~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

TOMOYO Linux Cross Reference
Linux/fs/ceph/super.c

Version: ~ [ linux-5.4-rc3 ] ~ [ linux-5.3.6 ] ~ [ linux-5.2.21 ] ~ [ linux-5.1.21 ] ~ [ linux-5.0.21 ] ~ [ linux-4.20.17 ] ~ [ linux-4.19.79 ] ~ [ linux-4.18.20 ] ~ [ linux-4.17.19 ] ~ [ linux-4.16.18 ] ~ [ linux-4.15.18 ] ~ [ linux-4.14.149 ] ~ [ linux-4.13.16 ] ~ [ linux-4.12.14 ] ~ [ linux-4.11.12 ] ~ [ linux-4.10.17 ] ~ [ linux-4.9.196 ] ~ [ linux-4.8.17 ] ~ [ linux-4.7.10 ] ~ [ linux-4.6.7 ] ~ [ linux-4.5.7 ] ~ [ linux-4.4.196 ] ~ [ linux-4.3.6 ] ~ [ linux-4.2.8 ] ~ [ linux-4.1.52 ] ~ [ linux-4.0.9 ] ~ [ linux-3.19.8 ] ~ [ linux-3.18.140 ] ~ [ linux-3.17.8 ] ~ [ linux-3.16.75 ] ~ [ linux-3.15.10 ] ~ [ linux-3.14.79 ] ~ [ linux-3.13.11 ] ~ [ linux-3.12.74 ] ~ [ linux-3.11.10 ] ~ [ linux-3.10.108 ] ~ [ linux-3.9.11 ] ~ [ linux-3.8.13 ] ~ [ linux-3.7.10 ] ~ [ linux-3.6.11 ] ~ [ linux-3.5.7 ] ~ [ linux-3.4.113 ] ~ [ linux-3.3.8 ] ~ [ linux-3.2.102 ] ~ [ linux-3.1.10 ] ~ [ linux-3.0.101 ] ~ [ linux-2.6.32.71 ] ~ [ linux-2.6.0 ] ~ [ linux-2.4.37.11 ] ~ [ unix-v6-master ] ~ [ ccs-tools-1.8.5 ] ~ [ policy-sample ] ~
Architecture: ~ [ i386 ] ~ [ alpha ] ~ [ m68k ] ~ [ mips ] ~ [ ppc ] ~ [ sparc ] ~ [ sparc64 ] ~

  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/decode.h>
 22 #include <linux/ceph/mon_client.h>
 23 #include <linux/ceph/auth.h>
 24 #include <linux/ceph/debugfs.h>
 25 
 26 /*
 27  * Ceph superblock operations
 28  *
 29  * Handle the basics of mounting, unmounting.
 30  */
 31 
 32 /*
 33  * super ops
 34  */
 35 static void ceph_put_super(struct super_block *s)
 36 {
 37         struct ceph_fs_client *fsc = ceph_sb_to_client(s);
 38 
 39         dout("put_super\n");
 40         ceph_mdsc_close_sessions(fsc->mdsc);
 41 
 42         /*
 43          * ensure we release the bdi before put_anon_super releases
 44          * the device name.
 45          */
 46         if (s->s_bdi == &fsc->backing_dev_info) {
 47                 bdi_unregister(&fsc->backing_dev_info);
 48                 s->s_bdi = NULL;
 49         }
 50 
 51         return;
 52 }
 53 
 54 static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
 55 {
 56         struct ceph_fs_client *fsc = ceph_inode_to_client(dentry->d_inode);
 57         struct ceph_monmap *monmap = fsc->client->monc.monmap;
 58         struct ceph_statfs st;
 59         u64 fsid;
 60         int err;
 61 
 62         dout("statfs\n");
 63         err = ceph_monc_do_statfs(&fsc->client->monc, &st);
 64         if (err < 0)
 65                 return err;
 66 
 67         /* fill in kstatfs */
 68         buf->f_type = CEPH_SUPER_MAGIC;  /* ?? */
 69 
 70         /*
 71          * express utilization in terms of large blocks to avoid
 72          * overflow on 32-bit machines.
 73          *
 74          * NOTE: for the time being, we make bsize == frsize to humor
 75          * not-yet-ancient versions of glibc that are broken.
 76          * Someday, we will probably want to report a real block
 77          * size...  whatever that may mean for a network file system!
