super.c 26.1 KB
Newer Older
S
Sage Weil 已提交
1

2
#include <linux/ceph/ceph_debug.h>
S
Sage Weil 已提交
3 4

#include <linux/backing-dev.h>
S
Sage Weil 已提交
5
#include <linux/ctype.h>
S
Sage Weil 已提交
6 7 8 9 10 11 12 13
#include <linux/fs.h>
#include <linux/inet.h>
#include <linux/in6.h>
#include <linux/module.h>
#include <linux/mount.h>
#include <linux/parser.h>
#include <linux/sched.h>
#include <linux/seq_file.h>
14
#include <linux/slab.h>
S
Sage Weil 已提交
15 16 17 18
#include <linux/statfs.h>
#include <linux/string.h>

#include "super.h"
19
#include "mds_client.h"
20
#include "cache.h"
21

22
#include <linux/ceph/ceph_features.h>
23 24 25 26
#include <linux/ceph/decode.h>
#include <linux/ceph/mon_client.h>
#include <linux/ceph/auth.h>
#include <linux/ceph/debugfs.h>
S
Sage Weil 已提交
27 28 29 30 31 32 33 34 35 36 37 38

/*
 * Ceph superblock operations
 *
 * Handle the basics of mounting, unmounting.
 */

/*
 * super ops
 */
static void ceph_put_super(struct super_block *s)
{
39
	struct ceph_fs_client *fsc = ceph_sb_to_client(s);
S
Sage Weil 已提交
40 41

	dout("put_super\n");
42
	ceph_mdsc_close_sessions(fsc->mdsc);
S
Sage Weil 已提交
43 44 45 46
}

static int ceph_statfs(struct dentry *dentry, struct kstatfs *buf)
{
47
	struct ceph_fs_client *fsc = ceph_inode_to_client(d_inode(dentry));
48
	struct ceph_monmap *monmap = fsc->client->monc.monmap;
S
Sage Weil 已提交
49 50 51 52 53
	struct ceph_statfs st;
	u64 fsid;
	int err;

	dout("statfs\n");
54
	err = ceph_monc_do_statfs(&fsc->client->monc, &st);
S
Sage Weil 已提交
55 56 57 58 59 60 61 62 63
	if (err < 0)
		return err;

	/* fill in kstatfs */
	buf->f_type = CEPH_SUPER_MAGIC;  /* ?? */

	/*
	 * express utilization in terms of large blocks to avoid
	 * overflow on 32-bit machines.
S
Sage Weil 已提交
64 65 66 67 68
	 *
	 * NOTE: for the time being, we make bsize == frsize to humor
	 * not-yet-ancient versions of glibc that are broken.
	 * Someday, we will probably want to report a real block
	 * size...  whatever that may mean for a network file system!
S
Sage Weil 已提交
69 70
	 */
	buf->f_bsize = 1 << CEPH_BLOCK_SHIFT;
S
Sage Weil 已提交
71
	buf->f_frsize = 1 << CEPH_BLOCK_SHIFT;
S
Sage Weil 已提交
72
	buf->f_blocks = le64_to_cpu(st.kb) >> (CEPH_BLOCK_SHIFT-10);
73
	buf->f_bfree = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);
S
Sage Weil 已提交
74 75 76 77
	buf->f_bavail = le64_to_cpu(st.kb_avail) >> (CEPH_BLOCK_SHIFT-10);

	buf->f_files = le64_to_cpu(st.num_objects);
	buf->f_ffree = -1;
78
	buf->f_namelen = NAME_MAX;
S
Sage Weil 已提交
79 80 81 82 83 84 85 86 87 88

	/* leave fsid little-endian, regardless of host endianness */
	fsid = *(u64 *)(&monmap->fsid) ^ *((u64 *)&monmap->fsid + 1);
	buf->f_fsid.val[0] = fsid & 0xffffffff;
	buf->f_fsid.val[1] = fsid >> 32;

	return 0;
}


89
static int ceph_sync_fs(struct super_block *sb, int wait)
S
Sage Weil 已提交
90
{
91
	struct ceph_fs_client *fsc = ceph_sb_to_client(sb);
92 93 94

	if (!wait) {
		dout("sync_fs (non-blocking)\n");
95
		ceph_flush_dirty_caps(fsc->mdsc);
96 97 98 99 100
		dout("sync_fs (non-blocking) done\n");
		return 0;
	}

	dout("sync_fs (blocking)\n");
101 102
	ceph_osdc_sync(&fsc->client->osdc);
	ceph_mdsc_sync(fsc->mdsc);
103
	dout("sync_fs (blocking) done\n");
S
Sage Weil 已提交
104 105 106 107 108 109 110 111 112
	return 0;
}

/*
 * mount options
 */
enum {
	Opt_wsize,
	Opt_rsize,
S
Sage Weil 已提交
113
	Opt_rasize,
S
Sage Weil 已提交
114 115
	Opt_caps_wanted_delay_min,
	Opt_caps_wanted_delay_max,
116
	Opt_cap_release_safety,
S
Sage Weil 已提交
117
	Opt_readdir_max_entries,
118
	Opt_readdir_max_bytes,
Y
Yehuda Sadeh 已提交
119
	Opt_congestion_kb,
120
	Opt_last_int,
S
Sage Weil 已提交
121 122
	/* int args above */
	Opt_snapdirname,
123
	Opt_mds_namespace,
124
	Opt_last_string,
S
Sage Weil 已提交
125 126 127 128 129
	/* string args above */
	Opt_dirstat,
	Opt_nodirstat,
	Opt_rbytes,
	Opt_norbytes,
130
	Opt_asyncreaddir,
S
Sage Weil 已提交
131
	Opt_noasyncreaddir,
132 133
	Opt_dcache,
	Opt_nodcache,
Y
Yehuda Sadeh 已提交
134
	Opt_ino32,
135
	Opt_noino32,
136
	Opt_fscache,
137
	Opt_nofscache,
138 139
	Opt_poolperm,
	Opt_nopoolperm,
140 141 142
#ifdef CONFIG_CEPH_FS_POSIX_ACL
	Opt_acl,
#endif
143
	Opt_noacl,
S
Sage Weil 已提交
144 145
};

