1
2#include <linux/ceph/ceph_debug.h>
3
4#include <linux/backing-dev.h>
5#include <linux/ctype.h>
6#include <linux/fs.h>
7#include <linux/inet.h>
8#include <linux/in6.h>
9#include <linux/module.h>
10#include <linux/mount.h>
11#include <linux/parser.h>
12#include <linux/sched.h>
13#include <linux/seq_file.h>
14#include <linux/slab.h>
15#include <linux/statfs.h>
16#include <linux/string.h>
17
18#include "super.h"
19#include "mds_client.h"
20
21#include <linux/ceph/ceph_features.h>
22#include <linux/ceph/decode.h>
23#include <linux/ceph/mon_client.h>
24#include <linux/ceph/auth.h>
25#include <linux/ceph/debugfs.h>
26
27
28
29
30
31
32
33
34
35
36static void ceph_put_super(struct super_block *s)
37{
38 struct ceph_fs_client *fsc = ceph_sb_to_client(s);
39
40 dout("put_super\n");
41 ceph_mdsc_close_sessions(fsc->mdsc);
42
43
44
45
46
47 if (s->s_bdi == &fsc->backing_dev_info) {
48 bdi_unregister(&fsc->backing_dev_info);
49 s->s_bdi = NULL;
50 }
51
52 return;
53}
54
55static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
56{
57 struct ceph_fs_client *fsc = ceph_inode_to_client(dentry->d_inode);
58 struct ceph_monmap *monmap = fsc->client->monc.monmap;
59 struct ceph_statfs st;
60 u64 fsid;
61 int err;
62 u64 data_pool;
63
64 if (fsc->mdsc->mdsmap->m_num_data_pg_pools == 1) {
65 data_pool = fsc->mdsc->mdsmap->m_data_pg_pools[0];
66 } else {
67 data_pool = CEPH_NOPOOL;
68 }
69
70 dout("statfs\n");
71 err = ceph_monc_do_statfs(&fsc->client->monc, data_pool, &st);
72 if (err < 0)
73 return err;
74
75
76 buf->f_type = CEPH_SUPER_MAGIC;
77
78
79
80
81
82
83
84
85
86
87 buf->f_bsize = 1 << CEPH_BLOCK_SHIFT;
88 buf->f_frsize = 1 << CEPH_BLOCK_SHIFT;
89
90
91
92
93
94
95 if (ceph_test_mount_opt(fsc, NOQUOTADF) ||
96 !ceph_quota_update_statfs(fsc, buf)) {
97 buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10);
98 buf->f_bfree = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
99 buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
100 }
101
102 buf->f_files = le64_to_cpu(st.num_objects);
103 buf->f_ffree = -1;
104 buf->f_namelen = NAME_MAX;
105
106
107 fsid = le64_to_cpu(*(__le64 *)(&monmap->fsid)) ^
108 le64_to_cpu(*((__le64 *)&monmap->fsid + 1));
109 buf->f_fsid.val[0] = fsid & 0xffffffff;
110 buf->f_fsid.val[1] = fsid >> 32;
111
112 return 0;
113}
114
115
116static int ceph_sync_fs(struct super_block *sb, int wait)
117{
118 struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
119
120 if (!wait) {
121 dout("sync_fs (non-blocking)\n");
122 ceph_flush_dirty_caps(fsc->mdsc);
123 dout("sync_fs (non-blocking) done\n");
124 return 0;
125 }
126
127 dout("sync_fs (blocking)\n");
128 ceph_osdc_sync(&fsc->client->osdc);
129 ceph_mdsc_sync(fsc->mdsc);
130 dout("sync_fs (blocking) done\n");
131 return 0;
132}
133
134
135
136
137enum {
138 Opt_wsize,
139 Opt_rsize,
140 Opt_rasize,
141 Opt_caps_wanted_delay_min,
142 Opt_caps_wanted_delay_max,
143 Opt_readdir_max_entries,
144 Opt_readdir_max_bytes,
145 Opt_congestion_kb,
146 Opt_last_int,
147
148 Opt_snapdirname,
149 Opt_mds_namespace,
150 Opt_last_string,
151
152 Opt_dirstat,
153 Opt_nodirstat,
154 Opt_rbytes,
155 Opt_norbytes,
156 Opt_asyncreaddir,
157 Opt_noasyncreaddir,
158 Opt_dcache,
159 Opt_nodcache,
160 Opt_ino32,
161 Opt_noino32,
162 Opt_poolperm,
163 Opt_nopoolperm,
164 Opt_require_active_mds,
165 Opt_norequire_active_mds,
166#ifdef CONFIG_CEPH_FS_POSIX_ACL
167 Opt_acl,
168#endif
169 Opt_noacl,
170 Opt_quotadf,
171 Opt_noquotadf,
