2 #include "ceph_debug.h"
4 #include <linux/backing-dev.h>
6 #include <linux/inet.h>
8 #include <linux/module.h>
9 #include <linux/mount.h>
10 #include <linux/parser.h>
11 #include <linux/sched.h>
12 #include <linux/seq_file.h>
13 #include <linux/slab.h>
14 #include <linux/statfs.h>
15 #include <linux/string.h>
19 #include "mon_client.h"
23 * Ceph superblock operations
25 * Handle the basics of mounting, unmounting.
30 * find filename portion of a path (/foo/bar/baz -> baz)
32 const char *ceph_file_part(const char *s, int len)
34 const char *e = s + len;
36 while (e != s && *(e-1) != '/')
45 static void ceph_put_super(struct super_block *s)
47 struct ceph_client *client = ceph_sb_to_client(s);
50 ceph_mdsc_close_sessions(&client->mdsc);
53 * ensure we release the bdi before put_anon_super releases
56 if (s->s_bdi == &client->backing_dev_info) {
57 bdi_unregister(&client->backing_dev_info);
64 static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
66 struct ceph_client *client = ceph_inode_to_client(dentry->d_inode);
67 struct ceph_monmap *monmap = client->monc.monmap;
68 struct ceph_statfs st;
73 err = ceph_monc_do_statfs(&client->monc, &st);
78 buf->f_type = CEPH_SUPER_MAGIC; /* ?? */
81 * express utilization in terms of large blocks to avoid
82 * overflow on 32-bit machines.
84 buf->f_bsize = 1 << CEPH_BLOCK_SHIFT;
85 buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10);
86 buf->f_bfree = (le64_to_cpu(st.kb) - le64_to_cpu(st.kb_used)) >>
87 (CEPH_BLOCK_SHIFT-10);
88 buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
90 buf->f_files = le64_to_cpu(st.num_objects);
92 buf->f_namelen = NAME_MAX;
93 buf->f_frsize = PAGE_CACHE_SIZE;
95 /* leave fsid little-endian, regardless of host endianness */
96 fsid = *(u64 *)(&monmap->fsid) ^ *((u64 *)&monmap->fsid + 1);
97 buf->f_fsid.val[0] = fsid & 0xffffffff;
98 buf->f_fsid.val[1] = fsid >> 32;
104 static int ceph_syncfs(struct super_block *sb, int wait)
106 dout("sync_fs %d\n", wait);
107 ceph_osdc_sync(&ceph_sb_to_client(sb)->osdc);
108 ceph_mdsc_sync(&ceph_sb_to_client(sb)->mdsc);
109 dout("sync_fs %d done\n", wait);
113 static int default_congestion_kb(void)
120 * congestion size, scale with available memory.
132 * This allows larger machines to have larger/more transfers.
133 * Limit the default to 256M
135 congestion_kb = (16*int_sqrt(totalram_pages)) << (PAGE_SHIFT-10);
136 if (congestion_kb > 256*1024)
137 congestion_kb = 256*1024;
139 return congestion_kb;
143 * ceph_show_options - Show mount options in /proc/mounts
144 * @m: seq_file to write to
145 * @mnt: mount descriptor
147 static int ceph_show_options(struct seq_file *m, struct vfsmount *mnt)
149 struct ceph_client *client = ceph_sb_to_client(mnt->mnt_sb);
150 struct ceph_mount_args *args = client->mount_args;
152 if (args->flags & CEPH_OPT_FSID)
153 seq_printf(m, ",fsidmajor=%llu,fsidminor%llu",
154 le64_to_cpu(*(__le64 *)&args->fsid.fsid[0]),
155 le64_to_cpu(*(__le64 *)&args->fsid.fsid[8]));
156 if (args->flags & CEPH_OPT_NOSHARE)
157 seq_puts(m, ",noshare");
158 if (args->flags & CEPH_OPT_DIRSTAT)
159 seq_puts(m, ",dirstat");
160 if ((args->flags & CEPH_OPT_RBYTES) == 0)
161 seq_puts(m, ",norbytes");
162 if (args->flags & CEPH_OPT_NOCRC)
163 seq_puts(m, ",nocrc");
