2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com) (avishay@il.ibm.com)
4 * Copyright (C) 2005, 2006
5 * International Business Machines
6 * Copyright (C) 2008, 2009
7 * Boaz Harrosh <bharrosh@panasas.com>
9 * Copyrights for code taken from ext2:
10 * Copyright (C) 1992, 1993, 1994, 1995
11 * Remy Card (card@masi.ibp.fr)
12 * Laboratoire MASI - Institut Blaise Pascal
13 * Universite Pierre et Marie Curie (Paris VI)
15 * linux/fs/minix/inode.c
16 * Copyright (C) 1991, 1992 Linus Torvalds
18 * This file is part of exofs.
20 * exofs is free software; you can redistribute it and/or modify
21 * it under the terms of the GNU General Public License as published by
22 * the Free Software Foundation. Since it is based on ext2, and the only
23 * valid version of GPL for the Linux kernel is version 2, the only valid
24 * version of GPL for exofs is version 2.
26 * exofs is distributed in the hope that it will be useful,
27 * but WITHOUT ANY WARRANTY; without even the implied warranty of
28 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
29 * GNU General Public License for more details.
31 * You should have received a copy of the GNU General Public License
32 * along with exofs; if not, write to the Free Software
33 * Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
36 #include <linux/string.h>
37 #include <linux/parser.h>
38 #include <linux/vfs.h>
39 #include <linux/random.h>
40 #include <linux/exportfs.h>
44 /******************************************************************************
46 *****************************************************************************/
49 * struct to hold what we get from mount options
51 struct exofs_mountopt
{
58 * exofs-specific mount-time options.
60 enum { Opt_pid
, Opt_to
, Opt_mkfs
, Opt_format
, Opt_err
};
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.
67 static match_table_t tokens
= {
74 * The main option parsing method. Also makes sure that all of the mandatory
75 * mount options were set.
77 static int parse_options(char *options
, struct exofs_mountopt
*opts
)
80 substring_t args
[MAX_OPT_ARGS
];
84 EXOFS_DBGMSG("parse_options %s\n", options
);
86 memset(opts
, 0, sizeof(*opts
));
87 opts
->timeout
= BLK_DEFAULT_SG_TIMEOUT
;
89 while ((p
= strsep(&options
, ",")) != NULL
) {
96 token
= match_token(p
, tokens
, args
);
99 if (0 == match_strlcpy(str
, &args
[0], sizeof(str
)))
101 opts
->pid
= simple_strtoull(str
, NULL
, 0);
102 if (opts
->pid
< EXOFS_MIN_PID
) {
103 EXOFS_ERR("Partition ID must be >= %u",
110 if (match_int(&args
[0], &option
))
113 EXOFS_ERR("Timout must be > 0");
116 opts
->timeout
= option
* HZ
;
122 EXOFS_ERR("Need to specify the following options:\n");
123 EXOFS_ERR(" -o pid=pid_no_to_use\n");
130 /******************************************************************************
132 *****************************************************************************/
135 * Our inode cache. Isn't it pretty?
137 static struct kmem_cache
*exofs_inode_cachep
;
140 * Allocate an inode in the cache
142 static struct inode
*exofs_alloc_inode(struct super_block
*sb
)
144 struct exofs_i_info
*oi
;
146 oi
= kmem_cache_alloc(exofs_inode_cachep
, GFP_KERNEL
);
150 oi
->vfs_inode
.i_version
= 1;
151 return &oi
->vfs_inode
;
155 * Remove an inode from the cache
157 static void exofs_destroy_inode(struct inode
*inode
)
159 kmem_cache_free(exofs_inode_cachep
, exofs_i(inode
));
163 * Initialize the inode
165 static void exofs_init_once(void *foo
)
167 struct exofs_i_info
*oi
= foo
;
169 inode_init_once(&oi
->vfs_inode
);
173 * Create and initialize the inode cache
175 static int init_inodecache(void)
177 exofs_inode_cachep
= kmem_cache_create("exofs_inode_cache",
178 sizeof(struct exofs_i_info
), 0,
179 SLAB_RECLAIM_ACCOUNT
| SLAB_MEM_SPREAD
,
181 if (exofs_inode_cachep
== NULL
)
187 * Destroy the inode cache
189 static void destroy_inodecache(void)
191 kmem_cache_destroy(exofs_inode_cachep
);
194 /******************************************************************************
195 * SUPERBLOCK FUNCTIONS
196 *****************************************************************************/
197 static const struct super_operations exofs_sops
;
198 static const struct export_operations exofs_export_ops
;
201 * Write the superblock to the OSD
203 static void exofs_write_super(struct super_block
*sb
)
205 struct exofs_sb_info
*sbi
;
206 struct exofs_fscb
*fscb
;
207 struct osd_request
*or;
208 struct osd_obj_id obj
;
211 fscb
= kzalloc(sizeof(struct exofs_fscb
), GFP_KERNEL
);
213 EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
220 fscb
->s_nextid
= cpu_to_le64(sbi
->s_nextid
);
221 fscb
->s_numfiles
= cpu_to_le32(sbi
->s_numfiles
);
222 fscb
->s_magic
= cpu_to_le16(sb
->s_magic
);
225 or = osd_start_request(sbi
->s_dev
, GFP_KERNEL
);
227 EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");
231 obj
.partition
= sbi
->s_pid
;
232 obj
.id
= EXOFS_SUPER_ID
;
233 ret
= osd_req_write_kern(or, &obj
, 0, fscb
, sizeof(*fscb
));
235 EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
239 ret
= exofs_sync_op(or, sbi
->s_timeout
, sbi
->s_cred
);
241 EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
255 * This function is called when the vfs is freeing the superblock. We just
256 * need to free our own part.
