Commit | Line | Data |
---|---|---|
16725b9d SW |
1 | |
2 | #include "ceph_debug.h" | |
3 | ||
4 | #include <linux/backing-dev.h> | |
5 | #include <linux/fs.h> | |
6 | #include <linux/inet.h> | |
7 | #include <linux/in6.h> | |
8 | #include <linux/module.h> | |
9 | #include <linux/mount.h> | |
10 | #include <linux/parser.h> | |
11 | #include <linux/rwsem.h> | |
12 | #include <linux/sched.h> | |
13 | #include <linux/seq_file.h> | |
14 | #include <linux/statfs.h> | |
15 | #include <linux/string.h> | |
16 | #include <linux/version.h> | |
17 | #include <linux/vmalloc.h> | |
18 | ||
16725b9d SW |
19 | #include "decode.h" |
20 | #include "super.h" | |
21 | #include "mon_client.h" | |
22 | ||
23 | /* | |
24 | * Ceph superblock operations | |
25 | * | |
26 | * Handle the basics of mounting, unmounting. | |
27 | */ | |
28 | ||
29 | ||
30 | /* | |
31 | * find filename portion of a path (/foo/bar/baz -> baz) | |
32 | */ | |
33 | const char *ceph_file_part(const char *s, int len) | |
34 | { | |
35 | const char *e = s + len; | |
36 | ||
37 | while (e != s && *(e-1) != '/') | |
38 | e--; | |
39 | return e; | |
40 | } | |
41 | ||
42 | ||
43 | /* | |
44 | * super ops | |
45 | */ | |
46 | static void ceph_put_super(struct super_block *s) | |
47 | { | |
48 | struct ceph_client *cl = ceph_client(s); | |
49 | ||
50 | dout("put_super\n"); | |
51 | ceph_mdsc_close_sessions(&cl->mdsc); | |
52 | return; | |
53 | } | |
54 | ||
55 | static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf) | |
56 | { | |
57 | struct ceph_client *client = ceph_inode_to_client(dentry->d_inode); | |
58 | struct ceph_monmap *monmap = client->monc.monmap; | |
59 | struct ceph_statfs st; | |
60 | u64 fsid; | |
61 | int err; | |
62 | ||
63 | dout("statfs\n"); | |
64 | err = ceph_monc_do_statfs(&client->monc, &st); | |
65 | if (err < 0) | |
66 | return err; | |
67 | ||
68 | /* fill in kstatfs */ | |
69 | buf->f_type = CEPH_SUPER_MAGIC; /* ?? */ | |
70 | ||
71 | /* | |
72 | * express utilization in terms of large blocks to avoid | |
73 | * overflow on 32-bit machines. | |
74 | */ | |
75 | buf->f_bsize = 1 << CEPH_BLOCK_SHIFT; | |
76 | buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10); | |
77 | buf->f_bfree = (le64_to_cpu(st.kb) - le64_to_cpu(st.kb_used)) >> | |
78 | (CEPH_BLOCK_SHIFT-10); | |
79 | buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10); | |
80 | ||
81 | buf->f_files = le64_to_cpu(st.num_objects); | |
82 | buf->f_ffree = -1; | |
83 | buf->f_namelen = PATH_MAX; | |
84 | buf->f_frsize = PAGE_CACHE_SIZE; | |
85 | ||
86 | /* leave fsid little-endian, regardless of host endianness */ | |
87 | fsid = *(u64 *)(&monmap->fsid) ^ *((u64 *)&monmap->fsid + 1); | |
88 | buf->f_fsid.val[0] = fsid & 0xffffffff; | |
89 | buf->f_fsid.val[1] = fsid >> 32; | |
90 | ||
91 | return 0; | |
92 | } | |
93 | ||
94 | ||
95 | static int ceph_syncfs(struct super_block *sb, int wait) | |
96 | { | |
97 | dout("sync_fs %d\n", wait); | |
98 | ceph_osdc_sync(&ceph_client(sb)->osdc); | |
99 | ceph_mdsc_sync(&ceph_client(sb)->mdsc); | |
f2cf418c | 100 | dout("sync_fs %d done\n", wait); |
16725b9d SW |
101 | return 0; |
102 | } | |
103 | ||
104 | ||
105 | /** | |
106 | * ceph_show_options - Show mount options in /proc/mounts | |
107 | * @m: seq_file to write to | |
108 | * @mnt: mount descriptor | |
109 | */ | |
110 | static int ceph_show_options(struct seq_file *m, struct vfsmount *mnt) | |
111 | { | |
112 | struct ceph_client *client = ceph_sb_to_client(mnt->mnt_sb); | |
113 | struct ceph_mount_args *args = &client->mount_args; | |
114 | ||
115 | if (args->flags & CEPH_OPT_FSID) | |
116 | seq_printf(m, ",fsidmajor=%llu,fsidminor%llu", | |
117 | le64_to_cpu(*(__le64 *)&args->fsid.fsid[0]), | |
118 | le64_to_cpu(*(__le64 *)&args->fsid.fsid[8])); | |
119 | if (args->flags & CEPH_OPT_NOSHARE) | |
120 | seq_puts(m, ",noshare"); | |
121 | if (args->flags & CEPH_OPT_DIRSTAT) | |
122 | seq_puts(m, ",dirstat"); | |
123 | if ((args->flags & CEPH_OPT_RBYTES) == 0) | |