164 if (args->flags & CEPH_OPT_NOASYNCREADDIR)
165 seq_puts(m, ",noasyncreaddir");
167 if (args->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT)
168 seq_printf(m, ",mount_timeout=%d", args->mount_timeout);
169 if (args->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT)
170 seq_printf(m, ",osd_idle_ttl=%d", args->osd_idle_ttl);
171 if (args->osd_timeout != CEPH_OSD_TIMEOUT_DEFAULT)
172 seq_printf(m, ",osdtimeout=%d", args->osd_timeout);
173 if (args->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT)
174 seq_printf(m, ",osdkeepalivetimeout=%d",
175 args->osd_keepalive_timeout);
177 seq_printf(m, ",wsize=%d", args->wsize);
178 if (args->rsize != CEPH_MOUNT_RSIZE_DEFAULT)
179 seq_printf(m, ",rsize=%d", args->rsize);
180 if (args->congestion_kb != default_congestion_kb())
181 seq_printf(m, ",write_congestion_kb=%d", args->congestion_kb);
182 if (args->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
183 seq_printf(m, ",caps_wanted_delay_min=%d",
184 args->caps_wanted_delay_min);
185 if (args->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
186 seq_printf(m, ",caps_wanted_delay_max=%d",
187 args->caps_wanted_delay_max);
188 if (args->cap_release_safety != CEPH_CAP_RELEASE_SAFETY_DEFAULT)
189 seq_printf(m, ",cap_release_safety=%d",
190 args->cap_release_safety);
191 if (args->max_readdir != CEPH_MAX_READDIR_DEFAULT)
192 seq_printf(m, ",readdir_max_entries=%d", args->max_readdir);
193 if (args->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
194 seq_printf(m, ",readdir_max_bytes=%d", args->max_readdir_bytes);
195 if (strcmp(args->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
196 seq_printf(m, ",snapdirname=%s", args->snapdir_name);
198 seq_printf(m, ",name=%s", args->name);
200 seq_puts(m, ",secret=<hidden>");
207 struct kmem_cache *ceph_inode_cachep;
208 struct kmem_cache *ceph_cap_cachep;
209 struct kmem_cache *ceph_dentry_cachep;
210 struct kmem_cache *ceph_file_cachep;
212 static void ceph_inode_init_once(void *foo)
214 struct ceph_inode_info *ci = foo;
215 inode_init_once(&ci->vfs_inode);
218 static int __init init_caches(void)
220 ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
221 sizeof(struct ceph_inode_info),
222 __alignof__(struct ceph_inode_info),
223 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
224 ceph_inode_init_once);
225 if (ceph_inode_cachep == NULL)
228 ceph_cap_cachep = KMEM_CACHE(ceph_cap,
229 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
230 if (ceph_cap_cachep == NULL)
233 ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
234 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
235 if (ceph_dentry_cachep == NULL)
238 ceph_file_cachep = KMEM_CACHE(ceph_file_info,
239 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
240 if (ceph_file_cachep == NULL)
246 kmem_cache_destroy(ceph_dentry_cachep);
248 kmem_cache_destroy(ceph_cap_cachep);
250 kmem_cache_destroy(ceph_inode_cachep);
254 static void destroy_caches(void)
256 kmem_cache_destroy(ceph_inode_cachep);
257 kmem_cache_destroy(ceph_cap_cachep);
258 kmem_cache_destroy(ceph_dentry_cachep);
259 kmem_cache_destroy(ceph_file_cachep);
264 * ceph_umount_begin - initiate forced umount. Tear down down the
265 * mount, skipping steps that may hang while waiting for server(s).