 78          */
 79         buf->f_bsize = 1 << CEPH_BLOCK_SHIFT;
 80         buf->f_frsize = 1 << CEPH_BLOCK_SHIFT;
 81         buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10);
 82         buf->f_bfree = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
 83         buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
 84 
 85         buf->f_files = le64_to_cpu(st.num_objects);
 86         buf->f_ffree = -1;
 87         buf->f_namelen = NAME_MAX;
 88 
 89         /* leave fsid little-endian, regardless of host endianness */
 90         fsid = *(u64 *)(&monmap->fsid) ^ *((u64 *)&monmap->fsid + 1);
 91         buf->f_fsid.val[0] = fsid & 0xffffffff;
 92         buf->f_fsid.val[1] = fsid >> 32;
 93 
 94         return 0;
 95 }
 96 
 97 
 98 static int ceph_sync_fs(struct super_block *sb, int wait)
 99 {
100         struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
101 
102         if (!wait) {
103                 dout("sync_fs (non-blocking)\n");
104                 ceph_flush_dirty_caps(fsc->mdsc);
105                 dout("sync_fs (non-blocking) done\n");
106                 return 0;
107         }
108 
109         dout("sync_fs (blocking)\n");
110         ceph_osdc_sync(&fsc->client->osdc);
111         ceph_mdsc_sync(fsc->mdsc);
112         dout("sync_fs (blocking) done\n");
113         return 0;
114 }
115 
116 /*
117  * mount options
118  */
119 enum {
120         Opt_wsize,
121         Opt_rsize,
122         Opt_rasize,
123         Opt_caps_wanted_delay_min,
124         Opt_caps_wanted_delay_max,
125         Opt_cap_release_safety,
126         Opt_readdir_max_entries,
127         Opt_readdir_max_bytes,
128         Opt_congestion_kb,
129         Opt_last_int,
130         /* int args above */
131         Opt_snapdirname,
132         Opt_last_string,
133         /* string args above */
134         Opt_dirstat,
135         Opt_nodirstat,
136         Opt_rbytes,
137         Opt_norbytes,
138         Opt_noasyncreaddir,
139         Opt_ino32,
140 };
141 
142 static match_table_t fsopt_tokens = {
143         {Opt_wsize, "wsize=%d"},
144         {Opt_rsize, "rsize=%d"},
145         {Opt_rasize, "rasize=%d"},
146         {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
147         {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"},
148         {Opt_cap_release_safety, "cap_release_safety=%d"},
149         {Opt_readdir_max_entries, "readdir_max_entries=%d"},
150         {Opt_readdir_max_bytes, "readdir_max_bytes=%d"},
151         {Opt_congestion_kb, "write_congestion_kb=%d"},
152         /* int args above */
153         {Opt_snapdirname, "snapdirname=%s"},
154         /* string args above */
155         {Opt_dirstat, "dirstat"},
156         {Opt_nodirstat, "nodirstat"},
157         {Opt_rbytes, "rbytes"},
158         {Opt_norbytes, "norbytes"},
159         {Opt_noasyncreaddir, "noasyncreaddir"},
160         {Opt_ino32, "ino32"},
161         {-1, NULL}
162 };
163 
164 static int parse_fsopt_token(char *c, void *private)
165 {
166         struct ceph_mount_options *fsopt = private;
167         substring_t argstr[MAX_OPT_ARGS];
168         int token, intval, ret;
169 
170         token = match_token((char *)c, fsopt_tokens, argstr);
171         if (token < 0)
172                 return -EINVAL;
173 
174         if (token < Opt_last_int) {
175                 ret = match_int(&argstr[0], &intval);
176                 if (ret < 0) {
177                         pr_err("bad mount option arg (not int) "
178                                "at '%s'\n", c);
179                         return ret;
180                 }
181                 dout("got int token %d val %d\n", token, intval);
182         } else if (token > Opt_last_int && token < Opt_last_string) {
183                 dout("got string token %d val %s\n", token,
184                      argstr[0].from);
185         } else {
186                 dout("got token %d\n", token);
187         }
188 
189         switch (token) {
190         case Opt_snapdirname:
191                 kfree(fsopt->snapdir_name);
192                 fsopt->snapdir_name = kstrndup(argstr[0].from,
193                                                argstr[0].to-argstr[0].from,
194                                                GFP_KERNEL);
195                 if (!fsopt->snapdir_name)
196                         return -ENOMEM;
197                 break;
198 
199                 /* misc */
200         case Opt_wsize:
