blob: b382e5910eea8bf7cf7a4711fac79104e6466224 [file] [log] [blame]
Sage Weil16725b92009-10-06 11:31:07 -07001
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07002#include <linux/ceph/ceph_debug.h>
Sage Weil16725b92009-10-06 11:31:07 -07003
4#include <linux/backing-dev.h>
Sage Weilc309f0a2010-06-30 21:34:01 -07005#include <linux/ctype.h>
Sage Weil16725b92009-10-06 11:31:07 -07006#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>
Sage Weil16725b92009-10-06 11:31:07 -070012#include <linux/sched.h>
13#include <linux/seq_file.h>
Tejun Heo5a0e3ad2010-03-24 17:04:11 +090014#include <linux/slab.h>
Sage Weil16725b92009-10-06 11:31:07 -070015#include <linux/statfs.h>
16#include <linux/string.h>
Sage Weil16725b92009-10-06 11:31:07 -070017
Sage Weil16725b92009-10-06 11:31:07 -070018#include "super.h"
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070019#include "mds_client.h"
Milosz Tanski99ccbd22013-08-21 17:29:54 -040020#include "cache.h"
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070021
Sage Weil1fe60e52012-07-30 16:23:22 -070022#include <linux/ceph/ceph_features.h>
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070023#include <linux/ceph/decode.h>
24#include <linux/ceph/mon_client.h>
25#include <linux/ceph/auth.h>
26#include <linux/ceph/debugfs.h>
Sage Weil16725b92009-10-06 11:31:07 -070027
28/*
29 * Ceph superblock operations
30 *
31 * Handle the basics of mounting, unmounting.
32 */
33
Sage Weil16725b92009-10-06 11:31:07 -070034/*
35 * super ops
36 */
37static void ceph_put_super(struct super_block *s)
38{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070039 struct ceph_fs_client *fsc = ceph_sb_to_client(s);
Sage Weil16725b92009-10-06 11:31:07 -070040
41 dout("put_super\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070042 ceph_mdsc_close_sessions(fsc->mdsc);
Sage Weil16725b92009-10-06 11:31:07 -070043}
44
45static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
46{
David Howells2b0143b2015-03-17 22:25:59 +000047 struct ceph_fs_client *fsc = ceph_inode_to_client(d_inode(dentry));
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070048 struct ceph_monmap *monmap = fsc->client->monc.monmap;
Sage Weil16725b92009-10-06 11:31:07 -070049 struct ceph_statfs st;
50 u64 fsid;
51 int err;
52
53 dout("statfs\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070054 err = ceph_monc_do_statfs(&fsc->client->monc, &st);
Sage Weil16725b92009-10-06 11:31:07 -070055 if (err < 0)
56 return err;
57
58 /* fill in kstatfs */
59 buf->f_type = CEPH_SUPER_MAGIC; /* ?? */
60
61 /*
62 * express utilization in terms of large blocks to avoid
63 * overflow on 32-bit machines.
Sage Weil92a49fb2013-02-22 15:31:00 -080064 *
65 * NOTE: for the time being, we make bsize == frsize to humor
66 * not-yet-ancient versions of glibc that are broken.
67 * Someday, we will probably want to report a real block
68 * size... whatever that may mean for a network file system!
Sage Weil16725b92009-10-06 11:31:07 -070069 */
70 buf->f_bsize = 1 << CEPH_BLOCK_SHIFT;
Sage Weil92a49fb2013-02-22 15:31:00 -080071 buf->f_frsize = 1 << CEPH_BLOCK_SHIFT;
Sage Weil16725b92009-10-06 11:31:07 -070072 buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10);
Greg Farnum8f04d422011-07-26 11:26:54 -070073 buf->f_bfree = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
Sage Weil16725b92009-10-06 11:31:07 -070074 buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
75
76 buf->f_files = le64_to_cpu(st.num_objects);
77 buf->f_ffree = -1;
Sage Weil558d3492010-06-01 12:51:12 -070078 buf->f_namelen = NAME_MAX;
Sage Weil16725b92009-10-06 11:31:07 -070079
80 /* leave fsid little-endian, regardless of host endianness */
81 fsid = *(u64 *)(&monmap->fsid) ^ *((u64 *)&monmap->fsid + 1);
82 buf->f_fsid.val[0] = fsid & 0xffffffff;
83 buf->f_fsid.val[1] = fsid >> 32;
84
85 return 0;
86}
87
88
Sage Weil2d9c98a2010-07-30 09:38:13 -070089static int ceph_sync_fs(struct super_block *sb, int wait)
Sage Weil16725b92009-10-06 11:31:07 -070090{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070091 struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
Sage Weil2d9c98a2010-07-30 09:38:13 -070092
93 if (!wait) {
94 dout("sync_fs (non-blocking)\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -070095 ceph_flush_dirty_caps(fsc->mdsc);
Sage Weil2d9c98a2010-07-30 09:38:13 -070096 dout("sync_fs (non-blocking) done\n");
97 return 0;
98 }
99
100 dout("sync_fs (blocking)\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700101 ceph_osdc_sync(&fsc->client->osdc);
102 ceph_mdsc_sync(fsc->mdsc);
Sage Weil2d9c98a2010-07-30 09:38:13 -0700103 dout("sync_fs (blocking) done\n");
Sage Weil16725b92009-10-06 11:31:07 -0700104 return 0;
105}
106
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700107/*
108 * mount options
109 */
110enum {
111 Opt_wsize,
112 Opt_rsize,
Sage Weil83817e32011-08-04 08:03:44 -0700113 Opt_rasize,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700114 Opt_caps_wanted_delay_min,
115 Opt_caps_wanted_delay_max,
116 Opt_cap_release_safety,
117 Opt_readdir_max_entries,
118 Opt_readdir_max_bytes,
119 Opt_congestion_kb,
120 Opt_last_int,
121 /* int args above */
122 Opt_snapdirname,
Yan, Zheng430afba2016-07-08 11:25:38 +0800123 Opt_mds_namespace,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700124 Opt_last_string,
125 /* string args above */
126 Opt_dirstat,
127 Opt_nodirstat,
128 Opt_rbytes,
129 Opt_norbytes,
Alex Eldercffaba12012-02-15 07:43:54 -0600130 Opt_asyncreaddir,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700131 Opt_noasyncreaddir,
Sage Weila40dc6c2012-01-10 09:12:55 -0800132 Opt_dcache,
133 Opt_nodcache,
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800134 Opt_ino32,
Alex Eldercffaba12012-02-15 07:43:54 -0600135 Opt_noino32,
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400136 Opt_fscache,
Sage Weil45195e42014-02-16 10:05:29 -0800137 Opt_nofscache,
Yan, Zheng10183a62015-04-27 15:33:28 +0800138 Opt_poolperm,
139 Opt_nopoolperm,
Sage Weil45195e42014-02-16 10:05:29 -0800140#ifdef CONFIG_CEPH_FS_POSIX_ACL
141 Opt_acl,
142#endif
Yan, Zheng10183a62015-04-27 15:33:28 +0800143 Opt_noacl,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700144};
145
146static match_table_t fsopt_tokens = {
