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 | 19 | #include "mds_client.h" |
99ccbd22 | 20 | #include "cache.h" |
3d14c5d2 | 21 | |
1fe60e51 | 22 | #include <linux/ceph/ceph_features.h> |
3d14c5d2 YS |
23 | #include <linux/ceph/decode.h> |
24 | #include <linux/ceph/mon_client.h> | |
25 | #include <linux/ceph/auth.h> | |
26 | #include <linux/ceph/debugfs.h> | |
16725b9d SW |
27 | |
28 | /* | |
29 | * Ceph superblock operations | |
30 | * | |
31 | * Handle the basics of mounting, unmounting. | |
32 | */ | |
33 | ||
16725b9d SW |
34 | /* |
35 | * super ops | |
36 | */ | |
37 | static void ceph_put_super(struct super_block *s) | |
38 | { | |
3d14c5d2 | 39 | struct ceph_fs_client *fsc = ceph_sb_to_client(s); |
16725b9d SW |
40 | |
41 | dout("put_super\n"); | |
3d14c5d2 | 42 | ceph_mdsc_close_sessions(fsc->mdsc); |
16725b9d SW |
43 | } |
44 | ||
45 | static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf) | |
46 | { | |
2b0143b5 | 47 | struct ceph_fs_client *fsc = ceph_inode_to_client(d_inode(dentry)); |
73fb0949 | 48 | struct ceph_mon_client *monc = &fsc->client->monc; |
16725b9d SW |
49 | struct ceph_statfs st; |
50 | u64 fsid; | |
51 | int err; | |
06d74376 DF |
52 | u64 data_pool; |
53 | ||
54 | if (fsc->mdsc->mdsmap->m_num_data_pg_pools == 1) { | |
55 | data_pool = fsc->mdsc->mdsmap->m_data_pg_pools[0]; | |
56 | } else { | |
57 | data_pool = CEPH_NOPOOL; | |
58 | } | |
16725b9d SW |
59 | |
60 | dout("statfs\n"); | |
73fb0949 | 61 | err = ceph_monc_do_statfs(monc, data_pool, &st); |
16725b9d SW |
62 | if (err < 0) |
63 | return err; | |
64 | ||
65 | /* fill in kstatfs */ | |
66 | buf->f_type = CEPH_SUPER_MAGIC; /* ?? */ | |
67 | ||
68 | /* | |
69 | * express utilization in terms of large blocks to avoid | |
70 | * overflow on 32-bit machines. | |
92a49fb0 SW |
71 | * |
72 | * NOTE: for the time being, we make bsize == frsize to humor | |
73 | * not-yet-ancient versions of glibc that are broken. | |
74 | * Someday, we will probably want to report a real block | |
75 | * size... whatever that may mean for a network file system! | |
16725b9d SW |
76 | */ |
77 | buf->f_bsize = 1 << CEPH_BLOCK_SHIFT; | |
92a49fb0 | 78 | buf->f_frsize = 1 << CEPH_BLOCK_SHIFT; |
9122eed5 LH |
79 | |
80 | /* | |
81 | * By default use root quota for stats; fallback to overall filesystem | |
82 | * usage if using 'noquotadf' mount option or if the root dir doesn't | |
83 | * have max_bytes quota set. | |
84 | */ | |
85 | if (ceph_test_mount_opt(fsc, NOQUOTADF) || | |
86 | !ceph_quota_update_statfs(fsc, buf)) { | |
87 | buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10); | |
88 | buf->f_bfree = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10); | |
89 | buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10); | |
90 | } | |
16725b9d SW |
91 | |
92 | buf->f_files = le64_to_cpu(st.num_objects); | |
93 | buf->f_ffree = -1; | |
558d3499 | 94 | buf->f_namelen = NAME_MAX; |
16725b9d | 95 | |
080a330e | 96 | /* Must convert the fsid, for consistent values across arches */ |
73fb0949 LH |
97 | mutex_lock(&monc->mutex); |
98 | fsid = le64_to_cpu(*(__le64 *)(&monc->monmap->fsid)) ^ | |
99 | le64_to_cpu(*((__le64 *)&monc->monmap->fsid + 1)); | |
100 | mutex_unlock(&monc->mutex); | |
101 | ||
16725b9d SW |
102 | buf->f_fsid.val[0] = fsid & 0xffffffff; |
103 | buf->f_fsid.val[1] = fsid >> 32; | |
104 | ||
105 | return 0; | |
106 | } | |
107 | ||
108 | ||
2d9c98ae | 109 | static int ceph_sync_fs(struct super_block *sb, int wait) |
16725b9d | 110 | { |
3d14c5d2 | 111 | struct ceph_fs_client *fsc = ceph_sb_to_client(sb); |
2d9c98ae SW |
112 | |
113 | if (!wait) { | |
114 | dout("sync_fs (non-blocking)\n"); | |
3d14c5d2 | 115 | ceph_flush_dirty_caps(fsc->mdsc); |
2d9c98ae SW |
116 | dout("sync_fs (non-blocking) done\n"); |
117 | return 0; | |
118 | } | |
119 | ||
120 | dout("sync_fs (blocking)\n"); | |
3d14c5d2 YS |
121 | ceph_osdc_sync(&fsc->client->osdc); |
122 | ceph_mdsc_sync(fsc->mdsc); | |
2d9c98ae | 123 | dout("sync_fs (blocking) done\n"); |
16725b9d SW |
124 | return 0; |
125 | } | |
126 | ||
16725b9d SW |
127 | /* |
128 | * mount options | |
129 | */ | |
130 | enum { | |
16725b9d SW |
131 | Opt_wsize, |
132 | Opt_rsize, | |
83817e35 | 133 | Opt_rasize, |
16725b9d SW |
134 | Opt_caps_wanted_delay_min, |
135 | Opt_caps_wanted_delay_max, | |
136 | Opt_readdir_max_entries, | |
23804d91 | 137 | Opt_readdir_max_bytes, |
2baba250 | 138 | Opt_congestion_kb, |
e53c2fe0 | 139 | Opt_last_int, |
16725b9d SW |
140 | /* int args above */ |
141 | Opt_snapdirname, | |
430afbad | 142 | Opt_mds_namespace, |
1d8f8360 | 143 | Opt_fscache_uniq, |
e53c2fe0 | 144 | Opt_last_string, |
16725b9d | 145 | /* string args above */ |
16725b9d SW |
146 | Opt_dirstat, |
147 | Opt_nodirstat, | |
148 | Opt_rbytes, | |
149 | Opt_norbytes, | |
cffaba15 | 150 | Opt_asyncreaddir, |
16725b9d | 151 | Opt_noasyncreaddir, |
a40dc6cc SW |
152 | Opt_dcache, |
153 | Opt_nodcache, | |
ad1fee96 | 154 | Opt_ino32, |
cffaba15 | 155 | Opt_noino32, |
99ccbd22 | 156 | Opt_fscache, |
45195e42 | 157 | Opt_nofscache, |
10183a69 YZ |
158 | Opt_poolperm, |
159 | Opt_nopoolperm, | |
e9e427f0 YZ |
160 | Opt_require_active_mds, |
161 | Opt_norequire_active_mds, | |
45195e42 SW |
162 | #ifdef CONFIG_CEPH_FS_POSIX_ACL |
163 | Opt_acl, | |
164 | #endif | |
10183a69 | 165 | Opt_noacl, |
9122eed5 LH |
166 | Opt_quotadf, |
167 | Opt_noquotadf, | |
16725b9d SW |
168 | }; |
169 | ||
3d14c5d2 | 170 | static match_table_t fsopt_tokens = { |
16725b9d SW |