172};
173
174static match_table_t fsopt_tokens = {
175 {Opt_wsize, "wsize=%d"},
176 {Opt_rsize, "rsize=%d"},
177 {Opt_rasize, "rasize=%d"},
178 {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
179 {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"},
180 {Opt_readdir_max_entries, "readdir_max_entries=%d"},
181 {Opt_readdir_max_bytes, "readdir_max_bytes=%d"},
182 {Opt_congestion_kb, "write_congestion_kb=%d"},
183
184 {Opt_snapdirname, "snapdirname=%s"},
185 {Opt_mds_namespace, "mds_namespace=%s"},
186
187 {Opt_dirstat, "dirstat"},
188 {Opt_nodirstat, "nodirstat"},
189 {Opt_rbytes, "rbytes"},
190 {Opt_norbytes, "norbytes"},
191 {Opt_asyncreaddir, "asyncreaddir"},
192 {Opt_noasyncreaddir, "noasyncreaddir"},
193 {Opt_dcache, "dcache"},
194 {Opt_nodcache, "nodcache"},
195 {Opt_ino32, "ino32"},
196 {Opt_noino32, "noino32"},
197 {Opt_poolperm, "poolperm"},
198 {Opt_nopoolperm, "nopoolperm"},
199 {Opt_require_active_mds, "require_active_mds"},
200 {Opt_norequire_active_mds, "norequire_active_mds"},
201#ifdef CONFIG_CEPH_FS_POSIX_ACL
202 {Opt_acl, "acl"},
203#endif
204 {Opt_noacl, "noacl"},
205 {Opt_quotadf, "quotadf"},
206 {Opt_noquotadf, "noquotadf"},
207 {-1, NULL}
208};
209
210static int parse_fsopt_token(char *c, void *private)
211{
212 struct ceph_mount_options *fsopt = private;
213 substring_t argstr[MAX_OPT_ARGS];
214 int token, intval, ret;
215
216 token = match_token((char *)c, fsopt_tokens, argstr);
217 if (token < 0)
218 return -EINVAL;
219
220 if (token < Opt_last_int) {
221 ret = match_int(&argstr[0], &intval);
222 if (ret < 0) {
223 pr_err("bad mount option arg (not int) "
224 "at '%s'\n", c);
225 return ret;
226 }
227 dout("got int token %d val %d\n", token, intval);
228 } else if (token > Opt_last_int && token < Opt_last_string) {
229 dout("got string token %d val %s\n", token,
230 argstr[0].from);
231 } else {
232 dout("got token %d\n", token);
233 }
234
235 switch (token) {
236 case Opt_snapdirname:
237 kfree(fsopt->snapdir_name);
238 fsopt->snapdir_name = kstrndup(argstr[0].from,
239 argstr[0].to-argstr[0].from,
240 GFP_KERNEL);
241 if (!fsopt->snapdir_name)
242 return -ENOMEM;
243 break;
244 case Opt_mds_namespace:
245 kfree(fsopt->mds_namespace);
246 fsopt->mds_namespace = kstrndup(argstr[0].from,
247 argstr[0].to-argstr[0].from,
248 GFP_KERNEL);
249 if (!fsopt->mds_namespace)
250 return -ENOMEM;
251 break;
252
253 case Opt_wsize:
254 if (intval < PAGE_SIZE || intval > CEPH_MAX_WRITE_SIZE)
255 return -EINVAL;
256 fsopt->wsize = ALIGN(intval, PAGE_SIZE);
257 break;
258 case Opt_rsize:
259 if (intval < PAGE_SIZE || intval > CEPH_MAX_READ_SIZE)
260 return -EINVAL;
261 fsopt->rsize = ALIGN(intval, PAGE_SIZE);
262 break;
263 case Opt_rasize:
264 if (intval < 0)
265 return -EINVAL;
266 fsopt->rasize = ALIGN(intval + PAGE_SIZE - 1, PAGE_SIZE);
267 break;
268 case Opt_caps_wanted_delay_min:
269 if (intval < 1)
270 return -EINVAL;
271 fsopt->caps_wanted_delay_min = intval;
272 break;
273 case Opt_caps_wanted_delay_max:
274 if (intval < 1)
275 return -EINVAL;
276 fsopt->caps_wanted_delay_max = intval;
277 break;
278 case Opt_readdir_max_entries:
279 if (intval < 1)
280 return -EINVAL;
281 fsopt->max_readdir = intval;
282 break;
283 case Opt_readdir_max_bytes:
284 if (intval < PAGE_SIZE && intval != 0)
285 return -EINVAL;
286 fsopt->max_readdir_bytes = intval;
287 break;
288 case Opt_congestion_kb:
289 if (intval < 1024)
290 return -EINVAL;
291 fsopt->congestion_kb = intval;