124 | seq_puts(m, ",norbytes"); | |
125 | if (args->flags & CEPH_OPT_NOCRC) | |
126 | seq_puts(m, ",nocrc"); | |
127 | if (args->flags & CEPH_OPT_NOASYNCREADDIR) | |
128 | seq_puts(m, ",noasyncreaddir"); | |
129 | if (strcmp(args->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT)) | |
130 | seq_printf(m, ",snapdirname=%s", args->snapdir_name); | |
131 | if (args->secret) | |
132 | seq_puts(m, ",secret=<hidden>"); | |
133 | return 0; | |
134 | } | |
135 | ||
136 | /* | |
137 | * caches | |
138 | */ | |
139 | struct kmem_cache *ceph_inode_cachep; | |
140 | struct kmem_cache *ceph_cap_cachep; | |
141 | struct kmem_cache *ceph_dentry_cachep; | |
142 | struct kmem_cache *ceph_file_cachep; | |
143 | ||
144 | static void ceph_inode_init_once(void *foo) | |
145 | { | |
146 | struct ceph_inode_info *ci = foo; | |
147 | inode_init_once(&ci->vfs_inode); | |
148 | } | |
149 | ||
150 | static int __init init_caches(void) | |
151 | { | |
152 | ceph_inode_cachep = kmem_cache_create("ceph_inode_info", | |
153 | sizeof(struct ceph_inode_info), | |
154 | __alignof__(struct ceph_inode_info), | |
155 | (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), | |
156 | ceph_inode_init_once); | |
157 | if (ceph_inode_cachep == NULL) | |
158 | return -ENOMEM; | |
159 | ||
160 | ceph_cap_cachep = KMEM_CACHE(ceph_cap, | |
161 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
162 | if (ceph_cap_cachep == NULL) | |
163 | goto bad_cap; | |
164 | ||
165 | ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info, | |
166 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
167 | if (ceph_dentry_cachep == NULL) | |
168 | goto bad_dentry; | |
169 | ||
170 | ceph_file_cachep = KMEM_CACHE(ceph_file_info, | |
171 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
172 | if (ceph_file_cachep == NULL) | |
173 | goto bad_file; | |
174 | ||
175 | return 0; | |
176 | ||
177 | bad_file: | |
178 | kmem_cache_destroy(ceph_dentry_cachep); | |
179 | bad_dentry: | |
180 | kmem_cache_destroy(ceph_cap_cachep); | |
181 | bad_cap: | |
182 | kmem_cache_destroy(ceph_inode_cachep); | |
183 | return -ENOMEM; | |
184 | } | |
185 | ||
186 | static void destroy_caches(void) | |
187 | { | |
188 | kmem_cache_destroy(ceph_inode_cachep); | |
189 | kmem_cache_destroy(ceph_cap_cachep); | |
190 | kmem_cache_destroy(ceph_dentry_cachep); | |
191 | kmem_cache_destroy(ceph_file_cachep); | |
192 | } | |
193 | ||
194 | ||
195 | /* | |
196 | * ceph_umount_begin - initiate forced umount. Tear down down the | |
197 | * mount, skipping steps that may hang while waiting for server(s). | |
198 | */ | |
199 | static void ceph_umount_begin(struct super_block *sb) | |
200 | { | |
201 | struct ceph_client *client = ceph_sb_to_client(sb); | |
202 | ||
203 | dout("ceph_umount_begin - starting forced umount\n"); | |
204 | if (!client) | |
205 | return; | |
206 | client->mount_state = CEPH_MOUNT_SHUTDOWN; | |
207 | return; | |
208 | } | |
209 | ||
210 | static const struct super_operations ceph_super_ops = { | |
211 | .alloc_inode = ceph_alloc_inode, | |
212 | .destroy_inode = ceph_destroy_inode, | |
213 | .write_inode = ceph_write_inode, | |
214 | .sync_fs = ceph_syncfs, | |
215 | .put_super = ceph_put_super, | |
216 | .show_options = ceph_show_options, | |
217 | .statfs = ceph_statfs, | |
218 | .umount_begin = ceph_umount_begin, | |
219 | }; | |
220 | ||
221 | ||
222 | const char *ceph_msg_type_name(int type) | |
223 | { | |
224 | switch (type) { | |
225 | case CEPH_MSG_SHUTDOWN: return "shutdown"; | |
226 | case CEPH_MSG_PING: return "ping"; | |
227 | case CEPH_MSG_MON_MAP: return "mon_map"; | |
228 | case CEPH_MSG_MON_GET_MAP: return "mon_get_map"; | |
229 | case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe"; | |
230 | case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack"; | |
231 | case CEPH_MSG_CLIENT_MOUNT: return "client_mount"; | |
232 | case CEPH_MSG_CLIENT_MOUNT_ACK: return "client_mount_ack"; | |
