]>
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 BH |
304 | |
305 | exofs_free_sbi(sbi); | |
ba9e5e98 BH |
306 | sb->s_fs_info = NULL; |
307 | } | |
308 | ||
04dc1e88 BH |
309 | static int _read_and_match_data_map(struct exofs_sb_info *sbi, unsigned numdevs, |
310 | struct exofs_device_table *dt) | |
311 | { | |
5d952b83 BH |
312 | u64 stripe_length; |
313 | ||
04dc1e88 BH |
314 | sbi->data_map.odm_num_comps = |
315 | le32_to_cpu(dt->dt_data_map.cb_num_comps); | |
316 | sbi->data_map.odm_stripe_unit = | |
317 | le64_to_cpu(dt->dt_data_map.cb_stripe_unit); | |
318 | sbi->data_map.odm_group_width = | |
319 | le32_to_cpu(dt->dt_data_map.cb_group_width); | |
320 | sbi->data_map.odm_group_depth = | |
321 | le32_to_cpu(dt->dt_data_map.cb_group_depth); | |
322 | sbi->data_map.odm_mirror_cnt = | |
323 | le32_to_cpu(dt->dt_data_map.cb_mirror_cnt); | |
324 | sbi->data_map.odm_raid_algorithm = | |
325 | le32_to_cpu(dt->dt_data_map.cb_raid_algorithm); | |
326 | ||
50a76fd3 | 327 | /* FIXME: Only raid0 for now. if not so, do not mount */ |
5d952b83 BH |
328 | if (sbi->data_map.odm_num_comps != numdevs) { |
329 | EXOFS_ERR("odm_num_comps(%u) != numdevs(%u)\n", | |
330 | sbi->data_map.odm_num_comps, numdevs); | |
331 | return -EINVAL; | |
332 | } | |
333 | if (sbi->data_map.odm_raid_algorithm != PNFS_OSD_RAID_0) { | |
334 | EXOFS_ERR("Only RAID_0 for now\n"); | |
335 | return -EINVAL; | |
336 | } | |
337 | if (0 != (numdevs % (sbi->data_map.odm_mirror_cnt + 1))) { | |
338 | EXOFS_ERR("Data Map wrong, numdevs=%d mirrors=%d\n", | |
339 | numdevs, sbi->data_map.odm_mirror_cnt); | |
340 | return -EINVAL; | |
341 | } | |
342 | ||
5d952b83 BH |
343 | if (0 != (sbi->data_map.odm_stripe_unit & ~PAGE_MASK)) { |
344 | EXOFS_ERR("Stripe Unit(0x%llx)" | |
345 | " must be Multples of PAGE_SIZE(0x%lx)\n", | |
346 | _LLU(sbi->data_map.odm_stripe_unit), PAGE_SIZE); | |
347 | return -EINVAL; | |
348 | } | |
349 | ||
350 | sbi->layout.stripe_unit = sbi->data_map.odm_stripe_unit; | |
351 | sbi->layout.mirrors_p1 = sbi->data_map.odm_mirror_cnt + 1; | |
50a76fd3 BH |
352 | |
353 | if (sbi->data_map.odm_group_width) { | |
354 | sbi->layout.group_width = sbi->data_map.odm_group_width; | |
355 | sbi->layout.group_depth = sbi->data_map.odm_group_depth; | |
356 | if (!sbi->layout.group_depth) { | |
357 | EXOFS_ERR("group_depth == 0 && group_width != 0\n"); | |
358 | return -EINVAL; | |
359 | } | |
360 | sbi->layout.group_count = sbi->data_map.odm_num_comps / | |
361 | sbi->layout.mirrors_p1 / | |
362 | sbi->data_map.odm_group_width; | |
363 | } else { | |
364 | if (sbi->data_map.odm_group_depth) { | |
365 | printk(KERN_NOTICE "Warning: group_depth ignored " | |
366 | "group_width == 0 && group_depth == %d\n", | |
367 | sbi->data_map.odm_group_depth); | |
368 | sbi->data_map.odm_group_depth = 0; | |
369 | } | |
370 | sbi->layout.group_width = sbi->data_map.odm_num_comps / | |
5d952b83 | 371 | sbi->layout.mirrors_p1; |
50a76fd3 BH |
372 | sbi->layout.group_depth = -1; |
373 | sbi->layout.group_count = 1; | |
374 | } | |
375 | ||
376 | stripe_length = (u64)sbi->layout.group_width * sbi->layout.stripe_unit; | |
377 | if (stripe_length >= (1ULL << 32)) { | |
378 | EXOFS_ERR("Total Stripe length(0x%llx)" | |
379 | " >= 32bit is not supported\n", _LLU(stripe_length)); | |