201                 fsopt->wsize = intval;
202                 break;
203         case Opt_rsize:
204                 fsopt->rsize = intval;
205                 break;
206         case Opt_rasize:
207                 fsopt->rasize = intval;
208                 break;
209         case Opt_caps_wanted_delay_min:
210                 fsopt->caps_wanted_delay_min = intval;
211                 break;
212         case Opt_caps_wanted_delay_max:
213                 fsopt->caps_wanted_delay_max = intval;
214                 break;
215         case Opt_readdir_max_entries:
216                 fsopt->max_readdir = intval;
217                 break;
218         case Opt_readdir_max_bytes:
219                 fsopt->max_readdir_bytes = intval;
220                 break;
221         case Opt_congestion_kb:
222                 fsopt->congestion_kb = intval;
223                 break;
224         case Opt_dirstat:
225                 fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT;
226                 break;
227         case Opt_nodirstat:
228                 fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT;
229                 break;
230         case Opt_rbytes:
231                 fsopt->flags |= CEPH_MOUNT_OPT_RBYTES;
232                 break;
233         case Opt_norbytes:
234                 fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES;
235                 break;
236         case Opt_noasyncreaddir:
237                 fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
238                 break;
239         case Opt_ino32:
240                 fsopt->flags |= CEPH_MOUNT_OPT_INO32;
241                 break;
242         default:
243                 BUG_ON(token);
244         }
245         return 0;
246 }
247 
248 static void destroy_mount_options(struct ceph_mount_options *args)
249 {
250         dout("destroy_mount_options %p\n", args);
251         kfree(args->snapdir_name);
252         kfree(args);
253 }
254 
255 static int strcmp_null(const char *s1, const char *s2)
256 {
257         if (!s1 && !s2)
258                 return 0;
259         if (s1 && !s2)
260                 return -1;
261         if (!s1 && s2)
262                 return 1;
263         return strcmp(s1, s2);
264 }
265 
266 static int compare_mount_options(struct ceph_mount_options *new_fsopt,
267                                  struct ceph_options *new_opt,
268                                  struct ceph_fs_client *fsc)
269 {
270         struct ceph_mount_options *fsopt1 = new_fsopt;
271         struct ceph_mount_options *fsopt2 = fsc->mount_options;
272         int ofs = offsetof(struct ceph_mount_options, snapdir_name);
273         int ret;
274 
275         ret = memcmp(fsopt1, fsopt2, ofs);
276         if (ret)
277                 return ret;
278 
279         ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name);
280         if (ret)
281                 return ret;
282 
283         return ceph_compare_options(new_opt, fsc->client);
284 }
285 
286 static int parse_mount_options(struct ceph_mount_options **pfsopt,
287                                struct ceph_options **popt,
288                                int flags, char *options,
289                                const char *dev_name,
290                                const char **path)
291 {
292         struct ceph_mount_options *fsopt;
293         const char *dev_name_end;
294         int err = -ENOMEM;
295 
296         fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL);
297         if (!fsopt)
298                 return -ENOMEM;
299 
300         dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name);
301 
302         fsopt->sb_flags = flags;
303         fsopt->flags = CEPH_MOUNT_OPT_DEFAULT;
304 
305         fsopt->rsize = CEPH_RSIZE_DEFAULT;
306         fsopt->rasize = CEPH_RASIZE_DEFAULT;
307         fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
308         fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
309         fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
310         fsopt->cap_release_safety = CEPH_CAP_RELEASE_SAFETY_DEFAULT;
311         fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT;
312         fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
313         fsopt->congestion_kb = default_congestion_kb();
314 
315         /* ip1[:port1][,ip2[:port2]...]:/subdir/in/fs */
316         err = -EINVAL;
317         if (!dev_name)