147 {Opt_wsize, "wsize=%d"},
148 {Opt_rsize, "rsize=%d"},
Sage Weil83817e32011-08-04 08:03:44 -0700149 {Opt_rasize, "rasize=%d"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700150 {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
151 {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"},
152 {Opt_cap_release_safety, "cap_release_safety=%d"},
153 {Opt_readdir_max_entries, "readdir_max_entries=%d"},
154 {Opt_readdir_max_bytes, "readdir_max_bytes=%d"},
155 {Opt_congestion_kb, "write_congestion_kb=%d"},
156 /* int args above */
157 {Opt_snapdirname, "snapdirname=%s"},
Yan, Zheng430afba2016-07-08 11:25:38 +0800158 {Opt_mds_namespace, "mds_namespace=%s"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700159 /* string args above */
160 {Opt_dirstat, "dirstat"},
161 {Opt_nodirstat, "nodirstat"},
162 {Opt_rbytes, "rbytes"},
163 {Opt_norbytes, "norbytes"},
Alex Eldercffaba12012-02-15 07:43:54 -0600164 {Opt_asyncreaddir, "asyncreaddir"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700165 {Opt_noasyncreaddir, "noasyncreaddir"},
Sage Weila40dc6c2012-01-10 09:12:55 -0800166 {Opt_dcache, "dcache"},
167 {Opt_nodcache, "nodcache"},
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800168 {Opt_ino32, "ino32"},
Alex Eldercffaba12012-02-15 07:43:54 -0600169 {Opt_noino32, "noino32"},
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400170 {Opt_fscache, "fsc"},
171 {Opt_nofscache, "nofsc"},
Yan, Zheng10183a62015-04-27 15:33:28 +0800172 {Opt_poolperm, "poolperm"},
173 {Opt_nopoolperm, "nopoolperm"},
Sage Weil45195e42014-02-16 10:05:29 -0800174#ifdef CONFIG_CEPH_FS_POSIX_ACL
175 {Opt_acl, "acl"},
176#endif
177 {Opt_noacl, "noacl"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700178 {-1, NULL}
179};
180
181static int parse_fsopt_token(char *c, void *private)
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800182{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700183 struct ceph_mount_options *fsopt = private;
184 substring_t argstr[MAX_OPT_ARGS];
185 int token, intval, ret;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800186
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700187 token = match_token((char *)c, fsopt_tokens, argstr);
188 if (token < 0)
189 return -EINVAL;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800190
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700191 if (token < Opt_last_int) {
192 ret = match_int(&argstr[0], &intval);
193 if (ret < 0) {
194 pr_err("bad mount option arg (not int) "
195 "at '%s'\n", c);
196 return ret;
197 }
198 dout("got int token %d val %d\n", token, intval);
199 } else if (token > Opt_last_int && token < Opt_last_string) {
200 dout("got string token %d val %s\n", token,
201 argstr[0].from);
202 } else {
203 dout("got token %d\n", token);
204 }
205
206 switch (token) {
207 case Opt_snapdirname:
208 kfree(fsopt->snapdir_name);
209 fsopt->snapdir_name = kstrndup(argstr[0].from,
210 argstr[0].to-argstr[0].from,
211 GFP_KERNEL);
212 if (!fsopt->snapdir_name)
213 return -ENOMEM;
214 break;
Yan, Zheng235a0982016-03-30 17:18:34 +0800215 case Opt_mds_namespace:
Yan, Zheng430afba2016-07-08 11:25:38 +0800216 fsopt->mds_namespace = kstrndup(argstr[0].from,
217 argstr[0].to-argstr[0].from,
218 GFP_KERNEL);
219 if (!fsopt->mds_namespace)
220 return -ENOMEM;
Yan, Zheng235a0982016-03-30 17:18:34 +0800221 break;
Yan, Zheng430afba2016-07-08 11:25:38 +0800222 /* misc */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700223 case Opt_wsize:
224 fsopt->wsize = intval;
225 break;
226 case Opt_rsize:
227 fsopt->rsize = intval;
228 break;
Sage Weil83817e32011-08-04 08:03:44 -0700229 case Opt_rasize:
230 fsopt->rasize = intval;
231 break;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700232 case Opt_caps_wanted_delay_min:
233 fsopt->caps_wanted_delay_min = intval;
234 break;
235 case Opt_caps_wanted_delay_max:
236 fsopt->caps_wanted_delay_max = intval;
237 break;
238 case Opt_readdir_max_entries:
239 fsopt->max_readdir = intval;
240 break;
241 case Opt_readdir_max_bytes:
242 fsopt->max_readdir_bytes = intval;
243 break;
244 case Opt_congestion_kb:
245 fsopt->congestion_kb = intval;
246 break;
247 case Opt_dirstat:
248 fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT;
249 break;
250 case Opt_nodirstat:
251 fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT;
252 break;
253 case Opt_rbytes:
254 fsopt->flags |= CEPH_MOUNT_OPT_RBYTES;
255 break;
256 case Opt_norbytes:
257 fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES;
258 break;
Alex Eldercffaba12012-02-15 07:43:54 -0600259 case Opt_asyncreaddir:
260 fsopt->flags &= ~CEPH_MOUNT_OPT_NOASYNCREADDIR;
261 break;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700262 case Opt_noasyncreaddir:
263 fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
264 break;
Sage Weila40dc6c2012-01-10 09:12:55 -0800265 case Opt_dcache:
266 fsopt->flags |= CEPH_MOUNT_OPT_DCACHE;
267 break;
268 case Opt_nodcache:
269 fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
270 break;
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800271 case Opt_ino32:
272 fsopt->flags |= CEPH_MOUNT_OPT_INO32;
273 break;
Alex Eldercffaba12012-02-15 07:43:54 -0600274 case Opt_noino32:
275 fsopt->flags &= ~CEPH_MOUNT_OPT_INO32;
276 break;
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400277 case Opt_fscache:
278 fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE;
279 break;
280 case Opt_nofscache:
281 fsopt->flags &= ~CEPH_MOUNT_OPT_FSCACHE;
282 break;
Yan, Zheng10183a62015-04-27 15:33:28 +0800283 case Opt_poolperm:
284 fsopt->flags &= ~CEPH_MOUNT_OPT_NOPOOLPERM;
285 printk ("pool perm");
286 break;
287 case Opt_nopoolperm:
288 fsopt->flags |= CEPH_MOUNT_OPT_NOPOOLPERM;
289 break;
Sage Weil45195e42014-02-16 10:05:29 -0800290#ifdef CONFIG_CEPH_FS_POSIX_ACL
291 case Opt_acl:
292 fsopt->sb_flags |= MS_POSIXACL;
293 break;
294#endif
295 case Opt_noacl:
296 fsopt->sb_flags &= ~MS_POSIXACL;
297 break;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700298 default:
299 BUG_ON(token);
300 }
301 return 0;
302}
303
304static void destroy_mount_options(struct ceph_mount_options *args)
305{
306 dout("destroy_mount_options %p\n", args);
307 kfree(args->snapdir_name);