146
static match_table_t fsopt_tokens = {
S
Sage Weil 已提交
147 148
	{Opt_wsize, "wsize=%d"},
	{Opt_rsize, "rsize=%d"},
S
Sage Weil 已提交
149
	{Opt_rasize, "rasize=%d"},
S
Sage Weil 已提交
150 151
	{Opt_caps_wanted_delay_min, "caps_wanted_delay_min=%d"},
	{Opt_caps_wanted_delay_max, "caps_wanted_delay_max=%d"},
152
	{Opt_cap_release_safety, "cap_release_safety=%d"},
S
Sage Weil 已提交
153
	{Opt_readdir_max_entries, "readdir_max_entries=%d"},
154
	{Opt_readdir_max_bytes, "readdir_max_bytes=%d"},
Y
Yehuda Sadeh 已提交
155
	{Opt_congestion_kb, "write_congestion_kb=%d"},
S
Sage Weil 已提交
156 157
	/* int args above */
	{Opt_snapdirname, "snapdirname=%s"},
158
	{Opt_mds_namespace, "mds_namespace=%s"},
S
Sage Weil 已提交
159 160 161 162 163
	/* string args above */
	{Opt_dirstat, "dirstat"},
	{Opt_nodirstat, "nodirstat"},
	{Opt_rbytes, "rbytes"},
	{Opt_norbytes, "norbytes"},
164
	{Opt_asyncreaddir, "asyncreaddir"},
S
Sage Weil 已提交
165
	{Opt_noasyncreaddir, "noasyncreaddir"},
166 167
	{Opt_dcache, "dcache"},
	{Opt_nodcache, "nodcache"},
Y
Yehuda Sadeh 已提交
168
	{Opt_ino32, "ino32"},
169
	{Opt_noino32, "noino32"},
170 171
	{Opt_fscache, "fsc"},
	{Opt_nofscache, "nofsc"},
172 173
	{Opt_poolperm, "poolperm"},
	{Opt_nopoolperm, "nopoolperm"},
174 175 176 177
#ifdef CONFIG_CEPH_FS_POSIX_ACL
	{Opt_acl, "acl"},
#endif
	{Opt_noacl, "noacl"},
S
Sage Weil 已提交
178 179 180
	{-1, NULL}
};

181
static int parse_fsopt_token(char *c, void *private)
S
Sage Weil 已提交
182
{
183 184 185 186 187 188 189 190 191 192 193 194 195 196
	struct ceph_mount_options *fsopt = private;
	substring_t argstr[MAX_OPT_ARGS];
	int token, intval, ret;

	token = match_token((char *)c, fsopt_tokens, argstr);
	if (token < 0)
		return -EINVAL;

	if (token < Opt_last_int) {
		ret = match_int(&argstr[0], &intval);
		if (ret < 0) {
			pr_err("bad mount option arg (not int) "
			       "at '%s'\n", c);
			return ret;
S
Sage Weil 已提交
197
		}
198 199 200 201 202 203
		dout("got int token %d val %d\n", token, intval);
	} else if (token > Opt_last_int && token < Opt_last_string) {
		dout("got string token %d val %s\n", token,
		     argstr[0].from);
	} else {
		dout("got token %d\n", token);
S
Sage Weil 已提交
204 205
	}

206 207 208 209 210 211 212 213 214
	switch (token) {
	case Opt_snapdirname:
		kfree(fsopt->snapdir_name);
		fsopt->snapdir_name = kstrndup(argstr[0].from,
					       argstr[0].to-argstr[0].from,
					       GFP_KERNEL);
		if (!fsopt->snapdir_name)
			return -ENOMEM;
		break;
Y
Yan, Zheng 已提交
215
	case Opt_mds_namespace:
216 217 218 219 220
		fsopt->mds_namespace = kstrndup(argstr[0].from,
						argstr[0].to-argstr[0].from,
						GFP_KERNEL);
		if (!fsopt->mds_namespace)
			return -ENOMEM;
Y
Yan, Zheng 已提交
221
		break;
222
		/* misc */
223 224 225 226 227 228
	case Opt_wsize:
		fsopt->wsize = intval;
		break;
	case Opt_rsize:
		fsopt->rsize = intval;
		break;
S
Sage Weil 已提交
229 230 231
	case Opt_rasize:
		fsopt->rasize = intval;
		break;
232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258
	case Opt_caps_wanted_delay_min:
		fsopt->caps_wanted_delay_min = intval;
		break;
	case Opt_caps_wanted_delay_max:
		fsopt->caps_wanted_delay_max = intval;
		break;
	case Opt_readdir_max_entries:
		fsopt->max_readdir = intval;
		break;
	case Opt_readdir_max_bytes:
		fsopt->max_readdir_bytes = intval;
		break;
	case Opt_congestion_kb:
		fsopt->congestion_kb = intval;
		break;
	case Opt_dirstat:
		fsopt->flags |= CEPH_MOUNT_OPT_DIRSTAT;
		break;
	case Opt_nodirstat:
		fsopt->flags &= ~CEPH_MOUNT_OPT_DIRSTAT;
		break;
	case Opt_rbytes:
		fsopt->flags |= CEPH_MOUNT_OPT_RBYTES;
		break;
	case Opt_norbytes:
		fsopt->flags &= ~CEPH_MOUNT_OPT_RBYTES;
		break;
259 260 261
	case Opt_asyncreaddir:
		fsopt->flags &= ~CEPH_MOUNT_OPT_NOASYNCREADDIR;
		break;
262 263 264
	case Opt_noasyncreaddir:
		fsopt->flags |= CEPH_MOUNT_OPT_NOASYNCREADDIR;
		break;
265 266 267 268 269 270
	case Opt_dcache:
		fsopt->flags |= CEPH_MOUNT_OPT_DCACHE;
		break;
	case Opt_nodcache:
		fsopt->flags &= ~CEPH_MOUNT_OPT_DCACHE;
		break;
Y
Yehuda Sadeh 已提交
271 272 273
	case Opt_ino32:
		fsopt->flags |= CEPH_MOUNT_OPT_INO32;
		break;
274 275 276
	case Opt_noino32:
		fsopt->flags &= ~CEPH_MOUNT_OPT_INO32;
		break;
277 278 279 280 281 282
	case Opt_fscache:
		fsopt->flags |= CEPH_MOUNT_OPT_FSCACHE;
		break;
	case Opt_nofscache:
		fsopt->flags &= ~CEPH_MOUNT_OPT_FSCACHE;
		break;
283 284 285 286 287 288 289
	case Opt_poolperm:
		fsopt->flags &= ~CEPH_MOUNT_OPT_NOPOOLPERM;
		printk ("pool perm");
		break;
	case Opt_nopoolperm:
		fsopt->flags |= CEPH_MOUNT_OPT_NOPOOLPERM;
		break;
290 291 292 293 294 295 296 297
#ifdef CONFIG_CEPH_FS_POSIX_ACL
	case Opt_acl:
		fsopt->sb_flags |= MS_POSIXACL;
		break;
#endif
	case Opt_noacl:
		fsopt->sb_flags &= ~MS_POSIXACL;
		break;
298 299 300 301
	default:
		BUG_ON(token);
	}
	return 0;
S
Sage Weil 已提交
302
}
S
Sage Weil 已提交
303

