type = QEMU_OPT_STRING, | |
96 | .help = "Logfile path of libgfapi", | |
97 | }, | |
f70c50c8 PKK |
98 | { /* end of list */ } |
99 | } | |
100 | }; | |
101 | ||
102 | static QemuOptsList runtime_opts = { | |
103 | .name = "gluster", | |
104 | .head = QTAILQ_HEAD_INITIALIZER(runtime_opts.head), | |
105 | .desc = { | |
106 | { | |
107 | .name = GLUSTER_OPT_FILENAME, | |
108 | .type = QEMU_OPT_STRING, | |
109 | .help = "URL to the gluster image", | |
110 | }, | |
111 | { | |
112 | .name = GLUSTER_OPT_DEBUG, | |
113 | .type = QEMU_OPT_NUMBER, | |
114 | .help = "Gluster log level, valid range is 0-9", | |
115 | }, | |
e9db8ff3 PKK |
116 | { |
117 | .name = GLUSTER_OPT_LOGFILE, | |
118 | .type = QEMU_OPT_STRING, | |
119 | .help = "Logfile path of libgfapi", | |
120 | }, | |
f70c50c8 PKK |
121 | { /* end of list */ } |
122 | }, | |
123 | }; | |
124 | ||
6c7189bb PKK |
125 | static QemuOptsList runtime_json_opts = { |
126 | .name = "gluster_json", | |
127 | .head = QTAILQ_HEAD_INITIALIZER(runtime_json_opts.head), | |
128 | .desc = { | |
129 | { | |
130 | .name = GLUSTER_OPT_VOLUME, | |
131 | .type = QEMU_OPT_STRING, | |
132 | .help = "name of gluster volume where VM image resides", | |
133 | }, | |
134 | { | |
135 | .name = GLUSTER_OPT_PATH, | |
136 | .type = QEMU_OPT_STRING, | |
137 | .help = "absolute path to image file in gluster volume", | |
138 | }, | |
139 | { | |
140 | .name = GLUSTER_OPT_DEBUG, | |
141 | .type = QEMU_OPT_NUMBER, | |
142 | .help = "Gluster log level, valid range is 0-9", | |
143 | }, | |
144 | { /* end of list */ } | |
145 | }, | |
146 | }; | |
147 | ||
148 | static QemuOptsList runtime_type_opts = { | |
149 | .name = "gluster_type", | |
150 | .head = QTAILQ_HEAD_INITIALIZER(runtime_type_opts.head), | |
151 | .desc = { | |
152 | { | |
153 | .name = GLUSTER_OPT_TYPE, | |
154 | .type = QEMU_OPT_STRING, | |
c5f1ae3a | 155 | .help = "inet|unix", |
6c7189bb PKK |
156 | }, |
157 | { /* end of list */ } | |
158 | }, | |
159 | }; | |
160 | ||
161 | static QemuOptsList runtime_unix_opts = { | |
162 | .name = "gluster_unix", | |
163 | .head = QTAILQ_HEAD_INITIALIZER(runtime_unix_opts.head), | |
164 | .desc = { | |
165 | { | |
166 | .name = GLUSTER_OPT_SOCKET, | |
167 | .type = QEMU_OPT_STRING, | |
168 | .help = "socket file path)", | |
169 | }, | |
170 | { /* end of list */ } | |
171 | }, | |
172 | }; | |
173 | ||
c5f1ae3a MA |
174 | static QemuOptsList runtime_inet_opts = { |
175 | .name = "gluster_inet", | |
176 | .head = QTAILQ_HEAD_INITIALIZER(runtime_inet_opts.head), | |
6c7189bb PKK |
177 | .desc = { |
178 | { | |
179 | .name = GLUSTER_OPT_TYPE, | |
180 | .type = QEMU_OPT_STRING, | |
c5f1ae3a | 181 | .help = "inet|unix", |
6c7189bb PKK |
182 | }, |
183 | { | |
184 | .name = GLUSTER_OPT_HOST, | |
185 | .type = QEMU_OPT_STRING, | |
186 | .help = "host address (hostname/ipv4/ipv6 addresses)", | |
187 | }, | |
188 | { | |
189 | .name = GLUSTER_OPT_PORT, | |
53d9837f | 190 | .type = QEMU_OPT_STRING, |
6c7189bb PKK |
191 | .help = "port number on which glusterd is listening (default 24007)", |
192 | }, | |
193 | { | |
194 | .name = "to", | |
195 | .type = QEMU_OPT_NUMBER, | |
196 | .help = "max port number, not supported by gluster", | |
197 | }, | |
198 | { | |
199 | .name = "ipv4", | |
200 | .type = QEMU_OPT_BOOL, | |
201 | .help = "ipv4 bool value, not supported by gluster", | |
202 | }, | |
203 | { | |
204 | .name = "ipv6", | |
205 | .type = QEMU_OPT_BOOL, | |
206 | .help = "ipv6 bool value, not supported by gluster", | |
207 | }, | |
208 | { /* end of list */ } | |
209 | }, | |
210 | }; | |
f70c50c8 | 211 | |
6349c154 PKK |
212 | static void glfs_set_preopened(const char *volume, glfs_t *fs) |
213 | { | |
214 | ListElement *entry = NULL; | |
215 | ||
216 | entry = g_new(ListElement, 1); | |
217 | ||
218 | entry->saved.volume = g_strdup(volume); | |
219 | ||
220 | entry->saved.fs = fs; | |
221 | entry->saved.ref = 1; | |
222 | ||
223 | QLIST_INSERT_HEAD(&glfs_list, entry, list); | |
224 | } | |
225 | ||
226 | static glfs_t *glfs_find_preopened(const char *volume) | |
227 | { | |
228 | ListElement *entry = NULL; | |
229 | ||
230 | QLIST_FOREACH(entry, &glfs_list, list) { | |
231 | if (strcmp(entry->saved.volume, volume) == 0) { | |
232 | entry->saved.ref++; | |
233 | return entry->saved.fs; | |
234 | } | |
235 | } | |
236 | ||
237 | return NULL; | |
238 | } | |
239 | ||
240 | static void glfs_clear_preopened(glfs_t *fs) | |
241 | { | |
242 | ListElement *entry = NULL; | |
668c0e44 | 243 | ListElement *next; |
6349c154 PKK |
244 | |
245 | if (fs == NULL) { | |
246 | return; | |
247 | } | |
248 | ||
668c0e44 | 249 | QLIST_FOREACH_SAFE(entry, &glfs_list, list, next) { |
6349c154 PKK |
250 | if (entry->saved.fs == fs) { |
251 | if (--entry->saved.ref) { | |
252 | return; | |
253 | } | |
254 | ||
255 | QLIST_REMOVE(entry, list); | |
256 | ||
257 | glfs_fini(entry->saved.fs); | |
258 | g_free(entry->saved.volume); | |
259 | g_free(entry); | |
260 | } | |
261 | } | |
262 | } | |
263 | ||
7edac2dd | 264 | static int parse_volume_options(BlockdevOptionsGluster *gconf, char *path) |
8d6d89cb BR |
265 | { |
266 | char *p, *q; | |
267 | ||
268 | if (!path) { | |
269 | return -EINVAL; | |
270 | } | |
271 | ||
272 | /* volume */ | |
273 | p = q = path + strspn(path, "/"); | |
274 | p += strcspn(p, "/"); | |
275 | if (*p == '\0') { | |
276 | return -EINVAL; | |
277 | } | |
d5cf4079 | 278 | gconf->volume = g_strndup(q, p - q); |
8d6d89cb | 279 | |
d5cf4079 | 280 | /* path */ |
8d6d89cb BR |
281 | p += strspn(p, "/"); |
282 | if (*p == '\0') { | |
283 | return -EINVAL; | |
284 | } | |
d5cf4079 | 285 | gconf->path = g_strdup(p); |
8d6d89cb BR |
286 | return 0; |
287 | } | |
288 | ||
289 | /* | |
d5cf4079 | 290 | * file=gluster[+transport]://[host[:port]]/volume/path[?socket=...] |
8d6d89cb BR |
291 | * |
292 | * 'gluster' is the protocol. | |
293 | * | |
294 | * 'transport' specifies the transport type used to connect to gluster | |
295 | * management daemon (glusterd). Valid transport types are | |
0552ff24 | 296 | * tcp or unix. If a transport type isn't specified, then tcp type is assumed. |
8d6d89cb | 297 | * |
d5cf4079 | 298 | * 'host' specifies the host where the volume file specification for |
0552ff24 | 299 | * the given volume resides. This can be either hostname or ipv4 address. |
d5cf4079 | 300 | * If transport type is 'unix', then 'host' field should not be specified. |
8d6d89cb BR |
301 | * The 'socket' field needs to be populated with the path to unix domain |
302 | * socket. | |
303 | * | |
304 | * 'port' is the port number on which glusterd is listening. This is optional | |
305 | * and if not specified, QEMU will send 0 which will make gluster to use the | |
306 | * default port. If the transport type is unix, then 'port' should not be | |
307 | * specified. | |
308 | * | |