380 | return -EINVAL; | |
381 | } | |
5d952b83 BH |
382 | |
383 | return 0; | |
04dc1e88 BH |
384 | } |
385 | ||
386 | /* @odi is valid only as long as @fscb_dev is valid */ | |
387 | static int exofs_devs_2_odi(struct exofs_dt_device_info *dt_dev, | |
388 | struct osd_dev_info *odi) | |
389 | { | |
390 | odi->systemid_len = le32_to_cpu(dt_dev->systemid_len); | |
391 | memcpy(odi->systemid, dt_dev->systemid, odi->systemid_len); | |
392 | ||
393 | odi->osdname_len = le32_to_cpu(dt_dev->osdname_len); | |
394 | odi->osdname = dt_dev->osdname; | |
395 | ||
396 | /* FIXME support long names. Will need a _put function */ | |
397 | if (dt_dev->long_name_offset) | |
398 | return -EINVAL; | |
399 | ||
400 | /* Make sure osdname is printable! | |
401 | * mkexofs should give us space for a null-terminator else the | |
402 | * device-table is invalid. | |
403 | */ | |
404 | if (unlikely(odi->osdname_len >= sizeof(dt_dev->osdname))) | |
405 | odi->osdname_len = sizeof(dt_dev->osdname) - 1; | |
406 | dt_dev->osdname[odi->osdname_len] = 0; | |
407 | ||
408 | /* If it's all zeros something is bad we read past end-of-obj */ | |
409 | return !(odi->systemid_len || odi->osdname_len); | |
410 | } | |
411 | ||
412 | static int exofs_read_lookup_dev_table(struct exofs_sb_info **psbi, | |
413 | unsigned table_count) | |
414 | { | |
415 | struct exofs_sb_info *sbi = *psbi; | |
416 | struct osd_dev *fscb_od; | |
45d3abcb | 417 | struct osd_obj_id obj = {.partition = sbi->layout.s_pid, |
04dc1e88 BH |
418 | .id = EXOFS_DEVTABLE_ID}; |
419 | struct exofs_device_table *dt; | |
420 | unsigned table_bytes = table_count * sizeof(dt->dt_dev_table[0]) + | |
421 | sizeof(*dt); | |
422 | unsigned numdevs, i; | |
423 | int ret; | |
424 | ||
425 | dt = kmalloc(table_bytes, GFP_KERNEL); | |
426 | if (unlikely(!dt)) { | |
427 | EXOFS_ERR("ERROR: allocating %x bytes for device table\n", | |
428 | table_bytes); | |
429 | return -ENOMEM; | |
430 | } | |
431 | ||
45d3abcb BH |
432 | fscb_od = sbi->layout.s_ods[0]; |
433 | sbi->layout.s_ods[0] = NULL; | |
434 | sbi->layout.s_numdevs = 0; | |
04dc1e88 BH |
435 | ret = exofs_read_kern(fscb_od, sbi->s_cred, &obj, 0, dt, table_bytes); |
436 | if (unlikely(ret)) { | |
437 | EXOFS_ERR("ERROR: reading device table\n"); | |
438 | goto out; | |
439 | } | |
440 | ||
441 | numdevs = le64_to_cpu(dt->dt_num_devices); | |
442 | if (unlikely(!numdevs)) { | |
443 | ret = -EINVAL; | |
444 | goto out; | |
445 | } | |
446 | WARN_ON(table_count != numdevs); | |
447 | ||
448 | ret = _read_and_match_data_map(sbi, numdevs, dt); | |
449 | if (unlikely(ret)) | |
450 | goto out; | |
451 | ||
452 | if (likely(numdevs > 1)) { | |
45d3abcb | 453 | unsigned size = numdevs * sizeof(sbi->layout.s_ods[0]); |
04dc1e88 BH |
454 | |
455 | sbi = krealloc(sbi, sizeof(*sbi) + size, GFP_KERNEL); | |
456 | if (unlikely(!sbi)) { | |
457 | ret = -ENOMEM; | |
458 | goto out; | |
459 | } | |
45d3abcb BH |
460 | memset(&sbi->layout.s_ods[1], 0, |
461 | size - sizeof(sbi->layout.s_ods[0])); | |
04dc1e88 BH |
462 | *psbi = sbi; |
463 | } | |
464 | ||
465 | for (i = 0; i < numdevs; i++) { | |
466 | struct exofs_fscb fscb; | |
467 | struct osd_dev_info odi; | |
468 | struct osd_dev *od; | |
469 | ||
470 | if (exofs_devs_2_odi(&dt->dt_dev_table[i], &odi)) { | |