171 | {Opt_wsize, "wsize=%d"}, |
172 | {Opt_rsize, "rsize=%d"}, | |
83817e35 | 173 | {Opt_rasize, "rasize=%d"}, |
16725b9d SW |
174 | {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"}, |
175 | {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"}, | |
176 | {Opt_readdir_max_entries, "readdir_max_entries=%d"}, | |
23804d91 | 177 | {Opt_readdir_max_bytes, "readdir_max_bytes=%d"}, |
2baba250 | 178 | {Opt_congestion_kb, "write_congestion_kb=%d"}, |
16725b9d SW |
179 | /* int args above */ |
180 | {Opt_snapdirname, "snapdirname=%s"}, | |
430afbad | 181 | {Opt_mds_namespace, "mds_namespace=%s"}, |
1d8f8360 | 182 | {Opt_fscache_uniq, "fsc=%s"}, |
16725b9d | 183 | /* string args above */ |
16725b9d SW |
184 | {Opt_dirstat, "dirstat"}, |
185 | {Opt_nodirstat, "nodirstat"}, | |
186 | {Opt_rbytes, "rbytes"}, | |
187 | {Opt_norbytes, "norbytes"}, | |
cffaba15 | 188 | {Opt_asyncreaddir, "asyncreaddir"}, |
16725b9d | 189 | {Opt_noasyncreaddir, "noasyncreaddir"}, |
a40dc6cc SW |
190 | {Opt_dcache, "dcache"}, |
191 | {Opt_nodcache, "nodcache"}, | |
ad1fee96 | 192 | {Opt_ino32, "ino32"}, |
cffaba15 | 193 | {Opt_noino32, "noino32"}, |
99ccbd22 MT |
194 | {Opt_fscache, "fsc"}, |
195 | {Opt_nofscache, "nofsc"}, | |
10183a69 YZ |
196 | {Opt_poolperm, "poolperm"}, |
197 | {Opt_nopoolperm, "nopoolperm"}, | |
e9e427f0 YZ |
198 | {Opt_require_active_mds, "require_active_mds"}, |
199 | {Opt_norequire_active_mds, "norequire_active_mds"}, | |
45195e42 SW |
200 | #ifdef CONFIG_CEPH_FS_POSIX_ACL |
201 | {Opt_acl, "acl"}, | |
202 | #endif | |
203 | {Opt_noacl, "noacl"}, | |
9122eed5 LH |
204 | {Opt_quotadf, "quotadf"}, |
205 | {Opt_noquotadf, "noquotadf"}, | |
16725b9d SW |
206 | {-1, NULL} |
207 | }; | |
208 | ||
3d14c5d2 | 209 | static int parse_fsopt_token(char *c, void *private) |
c309f0ab | 210 | { |
3d14c5d2 YS |
211 | struct ceph_mount_options *fsopt = private; |
212 | substring_t argstr[MAX_OPT_ARGS]; | |
213 | int token, intval, ret; | |
214 | ||
215 | token = match_token((char *)c, fsopt_tokens, argstr); | |
216 | if (token < 0) | |
217 | return -EINVAL; | |
218 | ||
219 | if (token < Opt_last_int) { | |
220 | ret = match_int(&argstr[0], &intval); | |
221 | if (ret < 0) { | |
222 | pr_err("bad mount option arg (not int) " | |
223 | "at '%s'\n", c); | |
224 | return ret; | |
c309f0ab | 225 | } |
3d14c5d2 YS |
226 | dout("got int token %d val %d\n", token, intval); |
227 | } else if (token > Opt_last_int && token < Opt_last_string) { | |
228 | dout("got string token %d val %s\n", token, | |
229 | argstr[0].from); | |
230 | } else { | |
231 | dout("got token %d\n", token); | |
c309f0ab SW |
232 | } |
233 | ||
3d14c5d2 YS |
234 | switch (token) { |
235 | case Opt_snapdirname: | |
236 | kfree(fsopt->snapdir_name); | |
237 | fsopt->snapdir_name = kstrndup(argstr[0].from, | |
238 | argstr[0].to-argstr[0].from, | |
239 | GFP_KERNEL); | |
240 | if (!fsopt->snapdir_name) | |
241 | return -ENOMEM; | |
242 | break; | |
235a0982 | 243 | case Opt_mds_namespace: |
937441f3 | 244 | kfree(fsopt->mds_namespace); |
430afbad YZ |
245 | fsopt->mds_namespace = kstrndup(argstr[0].from, |
246 | argstr[0].to-argstr[0].from, | |
247 | GFP_KERNEL); | |
248 | if (!fsopt->mds_namespace) | |
249 | return -ENOMEM; | |
235a0982 | 250 | break; |
1d8f8360 | 251 | case Opt_fscache_uniq: |
937441f3 | 252 | kfree(fsopt->fscache_uniq); |
1d8f8360 YZ |
253 | fsopt->fscache_uniq = kstrndup(argstr[0].from, |
254 | argstr[0].to-argstr[0].from, | |
255 | GFP_KERNEL); | |
256 | if (!fsopt->fscache_uniq) | |
257 | return -ENOMEM; | |
258 | fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE; | |
259 | break; | |
430afbad | 260 | /* misc */ |
3d14c5d2 | 261 | case Opt_wsize: |
8db0c759 | 262 | if (intval < (int)PAGE_SIZE || intval > CEPH_MAX_WRITE_SIZE) |
95cca2b4 YZ |
263 | return -EINVAL; |
264 | fsopt->wsize = ALIGN(intval, PAGE_SIZE); | |
3d14c5d2 YS |
265 | break; |
266 | case Opt_rsize: | |
8db0c759 | 267 | if (intval < (int)PAGE_SIZE || intval > CEPH_MAX_READ_SIZE) |
aa187926 YZ |
268 | return -EINVAL; |
269 | fsopt->rsize = ALIGN(intval, PAGE_SIZE); | |
3d14c5d2 | 270 | break; |
83817e35 | 271 | case Opt_rasize: |
4214fb15 YZ |
272 | if (intval < 0) |
273 | return -EINVAL; | |
c36ed50d | 274 | fsopt->rasize = ALIGN(intval, PAGE_SIZE); |
83817e35 | 275 | break; |
3d14c5d2 | 276 | case Opt_caps_wanted_delay_min: |
4214fb15 YZ |
277 | if (intval < 1) |
278 | return -EINVAL; | |
3d14c5d2 YS |
279 | fsopt->caps_wanted_delay_min = intval; |
280 | break; | |
281 | case Opt_caps_wanted_delay_max: | |
4214fb15 YZ |
282 | if (intval < 1) |
283 | return -EINVAL; | |
3d14c5d2 YS |
284 | fsopt->caps_wanted_delay_max = intval; |
285 | break; | |
286 | case Opt_readdir_max_entries: | |
4214fb15 YZ |
287 | if (intval < 1) |
288 | return -EINVAL; | |
3d14c5d2 YS |
289 | fsopt->max_readdir = intval; |
290 | break; | |
291 | case Opt_readdir_max_bytes: | |
8db0c759 | 292 | if (intval < (int)PAGE_SIZE && intval != 0) |
4214fb15 | 293 | return -EINVAL; |
3d14c5d2 YS |
294 | fsopt->max_readdir_bytes = intval; |
295 | break; | |
296 | case Opt_congestion_kb: | |
4214fb15 YZ |
297 | if (intval < 1024) /* at least 1M */ |
298 | return -EINVAL; | |
3d14c5d2 YS |
299 | fsopt->congestion_kb = intval; |
300 | break; | |
301 | case Opt_dirstat: | |
302 | fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT; | |
303 | break; | |
304 | case Opt_nodirstat: | |
305 | fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT; | |
306 | break; | |
307 | case Opt_rbytes: | |