d5cf4079 | 309 | * 'volume' is the name of the gluster volume which contains the VM image. |
8d6d89cb | 310 | * |
d5cf4079 | 311 | * 'path' is the path to the actual VM image that resides on gluster volume. |
8d6d89cb BR |
312 | * |
313 | * Examples: | |
314 | * | |
315 | * file=gluster://1.2.3.4/testvol/a.img | |
316 | * file=gluster+tcp://1.2.3.4/testvol/a.img | |
317 | * file=gluster+tcp://1.2.3.4:24007/testvol/dir/a.img | |
d5cf4079 | 318 | * file=gluster+tcp://host.domain.com:24007/testvol/dir/a.img |
8d6d89cb | 319 | * file=gluster+unix:///testvol/dir/a.img?socket=/tmp/glusterd.socket |
8d6d89cb | 320 | */ |
7edac2dd PKK |
321 | static int qemu_gluster_parse_uri(BlockdevOptionsGluster *gconf, |
322 | const char *filename) | |
8d6d89cb | 323 | { |
62cf396b | 324 | SocketAddress *gsconf; |
8d6d89cb BR |
325 | URI *uri; |
326 | QueryParams *qp = NULL; | |
327 | bool is_unix = false; | |
328 | int ret = 0; | |
329 | ||
330 | uri = uri_parse(filename); | |
331 | if (!uri) { | |
332 | return -EINVAL; | |
333 | } | |
334 | ||
62cf396b MA |
335 | gconf->server = g_new0(SocketAddressList, 1); |
336 | gconf->server->value = gsconf = g_new0(SocketAddress, 1); | |
7edac2dd | 337 | |
8d6d89cb | 338 | /* transport */ |
24897a76 | 339 | if (!uri->scheme || !strcmp(uri->scheme, "gluster")) { |
62cf396b | 340 | gsconf->type = SOCKET_ADDRESS_TYPE_INET; |
8d6d89cb | 341 | } else if (!strcmp(uri->scheme, "gluster+tcp")) { |
62cf396b | 342 | gsconf->type = SOCKET_ADDRESS_TYPE_INET; |
8d6d89cb | 343 | } else if (!strcmp(uri->scheme, "gluster+unix")) { |
62cf396b | 344 | gsconf->type = SOCKET_ADDRESS_TYPE_UNIX; |
8d6d89cb BR |
345 | is_unix = true; |
346 | } else if (!strcmp(uri->scheme, "gluster+rdma")) { | |
62cf396b | 347 | gsconf->type = SOCKET_ADDRESS_TYPE_INET; |
3dc6f869 | 348 | warn_report("rdma feature is not supported, falling back to tcp"); |
8d6d89cb BR |
349 | } else { |
350 | ret = -EINVAL; | |
351 | goto out; | |
352 | } | |
353 | ||
354 | ret = parse_volume_options(gconf, uri->path); | |
355 | if (ret < 0) { | |
356 | goto out; | |
357 | } | |
358 | ||
359 | qp = query_params_parse(uri->query); | |
360 | if (qp->n > 1 || (is_unix && !qp->n) || (!is_unix && qp->n)) { | |
361 | ret = -EINVAL; | |
362 | goto out; | |
363 | } | |
364 | ||
365 | if (is_unix) { | |
366 | if (uri->server || uri->port) { | |
367 | ret = -EINVAL; | |
368 | goto out; | |
369 | } | |
370 | if (strcmp(qp->p[0].name, "socket")) { | |
371 | ret = -EINVAL; | |
372 | goto out; | |
373 | } | |
7edac2dd | 374 | gsconf->u.q_unix.path = g_strdup(qp->p[0].value); |
8d6d89cb | 375 | } else { |
c5f1ae3a | 376 | gsconf->u.inet.host = g_strdup(uri->server ? uri->server : "localhost"); |
7edac2dd | 377 | if (uri->port) { |
c5f1ae3a | 378 | gsconf->u.inet.port = g_strdup_printf("%d", uri->port); |
7edac2dd | 379 | } else { |
c5f1ae3a | 380 | gsconf->u.inet.port = g_strdup_printf("%d", GLUSTER_DEFAULT_PORT); |
7edac2dd | 381 | } |
8d6d89cb BR |
382 | } |
383 | ||
384 | out: | |
385 | if (qp) { | |
386 | query_params_free(qp); | |
387 | } | |
388 | uri_free(uri); | |
389 | return ret; | |
390 | } | |
391 | ||
6c7189bb PKK |
392 | static struct glfs *qemu_gluster_glfs_init(BlockdevOptionsGluster *gconf, |
393 | Error **errp) | |
8d6d89cb | 394 | { |
6c7189bb | 395 | struct glfs *glfs; |
8d6d89cb BR |
396 | int ret; |
397 | int old_errno; | |
62cf396b | 398 | SocketAddressList *server; |
c56ac33b | 399 | unsigned long long port; |
8d6d89cb | 400 | |
6349c154 PKK |
401 | glfs = glfs_find_preopened(gconf->volume); |
402 | if (glfs) { | |
403 | return glfs; | |
404 | } | |
405 | ||
d5cf4079 | 406 | glfs = glfs_new(gconf->volume); |
8d6d89cb BR |
407 | if (!glfs) { |
408 | goto out; | |
409 | } | |
410 | ||
6349c154 PKK |
411 | glfs_set_preopened(gconf->volume, glfs); |
412 | ||
6c7189bb | 413 | for (server = gconf->server; server; server = server->next) { |
fce5d538 | 414 | switch (server->value->type) { |
62cf396b | 415 | case SOCKET_ADDRESS_TYPE_UNIX: |
fc29458d | 416 | ret = glfs_set_volfile_server(glfs, "unix", |
6c7189bb | 417 | server->value->u.q_unix.path, 0); |
fce5d538 | 418 | break; |
62cf396b | 419 | case SOCKET_ADDRESS_TYPE_INET: |
c5f1ae3a | 420 | if (parse_uint_full(server->value->u.inet.port, &port, 10) < 0 || |
c56ac33b PKK |
421 | port > 65535) { |
422 | error_setg(errp, "'%s' is not a valid port number", | |
c5f1ae3a | 423 | server->value->u.inet.port); |
c56ac33b PKK |
424 | errno = EINVAL; |
425 | goto out; | |
426 | } | |
fc29458d | 427 | ret = glfs_set_volfile_server(glfs, "tcp", |
c5f1ae3a | 428 | server->value->u.inet.host, |
c56ac33b | 429 | (int)port); |
fce5d538 | 430 | break; |
62cf396b MA |
431 | case SOCKET_ADDRESS_TYPE_VSOCK: |
432 | case SOCKET_ADDRESS_TYPE_FD: | |
fce5d538 MA |
433 | default: |
434 | abort(); | |
6c7189bb PKK |
435 | } |
436 | ||
437 | if (ret < 0) { | |
438 | goto out; | |
439 | } | |
8d6d89cb BR |
440 | } |
441 | ||
1a417e46 | 442 | ret = glfs_set_logging(glfs, gconf->logfile, gconf->debug); |
8d6d89cb BR |
443 | if (ret < 0) { |
444 | goto out; | |
445 | } | |
446 | ||
447 | ret = glfs_init(glfs); | |
448 | if (ret) { | |
6c7189bb PKK |
449 | error_setg(errp, "Gluster connection for volume %s, path %s failed" |
450 | " to connect", gconf->volume, gconf->path); | |
451 | for (server = gconf->server; server; server = server->next) { | |
62cf396b | 452 | if (server->value->type == SOCKET_ADDRESS_TYPE_UNIX) { |
6c7189bb PKK |
453 | error_append_hint(errp, "hint: failed on socket %s ", |
454 | server->value->u.q_unix.path); | |
455 | } else { | |
456 | error_append_hint(errp, "hint: failed on host %s and port %s ", | |
c5f1ae3a MA |
457 | server->value->u.inet.host, |
458 | server->value->u.inet.port); | |
6c7189bb | 459 | } |
7edac2dd | 460 | } |
4557117d | 461 | |
6c7189bb PKK |
462 | error_append_hint(errp, "Please refer to gluster logs for more info\n"); |
463 | ||
4557117d | 464 | /* glfs_init sometimes doesn't set errno although docs suggest that */ |
7edac2dd | 465 | if (errno == 0) { |
4557117d | 466 | errno = EINVAL; |
7edac2dd | 467 | } |
4557117d | 468 | |
8d6d89cb BR |
469 | goto out; |
470 | } | |
471 | return glfs; | |
472 | ||
473 | out: | |
474 | if (glfs) { | |
475 | old_errno = errno; | |
6349c154 | 476 | glfs_clear_preopened(glfs); |
8d6d89cb BR |
477 | errno = old_errno; |
478 | } | |
479 | return NULL; | |
480 | } | |
481 | ||
6c7189bb PKK |
482 | /* |
483 | * Convert the json formatted command line into qapi. | |
484 | */ | |
485 | static int qemu_gluster_parse_json(BlockdevOptionsGluster *gconf, | |
486 | QDict *options, Error **errp) | |
487 | { | |
488 | QemuOpts *opts; | |
62cf396b MA |
489 | SocketAddress *gsconf = NULL; |
490 | SocketAddressList *curr = NULL; | |
6c7189bb PKK |
491 | QDict *backing_options = NULL; |
492 | Error *local_err = NULL; | |
