blob: f446afada328a45c2b70f648cbb48261e11e996c [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,
123 Opt_last_string,
124 /* string args above */
125 Opt_dirstat,
126 Opt_nodirstat,
127 Opt_rbytes,
128 Opt_norbytes,
Alex Eldercffaba12012-02-15 07:43:54 -0600129 Opt_asyncreaddir,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700130 Opt_noasyncreaddir,
Sage Weila40dc6c2012-01-10 09:12:55 -0800131 Opt_dcache,
132 Opt_nodcache,
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800133 Opt_ino32,
Alex Eldercffaba12012-02-15 07:43:54 -0600134 Opt_noino32,
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400135 Opt_fscache,
Sage Weil45195e42014-02-16 10:05:29 -0800136 Opt_nofscache,
Yan, Zheng10183a62015-04-27 15:33:28 +0800137 Opt_poolperm,
138 Opt_nopoolperm,
Sage Weil45195e42014-02-16 10:05:29 -0800139#ifdef CONFIG_CEPH_FS_POSIX_ACL
140 Opt_acl,
141#endif
Yan, Zheng10183a62015-04-27 15:33:28 +0800142 Opt_noacl,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700143};
144
145static match_table_t fsopt_tokens = {
146 {Opt_wsize, "wsize=%d"},
147 {Opt_rsize, "rsize=%d"},
Sage Weil83817e32011-08-04 08:03:44 -0700148 {Opt_rasize, "rasize=%d"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700149 {Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
150 {Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"},
151 {Opt_cap_release_safety, "cap_release_safety=%d"},
152 {Opt_readdir_max_entries, "readdir_max_entries=%d"},
153 {Opt_readdir_max_bytes, "readdir_max_bytes=%d"},
154 {Opt_congestion_kb, "write_congestion_kb=%d"},
155 /* int args above */
156 {Opt_snapdirname, "snapdirname=%s"},
157 /* string args above */
158 {Opt_dirstat, "dirstat"},
159 {Opt_nodirstat, "nodirstat"},
160 {Opt_rbytes, "rbytes"},
161 {Opt_norbytes, "norbytes"},
Alex Eldercffaba12012-02-15 07:43:54 -0600162 {Opt_asyncreaddir, "asyncreaddir"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700163 {Opt_noasyncreaddir, "noasyncreaddir"},
Sage Weila40dc6c2012-01-10 09:12:55 -0800164 {Opt_dcache, "dcache"},
165 {Opt_nodcache, "nodcache"},
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800166 {Opt_ino32, "ino32"},
Alex Eldercffaba12012-02-15 07:43:54 -0600167 {Opt_noino32, "noino32"},
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400168 {Opt_fscache, "fsc"},
169 {Opt_nofscache, "nofsc"},
Yan, Zheng10183a62015-04-27 15:33:28 +0800170 {Opt_poolperm, "poolperm"},
171 {Opt_nopoolperm, "nopoolperm"},
Sage Weil45195e42014-02-16 10:05:29 -0800172#ifdef CONFIG_CEPH_FS_POSIX_ACL
173 {Opt_acl, "acl"},
174#endif
175 {Opt_noacl, "noacl"},
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700176 {-1, NULL}
177};
178
179static int parse_fsopt_token(char *c, void *private)
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800180{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700181 struct ceph_mount_options *fsopt = private;
182 substring_t argstr[MAX_OPT_ARGS];
183 int token, intval, ret;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800184
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700185 token = match_token((char *)c, fsopt_tokens, argstr);
186 if (token < 0)
187 return -EINVAL;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800188
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700189 if (token < Opt_last_int) {
190 ret = match_int(&argstr[0], &intval);
191 if (ret < 0) {
192 pr_err("bad mount option arg (not int) "
193 "at '%s'\n", c);
194 return ret;
195 }
196 dout("got int token %d val %d\n", token, intval);
197 } else if (token > Opt_last_int && token < Opt_last_string) {
198 dout("got string token %d val %s\n", token,
199 argstr[0].from);
200 } else {
201 dout("got token %d\n", token);
202 }
203
204 switch (token) {
205 case Opt_snapdirname:
206 kfree(fsopt->snapdir_name);
207 fsopt->snapdir_name = kstrndup(argstr[0].from,
208 argstr[0].to-argstr[0].from,
209 GFP_KERNEL);
210 if (!fsopt->snapdir_name)
211 return -ENOMEM;
212 break;
213
214 /* misc */
215 case Opt_wsize:
216 fsopt->wsize = intval;
217 break;
218 case Opt_rsize:
219 fsopt->rsize = intval;
220 break;
Sage Weil83817e32011-08-04 08:03:44 -0700221 case Opt_rasize:
222 fsopt->rasize = intval;
223 break;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700224 case Opt_caps_wanted_delay_min:
225 fsopt->caps_wanted_delay_min = intval;
226 break;
227 case Opt_caps_wanted_delay_max:
228 fsopt->caps_wanted_delay_max = intval;
229 break;
230 case Opt_readdir_max_entries:
231 fsopt->max_readdir = intval;
232 break;
233 case Opt_readdir_max_bytes:
234 fsopt->max_readdir_bytes = intval;
235 break;
236 case Opt_congestion_kb:
