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
= PATH_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 (strcmp(args
->snapdir_name
, CEPH_SNAPDIRNAME_DEFAULT
))
194 seq_printf(m
, ",snapdirname=%s", args
->snapdir_name
);
196 seq_printf(m
, ",name=%s", args
->name
);
198 seq_puts(m
, ",secret=<hidden>");
205 struct kmem_cache
*ceph_inode_cachep
;
206 struct kmem_cache
*ceph_cap_cachep
;
207 struct kmem_cache
*ceph_dentry_cachep
;
208 struct kmem_cache
*ceph_file_cachep
;
210 static void ceph_inode_init_once(void *foo
)
212 struct ceph_inode_info
*ci
= foo
;
213 inode_init_once(&ci
->vfs_inode
);
216 static int __init
init_caches(void)
218 ceph_inode_cachep
= kmem_cache_create("ceph_inode_info",
219 sizeof(struct ceph_inode_info
),
220 __alignof__(struct ceph_inode_info
),
221 (SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
),
222 ceph_inode_init_once
);
223 if (ceph_inode_cachep
== NULL
)
226 ceph_cap_cachep
= KMEM_CACHE(ceph_cap
,
227 SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
);
228 if (ceph_cap_cachep
== NULL
)
231 ceph_dentry_cachep
= KMEM_CACHE(ceph_dentry_info
,
232 SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
);
233 if (ceph_dentry_cachep
== NULL
)
236 ceph_file_cachep
= KMEM_CACHE(ceph_file_info
,
237 SLAB_RECLAIM_ACCOUNT
|SLAB_MEM_SPREAD
);
238 if (ceph_file_cachep
== NULL
)
244 kmem_cache_destroy(ceph_dentry_cachep
);
246 kmem_cache_destroy(ceph_cap_cachep
);
248 kmem_cache_destroy(ceph_inode_cachep
);
252 static void destroy_caches(void)
254 kmem_cache_destroy(ceph_inode_cachep
);
255 kmem_cache_destroy(ceph_cap_cachep
);
256 kmem_cache_destroy(ceph_dentry_cachep
);
257 kmem_cache_destroy(ceph_file_cachep
);
262 * ceph_umount_begin - initiate forced umount. Tear down down the
263 * mount, skipping steps that may hang while waiting for server(s).
265 static void ceph_umount_begin(struct super_block
*sb
)
267 struct ceph_client
*client
= ceph_sb_to_client(sb
);
269 dout("ceph_umount_begin - starting forced umount\n");
272 client
->mount_state
= CEPH_MOUNT_SHUTDOWN
;
276 static const struct super_operations ceph_super_ops
= {
277 .alloc_inode
= ceph_alloc_inode
,
278 .destroy_inode
= ceph_destroy_inode
,
279 .write_inode
= ceph_write_inode
,
280 .sync_fs
= ceph_syncfs
,
281 .put_super
= ceph_put_super
,
282 .show_options
= ceph_show_options
,
283 .statfs
= ceph_statfs
,
284 .umount_begin
= ceph_umount_begin
,
288 const char *ceph_msg_type_name(int type
)
291 case CEPH_MSG_SHUTDOWN
: return "shutdown";
292 case CEPH_MSG_PING
: return "ping";
293 case CEPH_MSG_AUTH
: return "auth";
294 case CEPH_MSG_AUTH_REPLY
: return "auth_reply";
295 case CEPH_MSG_MON_MAP
: return "mon_map";
296 case CEPH_MSG_MON_GET_MAP
: return "mon_get_map";
297 case CEPH_MSG_MON_SUBSCRIBE
: return "mon_subscribe";
298 case CEPH_MSG_MON_SUBSCRIBE_ACK
: return "mon_subscribe_ack";
299 case CEPH_MSG_STATFS
: return "statfs";
300 case CEPH_MSG_STATFS_REPLY
: return "statfs_reply";
301 case CEPH_MSG_MDS_MAP
: return "mds_map";
302 case CEPH_MSG_CLIENT_SESSION
: return "client_session";
303 case CEPH_MSG_CLIENT_RECONNECT
: return "client_reconnect";
304 case CEPH_MSG_CLIENT_REQUEST
: return "client_request";
305 case CEPH_MSG_CLIENT_REQUEST_FORWARD
