Commit | Line | Data |
---|---|---|
ba9e5e98 BH |
1 | /* |
2 | * Copyright (C) 2005, 2006 | |
27d2e149 | 3 | * Avishay Traeger (avishay@gmail.com) |
ba9e5e98 BH |
4 | * Copyright (C) 2008, 2009 |
5 | * Boaz Harrosh <bharrosh@panasas.com> | |
6 | * | |
7 | * Copyrights for code taken from ext2: | |
8 | * Copyright (C) 1992, 1993, 1994, 1995 | |
9 | * Remy Card (card@masi.ibp.fr) | |
10 | * Laboratoire MASI - Institut Blaise Pascal | |
11 | * Universite Pierre et Marie Curie (Paris VI) | |
12 | * from | |
13 | * linux/fs/minix/inode.c | |
14 | * Copyright (C) 1991, 1992 Linus Torvalds | |
15 | * | |
16 | * This file is part of exofs. | |
17 | * | |
18 | * exofs is free software; you can redistribute it and/or modify | |
19 | * it under the terms of the GNU General Public License as published by | |
20 | * the Free Software Foundation. Since it is based on ext2, and the only | |
21 | * valid version of GPL for the Linux kernel is version 2, the only valid | |
22 | * version of GPL for exofs is version 2. | |
23 | * | |
24 | * exofs is distributed in the hope that it will be useful, | |
25 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
26 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the | |
27 | * GNU General Public License for more details. | |
28 | * | |
29 | * You should have received a copy of the GNU General Public License | |
30 | * along with exofs; if not, write to the Free Software | |
31 | * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA | |
32 | */ | |
33 | ||
34 | #include <linux/string.h> | |
35 | #include <linux/parser.h> | |
36 | #include <linux/vfs.h> | |
37 | #include <linux/random.h> | |
8cf74b39 | 38 | #include <linux/exportfs.h> |
5a0e3ad6 | 39 | #include <linux/slab.h> |
ba9e5e98 BH |
40 | |
41 | #include "exofs.h" | |
42 | ||
43 | /****************************************************************************** | |
44 | * MOUNT OPTIONS | |
45 | *****************************************************************************/ | |
46 | ||
47 | /* | |
48 | * struct to hold what we get from mount options | |
49 | */ | |
50 | struct exofs_mountopt { | |
51 | const char *dev_name; | |
52 | uint64_t pid; | |
53 | int timeout; | |
54 | }; | |
55 | ||
56 | /* | |
57 | * exofs-specific mount-time options. | |
58 | */ | |
59 | enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err }; | |
60 | ||
61 | /* | |
62 | * Our mount-time options. These should ideally be 64-bit unsigned, but the | |
63 | * kernel's parsing functions do not currently support that. 32-bit should be | |
64 | * sufficient for most applications now. | |
65 | */ | |
66 | static match_table_t tokens = { | |
67 | {Opt_pid, "pid=%u"}, | |
68 | {Opt_to, "to=%u"}, | |
69 | {Opt_err, NULL} | |
70 | }; | |
71 | ||
72 | /* | |
73 | * The main option parsing method. Also makes sure that all of the mandatory | |
74 | * mount options were set. | |
75 | */ | |
76 | static int parse_options(char *options, struct exofs_mountopt *opts) | |
77 | { | |
78 | char *p; | |
79 | substring_t args[MAX_OPT_ARGS]; | |
80 | int option; | |
81 | bool s_pid = false; | |
82 | ||
83 | EXOFS_DBGMSG("parse_options %s\n", options); | |
84 | /* defaults */ | |
85 | memset(opts, 0, sizeof(*opts)); | |
86 | opts->timeout = BLK_DEFAULT_SG_TIMEOUT; | |
87 | ||
88 | while ((p = strsep(&options, ",")) != NULL) { | |
89 | int token; | |
90 | char str[32]; | |
91 | ||
92 | if (!*p) | |
93 | continue; | |
94 | ||
95 | token = match_token(p, tokens, args); | |
96 | switch (token) { | |
97 | case Opt_pid: | |
98 | if (0 == match_strlcpy(str, &args[0], sizeof(str))) | |
99 | return -EINVAL; | |
100 | opts->pid = simple_strtoull(str, NULL, 0); | |
101 | if (opts->pid < EXOFS_MIN_PID) { | |
102 | EXOFS_ERR("Partition ID must be >= %u", | |
103 | EXOFS_MIN_PID); | |
104 | return -EINVAL; | |
105 | } | |
106 | s_pid = 1; | |
107 | break; | |
108 | case Opt_to: | |
109 | if (match_int(&args[0], &option)) | |