267 static void ceph_umount_begin(struct super_block *sb)
269 struct ceph_client *client = ceph_sb_to_client(sb);
271 dout("ceph_umount_begin - starting forced umount\n");
274 client->mount_state = CEPH_MOUNT_SHUTDOWN;
278 static const struct super_operations ceph_super_ops = {
279 .alloc_inode = ceph_alloc_inode,
280 .destroy_inode = ceph_destroy_inode,
281 .write_inode = ceph_write_inode,
282 .sync_fs = ceph_syncfs,
283 .put_super = ceph_put_super,
284 .show_options = ceph_show_options,
285 .statfs = ceph_statfs,
286 .umount_begin = ceph_umount_begin,
290 const char *ceph_msg_type_name(int type)
293 case CEPH_MSG_SHUTDOWN: return "shutdown";
294 case CEPH_MSG_PING: return "ping";
295 case CEPH_MSG_AUTH: return "auth";
296 case CEPH_MSG_AUTH_REPLY: return "auth_reply";
297 case CEPH_MSG_MON_MAP: return "mon_map";
298 case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
299 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
300 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
301 case CEPH_MSG_STATFS: return "statfs";
302 case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
303 case CEPH_MSG_MDS_MAP: return "mds_map";
304 case CEPH_MSG_CLIENT_SESSION: return "client_session";
305 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
306 case CEPH_MSG_CLIENT_REQUEST: return "client_request";
307 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
308 case CEPH_MSG_CLIENT_REPLY: return "client_reply";
309 case CEPH_MSG_CLIENT_CAPS: return "client_caps";
310 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
311 case CEPH_MSG_CLIENT_SNAP: return "client_snap";
312 case CEPH_MSG_CLIENT_LEASE: return "client_lease";
313 case CEPH_MSG_OSD_MAP: return "osd_map";
314 case CEPH_MSG_OSD_OP: return "osd_op";
315 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
316 default: return "unknown";
330 Opt_osdkeepalivetimeout,
333 Opt_caps_wanted_delay_min,
334 Opt_caps_wanted_delay_max,
335 Opt_cap_release_safety,
336 Opt_readdir_max_entries,
337 Opt_readdir_max_bytes,
345 /* string args above */
356 static match_table_t arg_tokens = {
357 {Opt_fsidmajor, "fsidmajor=%ld"},
358 {Opt_fsidminor, "fsidminor=%ld"},
359 {Opt_wsize, "wsize=%d"},
360 {Opt_rsize, "rsize=%d"},
361 {Opt_osdtimeout, "osdtimeout=%d"},
362 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
363 {Opt_mount_timeout, "mount_timeout=%d"},
364 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
365 {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
366 {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"},
367 {Opt_cap_release_safety, "cap_release_safety=%d"},
368 {Opt_readdir_max_entries, "readdir_max_entries=%d"},
369 {Opt_readdir_max_bytes, "readdir_max_bytes=%d"},
370 {Opt_congestion_kb, "write_congestion_kb=%d"},
372 {Opt_snapdirname, "snapdirname=%s"},
373 {Opt_name, "name=%s"},
374 {Opt_secret, "secret=%s"},
375 /* string args above */
377 {Opt_noshare, "noshare"},
378 {Opt_dirstat, "dirstat"},
379 {Opt_nodirstat, "nodirstat"},
380 {Opt_rbytes, "rbytes"},
381 {Opt_norbytes, "norbytes"},
382 {Opt_nocrc, "nocrc"},
383 {Opt_noasyncreaddir, "noasyncreaddir"},
388 static struct ceph_mount_args *parse_mount_args(int flags, char *options,
389 const char *dev_name,
392 struct ceph_mount_args *args;
395 substring_t argstr[MAX_OPT_ARGS];
397 args = kzalloc(sizeof(*args), GFP_KERNEL);
399 return ERR_PTR(-ENOMEM);
400 args->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*args->mon_addr),
405 dout("parse_mount_args %p, dev_name '%s'\n", args, dev_name);
407 /* start with defaults */
408 args->sb_flags = flags;
409 args->flags = CEPH_OPT_DEFAULT;