493 | char *str = NULL; | |
494 | const char *ptr; | |
56faeb9b | 495 | int i, type, num_servers; |
6c7189bb PKK |
496 | |
497 | /* create opts info from runtime_json_opts list */ | |
498 | opts = qemu_opts_create(&runtime_json_opts, NULL, 0, &error_abort); | |
499 | qemu_opts_absorb_qdict(opts, options, &local_err); | |
500 | if (local_err) { | |
501 | goto out; | |
502 | } | |
503 | ||
504 | num_servers = qdict_array_entries(options, GLUSTER_OPT_SERVER_PATTERN); | |
505 | if (num_servers < 1) { | |
506 | error_setg(&local_err, QERR_MISSING_PARAMETER, "server"); | |
507 | goto out; | |
508 | } | |
509 | ||
510 | ptr = qemu_opt_get(opts, GLUSTER_OPT_VOLUME); | |
511 | if (!ptr) { | |
512 | error_setg(&local_err, QERR_MISSING_PARAMETER, GLUSTER_OPT_VOLUME); | |
513 | goto out; | |
514 | } | |
515 | gconf->volume = g_strdup(ptr); | |
516 | ||
517 | ptr = qemu_opt_get(opts, GLUSTER_OPT_PATH); | |
518 | if (!ptr) { | |
519 | error_setg(&local_err, QERR_MISSING_PARAMETER, GLUSTER_OPT_PATH); | |
520 | goto out; | |
521 | } | |
522 | gconf->path = g_strdup(ptr); | |
523 | qemu_opts_del(opts); | |
524 | ||
525 | for (i = 0; i < num_servers; i++) { | |
526 | str = g_strdup_printf(GLUSTER_OPT_SERVER_PATTERN"%d.", i); | |
527 | qdict_extract_subqdict(options, &backing_options, str); | |
528 | ||
529 | /* create opts info from runtime_type_opts list */ | |
530 | opts = qemu_opts_create(&runtime_type_opts, NULL, 0, &error_abort); | |
531 | qemu_opts_absorb_qdict(opts, backing_options, &local_err); | |
532 | if (local_err) { | |
533 | goto out; | |
534 | } | |
535 | ||
536 | ptr = qemu_opt_get(opts, GLUSTER_OPT_TYPE); | |
6c7189bb PKK |
537 | if (!ptr) { |
538 | error_setg(&local_err, QERR_MISSING_PARAMETER, GLUSTER_OPT_TYPE); | |
539 | error_append_hint(&local_err, GERR_INDEX_HINT, i); | |
540 | goto out; | |
541 | ||
542 | } | |
62cf396b | 543 | gsconf = g_new0(SocketAddress, 1); |
c5f1ae3a MA |
544 | if (!strcmp(ptr, "tcp")) { |
545 | ptr = "inet"; /* accept legacy "tcp" */ | |
546 | } | |
06c60b6c | 547 | type = qapi_enum_parse(SocketAddressType_lookup, ptr, -1, NULL); |
62cf396b MA |
548 | if (type != SOCKET_ADDRESS_TYPE_INET |
549 | && type != SOCKET_ADDRESS_TYPE_UNIX) { | |
fce5d538 MA |
550 | error_setg(&local_err, |
551 | "Parameter '%s' may be 'inet' or 'unix'", | |
552 | GLUSTER_OPT_TYPE); | |
6c7189bb PKK |
553 | error_append_hint(&local_err, GERR_INDEX_HINT, i); |
554 | goto out; | |
555 | } | |
fce5d538 | 556 | gsconf->type = type; |
6c7189bb PKK |
557 | qemu_opts_del(opts); |
558 | ||
62cf396b | 559 | if (gsconf->type == SOCKET_ADDRESS_TYPE_INET) { |
c5f1ae3a MA |
560 | /* create opts info from runtime_inet_opts list */ |
561 | opts = qemu_opts_create(&runtime_inet_opts, NULL, 0, &error_abort); | |
6c7189bb PKK |
562 | qemu_opts_absorb_qdict(opts, backing_options, &local_err); |
563 | if (local_err) { | |
564 | goto out; | |
565 | } | |
566 | ||
567 | ptr = qemu_opt_get(opts, GLUSTER_OPT_HOST); | |
568 | if (!ptr) { | |
569 | error_setg(&local_err, QERR_MISSING_PARAMETER, | |
570 | GLUSTER_OPT_HOST); | |
571 | error_append_hint(&local_err, GERR_INDEX_HINT, i); | |
572 | goto out; | |
573 | } | |
c5f1ae3a | 574 | gsconf->u.inet.host = g_strdup(ptr); |
6c7189bb PKK |
575 | ptr = qemu_opt_get(opts, GLUSTER_OPT_PORT); |
576 | if (!ptr) { | |
577 | error_setg(&local_err, QERR_MISSING_PARAMETER, | |
578 | GLUSTER_OPT_PORT); | |
579 | error_append_hint(&local_err, GERR_INDEX_HINT, i); | |
580 | goto out; | |
581 | } | |
c5f1ae3a | 582 | gsconf->u.inet.port = g_strdup(ptr); |
6c7189bb PKK |
583 | |
584 | /* defend for unsupported fields in InetSocketAddress, | |
585 | * i.e. @ipv4, @ipv6 and @to | |
586 | */ | |
587 | ptr = qemu_opt_get(opts, GLUSTER_OPT_TO); | |
588 | if (ptr) { | |
c5f1ae3a | 589 | gsconf->u.inet.has_to = true; |
6c7189bb PKK |
590 | } |
591 | ptr = qemu_opt_get(opts, GLUSTER_OPT_IPV4); | |
592 | if (ptr) { | |
c5f1ae3a | 593 | gsconf->u.inet.has_ipv4 = true; |
6c7189bb PKK |
594 | } |
595 | ptr = qemu_opt_get(opts, GLUSTER_OPT_IPV6); | |
596 | if (ptr) { | |
c5f1ae3a | 597 | gsconf->u.inet.has_ipv6 = true; |
6c7189bb | 598 | } |
c5f1ae3a | 599 | if (gsconf->u.inet.has_to) { |
6c7189bb PKK |
600 | error_setg(&local_err, "Parameter 'to' not supported"); |
601 | goto out; | |
602 | } | |
c5f1ae3a | 603 | if (gsconf->u.inet.has_ipv4 || gsconf->u.inet.has_ipv6) { |
6c7189bb PKK |
604 | error_setg(&local_err, "Parameters 'ipv4/ipv6' not supported"); |
605 | goto out; | |
606 | } | |
607 | qemu_opts_del(opts); | |
608 | } else { | |
609 | /* create opts info from runtime_unix_opts list */ | |
610 | opts = qemu_opts_create(&runtime_unix_opts, NULL, 0, &error_abort); | |
611 | qemu_opts_absorb_qdict(opts, backing_options, &local_err); | |
612 | if (local_err) { | |
613 | goto out; | |
614 | } | |
615 | ||
616 | ptr = qemu_opt_get(opts, GLUSTER_OPT_SOCKET); | |
617 | if (!ptr) { | |
618 | error_setg(&local_err, QERR_MISSING_PARAMETER, | |
619 | GLUSTER_OPT_SOCKET); | |
620 | error_append_hint(&local_err, GERR_INDEX_HINT, i); | |
621 | goto out; | |
622 | } | |
623 | gsconf->u.q_unix.path = g_strdup(ptr); | |
624 | qemu_opts_del(opts); | |
625 | } | |
626 | ||
627 | if (gconf->server == NULL) { | |
62cf396b | 628 | gconf->server = g_new0(SocketAddressList, 1); |
6c7189bb PKK |
629 | gconf->server->value = gsconf; |
630 | curr = gconf->server; | |
631 | } else { | |
62cf396b | 632 | curr->next = g_new0(SocketAddressList, 1); |
6c7189bb PKK |
633 | curr->next->value = gsconf; |
634 | curr = curr->next; | |
635 | } | |
85a82e85 | 636 | gsconf = NULL; |
6c7189bb | 637 | |
85a82e85 MA |
638 | QDECREF(backing_options); |
639 | backing_options = NULL; | |
6c7189bb PKK |
640 | g_free(str); |
641 | str = NULL; | |
642 | } | |
643 | ||
644 | return 0; | |
645 | ||
646 | out: | |
647 | error_propagate(errp, local_err); | |
62cf396b | 648 | qapi_free_SocketAddress(gsconf); |
6c7189bb | 649 | qemu_opts_del(opts); |
85a82e85 MA |
650 | g_free(str); |
651 | QDECREF(backing_options); | |
6c7189bb PKK |
652 | errno = EINVAL; |
653 | return -errno; | |
654 | } | |
655 | ||
656 | static struct glfs *qemu_gluster_init(BlockdevOptionsGluster *gconf, | |
657 | const char *filename, | |
658 | QDict *options, Error **errp) | |
659 | { | |
660 | int ret; | |
661 | if (filename) { | |
662 | ret = qemu_gluster_parse_uri(gconf, filename); | |
663 | if (ret < 0) { | |
664 | error_setg(errp, "invalid URI"); | |
665 | error_append_hint(errp, "Usage: file=gluster[+transport]://" | |
e9db8ff3 PKK |
666 | "[host[:port]]volume/path[?socket=...]" |
667 | "[,file.debug=N]" | |
668 | "[,file.logfile=/path/filename.log]\n"); | |
6c7189bb PKK |
669 | errno = -ret; |
670 | return NULL; | |
671 | } | |
672 | } else { | |
673 | ret = qemu_gluster_parse_json(gconf, options, errp); | |
674 | if (ret < 0) { | |
675 | error_append_hint(errp, "Usage: " | |
