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 | ||
405f5571 | 34 | #include <linux/smp_lock.h> |
ba9e5e98 BH |
35 | #include <linux/string.h> |
36 | #include <linux/parser.h> | |
37 | #include <linux/vfs.h> | |
38 | #include <linux/random.h> | |
8cf74b39 | 39 | #include <linux/exportfs.h> |
5a0e3ad6 | 40 | #include <linux/slab.h> |
ba9e5e98 BH |
41 | |
42 | #include "exofs.h" | |
43 | ||
44 | /****************************************************************************** | |
45 | * MOUNT OPTIONS | |
46 | *****************************************************************************/ | |
47 | ||
48 | /* | |
49 | * struct to hold what we get from mount options | |
50 | */ | |
51 | struct exofs_mountopt { | |
52 | const char *dev_name; | |
53 | uint64_t pid; | |
54 | int timeout; | |
55 | }; | |
56 | ||
57 | /* | |
58 | * exofs-specific mount-time options. | |
59 | */ | |
60 | enum { Opt_pid, Opt_to, Opt_mkfs, Opt_format, Opt_err }; | |
61 | ||
62 | /* | |
63 | * Our mount-time options. These should ideally be 64-bit unsigned, but the | |
64 | * kernel's parsing functions do not currently support that. 32-bit should be | |
65 | * sufficient for most applications now. | |
66 | */ | |
67 | static match_table_t tokens = { | |
68 | {Opt_pid, "pid=%u"}, | |
69 | {Opt_to, "to=%u"}, | |
70 | {Opt_err, NULL} | |
71 | }; | |
72 | ||
73 | /* | |
74 | * The main option parsing method. Also makes sure that all of the mandatory | |
75 | * mount options were set. | |
76 | */ | |
77 | static int parse_options(char *options, struct exofs_mountopt *opts) | |
78 | { | |
79 | char *p; | |
80 | substring_t args[MAX_OPT_ARGS]; | |
81 | int option; | |
82 | bool s_pid = false; | |
83 | ||
84 | EXOFS_DBGMSG("parse_options %s\n", options); | |
85 | /* defaults */ | |
86 | memset(opts, 0, sizeof(*opts)); | |
87 | opts->timeout = BLK_DEFAULT_SG_TIMEOUT; | |
88 | ||
89 | while ((p = strsep(&options, ",")) != NULL) { | |
90 | int token; | |
91 | char str[32]; | |
92 | ||
93 | if (!*p) | |
94 | continue; | |
95 | ||
96 | token = match_token(p, tokens, args); | |
97 | switch (token) { | |
98 | case Opt_pid: | |
99 | if (0 == match_strlcpy(str, &args[0], sizeof(str))) | |
100 | return -EINVAL; | |
101 | opts->pid = simple_strtoull(str, NULL, 0); | |
102 | if (opts->pid < EXOFS_MIN_PID) { | |
103 | EXOFS_ERR("Partition ID must be >= %u", | |
104 | EXOFS_MIN_PID); | |
105 | return -EINVAL; | |
106 | } | |
107 | s_pid = 1; | |
108 | break; | |
109 | case Opt_to: | |
110 | if (match_int(&args[0], &option)) | |
111 | return -EINVAL; | |
112 | if (option <= 0) { | |
113 | EXOFS_ERR("Timout must be > 0"); | |
114 | return -EINVAL; | |
115 | } | |
116 | opts->timeout = option * HZ; | |
117 | break; | |
118 | } | |
119 | } | |
120 | ||
121 | if (!s_pid) { | |
122 | EXOFS_ERR("Need to specify the following options:\n"); | |
123 | EXOFS_ERR(" -o pid=pid_no_to_use\n"); | |
124 | return -EINVAL; | |
125 | } | |
126 | ||
127 | return 0; | |
128 | } | |
129 | ||
130 | /****************************************************************************** | |
131 | * INODE CACHE | |
132 | *****************************************************************************/ | |
133 | ||
134 | /* | |
135 | * Our inode cache. Isn't it pretty? | |
136 | */ | |
137 | static struct kmem_cache *exofs_inode_cachep; | |
138 | ||
139 | /* | |
140 | * Allocate an inode in the cache | |
141 | */ | |
142 | static struct inode *exofs_alloc_inode(struct super_block *sb) | |
143 | { | |
144 | struct exofs_i_info *oi; | |
145 | ||
146 | oi = kmem_cache_alloc(exofs_inode_cachep, GFP_KERNEL); | |
147 | if (!oi) | |
