Commit | Line | Data |
---|---|---|
16725b9d | 1 | |
3d14c5d2 | 2 | #include <linux/ceph/ceph_debug.h> |
16725b9d SW |
3 | |
4 | #include <linux/backing-dev.h> | |
c309f0ab | 5 | #include <linux/ctype.h> |
16725b9d SW |
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> | |
16725b9d SW |
12 | #include <linux/sched.h> |
13 | #include <linux/seq_file.h> | |
5a0e3ad6 | 14 | #include <linux/slab.h> |
16725b9d SW |
15 | #include <linux/statfs.h> |
16 | #include <linux/string.h> | |
16725b9d | 17 | |
16725b9d | 18 | #include "super.h" |
3d14c5d2 YS |
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> | |
16725b9d SW |
25 | |
26 | /* | |
27 | * Ceph superblock operations | |
28 | * | |
29 | * Handle the basics of mounting, unmounting. | |
30 | */ | |
31 | ||
16725b9d SW |
32 | /* |
33 | * super ops | |
34 | */ | |
35 | static void ceph_put_super(struct super_block *s) | |
36 | { | |
3d14c5d2 | 37 | struct ceph_fs_client *fsc = ceph_sb_to_client(s); |
16725b9d SW |
38 | |
39 | dout("put_super\n"); | |
3d14c5d2 | 40 | ceph_mdsc_close_sessions(fsc->mdsc); |
5dfc589a SW |
41 | |
42 | /* | |
43 | * ensure we release the bdi before put_anon_super releases | |
44 | * the device name. | |
45 | */ | |
3d14c5d2 YS |
46 | if (s->s_bdi == &fsc->backing_dev_info) { |
47 | bdi_unregister(&fsc->backing_dev_info); | |
5dfc589a SW |
48 | s->s_bdi = NULL; |
49 | } | |
50 | ||
16725b9d SW |
51 | return; |
52 | } | |
53 | ||
54 | static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf) | |
55 | { | |
3d14c5d2 YS |
56 | struct ceph_fs_client *fsc = ceph_inode_to_client(dentry->d_inode); |
57 | struct ceph_monmap *monmap = fsc->client->monc.monmap; | |
16725b9d SW |
58 | struct ceph_statfs st; |
59 | u64 fsid; | |
60 | int err; | |
61 | ||
62 | dout("statfs\n"); | |
3d14c5d2 | 63 | err = ceph_monc_do_statfs(&fsc->client->monc, &st); |
16725b9d SW |
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 | buf->f_bsize = 1 << CEPH_BLOCK_SHIFT; | |
75 | buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10); | |
76 | buf->f_bfree = (le64_to_cpu(st.kb) - le64_to_cpu(st.kb_used)) >> | |
77 | (CEPH_BLOCK_SHIFT-10); | |
78 | buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10); | |
79 | ||
80 | buf->f_files = le64_to_cpu(st.num_objects); | |
81 | buf->f_ffree = -1; | |
558d3499 | 82 | buf->f_namelen = NAME_MAX; |
16725b9d SW |
83 | buf->f_frsize = PAGE_CACHE_SIZE; |
84 | ||
85 | /* leave fsid little-endian, regardless of host endianness */ | |
86 | fsid = *(u64 *)(&monmap->fsid) ^ *((u64 *)&monmap->fsid + 1); | |
87 | buf->f_fsid.val[0] = fsid & 0xffffffff; | |
88 | buf->f_fsid.val[1] = fsid >> 32; | |
89 | ||
90 | return 0; | |
91 | } | |
92 | ||
93 | ||
2d9c98ae | 94 | static int ceph_sync_fs(struct super_block *sb, int wait) |
16725b9d | 95 | { |
3d14c5d2 | 96 | struct ceph_fs_client *fsc = ceph_sb_to_client(sb); |
2d9c98ae SW |
97 | |
98 | if (!wait) { | |
99 | dout("sync_fs (non-blocking)\n"); | |
3d14c5d2 | 100 | ceph_flush_dirty_caps(fsc->mdsc); |
2d9c98ae SW |
101 | dout("sync_fs (non-blocking) done\n"); |
102 | return 0; | |
103 | } | |
104 | ||
105 | dout("sync_fs (blocking)\n"); | |
3d14c5d2 YS |
106 | ceph_osdc_sync(&fsc->client->osdc); |
107 | ceph_mdsc_sync(fsc->mdsc); | |
2d9c98ae | 108 | dout("sync_fs (blocking) done\n"); |
16725b9d SW |
109 | return 0; |
110 | } | |
111 | ||
16725b9d SW |
112 | /* |
113 | * mount options | |
114 | */ | |
115 | enum { | |
16725b9d SW |
116 | Opt_wsize, |
117 | Opt_rsize, | |
16725b9d SW |
118 | Opt_caps_wanted_delay_min, |
119 | Opt_caps_wanted_delay_max, | |
6e19a16e | 120 | Opt_cap_release_safety, |
16725b9d | 121 | Opt_readdir_max_entries, |
23804d91 | 122 | Opt_readdir_max_bytes, |
2baba250 | 123 | Opt_congestion_kb, |
e53c2fe0 | 124 | Opt_last_int, |
16725b9d SW |
125 | /* int args above */ |
126 | Opt_snapdirname, | |
e53c2fe0 | 127 | Opt_last_string, |
16725b9d | 128 | /* string args above */ |
16725b9d SW |
129 | Opt_dirstat, |
130 | Opt_nodirstat, | |
131 | Opt_rbytes, | |
132 | Opt_norbytes, | |
16725b9d SW |
133 | Opt_noasyncreaddir, |
134 | }; | |
135 | ||