308 | fsopt->flags |= CEPH_MOUNT_OPT_RBYTES; | |
309 | break; | |
310 | case Opt_norbytes: | |
311 | fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES; | |
312 | break; | |
cffaba15 AE |
313 | case Opt_asyncreaddir: |
314 | fsopt->flags &= ~CEPH_MOUNT_OPT_NOASYNCREADDIR; | |
315 | break; | |
3d14c5d2 YS |
316 | case Opt_noasyncreaddir: |
317 | fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR; | |
318 | break; | |
a40dc6cc SW |
319 | case Opt_dcache: |
320 | fsopt->flags |= CEPH_MOUNT_OPT_DCACHE; | |
321 | break; | |
322 | case Opt_nodcache: | |
323 | fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE; | |
324 | break; | |
ad1fee96 YS |
325 | case Opt_ino32: |
326 | fsopt->flags |= CEPH_MOUNT_OPT_INO32; | |
327 | break; | |
cffaba15 AE |
328 | case Opt_noino32: |
329 | fsopt->flags &= ~CEPH_MOUNT_OPT_INO32; | |
330 | break; | |
99ccbd22 MT |
331 | case Opt_fscache: |
332 | fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE; | |
7ae7a828 CX |
333 | kfree(fsopt->fscache_uniq); |
334 | fsopt->fscache_uniq = NULL; | |
99ccbd22 MT |
335 | break; |
336 | case Opt_nofscache: | |
337 | fsopt->flags &= ~CEPH_MOUNT_OPT_FSCACHE; | |
7ae7a828 CX |
338 | kfree(fsopt->fscache_uniq); |
339 | fsopt->fscache_uniq = NULL; | |
99ccbd22 | 340 | break; |
10183a69 YZ |
341 | case Opt_poolperm: |
342 | fsopt->flags &= ~CEPH_MOUNT_OPT_NOPOOLPERM; | |
10183a69 YZ |
343 | break; |
344 | case Opt_nopoolperm: | |
345 | fsopt->flags |= CEPH_MOUNT_OPT_NOPOOLPERM; | |
346 | break; | |
e9e427f0 YZ |
347 | case Opt_require_active_mds: |
348 | fsopt->flags &= ~CEPH_MOUNT_OPT_MOUNTWAIT; | |
349 | break; | |
350 | case Opt_norequire_active_mds: | |
351 | fsopt->flags |= CEPH_MOUNT_OPT_MOUNTWAIT; | |
352 | break; | |
9122eed5 LH |
353 | case Opt_quotadf: |
354 | fsopt->flags &= ~CEPH_MOUNT_OPT_NOQUOTADF; | |
355 | break; | |
356 | case Opt_noquotadf: | |
357 | fsopt->flags |= CEPH_MOUNT_OPT_NOQUOTADF; | |
358 | break; | |
45195e42 SW |
359 | #ifdef CONFIG_CEPH_FS_POSIX_ACL |
360 | case Opt_acl: | |
1751e8a6 | 361 | fsopt->sb_flags |= SB_POSIXACL; |
45195e42 SW |
362 | break; |
363 | #endif | |
364 | case Opt_noacl: | |
1751e8a6 | 365 | fsopt->sb_flags &= ~SB_POSIXACL; |
45195e42 | 366 | break; |
3d14c5d2 YS |
367 | default: |
368 | BUG_ON(token); | |
369 | } | |
370 | return 0; | |
c309f0ab | 371 | } |
16725b9d | 372 | |
3d14c5d2 | 373 | static void destroy_mount_options(struct ceph_mount_options *args) |
16725b9d | 374 | { |
3d14c5d2 YS |
375 | dout("destroy_mount_options %p\n", args); |
376 | kfree(args->snapdir_name); | |
430afbad | 377 | kfree(args->mds_namespace); |
3f384954 | 378 | kfree(args->server_path); |
1d8f8360 | 379 | kfree(args->fscache_uniq); |
3d14c5d2 YS |
380 | kfree(args); |
381 | } | |
16725b9d | 382 | |
3d14c5d2 YS |
383 | static int strcmp_null(const char *s1, const char *s2) |
384 | { | |
385 | if (!s1 && !s2) | |
386 | return 0; | |
387 | if (s1 && !s2) | |
388 | return -1; | |
389 | if (!s1 && s2) | |
390 | return 1; | |
391 | return strcmp(s1, s2); | |
392 | } | |
16725b9d | 393 | |
3d14c5d2 YS |
394 | static int compare_mount_options(struct ceph_mount_options *new_fsopt, |
395 | struct ceph_options *new_opt, | |
396 | struct ceph_fs_client *fsc) | |
397 | { | |
398 | struct ceph_mount_options *fsopt1 = new_fsopt; | |
399 | struct ceph_mount_options *fsopt2 = fsc->mount_options; | |
400 | int ofs = offsetof(struct ceph_mount_options, snapdir_name); | |
401 | int ret; | |
16725b9d | 402 | |
3d14c5d2 YS |
403 | ret = memcmp(fsopt1, fsopt2, ofs); |
404 | if (ret) | |
405 | return ret; | |
406 | ||
407 | ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name); | |
430afbad YZ |
408 | if (ret) |
409 | return ret; | |
410 | ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace); | |
3d14c5d2 YS |
411 | if (ret) |
412 | return ret; | |
3f384954 | 413 | ret = strcmp_null(fsopt1->server_path, fsopt2->server_path); |
1d8f8360 YZ |
414 | if (ret) |
415 | return ret; | |
416 | ret = strcmp_null(fsopt1->fscache_uniq, fsopt2->fscache_uniq); | |
3f384954 YZ |
417 | if (ret) |
418 | return ret; | |
419 | ||
3d14c5d2 YS |
420 | return ceph_compare_options(new_opt, fsc->client); |
421 | } | |
422 | ||
423 | static int parse_mount_options(struct ceph_mount_options **pfsopt, | |
424 | struct ceph_options **popt, | |
425 | int flags, char *options, | |
3f384954 | 426 | const char *dev_name) |
3d14c5d2 YS |
427 | { |
428 | struct ceph_mount_options *fsopt; | |
429 | const char *dev_name_end; | |
c98f533c AE |
430 | int err; |
431 | ||
432 | if (!dev_name || !*dev_name) | |
433 | return -EINVAL; | |
3d14c5d2 YS |
434 | |
435 | fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL); | |
436 | if (!fsopt) | |
437 | return -ENOMEM; | |
438 | ||
439 | dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name); | |
440 | ||
80db8bea NW |
441 | fsopt->sb_flags = flags; |
442 | fsopt->flags = CEPH_MOUNT_OPT_DEFAULT; | |
3d14c5d2 | 443 | |
95cca2b4 | 444 | fsopt->wsize = CEPH_MAX_WRITE_SIZE; |
aa187926 | 445 | fsopt->rsize = CEPH_MAX_READ_SIZE; |
80db8bea NW |
446 | fsopt->rasize = CEPH_RASIZE_DEFAULT; |
447 | fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL); | |
a149bb9a SK |
448 | if (!fsopt->snapdir_name) { |
449 | err = -ENOMEM; | |
450 | goto out; | |
451 | } | |
452 | ||
50aac4fe SW |
453 | fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT; |
454 | fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT; | |
80db8bea NW |
455 | fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT; |