237 fsopt->congestion_kb = intval;
238 break;
239 case Opt_dirstat:
240 fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT;
241 break;
242 case Opt_nodirstat:
243 fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT;
244 break;
245 case Opt_rbytes:
246 fsopt->flags |= CEPH_MOUNT_OPT_RBYTES;
247 break;
248 case Opt_norbytes:
249 fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES;
250 break;
Alex Eldercffaba12012-02-15 07:43:54 -0600251 case Opt_asyncreaddir:
252 fsopt->flags &= ~CEPH_MOUNT_OPT_NOASYNCREADDIR;
253 break;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700254 case Opt_noasyncreaddir:
255 fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
256 break;
Sage Weila40dc6c2012-01-10 09:12:55 -0800257 case Opt_dcache:
258 fsopt->flags |= CEPH_MOUNT_OPT_DCACHE;
259 break;
260 case Opt_nodcache:
261 fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
262 break;
Yehuda Sadehad1fee92011-01-21 16:44:03 -0800263 case Opt_ino32:
264 fsopt->flags |= CEPH_MOUNT_OPT_INO32;
265 break;
Alex Eldercffaba12012-02-15 07:43:54 -0600266 case Opt_noino32:
267 fsopt->flags &= ~CEPH_MOUNT_OPT_INO32;
268 break;
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400269 case Opt_fscache:
270 fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE;
271 break;
272 case Opt_nofscache:
273 fsopt->flags &= ~CEPH_MOUNT_OPT_FSCACHE;
274 break;
Yan, Zheng10183a62015-04-27 15:33:28 +0800275 case Opt_poolperm:
276 fsopt->flags &= ~CEPH_MOUNT_OPT_NOPOOLPERM;
277 printk ("pool perm");
278 break;
279 case Opt_nopoolperm:
280 fsopt->flags |= CEPH_MOUNT_OPT_NOPOOLPERM;
281 break;
Sage Weil45195e42014-02-16 10:05:29 -0800282#ifdef CONFIG_CEPH_FS_POSIX_ACL
283 case Opt_acl:
284 fsopt->sb_flags |= MS_POSIXACL;
285 break;
286#endif
287 case Opt_noacl:
288 fsopt->sb_flags &= ~MS_POSIXACL;
289 break;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700290 default:
291 BUG_ON(token);
292 }
293 return 0;
294}
295
296static void destroy_mount_options(struct ceph_mount_options *args)
297{
298 dout("destroy_mount_options %p\n", args);
299 kfree(args->snapdir_name);
300 kfree(args);
301}
302
303static int strcmp_null(const char *s1, const char *s2)
304{
305 if (!s1 && !s2)
306 return 0;
307 if (s1 && !s2)
308 return -1;
309 if (!s1 && s2)
310 return 1;
311 return strcmp(s1, s2);
312}
313
314static int compare_mount_options(struct ceph_mount_options *new_fsopt,
315 struct ceph_options *new_opt,
316 struct ceph_fs_client *fsc)
317{
318 struct ceph_mount_options *fsopt1 = new_fsopt;
319 struct ceph_mount_options *fsopt2 = fsc->mount_options;
320 int ofs = offsetof(struct ceph_mount_options, snapdir_name);
321 int ret;
322
323 ret = memcmp(fsopt1, fsopt2, ofs);
324 if (ret)
325 return ret;
326
327 ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name);
328 if (ret)
329 return ret;
330
331 return ceph_compare_options(new_opt, fsc->client);
332}
333
334static int parse_mount_options(struct ceph_mount_options **pfsopt,
335 struct ceph_options **popt,
336 int flags, char *options,
337 const char *dev_name,
338 const char **path)
339{
340 struct ceph_mount_options *fsopt;
341 const char *dev_name_end;
Alex Elderc98f5332012-08-09 10:33:26 -0700342 int err;
343
344 if (!dev_name || !*dev_name)
345 return -EINVAL;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700346
347 fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL);
348 if (!fsopt)
349 return -ENOMEM;
350
351 dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name);
352
Noah Watkins80db8be2011-08-22 13:49:23 -0600353 fsopt->sb_flags = flags;
354 fsopt->flags = CEPH_MOUNT_OPT_DEFAULT;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700355
Noah Watkins80db8be2011-08-22 13:49:23 -0600356 fsopt->rsize = CEPH_RSIZE_DEFAULT;
357 fsopt->rasize = CEPH_RASIZE_DEFAULT;
358 fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
Sanidhya Kashyapa149bb92015-03-21 12:54:58 -0400359 if (!fsopt->snapdir_name) {
360 err = -ENOMEM;
361 goto out;
362 }
363
Sage Weil50aac4f2011-01-18 07:59:40 -0800364 fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
365 fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
Noah Watkins80db8be2011-08-22 13:49:23 -0600366 fsopt->cap_release_safety = CEPH_CAP_RELEASE_SAFETY_DEFAULT;
367 fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT;
368 fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
369 fsopt->congestion_kb = default_congestion_kb();
370
Alex Elderc98f5332012-08-09 10:33:26 -0700371 /*
372 * Distinguish the server list from the path in "dev_name".
373 * Internally we do not include the leading '/' in the path.