148 | return NULL; | |
149 | ||
150 | oi->vfs_inode.i_version = 1; | |
151 | return &oi->vfs_inode; | |
152 | } | |
153 | ||
154 | /* | |
155 | * Remove an inode from the cache | |
156 | */ | |
157 | static void exofs_destroy_inode(struct inode *inode) | |
158 | { | |
159 | kmem_cache_free(exofs_inode_cachep, exofs_i(inode)); | |
160 | } | |
161 | ||
162 | /* | |
163 | * Initialize the inode | |
164 | */ | |
165 | static void exofs_init_once(void *foo) | |
166 | { | |
167 | struct exofs_i_info *oi = foo; | |
168 | ||
169 | inode_init_once(&oi->vfs_inode); | |
170 | } | |
171 | ||
172 | /* | |
173 | * Create and initialize the inode cache | |
174 | */ | |
175 | static int init_inodecache(void) | |
176 | { | |
177 | exofs_inode_cachep = kmem_cache_create("exofs_inode_cache", | |
178 | sizeof(struct exofs_i_info), 0, | |
179 | SLAB_RECLAIM_ACCOUNT | SLAB_MEM_SPREAD, | |
180 | exofs_init_once); | |
181 | if (exofs_inode_cachep == NULL) | |
182 | return -ENOMEM; | |
183 | return 0; | |
184 | } | |
185 | ||
186 | /* | |
187 | * Destroy the inode cache | |
188 | */ | |
189 | static void destroy_inodecache(void) | |
190 | { | |
191 | kmem_cache_destroy(exofs_inode_cachep); | |
192 | } | |
193 | ||
194 | /****************************************************************************** | |
195 | * SUPERBLOCK FUNCTIONS | |
196 | *****************************************************************************/ | |
197 | static const struct super_operations exofs_sops; | |
8cf74b39 | 198 | static const struct export_operations exofs_export_ops; |
ba9e5e98 BH |
199 | |
200 | /* | |
201 | * Write the superblock to the OSD | |
202 | */ | |
baaf94cd | 203 | int exofs_sync_fs(struct super_block *sb, int wait) |
ba9e5e98 BH |
204 | { |
205 | struct exofs_sb_info *sbi; | |
206 | struct exofs_fscb *fscb; | |
06886a5a | 207 | struct exofs_io_state *ios; |
80e09fb9 | 208 | int ret = -ENOMEM; |
ba9e5e98 | 209 | |
ebc1ac16 | 210 | lock_super(sb); |
ba9e5e98 | 211 | sbi = sb->s_fs_info; |
06886a5a BH |
212 | fscb = &sbi->s_fscb; |
213 | ||
45d3abcb | 214 | ret = exofs_get_io_state(&sbi->layout, &ios); |
06886a5a BH |
215 | if (ret) |
216 | goto out; | |
217 | ||
04dc1e88 BH |
218 | /* Note: We only write the changing part of the fscb. .i.e upto the |
219 | * the fscb->s_dev_table_oid member. There is no read-modify-write | |
220 | * here. | |
221 | */ | |
222 | ios->length = offsetof(struct exofs_fscb, s_dev_table_oid); | |
06886a5a | 223 | memset(fscb, 0, ios->length); |
ba9e5e98 BH |
224 | fscb->s_nextid = cpu_to_le64(sbi->s_nextid); |
225 | fscb->s_numfiles = cpu_to_le32(sbi->s_numfiles); | |
226 | fscb->s_magic = cpu_to_le16(sb->s_magic); | |
227 | fscb->s_newfs = 0; | |
04dc1e88 | 228 | fscb->s_version = EXOFS_FSCB_VER; |
ba9e5e98 | 229 | |
06886a5a BH |
230 | ios->obj.id = EXOFS_SUPER_ID; |
231 | ios->offset = 0; | |
232 | ios->kern_buff = fscb; | |
233 | ios->cred = sbi->s_cred; | |
ba9e5e98 | 234 | |
06886a5a | 235 | ret = exofs_sbi_write(ios); |
ba9e5e98 | 236 | if (unlikely(ret)) { |
06886a5a | 237 | EXOFS_ERR("%s: exofs_sbi_write failed.\n", __func__); |
ba9e5e98 BH |
238 | goto out; |
239 | } | |
240 | sb->s_dirt = 0; | |
241 | ||
242 | out: | |
06886a5a BH |
243 | EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi->s_nextid), ret); |
244 | exofs_put_io_state(ios); | |
ebc1ac16 | 245 | unlock_super(sb); |
80e09fb9 CH |
246 | return ret; |
247 | } | |
248 | ||
249 | static void exofs_write_super(struct super_block *sb) | |
250 | { | |
251 | if (!(sb->s_flags & MS_RDONLY)) | |
252 | exofs_sync_fs(sb, 1); | |
253 | else | |
254 | sb->s_dirt = 0; | |
ba9e5e98 BH |
255 | } |