456 | fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT; | |
457 | fsopt->congestion_kb = default_congestion_kb(); | |
458 | ||
c98f533c AE |
459 | /* |
460 | * Distinguish the server list from the path in "dev_name". | |
461 | * Internally we do not include the leading '/' in the path. | |
462 | * | |
463 | * "dev_name" will look like: | |
464 | * <server_spec>[,<server_spec>...]:[<path>] | |
465 | * where | |
466 | * <server_spec> is <ip>[:<port>] | |
467 | * <path> is optional, but if present must begin with '/' | |
468 | */ | |
469 | dev_name_end = strchr(dev_name, '/'); | |
470 | if (dev_name_end) { | |
ce2728aa YZ |
471 | if (strlen(dev_name_end) > 1) { |
472 | fsopt->server_path = kstrdup(dev_name_end, GFP_KERNEL); | |
473 | if (!fsopt->server_path) { | |
474 | err = -ENOMEM; | |
475 | goto out; | |
476 | } | |
3f384954 | 477 | } |
c98f533c | 478 | } else { |
c98f533c | 479 | dev_name_end = dev_name + strlen(dev_name); |
c98f533c | 480 | } |
80db8bea | 481 | err = -EINVAL; |
c98f533c | 482 | dev_name_end--; /* back up to ':' separator */ |
54464296 | 483 | if (dev_name_end < dev_name || *dev_name_end != ':') { |
c98f533c | 484 | pr_err("device name is missing path (no : separator in %s)\n", |
80db8bea NW |
485 | dev_name); |
486 | goto out; | |
487 | } | |
3d14c5d2 | 488 | dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name); |
3f384954 YZ |
489 | if (fsopt->server_path) |
490 | dout("server path '%s'\n", fsopt->server_path); | |
16725b9d | 491 | |
ee57741c | 492 | *popt = ceph_parse_options(options, dev_name, dev_name_end, |
3d14c5d2 | 493 | parse_fsopt_token, (void *)fsopt); |
ee57741c AE |
494 | if (IS_ERR(*popt)) { |
495 | err = PTR_ERR(*popt); | |
3d14c5d2 | 496 | goto out; |
ee57741c | 497 | } |
3d14c5d2 YS |
498 | |
499 | /* success */ | |
500 | *pfsopt = fsopt; | |
501 | return 0; | |
16725b9d | 502 | |
7b813c46 | 503 | out: |
3d14c5d2 YS |
504 | destroy_mount_options(fsopt); |
505 | return err; | |
16725b9d SW |
506 | } |
507 | ||
3d14c5d2 YS |
508 | /** |
509 | * ceph_show_options - Show mount options in /proc/mounts | |
510 | * @m: seq_file to write to | |
34c80b1d | 511 | * @root: root of that (sub)tree |
3d14c5d2 | 512 | */ |
34c80b1d | 513 | static int ceph_show_options(struct seq_file *m, struct dentry *root) |
16725b9d | 514 | { |
34c80b1d | 515 | struct ceph_fs_client *fsc = ceph_sb_to_client(root->d_sb); |
3d14c5d2 | 516 | struct ceph_mount_options *fsopt = fsc->mount_options; |
ff40f9ae ID |
517 | size_t pos; |
518 | int ret; | |
519 | ||
520 | /* a comma between MNT/MS and client options */ | |
521 | seq_putc(m, ','); | |
522 | pos = m->count; | |
523 | ||
524 | ret = ceph_print_client_options(m, fsc->client); | |
525 | if (ret) | |
526 | return ret; | |
527 | ||
528 | /* retract our comma if no client options */ | |
529 | if (m->count == pos) | |
530 | m->count--; | |
3d14c5d2 YS |
531 | |
532 | if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT) | |
533 | seq_puts(m, ",dirstat"); | |
133e9156 YZ |
534 | if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES)) |
535 | seq_puts(m, ",rbytes"); | |
3d14c5d2 YS |
536 | if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR) |
537 | seq_puts(m, ",noasyncreaddir"); | |
ff7eeb82 | 538 | if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0) |
a40dc6cc | 539 | seq_puts(m, ",nodcache"); |
3619aa8b CX |
540 | if (fsopt->flags & CEPH_MOUNT_OPT_INO32) |
541 | seq_puts(m, ",ino32"); | |
1d8f8360 | 542 | if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) { |
4d8969af | 543 | seq_show_option(m, "fsc", fsopt->fscache_uniq); |
1d8f8360 | 544 | } |
10183a69 YZ |
545 | if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM) |
546 | seq_puts(m, ",nopoolperm"); | |
9122eed5 LH |
547 | if (fsopt->flags & CEPH_MOUNT_OPT_NOQUOTADF) |
548 | seq_puts(m, ",noquotadf"); | |
3d14c5d2 | 549 | |
45195e42 | 550 | #ifdef CONFIG_CEPH_FS_POSIX_ACL |
1751e8a6 | 551 | if (fsopt->sb_flags & SB_POSIXACL) |
45195e42 SW |
552 | seq_puts(m, ",acl"); |
553 | else | |
554 | seq_puts(m, ",noacl"); | |
555 | #endif | |
556 | ||
430afbad | 557 | if (fsopt->mds_namespace) |
4d8969af | 558 | seq_show_option(m, "mds_namespace", fsopt->mds_namespace); |
6dd4940b | 559 | if (fsopt->wsize != CEPH_MAX_WRITE_SIZE) |
3d14c5d2 | 560 | seq_printf(m, ",wsize=%d", fsopt->wsize); |
aa187926 | 561 | if (fsopt->rsize != CEPH_MAX_READ_SIZE) |
3d14c5d2 | 562 | seq_printf(m, ",rsize=%d", fsopt->rsize); |
83817e35 | 563 | if (fsopt->rasize != CEPH_RASIZE_DEFAULT) |
2151937d | 564 | seq_printf(m, ",rasize=%d", fsopt->rasize); |
3d14c5d2 YS |
565 | if (fsopt->congestion_kb != default_congestion_kb()) |
566 | seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb); | |
567 | if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT) | |
568 | seq_printf(m, ",caps_wanted_delay_min=%d", | |
569 | fsopt->caps_wanted_delay_min); | |
570 | if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT) | |
571 | seq_printf(m, ",caps_wanted_delay_max=%d", | |
572 | fsopt->caps_wanted_delay_max); | |
3d14c5d2 YS |
573 | if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT) |
574 | seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir); | |
575 | if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT) | |
576 | seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes); | |
577 | if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT)) | |
a068acf2 | 578 | seq_show_option(m, "snapdirname", fsopt->snapdir_name); |
ff40f9ae | 579 | |
