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/string.h>
35 #include <linux/parser.h>
36 #include <linux/vfs.h>
37 #include <linux/random.h>
38 #include <linux/module.h>
39 #include <linux/exportfs.h>
40 #include <linux/slab.h>
44 #define EXOFS_DBGMSG2(M...) do {} while (0)
46 /******************************************************************************
48 *****************************************************************************/
51 * struct to hold what we get from mount options
53 struct exofs_mountopt
{
61 * exofs-specific mount-time options.
63 enum { Opt_name
, Opt_pid
, Opt_to
, Opt_err
};
66 * Our mount-time options. These should ideally be 64-bit unsigned, but the
67 * kernel's parsing functions do not currently support that. 32-bit should be
68 * sufficient for most applications now.
70 static match_table_t tokens
= {
71 {Opt_name
, "osdname=%s"},
78 * The main option parsing method. Also makes sure that all of the mandatory
79 * mount options were set.
81 static int parse_options(char *options
, struct exofs_mountopt
*opts
)
84 substring_t args
[MAX_OPT_ARGS
];
88 EXOFS_DBGMSG("parse_options %s\n", options
);
90 memset(opts
, 0, sizeof(*opts
));
91 opts
->timeout
= BLK_DEFAULT_SG_TIMEOUT
;
93 while ((p
= strsep(&options
, ",")) != NULL
) {
100 token
= match_token(p
, tokens
, args
);
103 opts
->dev_name
= match_strdup(&args
[0]);
104 if (unlikely(!opts
->dev_name
)) {
105 EXOFS_ERR("Error allocating dev_name");
108 opts
->is_osdname
= true;
111 if (0 == match_strlcpy(str
, &args
[0], sizeof(str
)))
113 opts
->pid
= simple_strtoull(str
, NULL
, 0);
114 if (opts
->pid
< EXOFS_MIN_PID
) {
115 EXOFS_ERR("Partition ID must be >= %u",
122 if (match_int(&args
[0], &option
))
125 EXOFS_ERR("Timout must be > 0");
128 opts
->timeout
= option
* HZ
;
134 EXOFS_ERR("Need to specify the following options:\n");
135 EXOFS_ERR(" -o pid=pid_no_to_use\n");
142 /******************************************************************************
144 *****************************************************************************/
147 * Our inode cache. Isn't it pretty?
149 static struct kmem_cache
*exofs_inode_cachep
;
152 * Allocate an inode in the cache
154 static struct inode
*exofs_alloc_inode(struct super_block
*sb
)
156 struct exofs_i_info
*oi
;
158 oi
= kmem_cache_alloc(exofs_inode_cachep
, GFP_KERNEL
);
162 oi
->vfs_inode
.i_version
= 1;
163 return &oi
->vfs_inode
;
166 static void exofs_i_callback(struct rcu_head
*head
)
168 struct inode
*inode
= container_of(head
, struct inode
, i_rcu
);
169 kmem_cache_free(exofs_inode_cachep
, exofs_i(inode
));
173 * Remove an inode from the cache
175 static void exofs_destroy_inode(struct inode
*inode
)
177 call_rcu(&inode
->i_rcu
, exofs_i_callback
);
181 * Initialize the inode
183 static void exofs_init_once(void *foo
)
185 struct exofs_i_info
*oi
= foo
;
187 inode_init_once(&oi
->vfs_inode
);
191 * Create and initialize the inode cache
193 static int init_inodecache(void)
195 exofs_inode_cachep
= kmem_cache_create("exofs_inode_cache",
196 sizeof(struct exofs_i_info
), 0,
197 SLAB_RECLAIM_ACCOUNT
| SLAB_MEM_SPREAD
,
199 if (exofs_inode_cachep
== NULL
)
205 * Destroy the inode cache
207 static void destroy_inodecache(void)
209 kmem_cache_destroy(exofs_inode_cachep
);
212 /******************************************************************************
214 *****************************************************************************/
215 void exofs_make_credential(u8 cred_a
[OSD_CAP_LEN
], const struct osd_obj_id
*obj
)
217 osd_sec_init_nosec_doall_caps(cred_a
, obj
, false, true);
220 static int exofs_read_kern(struct osd_dev
*od
, u8
*cred
, struct osd_obj_id