233 | case CEPH_MSG_STATFS: return "statfs"; | |
234 | case CEPH_MSG_STATFS_REPLY: return "statfs_reply"; | |
16725b9d SW |
235 | case CEPH_MSG_MDS_MAP: return "mds_map"; |
236 | case CEPH_MSG_CLIENT_SESSION: return "client_session"; | |
237 | case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect"; | |
238 | case CEPH_MSG_CLIENT_REQUEST: return "client_request"; | |
239 | case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward"; | |
240 | case CEPH_MSG_CLIENT_REPLY: return "client_reply"; | |
241 | case CEPH_MSG_CLIENT_CAPS: return "client_caps"; | |
242 | case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release"; | |
243 | case CEPH_MSG_CLIENT_SNAP: return "client_snap"; | |
244 | case CEPH_MSG_CLIENT_LEASE: return "client_lease"; | |
16725b9d SW |
245 | case CEPH_MSG_OSD_MAP: return "osd_map"; |
246 | case CEPH_MSG_OSD_OP: return "osd_op"; | |
247 | case CEPH_MSG_OSD_OPREPLY: return "osd_opreply"; | |
248 | default: return "unknown"; | |
249 | } | |
250 | } | |
251 | ||
252 | ||
253 | /* | |
254 | * mount options | |
255 | */ | |
256 | enum { | |
257 | Opt_fsidmajor, | |
258 | Opt_fsidminor, | |
259 | Opt_monport, | |
260 | Opt_wsize, | |
261 | Opt_rsize, | |
262 | Opt_osdtimeout, | |
263 | Opt_mount_timeout, | |
264 | Opt_caps_wanted_delay_min, | |
265 | Opt_caps_wanted_delay_max, | |
266 | Opt_readdir_max_entries, | |
267 | /* int args above */ | |
268 | Opt_snapdirname, | |
269 | Opt_secret, | |
270 | /* string args above */ | |
271 | Opt_ip, | |
272 | Opt_noshare, | |
273 | Opt_dirstat, | |
274 | Opt_nodirstat, | |
275 | Opt_rbytes, | |
276 | Opt_norbytes, | |
277 | Opt_nocrc, | |
278 | Opt_noasyncreaddir, | |
279 | }; | |
280 | ||
281 | static match_table_t arg_tokens = { | |
282 | {Opt_fsidmajor, "fsidmajor=%ld"}, | |
283 | {Opt_fsidminor, "fsidminor=%ld"}, | |
284 | {Opt_monport, "monport=%d"}, | |
285 | {Opt_wsize, "wsize=%d"}, | |
286 | {Opt_rsize, "rsize=%d"}, | |
287 | {Opt_osdtimeout, "osdtimeout=%d"}, | |
288 | {Opt_mount_timeout, "mount_timeout=%d"}, | |
289 | {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"}, | |
290 | {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"}, | |
291 | {Opt_readdir_max_entries, "readdir_max_entries=%d"}, | |
292 | /* int args above */ | |
293 | {Opt_snapdirname, "snapdirname=%s"}, | |
294 | {Opt_secret, "secret=%s"}, | |
295 | /* string args above */ | |
296 | {Opt_ip, "ip=%s"}, | |
297 | {Opt_noshare, "noshare"}, | |
298 | {Opt_dirstat, "dirstat"}, | |
299 | {Opt_nodirstat, "nodirstat"}, | |
300 | {Opt_rbytes, "rbytes"}, | |
301 | {Opt_norbytes, "norbytes"}, | |
302 | {Opt_nocrc, "nocrc"}, | |
303 | {Opt_noasyncreaddir, "noasyncreaddir"}, | |
304 | {-1, NULL} | |
305 | }; | |
306 | ||
307 | ||
308 | static int parse_mount_args(struct ceph_client *client, | |
309 | int flags, char *options, const char *dev_name, | |
310 | const char **path) | |
311 | { | |
312 | struct ceph_mount_args *args = &client->mount_args; | |
313 | const char *c; | |
314 | int err; | |
315 | substring_t argstr[MAX_OPT_ARGS]; | |
316 | int num_mon; | |
7b813c46 | 317 | struct ceph_entity_addr *mon_addr; |
16725b9d SW |
318 | int i; |
319 | ||
320 | dout("parse_mount_args dev_name '%s'\n", dev_name); | |
321 | memset(args, 0, sizeof(*args)); | |
322 | ||
7b813c46 SW |
323 | mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*mon_addr), GFP_KERNEL); |
324 | if (!mon_addr) | |
325 | return -ENOMEM; | |
326 | ||
16725b9d SW |
327 | /* start with defaults */ |
328 | args->sb_flags = flags; | |
329 | args->flags = CEPH_OPT_DEFAULT; | |
330 | args->osd_timeout = 5; /* seconds */ | |
331 | args->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */ | |
332 | args->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT; | |
333 | args->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT; | |
8fa97655 | 334 | args->rsize = CEPH_MOUNT_RSIZE_DEFAULT; |
16725b9d SW |
335 | args->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL); |