3d14c5d2 | 580 | return 0; |
16725b9d SW |
581 | } |
582 | ||
583 | /* | |
3d14c5d2 YS |
584 | * handle any mon messages the standard library doesn't understand. |
585 | * return error if we don't either. | |
16725b9d | 586 | */ |
3d14c5d2 | 587 | static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg) |
16725b9d | 588 | { |
3d14c5d2 YS |
589 | struct ceph_fs_client *fsc = client->private; |
590 | int type = le16_to_cpu(msg->hdr.type); | |
591 | ||
592 | switch (type) { | |
593 | case CEPH_MSG_MDS_MAP: | |
430afbad YZ |
594 | ceph_mdsc_handle_mdsmap(fsc->mdsc, msg); |
595 | return 0; | |
596 | case CEPH_MSG_FS_MAP_USER: | |
597 | ceph_mdsc_handle_fsmap(fsc->mdsc, msg); | |
3d14c5d2 | 598 | return 0; |
3d14c5d2 YS |
599 | default: |
600 | return -1; | |
601 | } | |
602 | } | |
603 | ||
604 | /* | |
605 | * create a new fs client | |
606 | */ | |
0c6d4b4e | 607 | static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt, |
3d14c5d2 YS |
608 | struct ceph_options *opt) |
609 | { | |
610 | struct ceph_fs_client *fsc; | |
3bf53337 AE |
611 | int page_count; |
612 | size_t size; | |
16725b9d SW |
613 | int err = -ENOMEM; |
614 | ||
3d14c5d2 YS |
615 | fsc = kzalloc(sizeof(*fsc), GFP_KERNEL); |
616 | if (!fsc) | |
16725b9d SW |
617 | return ERR_PTR(-ENOMEM); |
618 | ||
74da4a0f | 619 | fsc->client = ceph_create_client(opt, fsc); |
3d14c5d2 YS |
620 | if (IS_ERR(fsc->client)) { |
621 | err = PTR_ERR(fsc->client); | |
622 | goto fail; | |
623 | } | |
c843d13c | 624 | |
3d14c5d2 | 625 | fsc->client->extra_mon_dispatch = extra_mon_dispatch; |
c843d13c | 626 | fsc->client->osdc.abort_on_full = true; |
430afbad | 627 | |
d37b1d99 | 628 | if (!fsopt->mds_namespace) { |
430afbad YZ |
629 | ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP, |
630 | 0, true); | |
631 | } else { | |
632 | ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_FSMAP, | |
633 | 0, false); | |
634 | } | |
16725b9d | 635 | |
3d14c5d2 | 636 | fsc->mount_options = fsopt; |
16725b9d | 637 | |
3d14c5d2 YS |
638 | fsc->sb = NULL; |
639 | fsc->mount_state = CEPH_MOUNT_MOUNTING; | |
16725b9d | 640 | |
3d14c5d2 | 641 | atomic_long_set(&fsc->writeback_count, 0); |
16725b9d SW |
642 | |
643 | err = -ENOMEM; | |
01e6acc4 TH |
644 | /* |
645 | * The number of concurrent works can be high but they don't need | |
646 | * to be processed in parallel, limit concurrency. | |
647 | */ | |
648 | fsc->wb_wq = alloc_workqueue("ceph-writeback", 0, 1); | |
d37b1d99 | 649 | if (!fsc->wb_wq) |
09dc9fc2 | 650 | goto fail_client; |
01e6acc4 | 651 | fsc->pg_inv_wq = alloc_workqueue("ceph-pg-invalid", 0, 1); |
d37b1d99 | 652 | if (!fsc->pg_inv_wq) |
16725b9d | 653 | goto fail_wb_wq; |
01e6acc4 | 654 | fsc->trunc_wq = alloc_workqueue("ceph-trunc", 0, 1); |
d37b1d99 | 655 | if (!fsc->trunc_wq) |
16725b9d SW |
656 | goto fail_pg_inv_wq; |
657 | ||
b9bfb93c SW |
658 | /* set up mempools */ |
659 | err = -ENOMEM; | |
09cbfeaf | 660 | page_count = fsc->mount_options->wsize >> PAGE_SHIFT; |
3bf53337 AE |
661 | size = sizeof (struct page *) * (page_count ? page_count : 1); |
662 | fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10, size); | |
3d14c5d2 | 663 | if (!fsc->wb_pagevec_pool) |
b9bfb93c SW |
664 | goto fail_trunc_wq; |
665 | ||
85ccce43 | 666 | /* caps */ |
3d14c5d2 YS |
667 | fsc->min_caps = fsopt->max_readdir; |
668 | ||
669 | return fsc; | |
b9bfb93c | 670 | |
16725b9d | 671 | fail_trunc_wq: |
3d14c5d2 | 672 | destroy_workqueue(fsc->trunc_wq); |
16725b9d | 673 | fail_pg_inv_wq: |
3d14c5d2 | 674 | destroy_workqueue(fsc->pg_inv_wq); |
16725b9d | 675 | fail_wb_wq: |
3d14c5d2 | 676 | destroy_workqueue(fsc->wb_wq); |
3d14c5d2 YS |
677 | fail_client: |
678 | ceph_destroy_client(fsc->client); | |
16725b9d | 679 | fail: |
3d14c5d2 | 680 | kfree(fsc); |
16725b9d SW |
681 | return ERR_PTR(err); |
682 | } | |
683 | ||
a57d9064 YZ |
684 | static void flush_fs_workqueues(struct ceph_fs_client *fsc) |
685 | { | |
686 | flush_workqueue(fsc->wb_wq); | |
687 | flush_workqueue(fsc->pg_inv_wq); | |
688 | flush_workqueue(fsc->trunc_wq); | |
689 | } | |
690 | ||
0c6d4b4e | 691 | static void destroy_fs_client(struct ceph_fs_client *fsc) |
16725b9d | 692 | { |
3d14c5d2 | 693 | dout("destroy_fs_client %p\n", fsc); |
16725b9d | 694 | |
3d14c5d2 YS |
695 | destroy_workqueue(fsc->wb_wq); |
696 | destroy_workqueue(fsc->pg_inv_wq); | |
697 | destroy_workqueue(fsc->trunc_wq); | |
16725b9d | 698 | |
3d14c5d2 | 699 | mempool_destroy(fsc->wb_pagevec_pool); |
16725b9d | 700 | |
3d14c5d2 | 701 | destroy_mount_options(fsc->mount_options); |
5dfc589a | 702 | |
3d14c5d2 | 703 | ceph_destroy_client(fsc->client); |
16725b9d | 704 | |
3d14c5d2 YS |
705 | kfree(fsc); |
706 | dout("destroy_fs_client %p done\n", fsc); | |
16725b9d SW |
707 | } |
708 | ||
0743304d | 709 | /* |
3d14c5d2 | 710 | * caches |
0743304d | 711 | */ |
3d14c5d2 YS |
712 | struct kmem_cache *ceph_inode_cachep; |
713 | struct kmem_cache *ceph_cap_cachep; | |
f66fd9f0 | 714 | struct kmem_cache *ceph_cap_flush_cachep; |
3d14c5d2 YS |
715 | struct kmem_cache *ceph_dentry_cachep; |
716 | struct kmem_cache *ceph_file_cachep; | |
bb48bd4d | 717 | struct kmem_cache *ceph_dir_file_cachep; |
3d14c5d2 YS |
718 | |
719 | static void ceph_inode_init_once(void *foo) | |
0743304d | 720 | { |
3d14c5d2 YS |
721 | struct ceph_inode_info *ci = foo; |
722 | inode_init_once(&ci->vfs_inode); | |
723 | } | |
724 | ||
725 | static int __init init_caches(void) | |
726 | { | |
99ccbd22 MT |
727 | int error = -ENOMEM; |