Yan, Zheng430afba2016-07-08 11:25:38 +0800308 kfree(args->mds_namespace);
Yan, Zheng3f384952016-04-21 11:09:55 +0800309 kfree(args->server_path);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700310 kfree(args);
311}
312
313static int strcmp_null(const char *s1, const char *s2)
314{
315 if (!s1 && !s2)
316 return 0;
317 if (s1 && !s2)
318 return -1;
319 if (!s1 && s2)
320 return 1;
321 return strcmp(s1, s2);
322}
323
324static int compare_mount_options(struct ceph_mount_options *new_fsopt,
325 struct ceph_options *new_opt,
326 struct ceph_fs_client *fsc)
327{
328 struct ceph_mount_options *fsopt1 = new_fsopt;
329 struct ceph_mount_options *fsopt2 = fsc->mount_options;
330 int ofs = offsetof(struct ceph_mount_options, snapdir_name);
331 int ret;
332
333 ret = memcmp(fsopt1, fsopt2, ofs);
334 if (ret)
335 return ret;
336
337 ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name);
338 if (ret)
339 return ret;
Yan, Zheng430afba2016-07-08 11:25:38 +0800340 ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace);
341 if (ret)
342 return ret;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700343
Yan, Zheng3f384952016-04-21 11:09:55 +0800344 ret = strcmp_null(fsopt1->server_path, fsopt2->server_path);
345 if (ret)
346 return ret;
347
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700348 return ceph_compare_options(new_opt, fsc->client);
349}
350
351static int parse_mount_options(struct ceph_mount_options **pfsopt,
352 struct ceph_options **popt,
353 int flags, char *options,
Yan, Zheng3f384952016-04-21 11:09:55 +0800354 const char *dev_name)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700355{
356 struct ceph_mount_options *fsopt;
357 const char *dev_name_end;
Alex Elderc98f5332012-08-09 10:33:26 -0700358 int err;
359
360 if (!dev_name || !*dev_name)
361 return -EINVAL;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700362
363 fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL);
364 if (!fsopt)
365 return -ENOMEM;
366
367 dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name);
368
Noah Watkins80db8be2011-08-22 13:49:23 -0600369 fsopt->sb_flags = flags;
370 fsopt->flags = CEPH_MOUNT_OPT_DEFAULT;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700371
Noah Watkins80db8be2011-08-22 13:49:23 -0600372 fsopt->rsize = CEPH_RSIZE_DEFAULT;
373 fsopt->rasize = CEPH_RASIZE_DEFAULT;
374 fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
Sanidhya Kashyapa149bb92015-03-21 12:54:58 -0400375 if (!fsopt->snapdir_name) {
376 err = -ENOMEM;
377 goto out;
378 }
379
Sage Weil50aac4f2011-01-18 07:59:40 -0800380 fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
381 fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
Noah Watkins80db8be2011-08-22 13:49:23 -0600382 fsopt->cap_release_safety = CEPH_CAP_RELEASE_SAFETY_DEFAULT;
383 fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT;
384 fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
385 fsopt->congestion_kb = default_congestion_kb();
386
Alex Elderc98f5332012-08-09 10:33:26 -0700387 /*
388 * Distinguish the server list from the path in "dev_name".
389 * Internally we do not include the leading '/' in the path.
390 *
391 * "dev_name" will look like:
392 * <server_spec>[,<server_spec>...]:[<path>]
393 * where
394 * <server_spec> is <ip>[:<port>]
395 * <path> is optional, but if present must begin with '/'
396 */
397 dev_name_end = strchr(dev_name, '/');
398 if (dev_name_end) {
Yan, Zhengce2728a2016-09-14 14:53:05 +0800399 if (strlen(dev_name_end) > 1) {
400 fsopt->server_path = kstrdup(dev_name_end, GFP_KERNEL);
401 if (!fsopt->server_path) {
402 err = -ENOMEM;
403 goto out;
404 }
Yan, Zheng3f384952016-04-21 11:09:55 +0800405 }
Alex Elderc98f5332012-08-09 10:33:26 -0700406 } else {
Alex Elderc98f5332012-08-09 10:33:26 -0700407 dev_name_end = dev_name + strlen(dev_name);
Alex Elderc98f5332012-08-09 10:33:26 -0700408 }
Noah Watkins80db8be2011-08-22 13:49:23 -0600409 err = -EINVAL;
Alex Elderc98f5332012-08-09 10:33:26 -0700410 dev_name_end--; /* back up to ':' separator */
Sasha Levin54464292013-07-01 18:33:39 -0400411 if (dev_name_end < dev_name || *dev_name_end != ':') {
Alex Elderc98f5332012-08-09 10:33:26 -0700412 pr_err("device name is missing path (no : separator in %s)\n",
Noah Watkins80db8be2011-08-22 13:49:23 -0600413 dev_name);
414 goto out;
415 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700416 dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name);
Yan, Zheng3f384952016-04-21 11:09:55 +0800417 if (fsopt->server_path)
418 dout("server path '%s'\n", fsopt->server_path);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700419
Alex Elderee577412012-01-24 10:08:36 -0600420 *popt = ceph_parse_options(options, dev_name, dev_name_end,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700421 parse_fsopt_token, (void *)fsopt);
Alex Elderee577412012-01-24 10:08:36 -0600422 if (IS_ERR(*popt)) {
423 err = PTR_ERR(*popt);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700424 goto out;
Alex Elderee577412012-01-24 10:08:36 -0600425 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700426
427 /* success */
428 *pfsopt = fsopt;
429 return 0;
430
431out:
432 destroy_mount_options(fsopt);
433 return err;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800434}
435
Sage Weil6e19a162010-04-29 16:38:32 -0700436/**
437 * ceph_show_options - Show mount options in /proc/mounts
438 * @m: seq_file to write to
Al Viro34c80b12011-12-08 21:32:45 -0500439 * @root: root of that (sub)tree
Sage Weil6e19a162010-04-29 16:38:32 -0700440 */
Al Viro34c80b12011-12-08 21:32:45 -0500441static int ceph_show_options(struct seq_file *m, struct dentry *root)
Sage Weil6e19a162010-04-29 16:38:32 -0700442{
Al Viro34c80b12011-12-08 21:32:45 -0500443 struct ceph_fs_client *fsc = ceph_sb_to_client(root->d_sb);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700444 struct ceph_mount_options *fsopt = fsc->mount_options;
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300445 size_t pos;
446 int ret;
Sage Weil6e19a162010-04-29 16:38:32 -0700447
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300448 /* a comma between MNT/MS and client options */
449 seq_putc(m, ',');
450 pos = m->count;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700451