292 break;
293 case Opt_dirstat:
294 fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT;
295 break;
296 case Opt_nodirstat:
297 fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT;
298 break;
299 case Opt_rbytes:
300 fsopt->flags |= CEPH_MOUNT_OPT_RBYTES;
301 break;
302 case Opt_norbytes:
303 fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES;
304 break;
305 case Opt_asyncreaddir:
306 fsopt->flags &= ~CEPH_MOUNT_OPT_NOASYNCREADDIR;
307 break;
308 case Opt_noasyncreaddir:
309 fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
310 break;
311 case Opt_dcache:
312 fsopt->flags |= CEPH_MOUNT_OPT_DCACHE;
313 break;
314 case Opt_nodcache:
315 fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
316 break;
317 case Opt_ino32:
318 fsopt->flags |= CEPH_MOUNT_OPT_INO32;
319 break;
320 case Opt_noino32:
321 fsopt->flags &= ~CEPH_MOUNT_OPT_INO32;
322 break;
323 case Opt_poolperm:
324 fsopt->flags &= ~CEPH_MOUNT_OPT_NOPOOLPERM;
325 break;
326 case Opt_nopoolperm:
327 fsopt->flags |= CEPH_MOUNT_OPT_NOPOOLPERM;
328 break;
329 case Opt_require_active_mds:
330 fsopt->flags &= ~CEPH_MOUNT_OPT_MOUNTWAIT;
331 break;
332 case Opt_norequire_active_mds:
333 fsopt->flags |= CEPH_MOUNT_OPT_MOUNTWAIT;
334 break;
335 case Opt_quotadf:
336 fsopt->flags &= ~CEPH_MOUNT_OPT_NOQUOTADF;
337 break;
338 case Opt_noquotadf:
339 fsopt->flags |= CEPH_MOUNT_OPT_NOQUOTADF;
340 break;
341#ifdef CONFIG_CEPH_FS_POSIX_ACL
342 case Opt_acl:
343 fsopt->sb_flags |= MS_POSIXACL;
344 break;
345#endif
346 case Opt_noacl:
347 fsopt->sb_flags &= ~MS_POSIXACL;
348 break;
349 default:
350 BUG_ON(token);
351 }
352 return 0;
353}
354
355static void destroy_mount_options(struct ceph_mount_options *args)
356{
357 dout("destroy_mount_options %p\n", args);
358 kfree(args->snapdir_name);
359 kfree(args->mds_namespace);
360 kfree(args->server_path);
361 kfree(args);
362}
363
364static int strcmp_null(const char *s1, const char *s2)
365{
366 if (!s1 && !s2)
367 return 0;
368 if (s1 && !s2)
369 return -1;
370 if (!s1 && s2)
371 return 1;
372 return strcmp(s1, s2);
373}
374
375static int compare_mount_options(struct ceph_mount_options *new_fsopt,
376 struct ceph_options *new_opt,
377 struct ceph_fs_client *fsc)
378{
379 struct ceph_mount_options *fsopt1 = new_fsopt;
380 struct ceph_mount_options *fsopt2 = fsc->mount_options;
381 int ofs = offsetof(struct ceph_mount_options, snapdir_name);
382 int ret;
383
384 ret = memcmp(fsopt1, fsopt2, ofs);
385 if (ret)
386 return ret;
387
388 ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name);
389 if (ret)
390 return ret;
391 ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace);
392 if (ret)
393 return ret;
394
395 ret = strcmp_null(fsopt1->server_path, fsopt2->server_path);
396 if (ret)
397 return ret;
398
399 return ceph_compare_options(new_opt, fsc->client);
400}
401
402static int parse_mount_options(struct ceph_mount_options **pfsopt,
403 struct ceph_options **popt,
404 int flags, char *options,
405 const char *dev_name)
406{
407 struct ceph_mount_options *fsopt;
408 const char *dev_name_end;
409 int err;
410
411 if (!dev_name || !*dev_name)
412 return -EINVAL;
413
414 fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL);
415 if (!fsopt)
416 return -ENOMEM;
417
418 dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name);
419
420 fsopt->sb_flags = flags;
421 fsopt->flags = CEPH_MOUNT_OPT_DEFAULT;
422
423 fsopt->wsize = CEPH_MAX_WRITE_SIZE;
424 fsopt->rsize = CEPH_MAX_READ_SIZE;