258 static void exofs_put_super(struct super_block
*sb
)
261 struct exofs_sb_info
*sbi
= sb
->s_fs_info
;
266 exofs_write_super(sb
);
268 /* make sure there are no pending commands */
269 for (num_pend
= atomic_read(&sbi
->s_curr_pending
); num_pend
> 0;
270 num_pend
= atomic_read(&sbi
->s_curr_pending
)) {
271 wait_queue_head_t wq
;
272 init_waitqueue_head(&wq
);
273 wait_event_timeout(wq
,
274 (atomic_read(&sbi
->s_curr_pending
) == 0),
275 msecs_to_jiffies(100));
278 osduld_put_device(sbi
->s_dev
);
279 kfree(sb
->s_fs_info
);
280 sb
->s_fs_info
= NULL
;
286 * Read the superblock from the OSD and fill in the fields
288 static int exofs_fill_super(struct super_block
*sb
, void *data
, int silent
)
291 struct exofs_mountopt
*opts
= data
;
292 struct exofs_sb_info
*sbi
; /*extended info */
293 struct exofs_fscb fscb
; /*on-disk superblock info */
294 struct osd_request
*or = NULL
;
295 struct osd_obj_id obj
;
298 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
303 /* use mount options to fill superblock */
304 sbi
->s_dev
= osduld_path_lookup(opts
->dev_name
);
305 if (IS_ERR(sbi
->s_dev
)) {
306 ret
= PTR_ERR(sbi
->s_dev
);
311 sbi
->s_pid
= opts
->pid
;
312 sbi
->s_timeout
= opts
->timeout
;
314 /* fill in some other data by hand */
315 memset(sb
->s_id
, 0, sizeof(sb
->s_id
));
316 strcpy(sb
->s_id
, "exofs");
317 sb
->s_blocksize
= EXOFS_BLKSIZE
;
318 sb
->s_blocksize_bits
= EXOFS_BLKSHIFT
;
319 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
320 atomic_set(&sbi
->s_curr_pending
, 0);
324 /* read data from on-disk superblock object */
325 obj
.partition
= sbi
->s_pid
;
326 obj
.id
= EXOFS_SUPER_ID
;
327 exofs_make_credential(sbi
->s_cred
, &obj
);
329 or = osd_start_request(sbi
->s_dev
, GFP_KERNEL
);
333 "exofs_fill_super: osd_start_request failed.\n");
337 ret
= osd_req_read_kern(or, &obj
, 0, &fscb
, sizeof(fscb
));
341 "exofs_fill_super: osd_req_read_kern failed.\n");
346 ret
= exofs_sync_op(or, sbi
->s_timeout
, sbi
->s_cred
);
349 EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
354 sb
->s_magic
= le16_to_cpu(fscb
.s_magic
);
355 sbi
->s_nextid
= le64_to_cpu(fscb
.s_nextid
);
356 sbi
->s_numfiles
= le32_to_cpu(fscb
.s_numfiles
);
358 /* make sure what we read from the object store is correct */
359 if (sb
->s_magic
!= EXOFS_SUPER_MAGIC
) {
361 EXOFS_ERR("ERROR: Bad magic value\n");
366 /* start generation numbers from a random point */
367 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
368 spin_lock_init(&sbi
->s_next_gen_lock
);
370 /* set up operation vectors */
371 sb
->s_op
= &exofs_sops
;
372 sb
->s_export_op
= &exofs_export_ops
;
373 root
= exofs_iget(sb
, EXOFS_ROOT_ID
- EXOFS_OBJ_OFF
);
375 EXOFS_ERR("ERROR: exofs_iget failed\n");
379 sb
->s_root
= d_alloc_root(root
);
382 EXOFS_ERR("ERROR: get root inode failed\n");
387 if (!S_ISDIR(root
->i_mode
)) {
390 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
403 osduld_put_device(sbi
->s_dev
); /* NULL safe */
409 * Set up the superblock (calls exofs_fill_super eventually)
411 static int exofs_get_sb(struct file_system_type
*type
,
412 int flags
, const char *dev_name
,
413 void *data
, struct vfsmount
*mnt
)
415 struct exofs_mountopt opts
;
418 ret
= parse_options(data
, &opts
);
422 opts
.dev_name
= dev_name
;
423 return get_sb_nodev(type
, flags
, &opts
, exofs_fill_super
, mnt
);
427 * Return information about the file system state in the buffer. This is used
428 * by the 'df' command, for example.