110 | return -EINVAL; | |
111 | if (option <= 0) { | |
112 | EXOFS_ERR("Timout must be > 0"); | |
113 | return -EINVAL; | |
114 | } | |
115 | opts->timeout = option * HZ; | |
116 | break; | |
117 | } | |
118 | } | |
119 | ||
120 | if (!s_pid) { | |
121 | EXOFS_ERR("Need to specify the following options:\n"); | |
122 | EXOFS_ERR(" -o pid=pid_no_to_use\n"); | |
123 | return -EINVAL; | |
124 | } | |
125 | ||
126 | return 0; | |
127 | } | |
128 | ||
129 | /****************************************************************************** | |
130 | * INODE CACHE | |
131 | *****************************************************************************/ | |
132 | ||
133 | /* | |
134 | * Our inode cache. Isn't it pretty? | |
135 | */ | |
136 | static struct kmem_cache *exofs_inode_cachep; | |
137 | ||
138 | /* | |
139 | * Allocate an inode in the cache | |
140 | */ | |
141 | static struct inode *exofs_alloc_inode(struct super_block *sb) | |
142 | { | |
143 | struct exofs_i_info *oi; | |
144 | ||
145 | oi = kmem_cache_alloc(exofs_inode_cachep, GFP_KERNEL); | |
146 | if (!oi) | |
147 | return NULL; | |
148 | ||
149 | oi->vfs_inode.i_version = 1; | |
150 | return &oi->vfs_inode; | |
151 | } | |
152 | ||
fa0d7e3d NP |
153 | static void exofs_i_callback(struct rcu_head *head) |
154 | { | |
155 | struct inode *inode = container_of(head, struct inode, i_rcu); | |
156 | INIT_LIST_HEAD(&inode->i_dentry); | |
157 | kmem_cache_free(exofs_inode_cachep, exofs_i(inode)); | |
158 | } | |
159 | ||
ba9e5e98 BH |
160 | /* |
161 | * Remove an inode from the cache | |
162 | */ | |
163 | static void exofs_destroy_inode(struct inode *inode) | |
164 | { | |
fa0d7e3d | 165 | call_rcu(&inode->i_rcu, exofs_i_callback); |
ba9e5e98 BH |
166 | } |
167 | ||
168 | /* | |
169 | * Initialize the inode | |
170 | */ | |
171 | static void exofs_init_once(void *foo) | |
172 | { | |
173 | struct exofs_i_info *oi = foo; | |
174 | ||
175 | inode_init_once(&oi->vfs_inode); | |
176 | } | |
177 | ||
178 | /* | |
179 | * Create and initialize the inode cache | |
180 | */ | |
181 | static int init_inodecache(void) | |
182 | { | |
183 | exofs_inode_cachep = kmem_cache_create("exofs_inode_cache", | |
184 | sizeof(struct exofs_i_info), 0, | |
185 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, | |
186 | exofs_init_once); | |
187 | if (exofs_inode_cachep == NULL) | |
188 | return -ENOMEM; | |
189 | return 0; | |
190 | } | |
191 | ||
192 | /* | |
193 | * Destroy the inode cache | |
194 | */ | |
195 | static void destroy_inodecache(void) | |
196 | { | |
197 | kmem_cache_destroy(exofs_inode_cachep); | |
198 | } | |
199 | ||
200 | /****************************************************************************** | |
201 | * SUPERBLOCK FUNCTIONS | |
202 | *****************************************************************************/ | |
203 | static const struct super_operations exofs_sops; | |
8cf74b39 | 204 | static const struct export_operations exofs_export_ops; |
ba9e5e98 BH |
205 | |
206 | /* | |
207 | * Write the superblock to the OSD | |
208 | */ | |
baaf94cd | 209 | int exofs_sync_fs(struct super_block *sb, int wait) |
ba9e5e98 BH |
210 | { |
211 | struct exofs_sb_info *sbi; | |
212 | struct exofs_fscb *fscb; | |
06886a5a | 213 | struct exofs_io_state *ios; |
80e09fb9 | 214 | int ret = -ENOMEM; |
ba9e5e98 | 215 | |
ebc1ac16 | 216 | lock_super(sb); |
ba9e5e98 | 217 | sbi = sb->s_fs_info; |
06886a5a BH |
218 | fscb = &sbi->s_fscb; |
219 | ||
45d3abcb | 220 | ret = exofs_get_io_state(&sbi->layout, &ios); |
06886a5a BH |
221 | if (ret) |
222 | goto out; | |
223 | ||
04dc1e88 BH |
224 | /* Note: We only write the changing part of the fscb. .i.e upto the |
225 | * the fscb->s_dev_table_oid member. There is no read-modify-write | |
226 | * here. | |
227 | */ | |
228 | ios->length = offsetof(struct exofs_fscb, s_dev_table_oid); | |
06886a5a | 229 | memset(fscb, 0, ios->length); |
ba9e5e98 BH |
230 | fscb->s_nextid = cpu_to_le64(sbi->s_nextid); |