425 fsopt->rasize = CEPH_RASIZE_DEFAULT;
426 fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
427 if (!fsopt->snapdir_name) {
428 err = -ENOMEM;
429 goto out;
430 }
431
432 fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
433 fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
434 fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT;
435 fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
436 fsopt->congestion_kb = default_congestion_kb();
437
438
439
440
441
442
443
444
445
446
447
448 dev_name_end = strchr(dev_name, '/');
449 if (dev_name_end) {
450 if (strlen(dev_name_end) > 1) {
451 fsopt->server_path = kstrdup(dev_name_end, GFP_KERNEL);
452 if (!fsopt->server_path) {
453 err = -ENOMEM;
454 goto out;
455 }
456 }
457 } else {
458 dev_name_end = dev_name + strlen(dev_name);
459 }
460 err = -EINVAL;
461 dev_name_end--;
462 if (dev_name_end < dev_name || *dev_name_end != ':') {
463 pr_err("device name is missing path (no : separator in %s)\n",
464 dev_name);
465 goto out;
466 }
467 dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name);
468 if (fsopt->server_path)
469 dout("server path '%s'\n", fsopt->server_path);
470
471 *popt = ceph_parse_options(options, dev_name, dev_name_end,
472 parse_fsopt_token, (void *)fsopt);
473 if (IS_ERR(*popt)) {
474 err = PTR_ERR(*popt);
475 goto out;
476 }
477
478
479 *pfsopt = fsopt;
480 return 0;
481
482out:
483 destroy_mount_options(fsopt);
484 return err;
485}
486
487
488
489
490
491
492static int ceph_show_options(struct seq_file *m, struct dentry *root)
493{
494 struct ceph_fs_client *fsc = ceph_sb_to_client(root->d_sb);
495 struct ceph_mount_options *fsopt = fsc->mount_options;
496 size_t pos;
497 int ret;
498
499
500 seq_putc(m, ',');
501 pos = m->count;
502
503 ret = ceph_print_client_options(m, fsc->client);
504 if (ret)
505 return ret;
506
507
508 if (m->count == pos)
509 m->count--;
510
511 if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT)
512 seq_puts(m, ",dirstat");
513 if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES))
514 seq_puts(m, ",rbytes");
515 if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
516 seq_puts(m, ",noasyncreaddir");
517 if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
518 seq_puts(m, ",nodcache");
519 if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
520 seq_puts(m, ",nopoolperm");
521 if (fsopt->flags & CEPH_MOUNT_OPT_NOQUOTADF)
522 seq_puts(m, ",noquotadf");
523
524#ifdef CONFIG_CEPH_FS_POSIX_ACL
525 if (fsopt->sb_flags & MS_POSIXACL)
526 seq_puts(m, ",acl");
527 else
528 seq_puts(m, ",noacl");
529#endif
530
531 if (fsopt->mds_namespace)
532 seq_printf(m, ",mds_namespace=%s", fsopt->mds_namespace);
533 if (fsopt->wsize)
534 seq_printf(m, ",wsize=%d", fsopt->wsize);
535 if (fsopt->rsize != CEPH_MAX_READ_SIZE)
536 seq_printf(m, ",rsize=%d", fsopt->rsize);
537 if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
538 seq_printf(m, ",rasize=%d", fsopt->rasize);
539 if (fsopt->congestion_kb != default_congestion_kb())
540 seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb);
541 if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
542 seq_printf(m, ",caps_wanted_delay_min=%d",
543 fsopt->caps_wanted_delay_min);
544 if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
545 seq_printf(m, ",caps_wanted_delay_max=%d",
546 fsopt->caps_wanted_delay_max);
547 if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT)