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300452 ret = ceph_print_client_options(m, fsc->client);
453 if (ret)
454 return ret;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700455
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300456 /* retract our comma if no client options */
457 if (m->count == pos)
458 m->count--;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700459
460 if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT)
461 seq_puts(m, ",dirstat");
Yan, Zheng133e9152016-01-25 10:44:33 +0800462 if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES))
463 seq_puts(m, ",rbytes");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700464 if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
Sage Weil6e19a162010-04-29 16:38:32 -0700465 seq_puts(m, ",noasyncreaddir");
Ilya Dryomovff7eeb82015-03-25 21:10:09 +0300466 if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
Sage Weila40dc6c2012-01-10 09:12:55 -0800467 seq_puts(m, ",nodcache");
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400468 if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE)
469 seq_puts(m, ",fsc");
Yan, Zheng10183a62015-04-27 15:33:28 +0800470 if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
471 seq_puts(m, ",nopoolperm");
Sage Weil6e19a162010-04-29 16:38:32 -0700472
Sage Weil45195e42014-02-16 10:05:29 -0800473#ifdef CONFIG_CEPH_FS_POSIX_ACL
474 if (fsopt->sb_flags & MS_POSIXACL)
475 seq_puts(m, ",acl");
476 else
477 seq_puts(m, ",noacl");
478#endif
479
Yan, Zheng430afba2016-07-08 11:25:38 +0800480 if (fsopt->mds_namespace)
481 seq_printf(m, ",mds_namespace=%s", fsopt->mds_namespace);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700482 if (fsopt->wsize)
483 seq_printf(m, ",wsize=%d", fsopt->wsize);
Sage Weil80456f82011-03-10 13:33:26 -0800484 if (fsopt->rsize != CEPH_RSIZE_DEFAULT)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700485 seq_printf(m, ",rsize=%d", fsopt->rsize);
Sage Weil83817e32011-08-04 08:03:44 -0700486 if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
Sage Weil21519372011-12-01 08:06:52 -0800487 seq_printf(m, ",rasize=%d", fsopt->rasize);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700488 if (fsopt->congestion_kb != default_congestion_kb())
489 seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb);
490 if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
Sage Weil6e19a162010-04-29 16:38:32 -0700491 seq_printf(m, ",caps_wanted_delay_min=%d",
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700492 fsopt->caps_wanted_delay_min);
493 if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
Sage Weil6e19a162010-04-29 16:38:32 -0700494 seq_printf(m, ",caps_wanted_delay_max=%d",
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700495 fsopt->caps_wanted_delay_max);
496 if (fsopt->cap_release_safety != CEPH_CAP_RELEASE_SAFETY_DEFAULT)
Sage Weil6e19a162010-04-29 16:38:32 -0700497 seq_printf(m, ",cap_release_safety=%d",
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700498 fsopt->cap_release_safety);
499 if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT)
500 seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir);
501 if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
502 seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes);
503 if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
Kees Cooka068acf2015-09-04 15:44:57 -0700504 seq_show_option(m, "snapdirname", fsopt->snapdir_name);
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300505
Sage Weil6e19a162010-04-29 16:38:32 -0700506 return 0;
507}
508
509/*
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700510 * handle any mon messages the standard library doesn't understand.
511 * return error if we don't either.
512 */
513static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg)
514{
515 struct ceph_fs_client *fsc = client->private;
516 int type = le16_to_cpu(msg->hdr.type);
517
518 switch (type) {
519 case CEPH_MSG_MDS_MAP:
Yan, Zheng430afba2016-07-08 11:25:38 +0800520 ceph_mdsc_handle_mdsmap(fsc->mdsc, msg);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700521 return 0;
Yan, Zheng430afba2016-07-08 11:25:38 +0800522 case CEPH_MSG_FS_MAP_USER:
523 ceph_mdsc_handle_fsmap(fsc->mdsc, msg);
524 return 0;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700525 default:
526 return -1;
527 }
528}
529
530/*
531 * create a new fs client
532 */
H Hartley Sweeten0c6d4b4e22011-09-23 11:53:30 -0700533static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700534 struct ceph_options *opt)
535{
536 struct ceph_fs_client *fsc;
Ilya Dryomov12b46292013-12-24 21:19:23 +0200537 const u64 supported_features =
Yan, Zhengd463a432016-03-31 15:53:01 +0800538 CEPH_FEATURE_FLOCK | CEPH_FEATURE_DIRLAYOUTHASH |
539 CEPH_FEATURE_MDSENC | CEPH_FEATURE_MDS_INLINE_DATA;
Ilya Dryomov12b46292013-12-24 21:19:23 +0200540 const u64 required_features = 0;
Alex Elder3bf53332013-04-01 10:48:40 -0500541 int page_count;
542 size_t size;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700543 int err = -ENOMEM;
544
545 fsc = kzalloc(sizeof(*fsc), GFP_KERNEL);
546 if (!fsc)
547 return ERR_PTR(-ENOMEM);
548
Sage Weil6ab00d42011-08-09 09:41:59 -0700549 fsc->client = ceph_create_client(opt, fsc, supported_features,
550 required_features);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700551 if (IS_ERR(fsc->client)) {
552 err = PTR_ERR(fsc->client);
553 goto fail;
554 }
555 fsc->client->extra_mon_dispatch = extra_mon_dispatch;
Yan, Zheng430afba2016-07-08 11:25:38 +0800556
557 if (fsopt->mds_namespace == NULL) {
558 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
559 0, true);
560 } else {
561 ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_FSMAP,
562 0, false);
563 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700564
565 fsc->mount_options = fsopt;
566
567 fsc->sb = NULL;
568 fsc->mount_state = CEPH_MOUNT_MOUNTING;
569
570 atomic_long_set(&fsc->writeback_count, 0);
571
572 err = bdi_init(&fsc->backing_dev_info);
573 if (err < 0)
574 goto fail_client;
575
576 err = -ENOMEM;
Tejun Heo01e6acc2011-01-03 14:49:45 +0100577 /*
578 * The number of concurrent works can be high but they don't need
579 * to be processed in parallel, limit concurrency.