374 *
375 * "dev_name" will look like:
376 * <server_spec>[,<server_spec>...]:[<path>]
377 * where
378 * <server_spec> is <ip>[:<port>]
379 * <path> is optional, but if present must begin with '/'
380 */
381 dev_name_end = strchr(dev_name, '/');
382 if (dev_name_end) {
383 /* skip over leading '/' for path */
384 *path = dev_name_end + 1;
385 } else {
386 /* path is empty */
387 dev_name_end = dev_name + strlen(dev_name);
388 *path = dev_name_end;
389 }
Noah Watkins80db8be2011-08-22 13:49:23 -0600390 err = -EINVAL;
Alex Elderc98f5332012-08-09 10:33:26 -0700391 dev_name_end--; /* back up to ':' separator */
Sasha Levin54464292013-07-01 18:33:39 -0400392 if (dev_name_end < dev_name || *dev_name_end != ':') {
Alex Elderc98f5332012-08-09 10:33:26 -0700393 pr_err("device name is missing path (no : separator in %s)\n",
Noah Watkins80db8be2011-08-22 13:49:23 -0600394 dev_name);
395 goto out;
396 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700397 dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700398 dout("server path '%s'\n", *path);
399
Alex Elderee577412012-01-24 10:08:36 -0600400 *popt = ceph_parse_options(options, dev_name, dev_name_end,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700401 parse_fsopt_token, (void *)fsopt);
Alex Elderee577412012-01-24 10:08:36 -0600402 if (IS_ERR(*popt)) {
403 err = PTR_ERR(*popt);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700404 goto out;
Alex Elderee577412012-01-24 10:08:36 -0600405 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700406
407 /* success */
408 *pfsopt = fsopt;
409 return 0;
410
411out:
412 destroy_mount_options(fsopt);
413 return err;
Yehuda Sadeh2baba252009-12-18 13:51:57 -0800414}
415
Sage Weil6e19a162010-04-29 16:38:32 -0700416/**
417 * ceph_show_options - Show mount options in /proc/mounts
418 * @m: seq_file to write to
Al Viro34c80b12011-12-08 21:32:45 -0500419 * @root: root of that (sub)tree
Sage Weil6e19a162010-04-29 16:38:32 -0700420 */
Al Viro34c80b12011-12-08 21:32:45 -0500421static int ceph_show_options(struct seq_file *m, struct dentry *root)
Sage Weil6e19a162010-04-29 16:38:32 -0700422{
Al Viro34c80b12011-12-08 21:32:45 -0500423 struct ceph_fs_client *fsc = ceph_sb_to_client(root->d_sb);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700424 struct ceph_mount_options *fsopt = fsc->mount_options;
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300425 size_t pos;
426 int ret;
Sage Weil6e19a162010-04-29 16:38:32 -0700427
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300428 /* a comma between MNT/MS and client options */
429 seq_putc(m, ',');
430 pos = m->count;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700431
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300432 ret = ceph_print_client_options(m, fsc->client);
433 if (ret)
434 return ret;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700435
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300436 /* retract our comma if no client options */
437 if (m->count == pos)
438 m->count--;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700439
440 if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT)
441 seq_puts(m, ",dirstat");
442 if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES) == 0)
443 seq_puts(m, ",norbytes");
444 if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
Sage Weil6e19a162010-04-29 16:38:32 -0700445 seq_puts(m, ",noasyncreaddir");
Ilya Dryomovff7eeb82015-03-25 21:10:09 +0300446 if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
Sage Weila40dc6c2012-01-10 09:12:55 -0800447 seq_puts(m, ",nodcache");
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400448 if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE)
449 seq_puts(m, ",fsc");
Yan, Zheng10183a62015-04-27 15:33:28 +0800450 if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
451 seq_puts(m, ",nopoolperm");
Sage Weil6e19a162010-04-29 16:38:32 -0700452
Sage Weil45195e42014-02-16 10:05:29 -0800453#ifdef CONFIG_CEPH_FS_POSIX_ACL
454 if (fsopt->sb_flags & MS_POSIXACL)
455 seq_puts(m, ",acl");
456 else
457 seq_puts(m, ",noacl");
458#endif
459
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700460 if (fsopt->wsize)
461 seq_printf(m, ",wsize=%d", fsopt->wsize);
Sage Weil80456f82011-03-10 13:33:26 -0800462 if (fsopt->rsize != CEPH_RSIZE_DEFAULT)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700463 seq_printf(m, ",rsize=%d", fsopt->rsize);
Sage Weil83817e32011-08-04 08:03:44 -0700464 if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
Sage Weil21519372011-12-01 08:06:52 -0800465 seq_printf(m, ",rasize=%d", fsopt->rasize);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700466 if (fsopt->congestion_kb != default_congestion_kb())
467 seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb);
468 if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
Sage Weil6e19a162010-04-29 16:38:32 -0700469 seq_printf(m, ",caps_wanted_delay_min=%d",
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700470 fsopt->caps_wanted_delay_min);
471 if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
Sage Weil6e19a162010-04-29 16:38:32 -0700472 seq_printf(m, ",caps_wanted_delay_max=%d",
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700473 fsopt->caps_wanted_delay_max);
474 if (fsopt->cap_release_safety != CEPH_CAP_RELEASE_SAFETY_DEFAULT)
Sage Weil6e19a162010-04-29 16:38:32 -0700475 seq_printf(m, ",cap_release_safety=%d",
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700476 fsopt->cap_release_safety);
477 if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT)
478 seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir);
479 if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
480 seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes);
481 if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
Kees Cooka068acf2015-09-04 15:44:57 -0700482 seq_show_option(m, "snapdirname", fsopt->snapdir_name);
Ilya Dryomovff40f9a2015-03-25 21:02:16 +0300483
Sage Weil6e19a162010-04-29 16:38:32 -0700484 return 0;
485}
486
487/*
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700488 * handle any mon messages the standard library doesn't understand.
489 * return error if we don't either.