336 | args->cap_release_safety = CEPH_CAPS_PER_RELEASE * 4; | |
337 | args->max_readdir = 1024; | |
338 | ||
339 | /* ip1[:port1][,ip2[:port2]...]:/subdir/in/fs */ | |
7b813c46 | 340 | err = -EINVAL; |
16725b9d | 341 | if (!dev_name) |
7b813c46 | 342 | goto out; |
16725b9d SW |
343 | *path = strstr(dev_name, ":/"); |
344 | if (*path == NULL) { | |
345 | pr_err("device name is missing path (no :/ in %s)\n", | |
346 | dev_name); | |
7b813c46 | 347 | goto out; |
16725b9d SW |
348 | } |
349 | ||
350 | /* get mon ip(s) */ | |
351 | err = ceph_parse_ips(dev_name, *path, mon_addr, | |
ecb19c46 | 352 | CEPH_MAX_MON, &num_mon); |
16725b9d | 353 | if (err < 0) |
7b813c46 | 354 | goto out; |
16725b9d SW |
355 | |
356 | /* build initial monmap */ | |
7b813c46 | 357 | err = -ENOMEM; |
16725b9d SW |
358 | client->monc.monmap = kzalloc(sizeof(*client->monc.monmap) + |
359 | num_mon*sizeof(client->monc.monmap->mon_inst[0]), | |
360 | GFP_KERNEL); | |
361 | if (!client->monc.monmap) | |
7b813c46 | 362 | goto out; |
16725b9d SW |
363 | for (i = 0; i < num_mon; i++) { |
364 | client->monc.monmap->mon_inst[i].addr = mon_addr[i]; | |
365 | client->monc.monmap->mon_inst[i].addr.erank = 0; | |
366 | client->monc.monmap->mon_inst[i].addr.nonce = 0; | |
367 | client->monc.monmap->mon_inst[i].name.type = | |
368 | CEPH_ENTITY_TYPE_MON; | |
369 | client->monc.monmap->mon_inst[i].name.num = cpu_to_le64(i); | |
370 | } | |
371 | client->monc.monmap->num_mon = num_mon; | |
372 | memset(&args->my_addr.in_addr, 0, sizeof(args->my_addr.in_addr)); | |
373 | ||
374 | /* path on server */ | |
375 | *path += 2; | |
376 | dout("server path '%s'\n", *path); | |
377 | ||
378 | /* parse mount options */ | |
379 | while ((c = strsep(&options, ",")) != NULL) { | |
380 | int token, intval, ret; | |
381 | if (!*c) | |
382 | continue; | |
7b813c46 | 383 | err = -EINVAL; |
16725b9d SW |
384 | token = match_token((char *)c, arg_tokens, argstr); |
385 | if (token < 0) { | |
386 | pr_err("bad mount option at '%s'\n", c); | |
7b813c46 | 387 | goto out; |
16725b9d SW |
388 | } |
389 | if (token < Opt_ip) { | |
390 | ret = match_int(&argstr[0], &intval); | |
391 | if (ret < 0) { | |
392 | pr_err("bad mount option arg (not int) " | |
393 | "at '%s'\n", c); | |
394 | continue; | |
395 | } | |
396 | dout("got token %d intval %d\n", token, intval); | |
397 | } | |
398 | switch (token) { | |
399 | case Opt_fsidmajor: | |
400 | *(__le64 *)&args->fsid.fsid[0] = cpu_to_le64(intval); | |
401 | break; | |
402 | case Opt_fsidminor: | |
403 | *(__le64 *)&args->fsid.fsid[8] = cpu_to_le64(intval); | |
404 | break; | |
405 | case Opt_ip: | |
406 | err = ceph_parse_ips(argstr[0].from, | |
407 | argstr[0].to, | |
408 | &args->my_addr, | |
409 | 1, NULL); | |
410 | if (err < 0) | |
411 | return err; | |
412 | args->flags |= CEPH_OPT_MYIP; | |
413 | break; | |
414 | ||
415 | case Opt_snapdirname: | |
416 | kfree(args->snapdir_name); | |
417 | args->snapdir_name = kstrndup(argstr[0].from, | |
418 | argstr[0].to-argstr[0].from, | |
419 | GFP_KERNEL); | |
420 | break; | |
421 | case Opt_secret: | |
422 | args->secret = kstrndup(argstr[0].from, | |
423 | argstr[0].to-argstr[0].from, | |
424 | GFP_KERNEL); | |
425 | break; | |
426 | ||
427 | /* misc */ | |
428 | case Opt_wsize: | |
429 | args->wsize = intval; | |
430 | break; | |
431 | case Opt_rsize: | |
432 | args->rsize = intval; | |
433 | break; | |
434 | case Opt_osdtimeout: | |
435 | args->osd_timeout = intval; | |
436 | break; | |
437 | case Opt_mount_timeout: | |
438 | args->mount_timeout = intval; | |
439 | break; | |
440 | case Opt_caps_wanted_delay_min: | |
441 | args->caps_wanted_delay_min = intval; | |
442 | break; | |
443 | case Opt_caps_wanted_delay_max: | |
444 | args->caps_wanted_delay_max = intval; | |
445 | break; | |
446 | case Opt_readdir_max_entries: | |
447 | args->max_readdir = intval; | |
448 | break; | |
449 | ||
450 | case Opt_noshare: | |
451 | args->flags |= CEPH_OPT_NOSHARE; | |
452 | break; | |
453 | ||
