2 * Copyright (C) 2005, 2006
3 * Avishay Traeger (avishay@gmail.com)
4 * Copyright (C) 2008, 2009
5 * Boaz Harrosh <bharrosh@panasas.com>
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)
13 * linux/fs/minix/inode.c
14 * Copyright (C) 1991, 1992 Linus Torvalds
16 * This file is part of exofs.
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.
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.
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
34 #include <linux/smp_lock.h>
35 #include <linux/string.h>
36 #include <linux/parser.h>
37 #include <linux/vfs.h>
38 #include <linux/random.h>
39 #include <linux/exportfs.h>
43 /******************************************************************************
45 *****************************************************************************/
48 * struct to hold what we get from mount options
50 struct exofs_mountopt
{
57 * exofs-specific mount-time options.
59 enum { Opt_pid
, Opt_to
, Opt_mkfs
, Opt_format
, Opt_err
};
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.
66 static match_table_t tokens
= {
73 * The main option parsing method. Also makes sure that all of the mandatory
74 * mount options were set.
76 static int parse_options(char *options
, struct exofs_mountopt
*opts
)
79 substring_t args
[MAX_OPT_ARGS
];
83 EXOFS_DBGMSG("parse_options %s\n", options
);
85 memset(opts
, 0, sizeof(*opts
));
86 opts
->timeout
= BLK_DEFAULT_SG_TIMEOUT
;
88 while ((p
= strsep(&options
, ",")) != NULL
) {
95 token
= match_token(p
, tokens
, args
);
98 if (0 == match_strlcpy(str
, &args
[0], sizeof(str
)))
100 opts
->pid
= simple_strtoull(str
, NULL
, 0);
101 if (opts
->pid
< EXOFS_MIN_PID
) {
102 EXOFS_ERR("Partition ID must be >= %u",
109 if (match_int(&args
[0], &option
))
112 EXOFS_ERR("Timout must be > 0");
115 opts
->timeout
= option
* HZ
;
121 EXOFS_ERR("Need to specify the following options:\n");
122 EXOFS_ERR(" -o pid=pid_no_to_use\n");
129 /******************************************************************************
131 *****************************************************************************/
134 * Our inode cache. Isn't it pretty?
136 static struct kmem_cache
*exofs_inode_cachep
;
139 * Allocate an inode in the cache
141 static struct inode
*exofs_alloc_inode(struct super_block
*sb
)
143 struct exofs_i_info
*oi
;
145 oi
= kmem_cache_alloc(exofs_inode_cachep
, GFP_KERNEL
);
149 oi
->vfs_inode
.i_version
= 1;
150 return &oi
->vfs_inode
;
154 * Remove an inode from the cache
156 static void exofs_destroy_inode(struct inode
*inode
)
158 kmem_cache_free(exofs_inode_cachep
, exofs_i(inode
));
162 * Initialize the inode
164 static void exofs_init_once(void *foo
)
166 struct exofs_i_info
*oi
= foo
;
168 inode_init_once(&oi
->vfs_inode
);
172 * Create and initialize the inode cache
174 static int init_inodecache(void)
176 exofs_inode_cachep
= kmem_cache_create("exofs_inode_cache",
177 sizeof(struct exofs_i_info
), 0,
178 SLAB_RECLAIM_ACCOUNT
| SLAB_MEM_SPREAD
,
180 if (exofs_inode_cachep
== NULL
)
186 * Destroy the inode cache
188 static void destroy_inodecache(void)
190 kmem_cache_destroy(exofs_inode_cachep
);
193 /******************************************************************************
194 * SUPERBLOCK FUNCTIONS
195 *****************************************************************************/
196 static const struct super_operations exofs_sops
;
197 static const struct export_operations exofs_export_ops
;
200 * Write the superblock to the OSD
202 int exofs_sync_fs(struct super_block
*sb
, int wait
)
204 struct exofs_sb_info
*sbi
;
205 struct exofs_fscb
*fscb
;
206 struct osd_request
*or;
207 struct osd_obj_id obj
;
210 fscb
= kzalloc(sizeof(struct exofs_fscb
), GFP_KERNEL
);
212 EXOFS_ERR("exofs_write_super: memory allocation failed.\n");
218 fscb
->s_nextid
= cpu_to_le64(sbi
->s_nextid
);
219 fscb
->s_numfiles
= cpu_to_le32(sbi
->s_numfiles
);
220 fscb
->s_magic
= cpu_to_le16(sb
->s_magic
);
223 or = osd_start_request(sbi
->s_dev
, GFP_KERNEL
);
225 EXOFS_ERR("exofs_write_super: osd_start_request failed.\n");
229 obj
.partition
= sbi
->s_pid
;
230 obj
.id
= EXOFS_SUPER_ID
;
231 ret
= osd_req_write_kern(or, &obj
, 0, fscb
, sizeof(*fscb
));
233 EXOFS_ERR("exofs_write_super: osd_req_write_kern failed.\n");
237 ret
= exofs_sync_op(or, sbi
->s_timeout
, sbi
->s_cred
);
239 EXOFS_ERR("exofs_write_super: exofs_sync_op failed.\n");
252 static void exofs_write_super(struct super_block
*sb
)
254 if (!(sb
->s_flags
& MS_RDONLY
))
255 exofs_sync_fs(sb
, 1);
260 static void _exofs_print_device(const char *msg
, const char *dev_path
,
261 struct osd_dev
*od
, u64 pid
)
263 const struct osd_dev_info
*odi
= osduld_device_info(od
);
265 printk(KERN_NOTICE
"exofs: %s %s osd_name-%s pid-0x%llx\n",
266 msg
, dev_path
?: "", odi
->osdname
, _LLU(pid
));
270 * This function is called when the vfs is freeing the superblock. We just
271 * need to free our own part.