304
static void destroy_mount_options(struct ceph_mount_options *args)
S
Sage Weil 已提交
305
{
306 307
	dout("destroy_mount_options %p\n", args);
	kfree(args->snapdir_name);
308
	kfree(args->mds_namespace);
309
	kfree(args->server_path);
310 311
	kfree(args);
}
S
Sage Weil 已提交
312

313 314 315 316 317 318 319 320 321 322
static int strcmp_null(const char *s1, const char *s2)
{
	if (!s1 && !s2)
		return 0;
	if (s1 && !s2)
		return -1;
	if (!s1 && s2)
		return 1;
	return strcmp(s1, s2);
}
S
Sage Weil 已提交
323

324 325 326 327 328 329 330 331
static int compare_mount_options(struct ceph_mount_options *new_fsopt,
				 struct ceph_options *new_opt,
				 struct ceph_fs_client *fsc)
{
	struct ceph_mount_options *fsopt1 = new_fsopt;
	struct ceph_mount_options *fsopt2 = fsc->mount_options;
	int ofs = offsetof(struct ceph_mount_options, snapdir_name);
	int ret;
S
Sage Weil 已提交
332

333 334 335 336 337
	ret = memcmp(fsopt1, fsopt2, ofs);
	if (ret)
		return ret;

	ret = strcmp_null(fsopt1->snapdir_name, fsopt2->snapdir_name);
338 339 340
	if (ret)
		return ret;
	ret = strcmp_null(fsopt1->mds_namespace, fsopt2->mds_namespace);
341 342 343
	if (ret)
		return ret;

344 345 346 347
	ret = strcmp_null(fsopt1->server_path, fsopt2->server_path);
	if (ret)
		return ret;

348 349 350 351 352 353
	return ceph_compare_options(new_opt, fsc->client);
}

static int parse_mount_options(struct ceph_mount_options **pfsopt,
			       struct ceph_options **popt,
			       int flags, char *options,
354
			       const char *dev_name)
355 356 357
{
	struct ceph_mount_options *fsopt;
	const char *dev_name_end;
358 359 360 361
	int err;

	if (!dev_name || !*dev_name)
		return -EINVAL;
362 363 364 365 366 367 368

	fsopt = kzalloc(sizeof(*fsopt), GFP_KERNEL);
	if (!fsopt)
		return -ENOMEM;

	dout("parse_mount_options %p, dev_name '%s'\n", fsopt, dev_name);

369 370
	fsopt->sb_flags = flags;
	fsopt->flags = CEPH_MOUNT_OPT_DEFAULT;
371

372 373 374
	fsopt->rsize = CEPH_RSIZE_DEFAULT;
	fsopt->rasize = CEPH_RASIZE_DEFAULT;
	fsopt->snapdir_name = kstrdup(CEPH_SNAPDIRNAME_DEFAULT, GFP_KERNEL);
375 376 377 378 379
	if (!fsopt->snapdir_name) {
		err = -ENOMEM;
		goto out;
	}

380 381
	fsopt->caps_wanted_delay_min = CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT;
	fsopt->caps_wanted_delay_max = CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT;
382 383 384 385 386
	fsopt->cap_release_safety = CEPH_CAP_RELEASE_SAFETY_DEFAULT;
	fsopt->max_readdir = CEPH_MAX_READDIR_DEFAULT;
	fsopt->max_readdir_bytes = CEPH_MAX_READDIR_BYTES_DEFAULT;
	fsopt->congestion_kb = default_congestion_kb();

387 388 389 390 391 392 393 394 395 396 397 398
	/*
	 * Distinguish the server list from the path in "dev_name".
	 * Internally we do not include the leading '/' in the path.
	 *
	 * "dev_name" will look like:
	 *     <server_spec>[,<server_spec>...]:[<path>]
	 * where
	 *     <server_spec> is <ip>[:<port>]
	 *     <path> is optional, but if present must begin with '/'
	 */
	dev_name_end = strchr(dev_name, '/');
	if (dev_name_end) {
399 400 401 402 403 404
		if (strlen(dev_name_end) > 1) {
			fsopt->server_path = kstrdup(dev_name_end, GFP_KERNEL);
			if (!fsopt->server_path) {
				err = -ENOMEM;
				goto out;
			}
405
		}
406 407 408
	} else {
		dev_name_end = dev_name + strlen(dev_name);
	}
409
	err = -EINVAL;
410
	dev_name_end--;		/* back up to ':' separator */
411
	if (dev_name_end < dev_name || *dev_name_end != ':') {
412
		pr_err("device name is missing path (no : separator in %s)\n",
413 414 415
				dev_name);
		goto out;
	}
416
	dout("device name '%.*s'\n", (int)(dev_name_end - dev_name), dev_name);
417 418
	if (fsopt->server_path)
		dout("server path '%s'\n", fsopt->server_path);
S
Sage Weil 已提交
419