548 seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir);
549 if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
550 seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes);
551 if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
552 seq_show_option(m, "snapdirname", fsopt->snapdir_name);
553
554 return 0;
555}
556
557
558
559
560
561static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg)
562{
563 struct ceph_fs_client *fsc = client->private;
564 int type = le16_to_cpu(msg->hdr.type);
565
566 switch (type) {
567 case CEPH_MSG_MDS_MAP:
568 ceph_mdsc_handle_mdsmap(fsc->mdsc, msg);
569 return 0;
570 case CEPH_MSG_FS_MAP_USER:
571 ceph_mdsc_handle_fsmap(fsc->mdsc, msg);
572 return 0;
573 default:
574 return -1;
575 }
576}
577
578
579
580
581static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
582 struct ceph_options *opt)
583{
584 struct ceph_fs_client *fsc;
585 int page_count;
586 size_t size;
587 int err = -ENOMEM;
588
589 fsc = kzalloc(sizeof(*fsc), GFP_KERNEL);
590 if (!fsc)
591 return ERR_PTR(-ENOMEM);
592
593 fsc->client = ceph_create_client(opt, fsc);
594 if (IS_ERR(fsc->client)) {
595 err = PTR_ERR(fsc->client);
596 goto fail;
597 }
598 fsc->client->extra_mon_dispatch = extra_mon_dispatch;
599
600 if (!fsopt->mds_namespace) {
601 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
602 0, true);
603 } else {
604 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_FSMAP,
605 0, false);
606 }
607
608 fsc->mount_options = fsopt;
609
610 fsc->sb = NULL;
611 fsc->mount_state = CEPH_MOUNT_MOUNTING;
612
613 atomic_long_set(&fsc->writeback_count, 0);
614
615 err = bdi_init(&fsc->backing_dev_info);
616 if (err < 0)
617 goto fail_client;
618
619 err = -ENOMEM;
620
621
622
623
624 fsc->wb_wq = alloc_workqueue("ceph-writeback", 0, 1);
625 if (!fsc->wb_wq)
626 goto fail_bdi;
627 fsc->pg_inv_wq = alloc_workqueue("ceph-pg-invalid", 0, 1);
628 if (!fsc->pg_inv_wq)
629 goto fail_wb_wq;
630 fsc->trunc_wq = alloc_workqueue("ceph-trunc", 0, 1);
631 if (!fsc->trunc_wq)
632 goto fail_pg_inv_wq;
633
634
635 err = -ENOMEM;
636 page_count = fsc->mount_options->wsize >> PAGE_CACHE_SHIFT;
637 size = sizeof (struct page *) * (page_count ? page_count : 1);
638 fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10, size);
639 if (!fsc->wb_pagevec_pool)
640 goto fail_trunc_wq;
641
642
643 fsc->min_caps = fsopt->max_readdir;
644
645 return fsc;
646
647fail_trunc_wq:
648 destroy_workqueue(fsc->trunc_wq);
649fail_pg_inv_wq:
650 destroy_workqueue(fsc->pg_inv_wq);
651fail_wb_wq:
652 destroy_workqueue(fsc->wb_wq);
653fail_bdi:
654 bdi_destroy(&fsc->backing_dev_info);
655fail_client:
656 ceph_destroy_client(fsc->client);
657fail:
658 kfree(fsc);
659 return ERR_PTR(err);
660}
661
662static void destroy_fs_client(struct ceph_fs_client *fsc)
663{
664 dout("destroy_fs_client %p\n", fsc);
665
666 destroy_workqueue(fsc->wb_wq);
667 destroy_workqueue(fsc->pg_inv_wq);
668 destroy_workqueue(fsc->trunc_wq);
669
670 bdi_destroy(&fsc->backing_dev_info);
671
672 mempool_destroy(fsc->wb_pagevec_pool);
673
674 destroy_mount_options(fsc->mount_options);
675
676 ceph_destroy_client(fsc->client);
677
678 kfree(fsc);
679 dout("destroy_fs_client %p done\n", fsc);
680}
681
682
683
684
685struct kmem_cache *ceph_inode_cachep;
686struct kmem_cache *ceph_cap_cachep;
687struct kmem_cache *ceph_cap_flush_cachep;
688struct kmem_cache *ceph_dentry_cachep;
689struct kmem_cache *ceph_file_cachep;
690struct kmem_cache *ceph_dir_file_cachep;
691