728 | ||
3d14c5d2 YS |
729 | ceph_inode_cachep = kmem_cache_create("ceph_inode_info", |
730 | sizeof(struct ceph_inode_info), | |
731 | __alignof__(struct ceph_inode_info), | |
5d097056 VD |
732 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD| |
733 | SLAB_ACCOUNT, ceph_inode_init_once); | |
d37b1d99 | 734 | if (!ceph_inode_cachep) |
3d14c5d2 YS |
735 | return -ENOMEM; |
736 | ||
bc4b5ad3 | 737 | ceph_cap_cachep = KMEM_CACHE(ceph_cap, SLAB_MEM_SPREAD); |
d37b1d99 | 738 | if (!ceph_cap_cachep) |
3d14c5d2 | 739 | goto bad_cap; |
f66fd9f0 YZ |
740 | ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush, |
741 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
d37b1d99 | 742 | if (!ceph_cap_flush_cachep) |
f66fd9f0 | 743 | goto bad_cap_flush; |
3d14c5d2 YS |
744 | |
745 | ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info, | |
746 | SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD); | |
d37b1d99 | 747 | if (!ceph_dentry_cachep) |
3d14c5d2 YS |
748 | goto bad_dentry; |
749 | ||
6b1a9a6c | 750 | ceph_file_cachep = KMEM_CACHE(ceph_file_info, SLAB_MEM_SPREAD); |
d37b1d99 | 751 | if (!ceph_file_cachep) |
3d14c5d2 YS |
752 | goto bad_file; |
753 | ||
bb48bd4d CX |
754 | ceph_dir_file_cachep = KMEM_CACHE(ceph_dir_file_info, SLAB_MEM_SPREAD); |
755 | if (!ceph_dir_file_cachep) | |
756 | goto bad_dir_file; | |
757 | ||
1c789249 CX |
758 | error = ceph_fscache_register(); |
759 | if (error) | |
760 | goto bad_fscache; | |
3d14c5d2 | 761 | |
99ccbd22 | 762 | return 0; |
1c789249 CX |
763 | |
764 | bad_fscache: | |
bb48bd4d CX |
765 | kmem_cache_destroy(ceph_dir_file_cachep); |
766 | bad_dir_file: | |
1c789249 | 767 | kmem_cache_destroy(ceph_file_cachep); |
3d14c5d2 YS |
768 | bad_file: |
769 | kmem_cache_destroy(ceph_dentry_cachep); | |
770 | bad_dentry: | |
f66fd9f0 YZ |
771 | kmem_cache_destroy(ceph_cap_flush_cachep); |
772 | bad_cap_flush: | |
3d14c5d2 YS |
773 | kmem_cache_destroy(ceph_cap_cachep); |
774 | bad_cap: | |
775 | kmem_cache_destroy(ceph_inode_cachep); | |
99ccbd22 | 776 | return error; |
0743304d SW |
777 | } |
778 | ||
3d14c5d2 YS |
779 | static void destroy_caches(void) |
780 | { | |
8c0a8537 KS |
781 | /* |
782 | * Make sure all delayed rcu free inodes are flushed before we | |
783 | * destroy cache. | |
784 | */ | |
785 | rcu_barrier(); | |
99ccbd22 | 786 | |
3d14c5d2 YS |
787 | kmem_cache_destroy(ceph_inode_cachep); |
788 | kmem_cache_destroy(ceph_cap_cachep); | |
f66fd9f0 | 789 | kmem_cache_destroy(ceph_cap_flush_cachep); |
3d14c5d2 YS |
790 | kmem_cache_destroy(ceph_dentry_cachep); |
791 | kmem_cache_destroy(ceph_file_cachep); | |
bb48bd4d | 792 | kmem_cache_destroy(ceph_dir_file_cachep); |
99ccbd22 MT |
793 | |
794 | ceph_fscache_unregister(); | |
3d14c5d2 YS |
795 | } |
796 | ||
797 | ||
16725b9d | 798 | /* |
3d14c5d2 YS |
799 | * ceph_umount_begin - initiate forced umount. Tear down down the |
800 | * mount, skipping steps that may hang while waiting for server(s). | |
16725b9d | 801 | */ |
3d14c5d2 | 802 | static void ceph_umount_begin(struct super_block *sb) |
16725b9d | 803 | { |
3d14c5d2 YS |
804 | struct ceph_fs_client *fsc = ceph_sb_to_client(sb); |
805 | ||
806 | dout("ceph_umount_begin - starting forced umount\n"); | |
807 | if (!fsc) | |
808 | return; | |
809 | fsc->mount_state = CEPH_MOUNT_SHUTDOWN; | |
12b69d5f | 810 | ceph_osdc_abort_requests(&fsc->client->osdc, -EIO); |
48fec5d0 | 811 | ceph_mdsc_force_umount(fsc->mdsc); |
3d14c5d2 | 812 | return; |
16725b9d SW |
813 | } |
814 | ||
3d14c5d2 YS |
815 | static const struct super_operations ceph_super_ops = { |
816 | .alloc_inode = ceph_alloc_inode, | |
817 | .destroy_inode = ceph_destroy_inode, | |
818 | .write_inode = ceph_write_inode, | |
9f12bd11 | 819 | .drop_inode = ceph_drop_inode, |
3d14c5d2 YS |
820 | .sync_fs = ceph_sync_fs, |
821 | .put_super = ceph_put_super, | |
822 | .show_options = ceph_show_options, | |
823 | .statfs = ceph_statfs, | |
824 | .umount_begin = ceph_umount_begin, | |
825 | }; | |
826 | ||
16725b9d SW |
827 | /* |
828 | * Bootstrap mount by opening the root directory. Note the mount | |
829 | * @started time from caller, and time out if this takes too long. | |
830 | */ | |
3d14c5d2 | 831 | static struct dentry *open_root_dentry(struct ceph_fs_client *fsc, |
16725b9d SW |
832 | const char *path, |
833 | unsigned long started) | |
834 | { | |
3d14c5d2 | 835 | struct ceph_mds_client *mdsc = fsc->mdsc; |
16725b9d SW |
836 | struct ceph_mds_request *req = NULL; |
837 | int err; | |
838 | struct dentry *root; | |
839 | ||
840 | /* open dir */ | |
841 | dout("open_root_inode opening '%s'\n", path); | |
842 | req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS); | |
843 | if (IS_ERR(req)) | |
7e34bc52 | 844 | return ERR_CAST(req); |
16725b9d | 845 | req->r_path1 = kstrdup(path, GFP_NOFS); |
a149bb9a SK |
846 | if (!req->r_path1) { |
847 | root = ERR_PTR(-ENOMEM); | |
848 | goto out; | |
849 | } | |
850 | ||
16725b9d SW |
851 | req->r_ino1.ino = CEPH_INO_ROOT; |
852 | req->r_ino1.snap = CEPH_NOSNAP; | |
853 | req->r_started = started; | |
a319bf56 | 854 | req->r_timeout = fsc->client->options->mount_timeout; |
16725b9d SW |
855 | req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE); |
856 | req->r_num_caps = 2; | |
857 | err = ceph_mdsc_do_request(mdsc, NULL, req); | |
858 | if (err == 0) { | |
3c5184ef AV |
859 | struct inode *inode = req->r_target_inode; |
860 | req->r_target_inode = NULL; | |
16725b9d | 861 | dout("open_root_inode success\n"); |