3d14c5d2 | 136 | static match_table_t fsopt_tokens = { |
16725b9d SW |
137 | {Opt_wsize, "wsize=%d"}, |
138 | {Opt_rsize, "rsize=%d"}, | |
16725b9d SW |
139 | {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"}, |
140 | {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"}, | |
6e19a16e | 141 | {Opt_cap_release_safety, "cap_release_safety=%d"}, |
16725b9d | 142 | {Opt_readdir_max_entries, "readdir_max_entries=%d"}, |
23804d91 | 143 | {Opt_readdir_max_bytes, "readdir_max_bytes=%d"}, |
2baba250 | 144 | {Opt_congestion_kb, "write_congestion_kb=%d"}, |
16725b9d SW |
145 | /* int args above */ |
146 | {Opt_snapdirname, "snapdirname=%s"}, | |
16725b9d | 147 | /* string args above */ |
16725b9d SW |
148 | {Opt_dirstat, "dirstat"}, |
149 | {Opt_nodirstat, "nodirstat"}, | |
150 | {Opt_rbytes, "rbytes"}, | |
151 | {Opt_norbytes, "norbytes"}, | |
16725b9d SW |
152 | {Opt_noasyncreaddir, "noasyncreaddir"}, |
153 | {-1, NULL} | |
154 | }; | |
155 | ||
3d14c5d2 | 156 | static int parse_fsopt_token(char *c, void *private) |
c309f0ab | 157 | { |
3d14c5d2 YS |
158 | struct ceph_mount_options *fsopt = private; |
159 | substring_t argstr[MAX_OPT_ARGS]; | |
160 | int token, intval, ret; | |
161 | ||
162 | token = match_token((char *)c, fsopt_tokens, argstr); | |
163 | if (token < 0) | |
164 | return -EINVAL; | |
165 | ||
166 | if (token < Opt_last_int) { | |
167 | ret = match_int(&argstr[0], &intval); | |
168 | if (ret < 0) { | |
169 | pr_err("bad mount option arg (not int) " | |
170 | "at '%s'\n", c); | |
171 | return ret; | |
c309f0ab | 172 | } |
3d14c5d2 YS |
173 | dout("got int token %d val %d\n", token, intval); |
174 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
175 | dout("got string token %d val %s\n", token, | |
176 | argstr[0].from); | |
177 | } else { | |
178 | dout("got token %d\n", token); | |
c309f0ab SW |
179 | } |
180 | ||
3d14c5d2 YS |
181 | switch (token) { |
182 | case Opt_snapdirname: | |
183 | kfree(fsopt->snapdir_name); | |
184 | fsopt->snapdir_name = kstrndup(argstr[0].from, | |
185 | argstr[0].to-argstr[0].from, | |
186 | GFP_KERNEL); | |
187 | if (!fsopt->snapdir_name) | |
188 | return -ENOMEM; | |
189 | break; | |
190 | ||
191 | /* misc */ | |
192 | case Opt_wsize: | |
193 | fsopt->wsize = intval; | |
194 | break; | |
195 | case Opt_rsize: | |
196 | fsopt->rsize = intval; | |
197 | break; | |
198 | case Opt_caps_wanted_delay_min: | |
199 | fsopt->caps_wanted_delay_min = intval; | |
200 | break; | |
201 | case Opt_caps_wanted_delay_max: | |
202 | fsopt->caps_wanted_delay_max = intval; | |
203 | break; | |
204 | case Opt_readdir_max_entries: | |
205 | fsopt->max_readdir = intval; | |
206 | break; | |
207 | case Opt_readdir_max_bytes: | |
208 | fsopt->max_readdir_bytes = intval; | |
209 | break; | |
210 | case Opt_congestion_kb: | |
211 | fsopt->congestion_kb = intval; | |
212 | break; | |
213 | case Opt_dirstat: | |
214 | fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT; | |
215 | break; | |
216 | case Opt_nodirstat: | |
217 | fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT; | |
218 | break; | |
219 | case Opt_rbytes: | |
220 | fsopt->flags |= CEPH_MOUNT_OPT_RBYTES; | |
221 | break; | |
222 | case Opt_norbytes: | |
223 | fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES; | |
224 | break; | |
225 | case Opt_noasyncreaddir: | |
226 | fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR; | |
227 | break; | |
228 | default: | |
229 | BUG_ON(token); | |
230 | } | |
231 | return 0; | |
c309f0ab | 232 | } |
16725b9d | 233 | |
3d14c5d2 | 234 | static void destroy_mount_options(struct ceph_mount_options *args) |
16725b9d | 235 | { |
3d14c5d2 YS |
236 | dout("destroy_mount_options %p\n", args); |
237 | kfree(args->snapdir_name); | |
238 | kfree(args); | |
239 | } | |
16725b9d | 240 | |
3d14c5d2 YS |
241 | static int strcmp_null(const char *s1, const char *s2) |
242 | { | |
243 | if (!s1 && !s2) | |
244 | return 0; | |
245 | if (s1 && !s2) | |
246 | return -1; | |
247 | if (!s1 && s2) | |
248 | return 1; | |
249 | return strcmp(s1, s2); | |
250 | } | |
16725b9d | 251 | |