676 | "-drive driver=qcow2,file.driver=gluster," | |
677 | "file.volume=testvol,file.path=/path/a.qcow2" | |
e9db8ff3 PKK |
678 | "[,file.debug=9]" |
679 | "[,file.logfile=/path/filename.log]," | |
c5f1ae3a | 680 | "file.server.0.type=inet," |
6c7189bb PKK |
681 | "file.server.0.host=1.2.3.4," |
682 | "file.server.0.port=24007," | |
683 | "file.server.1.transport=unix," | |
684 | "file.server.1.socket=/var/run/glusterd.socket ..." | |
685 | "\n"); | |
686 | errno = -ret; | |
687 | return NULL; | |
688 | } | |
689 | ||
690 | } | |
691 | ||
692 | return qemu_gluster_glfs_init(gconf, errp); | |
693 | } | |
694 | ||
7c815372 BR |
695 | /* |
696 | * AIO callback routine called from GlusterFS thread. | |
697 | */ | |
698 | static void gluster_finish_aiocb(struct glfs_fd *fd, ssize_t ret, void *arg) | |
699 | { | |
700 | GlusterAIOCB *acb = (GlusterAIOCB *)arg; | |
701 | ||
702 | if (!ret || ret == acb->size) { | |
703 | acb->ret = 0; /* Success */ | |
704 | } else if (ret < 0) { | |
a8827453 | 705 | acb->ret = -errno; /* Read/Write failed */ |
7c815372 BR |
706 | } else { |
707 | acb->ret = -EIO; /* Partial read/write - fail it */ | |
708 | } | |
709 | ||
1919631e | 710 | aio_co_schedule(acb->aio_context, acb->coroutine); |
7c815372 BR |
711 | } |
712 | ||
1b37b344 JC |
713 | static void qemu_gluster_parse_flags(int bdrv_flags, int *open_flags) |
714 | { | |
715 | assert(open_flags != NULL); | |
716 | ||
717 | *open_flags |= O_BINARY; | |
718 | ||
719 | if (bdrv_flags & BDRV_O_RDWR) { | |
720 | *open_flags |= O_RDWR; | |
721 | } else { | |
722 | *open_flags |= O_RDONLY; | |
723 | } | |
724 | ||
725 | if ((bdrv_flags & BDRV_O_NOCACHE)) { | |
726 | *open_flags |= O_DIRECT; | |
727 | } | |
728 | } | |
729 | ||
947eb203 NV |
730 | /* |
731 | * Do SEEK_DATA/HOLE to detect if it is functional. Older broken versions of | |
732 | * gfapi incorrectly return the current offset when SEEK_DATA/HOLE is used. | |
733 | * - Corrected versions return -1 and set errno to EINVAL. | |
734 | * - Versions that support SEEK_DATA/HOLE correctly, will return -1 and set | |
735 | * errno to ENXIO when SEEK_DATA is called with a position of EOF. | |
736 | */ | |
737 | static bool qemu_gluster_test_seek(struct glfs_fd *fd) | |
738 | { | |
d9b78974 JC |
739 | off_t ret = 0; |
740 | ||
741 | #if defined SEEK_HOLE && defined SEEK_DATA | |
742 | off_t eof; | |
947eb203 NV |
743 | |
744 | eof = glfs_lseek(fd, 0, SEEK_END); | |
745 | if (eof < 0) { | |
746 | /* this should never occur */ | |
747 | return false; | |
748 | } | |
749 | ||
750 | /* this should always fail with ENXIO if SEEK_DATA is supported */ | |
751 | ret = glfs_lseek(fd, eof, SEEK_DATA); | |
d9b78974 JC |
752 | #endif |
753 | ||
947eb203 NV |
754 | return (ret < 0) && (errno == ENXIO); |
755 | } | |
756 | ||
56d1b4d2 | 757 | static int qemu_gluster_open(BlockDriverState *bs, QDict *options, |
015a1036 | 758 | int bdrv_flags, Error **errp) |
8d6d89cb BR |
759 | { |
760 | BDRVGlusterState *s = bs->opaque; | |
1b37b344 | 761 | int open_flags = 0; |
8d6d89cb | 762 | int ret = 0; |
7edac2dd | 763 | BlockdevOptionsGluster *gconf = NULL; |
b4894776 KW |
764 | QemuOpts *opts; |
765 | Error *local_err = NULL; | |
e9db8ff3 | 766 | const char *filename, *logfile; |
b4894776 | 767 | |
87ea75d5 | 768 | opts = qemu_opts_create(&runtime_opts, NULL, 0, &error_abort); |
b4894776 | 769 | qemu_opts_absorb_qdict(opts, options, &local_err); |
84d18f06 | 770 | if (local_err) { |
a7451cb8 | 771 | error_propagate(errp, local_err); |
b4894776 KW |
772 | ret = -EINVAL; |
773 | goto out; | |
774 | } | |
775 | ||
7eac868a JC |
776 | filename = qemu_opt_get(opts, GLUSTER_OPT_FILENAME); |
777 | ||
1a417e46 PKK |
778 | s->debug = qemu_opt_get_number(opts, GLUSTER_OPT_DEBUG, |
779 | GLUSTER_DEBUG_DEFAULT); | |
780 | if (s->debug < 0) { | |
781 | s->debug = 0; | |
782 | } else if (s->debug > GLUSTER_DEBUG_MAX) { | |
783 | s->debug = GLUSTER_DEBUG_MAX; | |
7eac868a | 784 | } |
b4894776 | 785 | |
7edac2dd | 786 | gconf = g_new0(BlockdevOptionsGluster, 1); |
1a417e46 PKK |
787 | gconf->debug = s->debug; |
788 | gconf->has_debug = true; | |
e9db8ff3 PKK |
789 | |
790 | logfile = qemu_opt_get(opts, GLUSTER_OPT_LOGFILE); | |
791 | s->logfile = g_strdup(logfile ? logfile : GLUSTER_LOGFILE_DEFAULT); | |
792 | ||
793 | gconf->logfile = g_strdup(s->logfile); | |
794 | gconf->has_logfile = true; | |
795 | ||
6c7189bb | 796 | s->glfs = qemu_gluster_init(gconf, filename, options, errp); |
8d6d89cb BR |
797 | if (!s->glfs) { |
798 | ret = -errno; | |
799 | goto out; | |
800 | } | |
801 | ||
d85fa9eb JC |
802 | #ifdef CONFIG_GLUSTERFS_XLATOR_OPT |
803 | /* Without this, if fsync fails for a recoverable reason (for instance, | |
804 | * ENOSPC), gluster will dump its cache, preventing retries. This means | |
805 | * almost certain data loss. Not all gluster versions support the | |
806 | * 'resync-failed-syncs-after-fsync' key value, but there is no way to | |
807 | * discover during runtime if it is supported (this api returns success for | |
808 | * unknown key/value pairs) */ | |
809 | ret = glfs_set_xlator_option(s->glfs, "*-write-behind", | |
810 | "resync-failed-syncs-after-fsync", | |
811 | "on"); | |
812 | if (ret < 0) { | |
813 | error_setg_errno(errp, errno, "Unable to set xlator key/value pair"); | |
814 | ret = -errno; | |
815 | goto out; | |
816 | } | |
817 | #endif | |
818 | ||
1b37b344 | 819 | qemu_gluster_parse_flags(bdrv_flags, &open_flags); |
8d6d89cb | 820 | |
d5cf4079 | 821 | s->fd = glfs_open(s->glfs, gconf->path, open_flags); |
8d6d89cb BR |
822 | if (!s->fd) { |
823 | ret = -errno; | |
8d6d89cb | 824 | } |
8d6d89cb | 825 | |
947eb203 NV |
826 | s->supports_seek_data = qemu_gluster_test_seek(s->fd); |
827 | ||
8d6d89cb | 828 | out: |
b4894776 | 829 | qemu_opts_del(opts); |
7edac2dd | 830 | qapi_free_BlockdevOptionsGluster(gconf); |
8d6d89cb BR |
831 | if (!ret) { |
832 | return ret; | |
833 | } | |
e9db8ff3 | 834 | g_free(s->logfile); |
8d6d89cb BR |
835 | if (s->fd) { |
836 | glfs_close(s->fd); | |
837 | } | |
6349c154 PKK |
838 | |
839 | glfs_clear_preopened(s->glfs); | |
840 | ||
8d6d89cb BR |
841 | return ret; |
842 | } | |
843 | ||
adccfbcd JC |
844 | static int qemu_gluster_reopen_prepare(BDRVReopenState *state, |
845 | BlockReopenQueue *queue, Error **errp) | |
846 | { | |
847 | int ret = 0; | |
7eac868a | 848 | BDRVGlusterState *s; |
adccfbcd | 849 | BDRVGlusterReopenState *reop_s; |
7edac2dd | 850 | BlockdevOptionsGluster *gconf; |
adccfbcd JC |
851 | int open_flags = 0; |
852 | ||
853 | assert(state != NULL); | |
854 | assert(state->bs != NULL); | |
855 | ||
7eac868a JC |
856 | s = state->bs->opaque; |
857 | ||
5839e53b | 858 | state->opaque = g_new0(BDRVGlusterReopenState, 1); |
adccfbcd JC |
859 | reop_s = state->opaque; |
860 | ||