231 | fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles); | |
232 | fscb->s_magic = cpu_to_le16(sb->s_magic); | |
233 | fscb->s_newfs = 0; | |
04dc1e88 | 234 | fscb->s_version = EXOFS_FSCB_VER; |
ba9e5e98 | 235 | |
06886a5a BH |
236 | ios->obj.id = EXOFS_SUPER_ID; |
237 | ios->offset = 0; | |
238 | ios->kern_buff = fscb; | |
239 | ios->cred = sbi->s_cred; | |
ba9e5e98 | 240 | |
06886a5a | 241 | ret = exofs_sbi_write(ios); |
ba9e5e98 | 242 | if (unlikely(ret)) { |
06886a5a | 243 | EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__); |
ba9e5e98 BH |
244 | goto out; |
245 | } | |
246 | sb->s_dirt = 0; | |
247 | ||
248 | out: | |
06886a5a BH |
249 | EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret); |
250 | exofs_put_io_state(ios); | |
ebc1ac16 | 251 | unlock_super(sb); |
80e09fb9 CH |
252 | return ret; |
253 | } | |
254 | ||
255 | static void exofs_write_super(struct super_block *sb) | |
256 | { | |
257 | if (!(sb->s_flags & MS_RDONLY)) | |
258 | exofs_sync_fs(sb, 1); | |
259 | else | |
260 | sb->s_dirt = 0; | |
ba9e5e98 BH |
261 | } |
262 | ||
19fe294f BH |
263 | static void _exofs_print_device(const char *msg, const char *dev_path, |
264 | struct osd_dev *od, u64 pid) | |
265 | { | |
266 | const struct osd_dev_info *odi = osduld_device_info(od); | |
267 | ||
268 | printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n", | |
269 | msg, dev_path ?: "", odi->osdname, _LLU(pid)); | |
270 | } | |
271 | ||
04dc1e88 BH |
272 | void exofs_free_sbi(struct exofs_sb_info *sbi) |
273 | { | |
45d3abcb BH |
274 | while (sbi->layout.s_numdevs) { |
275 | int i = --sbi->layout.s_numdevs; | |
276 | struct osd_dev *od = sbi->layout.s_ods[i]; | |
04dc1e88 BH |
277 | |
278 | if (od) { | |
45d3abcb | 279 | sbi->layout.s_ods[i] = NULL; |
04dc1e88 BH |
280 | osduld_put_device(od); |
281 | } | |
282 | } | |
283 | kfree(sbi); | |
284 | } | |
285 | ||
ba9e5e98 BH |
286 | /* |
287 | * This function is called when the vfs is freeing the superblock. We just | |
288 | * need to free our own part. | |
289 | */ | |
290 | static void exofs_put_super(struct super_block *sb) | |
291 | { | |
292 | int num_pend; | |
293 | struct exofs_sb_info *sbi = sb->s_fs_info; | |
294 | ||
8c85e125 CH |
295 | if (sb->s_dirt) |
296 | exofs_write_super(sb); | |
297 | ||
ba9e5e98 BH |
298 | /* make sure there are no pending commands */ |
299 | for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0; | |
300 | num_pend = atomic_read(&sbi->s_curr_pending)) { | |
301 | wait_queue_head_t wq; | |
302 | init_waitqueue_head(&wq); | |
303 | wait_event_timeout(wq, | |
304 | (atomic_read(&sbi->s_curr_pending) == 0), | |
305 | msecs_to_jiffies(100)); | |
306 | } | |
307 | ||
45d3abcb BH |
308 | _exofs_print_device("Unmounting", NULL, sbi->layout.s_ods[0], |
309 | sbi->layout.s_pid); | |
04dc1e88 | 310 | |
b3d0ab7e | 311 | bdi_destroy(&sbi->bdi); |
04dc1e88 | 312 | exofs_free_sbi(sbi); |
ba9e5e98 BH |
313 | sb->s_fs_info = NULL; |
314 | } | |
315 | ||
04dc1e88 BH |
316 | static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs, |
317 | struct exofs_device_table *dt) | |
318 | { | |
5d952b83 BH |
319 | u64 stripe_length; |
320 | ||
04dc1e88 BH |
321 | sbi->data_map.odm_num_comps = |
322 | le32_to_cpu(dt->dt_data_map.cb_num_comps); | |
323 | sbi->data_map.odm_stripe_unit = | |
324 | le64_to_cpu(dt->dt_data_map.cb_stripe_unit); | |
325 | sbi->data_map.odm_group_width = | |
326 | le32_to_cpu(dt->dt_data_map.cb_group_width); | |
327 | sbi->data_map.odm_group_depth = | |
328 | le32_to_cpu(dt->dt_data_map.cb_group_depth); | |
329 | sbi->data_map.odm_mirror_cnt = | |
330 | le32_to_cpu(dt->dt_data_map.cb_mirror_cnt); | |
331 | sbi->data_map.odm_raid_algorithm = | |
332 | le32_to_cpu(dt->dt_data_map.cb_raid_algorithm); | |
333 | ||
50a76fd3 | 334 | /* FIXME: Only raid0 for now. if not so, do not mount */ |
5d952b83 BH |
335 | if (sbi->data_map.odm_num_comps != numdevs) { |