256 | ||
19fe294f BH |
257 | static void _exofs_print_device(const char *msg, const char *dev_path, |
258 | struct osd_dev *od, u64 pid) | |
259 | { | |
260 | const struct osd_dev_info *odi = osduld_device_info(od); | |
261 | ||
262 | printk(KERN_NOTICE "exofs: %s %s osd_name-%s pid-0x%llx\n", | |
263 | msg, dev_path ?: "", odi->osdname, _LLU(pid)); | |
264 | } | |
265 | ||
04dc1e88 BH |
266 | void exofs_free_sbi(struct exofs_sb_info *sbi) |
267 | { | |
45d3abcb BH |
268 | while (sbi->layout.s_numdevs) { |
269 | int i = --sbi->layout.s_numdevs; | |
270 | struct osd_dev *od = sbi->layout.s_ods[i]; | |
04dc1e88 BH |
271 | |
272 | if (od) { | |
45d3abcb | 273 | sbi->layout.s_ods[i] = NULL; |
04dc1e88 BH |
274 | osduld_put_device(od); |
275 | } | |
276 | } | |
277 | kfree(sbi); | |
278 | } | |
279 | ||
ba9e5e98 BH |
280 | /* |
281 | * This function is called when the vfs is freeing the superblock. We just | |
282 | * need to free our own part. | |
283 | */ | |
284 | static void exofs_put_super(struct super_block *sb) | |
285 | { | |
286 | int num_pend; | |
287 | struct exofs_sb_info *sbi = sb->s_fs_info; | |
288 | ||
8c85e125 CH |
289 | if (sb->s_dirt) |
290 | exofs_write_super(sb); | |
291 | ||
ba9e5e98 BH |
292 | /* make sure there are no pending commands */ |
293 | for (num_pend = atomic_read(&sbi->s_curr_pending); num_pend > 0; | |
294 | num_pend = atomic_read(&sbi->s_curr_pending)) { | |
295 | wait_queue_head_t wq; | |
296 | init_waitqueue_head(&wq); | |
297 | wait_event_timeout(wq, | |
298 | (atomic_read(&sbi->s_curr_pending) == 0), | |
299 | msecs_to_jiffies(100)); | |
300 | } | |
301 | ||
45d3abcb BH |
302 | _exofs_print_device("Unmounting", NULL, sbi->layout.s_ods[0], |
303 | sbi->layout.s_pid); | |
04dc1e88 | 304 | |
b3d0ab7e | 305 | bdi_destroy(&sbi->bdi); |
04dc1e88 | 306 | exofs_free_sbi(sbi); |
ba9e5e98 BH |
307 | sb->s_fs_info = NULL; |
308 | } | |
309 | ||
04dc1e88 BH |
310 | static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs, |
311 | struct exofs_device_table *dt) | |
312 | { | |
5d952b83 BH |
313 | u64 stripe_length; |
314 | ||
04dc1e88 BH |
315 | sbi->data_map.odm_num_comps = |
316 | le32_to_cpu(dt->dt_data_map.cb_num_comps); | |
317 | sbi->data_map.odm_stripe_unit = | |
318 | le64_to_cpu(dt->dt_data_map.cb_stripe_unit); | |
319 | sbi->data_map.odm_group_width = | |
320 | le32_to_cpu(dt->dt_data_map.cb_group_width); | |
321 | sbi->data_map.odm_group_depth = | |
322 | le32_to_cpu(dt->dt_data_map.cb_group_depth); | |
323 | sbi->data_map.odm_mirror_cnt = | |
324 | le32_to_cpu(dt->dt_data_map.cb_mirror_cnt); | |
325 | sbi->data_map.odm_raid_algorithm = | |
326 | le32_to_cpu(dt->dt_data_map.cb_raid_algorithm); | |
327 | ||
50a76fd3 | 328 | /* FIXME: Only raid0 for now. if not so, do not mount */ |
5d952b83 BH |
329 | if (sbi->data_map.odm_num_comps != numdevs) { |
330 | EXOFS_ERR("odm_num_comps(%u) != numdevs(%u)\n", | |
331 | sbi->data_map.odm_num_comps, numdevs); | |
332 | return -EINVAL; | |
333 | } | |
334 | if (sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) { | |
335 | EXOFS_ERR("Only RAID_0 for now\n"); | |
336 | return -EINVAL; | |
337 | } | |
338 | if (0 != (numdevs % (sbi->data_map.odm_mirror_cnt + 1))) { | |
339 | EXOFS_ERR("Data Map wrong, numdevs=%d mirrors=%d\n", | |
340 | numdevs, sbi->data_map.odm_mirror_cnt); | |
341 | return -EINVAL; | |
342 | } | |
343 | ||
5d952b83 BH |
344 | if (0 != (sbi->data_map.odm_stripe_unit & ~PAGE_MASK)) { |
345 | EXOFS_ERR("Stripe Unit(0x%llx)" | |
346 | " must be Multples of PAGE_SIZE(0x%lx)\n", | |
347 | _LLU(sbi->data_map.odm_stripe_unit), PAGE_SIZE); | |
348 | return -EINVAL; | |