410 args->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
411 args->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
412 args->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
413 args->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */
414 args->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
415 args->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
416 args->rsize = CEPH_MOUNT_RSIZE_DEFAULT;
417 args->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
418 args->cap_release_safety = CEPH_CAP_RELEASE_SAFETY_DEFAULT;
419 args->max_readdir = CEPH_MAX_READDIR_DEFAULT;
420 args->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
421 args->congestion_kb = default_congestion_kb();
423 /* ip1[:port1][,ip2[:port2]...]:/subdir/in/fs */
427 *path = strstr(dev_name, ":/");
429 pr_err("device name is missing path (no :/ in %s)\n",
435 err = ceph_parse_ips(dev_name, *path, args->mon_addr,
436 CEPH_MAX_MON, &args->num_mon);
442 dout("server path '%s'\n", *path);
444 /* parse mount options */
445 while ((c = strsep(&options, ",")) != NULL) {
446 int token, intval, ret;
450 token = match_token((char *)c, arg_tokens, argstr);
452 pr_err("bad mount option at '%s'\n", c);
455 if (token < Opt_last_int) {
456 ret = match_int(&argstr[0], &intval);
458 pr_err("bad mount option arg (not int) "
462 dout("got int token %d val %d\n", token, intval);
463 } else if (token > Opt_last_int && token < Opt_last_string) {
464 dout("got string token %d val %s\n", token,
467 dout("got token %d\n", token);
471 *(__le64 *)&args->fsid.fsid[0] = cpu_to_le64(intval);
474 *(__le64 *)&args->fsid.fsid[8] = cpu_to_le64(intval);
477 err = ceph_parse_ips(argstr[0].from,
483 args->flags |= CEPH_OPT_MYIP;
486 case Opt_snapdirname:
487 kfree(args->snapdir_name);
488 args->snapdir_name = kstrndup(argstr[0].from,
489 argstr[0].to-argstr[0].from,
493 args->name = kstrndup(argstr[0].from,
494 argstr[0].to-argstr[0].from,
498 args->secret = kstrndup(argstr[0].from,
499 argstr[0].to-argstr[0].from,
505 args->wsize = intval;
508 args->rsize = intval;
511 args->osd_timeout = intval;
513 case Opt_osdkeepalivetimeout:
514 args->osd_keepalive_timeout = intval;
516 case Opt_mount_timeout:
517 args->mount_timeout = intval;
519 case Opt_caps_wanted_delay_min:
520 args->caps_wanted_delay_min = intval;
522 case Opt_caps_wanted_delay_max:
523 args->caps_wanted_delay_max = intval;
525 case Opt_readdir_max_entries:
526 args->max_readdir = intval;
528 case Opt_readdir_max_bytes:
529 args->max_readdir_bytes = intval;
531 case Opt_congestion_kb:
532 args->congestion_kb = intval;
536 args->flags |= CEPH_OPT_NOSHARE;
540 args->flags |= CEPH_OPT_DIRSTAT;
543 args->flags &= ~CEPH_OPT_DIRSTAT;
546 args->flags |= CEPH_OPT_RBYTES;
549 args->flags &= ~CEPH_OPT_RBYTES;
552 args->flags |= CEPH_OPT_NOCRC;
554 case Opt_noasyncreaddir:
555 args->flags |= CEPH_OPT_NOASYNCREADDIR;
565 kfree(args->mon_addr);
570 static void destroy_mount_args(struct ceph_mount_args *args)
572 dout("destroy_mount_args %p\n", args);
573 kfree(args->snapdir_name);
574 args->snapdir_name = NULL;
583 * create a fresh client instance
585 static struct ceph_client *ceph_create_client(struct ceph_mount_args *args)
587 struct ceph_client *client;
590 client = kzalloc(sizeof(*client), GFP_KERNEL);
592 return ERR_PTR(-ENOMEM);
594 mutex_init(&client->mount_mutex);
596 init_waitqueue_head(&client->auth_wq);
599 client->mount_state = CEPH_MOUNT_MOUNTING;
600 client->mount_args = args;
604 client->auth_err = 0;
605 atomic_long_set(&client->writeback_count, 0);
607 err = bdi_init(&client->backing_dev_info);