*obj
,
221 u64 offset
, void *p
, unsigned length
)
223 struct osd_request
*or = osd_start_request(od
, GFP_KERNEL
);
224 /* struct osd_sense_info osi = {.key = 0};*/
228 EXOFS_DBGMSG("%s: osd_start_request failed.\n", __func__
);
231 ret
= osd_req_read_kern(or, obj
, offset
, p
, length
);
233 EXOFS_DBGMSG("%s: osd_req_read_kern failed.\n", __func__
);
237 ret
= osd_finalize_request(or, 0, cred
, NULL
);
239 EXOFS_DBGMSG("Failed to osd_finalize_request() => %d\n", ret
);
243 ret
= osd_execute_request(or);
245 EXOFS_DBGMSG("osd_execute_request() => %d\n", ret
);
246 /* osd_req_decode_sense(or, ret); */
250 EXOFS_DBGMSG2("read_kern(0x%llx) offset=0x%llx "
251 "length=0x%llx dev=%p ret=>%d\n",
252 _LLU(obj
->id
), _LLU(offset
), _LLU(length
), od
, ret
);
256 static const struct osd_attr g_attr_sb_stats
= ATTR_DEF(
259 sizeof(struct exofs_sb_stats
));
261 static int __sbi_read_stats(struct exofs_sb_info
*sbi
)
263 struct osd_attr attrs
[] = {
264 [0] = g_attr_sb_stats
,
266 struct ore_io_state
*ios
;
269 ret
= ore_get_io_state(&sbi
->layout
, &sbi
->oc
, &ios
);
271 EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__
);
275 ios
->in_attr
= attrs
;
276 ios
->in_attr_len
= ARRAY_SIZE(attrs
);
280 EXOFS_ERR("Error reading super_block stats => %d\n", ret
);
284 ret
= extract_attr_from_ios(ios
, &attrs
[0]);
286 EXOFS_ERR("%s: extract_attr of sb_stats failed\n", __func__
);
290 struct exofs_sb_stats
*ess
;
292 if (unlikely(attrs
[0].len
!= sizeof(*ess
))) {
293 EXOFS_ERR("%s: Wrong version of exofs_sb_stats "
294 "size(%d) != expected(%zd)\n",
295 __func__
, attrs
[0].len
, sizeof(*ess
));
299 ess
= attrs
[0].val_ptr
;
300 sbi
->s_nextid
= le64_to_cpu(ess
->s_nextid
);
301 sbi
->s_numfiles
= le32_to_cpu(ess
->s_numfiles
);
305 ore_put_io_state(ios
);
309 static void stats_done(struct ore_io_state
*ios
, void *p
)
311 ore_put_io_state(ios
);
312 /* Good thanks nothing to do anymore */
315 /* Asynchronously write the stats attribute */
316 int exofs_sbi_write_stats(struct exofs_sb_info
*sbi
)
318 struct osd_attr attrs
[] = {
319 [0] = g_attr_sb_stats
,
321 struct ore_io_state
*ios
;
324 ret
= ore_get_io_state(&sbi
->layout
, &sbi
->oc
, &ios
);
326 EXOFS_ERR("%s: ore_get_io_state failed.\n", __func__
);
330 sbi
->s_ess
.s_nextid
= cpu_to_le64(sbi
->s_nextid
);
331 sbi
->s_ess
.s_numfiles
= cpu_to_le64(sbi
->s_numfiles
);
332 attrs
[0].val_ptr
= &sbi
->s_ess
;
335 ios
->done
= stats_done
;
337 ios
->out_attr
= attrs
;
338 ios
->out_attr_len
= ARRAY_SIZE(attrs
);
340 ret
= ore_write(ios
);
342 EXOFS_ERR("%s: ore_write failed.\n", __func__
);
343 ore_put_io_state(ios
);
349 /******************************************************************************
350 * SUPERBLOCK FUNCTIONS
351 *****************************************************************************/
352 static const struct super_operations exofs_sops
;
353 static const struct export_operations exofs_export_ops
;
356 * Write the superblock to the OSD
358 static int exofs_sync_fs(struct super_block
*sb
, int wait
)
360 struct exofs_sb_info
*sbi
;
361 struct exofs_fscb
*fscb
;
362 struct ore_comp one_comp
;
363 struct ore_components oc
;
364 struct ore_io_state
*ios
;
367 fscb
= kmalloc(sizeof(*fscb
), GFP_KERNEL
);
373 /* NOTE: We no longer dirty the super_block anywhere in exofs. The
374 * reason we write the fscb here on unmount is so we can stay backwards
375 * compatible with fscb->s_version == 1. (What we are not compatible
376 * with is if a new version FS crashed and then we try to mount an old
377 * version). Otherwise the exofs_fscb is read-only from mkfs time. All
378 * the writeable info is set in exofs_sbi_write_stats() above.