318                 goto out;
319         *path = strstr(dev_name, ":/");
320         if (*path == NULL) {
321                 pr_err("device name is missing path (no :/ in %s)\n",
322                                 dev_name);
323                 goto out;
324         }
325         dev_name_end = *path;
326         dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name);
327 
328         /* path on server */
329         *path += 2;
330         dout("server path '%s'\n", *path);
331 
332         err = ceph_parse_options(popt, options, dev_name, dev_name_end,
333                                  parse_fsopt_token, (void *)fsopt);
334         if (err)
335                 goto out;
336 
337         /* success */
338         *pfsopt = fsopt;
339         return 0;
340 
341 out:
342         destroy_mount_options(fsopt);
343         return err;
344 }
345 
346 /**
347  * ceph_show_options - Show mount options in /proc/mounts
348  * @m: seq_file to write to
349  * @mnt: mount descriptor
350  */
351 static int ceph_show_options(struct seq_file *m, struct vfsmount *mnt)
352 {
353         struct ceph_fs_client *fsc = ceph_sb_to_client(mnt->mnt_sb);
354         struct ceph_mount_options *fsopt = fsc->mount_options;
355         struct ceph_options *opt = fsc->client->options;
356 
357         if (opt->flags & CEPH_OPT_FSID)
358                 seq_printf(m, ",fsid=%pU", &opt->fsid);
359         if (opt->flags & CEPH_OPT_NOSHARE)
360                 seq_puts(m, ",noshare");
361         if (opt->flags & CEPH_OPT_NOCRC)
362                 seq_puts(m, ",nocrc");
363 
364         if (opt->name) {
365                 seq_puts(m, ",name=");
366                 seq_escape(m, opt->name, ", \t\n\\");
367         }
368         if (opt->key)
369                 seq_puts(m, ",secret=<hidden>");
370 
371         if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
372                 seq_printf(m, ",mount_timeout=%d", opt->mount_timeout);
373         if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
374                 seq_printf(m, ",osd_idle_ttl=%d", opt->osd_idle_ttl);
375         if (opt->osd_timeout != CEPH_OSD_TIMEOUT_DEFAULT)
376                 seq_printf(m, ",osdtimeout=%d", opt->osd_timeout);
377         if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
378                 seq_printf(m, ",osdkeepalivetimeout=%d",
379                            opt->osd_keepalive_timeout);
380 
381         if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT)
382                 seq_puts(m, ",dirstat");
383         if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES) == 0)
384                 seq_puts(m, ",norbytes");
385         if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
386                 seq_puts(m, ",noasyncreaddir");
387 
388         if (fsopt->wsize)
389                 seq_printf(m, ",wsize=%d", fsopt->wsize);
390         if (fsopt->rsize != CEPH_RSIZE_DEFAULT)
391                 seq_printf(m, ",rsize=%d", fsopt->rsize);
392         if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
393                 seq_printf(m, ",rasize=%d", fsopt->rasize);
394         if (fsopt->congestion_kb != default_congestion_kb())
395                 seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb);
396         if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
397                 seq_printf(m, ",caps_wanted_delay_min=%d",
398                          fsopt->caps_wanted_delay_min);
399         if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
400                 seq_printf(m, ",caps_wanted_delay_max=%d",
401                            fsopt->caps_wanted_delay_max);
402         if (fsopt->cap_release_safety != CEPH_CAP_RELEASE_SAFETY_DEFAULT)
403                 seq_printf(m, ",cap_release_safety=%d",
404                            fsopt->cap_release_safety);
405         if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT)
406                 seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir);
407         if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
408                 seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes);
409         if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
410                 seq_show_option(m, "snapdirname", fsopt->snapdir_name);
411         return 0;
412 }
413 
414 /*
415  * handle any mon messages the standard library doesn't understand.
416  * return error if we don't either.