: return "client_request_forward";
306 case CEPH_MSG_CLIENT_REPLY
: return "client_reply";
307 case CEPH_MSG_CLIENT_CAPS
: return "client_caps";
308 case CEPH_MSG_CLIENT_CAPRELEASE
: return "client_cap_release";
309 case CEPH_MSG_CLIENT_SNAP
: return "client_snap";
310 case CEPH_MSG_CLIENT_LEASE
: return "client_lease";
311 case CEPH_MSG_OSD_MAP
: return "osd_map";
312 case CEPH_MSG_OSD_OP
: return "osd_op";
313 case CEPH_MSG_OSD_OPREPLY
: return "osd_opreply";
314 default: return "unknown";
329 Opt_osdkeepalivetimeout
,
332 Opt_caps_wanted_delay_min
,
333 Opt_caps_wanted_delay_max
,
334 Opt_cap_release_safety
,
335 Opt_readdir_max_entries
,
343 /* string args above */
354 static match_table_t arg_tokens
= {
355 {Opt_fsidmajor
, "fsidmajor=%ld"},
356 {Opt_fsidminor
, "fsidminor=%ld"},
357 {Opt_monport
, "monport=%d"},
358 {Opt_wsize
, "wsize=%d"},
359 {Opt_rsize
, "rsize=%d"},
360 {Opt_osdtimeout
, "osdtimeout=%d"},
361 {Opt_osdkeepalivetimeout
, "osdkeepalive=%d"},
362 {Opt_mount_timeout
, "mount_timeout=%d"},
363 {Opt_osd_idle_ttl
, "osd_idle_ttl=%d"},
364 {Opt_caps_wanted_delay_min
, "caps_wanted_delay_min=%d"},
365 {Opt_caps_wanted_delay_max
, "caps_wanted_delay_max=%d"},
366 {Opt_cap_release_safety
, "cap_release_safety=%d"},
367 {Opt_readdir_max_entries
, "readdir_max_entries=%d"},
368 {Opt_congestion_kb
, "write_congestion_kb=%d"},
370 {Opt_snapdirname
, "snapdirname=%s"},
371 {Opt_name
, "name=%s"},
372 {Opt_secret
, "secret=%s"},
373 /* string args above */
375 {Opt_noshare
, "noshare"},
376 {Opt_dirstat
, "dirstat"},
377 {Opt_nodirstat
, "nodirstat"},
378 {Opt_rbytes
, "rbytes"},
379 {Opt_norbytes
, "norbytes"},
380 {Opt_nocrc
, "nocrc"},
381 {Opt_noasyncreaddir
, "noasyncreaddir"},
386 static struct ceph_mount_args
*parse_mount_args(int flags
, char *options
,
387 const char *dev_name
,
390 struct ceph_mount_args
*args
;
393 substring_t argstr
[MAX_OPT_ARGS
];
395 args
= kzalloc(sizeof(*args
), GFP_KERNEL
);
397 return ERR_PTR(-ENOMEM
);
398 args
->mon_addr
= kcalloc(CEPH_MAX_MON
, sizeof(*args
->mon_addr
),
403 dout("parse_mount_args %p, dev_name '%s'\n", args
, dev_name
);
405 /* start with defaults */
406 args
->sb_flags
= flags
;
407 args
->flags
= CEPH_OPT_DEFAULT
;
408 args
->osd_timeout
= CEPH_OSD_TIMEOUT_DEFAULT
;
409 args
->osd_keepalive_timeout
= CEPH_OSD_KEEPALIVE_DEFAULT
;
410 args
->mount_timeout
= CEPH_MOUNT_TIMEOUT_DEFAULT
; /* seconds */
411 args
->osd_idle_ttl
= CEPH_OSD_IDLE_TTL_DEFAULT
; /* seconds */
412 args
->caps_wanted_delay_min
= CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT
;
413 args
->caps_wanted_delay_max
= CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT
;
414 args
->rsize
= CEPH_MOUNT_RSIZE_DEFAULT
;
415 args
->snapdir_name
= kstrdup(CEPH_SNAPDIRNAME_DEFAULT
, GFP_KERNEL
);
416 args
->cap_release_safety
= CEPH_CAP_RELEASE_SAFETY_DEFAULT
;
417 args
->max_readdir
= CEPH_MAX_READDIR_DEFAULT
;
418 args
->congestion_kb
= default_congestion_kb();
420 /* ip1[:port1][,ip2[:port2]...]:/subdir/in/fs */
424 *path
= strstr(dev_name
, ":/");
426 pr_err("device name is missing path (no :/ in %s)\n",
432 err
= ceph_parse_ips(dev_name
, *path
, args
->mon_addr
,
433 CEPH_MAX_MON
, &args
->num_mon
);