336 | EXOFS_ERR("odm_num_comps(%u) != numdevs(%u)\n", | |
337 | sbi->data_map.odm_num_comps, numdevs); | |
338 | return -EINVAL; | |
339 | } | |
340 | if (sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) { | |
341 | EXOFS_ERR("Only RAID_0 for now\n"); | |
342 | return -EINVAL; | |
343 | } | |
344 | if (0 != (numdevs % (sbi->data_map.odm_mirror_cnt + 1))) { | |
345 | EXOFS_ERR("Data Map wrong, numdevs=%d mirrors=%d\n", | |
346 | numdevs, sbi->data_map.odm_mirror_cnt); | |
347 | return -EINVAL; | |
348 | } | |
349 | ||
5d952b83 BH |
350 | if (0 != (sbi->data_map.odm_stripe_unit & ~PAGE_MASK)) { |
351 | EXOFS_ERR("Stripe Unit(0x%llx)" | |
352 | " must be Multples of PAGE_SIZE(0x%lx)\n", | |
353 | _LLU(sbi->data_map.odm_stripe_unit), PAGE_SIZE); | |
354 | return -EINVAL; | |
355 | } | |
356 | ||
357 | sbi->layout.stripe_unit = sbi->data_map.odm_stripe_unit; | |
358 | sbi->layout.mirrors_p1 = sbi->data_map.odm_mirror_cnt + 1; | |
50a76fd3 BH |
359 | |
360 | if (sbi->data_map.odm_group_width) { | |
361 | sbi->layout.group_width = sbi->data_map.odm_group_width; | |
362 | sbi->layout.group_depth = sbi->data_map.odm_group_depth; | |
363 | if (!sbi->layout.group_depth) { | |
364 | EXOFS_ERR("group_depth == 0 && group_width != 0\n"); | |
365 | return -EINVAL; | |
366 | } | |
367 | sbi->layout.group_count = sbi->data_map.odm_num_comps / | |
368 | sbi->layout.mirrors_p1 / | |
369 | sbi->data_map.odm_group_width; | |
370 | } else { | |
371 | if (sbi->data_map.odm_group_depth) { | |
372 | printk(KERN_NOTICE "Warning: group_depth ignored " | |
373 | "group_width == 0 && group_depth == %d\n", | |
374 | sbi->data_map.odm_group_depth); | |
375 | sbi->data_map.odm_group_depth = 0; | |
376 | } | |
377 | sbi->layout.group_width = sbi->data_map.odm_num_comps / | |
5d952b83 | 378 | sbi->layout.mirrors_p1; |
50a76fd3 BH |
379 | sbi->layout.group_depth = -1; |
380 | sbi->layout.group_count = 1; | |
381 | } | |
382 | ||
383 | stripe_length = (u64)sbi->layout.group_width * sbi->layout.stripe_unit; | |
384 | if (stripe_length >= (1ULL << 32)) { | |
385 | EXOFS_ERR("Total Stripe length(0x%llx)" | |
386 | " >= 32bit is not supported\n", _LLU(stripe_length)); | |
387 | return -EINVAL; | |
388 | } | |
5d952b83 BH |
389 | |
390 | return 0; | |
04dc1e88 BH |
391 | } |
392 | ||
66cd6cad | 393 | static unsigned __ra_pages(struct exofs_layout *layout) |
394 | { | |
395 | const unsigned _MIN_RA = 32; /* min 128K read-ahead */ | |
396 | unsigned ra_pages = layout->group_width * layout->stripe_unit / | |
397 | PAGE_SIZE; | |
398 | unsigned max_io_pages = exofs_max_io_pages(layout, ~0); | |
399 | ||
400 | ra_pages *= 2; /* two stripes */ | |
401 | if (ra_pages < _MIN_RA) | |
402 | ra_pages = roundup(_MIN_RA, ra_pages / 2); | |
403 | ||
404 | if (ra_pages > max_io_pages) | |
405 | ra_pages = max_io_pages; | |
406 | ||
407 | return ra_pages; | |
408 | } | |
409 | ||
04dc1e88 BH |
410 | /* @odi is valid only as long as @fscb_dev is valid */ |
411 | static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev, | |
412 | struct osd_dev_info *odi) | |
413 | { | |
414 | odi->systemid_len = le32_to_cpu(dt_dev->systemid_len); | |
415 | memcpy(odi->systemid, dt_dev->systemid, odi->systemid_len); | |
416 | ||
417 | odi->osdname_len = le32_to_cpu(dt_dev->osdname_len); | |
418 | odi->osdname = dt_dev->osdname; | |
419 | ||
420 | /* FIXME support long names. Will need a _put function */ | |
421 | if (dt_dev->long_name_offset) | |
422 | return -EINVAL; | |
423 | ||
424 | /* Make sure osdname is printable! | |
425 | * mkexofs should give us space for a null-terminator else the | |
426 | * device-table is invalid. | |
427 | */ | |
428 | if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname))) | |
429 | odi->osdname_len = sizeof(dt_dev->osdname) - 1; | |
430 | dt_dev->osdname[odi->osdname_len] = 0; | |
431 | ||
432 | /* If it's all zeros something is bad we read past end-of-obj */ | |