381 exofs_init_comps(&oc
, &one_comp
, sbi
, EXOFS_SUPER_ID
);
383 ret
= ore_get_io_state(&sbi
->layout
, &oc
, &ios
);
389 ios
->length
= offsetof(struct exofs_fscb
, s_dev_table_oid
);
390 memset(fscb
, 0, ios
->length
);
391 fscb
->s_nextid
= cpu_to_le64(sbi
->s_nextid
);
392 fscb
->s_numfiles
= cpu_to_le32(sbi
->s_numfiles
);
393 fscb
->s_magic
= cpu_to_le16(sb
->s_magic
);
395 fscb
->s_version
= EXOFS_FSCB_VER
;
398 ios
->kern_buff
= fscb
;
400 ret
= ore_write(ios
);
402 EXOFS_ERR("%s: ore_write failed.\n", __func__
);
409 EXOFS_DBGMSG("s_nextid=0x%llx ret=%d\n", _LLU(sbi
->s_nextid
), ret
);
410 ore_put_io_state(ios
);
415 static void exofs_write_super(struct super_block
*sb
)
417 if (!(sb
->s_flags
& MS_RDONLY
))
418 exofs_sync_fs(sb
, 1);
423 static void _exofs_print_device(const char *msg
, const char *dev_path
,
424 struct osd_dev
*od
, u64 pid
)
426 const struct osd_dev_info
*odi
= osduld_device_info(od
);
428 printk(KERN_NOTICE
"exofs: %s %s osd_name-%s pid-0x%llx\n",
429 msg
, dev_path
?: "", odi
->osdname
, _LLU(pid
));
432 static void exofs_free_sbi(struct exofs_sb_info
*sbi
)
434 unsigned numdevs
= sbi
->oc
.numdevs
;
437 unsigned i
= --numdevs
;
438 struct osd_dev
*od
= ore_comp_dev(&sbi
->oc
, i
);
441 ore_comp_set_dev(&sbi
->oc
, i
, NULL
);
442 osduld_put_device(od
);
450 * This function is called when the vfs is freeing the superblock. We just
451 * need to free our own part.
453 static void exofs_put_super(struct super_block
*sb
)
456 struct exofs_sb_info
*sbi
= sb
->s_fs_info
;
458 /* make sure there are no pending commands */
459 for (num_pend
= atomic_read(&sbi
->s_curr_pending
); num_pend
> 0;
460 num_pend
= atomic_read(&sbi
->s_curr_pending
)) {
461 wait_queue_head_t wq
;
463 printk(KERN_NOTICE
"%s: !!Pending operations in flight. "
464 "This is a BUG. please report to osd-dev@open-osd.org\n",
466 init_waitqueue_head(&wq
);
467 wait_event_timeout(wq
,
468 (atomic_read(&sbi
->s_curr_pending
) == 0),
469 msecs_to_jiffies(100));
472 _exofs_print_device("Unmounting", NULL
, ore_comp_dev(&sbi
->oc
, 0),
473 sbi
->one_comp
.obj
.partition
);
475 bdi_destroy(&sbi
->bdi
);
477 sb
->s_fs_info
= NULL
;
480 static int _read_and_match_data_map(struct exofs_sb_info
*sbi
, unsigned numdevs
,
481 struct exofs_device_table
*dt
)
485 sbi
->layout
.stripe_unit
=
486 le64_to_cpu(dt
->dt_data_map
.cb_stripe_unit
);
487 sbi
->layout
.group_width
=
488 le32_to_cpu(dt
->dt_data_map
.cb_group_width
);
489 sbi
->layout
.group_depth
=
490 le32_to_cpu(dt
->dt_data_map
.cb_group_depth
);
491 sbi
->layout
.mirrors_p1
=
492 le32_to_cpu(dt
->dt_data_map
.cb_mirror_cnt
) + 1;
493 sbi
->layout
.raid_algorithm
=
494 le32_to_cpu(dt
->dt_data_map
.cb_raid_algorithm
);
496 ret
= ore_verify_layout(numdevs
, &sbi
->layout
);
498 EXOFS_DBGMSG("exofs: layout: "
499 "num_comps=%u stripe_unit=0x%x group_width=%u "
500 "group_depth=0x%llx mirrors_p1=%u raid_algorithm=%u\n",
502 sbi
->layout
.stripe_unit
,
503 sbi
->layout
.group_width
,
504 _LLU(sbi
->layout
.group_depth
),
505 sbi
->layout
.mirrors_p1
,
506 sbi
->layout
.raid_algorithm
);
510 static unsigned __ra_pages(struct ore_layout
*layout
)
512 const unsigned _MIN_RA
= 32; /* min 128K read-ahead */
513 unsigned ra_pages
= layout
->group_width
* layout
->stripe_unit
/
515 unsigned max_io_pages
= exofs_max_io_pages(layout
, ~0);
517 ra_pages
*= 2; /* two stripes */
518 if (ra_pages
< _MIN_RA
)
519 ra_pages