439 dout("server path '%s'\n", *path
);
441 /* parse mount options */
442 while ((c
= strsep(&options
, ",")) != NULL
) {
443 int token
, intval
, ret
;
447 token
= match_token((char *)c
, arg_tokens
, argstr
);
449 pr_err("bad mount option at '%s'\n", c
);
452 if (token
< Opt_last_int
) {
453 ret
= match_int(&argstr
[0], &intval
);
455 pr_err("bad mount option arg (not int) "
459 dout("got int token %d val %d\n", token
, intval
);
460 } else if (token
> Opt_last_int
&& token
< Opt_last_string
) {
461 dout("got string token %d val %s\n", token
,
464 dout("got token %d\n", token
);
468 *(__le64
*)&args
->fsid
.fsid
[0] = cpu_to_le64(intval
);
471 *(__le64
*)&args
->fsid
.fsid
[8] = cpu_to_le64(intval
);
474 err
= ceph_parse_ips(argstr
[0].from
,
480 args
->flags
|= CEPH_OPT_MYIP
;
483 case Opt_snapdirname
:
484 kfree(args
->snapdir_name
);
485 args
->snapdir_name
= kstrndup(argstr
[0].from
,
486 argstr
[0].to
-argstr
[0].from
,
490 args
->name
= kstrndup(argstr
[0].from
,
491 argstr
[0].to
-argstr
[0].from
,
495 args
->secret
= kstrndup(argstr
[0].from
,
496 argstr
[0].to
-argstr
[0].from
,
502 args
->wsize
= intval
;
505 args
->rsize
= intval
;
508 args
->osd_timeout
= intval
;
510 case Opt_osdkeepalivetimeout
:
511 args
->osd_keepalive_timeout
= intval
;
513 case Opt_mount_timeout
:
514 args
->mount_timeout
= intval
;
516 case Opt_caps_wanted_delay_min
:
517 args
->caps_wanted_delay_min
= intval
;
519 case Opt_caps_wanted_delay_max
:
520 args
->caps_wanted_delay_max
= intval
;
522 case Opt_readdir_max_entries
:
523 args
->max_readdir
= intval
;
525 case Opt_congestion_kb
:
526 args
->congestion_kb
= intval
;
530 args
->flags
|= CEPH_OPT_NOSHARE
;
534 args
->flags
|= CEPH_OPT_DIRSTAT
;
537 args
->flags
&= ~CEPH_OPT_DIRSTAT
;
540 args
->flags
|= CEPH_OPT_RBYTES
;
543 args
->flags
&= ~CEPH_OPT_RBYTES
;
546 args
->flags
|= CEPH_OPT_NOCRC
;
548 case Opt_noasyncreaddir
:
549 args
->flags
|= CEPH_OPT_NOASYNCREADDIR
;
559 kfree(args
->mon_addr
);
564 static void destroy_mount_args(struct ceph_mount_args
*args
)
566 dout("destroy_mount_args %p\n", args
);
567 kfree(args
->snapdir_name
);
568 args
->snapdir_name
= NULL
;
577 * create a fresh client instance
579 static struct ceph_client
*ceph_create_client(struct ceph_mount_args
*args
)
581 struct ceph_client
*client
;
584 client
= kzalloc(sizeof(*client
), GFP_KERNEL
);
586 return ERR_PTR(-ENOMEM
);
588 mutex_init(&client
->mount_mutex
);
590 init_waitqueue_head(&client
->auth_wq
);
593 client
->mount_state
= CEPH_MOUNT_MOUNTING
;
594 client
->mount_args
= args
;
598 client
->auth_err
= 0;
599 atomic_long_set(&client
->writeback_count
, 0);
601 err
= bdi_init(&client
->backing_dev_info
);
606 client
->wb_wq
= create_workqueue("ceph-writeback");
607 if (client
->wb_wq
== NULL
)
609 client
->pg_inv_wq
= create_singlethread_workqueue("ceph-pg-invalid");
610 if (client
->pg_inv_wq
== NULL
)
612 client
->trunc_wq
= create_singlethread_workqueue("ceph-trunc");
613 if (client
->trunc_wq
== NULL
)
616 /* set up mempools */
618 client
->wb_pagevec_pool
= mempool_create_kmalloc_pool(10,
619 client
->mount_args
->wsize
>> PAGE_CACHE_SHIFT
);
620 if (!client
->wb_pagevec_pool
)
624 client
->min_caps
= args
->max_readdir
;
625 ceph_adjust_min_caps(client