3d14c5d2 YS |
252 | static int compare_mount_options(struct ceph_mount_options *new_fsopt, |
253 | struct ceph_options *new_opt, | |
254 | struct ceph_fs_client *fsc) | |
255 | { | |
256 | struct ceph_mount_options *fsopt1 = new_fsopt; | |
257 | struct ceph_mount_options *fsopt2 = fsc->mount_options; | |
258 | int ofs = offsetof(struct ceph_mount_options, snapdir_name); | |
259 | int ret; | |
16725b9d | 260 | |
3d14c5d2 YS |
261 | ret = memcmp(fsopt1, fsopt2, ofs); |
262 | if (ret) | |
263 | return ret; | |
264 | ||
265 | ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name); | |
266 | if (ret) | |
267 | return ret; | |
268 | ||
269 | return ceph_compare_options(new_opt, fsc->client); | |
270 | } | |
271 | ||
272 | static int parse_mount_options(struct ceph_mount_options **pfsopt, | |
273 | struct ceph_options **popt, | |
274 | int flags, char *options, | |
275 | const char *dev_name, | |
276 | const char **path) | |
277 | { | |
278 | struct ceph_mount_options *fsopt; | |
279 | const char *dev_name_end; | |
280 | int err = -ENOMEM; | |
281 | ||
282 | fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL); | |
283 | if (!fsopt) | |
284 | return -ENOMEM; | |
285 | ||
286 | dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name); | |
287 | ||
288 | fsopt->sb_flags = flags; | |
289 | fsopt->flags = CEPH_MOUNT_OPT_DEFAULT; | |
290 | ||
291 | fsopt->rsize = CEPH_MOUNT_RSIZE_DEFAULT; | |
292 | fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL); | |
293 | fsopt->cap_release_safety = CEPH_CAP_RELEASE_SAFETY_DEFAULT; | |
294 | fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT; | |
295 | fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT; | |
296 | fsopt->congestion_kb = default_congestion_kb(); | |
297 | ||
298 | /* ip1[:port1][,ip2[:port2]...]:/subdir/in/fs */ | |
299 | err = -EINVAL; | |
300 | if (!dev_name) | |
301 | goto out; | |
302 | *path = strstr(dev_name, ":/"); | |
303 | if (*path == NULL) { | |
304 | pr_err("device name is missing path (no :/ in %s)\n", | |
305 | dev_name); | |
306 | goto out; | |
307 | } | |
308 | dev_name_end = *path; | |
309 | dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name); | |
16725b9d | 310 | |
16725b9d SW |
311 | /* path on server */ |
312 | *path += 2; | |
313 | dout("server path '%s'\n", *path); | |
314 | ||
3d14c5d2 YS |
315 | err = ceph_parse_options(popt, options, dev_name, dev_name_end, |
316 | parse_fsopt_token, (void *)fsopt); | |
317 | if (err) | |
318 | goto out; | |
319 | ||
320 | /* success */ | |
321 | *pfsopt = fsopt; | |
322 | return 0; | |
16725b9d | 323 | |
7b813c46 | 324 | out: |
3d14c5d2 YS |
325 | destroy_mount_options(fsopt); |
326 | return err; | |
16725b9d SW |
327 | } |
328 | ||
3d14c5d2 YS |
329 | /** |
330 | * ceph_show_options - Show mount options in /proc/mounts | |
331 | * @m: seq_file to write to | |
332 | * @mnt: mount descriptor | |
333 | */ | |
334 | static int ceph_show_options(struct seq_file *m, struct vfsmount *mnt) | |
16725b9d | 335 | { |
3d14c5d2 YS |
336 | struct ceph_fs_client *fsc = ceph_sb_to_client(mnt->mnt_sb); |
337 | struct ceph_mount_options *fsopt = fsc->mount_options; | |
338 | struct ceph_options *opt = fsc->client->options; | |
339 | ||
340 | if (opt->flags & CEPH_OPT_FSID) | |
341 | seq_printf(m, ",fsid=%pU", &opt->fsid); | |
342 | if (opt->flags & CEPH_OPT_NOSHARE) | |
343 | seq_puts(m, ",noshare"); | |
344 | if (opt->flags & CEPH_OPT_NOCRC) | |
345 | seq_puts(m, ",nocrc"); | |
346 | ||
347 | if (opt->name) | |
348 | seq_printf(m, ",name=%s", opt->name); | |
349 | if (opt->secret) | |
350 | seq_puts(m, ",secret=<hidden>"); | |
351 | ||
352 | if (opt->mount_timeout != CEPH_MOUNT_TIMEOUT_DEFAULT) | |
353 | seq_printf(m, ",mount_timeout=%d", opt->mount_timeout); | |
354 | if (opt->osd_idle_ttl != CEPH_OSD_IDLE_TTL_DEFAULT) | |
355 | seq_printf(m, ",osd_idle_ttl=%d", opt->osd_idle_ttl); | |
356 | if (opt->osd_timeout != CEPH_OSD_TIMEOUT_DEFAULT) | |
357 | seq_printf(m, ",osdtimeout=%d", opt->osd_timeout); | |
358 | if (opt->osd_keepalive_timeout != CEPH_OSD_KEEPALIVE_DEFAULT) | |