692static void ceph_inode_init_once(void *foo)
693{
694 struct ceph_inode_info *ci = foo;
695 inode_init_once(&ci->vfs_inode);
696}
697
698static int __init init_caches(void)
699{
700 ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
701 sizeof(struct ceph_inode_info),
702 __alignof__(struct ceph_inode_info),
703 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
704 SLAB_ACCOUNT, ceph_inode_init_once);
705 if (!ceph_inode_cachep)
706 return -ENOMEM;
707
708 ceph_cap_cachep = KMEM_CACHE(ceph_cap, SLAB_MEM_SPREAD);
709 if (!ceph_cap_cachep)
710 goto bad_cap;
711 ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush,
712 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
713 if (!ceph_cap_flush_cachep)
714 goto bad_cap_flush;
715
716 ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
717 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
718 if (!ceph_dentry_cachep)
719 goto bad_dentry;
720
721 ceph_file_cachep = KMEM_CACHE(ceph_file_info, SLAB_MEM_SPREAD);
722
723 if (!ceph_file_cachep)
724 goto bad_file;
725
726 ceph_dir_file_cachep = KMEM_CACHE(ceph_dir_file_info, SLAB_MEM_SPREAD);
727 if (!ceph_dir_file_cachep)
728 goto bad_dir_file;
729
730 return 0;
731
732bad_dir_file:
733 kmem_cache_destroy(ceph_file_cachep);
734bad_file:
735 kmem_cache_destroy(ceph_dentry_cachep);
736bad_dentry:
737 kmem_cache_destroy(ceph_cap_flush_cachep);
738bad_cap_flush:
739 kmem_cache_destroy(ceph_cap_cachep);
740bad_cap:
741 kmem_cache_destroy(ceph_inode_cachep);
742 return -ENOMEM;
743}
744
745static void destroy_caches(void)
746{
747
748
749
750
751 rcu_barrier();
752 kmem_cache_destroy(ceph_inode_cachep);
753 kmem_cache_destroy(ceph_cap_cachep);
754 kmem_cache_destroy(ceph_cap_flush_cachep);
755 kmem_cache_destroy(ceph_dentry_cachep);
756 kmem_cache_destroy(ceph_file_cachep);
757 kmem_cache_destroy(ceph_dir_file_cachep);
758}
759
760
761
762
763
764
765static void ceph_umount_begin(struct super_block *sb)
766{
767 struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
768
769 dout("ceph_umount_begin - starting forced umount\n");
770 if (!fsc)
771 return;
772 fsc->mount_state = CEPH_MOUNT_SHUTDOWN;
773 ceph_mdsc_force_umount(fsc->mdsc);
774 return;
775}
776
777static const struct super_operations ceph_super_ops = {
778 .alloc_inode = ceph_alloc_inode,
779 .destroy_inode = ceph_destroy_inode,
780 .write_inode = ceph_write_inode,
781 .drop_inode = ceph_drop_inode,
782 .evict_inode = ceph_evict_inode,
783 .sync_fs = ceph_sync_fs,
784 .put_super = ceph_put_super,
785 .show_options = ceph_show_options,
786 .statfs = ceph_statfs,
787 .umount_begin = ceph_umount_begin,
788};
789
790
791
792
793
794static struct dentry *open_root_dentry(struct ceph_fs_client *fsc,
795 const char *path,
796 unsigned long started)
797{
798 struct ceph_mds_client *mdsc = fsc->mdsc;
799 struct ceph_mds_request *req = NULL;
800 int err;
801 struct dentry *root;
802
803
804 dout("open_root_inode opening '%s'\n", path);
805 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
806 if (IS_ERR(req))
807 return ERR_CAST(req);
808 req->r_path1 = kstrdup(path, GFP_NOFS);
809 if (!req->r_path1) {
810 root = ERR_PTR(-ENOMEM);
811 goto out;
812 }
813
814 req->r_ino1.ino = CEPH_INO_ROOT;
815 req->r_ino1.snap = CEPH_NOSNAP;
816 req->r_started = started;
817 req->r_timeout = fsc->client->options->mount_timeout;
818 req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
819 req->r_num_caps = 2;
820 err = ceph_mdsc_do_request(mdsc, NULL, req);