471 | EXOFS_ERR("ERROR: Read all-zeros device entry\n"); | |
472 | ret = -EINVAL; | |
473 | goto out; | |
474 | } | |
475 | ||
476 | printk(KERN_NOTICE "Add device[%d]: osd_name-%s\n", | |
477 | i, odi.osdname); | |
478 | ||
479 | /* On all devices the device table is identical. The user can | |
480 | * specify any one of the participating devices on the command | |
481 | * line. We always keep them in device-table order. | |
482 | */ | |
483 | if (fscb_od && osduld_device_same(fscb_od, &odi)) { | |
45d3abcb BH |
484 | sbi->layout.s_ods[i] = fscb_od; |
485 | ++sbi->layout.s_numdevs; | |
04dc1e88 BH |
486 | fscb_od = NULL; |
487 | continue; | |
488 | } | |
489 | ||
490 | od = osduld_info_lookup(&odi); | |
491 | if (unlikely(IS_ERR(od))) { | |
492 | ret = PTR_ERR(od); | |
493 | EXOFS_ERR("ERROR: device requested is not found " | |
494 | "osd_name-%s =>%d\n", odi.osdname, ret); | |
495 | goto out; | |
496 | } | |
497 | ||
45d3abcb BH |
498 | sbi->layout.s_ods[i] = od; |
499 | ++sbi->layout.s_numdevs; | |
04dc1e88 BH |
500 | |
501 | /* Read the fscb of the other devices to make sure the FS | |
502 | * partition is there. | |
503 | */ | |
504 | ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, | |
505 | sizeof(fscb)); | |
506 | if (unlikely(ret)) { | |
507 | EXOFS_ERR("ERROR: Malformed participating device " | |
508 | "error reading fscb osd_name-%s\n", | |
509 | odi.osdname); | |
510 | goto out; | |
511 | } | |
512 | ||
513 | /* TODO: verify other information is correct and FS-uuid | |
514 | * matches. Benny what did you say about device table | |
515 | * generation and old devices? | |
516 | */ | |
517 | } | |
518 | ||
519 | out: | |
520 | kfree(dt); | |
521 | if (unlikely(!ret && fscb_od)) { | |
522 | EXOFS_ERR( | |
523 | "ERROR: Bad device-table container device not present\n"); | |
524 | osduld_put_device(fscb_od); | |
525 | ret = -EINVAL; | |
526 | } | |
527 | ||
528 | return ret; | |
529 | } | |
530 | ||
ba9e5e98 BH |
531 | /* |
532 | * Read the superblock from the OSD and fill in the fields | |
533 | */ | |
534 | static int exofs_fill_super(struct super_block *sb, void *data, int silent) | |
535 | { | |
536 | struct inode *root; | |
537 | struct exofs_mountopt *opts = data; | |
538 | struct exofs_sb_info *sbi; /*extended info */ | |
06886a5a | 539 | struct osd_dev *od; /* Master device */ |
ba9e5e98 | 540 | struct exofs_fscb fscb; /*on-disk superblock info */ |
ba9e5e98 | 541 | struct osd_obj_id obj; |
04dc1e88 | 542 | unsigned table_count; |
ba9e5e98 BH |
543 | int ret; |
544 | ||
545 | sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); | |
546 | if (!sbi) | |
547 | return -ENOMEM; | |
ba9e5e98 BH |
548 | |
549 | /* use mount options to fill superblock */ | |
06886a5a BH |
550 | od = osduld_path_lookup(opts->dev_name); |
551 | if (IS_ERR(od)) { | |
552 | ret = PTR_ERR(od); | |
ba9e5e98 BH |
553 | goto free_sbi; |
554 | } | |
555 | ||
45d3abcb | 556 | /* Default layout in case we do not have a device-table */ |
5d952b83 BH |
557 | sbi->layout.stripe_unit = PAGE_SIZE; |
558 | sbi->layout.mirrors_p1 = 1; | |
559 | sbi->layout.group_width = 1; | |
50a76fd3 BH |
560 | sbi->layout.group_depth = -1; |
561 | sbi->layout.group_count = 1; | |
45d3abcb BH |
562 | sbi->layout.s_ods[0] = od; |
563 | sbi->layout.s_numdevs = 1; | |