580 */
581 fsc->wb_wq = alloc_workqueue("ceph-writeback", 0, 1);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700582 if (fsc->wb_wq == NULL)
583 goto fail_bdi;
Tejun Heo01e6acc2011-01-03 14:49:45 +0100584 fsc->pg_inv_wq = alloc_workqueue("ceph-pg-invalid", 0, 1);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700585 if (fsc->pg_inv_wq == NULL)
586 goto fail_wb_wq;
Tejun Heo01e6acc2011-01-03 14:49:45 +0100587 fsc->trunc_wq = alloc_workqueue("ceph-trunc", 0, 1);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700588 if (fsc->trunc_wq == NULL)
589 goto fail_pg_inv_wq;
590
591 /* set up mempools */
592 err = -ENOMEM;
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300593 page_count = fsc->mount_options->wsize >> PAGE_SHIFT;
Alex Elder3bf53332013-04-01 10:48:40 -0500594 size = sizeof (struct page *) * (page_count ? page_count : 1);
595 fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10, size);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700596 if (!fsc->wb_pagevec_pool)
597 goto fail_trunc_wq;
598
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400599 /* setup fscache */
600 if ((fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) &&
601 (ceph_fscache_register_fs(fsc) != 0))
602 goto fail_fscache;
603
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700604 /* caps */
605 fsc->min_caps = fsopt->max_readdir;
606
607 return fsc;
608
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400609fail_fscache:
610 ceph_fscache_unregister_fs(fsc);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700611fail_trunc_wq:
612 destroy_workqueue(fsc->trunc_wq);
613fail_pg_inv_wq:
614 destroy_workqueue(fsc->pg_inv_wq);
615fail_wb_wq:
616 destroy_workqueue(fsc->wb_wq);
617fail_bdi:
618 bdi_destroy(&fsc->backing_dev_info);
619fail_client:
620 ceph_destroy_client(fsc->client);
621fail:
622 kfree(fsc);
623 return ERR_PTR(err);
624}
625
H Hartley Sweeten0c6d4b4e22011-09-23 11:53:30 -0700626static void destroy_fs_client(struct ceph_fs_client *fsc)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700627{
628 dout("destroy_fs_client %p\n", fsc);
629
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400630 ceph_fscache_unregister_fs(fsc);
631
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700632 destroy_workqueue(fsc->wb_wq);
633 destroy_workqueue(fsc->pg_inv_wq);
634 destroy_workqueue(fsc->trunc_wq);
635
636 bdi_destroy(&fsc->backing_dev_info);
637
638 mempool_destroy(fsc->wb_pagevec_pool);
639
640 destroy_mount_options(fsc->mount_options);
641
642 ceph_fs_debugfs_cleanup(fsc);
643
644 ceph_destroy_client(fsc->client);
645
646 kfree(fsc);
647 dout("destroy_fs_client %p done\n", fsc);
648}
649
650/*
Sage Weil6e19a162010-04-29 16:38:32 -0700651 * caches
652 */
653struct kmem_cache *ceph_inode_cachep;
654struct kmem_cache *ceph_cap_cachep;
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800655struct kmem_cache *ceph_cap_flush_cachep;
Sage Weil6e19a162010-04-29 16:38:32 -0700656struct kmem_cache *ceph_dentry_cachep;
657struct kmem_cache *ceph_file_cachep;
658
659static void ceph_inode_init_once(void *foo)
660{
661 struct ceph_inode_info *ci = foo;
662 inode_init_once(&ci->vfs_inode);
663}
664
Sage Weil16725b92009-10-06 11:31:07 -0700665static int __init init_caches(void)
666{
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400667 int error = -ENOMEM;
668
Sage Weil16725b92009-10-06 11:31:07 -0700669 ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
670 sizeof(struct ceph_inode_info),
671 __alignof__(struct ceph_inode_info),
Vladimir Davydov5d097052016-01-14 15:18:21 -0800672 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
673 SLAB_ACCOUNT, ceph_inode_init_once);
Sage Weil16725b92009-10-06 11:31:07 -0700674 if (ceph_inode_cachep == NULL)
675 return -ENOMEM;
676
677 ceph_cap_cachep = KMEM_CACHE(ceph_cap,
678 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
679 if (ceph_cap_cachep == NULL)
680 goto bad_cap;
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800681 ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush,
682 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
683 if (ceph_cap_flush_cachep == NULL)
684 goto bad_cap_flush;
Sage Weil16725b92009-10-06 11:31:07 -0700685
686 ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
687 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
688 if (ceph_dentry_cachep == NULL)
689 goto bad_dentry;
690
Nikolay Borisov6b1a9a62016-07-25 20:12:13 +0300691 ceph_file_cachep = KMEM_CACHE(ceph_file_info, SLAB_MEM_SPREAD);
692
Sage Weil16725b92009-10-06 11:31:07 -0700693 if (ceph_file_cachep == NULL)
694 goto bad_file;
695
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400696 if ((error = ceph_fscache_register()))
697 goto bad_file;
Sage Weil16725b92009-10-06 11:31:07 -0700698
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400699 return 0;
Sage Weil16725b92009-10-06 11:31:07 -0700700bad_file:
701 kmem_cache_destroy(ceph_dentry_cachep);
702bad_dentry:
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800703 kmem_cache_destroy(ceph_cap_flush_cachep);
704bad_cap_flush:
Sage Weil16725b92009-10-06 11:31:07 -0700705 kmem_cache_destroy(ceph_cap_cachep);
706bad_cap:
707 kmem_cache_destroy(ceph_inode_cachep);
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400708 return error;
Sage Weil16725b92009-10-06 11:31:07 -0700709}
710
711static void destroy_caches(void)
712{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +1000713 /*
714 * Make sure all delayed rcu free inodes are flushed before we
715 * destroy cache.
716 */
717 rcu_barrier();
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400718
Sage Weil16725b92009-10-06 11:31:07 -0700719 kmem_cache_destroy(ceph_inode_cachep);
720 kmem_cache_destroy(ceph_cap_cachep);
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800721 kmem_cache_destroy(ceph_cap_flush_cachep);
Sage Weil16725b92009-10-06 11:31:07 -0700722 kmem_cache_destroy(ceph_dentry_cachep);
723 kmem_cache_destroy(ceph_file_cachep);
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400724
725 ceph_fscache_unregister();
Sage Weil16725b92009-10-06 11:31:07 -0700726}
727
728
729/*
730 * ceph_umount_begin - initiate forced umount. Tear down down the
731 * mount, skipping steps that may hang while waiting for server(s).