349 | } | |
350 | ||
351 | sbi->layout.stripe_unit = sbi->data_map.odm_stripe_unit; | |
352 | sbi->layout.mirrors_p1 = sbi->data_map.odm_mirror_cnt + 1; | |
50a76fd3 BH |
353 | |
354 | if (sbi->data_map.odm_group_width) { | |
355 | sbi->layout.group_width = sbi->data_map.odm_group_width; | |
356 | sbi->layout.group_depth = sbi->data_map.odm_group_depth; | |
357 | if (!sbi->layout.group_depth) { | |
358 | EXOFS_ERR("group_depth == 0 && group_width != 0\n"); | |
359 | return -EINVAL; | |
360 | } | |
361 | sbi->layout.group_count = sbi->data_map.odm_num_comps / | |
362 | sbi->layout.mirrors_p1 / | |
363 | sbi->data_map.odm_group_width; | |
364 | } else { | |
365 | if (sbi->data_map.odm_group_depth) { | |
366 | printk(KERN_NOTICE "Warning: group_depth ignored " | |
367 | "group_width == 0 && group_depth == %d\n", | |
368 | sbi->data_map.odm_group_depth); | |
369 | sbi->data_map.odm_group_depth = 0; | |
370 | } | |
371 | sbi->layout.group_width = sbi->data_map.odm_num_comps / | |
5d952b83 | 372 | sbi->layout.mirrors_p1; |
50a76fd3 BH |
373 | sbi->layout.group_depth = -1; |
374 | sbi->layout.group_count = 1; | |
375 | } | |
376 | ||
377 | stripe_length = (u64)sbi->layout.group_width * sbi->layout.stripe_unit; | |
378 | if (stripe_length >= (1ULL << 32)) { | |
379 | EXOFS_ERR("Total Stripe length(0x%llx)" | |
380 | " >= 32bit is not supported\n", _LLU(stripe_length)); | |
381 | return -EINVAL; | |
382 | } | |
5d952b83 BH |
383 | |
384 | return 0; | |
04dc1e88 BH |
385 | } |
386 | ||
387 | /* @odi is valid only as long as @fscb_dev is valid */ | |
388 | static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev, | |
389 | struct osd_dev_info *odi) | |
390 | { | |
391 | odi->systemid_len = le32_to_cpu(dt_dev->systemid_len); | |
392 | memcpy(odi->systemid, dt_dev->systemid, odi->systemid_len); | |
393 | ||
394 | odi->osdname_len = le32_to_cpu(dt_dev->osdname_len); | |
395 | odi->osdname = dt_dev->osdname; | |
396 | ||
397 | /* FIXME support long names. Will need a _put function */ | |
398 | if (dt_dev->long_name_offset) | |
399 | return -EINVAL; | |
400 | ||
401 | /* Make sure osdname is printable! | |
402 | * mkexofs should give us space for a null-terminator else the | |
403 | * device-table is invalid. | |
404 | */ | |
405 | if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname))) | |
406 | odi->osdname_len = sizeof(dt_dev->osdname) - 1; | |
407 | dt_dev->osdname[odi->osdname_len] = 0; | |
408 | ||
409 | /* If it's all zeros something is bad we read past end-of-obj */ | |
410 | return !(odi->systemid_len || odi->osdname_len); | |
411 | } | |
412 | ||
413 | static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi, | |
414 | unsigned table_count) | |
415 | { | |
416 | struct exofs_sb_info *sbi = *psbi; | |
417 | struct osd_dev *fscb_od; | |
45d3abcb | 418 | struct osd_obj_id obj = {.partition = sbi->layout.s_pid, |
04dc1e88 BH |
419 | .id = EXOFS_DEVTABLE_ID}; |
420 | struct exofs_device_table *dt; | |
421 | unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) + | |
422 | sizeof(*dt); | |
423 | unsigned numdevs, i; | |
424 | int ret; | |
425 | ||
426 | dt = kmalloc(table_bytes, GFP_KERNEL); | |
427 | if (unlikely(!dt)) { | |
428 | EXOFS_ERR("ERROR: allocating %x bytes for device table\n", | |
429 | table_bytes); | |
430 | return -ENOMEM; | |
431 | } | |
432 | ||
45d3abcb BH |
433 | fscb_od = sbi->layout.s_ods[0]; |
434 | sbi->layout.s_ods[0] = NULL; | |
435 | sbi->layout.s_numdevs = 0; | |
04dc1e88 BH |
436 | ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes); |
437 | if (unlikely(ret)) { | |
438 | EXOFS_ERR("ERROR: reading device table\n"); | |