612 client->wb_wq = create_workqueue("ceph-writeback");
613 if (client->wb_wq == NULL)
615 client->pg_inv_wq = create_singlethread_workqueue("ceph-pg-invalid");
616 if (client->pg_inv_wq == NULL)
618 client->trunc_wq = create_singlethread_workqueue("ceph-trunc");
619 if (client->trunc_wq == NULL)
622 /* set up mempools */
624 client->wb_pagevec_pool = mempool_create_kmalloc_pool(10,
625 client->mount_args->wsize >> PAGE_CACHE_SHIFT);
626 if (!client->wb_pagevec_pool)
630 client->min_caps = args->max_readdir;
633 err = ceph_monc_init(&client->monc, client);
636 err = ceph_osdc_init(&client->osdc, client);
639 err = ceph_mdsc_init(&client->mdsc, client);
645 ceph_osdc_stop(&client->osdc);
647 ceph_monc_stop(&client->monc);
649 mempool_destroy(client->wb_pagevec_pool);
651 destroy_workqueue(client->trunc_wq);
653 destroy_workqueue(client->pg_inv_wq);
655 destroy_workqueue(client->wb_wq);
657 bdi_destroy(&client->backing_dev_info);
663 static void ceph_destroy_client(struct ceph_client *client)
665 dout("destroy_client %p\n", client);
668 ceph_mdsc_stop(&client->mdsc);
669 ceph_osdc_stop(&client->osdc);
672 * make sure mds and osd connections close out before destroying
673 * the auth module, which is needed to free those connections'
678 ceph_monc_stop(&client->monc);
680 ceph_debugfs_client_cleanup(client);
681 destroy_workqueue(client->wb_wq);
682 destroy_workqueue(client->pg_inv_wq);
683 destroy_workqueue(client->trunc_wq);
685 bdi_destroy(&client->backing_dev_info);
688 ceph_messenger_destroy(client->msgr);
689 mempool_destroy(client->wb_pagevec_pool);
691 destroy_mount_args(client->mount_args);
694 dout("destroy_client %p done\n", client);
698 * Initially learn our fsid, or verify an fsid matches.
700 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
702 if (client->have_fsid) {
703 if (ceph_fsid_compare(&client->fsid, fsid)) {
704 pr_err("bad fsid, had " FSID_FORMAT " got " FSID_FORMAT,
705 PR_FSID(&client->fsid), PR_FSID(fsid));
709 pr_info("client%lld fsid " FSID_FORMAT "\n",
710 client->monc.auth->global_id, PR_FSID(fsid));
711 memcpy(&client->fsid, fsid, sizeof(*fsid));
712 ceph_debugfs_client_init(client);
713 client->have_fsid = true;
719 * true if we have the mon map (and have thus joined the cluster)
721 static int have_mon_and_osd_map(struct ceph_client *client)
723 return client->monc.monmap && client->monc.monmap->epoch &&
724 client->osdc.osdmap && client->osdc.osdmap->epoch;
728 * Bootstrap mount by opening the root directory. Note the mount
729 * @started time from caller, and time out if this takes too long.
731 static struct dentry *open_root_dentry(struct ceph_client *client,
733 unsigned long started)
735 struct ceph_mds_client *mdsc = &client->mdsc;
736 struct ceph_mds_request *req = NULL;
741 dout("open_root_inode opening '%s'\n", path);
742 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
744 return ERR_CAST(req);
745 req->r_path1 = kstrdup(path, GFP_NOFS);
746 req->r_ino1.ino = CEPH_INO_ROOT;
747 req->r_ino1.snap = CEPH_NOSNAP;
748 req->r_started = started;
749 req->r_timeout = client->mount_args->mount_timeout * HZ;
750 req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
752 err = ceph_mdsc_do_request(mdsc, NULL, req);
754 dout("open_root_inode success\n");
755 if (ceph_ino(req->r_target_inode) == CEPH_INO_ROOT &&
756 client->sb->s_root == NULL)
757 root = d_alloc_root(req->r_target_inode);
759 root = d_obtain_alias(req->r_target_inode);
760 req->r_target_inode = NULL;
761 dout("open_root_inode success, root dentry is %p\n", root);
765 ceph_mdsc_put_request(req);
770 * mount: join the ceph cluster, and open root directory.