433 | return !(odi->systemid_len || odi->osdname_len); | |
434 | } | |
435 | ||
436 | static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi, | |
437 | unsigned table_count) | |
438 | { | |
439 | struct exofs_sb_info *sbi = *psbi; | |
440 | struct osd_dev *fscb_od; | |
45d3abcb | 441 | struct osd_obj_id obj = {.partition = sbi->layout.s_pid, |
04dc1e88 BH |
442 | .id = EXOFS_DEVTABLE_ID}; |
443 | struct exofs_device_table *dt; | |
444 | unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) + | |
445 | sizeof(*dt); | |
446 | unsigned numdevs, i; | |
447 | int ret; | |
448 | ||
449 | dt = kmalloc(table_bytes, GFP_KERNEL); | |
450 | if (unlikely(!dt)) { | |
451 | EXOFS_ERR("ERROR: allocating %x bytes for device table\n", | |
452 | table_bytes); | |
453 | return -ENOMEM; | |
454 | } | |
455 | ||
45d3abcb BH |
456 | fscb_od = sbi->layout.s_ods[0]; |
457 | sbi->layout.s_ods[0] = NULL; | |
458 | sbi->layout.s_numdevs = 0; | |
04dc1e88 BH |
459 | ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes); |
460 | if (unlikely(ret)) { | |
461 | EXOFS_ERR("ERROR: reading device table\n"); | |
462 | goto out; | |
463 | } | |
464 | ||
465 | numdevs = le64_to_cpu(dt->dt_num_devices); | |
466 | if (unlikely(!numdevs)) { | |
467 | ret = -EINVAL; | |
468 | goto out; | |
469 | } | |
470 | WARN_ON(table_count != numdevs); | |
471 | ||
472 | ret = _read_and_match_data_map(sbi, numdevs, dt); | |
473 | if (unlikely(ret)) | |
474 | goto out; | |
475 | ||
476 | if (likely(numdevs > 1)) { | |
45d3abcb | 477 | unsigned size = numdevs * sizeof(sbi->layout.s_ods[0]); |
04dc1e88 BH |
478 | |
479 | sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL); | |
480 | if (unlikely(!sbi)) { | |
481 | ret = -ENOMEM; | |
482 | goto out; | |
483 | } | |
45d3abcb BH |
484 | memset(&sbi->layout.s_ods[1], 0, |
485 | size - sizeof(sbi->layout.s_ods[0])); | |
04dc1e88 BH |
486 | *psbi = sbi; |
487 | } | |
488 | ||
489 | for (i = 0; i < numdevs; i++) { | |
490 | struct exofs_fscb fscb; | |
491 | struct osd_dev_info odi; | |
492 | struct osd_dev *od; | |
493 | ||
494 | if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) { | |
495 | EXOFS_ERR("ERROR: Read all-zeros device entry\n"); | |
496 | ret = -EINVAL; | |
497 | goto out; | |
498 | } | |
499 | ||
500 | printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n", | |
501 | i, odi.osdname); | |
502 | ||
503 | /* On all devices the device table is identical. The user can | |
504 | * specify any one of the participating devices on the command | |
505 | * line. We always keep them in device-table order. | |
506 | */ | |
507 | if (fscb_od && osduld_device_same(fscb_od, &odi)) { | |
45d3abcb BH |
508 | sbi->layout.s_ods[i] = fscb_od; |
509 | ++sbi->layout.s_numdevs; | |
04dc1e88 BH |
510 | fscb_od = NULL; |
511 | continue; | |
512 | } | |
513 | ||
514 | od = osduld_info_lookup(&odi); | |
2c722c9a | 515 | if (IS_ERR(od)) { |
04dc1e88 BH |
516 | ret = PTR_ERR(od); |
517 | EXOFS_ERR("ERROR: device requested is not found " | |
518 | "osd_name-%s =>%d\n", odi.osdname, ret); | |
519 | goto out; | |
520 | } | |
521 | ||
45d3abcb BH |
522 | sbi->layout.s_ods[i] = od; |
523 | ++sbi->layout.s_numdevs; | |
04dc1e88 BH |
524 | |
525 | /* Read the fscb of the other devices to make sure the FS | |
526 | * partition is there. | |
527 | */ | |
528 | ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, | |
529 | sizeof(fscb)); | |
530 | if (unlikely(ret)) { | |
531 | EXOFS_ERR("ERROR: Malformed participating device " | |
532 | "error reading fscb osd_name-%s\n", | |
533 | odi.osdname); | |
534 | goto out; | |
535 | } | |
536 | ||
537 | /* TODO: verify other information is correct and FS-uuid | |
538 | * matches. Benny what did you say about device table | |
539 | * generation and old devices? | |
540 | */ | |
541 | } | |
542 | ||
543 | out: | |
544 | kfree(dt); | |
545 | if (unlikely(!ret && fscb_od)) { | |
546 | EXOFS_ERR( | |
547 | "ERROR: Bad device-table container device not present\n"); | |