490 */
491static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg)
492{
493 struct ceph_fs_client *fsc = client->private;
494 int type = le16_to_cpu(msg->hdr.type);
495
496 switch (type) {
497 case CEPH_MSG_MDS_MAP:
498 ceph_mdsc_handle_map(fsc->mdsc, msg);
499 return 0;
500
501 default:
502 return -1;
503 }
504}
505
506/*
507 * create a new fs client
508 */
H Hartley Sweeten0c6d4b4e22011-09-23 11:53:30 -0700509static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700510 struct ceph_options *opt)
511{
512 struct ceph_fs_client *fsc;
Ilya Dryomov12b46292013-12-24 21:19:23 +0200513 const u64 supported_features =
Sage Weil6ab00d42011-08-09 09:41:59 -0700514 CEPH_FEATURE_FLOCK |
Yan, Zheng65a22662014-11-17 10:01:03 +0800515 CEPH_FEATURE_DIRLAYOUTHASH |
516 CEPH_FEATURE_MDS_INLINE_DATA;
Ilya Dryomov12b46292013-12-24 21:19:23 +0200517 const u64 required_features = 0;
Alex Elder3bf53332013-04-01 10:48:40 -0500518 int page_count;
519 size_t size;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700520 int err = -ENOMEM;
521
522 fsc = kzalloc(sizeof(*fsc), GFP_KERNEL);
523 if (!fsc)
524 return ERR_PTR(-ENOMEM);
525
Sage Weil6ab00d42011-08-09 09:41:59 -0700526 fsc->client = ceph_create_client(opt, fsc, supported_features,
527 required_features);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700528 if (IS_ERR(fsc->client)) {
529 err = PTR_ERR(fsc->client);
530 goto fail;
531 }
532 fsc->client->extra_mon_dispatch = extra_mon_dispatch;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700533 fsc->client->monc.want_mdsmap = 1;
534
535 fsc->mount_options = fsopt;
536
537 fsc->sb = NULL;
538 fsc->mount_state = CEPH_MOUNT_MOUNTING;
539
540 atomic_long_set(&fsc->writeback_count, 0);
541
542 err = bdi_init(&fsc->backing_dev_info);
543 if (err < 0)
544 goto fail_client;
545
546 err = -ENOMEM;
Tejun Heo01e6acc2011-01-03 14:49:45 +0100547 /*
548 * The number of concurrent works can be high but they don't need
549 * to be processed in parallel, limit concurrency.
550 */
551 fsc->wb_wq = alloc_workqueue("ceph-writeback", 0, 1);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700552 if (fsc->wb_wq == NULL)
553 goto fail_bdi;
Tejun Heo01e6acc2011-01-03 14:49:45 +0100554 fsc->pg_inv_wq = alloc_workqueue("ceph-pg-invalid", 0, 1);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700555 if (fsc->pg_inv_wq == NULL)
556 goto fail_wb_wq;
Tejun Heo01e6acc2011-01-03 14:49:45 +0100557 fsc->trunc_wq = alloc_workqueue("ceph-trunc", 0, 1);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700558 if (fsc->trunc_wq == NULL)
559 goto fail_pg_inv_wq;
560
561 /* set up mempools */
562 err = -ENOMEM;
Alex Elder3bf53332013-04-01 10:48:40 -0500563 page_count = fsc->mount_options->wsize >> PAGE_CACHE_SHIFT;
564 size = sizeof (struct page *) * (page_count ? page_count : 1);
565 fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10, size);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700566 if (!fsc->wb_pagevec_pool)
567 goto fail_trunc_wq;
568
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400569 /* setup fscache */
570 if ((fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) &&
571 (ceph_fscache_register_fs(fsc) != 0))
572 goto fail_fscache;
573
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700574 /* caps */
575 fsc->min_caps = fsopt->max_readdir;
576
577 return fsc;
578
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400579fail_fscache:
580 ceph_fscache_unregister_fs(fsc);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700581fail_trunc_wq:
582 destroy_workqueue(fsc->trunc_wq);
583fail_pg_inv_wq:
584 destroy_workqueue(fsc->pg_inv_wq);
585fail_wb_wq:
586 destroy_workqueue(fsc->wb_wq);
587fail_bdi:
588 bdi_destroy(&fsc->backing_dev_info);
589fail_client:
590 ceph_destroy_client(fsc->client);
591fail:
592 kfree(fsc);
593 return ERR_PTR(err);
594}
595
H Hartley Sweeten0c6d4b4e22011-09-23 11:53:30 -0700596static void destroy_fs_client(struct ceph_fs_client *fsc)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700597{
598 dout("destroy_fs_client %p\n", fsc);
599
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400600 ceph_fscache_unregister_fs(fsc);
601
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700602 destroy_workqueue(fsc->wb_wq);
603 destroy_workqueue(fsc->pg_inv_wq);
604 destroy_workqueue(fsc->trunc_wq);
605
606 bdi_destroy(&fsc->backing_dev_info);
607
608 mempool_destroy(fsc->wb_pagevec_pool);
609
610 destroy_mount_options(fsc->mount_options);
611
612 ceph_fs_debugfs_cleanup(fsc);
613
614 ceph_destroy_client(fsc->client);
615
616 kfree(fsc);
617 dout("destroy_fs_client %p done\n", fsc);
618}
619
620/*
Sage Weil6e19a162010-04-29 16:38:32 -0700621 * caches
622 */
623struct kmem_cache *ceph_inode_cachep;
624struct kmem_cache *ceph_cap_cachep;
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800625struct kmem_cache *ceph_cap_flush_cachep;
Sage Weil6e19a162010-04-29 16:38:32 -0700626struct kmem_cache *ceph_dentry_cachep;
627struct kmem_cache *ceph_file_cachep;
628
629static void ceph_inode_init_once(void *foo)
630{
631 struct ceph_inode_info *ci = foo;
632 inode_init_once(&ci->vfs_inode);
633}
634
Sage Weil16725b92009-10-06 11:31:07 -0700635static int __init init_caches(void)
636{
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400637 int error = -ENOMEM;
638
Sage Weil16725b92009-10-06 11:31:07 -0700639 ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
640 sizeof(struct ceph_inode_info),
641 __alignof__(struct ceph_inode_info),
642 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
643 ceph_inode_init_once);
644 if (ceph_inode_cachep == NULL)
645 return -ENOMEM;
646
647 ceph_cap_cachep = KMEM_CACHE(ceph_cap,
648 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
649 if (ceph_cap_cachep == NULL)
650 goto bad_cap;
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800651 ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush,
652 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
653 if (ceph_cap_flush_cachep == NULL)
654 goto bad_cap_flush;
Sage Weil16725b92009-10-06 11:31:07 -0700655
656 ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
657 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
658 if (ceph_dentry_cachep == NULL)
659 goto bad_dentry;
660
661 ceph_file_cachep = KMEM_CACHE(ceph_file_info,
662 SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
663 if (ceph_file_cachep == NULL)
664 goto bad_file;
665
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400666 if ((error = ceph_fscache_register()))
667 goto bad_file;
Sage Weil16725b92009-10-06 11:31:07 -0700668
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400669 return 0;
Sage Weil16725b92009-10-06 11:31:07 -0700670bad_file:
671 kmem_cache_destroy(ceph_dentry_cachep);
672bad_dentry:
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800673 kmem_cache_destroy(ceph_cap_flush_cachep);
674bad_cap_flush:
Sage Weil16725b92009-10-06 11:31:07 -0700675 kmem_cache_destroy(ceph_cap_cachep);
676bad_cap:
677 kmem_cache_destroy(ceph_inode_cachep);
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400678 return error;
Sage Weil16725b92009-10-06 11:31:07 -0700679}
680
681static void destroy_caches(void)
682{
Kirill A. Shutemov8c0a8532012-09-26 11:33:07 +1000683 /*
684 * Make sure all delayed rcu free inodes are flushed before we
685 * destroy cache.