772 static int ceph_mount(struct ceph_client *client, struct vfsmount *mnt,
775 struct ceph_entity_addr *myaddr = NULL;
777 unsigned long timeout = client->mount_args->mount_timeout * HZ;
778 unsigned long started = jiffies; /* note the start time */
781 dout("mount start\n");
782 mutex_lock(&client->mount_mutex);
784 /* initialize the messenger */
785 if (client->msgr == NULL) {
786 if (ceph_test_opt(client, MYIP))
787 myaddr = &client->mount_args->my_addr;
788 client->msgr = ceph_messenger_create(myaddr);
789 if (IS_ERR(client->msgr)) {
790 err = PTR_ERR(client->msgr);
794 client->msgr->nocrc = ceph_test_opt(client, NOCRC);
797 /* open session, and wait for mon, mds, and osd maps */
798 err = ceph_monc_open_session(&client->monc);
802 while (!have_mon_and_osd_map(client)) {
804 if (timeout && time_after_eq(jiffies, started + timeout))
808 dout("mount waiting for mon_map\n");
809 err = wait_event_interruptible_timeout(client->auth_wq,
810 have_mon_and_osd_map(client) || (client->auth_err < 0),
812 if (err == -EINTR || err == -ERESTARTSYS)
814 if (client->auth_err < 0) {
815 err = client->auth_err;
820 dout("mount opening root\n");
821 root = open_root_dentry(client, "", started);
826 if (client->sb->s_root)
829 client->sb->s_root = root;
834 dout("mount opening base mountpoint\n");
835 root = open_root_dentry(client, path, started);
838 dput(client->sb->s_root);
839 client->sb->s_root = NULL;
844 mnt->mnt_root = root;
845 mnt->mnt_sb = client->sb;
847 client->mount_state = CEPH_MOUNT_MOUNTED;
848 dout("mount success\n");
852 mutex_unlock(&client->mount_mutex);
856 static int ceph_set_super(struct super_block *s, void *data)
858 struct ceph_client *client = data;
861 dout("set_super %p data %p\n", s, data);
863 s->s_flags = client->mount_args->sb_flags;
864 s->s_maxbytes = 1ULL << 40; /* temp value until we get mdsmap */
866 s->s_fs_info = client;
869 s->s_op = &ceph_super_ops;
870 s->s_export_op = &ceph_export_ops;
872 s->s_time_gran = 1000; /* 1000 ns == 1 us */
874 ret = set_anon_super(s, NULL); /* what is that second arg for? */
887 * share superblock if same fs AND options
889 static int ceph_compare_super(struct super_block *sb, void *data)
891 struct ceph_client *new = data;
892 struct ceph_mount_args *args = new->mount_args;
893 struct ceph_client *other = ceph_sb_to_client(sb);
896 dout("ceph_compare_super %p\n", sb);
897 if (args->flags & CEPH_OPT_FSID) {
898 if (ceph_fsid_compare(&args->fsid, &other->fsid)) {
899 dout("fsid doesn't match\n");
903 /* do we share (a) monitor? */
904 for (i = 0; i < new->monc.monmap->num_mon; i++)
905 if (ceph_monmap_contains(other->monc.monmap,
906 &new->monc.monmap->mon_inst[i].addr))
908 if (i == new->monc.monmap->num_mon) {
909 dout("mon ip not part of monmap\n");
912 dout("mon ip matches existing sb %p\n", sb);
914 if (args->sb_flags != other->mount_args->sb_flags) {
915 dout("flags differ\n");
922 * construct our own bdi so we can control readahead, etc.