564 | sbi->layout.s_pid = opts->pid; | |
ba9e5e98 BH |
565 | sbi->s_timeout = opts->timeout; |
566 | ||
567 | /* fill in some other data by hand */ | |
568 | memset(sb->s_id, 0, sizeof(sb->s_id)); | |
569 | strcpy(sb->s_id, "exofs"); | |
570 | sb->s_blocksize = EXOFS_BLKSIZE; | |
571 | sb->s_blocksize_bits = EXOFS_BLKSHIFT; | |
572 | sb->s_maxbytes = MAX_LFS_FILESIZE; | |
573 | atomic_set(&sbi->s_curr_pending, 0); | |
574 | sb->s_bdev = NULL; | |
575 | sb->s_dev = 0; | |
576 | ||
45d3abcb | 577 | obj.partition = sbi->layout.s_pid; |
ba9e5e98 BH |
578 | obj.id = EXOFS_SUPER_ID; |
579 | exofs_make_credential(sbi->s_cred, &obj); | |
580 | ||
06886a5a BH |
581 | ret = exofs_read_kern(od, sbi->s_cred, &obj, 0, &fscb, sizeof(fscb)); |
582 | if (unlikely(ret)) | |
ba9e5e98 | 583 | goto free_sbi; |
ba9e5e98 BH |
584 | |
585 | sb->s_magic = le16_to_cpu(fscb.s_magic); | |
586 | sbi->s_nextid = le64_to_cpu(fscb.s_nextid); | |
587 | sbi->s_numfiles = le32_to_cpu(fscb.s_numfiles); | |
588 | ||
589 | /* make sure what we read from the object store is correct */ | |
590 | if (sb->s_magic != EXOFS_SUPER_MAGIC) { | |
591 | if (!silent) | |
592 | EXOFS_ERR("ERROR: Bad magic value\n"); | |
593 | ret = -EINVAL; | |
594 | goto free_sbi; | |
595 | } | |
04dc1e88 BH |
596 | if (le32_to_cpu(fscb.s_version) != EXOFS_FSCB_VER) { |
597 | EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n", | |
598 | EXOFS_FSCB_VER, le32_to_cpu(fscb.s_version)); | |
599 | ret = -EINVAL; | |
600 | goto free_sbi; | |
601 | } | |
ba9e5e98 BH |
602 | |
603 | /* start generation numbers from a random point */ | |
604 | get_random_bytes(&sbi->s_next_generation, sizeof(u32)); | |
605 | spin_lock_init(&sbi->s_next_gen_lock); | |
606 | ||
04dc1e88 BH |
607 | table_count = le64_to_cpu(fscb.s_dev_table_count); |
608 | if (table_count) { | |
609 | ret = exofs_read_lookup_dev_table(&sbi, table_count); | |
610 | if (unlikely(ret)) | |
611 | goto free_sbi; | |
612 | } | |
613 | ||
ba9e5e98 | 614 | /* set up operation vectors */ |
06886a5a | 615 | sb->s_fs_info = sbi; |
ba9e5e98 | 616 | sb->s_op = &exofs_sops; |
8cf74b39 | 617 | sb->s_export_op = &exofs_export_ops; |
ba9e5e98 BH |
618 | root = exofs_iget(sb, EXOFS_ROOT_ID - EXOFS_OBJ_OFF); |
619 | if (IS_ERR(root)) { | |
620 | EXOFS_ERR("ERROR: exofs_iget failed\n"); | |
621 | ret = PTR_ERR(root); | |
622 | goto free_sbi; | |
623 | } | |
624 | sb->s_root = d_alloc_root(root); | |
625 | if (!sb->s_root) { | |
626 | iput(root); | |
627 | EXOFS_ERR("ERROR: get root inode failed\n"); | |
628 | ret = -ENOMEM; | |
629 | goto free_sbi; | |
630 | } | |
631 | ||
632 | if (!S_ISDIR(root->i_mode)) { | |
633 | dput(sb->s_root); | |
634 | sb->s_root = NULL; | |
635 | EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n", | |
636 | root->i_mode); | |
637 | ret = -EINVAL; | |
638 | goto free_sbi; | |
639 | } | |
640 | ||
45d3abcb BH |
641 | _exofs_print_device("Mounting", opts->dev_name, sbi->layout.s_ods[0], |
642 | sbi->layout.s_pid); | |
06886a5a | 643 | return 0; |
ba9e5e98 BH |
644 | |
645 | free_sbi: | |
06886a5a | 646 | EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n", |
45d3abcb | 647 | opts->dev_name, sbi->layout.s_pid, ret); |
04dc1e88 | 648 | exofs_free_sbi(sbi); |
06886a5a | 649 | return ret; |
ba9e5e98 BH |