548 | osduld_put_device(fscb_od); | |
549 | ret = -EINVAL; | |
550 | } | |
551 | ||
552 | return ret; | |
553 | } | |
554 | ||
ba9e5e98 BH |
555 | /* |
556 | * Read the superblock from the OSD and fill in the fields | |
557 | */ | |
558 | static int exofs_fill_super(struct super_block *sb, void *data, int silent) | |
559 | { | |
560 | struct inode *root; | |
561 | struct exofs_mountopt *opts = data; | |
562 | struct exofs_sb_info *sbi; /*extended info */ | |
06886a5a | 563 | struct osd_dev *od; /* Master device */ |
ba9e5e98 | 564 | struct exofs_fscb fscb; /*on-disk superblock info */ |
ba9e5e98 | 565 | struct osd_obj_id obj; |
04dc1e88 | 566 | unsigned table_count; |
ba9e5e98 BH |
567 | int ret; |
568 | ||
569 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
570 | if (!sbi) | |
571 | return -ENOMEM; | |
ba9e5e98 | 572 | |
b3d0ab7e JA |
573 | ret = bdi_setup_and_register(&sbi->bdi, "exofs", BDI_CAP_MAP_COPY); |
574 | if (ret) | |
575 | goto free_bdi; | |
576 | ||
ba9e5e98 | 577 | /* use mount options to fill superblock */ |
06886a5a BH |
578 | od = osduld_path_lookup(opts->dev_name); |
579 | if (IS_ERR(od)) { | |
580 | ret = PTR_ERR(od); | |
ba9e5e98 BH |
581 | goto free_sbi; |
582 | } | |
583 | ||
45d3abcb | 584 | /* Default layout in case we do not have a device-table */ |
5d952b83 BH |
585 | sbi->layout.stripe_unit = PAGE_SIZE; |
586 | sbi->layout.mirrors_p1 = 1; | |
587 | sbi->layout.group_width = 1; | |
50a76fd3 BH |
588 | sbi->layout.group_depth = -1; |
589 | sbi->layout.group_count = 1; | |
45d3abcb BH |
590 | sbi->layout.s_ods[0] = od; |
591 | sbi->layout.s_numdevs = 1; | |
592 | sbi->layout.s_pid = opts->pid; | |
ba9e5e98 BH |
593 | sbi->s_timeout = opts->timeout; |
594 | ||
595 | /* fill in some other data by hand */ | |
596 | memset(sb->s_id, 0, sizeof(sb->s_id)); | |
597 | strcpy(sb->s_id, "exofs"); | |
598 | sb->s_blocksize = EXOFS_BLKSIZE; | |
599 | sb->s_blocksize_bits = EXOFS_BLKSHIFT; | |
600 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
601 | atomic_set(&sbi->s_curr_pending, 0); | |
602 | sb->s_bdev = NULL; | |
603 | sb->s_dev = 0; | |
604 | ||
45d3abcb | 605 | obj.partition = sbi->layout.s_pid; |
ba9e5e98 BH |
606 | obj.id = EXOFS_SUPER_ID; |
607 | exofs_make_credential(sbi->s_cred, &obj); | |
608 | ||
06886a5a BH |
609 | ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb)); |
610 | if (unlikely(ret)) | |
ba9e5e98 | 611 | goto free_sbi; |
ba9e5e98 BH |
612 | |
613 | sb->s_magic = le16_to_cpu(fscb.s_magic); | |
614 | sbi->s_nextid = le64_to_cpu(fscb.s_nextid); | |
615 | sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles); | |
616 | ||
617 | /* make sure what we read from the object store is correct */ | |
618 | if (sb->s_magic != EXOFS_SUPER_MAGIC) { | |
619 | if (!silent) | |
620 | EXOFS_ERR("ERROR: Bad magic value\n"); | |
621 | ret = -EINVAL; | |
622 | goto free_sbi; | |
623 | } | |
04dc1e88 BH |
624 | if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) { |
625 | EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n", | |
626 | EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version)); | |
627 | ret = -EINVAL; | |
628 | goto free_sbi; | |
629 | } | |
ba9e5e98 BH |
630 | |
631 | /* start generation numbers from a random point */ | |
632 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
633 | spin_lock_init(&sbi->s_next_gen_lock); | |
634 | ||
04dc1e88 BH |
635 | table_count = le64_to_cpu(fscb.s_dev_table_count); |
636 | if (table_count) { | |
637 | ret = exofs_read_lookup_dev_table(&sbi, table_count); | |
638 | if (unlikely(ret)) | |
639 | goto free_sbi; | |
640 | } | |
641 | ||
ba9e5e98 | 642 | /* set up operation vectors */ |
66cd6cad | 643 | sbi->bdi.ra_pages = __ra_pages(&sbi->layout); |
b3d0ab7e | 644 | sb->s_bdi = &sbi->bdi; |
06886a5a | 645 | sb->s_fs_info = sbi; |
ba9e5e98 | 646 | sb->s_op = &exofs_sops; |
8cf74b39 | 647 | sb->s_export_op = &exofs_export_ops; |
ba9e5e98 BH |