359 | seq_printf(m, ",osdkeepalivetimeout=%d", | |
360 | opt->osd_keepalive_timeout); | |
361 | ||
362 | if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT) | |
363 | seq_puts(m, ",dirstat"); | |
364 | if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES) == 0) | |
365 | seq_puts(m, ",norbytes"); | |
366 | if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR) | |
367 | seq_puts(m, ",noasyncreaddir"); | |
368 | ||
369 | if (fsopt->wsize) | |
370 | seq_printf(m, ",wsize=%d", fsopt->wsize); | |
371 | if (fsopt->rsize != CEPH_MOUNT_RSIZE_DEFAULT) | |
372 | seq_printf(m, ",rsize=%d", fsopt->rsize); | |
373 | if (fsopt->congestion_kb != default_congestion_kb()) | |
374 | seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb); | |
375 | if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT) | |
376 | seq_printf(m, ",caps_wanted_delay_min=%d", | |
377 | fsopt->caps_wanted_delay_min); | |
378 | if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT) | |
379 | seq_printf(m, ",caps_wanted_delay_max=%d", | |
380 | fsopt->caps_wanted_delay_max); | |
381 | if (fsopt->cap_release_safety != CEPH_CAP_RELEASE_SAFETY_DEFAULT) | |
382 | seq_printf(m, ",cap_release_safety=%d", | |
383 | fsopt->cap_release_safety); | |
384 | if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT) | |
385 | seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir); | |
386 | if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT) | |
387 | seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes); | |
388 | if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT)) | |
389 | seq_printf(m, ",snapdirname=%s", fsopt->snapdir_name); | |
390 | return 0; | |
16725b9d SW |
391 | } |
392 | ||
393 | /* | |
3d14c5d2 YS |
394 | * handle any mon messages the standard library doesn't understand. |
395 | * return error if we don't either. | |
16725b9d | 396 | */ |
3d14c5d2 | 397 | static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg) |
16725b9d | 398 | { |
3d14c5d2 YS |
399 | struct ceph_fs_client *fsc = client->private; |
400 | int type = le16_to_cpu(msg->hdr.type); | |
401 | ||
402 | switch (type) { | |
403 | case CEPH_MSG_MDS_MAP: | |
404 | ceph_mdsc_handle_map(fsc->mdsc, msg); | |
405 | return 0; | |
406 | ||
407 | default: | |
408 | return -1; | |
409 | } | |
410 | } | |
411 | ||
412 | /* | |
413 | * create a new fs client | |
414 | */ | |
415 | struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt, | |
416 | struct ceph_options *opt) | |
417 | { | |
418 | struct ceph_fs_client *fsc; | |
16725b9d SW |
419 | int err = -ENOMEM; |
420 | ||
3d14c5d2 YS |
421 | fsc = kzalloc(sizeof(*fsc), GFP_KERNEL); |
422 | if (!fsc) | |
16725b9d SW |
423 | return ERR_PTR(-ENOMEM); |
424 | ||
3d14c5d2 YS |
425 | fsc->client = ceph_create_client(opt, fsc); |
426 | if (IS_ERR(fsc->client)) { | |
427 | err = PTR_ERR(fsc->client); | |
428 | goto fail; | |
429 | } | |
430 | fsc->client->extra_mon_dispatch = extra_mon_dispatch; | |
14303d20 SW |
431 | fsc->client->supported_features |= CEPH_FEATURE_FLOCK | |
432 | CEPH_FEATURE_DIRLAYOUTHASH; | |
3d14c5d2 | 433 | fsc->client->monc.want_mdsmap = 1; |
16725b9d | 434 | |
3d14c5d2 | 435 | fsc->mount_options = fsopt; |
16725b9d | 436 | |
3d14c5d2 YS |
437 | fsc->sb = NULL; |
438 | fsc->mount_state = CEPH_MOUNT_MOUNTING; | |
16725b9d | 439 | |
3d14c5d2 | 440 | atomic_long_set(&fsc->writeback_count, 0); |
16725b9d | 441 | |
3d14c5d2 | 442 | err = bdi_init(&fsc->backing_dev_info); |
859e7b14 | 443 | if (err < 0) |
3d14c5d2 | 444 | goto fail_client; |
859e7b14 | 445 | |
16725b9d | 446 | err = -ENOMEM; |
3d14c5d2 YS |
447 | fsc->wb_wq = create_workqueue("ceph-writeback"); |
448 | if (fsc->wb_wq == NULL) | |
859e7b14 | 449 | goto fail_bdi; |
3d14c5d2 YS |
450 | fsc->pg_inv_wq = create_singlethread_workqueue("ceph-pg-invalid"); |
451 | if (fsc->pg_inv_wq == NULL) | |
16725b9d | 452 | goto fail_wb_wq; |
3d14c5d2 YS |
453 | fsc->trunc_wq = create_singlethread_workqueue("ceph-trunc"); |
454 | if (fsc->trunc_wq == NULL) | |
16725b9d SW |
455 | goto fail_pg_inv_wq; |
456 | ||
b9bfb93c SW |