732 */
733static void ceph_umount_begin(struct super_block *sb)
734{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700735 struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
Sage Weil16725b92009-10-06 11:31:07 -0700736
737 dout("ceph_umount_begin - starting forced umount\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700738 if (!fsc)
Sage Weil16725b92009-10-06 11:31:07 -0700739 return;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700740 fsc->mount_state = CEPH_MOUNT_SHUTDOWN;
Yan, Zheng48fec5d2015-07-01 16:27:46 +0800741 ceph_mdsc_force_umount(fsc->mdsc);
Sage Weil16725b92009-10-06 11:31:07 -0700742 return;
743}
744
745static const struct super_operations ceph_super_ops = {
746 .alloc_inode = ceph_alloc_inode,
747 .destroy_inode = ceph_destroy_inode,
748 .write_inode = ceph_write_inode,
Yan, Zheng9f12bd12013-09-20 19:55:31 +0800749 .drop_inode = ceph_drop_inode,
Yan, Zheng9a5530c2016-06-15 16:29:18 +0800750 .evict_inode = ceph_evict_inode,
Sage Weil2d9c98a2010-07-30 09:38:13 -0700751 .sync_fs = ceph_sync_fs,
Sage Weil16725b92009-10-06 11:31:07 -0700752 .put_super = ceph_put_super,
753 .show_options = ceph_show_options,
754 .statfs = ceph_statfs,
755 .umount_begin = ceph_umount_begin,
756};
757
Sage Weil16725b92009-10-06 11:31:07 -0700758/*
759 * Bootstrap mount by opening the root directory. Note the mount
760 * @started time from caller, and time out if this takes too long.
761 */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700762static struct dentry *open_root_dentry(struct ceph_fs_client *fsc,
Sage Weil16725b92009-10-06 11:31:07 -0700763 const char *path,
764 unsigned long started)
765{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700766 struct ceph_mds_client *mdsc = fsc->mdsc;
Sage Weil16725b92009-10-06 11:31:07 -0700767 struct ceph_mds_request *req = NULL;
768 int err;
769 struct dentry *root;
770
771 /* open dir */
772 dout("open_root_inode opening '%s'\n", path);
773 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
774 if (IS_ERR(req))
Julia Lawall7e34bc52010-05-22 12:01:14 +0200775 return ERR_CAST(req);
Sage Weil16725b92009-10-06 11:31:07 -0700776 req->r_path1 = kstrdup(path, GFP_NOFS);
Sanidhya Kashyapa149bb92015-03-21 12:54:58 -0400777 if (!req->r_path1) {
778 root = ERR_PTR(-ENOMEM);
779 goto out;
780 }
781
Sage Weil16725b92009-10-06 11:31:07 -0700782 req->r_ino1.ino = CEPH_INO_ROOT;
783 req->r_ino1.snap = CEPH_NOSNAP;
784 req->r_started = started;
Ilya Dryomova319bf52015-05-15 12:02:17 +0300785 req->r_timeout = fsc->client->options->mount_timeout;
Sage Weil16725b92009-10-06 11:31:07 -0700786 req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
787 req->r_num_caps = 2;
788 err = ceph_mdsc_do_request(mdsc, NULL, req);
789 if (err == 0) {
Al Viro3c5184e2012-01-09 16:34:32 -0500790 struct inode *inode = req->r_target_inode;
791 req->r_target_inode = NULL;
Sage Weil16725b92009-10-06 11:31:07 -0700792 dout("open_root_inode success\n");
Yan, Zhengce2728a2016-09-14 14:53:05 +0800793 root = d_make_root(inode);
794 if (!root) {
795 root = ERR_PTR(-ENOMEM);
796 goto out;
Sage Weil774ac212011-11-11 09:48:08 -0800797 }
Linus Torvalds1a52bb02012-01-13 10:29:21 -0800798 ceph_init_dentry(root);
Sage Weil16725b92009-10-06 11:31:07 -0700799 dout("open_root_inode success, root dentry is %p\n", root);
800 } else {
801 root = ERR_PTR(err);
802 }
Al Viro3c5184e2012-01-09 16:34:32 -0500803out:
Sage Weil16725b92009-10-06 11:31:07 -0700804 ceph_mdsc_put_request(req);
805 return root;
806}
807
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700808
809
810
Sage Weil16725b92009-10-06 11:31:07 -0700811/*
812 * mount: join the ceph cluster, and open root directory.
813 */
Yan, Zheng3f384952016-04-21 11:09:55 +0800814static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc)
Sage Weil16725b92009-10-06 11:31:07 -0700815{
Sage Weil16725b92009-10-06 11:31:07 -0700816 int err;
Sage Weil16725b92009-10-06 11:31:07 -0700817 unsigned long started = jiffies; /* note the start time */
818 struct dentry *root;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700819 int first = 0; /* first vfsmount for this super_block */
Sage Weil16725b92009-10-06 11:31:07 -0700820
Yan, Zheng132ca7e2016-03-12 13:20:48 +0800821 dout("mount start %p\n", fsc);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700822 mutex_lock(&fsc->client->mount_mutex);
Sage Weil16725b92009-10-06 11:31:07 -0700823
Yan, Zheng132ca7e2016-03-12 13:20:48 +0800824 if (!fsc->sb->s_root) {
Yan, Zhengce2728a2016-09-14 14:53:05 +0800825 const char *path;
Yan, Zheng132ca7e2016-03-12 13:20:48 +0800826 err = __ceph_open_session(fsc->client, started);
827 if (err < 0)
828 goto out;
Sage Weil16725b92009-10-06 11:31:07 -0700829
Yan, Zhengce2728a2016-09-14 14:53:05 +0800830 if (!fsc->mount_options->server_path) {
831 path = "";
832 dout("mount opening path \\t\n");
833 } else {
834 path = fsc->mount_options->server_path + 1;
835 dout("mount opening path %s\n", path);
836 }
837 root = open_root_dentry(fsc, path, started);
Yan, Zheng132ca7e2016-03-12 13:20:48 +0800838 if (IS_ERR(root)) {
839 err = PTR_ERR(root);
840 goto out;
841 }
Yan, Zhengce2728a2016-09-14 14:53:05 +0800842 fsc->sb->s_root = dget(root);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700843 first = 1;
844
845 err = ceph_fs_debugfs_init(fsc);
846 if (err < 0)
847 goto fail;
Geert Uytterhoeven31ca5872016-10-13 17:15:37 +0200848 } else {
849 root = dget(fsc->sb->s_root);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700850 }
Sage Weil16725b92009-10-06 11:31:07 -0700851
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700852 fsc->mount_state = CEPH_MOUNT_MOUNTED;
Sage Weil16725b92009-10-06 11:31:07 -0700853 dout("mount success\n");