->min_caps
);
628 err
= ceph_monc_init(&client
->monc
, client
);
631 err
= ceph_osdc_init(&client
->osdc
, client
);
634 err
= ceph_mdsc_init(&client
->mdsc
, client
);
640 ceph_osdc_stop(&client
->osdc
);
642 ceph_monc_stop(&client
->monc
);
644 mempool_destroy(client
->wb_pagevec_pool
);
646 destroy_workqueue(client
->trunc_wq
);
648 destroy_workqueue(client
->pg_inv_wq
);
650 destroy_workqueue(client
->wb_wq
);
652 bdi_destroy(&client
->backing_dev_info
);
658 static void ceph_destroy_client(struct ceph_client
*client
)
660 dout("destroy_client %p\n", client
);
663 ceph_mdsc_stop(&client
->mdsc
);
664 ceph_monc_stop(&client
->monc
);
665 ceph_osdc_stop(&client
->osdc
);
667 ceph_adjust_min_caps(-client
->min_caps
);
669 ceph_debugfs_client_cleanup(client
);
670 destroy_workqueue(client
->wb_wq
);
671 destroy_workqueue(client
->pg_inv_wq
);
672 destroy_workqueue(client
->trunc_wq
);
674 bdi_destroy(&client
->backing_dev_info
);
677 ceph_messenger_destroy(client
->msgr
);
678 mempool_destroy(client
->wb_pagevec_pool
);
680 destroy_mount_args(client
->mount_args
);
683 dout("destroy_client %p done\n", client
);
687 * Initially learn our fsid, or verify an fsid matches.
689 int ceph_check_fsid(struct ceph_client
*client
, struct ceph_fsid
*fsid
)
691 if (client
->have_fsid
) {
692 if (ceph_fsid_compare(&client
->fsid
, fsid
)) {
693 pr_err("bad fsid, had " FSID_FORMAT
" got " FSID_FORMAT
,
694 PR_FSID(&client
->fsid
), PR_FSID(fsid
));
698 pr_info("client%lld fsid " FSID_FORMAT
"\n",
699 client
->monc
.auth
->global_id
, PR_FSID(fsid
));
700 memcpy(&client
->fsid
, fsid
, sizeof(*fsid
));
701 ceph_debugfs_client_init(client
);
702 client
->have_fsid
= true;
708 * true if we have the mon map (and have thus joined the cluster)
710 static int have_mon_and_osd_map(struct ceph_client
*client
)
712 return client
->monc
.monmap
&& client
->monc
.monmap
->epoch
&&
713 client
->osdc
.osdmap
&& client
->osdc
.osdmap
->epoch
;
717 * Bootstrap mount by opening the root directory. Note the mount
718 * @started time from caller, and time out if this takes too long.
720 static struct dentry
*open_root_dentry(struct ceph_client
*client
,
722 unsigned long started
)
724 struct ceph_mds_client
*mdsc
= &client
->mdsc
;
725 struct ceph_mds_request
*req
= NULL
;
730 dout("open_root_inode opening '%s'\n", path
);
731 req
= ceph_mdsc_create_request(mdsc
, CEPH_MDS_OP_GETATTR
, USE_ANY_MDS
);
733 return ERR_PTR(PTR_ERR(req
));
734 req
->r_path1
= kstrdup(path
, GFP_NOFS
);
735 req
->r_ino1
.ino
= CEPH_INO_ROOT
;
736 req
->r_ino1
.snap
= CEPH_NOSNAP
;
737 req
->r_started
= started
;
738 req
->r_timeout
= client
->mount_args
->mount_timeout
* HZ
;
739 req
->r_args
.getattr
.mask
= cpu_to_le32(CEPH_STAT_CAP_INODE
);
741 err
= ceph_mdsc_do_request(mdsc
, NULL
, req
);
743 dout("open_root_inode success\n");
744 if (ceph_ino(req
->r_target_inode
) == CEPH_INO_ROOT
&&
745 client
->sb
->s_root
== NULL
)
746 root
= d_alloc_root(req
->r_target_inode
);
748 root
= d_obtain_alias(req
->r_target_inode
);
749 req
->r_target_inode
= NULL
;
750 dout("open_root_inode success, root dentry is %p\n", root
);
754 ceph_mdsc_put_request(req
);
759 * mount: join the ceph cluster, and open root directory.