457 | /* set up mempools */ |
458 | err = -ENOMEM; | |
3d14c5d2 YS |
459 | fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10, |
460 | fsc->mount_options->wsize >> PAGE_CACHE_SHIFT); | |
461 | if (!fsc->wb_pagevec_pool) | |
b9bfb93c SW |
462 | goto fail_trunc_wq; |
463 | ||
85ccce43 | 464 | /* caps */ |
3d14c5d2 YS |
465 | fsc->min_caps = fsopt->max_readdir; |
466 | ||
467 | return fsc; | |
b9bfb93c | 468 | |
16725b9d | 469 | fail_trunc_wq: |
3d14c5d2 | 470 | destroy_workqueue(fsc->trunc_wq); |
16725b9d | 471 | fail_pg_inv_wq: |
3d14c5d2 | 472 | destroy_workqueue(fsc->pg_inv_wq); |
16725b9d | 473 | fail_wb_wq: |
3d14c5d2 | 474 | destroy_workqueue(fsc->wb_wq); |
859e7b14 | 475 | fail_bdi: |
3d14c5d2 YS |
476 | bdi_destroy(&fsc->backing_dev_info); |
477 | fail_client: | |
478 | ceph_destroy_client(fsc->client); | |
16725b9d | 479 | fail: |
3d14c5d2 | 480 | kfree(fsc); |
16725b9d SW |
481 | return ERR_PTR(err); |
482 | } | |
483 | ||
3d14c5d2 | 484 | void destroy_fs_client(struct ceph_fs_client *fsc) |
16725b9d | 485 | { |
3d14c5d2 | 486 | dout("destroy_fs_client %p\n", fsc); |
16725b9d | 487 | |
3d14c5d2 YS |
488 | destroy_workqueue(fsc->wb_wq); |
489 | destroy_workqueue(fsc->pg_inv_wq); | |
490 | destroy_workqueue(fsc->trunc_wq); | |
16725b9d | 491 | |
3d14c5d2 | 492 | bdi_destroy(&fsc->backing_dev_info); |
a922d38f | 493 | |
3d14c5d2 | 494 | mempool_destroy(fsc->wb_pagevec_pool); |
16725b9d | 495 | |
3d14c5d2 | 496 | destroy_mount_options(fsc->mount_options); |
5dfc589a | 497 | |
3d14c5d2 | 498 | ceph_fs_debugfs_cleanup(fsc); |
16725b9d | 499 | |
3d14c5d2 | 500 | ceph_destroy_client(fsc->client); |
16725b9d | 501 | |
3d14c5d2 YS |
502 | kfree(fsc); |
503 | dout("destroy_fs_client %p done\n", fsc); | |
16725b9d SW |
504 | } |
505 | ||
0743304d | 506 | /* |
3d14c5d2 | 507 | * caches |
0743304d | 508 | */ |
3d14c5d2 YS |
509 | struct kmem_cache *ceph_inode_cachep; |
510 | struct kmem_cache *ceph_cap_cachep; | |
511 | struct kmem_cache *ceph_dentry_cachep; | |
512 | struct kmem_cache *ceph_file_cachep; | |
513 | ||
514 | static void ceph_inode_init_once(void *foo) | |
0743304d | 515 | { |
3d14c5d2 YS |
516 | struct ceph_inode_info *ci = foo; |
517 | inode_init_once(&ci->vfs_inode); | |
518 | } | |
519 | ||
520 | static int __init init_caches(void) | |
521 | { | |
522 | ceph_inode_cachep = kmem_cache_create("ceph_inode_info", | |
523 | sizeof(struct ceph_inode_info), | |
524 | __alignof__(struct ceph_inode_info), | |
525 | (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), | |
526 | ceph_inode_init_once); | |
527 | if (ceph_inode_cachep == NULL) | |
528 | return -ENOMEM; | |
529 | ||
530 | ceph_cap_cachep = KMEM_CACHE(ceph_cap, | |
531 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
532 | if (ceph_cap_cachep == NULL) | |
533 | goto bad_cap; | |
534 | ||
535 | ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info, | |
536 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
537 | if (ceph_dentry_cachep == NULL) | |
538 | goto bad_dentry; | |
539 | ||
540 | ceph_file_cachep = KMEM_CACHE(ceph_file_info, | |
541 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
542 | if (ceph_file_cachep == NULL) | |
543 | goto bad_file; | |
544 | ||
0743304d | 545 | return 0; |
3d14c5d2 YS |
546 | |
547 | bad_file: | |
548 | kmem_cache_destroy(ceph_dentry_cachep); | |
549 | bad_dentry: | |
550 | kmem_cache_destroy(ceph_cap_cachep); | |
551 | bad_cap: | |
552 | kmem_cache_destroy(ceph_inode_cachep); | |
553 | return -ENOMEM; | |
0743304d SW |
554 | } |
555 | ||
3d14c5d2 YS |
556 | static void destroy_caches(void) |
557 | { | |
558 | kmem_cache_destroy(ceph_inode_cachep); | |
559 | kmem_cache_destroy(ceph_cap_cachep); | |
560 | kmem_cache_destroy(ceph_dentry_cachep); | |
561 | kmem_cache_destroy(ceph_file_cachep); | |
562 | } | |
563 | ||
564 | ||
16725b9d | 565 | /* |
3d14c5d2 YS |
566 | * ceph_umount_begin - initiate forced umount. Tear down down the |
567 | * mount, skipping steps that may hang while waiting for server(s). | |
16725b9d | 568 | */ |
3d14c5d2 | 569 | static void ceph_umount_begin(struct super_block *sb) |