650 | } |
651 | ||
652 | /* | |
653 | * Set up the superblock (calls exofs_fill_super eventually) | |
654 | */ | |
655 | static int exofs_get_sb(struct file_system_type *type, | |
656 | int flags, const char *dev_name, | |
657 | void *data, struct vfsmount *mnt) | |
658 | { | |
659 | struct exofs_mountopt opts; | |
660 | int ret; | |
661 | ||
662 | ret = parse_options(data, &opts); | |
663 | if (ret) | |
664 | return ret; | |
665 | ||
666 | opts.dev_name = dev_name; | |
667 | return get_sb_nodev(type, flags, &opts, exofs_fill_super, mnt); | |
668 | } | |
669 | ||
670 | /* | |
671 | * Return information about the file system state in the buffer. This is used | |
672 | * by the 'df' command, for example. | |
673 | */ | |
674 | static int exofs_statfs(struct dentry *dentry, struct kstatfs *buf) | |
675 | { | |
676 | struct super_block *sb = dentry->d_sb; | |
677 | struct exofs_sb_info *sbi = sb->s_fs_info; | |
06886a5a | 678 | struct exofs_io_state *ios; |
ba9e5e98 BH |
679 | struct osd_attr attrs[] = { |
680 | ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS, | |
681 | OSD_ATTR_PQ_CAPACITY_QUOTA, sizeof(__be64)), | |
682 | ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION, | |
683 | OSD_ATTR_PI_USED_CAPACITY, sizeof(__be64)), | |
684 | }; | |
685 | uint64_t capacity = ULLONG_MAX; | |
686 | uint64_t used = ULLONG_MAX; | |
ba9e5e98 BH |
687 | uint8_t cred_a[OSD_CAP_LEN]; |
688 | int ret; | |
689 | ||
45d3abcb | 690 | ret = exofs_get_io_state(&sbi->layout, &ios); |
06886a5a BH |
691 | if (ret) { |
692 | EXOFS_DBGMSG("exofs_get_io_state failed.\n"); | |
693 | return ret; | |
ba9e5e98 BH |
694 | } |
695 | ||
06886a5a BH |
696 | exofs_make_credential(cred_a, &ios->obj); |
697 | ios->cred = sbi->s_cred; | |
698 | ios->in_attr = attrs; | |
699 | ios->in_attr_len = ARRAY_SIZE(attrs); | |
700 | ||
701 | ret = exofs_sbi_read(ios); | |
ba9e5e98 BH |
702 | if (unlikely(ret)) |
703 | goto out; | |
704 | ||
06886a5a | 705 | ret = extract_attr_from_ios(ios, &attrs[0]); |
cae012d8 | 706 | if (likely(!ret)) { |
ba9e5e98 | 707 | capacity = get_unaligned_be64(attrs[0].val_ptr); |
cae012d8 BH |
708 | if (unlikely(!capacity)) |
709 | capacity = ULLONG_MAX; | |
710 | } else | |
ba9e5e98 BH |
711 | EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n"); |
712 | ||
06886a5a | 713 | ret = extract_attr_from_ios(ios, &attrs[1]); |
ba9e5e98 BH |
714 | if (likely(!ret)) |
715 | used = get_unaligned_be64(attrs[1].val_ptr); | |
716 | else | |
717 | EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n"); | |
718 | ||
719 | /* fill in the stats buffer */ | |
720 | buf->f_type = EXOFS_SUPER_MAGIC; | |
721 | buf->f_bsize = EXOFS_BLKSIZE; | |
cae012d8 BH |
722 | buf->f_blocks = capacity >> 9; |
723 | buf->f_bfree = (capacity - used) >> 9; | |
ba9e5e98 BH |
724 | buf->f_bavail = buf->f_bfree; |
725 | buf->f_files = sbi->s_numfiles; | |
726 | buf->f_ffree = EXOFS_MAX_ID - sbi->s_numfiles; | |
727 | buf->f_namelen = EXOFS_NAME_LEN; | |
728 | ||
729 | out: | |
06886a5a | 730 | exofs_put_io_state(ios); |
ba9e5e98 BH |
731 | return ret; |
732 | } | |
733 | ||
734 | static const struct super_operations exofs_sops = { | |
735 | .alloc_inode = exofs_alloc_inode, | |
736 | .destroy_inode = exofs_destroy_inode, | |
737 | .write_inode = exofs_write_inode, | |