417  */
418 static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg)
419 {
420         struct ceph_fs_client *fsc = client->private;
421         int type = le16_to_cpu(msg->hdr.type);
422 
423         switch (type) {
424         case CEPH_MSG_MDS_MAP:
425                 ceph_mdsc_handle_map(fsc->mdsc, msg);
426                 return 0;
427 
428         default:
429                 return -1;
430         }
431 }
432 
433 /*
434  * create a new fs client
435  */
436 static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
437                                         struct ceph_options *opt)
438 {
439         struct ceph_fs_client *fsc;
440         const unsigned supported_features =
441                 CEPH_FEATURE_FLOCK |
442                 CEPH_FEATURE_DIRLAYOUTHASH;
443         const unsigned required_features = 0;
444         int err = -ENOMEM;
445 
446         fsc = kzalloc(sizeof(*fsc), GFP_KERNEL);
447         if (!fsc)
448                 return ERR_PTR(-ENOMEM);
449 
450         fsc->client = ceph_create_client(opt, fsc, supported_features,
451                                          required_features);
452         if (IS_ERR(fsc->client)) {
453                 err = PTR_ERR(fsc->client);
454                 goto fail;
455         }
456         fsc->client->extra_mon_dispatch = extra_mon_dispatch;
457         fsc->client->monc.want_mdsmap = 1;
458 
459         fsc->mount_options = fsopt;
460 
461         fsc->sb = NULL;
462         fsc->mount_state = CEPH_MOUNT_MOUNTING;
463 
464         atomic_long_set(&fsc->writeback_count, 0);
465 
466         err = bdi_init(&fsc->backing_dev_info);
467         if (err < 0)
468                 goto fail_client;
469 
470         err = -ENOMEM;
471         /*
472          * The number of concurrent works can be high but they don't need
473          * to be processed in parallel, limit concurrency.
474          */
475         fsc->wb_wq = alloc_workqueue("ceph-writeback", 0, 1);
476         if (fsc->wb_wq == NULL)
477                 goto fail_bdi;
478         fsc->pg_inv_wq = alloc_workqueue("ceph-pg-invalid", 0, 1);
479         if (fsc->pg_inv_wq == NULL)
480                 goto fail_wb_wq;
481         fsc->trunc_wq = alloc_workqueue("ceph-trunc", 0, 1);
482         if (fsc->trunc_wq == NULL)
483                 goto fail_pg_inv_wq;
484 
485         /* set up mempools */
486         err = -ENOMEM;
487         fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10,
488                               fsc->mount_options->wsize >> PAGE_CACHE_SHIFT);
489         if (!fsc->wb_pagevec_pool)
490                 goto fail_trunc_wq;
491 
492         /* caps */
493         fsc->min_caps = fsopt->max_readdir;
494 
495         return fsc;
496 
497 fail_trunc_wq:
498         destroy_workqueue(fsc->trunc_wq);
499 fail_pg_inv_wq:
500         destroy_workqueue(fsc->pg_inv_wq);
501 fail_wb_wq:
502         destroy_workqueue(fsc->wb_wq);
503 fail_bdi:
504         bdi_destroy(&fsc->backing_dev_info);
505 fail_client:
506         ceph_destroy_client(fsc->client);
507 fail:
508         kfree(fsc);
509         return ERR_PTR(err);
510 }
511 
512 static void destroy_fs_client(struct ceph_fs_client *fsc)
513 {
514         dout("destroy_fs_client %p\n", fsc);
515 
516         destroy_workqueue(fsc->wb_wq);
517         destroy_workqueue(fsc->pg_inv_wq);
518         destroy_workqueue(fsc->trunc_wq);
519 
520         bdi_destroy(&fsc->backing_dev_info);
521 
522         mempool_destroy(fsc->wb_pagevec_pool);
523 
524         destroy_mount_options(fsc->mount_options);
525 
526         ceph_fs_debugfs_cleanup(fsc);
527 
528         ceph_destroy_client(fsc->client);
529 
530         kfree(fsc);
531         dout("destroy_fs_client %p done\n", fsc);
532 }
533 
534 /*
535  * caches
536  */
537 struct kmem_cache *ceph_inode_cachep;
538 struct kmem_cache *ceph_cap_cachep;
539 struct kmem_cache *ceph_dentry_cachep;
540 struct kmem_cache *ceph_file_cachep;
541 
542 static void ceph_inode_init_once(void *foo)
543 {
544         struct ceph_inode_info *ci = foo;
545         inode_init_once(&ci->vfs_inode);
546 }
547 
548 static int __init init_caches(void)
549 {
550         ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
551                                       sizeof(struct ceph_inode_info),
552                                       __alignof__(struct ceph_inode_info),
553                                       (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
554                                       ceph_inode_init_once);
555         if (ceph_inode_cachep == NULL)
556                 return -ENOMEM;
557 
558         ceph_cap_cachep = KMEM_CACHE(ceph_cap,
559                                      SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
560         if (ceph_cap_cachep == NULL)
561                 goto bad_cap;
562 
563         ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
564                                         SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
565         if (ceph_dentry_cachep == NULL)
566                 goto bad_dentry;
567 
568         ceph_file_cachep = KMEM_CACHE(ceph_file_info,
569                                       SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
570         if (ceph_file_cachep == NULL)
571                 goto bad_file;
572 
573         return 0;
574 
575 bad_file:
576         kmem_cache_destroy(ceph_dentry_cachep);
577 bad_dentry:
578         kmem_cache_destroy(ceph_cap_cachep);
579 bad_cap:
580         kmem_cache_destroy(ceph_inode_cachep);
581         return -ENOMEM;
582 }
583 
584 static void destroy_caches(void)
585 {
586         kmem_cache_destroy(ceph_inode_cachep);
587         kmem_cache_destroy(ceph_cap_cachep);
588         kmem_cache_destroy(ceph_dentry_cachep);
589         kmem_cache_destroy(ceph_file_cachep);
590 }
591 
592 
593 /*
594  * ceph_umount_begin - initiate forced umount.  Tear down down the
595  * mount, skipping steps that may hang while waiting for server(s).