430 static int exofs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
432 struct super_block
*sb
= dentry
->d_sb
;
433 struct exofs_sb_info
*sbi
= sb
->s_fs_info
;
434 struct osd_obj_id obj
= {sbi
->s_pid
, 0};
435 struct osd_attr attrs
[] = {
436 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS
,
437 OSD_ATTR_PQ_CAPACITY_QUOTA
, sizeof(__be64
)),
438 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION
,
439 OSD_ATTR_PI_USED_CAPACITY
, sizeof(__be64
)),
441 uint64_t capacity
= ULLONG_MAX
;
442 uint64_t used
= ULLONG_MAX
;
443 struct osd_request
*or;
444 uint8_t cred_a
[OSD_CAP_LEN
];
447 /* get used/capacity attributes */
448 exofs_make_credential(cred_a
, &obj
);
450 or = osd_start_request(sbi
->s_dev
, GFP_KERNEL
);
452 EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
456 osd_req_get_attributes(or, &obj
);
457 osd_req_add_get_attr_list(or, attrs
, ARRAY_SIZE(attrs
));
458 ret
= exofs_sync_op(or, sbi
->s_timeout
, cred_a
);
462 ret
= extract_attr_from_req(or, &attrs
[0]);
464 capacity
= get_unaligned_be64(attrs
[0].val_ptr
);
466 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
468 ret
= extract_attr_from_req(or, &attrs
[1]);
470 used
= get_unaligned_be64(attrs
[1].val_ptr
);
472 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
474 /* fill in the stats buffer */
475 buf
->f_type
= EXOFS_SUPER_MAGIC
;
476 buf
->f_bsize
= EXOFS_BLKSIZE
;
477 buf
->f_blocks
= (capacity
>> EXOFS_BLKSHIFT
);
478 buf
->f_bfree
= ((capacity
- used
) >> EXOFS_BLKSHIFT
);
479 buf
->f_bavail
= buf
->f_bfree
;
480 buf
->f_files
= sbi
->s_numfiles
;
481 buf
->f_ffree
= EXOFS_MAX_ID
- sbi
->s_numfiles
;
482 buf
->f_namelen
= EXOFS_NAME_LEN
;
489 static const struct super_operations exofs_sops
= {
490 .alloc_inode
= exofs_alloc_inode
,
491 .destroy_inode
= exofs_destroy_inode
,
492 .write_inode
= exofs_write_inode
,
493 .delete_inode
= exofs_delete_inode
,
494 .put_super
= exofs_put_super
,
495 .write_super
= exofs_write_super
,
496 .statfs
= exofs_statfs
,
499 /******************************************************************************
501 *****************************************************************************/
503 struct dentry
*exofs_get_parent(struct dentry
*child
)
505 unsigned long ino
= exofs_parent_ino(child
);
510 return d_obtain_alias(exofs_iget(child
->d_inode
->i_sb
, ino
));
513 static struct inode
*exofs_nfs_get_inode(struct super_block
*sb
,
514 u64 ino
, u32 generation
)
518 inode
= exofs_iget(sb
, ino
);
520 return ERR_CAST(inode
);
521 if (generation
&& inode
->i_generation
!= generation
) {
522 /* we didn't find the right inode.. */
524 return ERR_PTR(-ESTALE
);
529 static struct dentry
*exofs_fh_to_dentry(struct super_block
*sb
,
530 struct fid
*fid
, int fh_len
, int fh_type
)
532 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
533 exofs_nfs_get_inode
);
536 static struct dentry
*exofs_fh_to_parent(struct super_block
*sb
,
537 struct fid
*fid
, int fh_len
, int fh_type
)
539 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
540 exofs_nfs_get_inode
);
543 static const struct export_operations exofs_export_ops
= {
544 .fh_to_dentry
= exofs_fh_to_dentry
,
545 .fh_to_parent
= exofs_fh_to_parent
,
546 .get_parent
= exofs_get_parent
,
549 /******************************************************************************
551 *****************************************************************************/
554 * struct that describes this file system
556 static struct file_system_type exofs_type
= {
557 .owner
= THIS_MODULE
,
559 .get_sb
= exofs_get_sb
,
560 .kill_sb
= generic_shutdown_super
,
563 static int __init
init_exofs(void)
567 err
= init_inodecache();
571 err
= register_filesystem(&exofs_type
);
577 destroy_inodecache();
582 static void __exit
exit_exofs(void)
584 unregister_filesystem(&exofs_type
);
585 destroy_inodecache();
588 MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
589 MODULE_DESCRIPTION("exofs");
590 MODULE_LICENSE("GPL");
592 module_init(init_exofs
)
593 module_exit(exit_exofs
)