Al Viroa7f9fb22010-07-26 16:17:55 +0400854 mutex_unlock(&fsc->client->mount_mutex);
855 return root;
Sage Weil16725b92009-10-06 11:31:07 -0700856
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700857fail:
858 if (first) {
859 dput(fsc->sb->s_root);
860 fsc->sb->s_root = NULL;
861 }
Yan, Zheng132ca7e2016-03-12 13:20:48 +0800862out:
863 mutex_unlock(&fsc->client->mount_mutex);
864 return ERR_PTR(err);
Sage Weil16725b92009-10-06 11:31:07 -0700865}
866
867static int ceph_set_super(struct super_block *s, void *data)
868{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700869 struct ceph_fs_client *fsc = data;
Sage Weil16725b92009-10-06 11:31:07 -0700870 int ret;
871
872 dout("set_super %p data %p\n", s, data);
873
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700874 s->s_flags = fsc->mount_options->sb_flags;
Sage Weil16725b92009-10-06 11:31:07 -0700875 s->s_maxbytes = 1ULL << 40; /* temp value until we get mdsmap */
876
Guangliang Zhao7221fe42013-11-11 15:18:03 +0800877 s->s_xattr = ceph_xattr_handlers;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700878 s->s_fs_info = fsc;
879 fsc->sb = s;
Sage Weil16725b92009-10-06 11:31:07 -0700880
881 s->s_op = &ceph_super_ops;
882 s->s_export_op = &ceph_export_ops;
883
884 s->s_time_gran = 1000; /* 1000 ns == 1 us */
885
886 ret = set_anon_super(s, NULL); /* what is that second arg for? */
887 if (ret != 0)
888 goto fail;
889
890 return ret;
891
892fail:
893 s->s_fs_info = NULL;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700894 fsc->sb = NULL;
Sage Weil16725b92009-10-06 11:31:07 -0700895 return ret;
896}
897
898/*
899 * share superblock if same fs AND options
900 */
901static int ceph_compare_super(struct super_block *sb, void *data)
902{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700903 struct ceph_fs_client *new = data;
904 struct ceph_mount_options *fsopt = new->mount_options;
905 struct ceph_options *opt = new->client->options;
906 struct ceph_fs_client *other = ceph_sb_to_client(sb);
Sage Weil16725b92009-10-06 11:31:07 -0700907
908 dout("ceph_compare_super %p\n", sb);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700909
910 if (compare_mount_options(fsopt, opt, other)) {
911 dout("monitor(s)/mount options don't match\n");
912 return 0;
Sage Weil16725b92009-10-06 11:31:07 -0700913 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700914 if ((opt->flags & CEPH_OPT_FSID) &&
915 ceph_fsid_compare(&opt->fsid, &other->client->fsid)) {
916 dout("fsid doesn't match\n");
917 return 0;
918 }
919 if (fsopt->sb_flags != other->mount_options->sb_flags) {
Sage Weil16725b92009-10-06 11:31:07 -0700920 dout("flags differ\n");
921 return 0;
922 }
923 return 1;
924}
925
926/*
927 * construct our own bdi so we can control readahead, etc.
928 */
Jeff Mahoney00d56432010-06-10 11:13:58 -0400929static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
Sage Weil31e0cf82010-05-04 16:39:35 -0700930
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700931static int ceph_register_bdi(struct super_block *sb,
932 struct ceph_fs_client *fsc)
Sage Weil16725b92009-10-06 11:31:07 -0700933{
934 int err;
935
Sage Weil83817e32011-08-04 08:03:44 -0700936 /* set ra_pages based on rasize mount option? */
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300937 if (fsc->mount_options->rasize >= PAGE_SIZE)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700938 fsc->backing_dev_info.ra_pages =
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300939 (fsc->mount_options->rasize + PAGE_SIZE - 1)
Sage Weil16725b92009-10-06 11:31:07 -0700940 >> PAGE_SHIFT;
Yehuda Sadehe9852222011-07-22 11:12:28 -0700941 else
942 fsc->backing_dev_info.ra_pages =
Kirill A. Shutemov09cbfea2016-04-01 15:29:47 +0300943 VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
Yehuda Sadehe9852222011-07-22 11:12:28 -0700944
Joe Perchesd2cc4dd2012-11-29 08:37:03 -0600945 err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%ld",
Sage Weil31e0cf82010-05-04 16:39:35 -0700946 atomic_long_inc_return(&bdi_seq));
Sage Weil5dfc5892010-05-04 16:14:46 -0700947 if (!err)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700948 sb->s_bdi = &fsc->backing_dev_info;
Sage Weil16725b92009-10-06 11:31:07 -0700949 return err;
950}
951
Al Viroa7f9fb22010-07-26 16:17:55 +0400952static struct dentry *ceph_mount(struct file_system_type *fs_type,
953 int flags, const char *dev_name, void *data)
Sage Weil16725b92009-10-06 11:31:07 -0700954{
955 struct super_block *sb;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700956 struct ceph_fs_client *fsc;
Al Viroa7f9fb22010-07-26 16:17:55 +0400957 struct dentry *res;
Sage Weil16725b92009-10-06 11:31:07 -0700958 int err;
959 int (*compare_super)(struct super_block *, void *) = ceph_compare_super;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700960 struct ceph_mount_options *fsopt = NULL;
961 struct ceph_options *opt = NULL;
Sage Weil16725b92009-10-06 11:31:07 -0700962
Al Viroa7f9fb22010-07-26 16:17:55 +0400963 dout("ceph_mount\n");
Sage Weil45195e42014-02-16 10:05:29 -0800964
965#ifdef CONFIG_CEPH_FS_POSIX_ACL
966 flags |= MS_POSIXACL;
967#endif
Yan, Zheng3f384952016-04-21 11:09:55 +0800968 err = parse_mount_options(&fsopt, &opt, flags, data, dev_name);
Al Viroa7f9fb22010-07-26 16:17:55 +0400969 if (err < 0) {
970 res = ERR_PTR(err);
Sage Weil6b805182009-10-27 11:50:50 -0700971 goto out_final;
Al Viroa7f9fb22010-07-26 16:17:55 +0400972 }
Sage Weil16725b92009-10-06 11:31:07 -0700973
974 /* create client (which we may/may not use) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700975 fsc = create_fs_client(fsopt, opt);
976 if (IS_ERR(fsc)) {
Al Viroa7f9fb22010-07-26 16:17:55 +0400977 res = ERR_CAST(fsc);
Noah Watkins259a1872011-08-22 13:49:41 -0600978 destroy_mount_options(fsopt);