439 | goto out; | |
440 | } | |
441 | ||
442 | numdevs = le64_to_cpu(dt->dt_num_devices); | |
443 | if (unlikely(!numdevs)) { | |
444 | ret = -EINVAL; | |
445 | goto out; | |
446 | } | |
447 | WARN_ON(table_count != numdevs); | |
448 | ||
449 | ret = _read_and_match_data_map(sbi, numdevs, dt); | |
450 | if (unlikely(ret)) | |
451 | goto out; | |
452 | ||
453 | if (likely(numdevs > 1)) { | |
45d3abcb | 454 | unsigned size = numdevs * sizeof(sbi->layout.s_ods[0]); |
04dc1e88 BH |
455 | |
456 | sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL); | |
457 | if (unlikely(!sbi)) { | |
458 | ret = -ENOMEM; | |
459 | goto out; | |
460 | } | |
45d3abcb BH |
461 | memset(&sbi->layout.s_ods[1], 0, |
462 | size - sizeof(sbi->layout.s_ods[0])); | |
04dc1e88 BH |
463 | *psbi = sbi; |
464 | } | |
465 | ||
466 | for (i = 0; i < numdevs; i++) { | |
467 | struct exofs_fscb fscb; | |
468 | struct osd_dev_info odi; | |
469 | struct osd_dev *od; | |
470 | ||
471 | if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) { | |
472 | EXOFS_ERR("ERROR: Read all-zeros device entry\n"); | |
473 | ret = -EINVAL; | |
474 | goto out; | |
475 | } | |
476 | ||
477 | printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n", | |
478 | i, odi.osdname); | |
479 | ||
480 | /* On all devices the device table is identical. The user can | |
481 | * specify any one of the participating devices on the command | |
482 | * line. We always keep them in device-table order. | |
483 | */ | |
484 | if (fscb_od && osduld_device_same(fscb_od, &odi)) { | |
45d3abcb BH |
485 | sbi->layout.s_ods[i] = fscb_od; |
486 | ++sbi->layout.s_numdevs; | |
04dc1e88 BH |
487 | fscb_od = NULL; |
488 | continue; | |
489 | } | |
490 | ||
491 | od = osduld_info_lookup(&odi); | |
492 | if (unlikely(IS_ERR(od))) { | |
493 | ret = PTR_ERR(od); | |
494 | EXOFS_ERR("ERROR: device requested is not found " | |
495 | "osd_name-%s =>%d\n", odi.osdname, ret); | |
496 | goto out; | |
497 | } | |
498 | ||
45d3abcb BH |
499 | sbi->layout.s_ods[i] = od; |
500 | ++sbi->layout.s_numdevs; | |
04dc1e88 BH |
501 | |
502 | /* Read the fscb of the other devices to make sure the FS | |
503 | * partition is there. | |
504 | */ | |
505 | ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, | |
506 | sizeof(fscb)); | |
507 | if (unlikely(ret)) { | |
508 | EXOFS_ERR("ERROR: Malformed participating device " | |
509 | "error reading fscb osd_name-%s\n", | |
510 | odi.osdname); | |
511 | goto out; | |
512 | } | |
513 | ||
514 | /* TODO: verify other information is correct and FS-uuid | |
515 | * matches. Benny what did you say about device table | |
516 | * generation and old devices? | |
517 | */ | |
518 | } | |
519 | ||
520 | out: | |
521 | kfree(dt); | |
522 | if (unlikely(!ret && fscb_od)) { | |
523 | EXOFS_ERR( | |
524 | "ERROR: Bad device-table container device not present\n"); | |
525 | osduld_put_device(fscb_od); | |
526 | ret = -EINVAL; | |
527 | } | |
528 | ||
529 | return ret; | |
530 | } | |
531 | ||
ba9e5e98 BH |
532 | /* |
533 | * Read the superblock from the OSD and fill in the fields | |
534 | */ | |
535 | static int exofs_fill_super(struct super_block *sb, void *data, int silent) | |
536 | { | |
537 | struct inode *root; | |
538 | struct exofs_mountopt *opts = data; | |
539 | struct exofs_sb_info *sbi; /*extended info */ | |
06886a5a | 540 | struct osd_dev *od; /* Master device */ |
ba9e5e98 | 541 | struct exofs_fscb fscb; /*on-disk superblock info */ |
ba9e5e98 | 542 | struct osd_obj_id obj; |
04dc1e88 | 543 | unsigned table_count; |
ba9e5e98 BH |
544 | int ret; |
545 | ||
546 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