861 | qemu_gluster_parse_flags(state->flags, &open_flags); | |
862 | ||
7edac2dd | 863 | gconf = g_new0(BlockdevOptionsGluster, 1); |
1a417e46 PKK |
864 | gconf->debug = s->debug; |
865 | gconf->has_debug = true; | |
e9db8ff3 PKK |
866 | gconf->logfile = g_strdup(s->logfile); |
867 | gconf->has_logfile = true; | |
6c7189bb | 868 | reop_s->glfs = qemu_gluster_init(gconf, state->bs->filename, NULL, errp); |
adccfbcd JC |
869 | if (reop_s->glfs == NULL) { |
870 | ret = -errno; | |
871 | goto exit; | |
872 | } | |
873 | ||
d85fa9eb JC |
874 | #ifdef CONFIG_GLUSTERFS_XLATOR_OPT |
875 | ret = glfs_set_xlator_option(reop_s->glfs, "*-write-behind", | |
876 | "resync-failed-syncs-after-fsync", "on"); | |
877 | if (ret < 0) { | |
878 | error_setg_errno(errp, errno, "Unable to set xlator key/value pair"); | |
879 | ret = -errno; | |
880 | goto exit; | |
881 | } | |
882 | #endif | |
883 | ||
d5cf4079 | 884 | reop_s->fd = glfs_open(reop_s->glfs, gconf->path, open_flags); |
adccfbcd JC |
885 | if (reop_s->fd == NULL) { |
886 | /* reops->glfs will be cleaned up in _abort */ | |
887 | ret = -errno; | |
888 | goto exit; | |
889 | } | |
890 | ||
891 | exit: | |
892 | /* state->opaque will be freed in either the _abort or _commit */ | |
7edac2dd | 893 | qapi_free_BlockdevOptionsGluster(gconf); |
adccfbcd JC |
894 | return ret; |
895 | } | |
896 | ||
897 | static void qemu_gluster_reopen_commit(BDRVReopenState *state) | |
898 | { | |
899 | BDRVGlusterReopenState *reop_s = state->opaque; | |
900 | BDRVGlusterState *s = state->bs->opaque; | |
901 | ||
902 | ||
903 | /* close the old */ | |
904 | if (s->fd) { | |
905 | glfs_close(s->fd); | |
906 | } | |
6349c154 PKK |
907 | |
908 | glfs_clear_preopened(s->glfs); | |
adccfbcd JC |
909 | |
910 | /* use the newly opened image / connection */ | |
911 | s->fd = reop_s->fd; | |
912 | s->glfs = reop_s->glfs; | |
913 | ||
914 | g_free(state->opaque); | |
915 | state->opaque = NULL; | |
916 | ||
917 | return; | |
918 | } | |
919 | ||
920 | ||
921 | static void qemu_gluster_reopen_abort(BDRVReopenState *state) | |
922 | { | |
923 | BDRVGlusterReopenState *reop_s = state->opaque; | |
924 | ||
925 | if (reop_s == NULL) { | |
926 | return; | |
927 | } | |
928 | ||
929 | if (reop_s->fd) { | |
930 | glfs_close(reop_s->fd); | |
931 | } | |
932 | ||
6349c154 | 933 | glfs_clear_preopened(reop_s->glfs); |
adccfbcd JC |
934 | |
935 | g_free(state->opaque); | |
936 | state->opaque = NULL; | |
937 | ||
938 | return; | |
939 | } | |
940 | ||
7c815372 | 941 | #ifdef CONFIG_GLUSTERFS_ZEROFILL |
e88a36eb | 942 | static coroutine_fn int qemu_gluster_co_pwrite_zeroes(BlockDriverState *bs, |
f70c50c8 PKK |
943 | int64_t offset, |
944 | int size, | |
945 | BdrvRequestFlags flags) | |
7c815372 BR |
946 | { |
947 | int ret; | |
c833d1e8 | 948 | GlusterAIOCB acb; |
7c815372 | 949 | BDRVGlusterState *s = bs->opaque; |
7c815372 | 950 | |
c833d1e8 PB |
951 | acb.size = size; |
952 | acb.ret = 0; | |
953 | acb.coroutine = qemu_coroutine_self(); | |
954 | acb.aio_context = bdrv_get_aio_context(bs); | |
7c815372 | 955 | |
c833d1e8 | 956 | ret = glfs_zerofill_async(s->fd, offset, size, gluster_finish_aiocb, &acb); |
7c815372 | 957 | if (ret < 0) { |
c833d1e8 | 958 | return -errno; |
7c815372 BR |
959 | } |
960 | ||
961 | qemu_coroutine_yield(); | |
c833d1e8 | 962 | return acb.ret; |
7c815372 BR |
963 | } |
964 | #endif | |
965 | ||
8d6d89cb | 966 | static int qemu_gluster_create(const char *filename, |
90c772de | 967 | QemuOpts *opts, Error **errp) |
8d6d89cb | 968 | { |
7edac2dd | 969 | BlockdevOptionsGluster *gconf; |
8d6d89cb BR |
970 | struct glfs *glfs; |
971 | struct glfs_fd *fd; | |
972 | int ret = 0; | |
df3a429a | 973 | PreallocMode prealloc; |
8d6d89cb | 974 | int64_t total_size = 0; |
90c772de | 975 | char *tmp = NULL; |
df3a429a | 976 | Error *local_err = NULL; |
8d6d89cb | 977 | |
7edac2dd | 978 | gconf = g_new0(BlockdevOptionsGluster, 1); |
1a417e46 PKK |
979 | gconf->debug = qemu_opt_get_number_del(opts, GLUSTER_OPT_DEBUG, |
980 | GLUSTER_DEBUG_DEFAULT); | |
981 | if (gconf->debug < 0) { | |
982 | gconf->debug = 0; | |
983 | } else if (gconf->debug > GLUSTER_DEBUG_MAX) { | |
984 | gconf->debug = GLUSTER_DEBUG_MAX; | |
985 | } | |
986 | gconf->has_debug = true; | |
7eac868a | 987 | |
e9db8ff3 PKK |
988 | gconf->logfile = qemu_opt_get_del(opts, GLUSTER_OPT_LOGFILE); |
989 | if (!gconf->logfile) { | |
990 | gconf->logfile = g_strdup(GLUSTER_LOGFILE_DEFAULT); | |
991 | } | |
992 | gconf->has_logfile = true; | |
993 | ||
6c7189bb | 994 | glfs = qemu_gluster_init(gconf, filename, NULL, errp); |
8d6d89cb | 995 | if (!glfs) { |
4557117d | 996 | ret = -errno; |
8d6d89cb BR |
997 | goto out; |
998 | } | |
999 | ||
180e9526 HT |
1000 | total_size = ROUND_UP(qemu_opt_get_size_del(opts, BLOCK_OPT_SIZE, 0), |
1001 | BDRV_SECTOR_SIZE); | |
90c772de CL |
1002 | |
1003 | tmp = qemu_opt_get_del(opts, BLOCK_OPT_PREALLOC); | |
06c60b6c | 1004 | prealloc = qapi_enum_parse(PreallocMode_lookup, tmp, PREALLOC_MODE_OFF, |
df3a429a NV |
1005 | &local_err); |
1006 | g_free(tmp); | |
1007 | if (local_err) { | |
1008 | error_propagate(errp, local_err); | |
90c772de CL |
1009 | ret = -EINVAL; |
1010 | goto out; | |
8d6d89cb BR |
1011 | } |
1012 | ||
d5cf4079 | 1013 | fd = glfs_creat(glfs, gconf->path, |
f70c50c8 | 1014 | O_WRONLY | O_CREAT | O_TRUNC | O_BINARY, S_IRUSR | S_IWUSR); |
8d6d89cb BR |
1015 | if (!fd) { |
1016 | ret = -errno; | |
df3a429a NV |
1017 | goto out; |
1018 | } | |
1019 | ||
1020 | switch (prealloc) { | |
1021 | #ifdef CONFIG_GLUSTERFS_FALLOCATE | |
1022 | case PREALLOC_MODE_FALLOC: | |
1023 | if (glfs_fallocate(fd, 0, 0, total_size)) { | |
1024 | error_setg(errp, "Could not preallocate data for the new file"); | |
1025 | ret = -errno; | |
1026 | } | |
1027 | break; | |
1028 | #endif /* CONFIG_GLUSTERFS_FALLOCATE */ | |
1029 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
1030 | case PREALLOC_MODE_FULL: | |
180e9526 | 1031 | if (!glfs_ftruncate(fd, total_size)) { |
df3a429a NV |
1032 | if (glfs_zerofill(fd, 0, total_size)) { |
1033 | error_setg(errp, "Could not zerofill the new file"); | |
cf7f616b BR |
1034 | ret = -errno; |
1035 | } | |
1036 | } else { | |
df3a429a | 1037 | error_setg(errp, "Could not resize file"); |
8d6d89cb BR |
1038 | ret = -errno; |
1039 | } | |
df3a429a NV |
1040 | break; |
1041 | #endif /* CONFIG_GLUSTERFS_ZEROFILL */ | |
1042 | case PREALLOC_MODE_OFF: | |
1043 | if (glfs_ftruncate(fd, total_size) != 0) { | |
8d6d89cb | 1044 | ret = -errno; |
df3a429a | 1045 | error_setg(errp, "Could not resize file"); |
8d6d89cb | 1046 | } |
df3a429a NV |
1047 | break; |
1048 | default: | |
1049 | ret = -EINVAL; | |
1050 | error_setg(errp, "Unsupported preallocation mode: %s", | |
1051 | PreallocMode_lookup[prealloc]); | |
1052 | break; | |
1053 | } | |
1054 | ||
1055 | if (glfs_close(fd) != 0) { | |