ce2728aa YZ |
862 | root = d_make_root(inode); |
863 | if (!root) { | |
864 | root = ERR_PTR(-ENOMEM); | |
865 | goto out; | |
774ac21d | 866 | } |
16725b9d SW |
867 | dout("open_root_inode success, root dentry is %p\n", root); |
868 | } else { | |
869 | root = ERR_PTR(err); | |
870 | } | |
3c5184ef | 871 | out: |
16725b9d SW |
872 | ceph_mdsc_put_request(req); |
873 | return root; | |
874 | } | |
875 | ||
3d14c5d2 YS |
876 | |
877 | ||
878 | ||
16725b9d SW |
879 | /* |
880 | * mount: join the ceph cluster, and open root directory. | |
881 | */ | |
3f384954 | 882 | static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc) |
16725b9d | 883 | { |
16725b9d | 884 | int err; |
16725b9d SW |
885 | unsigned long started = jiffies; /* note the start time */ |
886 | struct dentry *root; | |
887 | ||
132ca7e1 | 888 | dout("mount start %p\n", fsc); |
3d14c5d2 | 889 | mutex_lock(&fsc->client->mount_mutex); |
16725b9d | 890 | |
132ca7e1 | 891 | if (!fsc->sb->s_root) { |
ce2728aa | 892 | const char *path; |
132ca7e1 YZ |
893 | err = __ceph_open_session(fsc->client, started); |
894 | if (err < 0) | |
895 | goto out; | |
16725b9d | 896 | |
1d8f8360 YZ |
897 | /* setup fscache */ |
898 | if (fsc->mount_options->flags & CEPH_MOUNT_OPT_FSCACHE) { | |
899 | err = ceph_fscache_register_fs(fsc); | |
900 | if (err < 0) | |
901 | goto out; | |
902 | } | |
903 | ||
ce2728aa YZ |
904 | if (!fsc->mount_options->server_path) { |
905 | path = ""; | |
906 | dout("mount opening path \\t\n"); | |
907 | } else { | |
908 | path = fsc->mount_options->server_path + 1; | |
909 | dout("mount opening path %s\n", path); | |
910 | } | |
18106734 CX |
911 | |
912 | err = ceph_fs_debugfs_init(fsc); | |
913 | if (err < 0) | |
914 | goto out; | |
915 | ||
ce2728aa | 916 | root = open_root_dentry(fsc, path, started); |
132ca7e1 YZ |
917 | if (IS_ERR(root)) { |
918 | err = PTR_ERR(root); | |
919 | goto out; | |
920 | } | |
ce2728aa | 921 | fsc->sb->s_root = dget(root); |
31ca5878 GU |
922 | } else { |
923 | root = dget(fsc->sb->s_root); | |
3d14c5d2 | 924 | } |
16725b9d | 925 | |
3d14c5d2 | 926 | fsc->mount_state = CEPH_MOUNT_MOUNTED; |
16725b9d | 927 | dout("mount success\n"); |
a7f9fb20 AV |
928 | mutex_unlock(&fsc->client->mount_mutex); |
929 | return root; | |
16725b9d | 930 | |
132ca7e1 YZ |
931 | out: |
932 | mutex_unlock(&fsc->client->mount_mutex); | |
933 | return ERR_PTR(err); | |
16725b9d SW |
934 | } |
935 | ||
936 | static int ceph_set_super(struct super_block *s, void *data) | |
937 | { | |
3d14c5d2 | 938 | struct ceph_fs_client *fsc = data; |
16725b9d SW |
939 | int ret; |
940 | ||
941 | dout("set_super %p data %p\n", s, data); | |
942 | ||
3d14c5d2 | 943 | s->s_flags = fsc->mount_options->sb_flags; |
16725b9d SW |
944 | s->s_maxbytes = 1ULL << 40; /* temp value until we get mdsmap */ |
945 | ||
7221fe4c | 946 | s->s_xattr = ceph_xattr_handlers; |
3d14c5d2 YS |
947 | s->s_fs_info = fsc; |
948 | fsc->sb = s; | |
16725b9d SW |
949 | |
950 | s->s_op = &ceph_super_ops; | |
18fc8abd | 951 | s->s_d_op = &ceph_dentry_ops; |
16725b9d SW |
952 | s->s_export_op = &ceph_export_ops; |
953 | ||
954 | s->s_time_gran = 1000; /* 1000 ns == 1 us */ | |
955 | ||
956 | ret = set_anon_super(s, NULL); /* what is that second arg for? */ | |
957 | if (ret != 0) | |
958 | goto fail; | |
959 | ||
960 | return ret; | |
961 | ||
962 | fail: | |
963 | s->s_fs_info = NULL; | |
3d14c5d2 | 964 | fsc->sb = NULL; |
16725b9d SW |
965 | return ret; |
966 | } | |
967 | ||
968 | /* | |
969 | * share superblock if same fs AND options | |
970 | */ | |
971 | static int ceph_compare_super(struct super_block *sb, void *data) | |
972 | { | |
3d14c5d2 YS |
973 | struct ceph_fs_client *new = data; |
974 | struct ceph_mount_options *fsopt = new->mount_options; | |
975 | struct ceph_options *opt = new->client->options; | |
976 | struct ceph_fs_client *other = ceph_sb_to_client(sb); | |
16725b9d SW |
977 | |
978 | dout("ceph_compare_super %p\n", sb); | |
3d14c5d2 YS |
979 | |
980 | if (compare_mount_options(fsopt, opt, other)) { | |
981 | dout("monitor(s)/mount options don't match\n"); | |
982 | return 0; | |
16725b9d | 983 | } |
3d14c5d2 YS |
984 | if ((opt->flags & CEPH_OPT_FSID) && |
985 | ceph_fsid_compare(&opt->fsid, &other->client->fsid)) { | |
986 | dout("fsid doesn't match\n"); | |
987 | return 0; | |
988 | } | |
989 | if (fsopt->sb_flags != other->mount_options->sb_flags) { | |
16725b9d SW |
990 | dout("flags differ\n"); |
991 | return 0; | |
992 | } | |
993 | return 1; | |
994 | } | |
995 | ||
996 | /* | |
997 | * construct our own bdi so we can control readahead, etc. | |
998 | */ | |
00d5643e | 999 | static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0); |
31e0cf8f | 1000 | |
09dc9fc2 | 1001 | static int ceph_setup_bdi(struct super_block *sb, struct ceph_fs_client *fsc) |
16725b9d SW |
1002 | { |
1003 | int err; | |
1004 | ||
09dc9fc2 JK |
1005 | err = super_setup_bdi_name(sb, "ceph-%ld", |
1006 | atomic_long_inc_return(&bdi_seq)); | |
1007 | if (err) | |
1008 | return err; | |
1009 | ||
83817e35 | 1010 | /* set ra_pages based on rasize mount option? */ |
4214fb15 | 1011 | sb->s_bdi->ra_pages = fsc->mount_options->rasize >> PAGE_SHIFT; |
e9852227 | 1012 | |
aa187926 YZ |
1013 | /* set io_pages based on max osd read size */ |
1014 | sb->s_bdi->io_pages = fsc->mount_options->rsize >> PAGE_SHIFT; | |
7c94ba27 | 1015 | |
09dc9fc2 | 1016 | return 0; |
16725b9d SW |
1017 | } |
1018 | ||
a7f9fb20 AV |
1019 | static struct dentry *ceph_mount(struct file_system_type *fs_type, |