= roundup(_MIN_RA
, ra_pages
/ 2);
521 if (ra_pages
> max_io_pages
)
522 ra_pages
= max_io_pages
;
527 /* @odi is valid only as long as @fscb_dev is valid */
528 static int exofs_devs_2_odi(struct exofs_dt_device_info
*dt_dev
,
529 struct osd_dev_info
*odi
)
531 odi
->systemid_len
= le32_to_cpu(dt_dev
->systemid_len
);
532 memcpy(odi
->systemid
, dt_dev
->systemid
, odi
->systemid_len
);
534 odi
->osdname_len
= le32_to_cpu(dt_dev
->osdname_len
);
535 odi
->osdname
= dt_dev
->osdname
;
537 /* FIXME support long names. Will need a _put function */
538 if (dt_dev
->long_name_offset
)
541 /* Make sure osdname is printable!
542 * mkexofs should give us space for a null-terminator else the
543 * device-table is invalid.
545 if (unlikely(odi
->osdname_len
>= sizeof(dt_dev
->osdname
)))
546 odi
->osdname_len
= sizeof(dt_dev
->osdname
) - 1;
547 dt_dev
->osdname
[odi
->osdname_len
] = 0;
549 /* If it's all zeros something is bad we read past end-of-obj */
550 return !(odi
->systemid_len
|| odi
->osdname_len
);
553 int __alloc_dev_table(struct exofs_sb_info
*sbi
, unsigned numdevs
,
554 struct exofs_dev
**peds
)
556 struct __alloc_ore_devs_and_exofs_devs
{
557 /* Twice bigger table: See exofs_init_comps() and comment at
558 * exofs_read_lookup_dev_table()
560 struct ore_dev
*oreds
[numdevs
* 2 - 1];
561 struct exofs_dev eds
[numdevs
];
563 struct exofs_dev
*eds
;
566 aoded
= kzalloc(sizeof(*aoded
), GFP_KERNEL
);
567 if (unlikely(!aoded
)) {
568 EXOFS_ERR("ERROR: faild allocating Device array[%d]\n",
573 sbi
->oc
.ods
= aoded
->oreds
;
574 *peds
= eds
= aoded
->eds
;
575 for (i
= 0; i
< numdevs
; ++i
)
576 aoded
->oreds
[i
] = &eds
[i
].ored
;
580 static int exofs_read_lookup_dev_table(struct exofs_sb_info
*sbi
,
581 struct osd_dev
*fscb_od
,
582 unsigned table_count
)
584 struct ore_comp comp
;
585 struct exofs_device_table
*dt
;
586 struct exofs_dev
*eds
;
587 unsigned table_bytes
= table_count
* sizeof(dt
->dt_dev_table
[0]) +
592 dt
= kmalloc(table_bytes
, GFP_KERNEL
);
594 EXOFS_ERR("ERROR: allocating %x bytes for device table\n",
601 comp
.obj
.partition
= sbi
->one_comp
.obj
.partition
;
602 comp
.obj
.id
= EXOFS_DEVTABLE_ID
;
603 exofs_make_credential(comp
.cred
, &comp
.obj
);
605 ret
= exofs_read_kern(fscb_od
, comp
.cred
, &comp
.obj
, 0, dt
,
608 EXOFS_ERR("ERROR: reading device table\n");
612 numdevs
= le64_to_cpu(dt
->dt_num_devices
);
613 if (unlikely(!numdevs
)) {
617 WARN_ON(table_count
!= numdevs
);
619 ret
= _read_and_match_data_map(sbi
, numdevs
, dt
);
623 ret
= __alloc_dev_table(sbi
, numdevs
, &eds
);
626 /* exofs round-robins the device table view according to inode
627 * number. We hold a: twice bigger table hence inodes can point
628 * to any device and have a sequential view of the table
629 * starting at this device. See exofs_init_comps()
631 memcpy(&sbi
->oc
.ods
[numdevs
], &sbi
->oc
.ods
[0],
632 (numdevs
- 1) * sizeof(sbi
->oc
.ods
[0]));
634 for (i
= 0; i
< numdevs
; i
++) {
635 struct exofs_fscb fscb
;
636 struct osd_dev_info odi
;
639 if (exofs_devs_2_odi(&dt
->dt_dev_table
[i
], &odi
)) {
640 EXOFS_ERR("ERROR: Read all-zeros device entry\n");
645 printk(KERN_NOTICE
"Add device[%d]: osd_name-%s\n",
648 /* the exofs id is currently the table index */
651 /* On all devices the device table is identical. The user can
652 * specify any one of the participating devices on the command
653 * line. We always keep them in device-table order.