420
	*popt = ceph_parse_options(options, dev_name, dev_name_end,
421
				 parse_fsopt_token, (void *)fsopt);
422 423
	if (IS_ERR(*popt)) {
		err = PTR_ERR(*popt);
424
		goto out;
425
	}
426 427 428 429

	/* success */
	*pfsopt = fsopt;
	return 0;
S
Sage Weil 已提交
430

431
out:
432 433
	destroy_mount_options(fsopt);
	return err;
S
Sage Weil 已提交
434 435
}

436 437 438
/**
 * ceph_show_options - Show mount options in /proc/mounts
 * @m: seq_file to write to
439
 * @root: root of that (sub)tree
440
 */
441
static int ceph_show_options(struct seq_file *m, struct dentry *root)
S
Sage Weil 已提交
442
{
443
	struct ceph_fs_client *fsc = ceph_sb_to_client(root->d_sb);
444
	struct ceph_mount_options *fsopt = fsc->mount_options;
445 446 447 448 449 450 451 452 453 454 455 456 457 458
	size_t pos;
	int ret;

	/* a comma between MNT/MS and client options */
	seq_putc(m, ',');
	pos = m->count;

	ret = ceph_print_client_options(m, fsc->client);
	if (ret)
		return ret;

	/* retract our comma if no client options */
	if (m->count == pos)
		m->count--;
459 460 461

	if (fsopt->flags & CEPH_MOUNT_OPT_DIRSTAT)
		seq_puts(m, ",dirstat");
462 463
	if ((fsopt->flags & CEPH_MOUNT_OPT_RBYTES))
		seq_puts(m, ",rbytes");
464 465
	if (fsopt->flags & CEPH_MOUNT_OPT_NOASYNCREADDIR)
		seq_puts(m, ",noasyncreaddir");
466
	if ((fsopt->flags & CEPH_MOUNT_OPT_DCACHE) == 0)
467
		seq_puts(m, ",nodcache");
468 469
	if (fsopt->flags & CEPH_MOUNT_OPT_FSCACHE)
		seq_puts(m, ",fsc");
470 471
	if (fsopt->flags & CEPH_MOUNT_OPT_NOPOOLPERM)
		seq_puts(m, ",nopoolperm");
472

473 474 475 476 477 478 479
#ifdef CONFIG_CEPH_FS_POSIX_ACL
	if (fsopt->sb_flags & MS_POSIXACL)
		seq_puts(m, ",acl");
	else
		seq_puts(m, ",noacl");
#endif

480 481
	if (fsopt->mds_namespace)
		seq_printf(m, ",mds_namespace=%s", fsopt->mds_namespace);
482 483
	if (fsopt->wsize)
		seq_printf(m, ",wsize=%d", fsopt->wsize);
484
	if (fsopt->rsize != CEPH_RSIZE_DEFAULT)
485
		seq_printf(m, ",rsize=%d", fsopt->rsize);
S
Sage Weil 已提交
486
	if (fsopt->rasize != CEPH_RASIZE_DEFAULT)
487
		seq_printf(m, ",rasize=%d", fsopt->rasize);
488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503
	if (fsopt->congestion_kb != default_congestion_kb())
		seq_printf(m, ",write_congestion_kb=%d", fsopt->congestion_kb);
	if (fsopt->caps_wanted_delay_min != CEPH_CAPS_WANTED_DELAY_MIN_DEFAULT)
		seq_printf(m, ",caps_wanted_delay_min=%d",
			 fsopt->caps_wanted_delay_min);
	if (fsopt->caps_wanted_delay_max != CEPH_CAPS_WANTED_DELAY_MAX_DEFAULT)
		seq_printf(m, ",caps_wanted_delay_max=%d",
			   fsopt->caps_wanted_delay_max);
	if (fsopt->cap_release_safety != CEPH_CAP_RELEASE_SAFETY_DEFAULT)
		seq_printf(m, ",cap_release_safety=%d",
			   fsopt->cap_release_safety);
	if (fsopt->max_readdir != CEPH_MAX_READDIR_DEFAULT)
		seq_printf(m, ",readdir_max_entries=%d", fsopt->max_readdir);
	if (fsopt->max_readdir_bytes != CEPH_MAX_READDIR_BYTES_DEFAULT)
		seq_printf(m, ",readdir_max_bytes=%d", fsopt->max_readdir_bytes);
	if (strcmp(fsopt->snapdir_name, CEPH_SNAPDIRNAME_DEFAULT))
504
		seq_show_option(m, "snapdirname", fsopt->snapdir_name);
505

506
	return 0;
S
Sage Weil 已提交
507 508 509
}

/*
510 511
 * handle any mon messages the standard library doesn't understand.
 * return error if we don't either.
S
Sage Weil 已提交
512
 */
513
static int extra_mon_dispatch(struct ceph_client *client, struct ceph_msg *msg)
S
Sage Weil 已提交
514
{
515 516 517 518 519
	struct ceph_fs_client *fsc = client->private;
	int type = le16_to_cpu(msg->hdr.type);

	switch (type) {
	case CEPH_MSG_MDS_MAP:
520 521 522 523
		ceph_mdsc_handle_mdsmap(fsc->mdsc, msg);
		return 0;
	case CEPH_MSG_FS_MAP_USER:
		ceph_mdsc_handle_fsmap(fsc->mdsc, msg);
524 525 526 527 528 529 530 531 532
		return 0;
	default:
		return -1;
	}
}