738 | .delete_inode = exofs_delete_inode, | |
739 | .put_super = exofs_put_super, | |
740 | .write_super = exofs_write_super, | |
80e09fb9 | 741 | .sync_fs = exofs_sync_fs, |
ba9e5e98 BH |
742 | .statfs = exofs_statfs, |
743 | }; | |
744 | ||
8cf74b39 BH |
745 | /****************************************************************************** |
746 | * EXPORT OPERATIONS | |
747 | *****************************************************************************/ | |
748 | ||
749 | struct dentry *exofs_get_parent(struct dentry *child) | |
750 | { | |
751 | unsigned long ino = exofs_parent_ino(child); | |
752 | ||
753 | if (!ino) | |
754 | return NULL; | |
755 | ||
756 | return d_obtain_alias(exofs_iget(child->d_inode->i_sb, ino)); | |
757 | } | |
758 | ||
759 | static struct inode *exofs_nfs_get_inode(struct super_block *sb, | |
760 | u64 ino, u32 generation) | |
761 | { | |
762 | struct inode *inode; | |
763 | ||
764 | inode = exofs_iget(sb, ino); | |
765 | if (IS_ERR(inode)) | |
766 | return ERR_CAST(inode); | |
767 | if (generation && inode->i_generation != generation) { | |
768 | /* we didn't find the right inode.. */ | |
769 | iput(inode); | |
770 | return ERR_PTR(-ESTALE); | |
771 | } | |
772 | return inode; | |
773 | } | |
774 | ||
775 | static struct dentry *exofs_fh_to_dentry(struct super_block *sb, | |
776 | struct fid *fid, int fh_len, int fh_type) | |
777 | { | |
778 | return generic_fh_to_dentry(sb, fid, fh_len, fh_type, | |
779 | exofs_nfs_get_inode); | |
780 | } | |
781 | ||
782 | static struct dentry *exofs_fh_to_parent(struct super_block *sb, | |
783 | struct fid *fid, int fh_len, int fh_type) | |
784 | { | |
785 | return generic_fh_to_parent(sb, fid, fh_len, fh_type, | |
786 | exofs_nfs_get_inode); | |
787 | } | |
788 | ||
789 | static const struct export_operations exofs_export_ops = { | |
790 | .fh_to_dentry = exofs_fh_to_dentry, | |
791 | .fh_to_parent = exofs_fh_to_parent, | |
792 | .get_parent = exofs_get_parent, | |
793 | }; | |
794 | ||
ba9e5e98 BH |
795 | /****************************************************************************** |
796 | * INSMOD/RMMOD | |
797 | *****************************************************************************/ | |
798 | ||
799 | /* | |
800 | * struct that describes this file system | |
801 | */ | |
802 | static struct file_system_type exofs_type = { | |
803 | .owner = THIS_MODULE, | |
804 | .name = "exofs", | |
805 | .get_sb = exofs_get_sb, | |
806 | .kill_sb = generic_shutdown_super, | |
807 | }; | |
808 | ||
809 | static int __init init_exofs(void) | |
810 | { | |
811 | int err; | |
812 | ||
813 | err = init_inodecache(); | |
814 | if (err) | |
815 | goto out; | |
816 | ||
817 | err = register_filesystem(&exofs_type); | |
818 | if (err) | |
819 | goto out_d; | |
820 | ||
821 | return 0; | |
822 | out_d: | |
823 | destroy_inodecache(); | |
824 | out: | |
825 | return err; | |
826 | } | |
827 | ||
828 | static void __exit exit_exofs(void) | |
829 | { | |
830 | unregister_filesystem(&exofs_type); | |
831 | destroy_inodecache(); | |
832 | } | |
833 | ||
834 | MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>"); | |
835 | MODULE_DESCRIPTION("exofs"); | |
836 | MODULE_LICENSE("GPL"); | |
837 | ||
838 | module_init(init_exofs) | |
839 | module_exit(exit_exofs) |