655 if (fscb_od
&& osduld_device_same(fscb_od
, &odi
)) {
656 eds
[i
].ored
.od
= fscb_od
;
662 od
= osduld_info_lookup(&odi
);
665 EXOFS_ERR("ERROR: device requested is not found "
666 "osd_name-%s =>%d\n", odi
.osdname
, ret
);
673 /* Read the fscb of the other devices to make sure the FS
674 * partition is there.
676 ret
= exofs_read_kern(od
, comp
.cred
, &comp
.obj
, 0, &fscb
,
679 EXOFS_ERR("ERROR: Malformed participating device "
680 "error reading fscb osd_name-%s\n",
685 /* TODO: verify other information is correct and FS-uuid
686 * matches. Benny what did you say about device table
687 * generation and old devices?
693 if (unlikely(fscb_od
&& !ret
)) {
694 EXOFS_ERR("ERROR: Bad device-table container device not present\n");
695 osduld_put_device(fscb_od
);
702 * Read the superblock from the OSD and fill in the fields
704 static int exofs_fill_super(struct super_block
*sb
, void *data
, int silent
)
707 struct exofs_mountopt
*opts
= data
;
708 struct exofs_sb_info
*sbi
; /*extended info */
709 struct osd_dev
*od
; /* Master device */
710 struct exofs_fscb fscb
; /*on-disk superblock info */
711 struct ore_comp comp
;
712 unsigned table_count
;
715 sbi
= kzalloc(sizeof(*sbi
), GFP_KERNEL
);
719 /* use mount options to fill superblock */
720 if (opts
->is_osdname
) {
721 struct osd_dev_info odi
= {.systemid_len
= 0};
723 odi
.osdname_len
= strlen(opts
->dev_name
);
724 odi
.osdname
= (u8
*)opts
->dev_name
;
725 od
= osduld_info_lookup(&odi
);
726 kfree(opts
->dev_name
);
727 opts
->dev_name
= NULL
;
729 od
= osduld_path_lookup(opts
->dev_name
);
736 /* Default layout in case we do not have a device-table */
737 sbi
->layout
.stripe_unit
= PAGE_SIZE
;
738 sbi
->layout
.mirrors_p1
= 1;
739 sbi
->layout
.group_width
= 1;
740 sbi
->layout
.group_depth
= -1;
741 sbi
->layout
.group_count
= 1;
742 sbi
->s_timeout
= opts
->timeout
;
744 sbi
->one_comp
.obj
.partition
= opts
->pid
;
745 sbi
->one_comp
.obj
.id
= 0;
746 exofs_make_credential(sbi
->one_comp
.cred
, &sbi
->one_comp
.obj
);
748 sbi
->oc
.single_comp
= EC_SINGLE_COMP
;
749 sbi
->oc
.comps
= &sbi
->one_comp
;
751 /* fill in some other data by hand */
752 memset(sb
->s_id
, 0, sizeof(sb
->s_id
));
753 strcpy(sb
->s_id
, "exofs");
754 sb
->s_blocksize
= EXOFS_BLKSIZE
;
755 sb
->s_blocksize_bits
= EXOFS_BLKSHIFT
;
756 sb
->s_maxbytes
= MAX_LFS_FILESIZE
;
757 sb
->s_max_links
= EXOFS_LINK_MAX
;
758 atomic_set(&sbi
->s_curr_pending
, 0);
762 comp
.obj
.partition
= sbi
->one_comp
.obj
.partition