686 */
687 rcu_barrier();
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400688
Sage Weil16725b92009-10-06 11:31:07 -0700689 kmem_cache_destroy(ceph_inode_cachep);
690 kmem_cache_destroy(ceph_cap_cachep);
Yan, Zhengf66fd9f2015-06-10 17:26:13 +0800691 kmem_cache_destroy(ceph_cap_flush_cachep);
Sage Weil16725b92009-10-06 11:31:07 -0700692 kmem_cache_destroy(ceph_dentry_cachep);
693 kmem_cache_destroy(ceph_file_cachep);
Milosz Tanski99ccbd22013-08-21 17:29:54 -0400694
695 ceph_fscache_unregister();
Sage Weil16725b92009-10-06 11:31:07 -0700696}
697
698
699/*
700 * ceph_umount_begin - initiate forced umount. Tear down down the
701 * mount, skipping steps that may hang while waiting for server(s).
702 */
703static void ceph_umount_begin(struct super_block *sb)
704{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700705 struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
Sage Weil16725b92009-10-06 11:31:07 -0700706
707 dout("ceph_umount_begin - starting forced umount\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700708 if (!fsc)
Sage Weil16725b92009-10-06 11:31:07 -0700709 return;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700710 fsc->mount_state = CEPH_MOUNT_SHUTDOWN;
Yan, Zheng48fec5d2015-07-01 16:27:46 +0800711 ceph_mdsc_force_umount(fsc->mdsc);
Sage Weil16725b92009-10-06 11:31:07 -0700712 return;
713}
714
715static const struct super_operations ceph_super_ops = {
716 .alloc_inode = ceph_alloc_inode,
717 .destroy_inode = ceph_destroy_inode,
718 .write_inode = ceph_write_inode,
Yan, Zheng9f12bd12013-09-20 19:55:31 +0800719 .drop_inode = ceph_drop_inode,
Sage Weil2d9c98a2010-07-30 09:38:13 -0700720 .sync_fs = ceph_sync_fs,
Sage Weil16725b92009-10-06 11:31:07 -0700721 .put_super = ceph_put_super,
722 .show_options = ceph_show_options,
723 .statfs = ceph_statfs,
724 .umount_begin = ceph_umount_begin,
725};
726
Sage Weil16725b92009-10-06 11:31:07 -0700727/*
728 * Bootstrap mount by opening the root directory. Note the mount
729 * @started time from caller, and time out if this takes too long.
730 */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700731static struct dentry *open_root_dentry(struct ceph_fs_client *fsc,
Sage Weil16725b92009-10-06 11:31:07 -0700732 const char *path,
733 unsigned long started)
734{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700735 struct ceph_mds_client *mdsc = fsc->mdsc;
Sage Weil16725b92009-10-06 11:31:07 -0700736 struct ceph_mds_request *req = NULL;
737 int err;
738 struct dentry *root;
739
740 /* open dir */
741 dout("open_root_inode opening '%s'\n", path);
742 req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
743 if (IS_ERR(req))
Julia Lawall7e34bc52010-05-22 12:01:14 +0200744 return ERR_CAST(req);
Sage Weil16725b92009-10-06 11:31:07 -0700745 req->r_path1 = kstrdup(path, GFP_NOFS);
Sanidhya Kashyapa149bb92015-03-21 12:54:58 -0400746 if (!req->r_path1) {
747 root = ERR_PTR(-ENOMEM);
748 goto out;
749 }
750
Sage Weil16725b92009-10-06 11:31:07 -0700751 req->r_ino1.ino = CEPH_INO_ROOT;
752 req->r_ino1.snap = CEPH_NOSNAP;
753 req->r_started = started;
Ilya Dryomova319bf52015-05-15 12:02:17 +0300754 req->r_timeout = fsc->client->options->mount_timeout;
Sage Weil16725b92009-10-06 11:31:07 -0700755 req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
756 req->r_num_caps = 2;
757 err = ceph_mdsc_do_request(mdsc, NULL, req);
758 if (err == 0) {
Al Viro3c5184e2012-01-09 16:34:32 -0500759 struct inode *inode = req->r_target_inode;
760 req->r_target_inode = NULL;
Sage Weil16725b92009-10-06 11:31:07 -0700761 dout("open_root_inode success\n");
Al Viro3c5184e2012-01-09 16:34:32 -0500762 if (ceph_ino(inode) == CEPH_INO_ROOT &&
Sage Weil774ac212011-11-11 09:48:08 -0800763 fsc->sb->s_root == NULL) {
Al Viro48fde702012-01-08 22:15:13 -0500764 root = d_make_root(inode);
Al Viro3c5184e2012-01-09 16:34:32 -0500765 if (!root) {
Al Viro3c5184e2012-01-09 16:34:32 -0500766 root = ERR_PTR(-ENOMEM);
767 goto out;
768 }
Sage Weil774ac212011-11-11 09:48:08 -0800769 } else {
J. Bruce Fields1a0a3972014-02-14 17:35:37 -0500770 root = d_obtain_root(inode);
Sage Weil774ac212011-11-11 09:48:08 -0800771 }
Linus Torvalds1a52bb02012-01-13 10:29:21 -0800772 ceph_init_dentry(root);
Sage Weil16725b92009-10-06 11:31:07 -0700773 dout("open_root_inode success, root dentry is %p\n", root);
774 } else {
775 root = ERR_PTR(err);
776 }
Al Viro3c5184e2012-01-09 16:34:32 -0500777out:
Sage Weil16725b92009-10-06 11:31:07 -0700778 ceph_mdsc_put_request(req);
779 return root;
780}
781
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700782
783
784
Sage Weil16725b92009-10-06 11:31:07 -0700785/*
786 * mount: join the ceph cluster, and open root directory.