273 static void exofs_put_super(struct super_block
*sb
)
276 struct exofs_sb_info
*sbi
= sb
->s_fs_info
;
279 exofs_write_super(sb
);
281 /* make sure there are no pending commands */
282 for (num_pend
= atomic_read(&sbi
->s_curr_pending
); num_pend
> 0;
283 num_pend
= atomic_read(&sbi
->s_curr_pending
)) {
284 wait_queue_head_t wq
;
285 init_waitqueue_head(&wq
);
286 wait_event_timeout(wq
,
287 (atomic_read(&sbi
->s_curr_pending
) == 0),
288 msecs_to_jiffies(100));
291 _exofs_print_device("Unmounting", NULL
, sbi
->s_dev
, sbi
->s_pid
);
292 osduld_put_device(sbi
->s_dev
);
293 kfree(sb
->s_fs_info
);
294 sb
->s_fs_info
= NULL
;
298 * Read the superblock from the OSD and fill in the fields
300 static int exofs_fill_super(struct super_block
*sb
, void *data
, int silent
)
303 struct exofs_mountopt
*opts
= data
;
304 struct exofs_sb_info
*sbi
; /*extended info */
305 struct exofs_fscb fscb
; /*on-disk superblock info */
306 struct osd_request
*or = NULL
;
307 struct osd_obj_id obj
;
310 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
315 /* use mount options to fill superblock */
316 sbi
->s_dev
= osduld_path_lookup(opts
->dev_name
);
317 if (IS_ERR(sbi
->s_dev
)) {
318 ret
= PTR_ERR(sbi
->s_dev
);
323 sbi
->s_pid
= opts
->pid
;
324 sbi
->s_timeout
= opts
->timeout
;
326 /* fill in some other data by hand */
327 memset(sb
->s_id
, 0, sizeof(sb
->s_id
));
328 strcpy(sb
->s_id
, "exofs");
329 sb
->s_blocksize
= EXOFS_BLKSIZE
;
330 sb
->s_blocksize_bits
= EXOFS_BLKSHIFT
;
331 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
332 atomic_set(&sbi
->s_curr_pending
, 0);
336 /* read data from on-disk superblock object */
337 obj
.partition
= sbi
->s_pid
;
338 obj
.id
= EXOFS_SUPER_ID
;
339 exofs_make_credential(sbi
->s_cred
, &obj
);
341 or = osd_start_request(sbi
->s_dev
, GFP_KERNEL
);
345 "exofs_fill_super: osd_start_request failed.\n");
349 ret
= osd_req_read_kern(or, &obj
, 0, &fscb
, sizeof(fscb
));
353 "exofs_fill_super: osd_req_read_kern failed.\n");
358 ret
= exofs_sync_op(or, sbi
->s_timeout
, sbi
->s_cred
);
361 EXOFS_ERR("exofs_fill_super: exofs_sync_op failed.\n");
366 sb
->s_magic
= le16_to_cpu(fscb
.s_magic
);
367 sbi
->s_nextid
= le64_to_cpu(fscb
.s_nextid
);
368 sbi
->s_numfiles
= le32_to_cpu(fscb
.s_numfiles
);
370 /* make sure what we read from the object store is correct */
371 if (sb
->s_magic
!= EXOFS_SUPER_MAGIC
) {
373 EXOFS_ERR("ERROR: Bad magic value\n");
378 /* start generation numbers from a random point */
379 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
380 spin_lock_init(&sbi
->s_next_gen_lock
);
382 /* set up operation vectors */
383 sb
->s_op
= &exofs_sops
;
384 sb
->s_export_op
= &exofs_export_ops
;
385 root
= exofs_iget(sb
, EXOFS_ROOT_ID
- EXOFS_OBJ_OFF
);
387 EXOFS_ERR("ERROR: exofs_iget failed\n");
391 sb
->s_root
= d_alloc_root(root
);
394 EXOFS_ERR("ERROR: get root inode failed\n");
399 if (!S_ISDIR(root
->i_mode
)) {
402 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
408 _exofs_print_device("Mounting", opts
->dev_name
, sbi
->s_dev
, sbi
->s_pid
);
416 osduld_put_device(sbi
->s_dev
); /* NULL safe */
422 * Set up the superblock (calls exofs_fill_super eventually)
424 static int exofs_get_sb(struct file_system_type
*type
,
425 int flags
, const char *dev_name
,
426 void *data
, struct vfsmount
*mnt
)
428 struct exofs_mountopt opts
;
431 ret
= parse_options(data
, &opts
);
435 opts
.dev_name
= dev_name
;
436 return get_sb_nodev(type
, flags
, &opts
, exofs_fill_super
, mnt
);
440 * Return information about the file system state in the buffer. This is used
441 * by the 'df' command, for example.