454 | case Opt_dirstat: | |
455 | args->flags |= CEPH_OPT_DIRSTAT; | |
456 | break; | |
457 | case Opt_nodirstat: | |
458 | args->flags &= ~CEPH_OPT_DIRSTAT; | |
459 | break; | |
460 | case Opt_rbytes: | |
461 | args->flags |= CEPH_OPT_RBYTES; | |
462 | break; | |
463 | case Opt_norbytes: | |
464 | args->flags &= ~CEPH_OPT_RBYTES; | |
465 | break; | |
466 | case Opt_nocrc: | |
467 | args->flags |= CEPH_OPT_NOCRC; | |
468 | break; | |
469 | case Opt_noasyncreaddir: | |
470 | args->flags |= CEPH_OPT_NOASYNCREADDIR; | |
471 | break; | |
472 | ||
473 | default: | |
474 | BUG_ON(token); | |
475 | } | |
476 | } | |
7b813c46 | 477 | err = 0; |
16725b9d | 478 | |
7b813c46 SW |
479 | out: |
480 | kfree(mon_addr); | |
481 | return err; | |
16725b9d SW |
482 | } |
483 | ||
484 | static void release_mount_args(struct ceph_mount_args *args) | |
485 | { | |
486 | kfree(args->snapdir_name); | |
487 | args->snapdir_name = NULL; | |
488 | kfree(args->secret); | |
489 | args->secret = NULL; | |
490 | } | |
491 | ||
492 | /* | |
493 | * create a fresh client instance | |
494 | */ | |
495 | static struct ceph_client *ceph_create_client(void) | |
496 | { | |
497 | struct ceph_client *client; | |
498 | int err = -ENOMEM; | |
499 | ||
500 | client = kzalloc(sizeof(*client), GFP_KERNEL); | |
501 | if (client == NULL) | |
502 | return ERR_PTR(-ENOMEM); | |
503 | ||
504 | mutex_init(&client->mount_mutex); | |
505 | ||
506 | init_waitqueue_head(&client->mount_wq); | |
507 | ||
508 | client->sb = NULL; | |
509 | client->mount_state = CEPH_MOUNT_MOUNTING; | |
510 | client->whoami = -1; | |
511 | ||
512 | client->msgr = NULL; | |
513 | ||
514 | client->mount_err = 0; | |
515 | client->signed_ticket = NULL; | |
516 | client->signed_ticket_len = 0; | |
517 | ||
518 | err = -ENOMEM; | |
519 | client->wb_wq = create_workqueue("ceph-writeback"); | |
520 | if (client->wb_wq == NULL) | |
521 | goto fail; | |
522 | client->pg_inv_wq = create_singlethread_workqueue("ceph-pg-invalid"); | |
523 | if (client->pg_inv_wq == NULL) | |
524 | goto fail_wb_wq; | |
525 | client->trunc_wq = create_singlethread_workqueue("ceph-trunc"); | |
526 | if (client->trunc_wq == NULL) | |
527 | goto fail_pg_inv_wq; | |
528 | ||
529 | /* subsystems */ | |
530 | err = ceph_monc_init(&client->monc, client); | |
531 | if (err < 0) | |
532 | goto fail_trunc_wq; | |
533 | err = ceph_osdc_init(&client->osdc, client); | |
534 | if (err < 0) | |
535 | goto fail_monc; | |
536 | ceph_mdsc_init(&client->mdsc, client); | |
537 | return client; | |
538 | ||
539 | fail_monc: | |
540 | ceph_monc_stop(&client->monc); | |
541 | fail_trunc_wq: | |
542 | destroy_workqueue(client->trunc_wq); | |
543 | fail_pg_inv_wq: | |
544 | destroy_workqueue(client->pg_inv_wq); | |
545 | fail_wb_wq: | |
546 | destroy_workqueue(client->wb_wq); | |
547 | fail: | |
548 | kfree(client); | |
549 | return ERR_PTR(err); | |
550 | } | |
551 | ||
552 | static void ceph_destroy_client(struct ceph_client *client) | |
553 | { | |
554 | dout("destroy_client %p\n", client); | |
555 | ||
556 | /* unmount */ | |
557 | ceph_mdsc_stop(&client->mdsc); | |
558 | ceph_monc_stop(&client->monc); | |
559 | ceph_osdc_stop(&client->osdc); | |
560 | ||
561 | kfree(client->signed_ticket); | |
562 | ||
563 | ceph_debugfs_client_cleanup(client); | |
564 | destroy_workqueue(client->wb_wq); | |
565 | destroy_workqueue(client->pg_inv_wq); | |
566 | destroy_workqueue(client->trunc_wq); | |
567 | ||
568 | if (client->msgr) | |
569 | ceph_messenger_destroy(client->msgr); | |
570 | if (client->wb_pagevec_pool) | |
571 | mempool_destroy(client->wb_pagevec_pool); | |
572 | ||
573 | release_mount_args(&client->mount_args); | |
574 | ||
575 | kfree(client); | |
576 | dout("destroy_client %p done\n", client); | |
577 | } | |
578 | ||
579 | /* | |
580 | * true if we have the mon map (and have thus joined the cluster) | |
581 | */ | |
582 | static int have_mon_map(struct ceph_client *client) | |
583 | { | |