787 */
Al Viroa7f9fb22010-07-26 16:17:55 +0400788static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc,
Sage Weil16725b92009-10-06 11:31:07 -0700789 const char *path)
790{
Sage Weil16725b92009-10-06 11:31:07 -0700791 int err;
Sage Weil16725b92009-10-06 11:31:07 -0700792 unsigned long started = jiffies; /* note the start time */
793 struct dentry *root;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700794 int first = 0; /* first vfsmount for this super_block */
Sage Weil16725b92009-10-06 11:31:07 -0700795
796 dout("mount start\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700797 mutex_lock(&fsc->client->mount_mutex);
Sage Weil16725b92009-10-06 11:31:07 -0700798
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700799 err = __ceph_open_session(fsc->client, started);
Sage Weil16725b92009-10-06 11:31:07 -0700800 if (err < 0)
801 goto out;
802
Sage Weil16725b92009-10-06 11:31:07 -0700803 dout("mount opening root\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700804 root = open_root_dentry(fsc, "", started);
Sage Weil16725b92009-10-06 11:31:07 -0700805 if (IS_ERR(root)) {
806 err = PTR_ERR(root);
807 goto out;
808 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700809 if (fsc->sb->s_root) {
Sage Weil16725b92009-10-06 11:31:07 -0700810 dput(root);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700811 } else {
812 fsc->sb->s_root = root;
813 first = 1;
814
815 err = ceph_fs_debugfs_init(fsc);
816 if (err < 0)
817 goto fail;
818 }
Sage Weil16725b92009-10-06 11:31:07 -0700819
820 if (path[0] == 0) {
821 dget(root);
822 } else {
823 dout("mount opening base mountpoint\n");
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700824 root = open_root_dentry(fsc, path, started);
Sage Weil16725b92009-10-06 11:31:07 -0700825 if (IS_ERR(root)) {
826 err = PTR_ERR(root);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700827 goto fail;
Sage Weil16725b92009-10-06 11:31:07 -0700828 }
829 }
830
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700831 fsc->mount_state = CEPH_MOUNT_MOUNTED;
Sage Weil16725b92009-10-06 11:31:07 -0700832 dout("mount success\n");
Al Viroa7f9fb22010-07-26 16:17:55 +0400833 mutex_unlock(&fsc->client->mount_mutex);
834 return root;
Sage Weil16725b92009-10-06 11:31:07 -0700835
836out:
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700837 mutex_unlock(&fsc->client->mount_mutex);
Al Viroa7f9fb22010-07-26 16:17:55 +0400838 return ERR_PTR(err);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700839
840fail:
841 if (first) {
842 dput(fsc->sb->s_root);
843 fsc->sb->s_root = NULL;
844 }
845 goto out;
Sage Weil16725b92009-10-06 11:31:07 -0700846}
847
848static int ceph_set_super(struct super_block *s, void *data)
849{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700850 struct ceph_fs_client *fsc = data;
Sage Weil16725b92009-10-06 11:31:07 -0700851 int ret;
852
853 dout("set_super %p data %p\n", s, data);
854
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700855 s->s_flags = fsc->mount_options->sb_flags;
Sage Weil16725b92009-10-06 11:31:07 -0700856 s->s_maxbytes = 1ULL << 40; /* temp value until we get mdsmap */
857
Guangliang Zhao7221fe42013-11-11 15:18:03 +0800858 s->s_xattr = ceph_xattr_handlers;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700859 s->s_fs_info = fsc;
860 fsc->sb = s;
Sage Weil16725b92009-10-06 11:31:07 -0700861
862 s->s_op = &ceph_super_ops;
863 s->s_export_op = &ceph_export_ops;
864
865 s->s_time_gran = 1000; /* 1000 ns == 1 us */
866
867 ret = set_anon_super(s, NULL); /* what is that second arg for? */
868 if (ret != 0)
869 goto fail;
870
871 return ret;
872
873fail:
874 s->s_fs_info = NULL;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700875 fsc->sb = NULL;
Sage Weil16725b92009-10-06 11:31:07 -0700876 return ret;
877}
878
879/*
880 * share superblock if same fs AND options
881 */
882static int ceph_compare_super(struct super_block *sb, void *data)
883{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700884 struct ceph_fs_client *new = data;
885 struct ceph_mount_options *fsopt = new->mount_options;
886 struct ceph_options *opt = new->client->options;
887 struct ceph_fs_client *other = ceph_sb_to_client(sb);
Sage Weil16725b92009-10-06 11:31:07 -0700888
889 dout("ceph_compare_super %p\n", sb);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700890
891 if (compare_mount_options(fsopt, opt, other)) {
892 dout("monitor(s)/mount options don't match\n");
893 return 0;
Sage Weil16725b92009-10-06 11:31:07 -0700894 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700895 if ((opt->flags & CEPH_OPT_FSID) &&
896 ceph_fsid_compare(&opt->fsid, &other->client->fsid)) {
897 dout("fsid doesn't match\n");
898 return 0;
899 }
900 if (fsopt->sb_flags != other->mount_options->sb_flags) {
Sage Weil16725b92009-10-06 11:31:07 -0700901 dout("flags differ\n");
902 return 0;
903 }
904 return 1;
905}
906
907/*
908 * construct our own bdi so we can control readahead, etc.