;
763 comp
.obj
.id
= EXOFS_SUPER_ID
;
764 exofs_make_credential(comp
.cred
, &comp
.obj
);
766 ret
= exofs_read_kern(od
, comp
.cred
, &comp
.obj
, 0, &fscb
, sizeof(fscb
));
770 sb
->s_magic
= le16_to_cpu(fscb
.s_magic
);
771 /* NOTE: we read below to be backward compatible with old versions */
772 sbi
->s_nextid
= le64_to_cpu(fscb
.s_nextid
);
773 sbi
->s_numfiles
= le32_to_cpu(fscb
.s_numfiles
);
775 /* make sure what we read from the object store is correct */
776 if (sb
->s_magic
!= EXOFS_SUPER_MAGIC
) {
778 EXOFS_ERR("ERROR: Bad magic value\n");
782 if (le32_to_cpu(fscb
.s_version
) > EXOFS_FSCB_VER
) {
783 EXOFS_ERR("ERROR: Bad FSCB version expected-%d got-%d\n",
784 EXOFS_FSCB_VER
, le32_to_cpu(fscb
.s_version
));
789 /* start generation numbers from a random point */
790 get_random_bytes(&sbi
->s_next_generation
, sizeof(u32
));
791 spin_lock_init(&sbi
->s_next_gen_lock
);
793 table_count
= le64_to_cpu(fscb
.s_dev_table_count
);
795 ret
= exofs_read_lookup_dev_table(sbi
, od
, table_count
);
799 struct exofs_dev
*eds
;
801 ret
= __alloc_dev_table(sbi
, 1, &eds
);
805 ore_comp_set_dev(&sbi
->oc
, 0, od
);
808 __sbi_read_stats(sbi
);
810 /* set up operation vectors */
811 sbi
->bdi
.ra_pages
= __ra_pages(&sbi
->layout
);
812 sb
->s_bdi
= &sbi
->bdi
;
814 sb
->s_op
= &exofs_sops
;
815 sb
->s_export_op
= &exofs_export_ops
;
816 root
= exofs_iget(sb
, EXOFS_ROOT_ID
- EXOFS_OBJ_OFF
);
818 EXOFS_ERR("ERROR: exofs_iget failed\n");
822 sb
->s_root
= d_alloc_root(root
);
825 EXOFS_ERR("ERROR: get root inode failed\n");
830 if (!S_ISDIR(root
->i_mode
)) {
833 EXOFS_ERR("ERROR: corrupt root inode (mode = %hd)\n",
839 ret
= bdi_setup_and_register(&sbi
->bdi
, "exofs", BDI_CAP_MAP_COPY
);
841 EXOFS_DBGMSG("Failed to bdi_setup_and_register\n");
847 _exofs_print_device("Mounting", opts
->dev_name
,
848 ore_comp_dev(&sbi
->oc
, 0),
849 sbi
->one_comp
.obj
.partition
);
853 EXOFS_ERR("Unable to mount exofs on %s pid=0x%llx err=%d\n",
854 opts
->dev_name
, sbi
->one_comp
.obj
.partition
, ret
);
860 * Set up the superblock (calls exofs_fill_super eventually)
862 static struct dentry
*exofs_mount(struct file_system_type
*type
,
863 int flags
, const char *dev_name
,
866 struct exofs_mountopt opts
;
869 ret
= parse_options(data
, &opts
);
874 opts
.dev_name
= dev_name
;
875 return mount_nodev(type
, flags
, &opts
, exofs_fill_super
);
879 * Return information about the file system state in the buffer. This is used
880 * by the 'df' command, for example.