979 ceph_destroy_options(opt);
Sage Weil6b805182009-10-27 11:50:50 -0700980 goto out_final;
981 }
Sage Weil16725b92009-10-06 11:31:07 -0700982
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700983 err = ceph_mdsc_init(fsc);
Al Viroa7f9fb22010-07-26 16:17:55 +0400984 if (err < 0) {
985 res = ERR_PTR(err);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700986 goto out;
Al Viroa7f9fb22010-07-26 16:17:55 +0400987 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700988
989 if (ceph_test_opt(fsc->client, NOSHARE))
Sage Weil16725b92009-10-06 11:31:07 -0700990 compare_super = NULL;
David Howells9249e172012-06-25 12:55:37 +0100991 sb = sget(fs_type, compare_super, ceph_set_super, flags, fsc);
Sage Weil16725b92009-10-06 11:31:07 -0700992 if (IS_ERR(sb)) {
Al Viroa7f9fb22010-07-26 16:17:55 +0400993 res = ERR_CAST(sb);
Sage Weil16725b92009-10-06 11:31:07 -0700994 goto out;
995 }
996
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700997 if (ceph_sb_to_client(sb) != fsc) {
998 ceph_mdsc_destroy(fsc);
999 destroy_fs_client(fsc);
1000 fsc = ceph_sb_to_client(sb);
1001 dout("get_sb got existing client %p\n", fsc);
Sage Weil16725b92009-10-06 11:31:07 -07001002 } else {
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001003 dout("get_sb using new client %p\n", fsc);
1004 err = ceph_register_bdi(sb, fsc);
Al Viroa7f9fb22010-07-26 16:17:55 +04001005 if (err < 0) {
1006 res = ERR_PTR(err);
Sage Weil16725b92009-10-06 11:31:07 -07001007 goto out_splat;
Al Viroa7f9fb22010-07-26 16:17:55 +04001008 }
Sage Weil16725b92009-10-06 11:31:07 -07001009 }
1010
Yan, Zheng3f384952016-04-21 11:09:55 +08001011 res = ceph_real_mount(fsc);
Al Viroa7f9fb22010-07-26 16:17:55 +04001012 if (IS_ERR(res))
Sage Weil16725b92009-10-06 11:31:07 -07001013 goto out_splat;
Al Viroa7f9fb22010-07-26 16:17:55 +04001014 dout("root %p inode %p ino %llx.%llx\n", res,
David Howells2b0143b2015-03-17 22:25:59 +00001015 d_inode(res), ceph_vinop(d_inode(res)));
Al Viroa7f9fb22010-07-26 16:17:55 +04001016 return res;
Sage Weil16725b92009-10-06 11:31:07 -07001017
1018out_splat:
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001019 ceph_mdsc_close_sessions(fsc->mdsc);
Al Viro3981f2e2010-03-21 19:22:29 -04001020 deactivate_locked_super(sb);
Sage Weil16725b92009-10-06 11:31:07 -07001021 goto out_final;
1022
1023out:
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001024 ceph_mdsc_destroy(fsc);
1025 destroy_fs_client(fsc);
Sage Weil16725b92009-10-06 11:31:07 -07001026out_final:
Al Viroa7f9fb22010-07-26 16:17:55 +04001027 dout("ceph_mount fail %ld\n", PTR_ERR(res));
1028 return res;
Sage Weil16725b92009-10-06 11:31:07 -07001029}
1030
1031static void ceph_kill_sb(struct super_block *s)
1032{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001033 struct ceph_fs_client *fsc = ceph_sb_to_client(s);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001034 dev_t dev = s->s_dev;
1035
Sage Weil16725b92009-10-06 11:31:07 -07001036 dout("kill_sb %p\n", s);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001037
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001038 ceph_mdsc_pre_umount(fsc->mdsc);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001039 generic_shutdown_super(s);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001040 ceph_mdsc_destroy(fsc);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001041
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001042 destroy_fs_client(fsc);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001043 free_anon_bdev(dev);
Sage Weil16725b92009-10-06 11:31:07 -07001044}
1045
1046static struct file_system_type ceph_fs_type = {
1047 .owner = THIS_MODULE,
1048 .name = "ceph",
Al Viroa7f9fb22010-07-26 16:17:55 +04001049 .mount = ceph_mount,
Sage Weil16725b92009-10-06 11:31:07 -07001050 .kill_sb = ceph_kill_sb,
1051 .fs_flags = FS_RENAME_DOES_D_MOVE,
1052};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08001053MODULE_ALIAS_FS("ceph");
Sage Weil16725b92009-10-06 11:31:07 -07001054
Sage Weil16725b92009-10-06 11:31:07 -07001055static int __init init_ceph(void)
1056{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001057 int ret = init_caches();
Sage Weil16725b92009-10-06 11:31:07 -07001058 if (ret)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001059 goto out;
Sage Weil16725b92009-10-06 11:31:07 -07001060
Yan, Zhengeb13e832014-03-09 23:16:40 +08001061 ceph_flock_init();
Alex Elder3ce6cd12012-01-23 15:49:28 -06001062 ceph_xattr_init();
Sage Weil16725b92009-10-06 11:31:07 -07001063 ret = register_filesystem(&ceph_fs_type);
1064 if (ret)
Ilya Dryomov34b759b2016-02-16 15:00:24 +01001065 goto out_xattr;
Sage Weil16725b92009-10-06 11:31:07 -07001066
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001067 pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
1068
Sage Weil16725b92009-10-06 11:31:07 -07001069 return 0;
1070
Yan, Zheng97c85a82014-11-06 15:09:41 +08001071out_xattr:
Alex Elder3ce6cd12012-01-23 15:49:28 -06001072 ceph_xattr_exit();
Sage Weil16725b92009-10-06 11:31:07 -07001073 destroy_caches();
Sage Weil16725b92009-10-06 11:31:07 -07001074out:
1075 return ret;
1076}
1077
1078static void __exit exit_ceph(void)
1079{
1080 dout("exit_ceph\n");
1081 unregister_filesystem(&ceph_fs_type);
Alex Elder3ce6cd12012-01-23 15:49:28 -06001082 ceph_xattr_exit();
Sage Weil16725b92009-10-06 11:31:07 -07001083 destroy_caches();
Sage Weil16725b92009-10-06 11:31:07 -07001084}
1085
1086module_init(init_ceph);
1087module_exit(exit_ceph);
1088
1089MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
1090MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
1091MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
1092MODULE_DESCRIPTION("Ceph filesystem for Linux");
1093MODULE_LICENSE("GPL");