761 static int ceph_mount(struct ceph_client
*client
, struct vfsmount
*mnt
,
764 struct ceph_entity_addr
*myaddr
= NULL
;
766 unsigned long timeout
= client
->mount_args
->mount_timeout
* HZ
;
767 unsigned long started
= jiffies
; /* note the start time */
770 dout("mount start\n");
771 mutex_lock(&client
->mount_mutex
);
773 /* initialize the messenger */
774 if (client
->msgr
== NULL
) {
775 if (ceph_test_opt(client
, MYIP
))
776 myaddr
= &client
->mount_args
->my_addr
;
777 client
->msgr
= ceph_messenger_create(myaddr
);
778 if (IS_ERR(client
->msgr
)) {
779 err
= PTR_ERR(client
->msgr
);
783 client
->msgr
->nocrc
= ceph_test_opt(client
, NOCRC
);
786 /* open session, and wait for mon, mds, and osd maps */
787 err
= ceph_monc_open_session(&client
->monc
);
791 while (!have_mon_and_osd_map(client
)) {
793 if (timeout
&& time_after_eq(jiffies
, started
+ timeout
))
797 dout("mount waiting for mon_map\n");
798 err
= wait_event_interruptible_timeout(client
->auth_wq
,
799 have_mon_and_osd_map(client
) || (client
->auth_err
< 0),
801 if (err
== -EINTR
|| err
== -ERESTARTSYS
)
803 if (client
->auth_err
< 0) {
804 err
= client
->auth_err
;
809 dout("mount opening root\n");
810 root
= open_root_dentry(client
, "", started
);
815 if (client
->sb
->s_root
)
818 client
->sb
->s_root
= root
;
823 dout("mount opening base mountpoint\n");
824 root
= open_root_dentry(client
, path
, started
);
827 dput(client
->sb
->s_root
);
828 client
->sb
->s_root
= NULL
;
833 mnt
->mnt_root
= root
;
834 mnt
->mnt_sb
= client
->sb
;
836 client
->mount_state
= CEPH_MOUNT_MOUNTED
;
837 dout("mount success\n");
841 mutex_unlock(&client
->mount_mutex
);
845 static int ceph_set_super(struct super_block
*s
, void *data
)
847 struct ceph_client
*client
= data
;
850 dout("set_super %p data %p\n", s
, data
);
852 s
->s_flags
= client
->mount_args
->sb_flags
;
853 s
->s_maxbytes
= 1ULL << 40; /* temp value until we get mdsmap */
855 s
->s_fs_info
= client
;
858 s
->s_op
= &ceph_super_ops
;
859 s
->s_export_op
= &ceph_export_ops
;
861 s
->s_time_gran
= 1000; /* 1000 ns == 1 us */
863 ret
= set_anon_super(s
, NULL
); /* what is that second arg for? */
876 * share superblock if same fs AND options
878 static int ceph_compare_super(struct super_block
*sb
, void *data
)
880 struct ceph_client
*new = data
;
881 struct ceph_mount_args
*args
= new->mount_args
;
882 struct ceph_client
*other
= ceph_sb_to_client(sb
);
885 dout("ceph_compare_super %p\n", sb
);
886 if (args
->flags
& CEPH_OPT_FSID
) {
887 if (ceph_fsid_compare(&args
->fsid
, &other
->fsid
)) {
888 dout("fsid doesn't match\n");
892 /* do we share (a) monitor? */
893 for (i
= 0; i
< new->monc
.monmap
->num_mon
; i
++)
894 if (ceph_monmap_contains(other
->monc
.monmap
,
895 &new->monc
.monmap
->mon_inst
[i
].addr
))
897 if (i
== new->monc
.monmap
->num_mon
) {
898 dout("mon ip not part of monmap\n");
901 dout("mon ip matches existing sb %p\n", sb
);
903 if (args
->sb_flags
!= other
->mount_args
->sb_flags
) {
904 dout("flags differ\n");
911 * construct our own bdi so we can control readahead, etc.