/*
 * create a new fs client
 */
533
static struct ceph_fs_client *create_fs_client(struct ceph_mount_options *fsopt,
534 535 536
					struct ceph_options *opt)
{
	struct ceph_fs_client *fsc;
537
	const u64 supported_features =
Y
Yan, Zheng 已提交
538 539
		CEPH_FEATURE_FLOCK | CEPH_FEATURE_DIRLAYOUTHASH |
		CEPH_FEATURE_MDSENC | CEPH_FEATURE_MDS_INLINE_DATA;
540
	const u64 required_features = 0;
541 542
	int page_count;
	size_t size;
S
Sage Weil 已提交
543 544
	int err = -ENOMEM;

545 546
	fsc = kzalloc(sizeof(*fsc), GFP_KERNEL);
	if (!fsc)
S
Sage Weil 已提交
547 548
		return ERR_PTR(-ENOMEM);

S
Sage Weil 已提交
549 550
	fsc->client = ceph_create_client(opt, fsc, supported_features,
					 required_features);
551 552 553 554 555
	if (IS_ERR(fsc->client)) {
		err = PTR_ERR(fsc->client);
		goto fail;
	}
	fsc->client->extra_mon_dispatch = extra_mon_dispatch;
556 557 558 559 560 561 562 563

	if (fsopt->mds_namespace == NULL) {
		ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_MDSMAP,
				   0, true);
	} else {
		ceph_monc_want_map(&fsc->client->monc, CEPH_SUB_FSMAP,
				   0, false);
	}
S
Sage Weil 已提交
564

565
	fsc->mount_options = fsopt;
S
Sage Weil 已提交
566

567 568
	fsc->sb = NULL;
	fsc->mount_state = CEPH_MOUNT_MOUNTING;
S
Sage Weil 已提交
569

570
	atomic_long_set(&fsc->writeback_count, 0);
S
Sage Weil 已提交
571

572
	err = bdi_init(&fsc->backing_dev_info);
573
	if (err < 0)
574
		goto fail_client;
575

S
Sage Weil 已提交
576
	err = -ENOMEM;
577 578 579 580 581
	/*
	 * The number of concurrent works can be high but they don't need
	 * to be processed in parallel, limit concurrency.
	 */
	fsc->wb_wq = alloc_workqueue("ceph-writeback", 0, 1);
582
	if (fsc->wb_wq == NULL)
583
		goto fail_bdi;
584
	fsc->pg_inv_wq = alloc_workqueue("ceph-pg-invalid", 0, 1);
585
	if (fsc->pg_inv_wq == NULL)
S
Sage Weil 已提交
586
		goto fail_wb_wq;
587
	fsc->trunc_wq = alloc_workqueue("ceph-trunc", 0, 1);
588
	if (fsc->trunc_wq == NULL)
S
Sage Weil 已提交
589 590
		goto fail_pg_inv_wq;

591 592
	/* set up mempools */
	err = -ENOMEM;
593
	page_count = fsc->mount_options->wsize >> PAGE_SHIFT;
594 595
	size = sizeof (struct page *) * (page_count ? page_count : 1);
	fsc->wb_pagevec_pool = mempool_create_kmalloc_pool(10, size);
596
	if (!fsc->wb_pagevec_pool)
597 598
		goto fail_trunc_wq;

599 600 601 602 603
	/* setup fscache */
	if ((fsopt->flags & CEPH_MOUNT_OPT_FSCACHE) &&
	    (ceph_fscache_register_fs(fsc) != 0))
		goto fail_fscache;

604
	/* caps */
605 606 607
	fsc->min_caps = fsopt->max_readdir;

	return fsc;
608

609 610
fail_fscache:
	ceph_fscache_unregister_fs(fsc);
S
Sage Weil 已提交
611
fail_trunc_wq:
612
	destroy_workqueue(fsc->trunc_wq);
S
Sage Weil 已提交
613
fail_pg_inv_wq:
614
	destroy_workqueue(fsc->pg_inv_wq);
S
Sage Weil 已提交
615
fail_wb_wq:
616
	destroy_workqueue(fsc->wb_wq);
617
fail_bdi:
618 619 620
	bdi_destroy(&fsc->backing_dev_info);
fail_client:
	ceph_destroy_client(fsc->client);
S
Sage Weil 已提交
621
fail:
622
	kfree(fsc);
S
Sage Weil 已提交
623 624 625
	return ERR_PTR(err);
}

626
static void destroy_fs_client(struct ceph_fs_client *fsc)
S
Sage Weil 已提交
627
{
628
	dout("destroy_fs_client %p\n", fsc);
S
Sage Weil 已提交
629

630 631
	ceph_fscache_unregister_fs(fsc);

632 633 634
	destroy_workqueue(fsc->wb_wq);
	destroy_workqueue(fsc->pg_inv_wq);
	destroy_workqueue(fsc->trunc_wq);
S
Sage Weil 已提交
635

636
	bdi_destroy(&fsc->backing_dev_info);
637

638
	mempool_destroy(fsc->wb_pagevec_pool);
S
Sage Weil 已提交
639

640
	destroy_mount_options(fsc->mount_options);
641

642
	ceph_fs_debugfs_cleanup(fsc);
S
Sage Weil 已提交
643

644
	ceph_destroy_client(fsc->client);
S
Sage Weil 已提交
645

646 647
	kfree(fsc);
	dout("destroy_fs_client %p done\n", fsc);
S
Sage Weil 已提交
648 649
}

650
/*
651
 * caches
652
 */
653 654
struct kmem_cache *ceph_inode_cachep;
struct kmem_cache *ceph_cap_cachep;
655
struct kmem_cache *ceph_cap_flush_cachep;
656 657 658 659
struct kmem_cache *ceph_dentry_cachep;
struct kmem_cache *ceph_file_cachep;

static void ceph_inode_init_once(void *foo)
660
{
661 662 663 664 665 666
	struct ceph_inode_info *ci = foo;
	inode_init_once(&ci->vfs_inode);
}

static int __init init_caches(void)
{
667 668
	int error = -ENOMEM;