584 | return client->monc.monmap && client->monc.monmap->epoch; | |
585 | } | |
586 | ||
587 | /* | |
588 | * Bootstrap mount by opening the root directory. Note the mount | |
589 | * @started time from caller, and time out if this takes too long. | |
590 | */ | |
591 | static struct dentry *open_root_dentry(struct ceph_client *client, | |
592 | const char *path, | |
593 | unsigned long started) | |
594 | { | |
595 | struct ceph_mds_client *mdsc = &client->mdsc; | |
596 | struct ceph_mds_request *req = NULL; | |
597 | int err; | |
598 | struct dentry *root; | |
599 | ||
600 | /* open dir */ | |
601 | dout("open_root_inode opening '%s'\n", path); | |
602 | req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS); | |
603 | if (IS_ERR(req)) | |
604 | return ERR_PTR(PTR_ERR(req)); | |
605 | req->r_path1 = kstrdup(path, GFP_NOFS); | |
606 | req->r_ino1.ino = CEPH_INO_ROOT; | |
607 | req->r_ino1.snap = CEPH_NOSNAP; | |
608 | req->r_started = started; | |
609 | req->r_timeout = client->mount_args.mount_timeout * HZ; | |
610 | req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE); | |
611 | req->r_num_caps = 2; | |
612 | err = ceph_mdsc_do_request(mdsc, NULL, req); | |
613 | if (err == 0) { | |
614 | dout("open_root_inode success\n"); | |
615 | if (ceph_ino(req->r_target_inode) == CEPH_INO_ROOT && | |
616 | client->sb->s_root == NULL) | |
617 | root = d_alloc_root(req->r_target_inode); | |
618 | else | |
619 | root = d_obtain_alias(req->r_target_inode); | |
620 | req->r_target_inode = NULL; | |
621 | dout("open_root_inode success, root dentry is %p\n", root); | |
622 | } else { | |
623 | root = ERR_PTR(err); | |
624 | } | |
625 | ceph_mdsc_put_request(req); | |
626 | return root; | |
627 | } | |
628 | ||
629 | /* | |
630 | * mount: join the ceph cluster, and open root directory. | |
631 | */ | |
632 | static int ceph_mount(struct ceph_client *client, struct vfsmount *mnt, | |
633 | const char *path) | |
634 | { | |
635 | struct ceph_entity_addr *myaddr = NULL; | |
636 | int err; | |
637 | unsigned long timeout = client->mount_args.mount_timeout * HZ; | |
638 | unsigned long started = jiffies; /* note the start time */ | |
639 | struct dentry *root; | |
640 | ||
641 | dout("mount start\n"); | |
642 | mutex_lock(&client->mount_mutex); | |
643 | ||
644 | /* initialize the messenger */ | |
645 | if (client->msgr == NULL) { | |
646 | if (ceph_test_opt(client, MYIP)) | |
647 | myaddr = &client->mount_args.my_addr; | |
648 | client->msgr = ceph_messenger_create(myaddr); | |
649 | if (IS_ERR(client->msgr)) { | |
650 | err = PTR_ERR(client->msgr); | |
651 | client->msgr = NULL; | |
652 | goto out; | |
653 | } | |
654 | client->msgr->nocrc = ceph_test_opt(client, NOCRC); | |
655 | } | |
656 | ||
657 | /* send mount request, and wait for mon, mds, and osd maps */ | |
658 | err = ceph_monc_request_mount(&client->monc); | |
659 | if (err < 0) | |
660 | goto out; | |
661 | ||
662 | while (!have_mon_map(client) && !client->mount_err) { | |
663 | err = -EIO; | |
664 | if (timeout && time_after_eq(jiffies, started + timeout)) | |
665 | goto out; | |
666 | ||
667 | /* wait */ | |
668 | dout("mount waiting for mount\n"); | |
669 | err = wait_event_interruptible_timeout(client->mount_wq, | |
670 | client->mount_err || have_mon_map(client), | |
671 | timeout); | |
672 | if (err == -EINTR || err == -ERESTARTSYS) | |
673 | goto out; | |
674 | if (client->mount_err) { | |
675 | err = client->mount_err; | |
676 | goto out; | |
677 | } | |
678 | } | |
679 | ||
680 | dout("mount opening root\n"); | |
681 | root = open_root_dentry(client, "", started); | |
682 | if (IS_ERR(root)) { | |
683 | err = PTR_ERR(root); | |
684 | goto out; | |
685 | } | |
686 | if (client->sb->s_root) | |
687 | dput(root); | |
688 | else | |
689 | client->sb->s_root = root; | |
690 | ||
691 | if (path[0] == 0) { | |
692 | dget(root); | |
693 | } else { | |
694 | dout("mount opening base mountpoint\n"); | |
695 | root = open_root_dentry(client, path, started); | |