1056 | ret = -errno; | |
8d6d89cb BR |
1057 | } |
1058 | out: | |
7edac2dd | 1059 | qapi_free_BlockdevOptionsGluster(gconf); |
6349c154 | 1060 | glfs_clear_preopened(glfs); |
8d6d89cb BR |
1061 | return ret; |
1062 | } | |
1063 | ||
15744b0b | 1064 | static coroutine_fn int qemu_gluster_co_rw(BlockDriverState *bs, |
f70c50c8 PKK |
1065 | int64_t sector_num, int nb_sectors, |
1066 | QEMUIOVector *qiov, int write) | |
8d6d89cb BR |
1067 | { |
1068 | int ret; | |
c833d1e8 | 1069 | GlusterAIOCB acb; |
8d6d89cb | 1070 | BDRVGlusterState *s = bs->opaque; |
15744b0b BR |
1071 | size_t size = nb_sectors * BDRV_SECTOR_SIZE; |
1072 | off_t offset = sector_num * BDRV_SECTOR_SIZE; | |
8d6d89cb | 1073 | |
c833d1e8 PB |
1074 | acb.size = size; |
1075 | acb.ret = 0; | |
1076 | acb.coroutine = qemu_coroutine_self(); | |
1077 | acb.aio_context = bdrv_get_aio_context(bs); | |
8d6d89cb BR |
1078 | |
1079 | if (write) { | |
1080 | ret = glfs_pwritev_async(s->fd, qiov->iov, qiov->niov, offset, 0, | |
f70c50c8 | 1081 | gluster_finish_aiocb, &acb); |
8d6d89cb BR |
1082 | } else { |
1083 | ret = glfs_preadv_async(s->fd, qiov->iov, qiov->niov, offset, 0, | |
f70c50c8 | 1084 | gluster_finish_aiocb, &acb); |
8d6d89cb BR |
1085 | } |
1086 | ||
1087 | if (ret < 0) { | |
c833d1e8 | 1088 | return -errno; |
8d6d89cb | 1089 | } |
15744b0b BR |
1090 | |
1091 | qemu_coroutine_yield(); | |
c833d1e8 | 1092 | return acb.ret; |
8d6d89cb BR |
1093 | } |
1094 | ||
4bff28b8 | 1095 | static int qemu_gluster_truncate(BlockDriverState *bs, int64_t offset, |
8243ccb7 | 1096 | PreallocMode prealloc, Error **errp) |
42ec24e2 PB |
1097 | { |
1098 | int ret; | |
1099 | BDRVGlusterState *s = bs->opaque; | |
1100 | ||
8243ccb7 HR |
1101 | if (prealloc != PREALLOC_MODE_OFF) { |
1102 | error_setg(errp, "Unsupported preallocation mode '%s'", | |
1103 | PreallocMode_lookup[prealloc]); | |
1104 | return -ENOTSUP; | |
1105 | } | |
1106 | ||
42ec24e2 PB |
1107 | ret = glfs_ftruncate(s->fd, offset); |
1108 | if (ret < 0) { | |
f59adb32 HR |
1109 | ret = -errno; |
1110 | error_setg_errno(errp, -ret, "Failed to truncate file"); | |
1111 | return ret; | |
42ec24e2 PB |
1112 | } |
1113 | ||
1114 | return 0; | |
1115 | } | |
1116 | ||
15744b0b | 1117 | static coroutine_fn int qemu_gluster_co_readv(BlockDriverState *bs, |
f70c50c8 PKK |
1118 | int64_t sector_num, |
1119 | int nb_sectors, | |
1120 | QEMUIOVector *qiov) | |
8d6d89cb | 1121 | { |
15744b0b | 1122 | return qemu_gluster_co_rw(bs, sector_num, nb_sectors, qiov, 0); |
8d6d89cb BR |
1123 | } |
1124 | ||
15744b0b | 1125 | static coroutine_fn int qemu_gluster_co_writev(BlockDriverState *bs, |
f70c50c8 PKK |
1126 | int64_t sector_num, |
1127 | int nb_sectors, | |
1128 | QEMUIOVector *qiov) | |
8d6d89cb | 1129 | { |
15744b0b | 1130 | return qemu_gluster_co_rw(bs, sector_num, nb_sectors, qiov, 1); |
8d6d89cb BR |
1131 | } |
1132 | ||
5d4343e6 JC |
1133 | static void qemu_gluster_close(BlockDriverState *bs) |
1134 | { | |
1135 | BDRVGlusterState *s = bs->opaque; | |
1136 | ||
e9db8ff3 | 1137 | g_free(s->logfile); |
5d4343e6 JC |
1138 | if (s->fd) { |
1139 | glfs_close(s->fd); | |
1140 | s->fd = NULL; | |
1141 | } | |
6349c154 | 1142 | glfs_clear_preopened(s->glfs); |
5d4343e6 JC |
1143 | } |
1144 | ||
15744b0b | 1145 | static coroutine_fn int qemu_gluster_co_flush_to_disk(BlockDriverState *bs) |
8d6d89cb BR |
1146 | { |
1147 | int ret; | |
c833d1e8 | 1148 | GlusterAIOCB acb; |
8d6d89cb BR |
1149 | BDRVGlusterState *s = bs->opaque; |
1150 | ||
c833d1e8 PB |
1151 | acb.size = 0; |
1152 | acb.ret = 0; | |
1153 | acb.coroutine = qemu_coroutine_self(); | |
1154 | acb.aio_context = bdrv_get_aio_context(bs); | |
8d6d89cb | 1155 | |
c833d1e8 | 1156 | ret = glfs_fsync_async(s->fd, gluster_finish_aiocb, &acb); |
8d6d89cb | 1157 | if (ret < 0) { |
d85fa9eb JC |
1158 | ret = -errno; |
1159 | goto error; | |
8d6d89cb | 1160 | } |
15744b0b BR |
1161 | |
1162 | qemu_coroutine_yield(); | |
d85fa9eb JC |
1163 | if (acb.ret < 0) { |
1164 | ret = acb.ret; | |
1165 | goto error; | |
1166 | } | |
1167 | ||
c833d1e8 | 1168 | return acb.ret; |
d85fa9eb JC |
1169 | |
1170 | error: | |
1171 | /* Some versions of Gluster (3.5.6 -> 3.5.8?) will not retain its cache | |
1172 | * after a fsync failure, so we have no way of allowing the guest to safely | |
1173 | * continue. Gluster versions prior to 3.5.6 don't retain the cache | |
1174 | * either, but will invalidate the fd on error, so this is again our only | |
1175 | * option. | |
1176 | * | |
1177 | * The 'resync-failed-syncs-after-fsync' xlator option for the | |
1178 | * write-behind cache will cause later gluster versions to retain its | |
1179 | * cache after error, so long as the fd remains open. However, we | |
1180 | * currently have no way of knowing if this option is supported. | |
1181 | * | |
1182 | * TODO: Once gluster provides a way for us to determine if the option | |
1183 | * is supported, bypass the closure and setting drv to NULL. */ | |
1184 | qemu_gluster_close(bs); | |
1185 | bs->drv = NULL; | |
1186 | return ret; | |
8d6d89cb BR |
1187 | } |
1188 | ||
0c14fb47 | 1189 | #ifdef CONFIG_GLUSTERFS_DISCARD |
1014170b EB |
1190 | static coroutine_fn int qemu_gluster_co_pdiscard(BlockDriverState *bs, |
1191 | int64_t offset, int size) | |
0c14fb47 BR |
1192 | { |
1193 | int ret; | |
c833d1e8 | 1194 | GlusterAIOCB acb; |
0c14fb47 | 1195 | BDRVGlusterState *s = bs->opaque; |
0c14fb47 | 1196 | |
c833d1e8 PB |
1197 | acb.size = 0; |
1198 | acb.ret = 0; | |
1199 | acb.coroutine = qemu_coroutine_self(); | |
1200 | acb.aio_context = bdrv_get_aio_context(bs); | |
0c14fb47 | 1201 | |
c833d1e8 | 1202 | ret = glfs_discard_async(s->fd, offset, size, gluster_finish_aiocb, &acb); |
0c14fb47 | 1203 | if (ret < 0) { |
c833d1e8 | 1204 | return -errno; |
0c14fb47 | 1205 | } |
15744b0b BR |
1206 | |
1207 | qemu_coroutine_yield(); | |
c833d1e8 | 1208 | return acb.ret; |
0c14fb47 BR |
1209 | } |
1210 | #endif | |
1211 | ||
8d6d89cb BR |
1212 | static int64_t qemu_gluster_getlength(BlockDriverState *bs) |
1213 | { | |
1214 | BDRVGlusterState *s = bs->opaque; | |
1215 | int64_t ret; | |
1216 | ||
1217 | ret = glfs_lseek(s->fd, 0, SEEK_END); | |
1218 | if (ret < 0) { | |
1219 | return -errno; | |
1220 | } else { | |
1221 | return ret; | |
1222 | } | |
1223 | } | |
1224 | ||
1225 | static int64_t qemu_gluster_allocated_file_size(BlockDriverState *bs) | |
1226 | { | |
1227 | BDRVGlusterState *s = bs->opaque; | |
1228 | struct stat st; | |
1229 | int ret; | |
1230 | ||
1231 | ret = glfs_fstat(s->fd, &st); | |
1232 | if (ret < 0) { | |
1233 | return -errno; | |
1234 | } else { | |
1235 | return st.st_blocks * 512; | |
1236 | } | |
1237 | } | |
1238 | ||
8ab6feec KW |
1239 | static int qemu_gluster_has_zero_init(BlockDriverState *bs) |
1240 | { | |