1020 | int flags, const char *dev_name, void *data) | |
16725b9d SW |
1021 | { |
1022 | struct super_block *sb; | |
3d14c5d2 | 1023 | struct ceph_fs_client *fsc; |
a7f9fb20 | 1024 | struct dentry *res; |
16725b9d SW |
1025 | int err; |
1026 | int (*compare_super)(struct super_block *, void *) = ceph_compare_super; | |
3d14c5d2 YS |
1027 | struct ceph_mount_options *fsopt = NULL; |
1028 | struct ceph_options *opt = NULL; | |
16725b9d | 1029 | |
a7f9fb20 | 1030 | dout("ceph_mount\n"); |
45195e42 SW |
1031 | |
1032 | #ifdef CONFIG_CEPH_FS_POSIX_ACL | |
1751e8a6 | 1033 | flags |= SB_POSIXACL; |
45195e42 | 1034 | #endif |
3f384954 | 1035 | err = parse_mount_options(&fsopt, &opt, flags, data, dev_name); |
a7f9fb20 AV |
1036 | if (err < 0) { |
1037 | res = ERR_PTR(err); | |
6b805185 | 1038 | goto out_final; |
a7f9fb20 | 1039 | } |
16725b9d SW |
1040 | |
1041 | /* create client (which we may/may not use) */ | |
3d14c5d2 YS |
1042 | fsc = create_fs_client(fsopt, opt); |
1043 | if (IS_ERR(fsc)) { | |
a7f9fb20 | 1044 | res = ERR_CAST(fsc); |
259a187a NW |
1045 | destroy_mount_options(fsopt); |
1046 | ceph_destroy_options(opt); | |
6b805185 SW |
1047 | goto out_final; |
1048 | } | |
16725b9d | 1049 | |
3d14c5d2 | 1050 | err = ceph_mdsc_init(fsc); |
a7f9fb20 AV |
1051 | if (err < 0) { |
1052 | res = ERR_PTR(err); | |
3d14c5d2 | 1053 | goto out; |
a7f9fb20 | 1054 | } |
3d14c5d2 YS |
1055 | |
1056 | if (ceph_test_opt(fsc->client, NOSHARE)) | |
16725b9d | 1057 | compare_super = NULL; |
9249e17f | 1058 | sb = sget(fs_type, compare_super, ceph_set_super, flags, fsc); |
16725b9d | 1059 | if (IS_ERR(sb)) { |
a7f9fb20 | 1060 | res = ERR_CAST(sb); |
16725b9d SW |
1061 | goto out; |
1062 | } | |
1063 | ||
3d14c5d2 YS |
1064 | if (ceph_sb_to_client(sb) != fsc) { |
1065 | ceph_mdsc_destroy(fsc); | |
1066 | destroy_fs_client(fsc); | |
1067 | fsc = ceph_sb_to_client(sb); | |
1068 | dout("get_sb got existing client %p\n", fsc); | |
16725b9d | 1069 | } else { |
3d14c5d2 | 1070 | dout("get_sb using new client %p\n", fsc); |
09dc9fc2 | 1071 | err = ceph_setup_bdi(sb, fsc); |
a7f9fb20 AV |
1072 | if (err < 0) { |
1073 | res = ERR_PTR(err); | |
16725b9d | 1074 | goto out_splat; |
a7f9fb20 | 1075 | } |
16725b9d SW |
1076 | } |
1077 | ||
3f384954 | 1078 | res = ceph_real_mount(fsc); |
a7f9fb20 | 1079 | if (IS_ERR(res)) |
16725b9d | 1080 | goto out_splat; |
a7f9fb20 | 1081 | dout("root %p inode %p ino %llx.%llx\n", res, |
2b0143b5 | 1082 | d_inode(res), ceph_vinop(d_inode(res))); |
a7f9fb20 | 1083 | return res; |
16725b9d SW |
1084 | |
1085 | out_splat: | |
3d14c5d2 | 1086 | ceph_mdsc_close_sessions(fsc->mdsc); |
3981f2e2 | 1087 | deactivate_locked_super(sb); |
16725b9d SW |
1088 | goto out_final; |
1089 | ||
1090 | out: | |
3d14c5d2 YS |
1091 | ceph_mdsc_destroy(fsc); |
1092 | destroy_fs_client(fsc); | |
16725b9d | 1093 | out_final: |
a7f9fb20 AV |
1094 | dout("ceph_mount fail %ld\n", PTR_ERR(res)); |
1095 | return res; | |
16725b9d SW |
1096 | } |
1097 | ||
1098 | static void ceph_kill_sb(struct super_block *s) | |
1099 | { | |
3d14c5d2 | 1100 | struct ceph_fs_client *fsc = ceph_sb_to_client(s); |
e4d27509 CH |
1101 | dev_t dev = s->s_dev; |
1102 | ||
16725b9d | 1103 | dout("kill_sb %p\n", s); |
e4d27509 | 1104 | |
3d14c5d2 | 1105 | ceph_mdsc_pre_umount(fsc->mdsc); |
a57d9064 YZ |
1106 | flush_fs_workqueues(fsc); |
1107 | ||
e4d27509 | 1108 | generic_shutdown_super(s); |
62a65f36 YZ |
1109 | |
1110 | fsc->client->extra_mon_dispatch = NULL; | |
1111 | ceph_fs_debugfs_cleanup(fsc); | |
1112 | ||
1d8f8360 YZ |
1113 | ceph_fscache_unregister_fs(fsc); |
1114 | ||
3d14c5d2 | 1115 | ceph_mdsc_destroy(fsc); |
e4d27509 | 1116 | |
3d14c5d2 | 1117 | destroy_fs_client(fsc); |
e4d27509 | 1118 | free_anon_bdev(dev); |
16725b9d SW |
1119 | } |
1120 | ||
1121 | static struct file_system_type ceph_fs_type = { | |
1122 | .owner = THIS_MODULE, | |
1123 | .name = "ceph", | |
a7f9fb20 | 1124 | .mount = ceph_mount, |
16725b9d SW |
1125 | .kill_sb = ceph_kill_sb, |
1126 | .fs_flags = FS_RENAME_DOES_D_MOVE, | |
1127 | }; | |
7f78e035 | 1128 | MODULE_ALIAS_FS("ceph"); |
16725b9d | 1129 | |
16725b9d SW |
1130 | static int __init init_ceph(void) |
1131 | { | |
3d14c5d2 | 1132 | int ret = init_caches(); |
16725b9d | 1133 | if (ret) |
3d14c5d2 | 1134 | goto out; |
16725b9d | 1135 | |
eb13e832 | 1136 | ceph_flock_init(); |
3ce6cd12 | 1137 | ceph_xattr_init(); |
16725b9d SW |
1138 | ret = register_filesystem(&ceph_fs_type); |
1139 | if (ret) | |
34b759b4 | 1140 | goto out_xattr; |
16725b9d | 1141 | |
3d14c5d2 YS |
1142 | pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL); |
1143 | ||
16725b9d SW |
1144 | return 0; |
1145 | ||
97c85a82 | 1146 | out_xattr: |
3ce6cd12 | 1147 | ceph_xattr_exit(); |
16725b9d | 1148 | destroy_caches(); |
16725b9d SW |
1149 | out: |
1150 | return ret; | |
1151 | } | |
1152 | ||
1153 | static void __exit exit_ceph(void) | |
1154 | { | |
1155 | dout("exit_ceph\n"); | |
1156 | unregister_filesystem(&ceph_fs_type); | |
3ce6cd12 | 1157 | ceph_xattr_exit(); |
16725b9d | 1158 | destroy_caches(); |
16725b9d SW |
1159 | } |
1160 | ||
1161 | module_init(init_ceph); | |
1162 | module_exit(exit_ceph); | |
1163 | ||
1164 | MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); | |
1165 | MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); | |
1166 | MODULE_AUTHOR("Patience Warnick <patience@newdream.net>"); | |
1167 | MODULE_DESCRIPTION("Ceph filesystem for Linux"); | |
1168 | MODULE_LICENSE("GPL"); |