669 670 671
	ceph_inode_cachep = kmem_cache_create("ceph_inode_info",
				      sizeof(struct ceph_inode_info),
				      __alignof__(struct ceph_inode_info),
672 673
				      SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
				      SLAB_ACCOUNT, ceph_inode_init_once);
674 675 676 677 678 679 680
	if (ceph_inode_cachep == NULL)
		return -ENOMEM;

	ceph_cap_cachep = KMEM_CACHE(ceph_cap,
				     SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
	if (ceph_cap_cachep == NULL)
		goto bad_cap;
681 682 683 684
	ceph_cap_flush_cachep = KMEM_CACHE(ceph_cap_flush,
					   SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
	if (ceph_cap_flush_cachep == NULL)
		goto bad_cap_flush;
685 686 687 688 689 690

	ceph_dentry_cachep = KMEM_CACHE(ceph_dentry_info,
					SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD);
	if (ceph_dentry_cachep == NULL)
		goto bad_dentry;

691 692
	ceph_file_cachep = KMEM_CACHE(ceph_file_info, SLAB_MEM_SPREAD);

693 694 695
	if (ceph_file_cachep == NULL)
		goto bad_file;

696 697
	if ((error = ceph_fscache_register()))
		goto bad_file;
698

699
	return 0;
700 701 702
bad_file:
	kmem_cache_destroy(ceph_dentry_cachep);
bad_dentry:
703 704
	kmem_cache_destroy(ceph_cap_flush_cachep);
bad_cap_flush:
705 706 707
	kmem_cache_destroy(ceph_cap_cachep);
bad_cap:
	kmem_cache_destroy(ceph_inode_cachep);
708
	return error;
709 710
}

711 712
static void destroy_caches(void)
{
713 714 715 716 717
	/*
	 * Make sure all delayed rcu free inodes are flushed before we
	 * destroy cache.
	 */
	rcu_barrier();
718

719 720
	kmem_cache_destroy(ceph_inode_cachep);
	kmem_cache_destroy(ceph_cap_cachep);
721
	kmem_cache_destroy(ceph_cap_flush_cachep);
722 723
	kmem_cache_destroy(ceph_dentry_cachep);
	kmem_cache_destroy(ceph_file_cachep);
724 725

	ceph_fscache_unregister();
726 727 728
}


S
Sage Weil 已提交
729
/*
730 731
 * ceph_umount_begin - initiate forced umount.  Tear down down the
 * mount, skipping steps that may hang while waiting for server(s).
S
Sage Weil 已提交
732
 */
733
static void ceph_umount_begin(struct super_block *sb)
S
Sage Weil 已提交
734
{
735 736 737 738 739 740
	struct ceph_fs_client *fsc = ceph_sb_to_client(sb);

	dout("ceph_umount_begin - starting forced umount\n");
	if (!fsc)
		return;
	fsc->mount_state = CEPH_MOUNT_SHUTDOWN;
741
	ceph_mdsc_force_umount(fsc->mdsc);
742
	return;
S
Sage Weil 已提交
743 744
}

745 746 747 748
static const struct super_operations ceph_super_ops = {
	.alloc_inode	= ceph_alloc_inode,
	.destroy_inode	= ceph_destroy_inode,
	.write_inode    = ceph_write_inode,
Y
Yan, Zheng 已提交
749
	.drop_inode	= ceph_drop_inode,
750
	.evict_inode	= ceph_evict_inode,
751 752 753 754 755 756 757
	.sync_fs        = ceph_sync_fs,
	.put_super	= ceph_put_super,
	.show_options   = ceph_show_options,
	.statfs		= ceph_statfs,
	.umount_begin   = ceph_umount_begin,
};

S
Sage Weil 已提交
758 759 760 761
/*
 * Bootstrap mount by opening the root directory.  Note the mount
 * @started time from caller, and time out if this takes too long.
 */
762
static struct dentry *open_root_dentry(struct ceph_fs_client *fsc,
S
Sage Weil 已提交
763 764 765
				       const char *path,
				       unsigned long started)
{
766
	struct ceph_mds_client *mdsc = fsc->mdsc;
S
Sage Weil 已提交
767 768 769 770 771 772 773 774
	struct ceph_mds_request *req = NULL;
	int err;
	struct dentry *root;

	/* open dir */
	dout("open_root_inode opening '%s'\n", path);
	req = ceph_mdsc_create_request(mdsc, CEPH_MDS_OP_GETATTR, USE_ANY_MDS);
	if (IS_ERR(req))
J
Julia Lawall 已提交
775
		return ERR_CAST(req);
S
Sage Weil 已提交
776
	req->r_path1 = kstrdup(path, GFP_NOFS);
777 778 779 780 781
	if (!req->r_path1) {
		root = ERR_PTR(-ENOMEM);
		goto out;
	}

S
Sage Weil 已提交
782 783 784
	req->r_ino1.ino = CEPH_INO_ROOT;
	req->r_ino1.snap = CEPH_NOSNAP;
	req->r_started = started;
785
	req->r_timeout = fsc->client->options->mount_timeout;
S
Sage Weil 已提交
786 787 788 789
	req->r_args.getattr.mask = cpu_to_le32(CEPH_STAT_CAP_INODE);
	req->r_num_caps = 2;
	err = ceph_mdsc_do_request(mdsc, NULL, req);
	if (err == 0) {
A
Al Viro 已提交
790 791
		struct inode *inode = req->r_target_inode;
		req->r_target_inode = NULL;
S
Sage Weil 已提交
792
		dout("open_root_inode success\n");
793 794 795 796
		root = d_make_root(inode);
		if (!root) {
			root = ERR_PTR(-ENOMEM);
			goto out;
S
Sage Weil 已提交
797
		}
798
		ceph_init_dentry(root);
S
Sage Weil 已提交
799 800 801 802
		dout("open_root_inode success, root dentry is %p\n", root);
	} else {
		root = ERR_PTR(err);
	}
A
Al Viro 已提交
803
out:
S
Sage Weil 已提交
804 805 806 807
	ceph_mdsc_put_request(req);
	return root;
}