924 static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
926 static int ceph_register_bdi(struct super_block *sb, struct ceph_client *client)
930 /* set ra_pages based on rsize mount option? */
931 if (client->mount_args->rsize >= PAGE_CACHE_SIZE)
932 client->backing_dev_info.ra_pages =
933 (client->mount_args->rsize + PAGE_CACHE_SIZE - 1)
935 err = bdi_register(&client->backing_dev_info, NULL, "ceph-%d",
936 atomic_long_inc_return(&bdi_seq));
938 sb->s_bdi = &client->backing_dev_info;
942 static int ceph_get_sb(struct file_system_type *fs_type,
943 int flags, const char *dev_name, void *data,
944 struct vfsmount *mnt)
946 struct super_block *sb;
947 struct ceph_client *client;
949 int (*compare_super)(struct super_block *, void *) = ceph_compare_super;
950 const char *path = NULL;
951 struct ceph_mount_args *args;
953 dout("ceph_get_sb\n");
954 args = parse_mount_args(flags, data, dev_name, &path);
960 /* create client (which we may/may not use) */
961 client = ceph_create_client(args);
962 if (IS_ERR(client)) {
963 err = PTR_ERR(client);
967 if (client->mount_args->flags & CEPH_OPT_NOSHARE)
968 compare_super = NULL;
969 sb = sget(fs_type, compare_super, ceph_set_super, client);
975 if (ceph_sb_to_client(sb) != client) {
976 ceph_destroy_client(client);
977 client = ceph_sb_to_client(sb);
978 dout("get_sb got existing client %p\n", client);
980 dout("get_sb using new client %p\n", client);
981 err = ceph_register_bdi(sb, client);
986 err = ceph_mount(client, mnt, path);
989 dout("root %p inode %p ino %llx.%llx\n", mnt->mnt_root,
990 mnt->mnt_root->d_inode, ceph_vinop(mnt->mnt_root->d_inode));
994 ceph_mdsc_close_sessions(&client->mdsc);
995 deactivate_locked_super(sb);
999 ceph_destroy_client(client);
1001 dout("ceph_get_sb fail %d\n", err);
1005 static void ceph_kill_sb(struct super_block *s)
1007 struct ceph_client *client = ceph_sb_to_client(s);
1008 dout("kill_sb %p\n", s);
1009 ceph_mdsc_pre_umount(&client->mdsc);
1010 kill_anon_super(s); /* will call put_super after sb is r/o */
1011 ceph_destroy_client(client);
1014 static struct file_system_type ceph_fs_type = {
1015 .owner = THIS_MODULE,
1017 .get_sb = ceph_get_sb,
1018 .kill_sb = ceph_kill_sb,
1019 .fs_flags = FS_RENAME_DOES_D_MOVE,
1022 #define _STRINGIFY(x) #x
1023 #define STRINGIFY(x) _STRINGIFY(x)
1025 static int __init init_ceph(void)
1029 ret = ceph_debugfs_init();
1033 ret = ceph_msgr_init();
1037 ret = init_caches();
1041 ret = register_filesystem(&ceph_fs_type);
1045 pr_info("loaded (mon/mds/osd proto %d/%d/%d, osdmap %d/%d %d/%d)\n",
1046 CEPH_MONC_PROTOCOL, CEPH_MDSC_PROTOCOL, CEPH_OSDC_PROTOCOL,
1047 CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
1048 CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
1056 ceph_debugfs_cleanup();
1061 static void __exit exit_ceph(void)
1063 dout("exit_ceph\n");
1064 unregister_filesystem(&ceph_fs_type);
1067 ceph_debugfs_cleanup();
1070 module_init(init_ceph);
1071 module_exit(exit_ceph);
1073 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
1074 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
1075 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
1076 MODULE_DESCRIPTION("Ceph filesystem for Linux");
1077 MODULE_LICENSE("GPL");