596  */
597 static void ceph_umount_begin(struct super_block *sb)
598 {
599         struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
600 
601         dout("ceph_umount_begin - starting forced umount\n");
602         if (!fsc)
603                 return;
604         fsc->mount_state = CEPH_MOUNT_SHUTDOWN;
605         return;
606 }
607 
608 static const struct super_operations ceph_super_ops = {
609         .alloc_inode    = ceph_alloc_inode,
610         .destroy_inode  = ceph_destroy_inode,
611         .write_inode    = ceph_write_inode,
612         .sync_fs        = ceph_sync_fs,
613         .put_super      = ceph_put_super,
614         .show_options   = ceph_show_options,
615         .statfs         = ceph_statfs,
616         .umount_begin   = ceph_umount_begin,
617 };
618 
619 /*
620  * Bootstrap mount by opening the root directory.  Note the mount
621  * @started time from caller, and time out if this takes too long.
622  */
623 static struct dentry *open_root_dentry(struct ceph_fs_client *fsc,
624                                        const char *path,
625                                        unsigned long started)
626 {
627         struct ceph_mds_client *mdsc = fsc->mdsc;
628         struct ceph_mds_request *req = NULL;
629         int err;
630         struct dentry *root;
631 
632         /* open dir */
633         dout("open_root_inode opening '%s'\n", path);
634         req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
635         if (IS_ERR(req))
636                 return ERR_CAST(req);
637         req->r_path1 = kstrdup(path, GFP_NOFS);
638         req->r_ino1.ino = CEPH_INO_ROOT;
639         req->r_ino1.snap = CEPH_NOSNAP;
640         req->r_started = started;
641         req->r_timeout = fsc->client->options->mount_timeout * HZ;
642         req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
643         req->r_num_caps = 2;
644         err = ceph_mdsc_do_request(mdsc, NULL, req);
645         if (err == 0) {
646                 dout("open_root_inode success\n");
647                 if (ceph_ino(req->r_target_inode) == CEPH_INO_ROOT &&
648                     fsc->sb->s_root == NULL) {
649                         root = d_alloc_root(req->r_target_inode);
650                         ceph_init_dentry(root);
651                 } else {
652                         root = d_obtain_alias(req->r_target_inode);
653                 }
654                 req->r_target_inode = NULL;
655                 dout("open_root_inode success, root dentry is %p\n", root);
656         } else {
657                 root = ERR_PTR(err);
658         }
659         ceph_mdsc_put_request(req);
660         return root;
661 }
662 
663 
664 
665 
666 /*
667  * mount: join the ceph cluster, and open root directory.