909 */
Jeff Mahoney00d56432010-06-10 11:13:58 -0400910static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
Sage Weil31e0cf82010-05-04 16:39:35 -0700911
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700912static int ceph_register_bdi(struct super_block *sb,
913 struct ceph_fs_client *fsc)
Sage Weil16725b92009-10-06 11:31:07 -0700914{
915 int err;
916
Sage Weil83817e32011-08-04 08:03:44 -0700917 /* set ra_pages based on rasize mount option? */
918 if (fsc->mount_options->rasize >= PAGE_CACHE_SIZE)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700919 fsc->backing_dev_info.ra_pages =
Sage Weil83817e32011-08-04 08:03:44 -0700920 (fsc->mount_options->rasize + PAGE_CACHE_SIZE - 1)
Sage Weil16725b92009-10-06 11:31:07 -0700921 >> PAGE_SHIFT;
Yehuda Sadehe9852222011-07-22 11:12:28 -0700922 else
923 fsc->backing_dev_info.ra_pages =
Christoph Hellwigdf0ce262015-01-14 10:42:41 +0100924 VM_MAX_READAHEAD * 1024 / PAGE_CACHE_SIZE;
Yehuda Sadehe9852222011-07-22 11:12:28 -0700925
Joe Perchesd2cc4dd2012-11-29 08:37:03 -0600926 err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%ld",
Sage Weil31e0cf82010-05-04 16:39:35 -0700927 atomic_long_inc_return(&bdi_seq));
Sage Weil5dfc5892010-05-04 16:14:46 -0700928 if (!err)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700929 sb->s_bdi = &fsc->backing_dev_info;
Sage Weil16725b92009-10-06 11:31:07 -0700930 return err;
931}
932
Al Viroa7f9fb22010-07-26 16:17:55 +0400933static struct dentry *ceph_mount(struct file_system_type *fs_type,
934 int flags, const char *dev_name, void *data)
Sage Weil16725b92009-10-06 11:31:07 -0700935{
936 struct super_block *sb;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700937 struct ceph_fs_client *fsc;
Al Viroa7f9fb22010-07-26 16:17:55 +0400938 struct dentry *res;
Sage Weil16725b92009-10-06 11:31:07 -0700939 int err;
940 int (*compare_super)(struct super_block *, void *) = ceph_compare_super;
Sage Weil6a18be12009-11-04 11:40:05 -0800941 const char *path = NULL;
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700942 struct ceph_mount_options *fsopt = NULL;
943 struct ceph_options *opt = NULL;
Sage Weil16725b92009-10-06 11:31:07 -0700944
Al Viroa7f9fb22010-07-26 16:17:55 +0400945 dout("ceph_mount\n");
Sage Weil45195e42014-02-16 10:05:29 -0800946
947#ifdef CONFIG_CEPH_FS_POSIX_ACL
948 flags |= MS_POSIXACL;
949#endif
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700950 err = parse_mount_options(&fsopt, &opt, flags, data, dev_name, &path);
Al Viroa7f9fb22010-07-26 16:17:55 +0400951 if (err < 0) {
952 res = ERR_PTR(err);
Sage Weil6b805182009-10-27 11:50:50 -0700953 goto out_final;
Al Viroa7f9fb22010-07-26 16:17:55 +0400954 }
Sage Weil16725b92009-10-06 11:31:07 -0700955
956 /* create client (which we may/may not use) */
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700957 fsc = create_fs_client(fsopt, opt);
958 if (IS_ERR(fsc)) {
Al Viroa7f9fb22010-07-26 16:17:55 +0400959 res = ERR_CAST(fsc);
Noah Watkins259a1872011-08-22 13:49:41 -0600960 destroy_mount_options(fsopt);
961 ceph_destroy_options(opt);
Sage Weil6b805182009-10-27 11:50:50 -0700962 goto out_final;
963 }
Sage Weil16725b92009-10-06 11:31:07 -0700964
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700965 err = ceph_mdsc_init(fsc);
Al Viroa7f9fb22010-07-26 16:17:55 +0400966 if (err < 0) {
967 res = ERR_PTR(err);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700968 goto out;
Al Viroa7f9fb22010-07-26 16:17:55 +0400969 }
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700970
971 if (ceph_test_opt(fsc->client, NOSHARE))
Sage Weil16725b92009-10-06 11:31:07 -0700972 compare_super = NULL;
David Howells9249e172012-06-25 12:55:37 +0100973 sb = sget(fs_type, compare_super, ceph_set_super, flags, fsc);
Sage Weil16725b92009-10-06 11:31:07 -0700974 if (IS_ERR(sb)) {
Al Viroa7f9fb22010-07-26 16:17:55 +0400975 res = ERR_CAST(sb);
Sage Weil16725b92009-10-06 11:31:07 -0700976 goto out;
977 }
978
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700979 if (ceph_sb_to_client(sb) != fsc) {
980 ceph_mdsc_destroy(fsc);
981 destroy_fs_client(fsc);
982 fsc = ceph_sb_to_client(sb);
983 dout("get_sb got existing client %p\n", fsc);
Sage Weil16725b92009-10-06 11:31:07 -0700984 } else {
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -0700985 dout("get_sb using new client %p\n", fsc);
986 err = ceph_register_bdi(sb, fsc);
Al Viroa7f9fb22010-07-26 16:17:55 +0400987 if (err < 0) {
988 res = ERR_PTR(err);