821 if (err == 0) {
822 struct inode *inode = req->r_target_inode;
823 req->r_target_inode = NULL;
824 dout("open_root_inode success\n");
825 root = d_make_root(inode);
826 if (!root) {
827 root = ERR_PTR(-ENOMEM);
828 goto out;
829 }
830 ceph_init_dentry(root);
831 dout("open_root_inode success, root dentry is %p\n", root);
832 } else {
833 root = ERR_PTR(err);
834 }
835out:
836 ceph_mdsc_put_request(req);
837 return root;
838}
839
840
841
842
843
844
845
846static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc)
847{
848 int err;
849 unsigned long started = jiffies;
850 struct dentry *root;
851
852 dout("mount start %p\n", fsc);
853 mutex_lock(&fsc->client->mount_mutex);
854
855 if (!fsc->sb->s_root) {
856 const char *path;
857 err = __ceph_open_session(fsc->client, started);
858 if (err < 0)
859 goto out;
860
861 if (!fsc->mount_options->server_path) {
862 path = "";
863 dout("mount opening path \\t\n");
864 } else {
865 path = fsc->mount_options->server_path + 1;
866 dout("mount opening path %s\n", path);
867 }
868
869 err = ceph_fs_debugfs_init(fsc);
870 if (err < 0)
871 goto out;
872
873 root = open_root_dentry(fsc, path, started);
874 if (IS_ERR(root)) {
875 err = PTR_ERR(root);
876 goto out;
877 }
878 fsc->sb->s_root = dget(root);
879 } else {
880 root = dget(fsc->sb->s_root);
881 }
882
883 fsc->mount_state = CEPH_MOUNT_MOUNTED;
884 dout("mount success\n");
885 mutex_unlock(&fsc->client->mount_mutex);
886 return root;
887
888out:
889 mutex_unlock(&fsc->client->mount_mutex);
890 return ERR_PTR(err);
891}
892
893static int ceph_set_super(struct super_block *s, void *data)
894{
895 struct ceph_fs_client *fsc = data;
896 int ret;
897
898 dout("set_super %p data %p\n", s, data);
899
900 s->s_flags = fsc->mount_options->sb_flags;
901 s->s_maxbytes = 1ULL << 40;
902
903 s->s_xattr = ceph_xattr_handlers;
904 s->s_fs_info = fsc;
905 fsc->sb = s;
906
907 s->s_op = &ceph_super_ops;
908 s->s_export_op = &ceph_export_ops;
909
910 s->s_time_gran = 1000;
911
912 ret = set_anon_super(s, NULL);
913 if (ret != 0)
914 goto fail;
915
916 return ret;
917
918fail:
919 s->s_fs_info = NULL;
920 fsc->sb = NULL;
921 return ret;
922}
923
924
925
926
927static int ceph_compare_super(struct super_block *sb, void *data)
928{
929 struct ceph_fs_client *new = data;
930 struct ceph_mount_options *fsopt = new->mount_options;
931 struct ceph_options *opt = new->client->options;
932 struct ceph_fs_client *other = ceph_sb_to_client(sb);
933
934 dout("ceph_compare_super %p\n", sb);
935
936 if (compare_mount_options(fsopt, opt, other)) {
937 dout("monitor(s)/mount options don't match\n");
938 return 0;
939 }
940 if ((opt->flags & CEPH_OPT_FSID) &&
941 ceph_fsid_compare(&opt->fsid, &other->client->fsid)) {
942 dout("fsid doesn't match\n");
943 return 0;
944 }
945 if (fsopt->sb_flags != other->mount_options->sb_flags) {
946 dout("flags differ\n");
947 return 0;
948 }
949 return 1;
950}
951
952
953
954
955static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
956
957static int ceph_register_bdi(struct super_block *sb,
958 struct ceph_fs_client *fsc)
959{
960 int err;
961
962
963 fsc->backing_dev_info.ra_pages = fsc->mount_options->rasize >> PAGE_SHIFT;
964
965 err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%ld",
966 atomic_long_inc_return(&bdi_seq));
967 if (!err)
968 sb->s_bdi = &fsc->backing_dev_info;
969 return err;
970}
971
972static struct dentry *ceph_mount(struct file_system_type *fs_type,