808 809 810



S
Sage Weil 已提交
811 812 813
/*
 * mount: join the ceph cluster, and open root directory.
 */
814
static struct dentry *ceph_real_mount(struct ceph_fs_client *fsc)
S
Sage Weil 已提交
815 816 817 818
{
	int err;
	unsigned long started = jiffies;  /* note the start time */
	struct dentry *root;
819
	int first = 0;   /* first vfsmount for this super_block */
S
Sage Weil 已提交
820

821
	dout("mount start %p\n", fsc);
822
	mutex_lock(&fsc->client->mount_mutex);
S
Sage Weil 已提交
823

824
	if (!fsc->sb->s_root) {
825
		const char *path;
826 827 828
		err = __ceph_open_session(fsc->client, started);
		if (err < 0)
			goto out;
S
Sage Weil 已提交
829

830 831 832 833 834 835 836 837
		if (!fsc->mount_options->server_path) {
			path = "";
			dout("mount opening path \\t\n");
		} else {
			path = fsc->mount_options->server_path + 1;
			dout("mount opening path %s\n", path);
		}
		root = open_root_dentry(fsc, path, started);
838 839 840 841
		if (IS_ERR(root)) {
			err = PTR_ERR(root);
			goto out;
		}
842
		fsc->sb->s_root = dget(root);
843 844 845 846 847 848
		first = 1;

		err = ceph_fs_debugfs_init(fsc);
		if (err < 0)
			goto fail;
	}
S
Sage Weil 已提交
849

850
	fsc->mount_state = CEPH_MOUNT_MOUNTED;
S
Sage Weil 已提交
851
	dout("mount success\n");
A
Al Viro 已提交
852 853
	mutex_unlock(&fsc->client->mount_mutex);
	return root;
S
Sage Weil 已提交
854

855 856 857 858 859
fail:
	if (first) {
		dput(fsc->sb->s_root);
		fsc->sb->s_root = NULL;
	}
860 861 862
out:
	mutex_unlock(&fsc->client->mount_mutex);
	return ERR_PTR(err);
S
Sage Weil 已提交
863 864 865 866
}

static int ceph_set_super(struct super_block *s, void *data)
{
867
	struct ceph_fs_client *fsc = data;
S
Sage Weil 已提交
868 869 870 871
	int ret;

	dout("set_super %p data %p\n", s, data);

872
	s->s_flags = fsc->mount_options->sb_flags;
S
Sage Weil 已提交
873 874
	s->s_maxbytes = 1ULL << 40;  /* temp value until we get mdsmap */

G
Guangliang Zhao 已提交
875
	s->s_xattr = ceph_xattr_handlers;
876 877
	s->s_fs_info = fsc;
	fsc->sb = s;
S
Sage Weil 已提交
878 879

	s->s_op = &ceph_super_ops;
A
Al Viro 已提交
880
	s->s_d_op = &ceph_dentry_ops;
S
Sage Weil 已提交
881 882 883 884 885 886 887 888 889 890 891 892
	s->s_export_op = &ceph_export_ops;

	s->s_time_gran = 1000;  /* 1000 ns == 1 us */

	ret = set_anon_super(s, NULL);  /* what is that second arg for? */
	if (ret != 0)
		goto fail;

	return ret;

fail:
	s->s_fs_info = NULL;
893
	fsc->sb = NULL;
S
Sage Weil 已提交
894 895 896 897 898 899 900 901
	return ret;
}

/*
 * share superblock if same fs AND options
 */
static int ceph_compare_super(struct super_block *sb, void *data)
{
902 903 904 905
	struct ceph_fs_client *new = data;
	struct ceph_mount_options *fsopt = new->mount_options;
	struct ceph_options *opt = new->client->options;
	struct ceph_fs_client *other = ceph_sb_to_client(sb);
S
Sage Weil 已提交
906 907

	dout("ceph_compare_super %p\n", sb);
908 909 910 911

	if (compare_mount_options(fsopt, opt, other)) {
		dout("monitor(s)/mount options don't match\n");
		return 0;
S
Sage Weil 已提交
912
	}
913 914 915 916 917 918
	if ((opt->flags & CEPH_OPT_FSID) &&
	    ceph_fsid_compare(&opt->fsid, &other->client->fsid)) {
		dout("fsid doesn't match\n");
		return 0;
	}
	if (fsopt->sb_flags != other->mount_options->sb_flags) {
S
Sage Weil 已提交
919 920 921 922 923 924 925 926 927
		dout("flags differ\n");
		return 0;
	}
	return 1;
}

/*
 * construct our own bdi so we can control readahead, etc.
 */
928
static atomic_long_t bdi_seq = ATOMIC_LONG_INIT(0);
929

930 931
static int ceph_register_bdi(struct super_block *sb,
			     struct ceph_fs_client *fsc)
S
Sage Weil 已提交
932 933 934
{
	int err;

S
Sage Weil 已提交
935
	/* set ra_pages based on rasize mount option? */
936
	if (fsc->mount_options->rasize >= PAGE_SIZE)
937
		fsc->backing_dev_info.ra_pages =
938
			(fsc->mount_options->rasize + PAGE_SIZE - 1)
S
Sage Weil 已提交
939
			>> PAGE_SHIFT;
940 941
	else
		fsc->backing_dev_info.ra_pages =
942
			VM_MAX_READAHEAD * 1024 / PAGE_SIZE;
943

944
	err = bdi_register(&fsc->backing_dev_info, NULL, "ceph-%ld",
945
			   atomic_long_inc_return(&bdi_seq));
946
	if (!err)
947
		sb->s_bdi = &fsc->backing_dev_info;
S
Sage Weil 已提交
948 949 950
	return err;
}

A
Al Viro 已提交
951 952
static struct dentry *ceph_mount(struct file_system_type *fs_type,
		       int flags, const char *dev_name, void *data)
S
Sage Weil 已提交
953 954
{
	struct super_block *sb;
955
	struct ceph_fs_client *fsc;
A
Al Viro 已提交
956
	struct dentry *res;
S
Sage Weil 已提交
957 958
	int err;
	int (*compare_super)(struct super_block *, void *) = ceph_compare_super;
959 960
	struct ceph_mount_options *fsopt = NULL;
	struct ceph_options *opt = NULL;
S
Sage Weil 已提交
961