16725b9d | 570 | { |
3d14c5d2 YS |
571 | struct ceph_fs_client *fsc = ceph_sb_to_client(sb); |
572 | ||
573 | dout("ceph_umount_begin - starting forced umount\n"); | |
574 | if (!fsc) | |
575 | return; | |
576 | fsc->mount_state = CEPH_MOUNT_SHUTDOWN; | |
577 | return; | |
16725b9d SW |
578 | } |
579 | ||
3d14c5d2 YS |
580 | static const struct super_operations ceph_super_ops = { |
581 | .alloc_inode = ceph_alloc_inode, | |
582 | .destroy_inode = ceph_destroy_inode, | |
583 | .write_inode = ceph_write_inode, | |
584 | .sync_fs = ceph_sync_fs, | |
585 | .put_super = ceph_put_super, | |
586 | .show_options = ceph_show_options, | |
587 | .statfs = ceph_statfs, | |
588 | .umount_begin = ceph_umount_begin, | |
589 | }; | |
590 | ||
16725b9d SW |
591 | /* |
592 | * Bootstrap mount by opening the root directory. Note the mount | |
593 | * @started time from caller, and time out if this takes too long. | |
594 | */ | |
3d14c5d2 | 595 | static struct dentry *open_root_dentry(struct ceph_fs_client *fsc, |
16725b9d SW |
596 | const char *path, |
597 | unsigned long started) | |
598 | { | |
3d14c5d2 | 599 | struct ceph_mds_client *mdsc = fsc->mdsc; |
16725b9d SW |
600 | struct ceph_mds_request *req = NULL; |
601 | int err; | |
602 | struct dentry *root; | |
603 | ||
604 | /* open dir */ | |
605 | dout("open_root_inode opening '%s'\n", path); | |
606 | req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS); | |
607 | if (IS_ERR(req)) | |
7e34bc52 | 608 | return ERR_CAST(req); |
16725b9d SW |
609 | req->r_path1 = kstrdup(path, GFP_NOFS); |
610 | req->r_ino1.ino = CEPH_INO_ROOT; | |
611 | req->r_ino1.snap = CEPH_NOSNAP; | |
612 | req->r_started = started; | |
3d14c5d2 | 613 | req->r_timeout = fsc->client->options->mount_timeout * HZ; |
16725b9d SW |
614 | req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE); |
615 | req->r_num_caps = 2; | |
616 | err = ceph_mdsc_do_request(mdsc, NULL, req); | |
617 | if (err == 0) { | |
618 | dout("open_root_inode success\n"); | |
619 | if (ceph_ino(req->r_target_inode) == CEPH_INO_ROOT && | |
3d14c5d2 | 620 | fsc->sb->s_root == NULL) |
16725b9d SW |
621 | root = d_alloc_root(req->r_target_inode); |
622 | else | |
623 | root = d_obtain_alias(req->r_target_inode); | |
624 | req->r_target_inode = NULL; | |
625 | dout("open_root_inode success, root dentry is %p\n", root); | |
626 | } else { | |
627 | root = ERR_PTR(err); | |
628 | } | |
629 | ceph_mdsc_put_request(req); | |
630 | return root; | |
631 | } | |
632 | ||
3d14c5d2 YS |
633 | |
634 | ||
635 | ||
16725b9d SW |
636 | /* |
637 | * mount: join the ceph cluster, and open root directory. | |
638 | */ | |
a7f9fb20 | 639 | static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc, |
16725b9d SW |
640 | const char *path) |
641 | { | |
16725b9d | 642 | int err; |
16725b9d SW |
643 | unsigned long started = jiffies; /* note the start time */ |
644 | struct dentry *root; | |
3d14c5d2 | 645 | int first = 0; /* first vfsmount for this super_block */ |
16725b9d SW |
646 | |
647 | dout("mount start\n"); | |
3d14c5d2 | 648 | mutex_lock(&fsc->client->mount_mutex); |
16725b9d | 649 | |
3d14c5d2 | 650 | err = __ceph_open_session(fsc->client, started); |
16725b9d SW |
651 | if (err < 0) |
652 | goto out; | |
653 | ||
16725b9d | 654 | dout("mount opening root\n"); |
3d14c5d2 | 655 | root = open_root_dentry(fsc, "", started); |
16725b9d SW |
656 | if (IS_ERR(root)) { |
657 | err = PTR_ERR(root); | |
658 | goto out; | |
659 | } | |
3d14c5d2 | 660 | if (fsc->sb->s_root) { |
16725b9d | 661 | dput(root); |
3d14c5d2 YS |
662 | } else { |
663 | fsc->sb->s_root = root; | |
664 | first = 1; | |
665 | ||
666 | err = ceph_fs_debugfs_init(fsc); | |
667 | if (err < 0) | |
668 | goto fail; | |
669 | } | |
16725b9d SW |
670 | |
671 | if (path[0] == 0) { | |
672 | dget(root); | |
673 | } else { | |
674 | dout("mount opening base mountpoint\n"); | |
3d14c5d2 | 675 | root = open_root_dentry(fsc, path, started); |
16725b9d SW |
676 | if (IS_ERR(root)) { |
677 | err = PTR_ERR(root); | |
3d14c5d2 | 678 | goto fail; |
16725b9d SW |
679 | } |