547 | if (!sbi) | |
548 | return -ENOMEM; | |
ba9e5e98 | 549 | |
b3d0ab7e JA |
550 | ret = bdi_setup_and_register(&sbi->bdi, "exofs", BDI_CAP_MAP_COPY); |
551 | if (ret) | |
552 | goto free_bdi; | |
553 | ||
ba9e5e98 | 554 | /* use mount options to fill superblock */ |
06886a5a BH |
555 | od = osduld_path_lookup(opts->dev_name); |
556 | if (IS_ERR(od)) { | |
557 | ret = PTR_ERR(od); | |
ba9e5e98 BH |
558 | goto free_sbi; |
559 | } | |
560 | ||
45d3abcb | 561 | /* Default layout in case we do not have a device-table */ |
5d952b83 BH |
562 | sbi->layout.stripe_unit = PAGE_SIZE; |
563 | sbi->layout.mirrors_p1 = 1; | |
564 | sbi->layout.group_width = 1; | |
50a76fd3 BH |
565 | sbi->layout.group_depth = -1; |
566 | sbi->layout.group_count = 1; | |
45d3abcb BH |
567 | sbi->layout.s_ods[0] = od; |
568 | sbi->layout.s_numdevs = 1; | |
569 | sbi->layout.s_pid = opts->pid; | |
ba9e5e98 BH |
570 | sbi->s_timeout = opts->timeout; |
571 | ||
572 | /* fill in some other data by hand */ | |
573 | memset(sb->s_id, 0, sizeof(sb->s_id)); | |
574 | strcpy(sb->s_id, "exofs"); | |
575 | sb->s_blocksize = EXOFS_BLKSIZE; | |
576 | sb->s_blocksize_bits = EXOFS_BLKSHIFT; | |
577 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
578 | atomic_set(&sbi->s_curr_pending, 0); | |
579 | sb->s_bdev = NULL; | |
580 | sb->s_dev = 0; | |
581 | ||
45d3abcb | 582 | obj.partition = sbi->layout.s_pid; |
ba9e5e98 BH |
583 | obj.id = EXOFS_SUPER_ID; |
584 | exofs_make_credential(sbi->s_cred, &obj); | |
585 | ||
06886a5a BH |
586 | ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb)); |
587 | if (unlikely(ret)) | |
ba9e5e98 | 588 | goto free_sbi; |
ba9e5e98 BH |
589 | |
590 | sb->s_magic = le16_to_cpu(fscb.s_magic); | |
591 | sbi->s_nextid = le64_to_cpu(fscb.s_nextid); | |
592 | sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles); | |
593 | ||
594 | /* make sure what we read from the object store is correct */ | |
595 | if (sb->s_magic != EXOFS_SUPER_MAGIC) { | |
596 | if (!silent) | |
597 | EXOFS_ERR("ERROR: Bad magic value\n"); | |
598 | ret = -EINVAL; | |
599 | goto free_sbi; | |
600 | } | |
04dc1e88 BH |
601 | if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) { |
602 | EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n", | |
603 | EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version)); | |
604 | ret = -EINVAL; | |
605 | goto free_sbi; | |
606 | } | |
ba9e5e98 BH |
607 | |
608 | /* start generation numbers from a random point */ | |
609 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
610 | spin_lock_init(&sbi->s_next_gen_lock); | |
611 | ||
04dc1e88 BH |
612 | table_count = le64_to_cpu(fscb.s_dev_table_count); |
613 | if (table_count) { | |
614 | ret = exofs_read_lookup_dev_table(&sbi, table_count); | |
615 | if (unlikely(ret)) | |
616 | goto free_sbi; | |
617 | } | |
618 | ||
ba9e5e98 | 619 | /* set up operation vectors */ |
b3d0ab7e | 620 | sb->s_bdi = &sbi->bdi; |
06886a5a | 621 | sb->s_fs_info = sbi; |
ba9e5e98 | 622 | sb->s_op = &exofs_sops; |
8cf74b39 | 623 | sb->s_export_op = &exofs_export_ops; |
ba9e5e98 BH |
624 | root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF); |
625 | if (IS_ERR(root)) { | |
626 | EXOFS_ERR("ERROR: exofs_iget failed\n"); | |
627 | ret = PTR_ERR(root); | |
628 | goto free_sbi; | |
629 | } | |
630 | sb->s_root = d_alloc_root(root); | |
631 | if (!sb->s_root) { | |
632 | iput(root); | |
633 | EXOFS_ERR("ERROR: get root inode failed\n"); | |
634 | ret = -ENOMEM; | |
635 | goto free_sbi; | |
636 | } | |
637 | ||
638 | if (!S_ISDIR(root->i_mode)) { | |
639 | dput(sb->s_root); | |