443 static int exofs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
445 struct super_block
*sb
= dentry
->d_sb
;
446 struct exofs_sb_info
*sbi
= sb
->s_fs_info
;
447 struct osd_obj_id obj
= {sbi
->s_pid
, 0};
448 struct osd_attr attrs
[] = {
449 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS
,
450 OSD_ATTR_PQ_CAPACITY_QUOTA
, sizeof(__be64
)),
451 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION
,
452 OSD_ATTR_PI_USED_CAPACITY
, sizeof(__be64
)),
454 uint64_t capacity
= ULLONG_MAX
;
455 uint64_t used
= ULLONG_MAX
;
456 struct osd_request
*or;
457 uint8_t cred_a
[OSD_CAP_LEN
];
460 /* get used/capacity attributes */
461 exofs_make_credential(cred_a
, &obj
);
463 or = osd_start_request(sbi
->s_dev
, GFP_KERNEL
);
465 EXOFS_DBGMSG("exofs_statfs: osd_start_request failed.\n");
469 osd_req_get_attributes(or, &obj
);
470 osd_req_add_get_attr_list(or, attrs
, ARRAY_SIZE(attrs
));
471 ret
= exofs_sync_op(or, sbi
->s_timeout
, cred_a
);
475 ret
= extract_attr_from_req(or, &attrs
[0]);
477 capacity
= get_unaligned_be64(attrs
[0].val_ptr
);
479 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
481 ret
= extract_attr_from_req(or, &attrs
[1]);
483 used
= get_unaligned_be64(attrs
[1].val_ptr
);
485 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
487 /* fill in the stats buffer */
488 buf
->f_type
= EXOFS_SUPER_MAGIC
;
489 buf
->f_bsize
= EXOFS_BLKSIZE
;
490 buf
->f_blocks
= (capacity
>> EXOFS_BLKSHIFT
);
491 buf
->f_bfree
= ((capacity
- used
) >> EXOFS_BLKSHIFT
);
492 buf
->f_bavail
= buf
->f_bfree
;
493 buf
->f_files
= sbi
->s_numfiles
;
494 buf
->f_ffree
= EXOFS_MAX_ID
- sbi
->s_numfiles
;
495 buf
->f_namelen
= EXOFS_NAME_LEN
;
502 static const struct super_operations exofs_sops
= {
503 .alloc_inode
= exofs_alloc_inode
,
504 .destroy_inode
= exofs_destroy_inode
,
505 .write_inode
= exofs_write_inode
,
506 .delete_inode
= exofs_delete_inode
,
507 .put_super
= exofs_put_super
,
508 .write_super
= exofs_write_super
,
509 .sync_fs
= exofs_sync_fs
,
510 .statfs
= exofs_statfs
,
513 /******************************************************************************
515 *****************************************************************************/
517 struct dentry
*exofs_get_parent(struct dentry
*child
)
519 unsigned long ino
= exofs_parent_ino(child
);
524 return d_obtain_alias(exofs_iget(child
->d_inode
->i_sb
, ino
));
527 static struct inode
*exofs_nfs_get_inode(struct super_block
*sb
,
528 u64 ino
, u32 generation
)
532 inode
= exofs_iget(sb
, ino
);
534 return ERR_CAST(inode
);
535 if (generation
&& inode
->i_generation
!= generation
) {
536 /* we didn't find the right inode.. */
538 return ERR_PTR(-ESTALE
);
543 static struct dentry
*exofs_fh_to_dentry(struct super_block
*sb
,
544 struct fid
*fid
, int fh_len
, int fh_type
)
546 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
547 exofs_nfs_get_inode
);
550 static struct dentry
*exofs_fh_to_parent(struct super_block
*sb
,
551 struct fid
*fid
, int fh_len
, int fh_type
)
553 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
554 exofs_nfs_get_inode
);
557 static const struct export_operations exofs_export_ops
= {
558 .fh_to_dentry
= exofs_fh_to_dentry
,
559 .fh_to_parent
= exofs_fh_to_parent
,
560 .get_parent
= exofs_get_parent
,
563 /******************************************************************************
565 *****************************************************************************/
568 * struct that describes this file system
570 static struct file_system_type exofs_type
= {
571 .owner
= THIS_MODULE
,
573 .get_sb
= exofs_get_sb
,
574 .kill_sb
= generic_shutdown_super
,
577 static int __init
init_exofs(void)
581 err
= init_inodecache();
585 err
= register_filesystem(&exofs_type
);
591 destroy_inodecache();
596 static void __exit
exit_exofs(void)
598 unregister_filesystem(&exofs_type
);
599 destroy_inodecache();
602 MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
603 MODULE_DESCRIPTION("exofs");
604 MODULE_LICENSE("GPL");
606 module_init(init_exofs
)
607 module_exit(exit_exofs
)