680 | } | |
681 | ||
3d14c5d2 | 682 | fsc->mount_state = CEPH_MOUNT_MOUNTED; |
16725b9d | 683 | dout("mount success\n"); |
a7f9fb20 AV |
684 | mutex_unlock(&fsc->client->mount_mutex); |
685 | return root; | |
16725b9d SW |
686 | |
687 | out: | |
3d14c5d2 | 688 | mutex_unlock(&fsc->client->mount_mutex); |
a7f9fb20 | 689 | return ERR_PTR(err); |
3d14c5d2 YS |
690 | |
691 | fail: | |
692 | if (first) { | |
693 | dput(fsc->sb->s_root); | |
694 | fsc->sb->s_root = NULL; | |
695 | } | |
696 | goto out; | |
16725b9d SW |
697 | } |
698 | ||
699 | static int ceph_set_super(struct super_block *s, void *data) | |
700 | { | |
3d14c5d2 | 701 | struct ceph_fs_client *fsc = data; |
16725b9d SW |
702 | int ret; |
703 | ||
704 | dout("set_super %p data %p\n", s, data); | |
705 | ||
3d14c5d2 | 706 | s->s_flags = fsc->mount_options->sb_flags; |
16725b9d SW |
707 | s->s_maxbytes = 1ULL << 40; /* temp value until we get mdsmap */ |
708 | ||
3d14c5d2 YS |
709 | s->s_fs_info = fsc; |
710 | fsc->sb = s; | |
16725b9d SW |
711 | |
712 | s->s_op = &ceph_super_ops; | |
713 | s->s_export_op = &ceph_export_ops; | |
714 | ||
715 | s->s_time_gran = 1000; /* 1000 ns == 1 us */ | |
716 | ||
717 | ret = set_anon_super(s, NULL); /* what is that second arg for? */ | |
718 | if (ret != 0) | |
719 | goto fail; | |
720 | ||
721 | return ret; | |
722 | ||
723 | fail: | |
724 | s->s_fs_info = NULL; | |
3d14c5d2 | 725 | fsc->sb = NULL; |
16725b9d SW |
726 | return ret; |
727 | } | |
728 | ||
729 | /* | |
730 | * share superblock if same fs AND options | |
731 | */ | |
732 | static int ceph_compare_super(struct super_block *sb, void *data) | |
733 | { | |
3d14c5d2 YS |
734 | struct ceph_fs_client *new = data; |
735 | struct ceph_mount_options *fsopt = new->mount_options; | |
736 | struct ceph_options *opt = new->client->options; | |
737 | struct ceph_fs_client *other = ceph_sb_to_client(sb); | |
16725b9d SW |
738 | |
739 | dout("ceph_compare_super %p\n", sb); | |
3d14c5d2 YS |
740 | |
741 | if (compare_mount_options(fsopt, opt, other)) { | |
742 | dout("monitor(s)/mount options don't match\n"); | |
743 | return 0; | |
16725b9d | 744 | } |
3d14c5d2 YS |
745 | if ((opt->flags & CEPH_OPT_FSID) && |
746 | ceph_fsid_compare(&opt->fsid, &other->client->fsid)) { | |
747 | dout("fsid doesn't match\n"); | |
748 | return 0; | |
749 | } | |
750 | if (fsopt->sb_flags != other->mount_options->sb_flags) { | |
16725b9d SW |
751 | dout("flags differ\n"); |
752 | return 0; | |
753 | } | |
754 | return 1; | |
755 | } | |
756 | ||
757 | /* | |
758 | * construct our own bdi so we can control readahead, etc. | |
759 | */ | |
00d5643e | 760 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); |
31e0cf8f | 761 | |
3d14c5d2 YS |
762 | static int ceph_register_bdi(struct super_block *sb, |
763 | struct ceph_fs_client *fsc) | |
16725b9d SW |
764 | { |
765 | int err; | |
766 | ||
16725b9d | 767 | /* set ra_pages based on rsize mount option? */ |
3d14c5d2 YS |
768 | if (fsc->mount_options->rsize >= PAGE_CACHE_SIZE) |
769 | fsc->backing_dev_info.ra_pages = | |
770 | (fsc->mount_options->rsize + PAGE_CACHE_SIZE - 1) | |
16725b9d | 771 | >> PAGE_SHIFT; |
3d14c5d2 | 772 | err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%d", |
31e0cf8f | 773 | atomic_long_inc_return(&bdi_seq)); |
5dfc589a | 774 | if (!err) |
3d14c5d2 | 775 | sb->s_bdi = &fsc->backing_dev_info; |
16725b9d SW |
776 | return err; |
777 | } | |
778 | ||
a7f9fb20 AV |
779 | static struct dentry *ceph_mount(struct file_system_type *fs_type, |
780 | int flags, const char *dev_name, void *data) | |
16725b9d SW |
781 | { |
782 | struct super_block *sb; | |
3d14c5d2 | 783 | struct ceph_fs_client *fsc; |
a7f9fb20 | 784 | struct dentry *res; |
16725b9d SW |
785 | int err; |
786 | int (*compare_super)(struct super_block *, void *) = ceph_compare_super; | |
6a18be16 | 787 | const char *path = NULL; |
3d14c5d2 YS |
788 | struct ceph_mount_options *fsopt = NULL; |
789 | struct ceph_options *opt = NULL; | |
16725b9d | 790 | |
a7f9fb20 | 791 | dout("ceph_mount\n"); |
3d14c5d2 | 792 | err = parse_mount_options(&fsopt, &opt, flags, data, dev_name, &path); |