648 | root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF); |
649 | if (IS_ERR(root)) { | |
650 | EXOFS_ERR("ERROR: exofs_iget failed\n"); | |
651 | ret = PTR_ERR(root); | |
652 | goto free_sbi; | |
653 | } | |
654 | sb->s_root = d_alloc_root(root); | |
655 | if (!sb->s_root) { | |
656 | iput(root); | |
657 | EXOFS_ERR("ERROR: get root inode failed\n"); | |
658 | ret = -ENOMEM; | |
659 | goto free_sbi; | |
660 | } | |
661 | ||
662 | if (!S_ISDIR(root->i_mode)) { | |
663 | dput(sb->s_root); | |
664 | sb->s_root = NULL; | |
665 | EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n", | |
666 | root->i_mode); | |
667 | ret = -EINVAL; | |
668 | goto free_sbi; | |
669 | } | |
670 | ||
45d3abcb BH |
671 | _exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0], |
672 | sbi->layout.s_pid); | |
06886a5a | 673 | return 0; |
ba9e5e98 BH |
674 | |
675 | free_sbi: | |
b3d0ab7e JA |
676 | bdi_destroy(&sbi->bdi); |
677 | free_bdi: | |
06886a5a | 678 | EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n", |
45d3abcb | 679 | opts->dev_name, sbi->layout.s_pid, ret); |
04dc1e88 | 680 | exofs_free_sbi(sbi); |
06886a5a | 681 | return ret; |
ba9e5e98 BH |
682 | } |
683 | ||
684 | /* | |
685 | * Set up the superblock (calls exofs_fill_super eventually) | |
686 | */ | |
3c26ff6e | 687 | static struct dentry *exofs_mount(struct file_system_type *type, |
ba9e5e98 | 688 | int flags, const char *dev_name, |
3c26ff6e | 689 | void *data) |
ba9e5e98 BH |
690 | { |
691 | struct exofs_mountopt opts; | |
692 | int ret; | |
693 | ||
694 | ret = parse_options(data, &opts); | |
695 | if (ret) | |
3c26ff6e | 696 | return ERR_PTR(ret); |
ba9e5e98 BH |
697 | |
698 | opts.dev_name = dev_name; | |
3c26ff6e | 699 | return mount_nodev(type, flags, &opts, exofs_fill_super); |
ba9e5e98 BH |
700 | } |
701 | ||
702 | /* | |
703 | * Return information about the file system state in the buffer. This is used | |
704 | * by the 'df' command, for example. | |
705 | */ | |
706 | static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
707 | { | |
708 | struct super_block *sb = dentry->d_sb; | |
709 | struct exofs_sb_info *sbi = sb->s_fs_info; | |
06886a5a | 710 | struct exofs_io_state *ios; |
ba9e5e98 BH |
711 | struct osd_attr attrs[] = { |
712 | ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS, | |
713 | OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)), | |
714 | ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION, | |
715 | OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)), | |
716 | }; | |
717 | uint64_t capacity = ULLONG_MAX; | |
718 | uint64_t used = ULLONG_MAX; | |
ba9e5e98 BH |
719 | uint8_t cred_a[OSD_CAP_LEN]; |
720 | int ret; | |
721 | ||
45d3abcb | 722 | ret = exofs_get_io_state(&sbi->layout, &ios); |
06886a5a BH |
723 | if (ret) { |
724 | EXOFS_DBGMSG("exofs_get_io_state failed.\n"); | |
725 | return ret; | |
ba9e5e98 BH |
726 | } |
727 | ||
06886a5a BH |
728 | exofs_make_credential(cred_a, &ios->obj); |
729 | ios->cred = sbi->s_cred; | |
730 | ios->in_attr = attrs; | |
731 | ios->in_attr_len = ARRAY_SIZE(attrs); | |
732 | ||
733 | ret = exofs_sbi_read(ios); | |
ba9e5e98 BH |
734 | if (unlikely(ret)) |
735 | goto out; | |
736 | ||
06886a5a | 737 | ret = extract_attr_from_ios(ios, &attrs[0]); |
cae012d8 | 738 | if (likely(!ret)) { |
ba9e5e98 | 739 | capacity = get_unaligned_be64(attrs[0].val_ptr); |
cae012d8 BH |
740 | if (unlikely(!capacity)) |
741 | capacity = ULLONG_MAX; | |
742 | } else | |
ba9e5e98 BH |
743 | EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n"); |
744 | ||
06886a5a | 745 | ret = extract_attr_from_ios(ios, &attrs[1]); |
ba9e5e98 BH |
746 | if (likely(!ret)) |
747 | used = get_unaligned_be64(attrs[1].val_ptr); | |
748 | else | |
749 | EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n"); | |
750 | ||
751 | /* fill in the stats buffer */ | |
752 | buf->f_type = EXOFS_SUPER_MAGIC; | |
753 | buf->f_bsize = EXOFS_BLKSIZE; | |