Sage Weil16725b92009-10-06 11:31:07 -0700989 goto out_splat;
Al Viroa7f9fb22010-07-26 16:17:55 +0400990 }
Sage Weil16725b92009-10-06 11:31:07 -0700991 }
992
Al Viroa7f9fb22010-07-26 16:17:55 +0400993 res = ceph_real_mount(fsc, path);
994 if (IS_ERR(res))
Sage Weil16725b92009-10-06 11:31:07 -0700995 goto out_splat;
Al Viroa7f9fb22010-07-26 16:17:55 +0400996 dout("root %p inode %p ino %llx.%llx\n", res,
David Howells2b0143b2015-03-17 22:25:59 +0000997 d_inode(res), ceph_vinop(d_inode(res)));
Al Viroa7f9fb22010-07-26 16:17:55 +0400998 return res;
Sage Weil16725b92009-10-06 11:31:07 -0700999
1000out_splat:
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001001 ceph_mdsc_close_sessions(fsc->mdsc);
Al Viro3981f2e2010-03-21 19:22:29 -04001002 deactivate_locked_super(sb);
Sage Weil16725b92009-10-06 11:31:07 -07001003 goto out_final;
1004
1005out:
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001006 ceph_mdsc_destroy(fsc);
1007 destroy_fs_client(fsc);
Sage Weil16725b92009-10-06 11:31:07 -07001008out_final:
Al Viroa7f9fb22010-07-26 16:17:55 +04001009 dout("ceph_mount fail %ld\n", PTR_ERR(res));
1010 return res;
Sage Weil16725b92009-10-06 11:31:07 -07001011}
1012
1013static void ceph_kill_sb(struct super_block *s)
1014{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001015 struct ceph_fs_client *fsc = ceph_sb_to_client(s);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001016 dev_t dev = s->s_dev;
1017
Sage Weil16725b92009-10-06 11:31:07 -07001018 dout("kill_sb %p\n", s);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001019
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001020 ceph_mdsc_pre_umount(fsc->mdsc);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001021 generic_shutdown_super(s);
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001022 ceph_mdsc_destroy(fsc);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001023
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001024 destroy_fs_client(fsc);
Christoph Hellwige4d27502015-01-14 10:42:38 +01001025 free_anon_bdev(dev);
Sage Weil16725b92009-10-06 11:31:07 -07001026}
1027
1028static struct file_system_type ceph_fs_type = {
1029 .owner = THIS_MODULE,
1030 .name = "ceph",
Al Viroa7f9fb22010-07-26 16:17:55 +04001031 .mount = ceph_mount,
Sage Weil16725b92009-10-06 11:31:07 -07001032 .kill_sb = ceph_kill_sb,
1033 .fs_flags = FS_RENAME_DOES_D_MOVE,
1034};
Eric W. Biederman7f78e032013-03-02 19:39:14 -08001035MODULE_ALIAS_FS("ceph");
Sage Weil16725b92009-10-06 11:31:07 -07001036
Sage Weil16725b92009-10-06 11:31:07 -07001037static int __init init_ceph(void)
1038{
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001039 int ret = init_caches();
Sage Weil16725b92009-10-06 11:31:07 -07001040 if (ret)
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001041 goto out;
Sage Weil16725b92009-10-06 11:31:07 -07001042
Yan, Zhengeb13e832014-03-09 23:16:40 +08001043 ceph_flock_init();
Alex Elder3ce6cd12012-01-23 15:49:28 -06001044 ceph_xattr_init();
Yan, Zheng97c85a82014-11-06 15:09:41 +08001045 ret = ceph_snap_init();
1046 if (ret)
1047 goto out_xattr;
Sage Weil16725b92009-10-06 11:31:07 -07001048 ret = register_filesystem(&ceph_fs_type);
1049 if (ret)
Yan, Zheng97c85a82014-11-06 15:09:41 +08001050 goto out_snap;
Sage Weil16725b92009-10-06 11:31:07 -07001051
Yehuda Sadeh3d14c5d2010-04-06 15:14:15 -07001052 pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);
1053
Sage Weil16725b92009-10-06 11:31:07 -07001054 return 0;
1055
Yan, Zheng97c85a82014-11-06 15:09:41 +08001056out_snap:
1057 ceph_snap_exit();
1058out_xattr:
Alex Elder3ce6cd12012-01-23 15:49:28 -06001059 ceph_xattr_exit();
Sage Weil16725b92009-10-06 11:31:07 -07001060 destroy_caches();
Sage Weil16725b92009-10-06 11:31:07 -07001061out:
1062 return ret;
1063}
1064
1065static void __exit exit_ceph(void)
1066{
1067 dout("exit_ceph\n");
1068 unregister_filesystem(&ceph_fs_type);
Yan, Zheng97c85a82014-11-06 15:09:41 +08001069 ceph_snap_exit();
Alex Elder3ce6cd12012-01-23 15:49:28 -06001070 ceph_xattr_exit();
Sage Weil16725b92009-10-06 11:31:07 -07001071 destroy_caches();
Sage Weil16725b92009-10-06 11:31:07 -07001072}
1073
1074module_init(init_ceph);
1075module_exit(exit_ceph);
1076
1077MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
1078MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
1079MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
1080MODULE_DESCRIPTION("Ceph filesystem for Linux");
1081MODULE_LICENSE("GPL");