a7f9fb20 AV |
793 | if (err < 0) { |
794 | res = ERR_PTR(err); | |
6b805185 | 795 | goto out_final; |
a7f9fb20 | 796 | } |
16725b9d SW |
797 | |
798 | /* create client (which we may/may not use) */ | |
3d14c5d2 YS |
799 | fsc = create_fs_client(fsopt, opt); |
800 | if (IS_ERR(fsc)) { | |
a7f9fb20 | 801 | res = ERR_CAST(fsc); |
3d14c5d2 YS |
802 | kfree(fsopt); |
803 | kfree(opt); | |
6b805185 SW |
804 | goto out_final; |
805 | } | |
16725b9d | 806 | |
3d14c5d2 | 807 | err = ceph_mdsc_init(fsc); |
a7f9fb20 AV |
808 | if (err < 0) { |
809 | res = ERR_PTR(err); | |
3d14c5d2 | 810 | goto out; |
a7f9fb20 | 811 | } |
3d14c5d2 YS |
812 | |
813 | if (ceph_test_opt(fsc->client, NOSHARE)) | |
16725b9d | 814 | compare_super = NULL; |
3d14c5d2 | 815 | sb = sget(fs_type, compare_super, ceph_set_super, fsc); |
16725b9d | 816 | if (IS_ERR(sb)) { |
a7f9fb20 | 817 | res = ERR_CAST(sb); |
16725b9d SW |
818 | goto out; |
819 | } | |
820 | ||
3d14c5d2 YS |
821 | if (ceph_sb_to_client(sb) != fsc) { |
822 | ceph_mdsc_destroy(fsc); | |
823 | destroy_fs_client(fsc); | |
824 | fsc = ceph_sb_to_client(sb); | |
825 | dout("get_sb got existing client %p\n", fsc); | |
16725b9d | 826 | } else { |
3d14c5d2 YS |
827 | dout("get_sb using new client %p\n", fsc); |
828 | err = ceph_register_bdi(sb, fsc); | |
a7f9fb20 AV |
829 | if (err < 0) { |
830 | res = ERR_PTR(err); | |
16725b9d | 831 | goto out_splat; |
a7f9fb20 | 832 | } |
16725b9d SW |
833 | } |
834 | ||
a7f9fb20 AV |
835 | res = ceph_real_mount(fsc, path); |
836 | if (IS_ERR(res)) | |
16725b9d | 837 | goto out_splat; |
a7f9fb20 AV |
838 | dout("root %p inode %p ino %llx.%llx\n", res, |
839 | res->d_inode, ceph_vinop(res->d_inode)); | |
840 | return res; | |
16725b9d SW |
841 | |
842 | out_splat: | |
3d14c5d2 | 843 | ceph_mdsc_close_sessions(fsc->mdsc); |
3981f2e2 | 844 | deactivate_locked_super(sb); |
16725b9d SW |
845 | goto out_final; |
846 | ||
847 | out: | |
3d14c5d2 YS |
848 | ceph_mdsc_destroy(fsc); |
849 | destroy_fs_client(fsc); | |
16725b9d | 850 | out_final: |
a7f9fb20 AV |
851 | dout("ceph_mount fail %ld\n", PTR_ERR(res)); |
852 | return res; | |
16725b9d SW |
853 | } |
854 | ||
855 | static void ceph_kill_sb(struct super_block *s) | |
856 | { | |
3d14c5d2 | 857 | struct ceph_fs_client *fsc = ceph_sb_to_client(s); |
16725b9d | 858 | dout("kill_sb %p\n", s); |
3d14c5d2 | 859 | ceph_mdsc_pre_umount(fsc->mdsc); |
16725b9d | 860 | kill_anon_super(s); /* will call put_super after sb is r/o */ |
3d14c5d2 YS |
861 | ceph_mdsc_destroy(fsc); |
862 | destroy_fs_client(fsc); | |
16725b9d SW |
863 | } |
864 | ||
865 | static struct file_system_type ceph_fs_type = { | |
866 | .owner = THIS_MODULE, | |
867 | .name = "ceph", | |
a7f9fb20 | 868 | .mount = ceph_mount, |
16725b9d SW |
869 | .kill_sb = ceph_kill_sb, |
870 | .fs_flags = FS_RENAME_DOES_D_MOVE, | |
871 | }; | |
872 | ||
873 | #define _STRINGIFY(x) #x | |
874 | #define STRINGIFY(x) _STRINGIFY(x) | |
875 | ||
876 | static int __init init_ceph(void) | |
877 | { | |
3d14c5d2 | 878 | int ret = init_caches(); |
16725b9d | 879 | if (ret) |
3d14c5d2 | 880 | goto out; |
16725b9d | 881 | |
16725b9d SW |
882 | ret = register_filesystem(&ceph_fs_type); |
883 | if (ret) | |
884 | goto out_icache; | |
885 | ||
3d14c5d2 YS |
886 | pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL); |
887 | ||
16725b9d SW |
888 | return 0; |
889 | ||
890 | out_icache: | |
891 | destroy_caches(); | |
16725b9d SW |
892 | out: |
893 | return ret; | |
894 | } | |
895 | ||
896 | static void __exit exit_ceph(void) | |
897 | { | |
898 | dout("exit_ceph\n"); | |
899 | unregister_filesystem(&ceph_fs_type); | |
16725b9d | 900 | destroy_caches(); |
16725b9d SW |
901 | } |
902 | ||
903 | module_init(init_ceph); | |
904 | module_exit(exit_ceph); | |
905 | ||
906 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
907 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
908 | MODULE_AUTHOR("Patience Warnick <patience@newdream.net>"); | |
909 | MODULE_DESCRIPTION("Ceph filesystem for Linux"); | |
910 | MODULE_LICENSE("GPL"); |