A
Al Viro 已提交
962
	dout("ceph_mount\n");
963 964 965 966

#ifdef CONFIG_CEPH_FS_POSIX_ACL
	flags |= MS_POSIXACL;
#endif
967
	err = parse_mount_options(&fsopt, &opt, flags, data, dev_name);
A
Al Viro 已提交
968 969
	if (err < 0) {
		res = ERR_PTR(err);
970
		goto out_final;
A
Al Viro 已提交
971
	}
S
Sage Weil 已提交
972 973

	/* create client (which we may/may not use) */
974 975
	fsc = create_fs_client(fsopt, opt);
	if (IS_ERR(fsc)) {
A
Al Viro 已提交
976
		res = ERR_CAST(fsc);
N
Noah Watkins 已提交
977 978
		destroy_mount_options(fsopt);
		ceph_destroy_options(opt);
979 980
		goto out_final;
	}
S
Sage Weil 已提交
981

982
	err = ceph_mdsc_init(fsc);
A
Al Viro 已提交
983 984
	if (err < 0) {
		res = ERR_PTR(err);
985
		goto out;
A
Al Viro 已提交
986
	}
987 988

	if (ceph_test_opt(fsc->client, NOSHARE))
S
Sage Weil 已提交
989
		compare_super = NULL;
D
David Howells 已提交
990
	sb = sget(fs_type, compare_super, ceph_set_super, flags, fsc);
S
Sage Weil 已提交
991
	if (IS_ERR(sb)) {
A
Al Viro 已提交
992
		res = ERR_CAST(sb);
S
Sage Weil 已提交
993 994 995
		goto out;
	}

996 997 998 999 1000
	if (ceph_sb_to_client(sb) != fsc) {
		ceph_mdsc_destroy(fsc);
		destroy_fs_client(fsc);
		fsc = ceph_sb_to_client(sb);
		dout("get_sb got existing client %p\n", fsc);
S
Sage Weil 已提交
1001
	} else {
1002 1003
		dout("get_sb using new client %p\n", fsc);
		err = ceph_register_bdi(sb, fsc);
A
Al Viro 已提交
1004 1005
		if (err < 0) {
			res = ERR_PTR(err);
S
Sage Weil 已提交
1006
			goto out_splat;
A
Al Viro 已提交
1007
		}
S
Sage Weil 已提交
1008 1009
	}

1010
	res = ceph_real_mount(fsc);
A
Al Viro 已提交
1011
	if (IS_ERR(res))
S
Sage Weil 已提交
1012
		goto out_splat;
A
Al Viro 已提交
1013
	dout("root %p inode %p ino %llx.%llx\n", res,
1014
	     d_inode(res), ceph_vinop(d_inode(res)));
A
Al Viro 已提交
1015
	return res;
S
Sage Weil 已提交
1016 1017

out_splat:
1018
	ceph_mdsc_close_sessions(fsc->mdsc);
1019
	deactivate_locked_super(sb);
S
Sage Weil 已提交
1020 1021 1022
	goto out_final;

out:
1023 1024
	ceph_mdsc_destroy(fsc);
	destroy_fs_client(fsc);
S
Sage Weil 已提交
1025
out_final:
A
Al Viro 已提交
1026 1027
	dout("ceph_mount fail %ld\n", PTR_ERR(res));
	return res;
S
Sage Weil 已提交
1028 1029 1030 1031
}

static void ceph_kill_sb(struct super_block *s)
{
1032
	struct ceph_fs_client *fsc = ceph_sb_to_client(s);
1033 1034
	dev_t dev = s->s_dev;

S
Sage Weil 已提交
1035
	dout("kill_sb %p\n", s);
1036

1037
	ceph_mdsc_pre_umount(fsc->mdsc);
1038
	generic_shutdown_super(s);
1039
	ceph_mdsc_destroy(fsc);
1040

1041
	destroy_fs_client(fsc);
1042
	free_anon_bdev(dev);
S
Sage Weil 已提交
1043 1044 1045 1046 1047
}

static struct file_system_type ceph_fs_type = {
	.owner		= THIS_MODULE,
	.name		= "ceph",
A
Al Viro 已提交
1048
	.mount		= ceph_mount,
S
Sage Weil 已提交
1049 1050 1051
	.kill_sb	= ceph_kill_sb,
	.fs_flags	= FS_RENAME_DOES_D_MOVE,
};
1052
MODULE_ALIAS_FS("ceph");
S
Sage Weil 已提交
1053 1054 1055

static int __init init_ceph(void)
{
1056
	int ret = init_caches();
S
Sage Weil 已提交
1057
	if (ret)
1058
		goto out;
S
Sage Weil 已提交
1059

1060
	ceph_flock_init();
1061
	ceph_xattr_init();
S
Sage Weil 已提交
1062 1063
	ret = register_filesystem(&ceph_fs_type);
	if (ret)
I
Ilya Dryomov 已提交
1064
		goto out_xattr;
S
Sage Weil 已提交
1065

1066 1067
	pr_info("loaded (mds proto %d)\n", CEPH_MDSC_PROTOCOL);

S
Sage Weil 已提交
1068 1069
	return 0;

1070
out_xattr:
1071
	ceph_xattr_exit();
S
Sage Weil 已提交
1072 1073 1074 1075 1076 1077 1078 1079 1080
	destroy_caches();
out:
	return ret;
}

static void __exit exit_ceph(void)
{
	dout("exit_ceph\n");
	unregister_filesystem(&ceph_fs_type);
1081
	ceph_xattr_exit();
S
Sage Weil 已提交
1082 1083 1084 1085 1086 1087 1088 1089 1090 1091 1092
	destroy_caches();
}

module_init(init_ceph);
module_exit(exit_ceph);

MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
MODULE_DESCRIPTION("Ceph filesystem for Linux");
MODULE_LICENSE("GPL");