913 static atomic_long_t bdi_seq
= ATOMIC_INIT(0);
915 static int ceph_register_bdi(struct super_block
*sb
, struct ceph_client
*client
)
919 /* set ra_pages based on rsize mount option? */
920 if (client
->mount_args
->rsize
>= PAGE_CACHE_SIZE
)
921 client
->backing_dev_info
.ra_pages
=
922 (client
->mount_args
->rsize
+ PAGE_CACHE_SIZE
- 1)
924 err
= bdi_register(&client
->backing_dev_info
, NULL
, "ceph-%d",
925 atomic_long_inc_return(&bdi_seq
));
927 sb
->s_bdi
= &client
->backing_dev_info
;
931 static int ceph_get_sb(struct file_system_type
*fs_type
,
932 int flags
, const char *dev_name
, void *data
,
933 struct vfsmount
*mnt
)
935 struct super_block
*sb
;
936 struct ceph_client
*client
;
938 int (*compare_super
)(struct super_block
*, void *) = ceph_compare_super
;
939 const char *path
= NULL
;
940 struct ceph_mount_args
*args
;
942 dout("ceph_get_sb\n");
943 args
= parse_mount_args(flags
, data
, dev_name
, &path
);
949 /* create client (which we may/may not use) */
950 client
= ceph_create_client(args
);
951 if (IS_ERR(client
)) {
952 err
= PTR_ERR(client
);
956 if (client
->mount_args
->flags
& CEPH_OPT_NOSHARE
)
957 compare_super
= NULL
;
958 sb
= sget(fs_type
, compare_super
, ceph_set_super
, client
);
964 if (ceph_sb_to_client(sb
) != client
) {
965 ceph_destroy_client(client
);
966 client
= ceph_sb_to_client(sb
);
967 dout("get_sb got existing client %p\n", client
);
969 dout("get_sb using new client %p\n", client
);
970 err
= ceph_register_bdi(sb
, client
);
975 err
= ceph_mount(client
, mnt
, path
);
978 dout("root %p inode %p ino %llx.%llx\n", mnt
->mnt_root
,
979 mnt
->mnt_root
->d_inode
, ceph_vinop(mnt
->mnt_root
->d_inode
));
983 ceph_mdsc_close_sessions(&client
->mdsc
);
984 up_write(&sb
->s_umount
);
985 deactivate_super(sb
);
989 ceph_destroy_client(client
);
991 dout("ceph_get_sb fail %d\n", err
);
995 static void ceph_kill_sb(struct super_block
*s
)
997 struct ceph_client
*client
= ceph_sb_to_client(s
);
998 dout("kill_sb %p\n", s
);
999 ceph_mdsc_pre_umount(&client
->mdsc
);
1000 kill_anon_super(s
); /* will call put_super after sb is r/o */
1001 ceph_destroy_client(client
);
1004 static struct file_system_type ceph_fs_type
= {
1005 .owner
= THIS_MODULE
,
1007 .get_sb
= ceph_get_sb
,
1008 .kill_sb
= ceph_kill_sb
,
1009 .fs_flags
= FS_RENAME_DOES_D_MOVE
,
1012 #define _STRINGIFY(x) #x
1013 #define STRINGIFY(x) _STRINGIFY(x)
1015 static int __init
init_ceph(void)
1019 ret
= ceph_debugfs_init();
1023 ret
= ceph_msgr_init();
1027 ret
= init_caches();
1033 ret
= register_filesystem(&ceph_fs_type
);
1037 pr_info("loaded (mon/mds/osd proto %d/%d/%d, osdmap %d/%d %d/%d)\n",
1038 CEPH_MONC_PROTOCOL
, CEPH_MDSC_PROTOCOL
, CEPH_OSDC_PROTOCOL
,
1039 CEPH_OSDMAP_VERSION
, CEPH_OSDMAP_VERSION_EXT
,
1040 CEPH_OSDMAP_INC_VERSION
, CEPH_OSDMAP_INC_VERSION_EXT
);
1048 ceph_debugfs_cleanup();
1053 static void __exit
exit_ceph(void)
1055 dout("exit_ceph\n");
1056 unregister_filesystem(&ceph_fs_type
);
1057 ceph_caps_finalize();
1060 ceph_debugfs_cleanup();
1063 module_init(init_ceph
);
1064 module_exit(exit_ceph
);
1066 MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
1067 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
1068 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
1069 MODULE_DESCRIPTION("Ceph filesystem for Linux");
1070 MODULE_LICENSE("GPL");