1241 | /* GlusterFS volume could be backed by a block device */ | |
1242 | return 0; | |
1243 | } | |
1244 | ||
947eb203 NV |
1245 | /* |
1246 | * Find allocation range in @bs around offset @start. | |
1247 | * May change underlying file descriptor's file offset. | |
1248 | * If @start is not in a hole, store @start in @data, and the | |
1249 | * beginning of the next hole in @hole, and return 0. | |
1250 | * If @start is in a non-trailing hole, store @start in @hole and the | |
1251 | * beginning of the next non-hole in @data, and return 0. | |
1252 | * If @start is in a trailing hole or beyond EOF, return -ENXIO. | |
1253 | * If we can't find out, return a negative errno other than -ENXIO. | |
1254 | * | |
c1bb86cd | 1255 | * (Shamefully copied from file-posix.c, only miniscule adaptions.) |
947eb203 NV |
1256 | */ |
1257 | static int find_allocation(BlockDriverState *bs, off_t start, | |
1258 | off_t *data, off_t *hole) | |
1259 | { | |
1260 | BDRVGlusterState *s = bs->opaque; | |
947eb203 NV |
1261 | |
1262 | if (!s->supports_seek_data) { | |
d9b78974 | 1263 | goto exit; |
947eb203 NV |
1264 | } |
1265 | ||
d9b78974 JC |
1266 | #if defined SEEK_HOLE && defined SEEK_DATA |
1267 | off_t offs; | |
1268 | ||
947eb203 NV |
1269 | /* |
1270 | * SEEK_DATA cases: | |
1271 | * D1. offs == start: start is in data | |
1272 | * D2. offs > start: start is in a hole, next data at offs | |
1273 | * D3. offs < 0, errno = ENXIO: either start is in a trailing hole | |
1274 | * or start is beyond EOF | |
1275 | * If the latter happens, the file has been truncated behind | |
1276 | * our back since we opened it. All bets are off then. | |
1277 | * Treating like a trailing hole is simplest. | |
1278 | * D4. offs < 0, errno != ENXIO: we learned nothing | |
1279 | */ | |
1280 | offs = glfs_lseek(s->fd, start, SEEK_DATA); | |
1281 | if (offs < 0) { | |
1282 | return -errno; /* D3 or D4 */ | |
1283 | } | |
223a23c1 JC |
1284 | |
1285 | if (offs < start) { | |
1286 | /* This is not a valid return by lseek(). We are safe to just return | |
1287 | * -EIO in this case, and we'll treat it like D4. Unfortunately some | |
1288 | * versions of gluster server will return offs < start, so an assert | |
1289 | * here will unnecessarily abort QEMU. */ | |
1290 | return -EIO; | |
1291 | } | |
947eb203 NV |
1292 | |
1293 | if (offs > start) { | |
1294 | /* D2: in hole, next data at offs */ | |
1295 | *hole = start; | |
1296 | *data = offs; | |
1297 | return 0; | |
1298 | } | |
1299 | ||
1300 | /* D1: in data, end not yet known */ | |
1301 | ||
1302 | /* | |
1303 | * SEEK_HOLE cases: | |
1304 | * H1. offs == start: start is in a hole | |
1305 | * If this happens here, a hole has been dug behind our back | |
1306 | * since the previous lseek(). | |
1307 | * H2. offs > start: either start is in data, next hole at offs, | |
1308 | * or start is in trailing hole, EOF at offs | |
1309 | * Linux treats trailing holes like any other hole: offs == | |
1310 | * start. Solaris seeks to EOF instead: offs > start (blech). | |
1311 | * If that happens here, a hole has been dug behind our back | |
1312 | * since the previous lseek(). | |
1313 | * H3. offs < 0, errno = ENXIO: start is beyond EOF | |
1314 | * If this happens, the file has been truncated behind our | |
1315 | * back since we opened it. Treat it like a trailing hole. | |
1316 | * H4. offs < 0, errno != ENXIO: we learned nothing | |
1317 | * Pretend we know nothing at all, i.e. "forget" about D1. | |
1318 | */ | |
1319 | offs = glfs_lseek(s->fd, start, SEEK_HOLE); | |
1320 | if (offs < 0) { | |
1321 | return -errno; /* D1 and (H3 or H4) */ | |
1322 | } | |
223a23c1 JC |
1323 | |
1324 | if (offs < start) { | |
1325 | /* This is not a valid return by lseek(). We are safe to just return | |
1326 | * -EIO in this case, and we'll treat it like H4. Unfortunately some | |
1327 | * versions of gluster server will return offs < start, so an assert | |
1328 | * here will unnecessarily abort QEMU. */ | |
1329 | return -EIO; | |
1330 | } | |
947eb203 NV |
1331 | |
1332 | if (offs > start) { | |
1333 | /* | |
1334 | * D1 and H2: either in data, next hole at offs, or it was in | |
1335 | * data but is now in a trailing hole. In the latter case, | |
1336 | * all bets are off. Treating it as if it there was data all | |
1337 | * the way to EOF is safe, so simply do that. | |
1338 | */ | |
1339 | *data = start; | |
1340 | *hole = offs; | |
1341 | return 0; | |
1342 | } | |
1343 | ||
1344 | /* D1 and H1 */ | |
1345 | return -EBUSY; | |
d9b78974 JC |
1346 | #endif |
1347 | ||
1348 | exit: | |
1349 | return -ENOTSUP; | |
947eb203 NV |
1350 | } |
1351 | ||
1352 | /* | |
1353 | * Returns the allocation status of the specified sectors. | |
1354 | * | |
1355 | * If 'sector_num' is beyond the end of the disk image the return value is 0 | |
1356 | * and 'pnum' is set to 0. | |
1357 | * | |
1358 | * 'pnum' is set to the number of sectors (including and immediately following | |
1359 | * the specified sector) that are known to be in the same | |
1360 | * allocated/unallocated state. | |
1361 | * | |
1362 | * 'nb_sectors' is the max value 'pnum' should be set to. If nb_sectors goes | |
1363 | * beyond the end of the disk image it will be clamped. | |
1364 | * | |
c1bb86cd | 1365 | * (Based on raw_co_get_block_status() from file-posix.c.) |
947eb203 NV |
1366 | */ |
1367 | static int64_t coroutine_fn qemu_gluster_co_get_block_status( | |
1368 | BlockDriverState *bs, int64_t sector_num, int nb_sectors, int *pnum, | |
1369 | BlockDriverState **file) | |
1370 | { | |
1371 | BDRVGlusterState *s = bs->opaque; | |
1372 | off_t start, data = 0, hole = 0; | |
1373 | int64_t total_size; | |
1374 | int ret = -EINVAL; | |
1375 | ||
1376 | if (!s->fd) { | |
1377 | return ret; | |
1378 | } | |
1379 | ||
1380 | start = sector_num * BDRV_SECTOR_SIZE; | |
1381 | total_size = bdrv_getlength(bs); | |
1382 | if (total_size < 0) { | |
1383 | return total_size; | |
1384 | } else if (start >= total_size) { | |
1385 | *pnum = 0; | |
1386 | return 0; | |
1387 | } else if (start + nb_sectors * BDRV_SECTOR_SIZE > total_size) { | |
1388 | nb_sectors = DIV_ROUND_UP(total_size - start, BDRV_SECTOR_SIZE); | |
1389 | } | |
1390 | ||
1391 | ret = find_allocation(bs, start, &data, &hole); | |
1392 | if (ret == -ENXIO) { | |
1393 | /* Trailing hole */ | |
1394 | *pnum = nb_sectors; | |
1395 | ret = BDRV_BLOCK_ZERO; | |
1396 | } else if (ret < 0) { | |
1397 | /* No info available, so pretend there are no holes */ | |
1398 | *pnum = nb_sectors; | |
1399 | ret = BDRV_BLOCK_DATA; | |
1400 | } else if (data == start) { | |
1401 | /* On a data extent, compute sectors to the end of the extent, | |
1402 | * possibly including a partial sector at EOF. */ | |
1403 | *pnum = MIN(nb_sectors, DIV_ROUND_UP(hole - start, BDRV_SECTOR_SIZE)); | |
1404 | ret = BDRV_BLOCK_DATA; | |