640 | sb->s_root = NULL; | |
641 | EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n", | |
642 | root->i_mode); | |
643 | ret = -EINVAL; | |
644 | goto free_sbi; | |
645 | } | |
646 | ||
45d3abcb BH |
647 | _exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0], |
648 | sbi->layout.s_pid); | |
06886a5a | 649 | return 0; |
ba9e5e98 BH |
650 | |
651 | free_sbi: | |
b3d0ab7e JA |
652 | bdi_destroy(&sbi->bdi); |
653 | free_bdi: | |
06886a5a | 654 | EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n", |
45d3abcb | 655 | opts->dev_name, sbi->layout.s_pid, ret); |
04dc1e88 | 656 | exofs_free_sbi(sbi); |
06886a5a | 657 | return ret; |
ba9e5e98 BH |
658 | } |
659 | ||
660 | /* | |
661 | * Set up the superblock (calls exofs_fill_super eventually) | |
662 | */ | |
663 | static int exofs_get_sb(struct file_system_type *type, | |
664 | int flags, const char *dev_name, | |
665 | void *data, struct vfsmount *mnt) | |
666 | { | |
667 | struct exofs_mountopt opts; | |
668 | int ret; | |
669 | ||
670 | ret = parse_options(data, &opts); | |
671 | if (ret) | |
672 | return ret; | |
673 | ||
674 | opts.dev_name = dev_name; | |
675 | return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt); | |
676 | } | |
677 | ||
678 | /* | |
679 | * Return information about the file system state in the buffer. This is used | |
680 | * by the 'df' command, for example. | |
681 | */ | |
682 | static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
683 | { | |
684 | struct super_block *sb = dentry->d_sb; | |
685 | struct exofs_sb_info *sbi = sb->s_fs_info; | |
06886a5a | 686 | struct exofs_io_state *ios; |
ba9e5e98 BH |
687 | struct osd_attr attrs[] = { |
688 | ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS, | |
689 | OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)), | |
690 | ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION, | |
691 | OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)), | |
692 | }; | |
693 | uint64_t capacity = ULLONG_MAX; | |
694 | uint64_t used = ULLONG_MAX; | |
ba9e5e98 BH |
695 | uint8_t cred_a[OSD_CAP_LEN]; |
696 | int ret; | |
697 | ||
45d3abcb | 698 | ret = exofs_get_io_state(&sbi->layout, &ios); |
06886a5a BH |
699 | if (ret) { |
700 | EXOFS_DBGMSG("exofs_get_io_state failed.\n"); | |
701 | return ret; | |
ba9e5e98 BH |
702 | } |
703 | ||
06886a5a BH |
704 | exofs_make_credential(cred_a, &ios->obj); |
705 | ios->cred = sbi->s_cred; | |
706 | ios->in_attr = attrs; | |
707 | ios->in_attr_len = ARRAY_SIZE(attrs); | |
708 | ||
709 | ret = exofs_sbi_read(ios); | |
ba9e5e98 BH |
710 | if (unlikely(ret)) |
711 | goto out; | |
712 | ||
06886a5a | 713 | ret = extract_attr_from_ios(ios, &attrs[0]); |
cae012d8 | 714 | if (likely(!ret)) { |
ba9e5e98 | 715 | capacity = get_unaligned_be64(attrs[0].val_ptr); |
cae012d8 BH |
716 | if (unlikely(!capacity)) |
717 | capacity = ULLONG_MAX; | |
718 | } else | |
ba9e5e98 BH |
719 | EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n"); |
720 | ||
06886a5a | 721 | ret = extract_attr_from_ios(ios, &attrs[1]); |
ba9e5e98 BH |
722 | if (likely(!ret)) |
723 | used = get_unaligned_be64(attrs[1].val_ptr); | |
724 | else | |
725 | EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n"); | |
726 | ||
727 | /* fill in the stats buffer */ | |
728 | buf->f_type = EXOFS_SUPER_MAGIC; | |
729 | buf->f_bsize = EXOFS_BLKSIZE; | |
cae012d8 BH |
730 | buf->f_blocks = capacity >> 9; |
731 | buf->f_bfree = (capacity - used) >> 9; | |
ba9e5e98 BH |
732 | buf->f_bavail = buf->f_bfree; |
733 | buf->f_files = sbi->s_numfiles; | |
734 | buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles; | |