882 static int exofs_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
884 struct super_block
*sb
= dentry
->d_sb
;
885 struct exofs_sb_info
*sbi
= sb
->s_fs_info
;
886 struct ore_io_state
*ios
;
887 struct osd_attr attrs
[] = {
888 ATTR_DEF(OSD_APAGE_PARTITION_QUOTAS
,
889 OSD_ATTR_PQ_CAPACITY_QUOTA
, sizeof(__be64
)),
890 ATTR_DEF(OSD_APAGE_PARTITION_INFORMATION
,
891 OSD_ATTR_PI_USED_CAPACITY
, sizeof(__be64
)),
893 uint64_t capacity
= ULLONG_MAX
;
894 uint64_t used
= ULLONG_MAX
;
897 ret
= ore_get_io_state(&sbi
->layout
, &sbi
->oc
, &ios
);
899 EXOFS_DBGMSG("ore_get_io_state failed.\n");
903 ios
->in_attr
= attrs
;
904 ios
->in_attr_len
= ARRAY_SIZE(attrs
);
910 ret
= extract_attr_from_ios(ios
, &attrs
[0]);
912 capacity
= get_unaligned_be64(attrs
[0].val_ptr
);
913 if (unlikely(!capacity
))
914 capacity
= ULLONG_MAX
;
916 EXOFS_DBGMSG("exofs_statfs: get capacity failed.\n");
918 ret
= extract_attr_from_ios(ios
, &attrs
[1]);
920 used
= get_unaligned_be64(attrs
[1].val_ptr
);
922 EXOFS_DBGMSG("exofs_statfs: get used-space failed.\n");
924 /* fill in the stats buffer */
925 buf
->f_type
= EXOFS_SUPER_MAGIC
;
926 buf
->f_bsize
= EXOFS_BLKSIZE
;
927 buf
->f_blocks
= capacity
>> 9;
928 buf
->f_bfree
= (capacity
- used
) >> 9;
929 buf
->f_bavail
= buf
->f_bfree
;
930 buf
->f_files
= sbi
->s_numfiles
;
931 buf
->f_ffree
= EXOFS_MAX_ID
- sbi
->s_numfiles
;
932 buf
->f_namelen
= EXOFS_NAME_LEN
;
935 ore_put_io_state(ios
);
939 static const struct super_operations exofs_sops
= {
940 .alloc_inode
= exofs_alloc_inode
,
941 .destroy_inode
= exofs_destroy_inode
,
942 .write_inode
= exofs_write_inode
,
943 .evict_inode
= exofs_evict_inode
,
944 .put_super
= exofs_put_super
,
945 .write_super
= exofs_write_super
,
946 .sync_fs
= exofs_sync_fs
,
947 .statfs
= exofs_statfs
,
950 /******************************************************************************
952 *****************************************************************************/
954 static struct dentry
*exofs_get_parent(struct dentry
*child
)
956 unsigned long ino
= exofs_parent_ino(child
);
959 return ERR_PTR(-ESTALE
);
961 return d_obtain_alias(exofs_iget(child
->d_inode
->i_sb
, ino
));
964 static struct inode
*exofs_nfs_get_inode(struct super_block
*sb
,
965 u64 ino
, u32 generation
)
969 inode
= exofs_iget(sb
, ino
);
971 return ERR_CAST(inode
);
972 if (generation
&& inode
->i_generation
!= generation
) {
973 /* we didn't find the right inode.. */
975 return ERR_PTR(-ESTALE
);
980 static struct dentry
*exofs_fh_to_dentry(struct super_block
*sb
,
981 struct fid
*fid
, int fh_len
, int fh_type
)
983 return generic_fh_to_dentry(sb
, fid
, fh_len
, fh_type
,
984 exofs_nfs_get_inode
);
987 static struct dentry
*exofs_fh_to_parent(struct super_block
*sb
,
988 struct fid
*fid
, int fh_len
, int fh_type
)
990 return generic_fh_to_parent(sb
, fid
, fh_len
, fh_type
,
991 exofs_nfs_get_inode
);
994 static const struct export_operations exofs_export_ops
= {
995 .fh_to_dentry
= exofs_fh_to_dentry
,
996 .fh_to_parent
= exofs_fh_to_parent
,
997 .get_parent
= exofs_get_parent
,
1000 /******************************************************************************
1002 *****************************************************************************/
1005 * struct that describes this file system
1007 static struct file_system_type exofs_type
= {
1008 .owner
= THIS_MODULE
,
1010 .mount
= exofs_mount
,
1011 .kill_sb
= generic_shutdown_super
,
1014 static int __init
init_exofs(void)
1018 err
= init_inodecache();
1022 err
= register_filesystem(&exofs_type
);
1028 destroy_inodecache();
1033 static void __exit
exit_exofs(void)
1035 unregister_filesystem(&exofs_type
);
1036 destroy_inodecache();
1039 MODULE_AUTHOR("Avishay Traeger <avishay@gmail.com>");
1040 MODULE_DESCRIPTION("exofs");
1041 MODULE_LICENSE("GPL");
1043 module_init(init_exofs
)
1044 module_exit(exit_exofs
)