696 | if (IS_ERR(root)) { | |
697 | err = PTR_ERR(root); | |
698 | dput(client->sb->s_root); | |
699 | client->sb->s_root = NULL; | |
700 | goto out; | |
701 | } | |
702 | } | |
703 | ||
704 | mnt->mnt_root = root; | |
705 | mnt->mnt_sb = client->sb; | |
706 | ||
707 | client->mount_state = CEPH_MOUNT_MOUNTED; | |
708 | dout("mount success\n"); | |
709 | err = 0; | |
710 | ||
711 | out: | |
712 | mutex_unlock(&client->mount_mutex); | |
713 | return err; | |
714 | } | |
715 | ||
716 | static int ceph_set_super(struct super_block *s, void *data) | |
717 | { | |
718 | struct ceph_client *client = data; | |
719 | int ret; | |
720 | ||
721 | dout("set_super %p data %p\n", s, data); | |
722 | ||
723 | s->s_flags = client->mount_args.sb_flags; | |
724 | s->s_maxbytes = 1ULL << 40; /* temp value until we get mdsmap */ | |
725 | ||
726 | s->s_fs_info = client; | |
727 | client->sb = s; | |
728 | ||
729 | s->s_op = &ceph_super_ops; | |
730 | s->s_export_op = &ceph_export_ops; | |
731 | ||
732 | s->s_time_gran = 1000; /* 1000 ns == 1 us */ | |
733 | ||
734 | ret = set_anon_super(s, NULL); /* what is that second arg for? */ | |
735 | if (ret != 0) | |
736 | goto fail; | |
737 | ||
738 | return ret; | |
739 | ||
740 | fail: | |
741 | s->s_fs_info = NULL; | |
742 | client->sb = NULL; | |
743 | return ret; | |
744 | } | |
745 | ||
746 | /* | |
747 | * share superblock if same fs AND options | |
748 | */ | |
749 | static int ceph_compare_super(struct super_block *sb, void *data) | |
750 | { | |
751 | struct ceph_client *new = data; | |
752 | struct ceph_mount_args *args = &new->mount_args; | |
753 | struct ceph_client *other = ceph_sb_to_client(sb); | |
754 | int i; | |
755 | ||
756 | dout("ceph_compare_super %p\n", sb); | |
757 | if (args->flags & CEPH_OPT_FSID) { | |
758 | if (ceph_fsid_compare(&args->fsid, &other->fsid)) { | |
759 | dout("fsid doesn't match\n"); | |
760 | return 0; | |
761 | } | |
762 | } else { | |
763 | /* do we share (a) monitor? */ | |
764 | for (i = 0; i < new->monc.monmap->num_mon; i++) | |
765 | if (ceph_monmap_contains(other->monc.monmap, | |
766 | &new->monc.monmap->mon_inst[i].addr)) | |
767 | break; | |
768 | if (i == new->monc.monmap->num_mon) { | |
769 | dout("mon ip not part of monmap\n"); | |
770 | return 0; | |
771 | } | |
772 | dout("mon ip matches existing sb %p\n", sb); | |
773 | } | |
774 | if (args->sb_flags != other->mount_args.sb_flags) { | |
775 | dout("flags differ\n"); | |
776 | return 0; | |
777 | } | |
778 | return 1; | |
779 | } | |
780 | ||
781 | /* | |
782 | * construct our own bdi so we can control readahead, etc. | |
783 | */ | |
784 | static int ceph_init_bdi(struct super_block *sb, struct ceph_client *client) | |
785 | { | |
786 | int err; | |
787 | ||
788 | err = bdi_init(&client->backing_dev_info); | |
789 | if (err < 0) | |
790 | return err; | |
f2cf418c | 791 | sb->s_bdi = &client->backing_dev_info; |
16725b9d SW |
792 | |
793 | /* set ra_pages based on rsize mount option? */ | |
794 | if (client->mount_args.rsize >= PAGE_CACHE_SIZE) | |
795 | client->backing_dev_info.ra_pages = | |
796 | (client->mount_args.rsize + PAGE_CACHE_SIZE - 1) | |
797 | >> PAGE_SHIFT; | |
798 | ||
799 | err = bdi_register_dev(&client->backing_dev_info, sb->s_dev); | |
800 | return err; | |
801 | } | |
802 | ||
803 | static int ceph_get_sb(struct file_system_type *fs_type, | |
804 | int flags, const char *dev_name, void *data, | |
805 | struct vfsmount *mnt) | |
806 | { | |
807 | struct super_block *sb; | |
808 | struct ceph_client *client; | |
809 | int err; | |
810 | int (*compare_super)(struct super_block *, void *) = ceph_compare_super; | |
6ca874e9 | 811 | const char *path = 0; |
16725b9d SW |
812 | |
813 | dout("ceph_get_sb\n"); | |
814 | ||
815 | /* create client (which we may/may not use) */ | |
816 | client = ceph_create_client(); | |
817 | if (IS_ERR(client)) | |
818 | return PTR_ERR(client); | |
819 | ||
820 | err = parse_mount_args(client, flags, data, dev_name, &path); | |