735 | buf->f_namelen = EXOFS_NAME_LEN; | |
736 | ||
737 | out: | |
06886a5a | 738 | exofs_put_io_state(ios); |
ba9e5e98 BH |
739 | return ret; |
740 | } | |
741 | ||
742 | static const struct super_operations exofs_sops = { | |
743 | .alloc_inode = exofs_alloc_inode, | |
744 | .destroy_inode = exofs_destroy_inode, | |
745 | .write_inode = exofs_write_inode, | |
746 | .delete_inode = exofs_delete_inode, | |
747 | .put_super = exofs_put_super, | |
748 | .write_super = exofs_write_super, | |
80e09fb9 | 749 | .sync_fs = exofs_sync_fs, |
ba9e5e98 BH |
750 | .statfs = exofs_statfs, |
751 | }; | |
752 | ||
8cf74b39 BH |
753 | /****************************************************************************** |
754 | * EXPORT OPERATIONS | |
755 | *****************************************************************************/ | |
756 | ||
757 | struct dentry *exofs_get_parent(struct dentry *child) | |
758 | { | |
759 | unsigned long ino = exofs_parent_ino(child); | |
760 | ||
761 | if (!ino) | |
762 | return NULL; | |
763 | ||
764 | return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino)); | |
765 | } | |
766 | ||
767 | static struct inode *exofs_nfs_get_inode(struct super_block *sb, | |
768 | u64 ino, u32 generation) | |
769 | { | |
770 | struct inode *inode; | |
771 | ||
772 | inode = exofs_iget(sb, ino); | |
773 | if (IS_ERR(inode)) | |
774 | return ERR_CAST(inode); | |
775 | if (generation && inode->i_generation != generation) { | |
776 | /* we didn't find the right inode.. */ | |
777 | iput(inode); | |
778 | return ERR_PTR(-ESTALE); | |
779 | } | |
780 | return inode; | |
781 | } | |
782 | ||
783 | static struct dentry *exofs_fh_to_dentry(struct super_block *sb, | |
784 | struct fid *fid, int fh_len, int fh_type) | |
785 | { | |
786 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
787 | exofs_nfs_get_inode); | |
788 | } | |
789 | ||
790 | static struct dentry *exofs_fh_to_parent(struct super_block *sb, | |
791 | struct fid *fid, int fh_len, int fh_type) | |
792 | { | |
793 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
794 | exofs_nfs_get_inode); | |
795 | } | |
796 | ||
797 | static const struct export_operations exofs_export_ops = { | |
798 | .fh_to_dentry = exofs_fh_to_dentry, | |
799 | .fh_to_parent = exofs_fh_to_parent, | |
800 | .get_parent = exofs_get_parent, | |
801 | }; | |
802 | ||
ba9e5e98 BH |
803 | /****************************************************************************** |
804 | * INSMOD/RMMOD | |
805 | *****************************************************************************/ | |
806 | ||
807 | /* | |
808 | * struct that describes this file system | |
809 | */ | |
810 | static struct file_system_type exofs_type = { | |
811 | .owner = THIS_MODULE, | |
812 | .name = "exofs", | |
813 | .get_sb = exofs_get_sb, | |
814 | .kill_sb = generic_shutdown_super, | |
815 | }; | |
816 | ||
817 | static int __init init_exofs(void) | |
818 | { | |
819 | int err; | |
820 | ||
821 | err = init_inodecache(); | |
822 | if (err) | |
823 | goto out; | |
824 | ||
825 | err = register_filesystem(&exofs_type); | |
826 | if (err) | |
827 | goto out_d; | |
828 | ||
829 | return 0; | |
830 | out_d: | |
831 | destroy_inodecache(); | |
832 | out: | |
833 | return err; | |
834 | } | |
835 | ||
836 | static void __exit exit_exofs(void) | |
837 | { | |
838 | unregister_filesystem(&exofs_type); | |
839 | destroy_inodecache(); | |
840 | } | |
841 | ||
842 | MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>"); | |
843 | MODULE_DESCRIPTION("exofs"); | |
844 | MODULE_LICENSE("GPL"); | |
845 | ||
846 | module_init(init_exofs) | |
847 | module_exit(exit_exofs) |