668  */
669 static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc,
670                       const char *path)
671 {
672         int err;
673         unsigned long started = jiffies;  /* note the start time */
674         struct dentry *root;
675         int first = 0;   /* first vfsmount for this super_block */
676 
677         dout("mount start\n");
678         mutex_lock(&fsc->client->mount_mutex);
679 
680         err = __ceph_open_session(fsc->client, started);
681         if (err < 0)
682                 goto out;
683 
684         dout("mount opening root\n");
685         root = open_root_dentry(fsc, "", started);
686         if (IS_ERR(root)) {
687                 err = PTR_ERR(root);
688                 goto out;
689         }
690         if (fsc->sb->s_root) {
691                 dput(root);
692         } else {
693                 fsc->sb->s_root = root;
694                 first = 1;
695 
696                 err = ceph_fs_debugfs_init(fsc);
697                 if (err < 0)
698                         goto fail;
699         }
700 
701         if (path[0] == 0) {
702                 dget(root);
703         } else {
704                 dout("mount opening base mountpoint\n");
705                 root = open_root_dentry(fsc, path, started);
706                 if (IS_ERR(root)) {
707                         err = PTR_ERR(root);
708                         goto fail;
709                 }
710         }
711 
712         fsc->mount_state = CEPH_MOUNT_MOUNTED;
713         dout("mount success\n");
714         mutex_unlock(&fsc->client->mount_mutex);
715         return root;
716 
717 out:
718         mutex_unlock(&fsc->client->mount_mutex);
719         return ERR_PTR(err);
720 
721 fail:
722         if (first) {
723                 dput(fsc->sb->s_root);
724                 fsc->sb->s_root = NULL;
725         }
726         goto out;
727 }
728 
729 static int ceph_set_super(struct super_block *s, void *data)
730 {
731         struct ceph_fs_client *fsc = data;
732         int ret;
733 
734         dout("set_super %p data %p\n", s, data);
735 
736         s->s_flags = fsc->mount_options->sb_flags;
737         s->s_maxbytes = 1ULL << 40;  /* temp value until we get mdsmap */
738 
739         s->s_fs_info = fsc;
740         fsc->sb = s;
741 
742         s->s_op = &ceph_super_ops;
743         s->s_export_op = &ceph_export_ops;
744 
745         s->s_time_gran = 1000;  /* 1000 ns == 1 us */
746 
747         ret = set_anon_super(s, NULL);  /* what is that second arg for? */
748         if (ret != 0)
749                 goto fail;
750 
751         return ret;
752 
753 fail:
754         s->s_fs_info = NULL;
755         fsc->sb = NULL;
756         return ret;
757 }
758 
759 /*
760  * share superblock if same fs AND options
761  */
762 static int ceph_compare_super(struct super_block *sb, void *data)
763 {
764         struct ceph_fs_client *new = data;
765         struct ceph_mount_options *fsopt = new->mount_options;
766         struct ceph_options *opt = new->client->options;
767         struct ceph_fs_client *other = ceph_sb_to_client(sb);
768 
769         dout("ceph_compare_super %p\n", sb);
770 
771         if (compare_mount_options(fsopt, opt, other)) {
772                 dout("monitor(s)/mount options don't match\n");
773                 return 0;
774         }
775         if ((opt->flags & CEPH_OPT_FSID) &&
776             ceph_fsid_compare(&opt->fsid, &other->client->fsid)) {
777                 dout("fsid doesn't match\n");
778                 return 0;
779         }
780         if (fsopt->sb_flags != other->mount_options->sb_flags) {
781                 dout("flags differ\n");
782                 return 0;
783         }
784         return 1;
785 }
786 
787 /*
788  * construct our own bdi so we can control readahead, etc.
789  */
790 static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
791 
792 static int ceph_register_bdi(struct super_block *sb,
793                              struct ceph_fs_client *fsc)
794 {
795         int err;
796 
797         /* set ra_pages based on rasize mount option? */
798         if (fsc->mount_options->rasize >= PAGE_CACHE_SIZE)
799                 fsc->backing_dev_info.ra_pages =
800                         (fsc->mount_options->rasize + PAGE_CACHE_SIZE - 1)
801                         >> PAGE_SHIFT;
802         else
803                 fsc->backing_dev_info.ra_pages =
804                         default_backing_dev_info.ra_pages;
805 
806         err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%d",
807                            atomic_long_inc_return(&bdi_seq));
808         if (!err)
809                 sb->s_bdi = &fsc->backing_dev_info;
810         return err;
811 }
812 