cae012d8 BH |
754 | buf->f_blocks = capacity >> 9; |
755 | buf->f_bfree = (capacity - used) >> 9; | |
ba9e5e98 BH |
756 | buf->f_bavail = buf->f_bfree; |
757 | buf->f_files = sbi->s_numfiles; | |
758 | buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles; | |
759 | buf->f_namelen = EXOFS_NAME_LEN; | |
760 | ||
761 | out: | |
06886a5a | 762 | exofs_put_io_state(ios); |
ba9e5e98 BH |
763 | return ret; |
764 | } | |
765 | ||
766 | static const struct super_operations exofs_sops = { | |
767 | .alloc_inode = exofs_alloc_inode, | |
768 | .destroy_inode = exofs_destroy_inode, | |
769 | .write_inode = exofs_write_inode, | |
4ec70c9b | 770 | .evict_inode = exofs_evict_inode, |
ba9e5e98 BH |
771 | .put_super = exofs_put_super, |
772 | .write_super = exofs_write_super, | |
80e09fb9 | 773 | .sync_fs = exofs_sync_fs, |
ba9e5e98 BH |
774 | .statfs = exofs_statfs, |
775 | }; | |
776 | ||
8cf74b39 BH |
777 | /****************************************************************************** |
778 | * EXPORT OPERATIONS | |
779 | *****************************************************************************/ | |
780 | ||
781 | struct dentry *exofs_get_parent(struct dentry *child) | |
782 | { | |
783 | unsigned long ino = exofs_parent_ino(child); | |
784 | ||
785 | if (!ino) | |
786 | return NULL; | |
787 | ||
788 | return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino)); | |
789 | } | |
790 | ||
791 | static struct inode *exofs_nfs_get_inode(struct super_block *sb, | |
792 | u64 ino, u32 generation) | |
793 | { | |
794 | struct inode *inode; | |
795 | ||
796 | inode = exofs_iget(sb, ino); | |
797 | if (IS_ERR(inode)) | |
798 | return ERR_CAST(inode); | |
799 | if (generation && inode->i_generation != generation) { | |
800 | /* we didn't find the right inode.. */ | |
801 | iput(inode); | |
802 | return ERR_PTR(-ESTALE); | |
803 | } | |
804 | return inode; | |
805 | } | |
806 | ||
807 | static struct dentry *exofs_fh_to_dentry(struct super_block *sb, | |
808 | struct fid *fid, int fh_len, int fh_type) | |
809 | { | |
810 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
811 | exofs_nfs_get_inode); | |
812 | } | |
813 | ||
814 | static struct dentry *exofs_fh_to_parent(struct super_block *sb, | |
815 | struct fid *fid, int fh_len, int fh_type) | |
816 | { | |
817 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
818 | exofs_nfs_get_inode); | |
819 | } | |
820 | ||
821 | static const struct export_operations exofs_export_ops = { | |
822 | .fh_to_dentry = exofs_fh_to_dentry, | |
823 | .fh_to_parent = exofs_fh_to_parent, | |
824 | .get_parent = exofs_get_parent, | |
825 | }; | |
826 | ||
ba9e5e98 BH |
827 | /****************************************************************************** |
828 | * INSMOD/RMMOD | |
829 | *****************************************************************************/ | |
830 | ||
831 | /* | |
832 | * struct that describes this file system | |
833 | */ | |
834 | static struct file_system_type exofs_type = { | |
835 | .owner = THIS_MODULE, | |
836 | .name = "exofs", | |
3c26ff6e | 837 | .mount = exofs_mount, |
ba9e5e98 BH |
838 | .kill_sb = generic_shutdown_super, |
839 | }; | |
840 | ||
841 | static int __init init_exofs(void) | |
842 | { | |
843 | int err; | |
844 | ||
845 | err = init_inodecache(); | |
846 | if (err) | |
847 | goto out; | |
848 | ||
849 | err = register_filesystem(&exofs_type); | |
850 | if (err) | |
851 | goto out_d; | |
852 | ||
853 | return 0; | |
854 | out_d: | |
855 | destroy_inodecache(); | |
856 | out: | |
857 | return err; | |
858 | } | |
859 | ||
860 | static void __exit exit_exofs(void) | |
861 | { | |
862 | unregister_filesystem(&exofs_type); | |
863 | destroy_inodecache(); | |
864 | } | |
865 | ||
866 | MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>"); | |
867 | MODULE_DESCRIPTION("exofs"); | |
868 | MODULE_LICENSE("GPL"); | |
869 | ||
870 | module_init(init_exofs) | |
871 | module_exit(exit_exofs) |