973 int flags, const char *dev_name, void *data)
974{
975 struct super_block *sb;
976 struct ceph_fs_client *fsc;
977 struct dentry *res;
978 int err;
979 int (*compare_super)(struct super_block *, void *) = ceph_compare_super;
980 struct ceph_mount_options *fsopt = NULL;
981 struct ceph_options *opt = NULL;
982
983 dout("ceph_mount\n");
984
985#ifdef CONFIG_CEPH_FS_POSIX_ACL
986 flags |= MS_POSIXACL;
987#endif
988 err = parse_mount_options(&fsopt, &opt, flags, data, dev_name);
989 if (err < 0) {
990 res = ERR_PTR(err);
991 goto out_final;
992 }
993
994
995 fsc = create_fs_client(fsopt, opt);
996 if (IS_ERR(fsc)) {
997 res = ERR_CAST(fsc);
998 destroy_mount_options(fsopt);
999 ceph_destroy_options(opt);
1000 goto out_final;
1001 }
1002
1003 err = ceph_mdsc_init(fsc);
1004 if (err < 0) {
1005 res = ERR_PTR(err);
1006 goto out;
1007 }
1008
1009 if (ceph_test_opt(fsc->client, NOSHARE))
1010 compare_super = NULL;
1011 sb = sget(fs_type, compare_super, ceph_set_super, flags, fsc);
1012 if (IS_ERR(sb)) {
1013 res = ERR_CAST(sb);
1014 goto out;
1015 }
1016
1017 if (ceph_sb_to_client(sb) != fsc) {
1018 ceph_mdsc_destroy(fsc);
1019 destroy_fs_client(fsc);
1020 fsc = ceph_sb_to_client(sb);
1021 dout("get_sb got existing client %p\n", fsc);
1022 } else {
1023 dout("get_sb using new client %p\n", fsc);
1024 err = ceph_register_bdi(sb, fsc);
1025 if (err < 0) {
1026 res = ERR_PTR(err);
1027 goto out_splat;
1028 }
1029 }
1030
1031 res = ceph_real_mount(fsc);
1032 if (IS_ERR(res))
1033 goto out_splat;
1034 dout("root %p inode %p ino %llx.%llx\n", res,
1035 res->d_inode, ceph_vinop(res->d_inode));
1036 return res;
1037
1038out_splat:
1039 ceph_mdsc_close_sessions(fsc->mdsc);
1040 deactivate_locked_super(sb);
1041 goto out_final;
1042
1043out:
1044 ceph_mdsc_destroy(fsc);
1045 destroy_fs_client(fsc);
1046out_final:
1047 dout("ceph_mount fail %ld\n", PTR_ERR(res));
1048 return res;
1049}
1050
1051static void ceph_kill_sb(struct super_block *s)
1052{
1053 struct ceph_fs_client *fsc = ceph_sb_to_client(s);
1054 dout("kill_sb %p\n", s);
1055 ceph_mdsc_pre_umount(fsc->mdsc);
1056 kill_anon_super(s);
1057
1058 fsc->client->extra_mon_dispatch = NULL;
1059 ceph_fs_debugfs_cleanup(fsc);
1060
1061 ceph_mdsc_destroy(fsc);
1062 destroy_fs_client(fsc);
1063}
1064
1065static struct file_system_type ceph_fs_type = {
1066 .owner = THIS_MODULE,
1067 .name = "ceph",
1068 .mount = ceph_mount,
1069 .kill_sb = ceph_kill_sb,
1070 .fs_flags = FS_RENAME_DOES_D_MOVE,
1071};
1072MODULE_ALIAS_FS("ceph");
1073
1074static int __init init_ceph(void)
1075{
1076 int ret = init_caches();
1077 if (ret)
1078 goto out;
1079
1080 ceph_flock_init();
1081 ceph_xattr_init();
1082 ret = register_filesystem(&ceph_fs_type);
1083 if (ret)
1084 goto out_xattr;
1085
1086 pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
1087
1088 return 0;
1089
1090out_xattr:
1091 ceph_xattr_exit();
1092 destroy_caches();
1093out:
1094 return ret;
1095}
1096
1097static void __exit exit_ceph(void)
1098{
1099 dout("exit_ceph\n");
1100 unregister_filesystem(&ceph_fs_type);
1101 ceph_xattr_exit();
1102 destroy_caches();
1103}
1104
1105module_init(init_ceph);
1106module_exit(exit_ceph);
1107
1108MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
1109MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
1110MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
1111MODULE_DESCRIPTION("Ceph filesystem for Linux");
1112MODULE_LICENSE("GPL");
1113