813 static struct dentry *ceph_mount(struct file_system_type *fs_type,
814                        int flags, const char *dev_name, void *data)
815 {
816         struct super_block *sb;
817         struct ceph_fs_client *fsc;
818         struct dentry *res;
819         int err;
820         int (*compare_super)(struct super_block *, void *) = ceph_compare_super;
821         const char *path = NULL;
822         struct ceph_mount_options *fsopt = NULL;
823         struct ceph_options *opt = NULL;
824 
825         dout("ceph_mount\n");
826         err = parse_mount_options(&fsopt, &opt, flags, data, dev_name, &path);
827         if (err < 0) {
828                 res = ERR_PTR(err);
829                 goto out_final;
830         }
831 
832         /* create client (which we may/may not use) */
833         fsc = create_fs_client(fsopt, opt);
834         if (IS_ERR(fsc)) {
835                 res = ERR_CAST(fsc);
836                 destroy_mount_options(fsopt);
837                 ceph_destroy_options(opt);
838                 goto out_final;
839         }
840 
841         err = ceph_mdsc_init(fsc);
842         if (err < 0) {
843                 res = ERR_PTR(err);
844                 goto out;
845         }
846 
847         if (ceph_test_opt(fsc->client, NOSHARE))
848                 compare_super = NULL;
849         sb = sget(fs_type, compare_super, ceph_set_super, fsc);
850         if (IS_ERR(sb)) {
851                 res = ERR_CAST(sb);
852                 goto out;
853         }
854 
855         if (ceph_sb_to_client(sb) != fsc) {
856                 ceph_mdsc_destroy(fsc);
857                 destroy_fs_client(fsc);
858                 fsc = ceph_sb_to_client(sb);
859                 dout("get_sb got existing client %p\n", fsc);
860         } else {
861                 dout("get_sb using new client %p\n", fsc);
862                 err = ceph_register_bdi(sb, fsc);
863                 if (err < 0) {
864                         res = ERR_PTR(err);
865                         goto out_splat;
866                 }
867         }
868 
869         res = ceph_real_mount(fsc, path);
870         if (IS_ERR(res))
871                 goto out_splat;
872         dout("root %p inode %p ino %llx.%llx\n", res,
873              res->d_inode, ceph_vinop(res->d_inode));
874         return res;
875 
876 out_splat:
877         ceph_mdsc_close_sessions(fsc->mdsc);
878         deactivate_locked_super(sb);
879         goto out_final;
880 
881 out:
882         ceph_mdsc_destroy(fsc);
883         destroy_fs_client(fsc);
884 out_final:
885         dout("ceph_mount fail %ld\n", PTR_ERR(res));
886         return res;
887 }
888 
889 static void ceph_kill_sb(struct super_block *s)
890 {
891         struct ceph_fs_client *fsc = ceph_sb_to_client(s);
892         dout("kill_sb %p\n", s);
893         ceph_mdsc_pre_umount(fsc->mdsc);
894         kill_anon_super(s);    /* will call put_super after sb is r/o */
895         ceph_mdsc_destroy(fsc);
896         destroy_fs_client(fsc);
897 }
898 
899 static struct file_system_type ceph_fs_type = {
900         .owner          = THIS_MODULE,
901         .name           = "ceph",
902         .mount          = ceph_mount,
903         .kill_sb        = ceph_kill_sb,
904         .fs_flags       = FS_RENAME_DOES_D_MOVE,
905 };
906 
907 #define _STRINGIFY(x) #x
908 #define STRINGIFY(x) _STRINGIFY(x)
909 
910 static int __init init_ceph(void)
911 {
912         int ret = init_caches();
913         if (ret)
914                 goto out;
915 
916         ret = ceph_snap_init();
917         if (ret)
918                 goto out_icache;
919 
920         ret = register_filesystem(&ceph_fs_type);
921         if (ret)
922                 goto out_snap;
923 
924         pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
925 
926         return 0;
927 
928 out_snap:
929         ceph_snap_exit();
930 out_icache:
931         destroy_caches();
932 out:
933         return ret;
934 }
935 
936 static void __exit exit_ceph(void)
937 {
938         dout("exit_ceph\n");
939         unregister_filesystem(&ceph_fs_type);
940         ceph_snap_exit();
941         destroy_caches();
942 }
943 
944 module_init(init_ceph);
945 module_exit(exit_ceph);
946 
947 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
948 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
949 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
950 MODULE_DESCRIPTION("Ceph filesystem for Linux");
951 MODULE_LICENSE("GPL");
952 

~ [ source navigation ] ~ [ diff markup ] ~ [ identifier search ] ~

kernel.org | git.kernel.org | LWN.net | Project Home | Wiki (Japanese) | Wiki (English) | SVN repository | Mail admin

Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.

osdn.jp