1405 | } else { | |
1406 | /* On a hole, compute sectors to the beginning of the next extent. */ | |
1407 | assert(hole == start); | |
1408 | *pnum = MIN(nb_sectors, (data - start) / BDRV_SECTOR_SIZE); | |
1409 | ret = BDRV_BLOCK_ZERO; | |
1410 | } | |
1411 | ||
1412 | *file = bs; | |
1413 | ||
1414 | return ret | BDRV_BLOCK_OFFSET_VALID | start; | |
1415 | } | |
1416 | ||
1417 | ||
8d6d89cb BR |
1418 | static BlockDriver bdrv_gluster = { |
1419 | .format_name = "gluster", | |
1420 | .protocol_name = "gluster", | |
1421 | .instance_size = sizeof(BDRVGlusterState), | |
6c7189bb | 1422 | .bdrv_needs_filename = false, |
8d6d89cb | 1423 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1424 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1425 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1426 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1427 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1428 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1429 | .bdrv_getlength = qemu_gluster_getlength, |
1430 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1431 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1432 | .bdrv_co_readv = qemu_gluster_co_readv, |
1433 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1434 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1435 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1436 | #ifdef CONFIG_GLUSTERFS_DISCARD |
1014170b | 1437 | .bdrv_co_pdiscard = qemu_gluster_co_pdiscard, |
7c815372 BR |
1438 | #endif |
1439 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1440 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1441 | #endif |
947eb203 | 1442 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1443 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1444 | }; |
1445 | ||
1446 | static BlockDriver bdrv_gluster_tcp = { | |
1447 | .format_name = "gluster", | |
1448 | .protocol_name = "gluster+tcp", | |
1449 | .instance_size = sizeof(BDRVGlusterState), | |
6c7189bb | 1450 | .bdrv_needs_filename = false, |
8d6d89cb | 1451 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1452 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1453 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1454 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1455 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1456 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1457 | .bdrv_getlength = qemu_gluster_getlength, |
1458 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1459 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1460 | .bdrv_co_readv = qemu_gluster_co_readv, |
1461 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1462 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1463 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1464 | #ifdef CONFIG_GLUSTERFS_DISCARD |
1014170b | 1465 | .bdrv_co_pdiscard = qemu_gluster_co_pdiscard, |
7c815372 BR |
1466 | #endif |
1467 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1468 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1469 | #endif |
947eb203 | 1470 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1471 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1472 | }; |
1473 | ||
1474 | static BlockDriver bdrv_gluster_unix = { | |
1475 | .format_name = "gluster", | |
1476 | .protocol_name = "gluster+unix", | |
1477 | .instance_size = sizeof(BDRVGlusterState), | |
030be321 | 1478 | .bdrv_needs_filename = true, |
8d6d89cb | 1479 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1480 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1481 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1482 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1483 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1484 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1485 | .bdrv_getlength = qemu_gluster_getlength, |
1486 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1487 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1488 | .bdrv_co_readv = qemu_gluster_co_readv, |
1489 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1490 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1491 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1492 | #ifdef CONFIG_GLUSTERFS_DISCARD |
1014170b | 1493 | .bdrv_co_pdiscard = qemu_gluster_co_pdiscard, |
7c815372 BR |
1494 | #endif |
1495 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1496 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1497 | #endif |
947eb203 | 1498 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1499 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1500 | }; |
1501 | ||
0552ff24 PKK |
1502 | /* rdma is deprecated (actually never supported for volfile fetch). |
1503 | * Let's maintain it for the protocol compatibility, to make sure things | |
1504 | * won't break immediately. For now, gluster+rdma will fall back to gluster+tcp | |
1505 | * protocol with a warning. | |
1506 | * TODO: remove gluster+rdma interface support | |
1507 | */ | |
8d6d89cb BR |
1508 | static BlockDriver bdrv_gluster_rdma = { |
1509 | .format_name = "gluster", | |
1510 | .protocol_name = "gluster+rdma", | |
1511 | .instance_size = sizeof(BDRVGlusterState), | |
030be321 | 1512 | .bdrv_needs_filename = true, |
8d6d89cb | 1513 | .bdrv_file_open = qemu_gluster_open, |
adccfbcd JC |
1514 | .bdrv_reopen_prepare = qemu_gluster_reopen_prepare, |
1515 | .bdrv_reopen_commit = qemu_gluster_reopen_commit, | |
1516 | .bdrv_reopen_abort = qemu_gluster_reopen_abort, | |
8d6d89cb | 1517 | .bdrv_close = qemu_gluster_close, |
c282e1fd | 1518 | .bdrv_create = qemu_gluster_create, |
8d6d89cb BR |
1519 | .bdrv_getlength = qemu_gluster_getlength, |
1520 | .bdrv_get_allocated_file_size = qemu_gluster_allocated_file_size, | |
42ec24e2 | 1521 | .bdrv_truncate = qemu_gluster_truncate, |
15744b0b BR |
1522 | .bdrv_co_readv = qemu_gluster_co_readv, |
1523 | .bdrv_co_writev = qemu_gluster_co_writev, | |
1524 | .bdrv_co_flush_to_disk = qemu_gluster_co_flush_to_disk, | |
8ab6feec | 1525 | .bdrv_has_zero_init = qemu_gluster_has_zero_init, |
0c14fb47 | 1526 | #ifdef CONFIG_GLUSTERFS_DISCARD |
1014170b | 1527 | .bdrv_co_pdiscard = qemu_gluster_co_pdiscard, |
7c815372 BR |
1528 | #endif |
1529 | #ifdef CONFIG_GLUSTERFS_ZEROFILL | |
e88a36eb | 1530 | .bdrv_co_pwrite_zeroes = qemu_gluster_co_pwrite_zeroes, |
0c14fb47 | 1531 | #endif |
947eb203 | 1532 | .bdrv_co_get_block_status = qemu_gluster_co_get_block_status, |
90c772de | 1533 | .create_opts = &qemu_gluster_create_opts, |
8d6d89cb BR |
1534 | }; |
1535 | ||
1536 | static void bdrv_gluster_init(void) | |
1537 | { | |
1538 | bdrv_register(&bdrv_gluster_rdma); | |
1539 | bdrv_register(&bdrv_gluster_unix); | |
1540 | bdrv_register(&bdrv_gluster_tcp); | |
1541 | bdrv_register(&bdrv_gluster); | |
1542 | } | |
1543 | ||
1544 | block_init(bdrv_gluster_init); |