821 | if (err < 0) | |
822 | goto out; | |
823 | ||
824 | if (client->mount_args.flags & CEPH_OPT_NOSHARE) | |
825 | compare_super = NULL; | |
826 | sb = sget(fs_type, compare_super, ceph_set_super, client); | |
827 | if (IS_ERR(sb)) { | |
828 | err = PTR_ERR(sb); | |
829 | goto out; | |
830 | } | |
831 | ||
832 | if (ceph_client(sb) != client) { | |
833 | ceph_destroy_client(client); | |
834 | client = ceph_client(sb); | |
835 | dout("get_sb got existing client %p\n", client); | |
836 | } else { | |
837 | dout("get_sb using new client %p\n", client); | |
838 | ||
839 | /* set up mempools */ | |
840 | err = -ENOMEM; | |
841 | client->wb_pagevec_pool = mempool_create_kmalloc_pool(10, | |
842 | client->mount_args.wsize >> PAGE_CACHE_SHIFT); | |
843 | if (!client->wb_pagevec_pool) | |
844 | goto out_splat; | |
845 | ||
846 | err = ceph_init_bdi(sb, client); | |
847 | if (err < 0) | |
848 | goto out_splat; | |
849 | } | |
850 | ||
851 | err = ceph_mount(client, mnt, path); | |
852 | if (err < 0) | |
853 | goto out_splat; | |
854 | dout("root %p inode %p ino %llx.%llx\n", mnt->mnt_root, | |
855 | mnt->mnt_root->d_inode, ceph_vinop(mnt->mnt_root->d_inode)); | |
856 | return 0; | |
857 | ||
858 | out_splat: | |
859 | ceph_mdsc_close_sessions(&client->mdsc); | |
860 | up_write(&sb->s_umount); | |
861 | deactivate_super(sb); | |
862 | goto out_final; | |
863 | ||
864 | out: | |
865 | ceph_destroy_client(client); | |
866 | out_final: | |
867 | dout("ceph_get_sb fail %d\n", err); | |
868 | return err; | |
869 | } | |
870 | ||
871 | static void ceph_kill_sb(struct super_block *s) | |
872 | { | |
873 | struct ceph_client *client = ceph_sb_to_client(s); | |
874 | dout("kill_sb %p\n", s); | |
875 | ceph_mdsc_pre_umount(&client->mdsc); | |
16725b9d | 876 | kill_anon_super(s); /* will call put_super after sb is r/o */ |
f2cf418c | 877 | bdi_unregister(&client->backing_dev_info); |
16725b9d SW |
878 | bdi_destroy(&client->backing_dev_info); |
879 | ceph_destroy_client(client); | |
880 | } | |
881 | ||
882 | static struct file_system_type ceph_fs_type = { | |
883 | .owner = THIS_MODULE, | |
884 | .name = "ceph", | |
885 | .get_sb = ceph_get_sb, | |
886 | .kill_sb = ceph_kill_sb, | |
887 | .fs_flags = FS_RENAME_DOES_D_MOVE, | |
888 | }; | |
889 | ||
890 | #define _STRINGIFY(x) #x | |
891 | #define STRINGIFY(x) _STRINGIFY(x) | |
892 | ||
893 | static int __init init_ceph(void) | |
894 | { | |
895 | int ret = 0; | |
896 | ||
897 | ret = ceph_debugfs_init(); | |
898 | if (ret < 0) | |
899 | goto out; | |
900 | ||
901 | ret = ceph_msgr_init(); | |
902 | if (ret < 0) | |
903 | goto out_debugfs; | |
904 | ||
905 | ret = init_caches(); | |
906 | if (ret) | |
907 | goto out_msgr; | |
908 | ||
909 | ceph_caps_init(); | |
910 | ||
911 | ret = register_filesystem(&ceph_fs_type); | |
912 | if (ret) | |
913 | goto out_icache; | |
914 | ||
fa0b72e9 SW |
915 | pr_info("loaded %d.%d.%d (mon/mds/osd proto %d/%d/%d)\n", |
916 | CEPH_VERSION_MAJOR, CEPH_VERSION_MINOR, CEPH_VERSION_PATCH, | |
917 | CEPH_MONC_PROTOCOL, CEPH_MDSC_PROTOCOL, CEPH_OSDC_PROTOCOL); | |
16725b9d SW |
918 | return 0; |
919 | ||
920 | out_icache: | |
921 | destroy_caches(); | |
922 | out_msgr: | |
923 | ceph_msgr_exit(); | |
924 | out_debugfs: | |
925 | ceph_debugfs_cleanup(); | |
926 | out: | |
927 | return ret; | |
928 | } | |
929 | ||
930 | static void __exit exit_ceph(void) | |
931 | { | |
932 | dout("exit_ceph\n"); | |
933 | unregister_filesystem(&ceph_fs_type); | |
934 | ceph_caps_finalize(); | |
935 | destroy_caches(); | |
936 | ceph_msgr_exit(); | |
937 | ceph_debugfs_cleanup(); | |
938 | } | |
939 | ||
940 | module_init(init_ceph); | |
941 | module_exit(exit_ceph); | |
942 | ||
943 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
944 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
945 | MODULE_AUTHOR("Patience Warnick <patience@newdream.net>"); | |
946 | MODULE_DESCRIPTION("Ceph filesystem for Linux"); | |
947 | MODULE_LICENSE("GPL"); |