2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License version 2.
10 #include <linux/slab.h>
11 #include <linux/spinlock.h>
12 #include <linux/completion.h>
13 #include <linux/buffer_head.h>
14 #include <linux/pagemap.h>
15 #include <linux/uio.h>
16 #include <linux/blkdev.h>
18 #include <linux/mount.h>
20 #include <linux/gfs2_ondisk.h>
21 #include <linux/ext2_fs.h>
22 #include <linux/crc32.h>
23 #include <linux/lm_interface.h>
24 #include <linux/writeback.h>
25 #include <asm/uaccess.h>
41 #include "ops_address.h"
42 #include "ops_inode.h"
45 * gfs2_llseek - seek to a location in a file
48 * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END)
50 * SEEK_END requires the glock for the file because it references the
53 * Returns: The new offset, or errno
56 static loff_t
gfs2_llseek(struct file
*file
, loff_t offset
, int origin
)
58 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
59 struct gfs2_holder i_gh
;
63 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, LM_FLAG_ANY
,
66 error
= generic_file_llseek_unlocked(file
, offset
, origin
);
67 gfs2_glock_dq_uninit(&i_gh
);
70 error
= generic_file_llseek_unlocked(file
, offset
, origin
);
76 * gfs2_readdir - Read directory entries from a directory
77 * @file: The directory to read from
78 * @dirent: Buffer for dirents
79 * @filldir: Function used to do the copying
84 static int gfs2_readdir(struct file
*file
, void *dirent
, filldir_t filldir
)
86 struct inode
*dir
= file
->f_mapping
->host
;
87 struct gfs2_inode
*dip
= GFS2_I(dir
);
88 struct gfs2_holder d_gh
;
89 u64 offset
= file
->f_pos
;
92 gfs2_holder_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
93 error
= gfs2_glock_nq(&d_gh
);
95 gfs2_holder_uninit(&d_gh
);
99 error
= gfs2_dir_read(dir
, &offset
, dirent
, filldir
);
101 gfs2_glock_dq_uninit(&d_gh
);
103 file
->f_pos
= offset
;
110 * @table: A table of 32 u32 flags
111 * @val: a 32 bit value to convert
113 * This function can be used to convert between fsflags values and
114 * GFS2's own flags values.
116 * Returns: the converted flags
118 static u32
fsflags_cvt(const u32
*table
, u32 val
)
130 static const u32 fsflags_to_gfs2
[32] = {
132 [4] = GFS2_DIF_IMMUTABLE
,
133 [5] = GFS2_DIF_APPENDONLY
,
134 [7] = GFS2_DIF_NOATIME
,
135 [12] = GFS2_DIF_EXHASH
,
136 [14] = GFS2_DIF_INHERIT_JDATA
,
139 static const u32 gfs2_to_fsflags
[32] = {
140 [gfs2fl_Sync
] = FS_SYNC_FL
,
141 [gfs2fl_Immutable
] = FS_IMMUTABLE_FL
,
142 [gfs2fl_AppendOnly
] = FS_APPEND_FL
,
143 [gfs2fl_NoAtime
] = FS_NOATIME_FL
,
144 [gfs2fl_ExHash
] = FS_INDEX_FL
,
145 [gfs2fl_InheritJdata
] = FS_JOURNAL_DATA_FL
,
148 static int gfs2_get_flags(struct file
*filp
, u32 __user
*ptr
)
150 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
151 struct gfs2_inode
*ip
= GFS2_I(inode
);
152 struct gfs2_holder gh
;
156 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, 0, &gh
);
157 error
= gfs2_glock_nq(&gh
);
161 fsflags
= fsflags_cvt(gfs2_to_fsflags
, ip
->i_di
.di_flags
);
162 if (!S_ISDIR(inode
->i_mode
) && ip
->i_di
.di_flags
& GFS2_DIF_JDATA
)
163 fsflags
|= FS_JOURNAL_DATA_FL
;
164 if (put_user(fsflags
, ptr
))
168 gfs2_holder_uninit(&gh
);
172 void gfs2_set_inode_flags(struct inode
*inode
)
174 struct gfs2_inode
*ip
= GFS2_I(inode
);
175 struct gfs2_dinode_host
*di
= &ip
->i_di
;
176 unsigned int flags
= inode
->i_flags
;
178 flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
179 if (di
->di_flags
& GFS2_DIF_IMMUTABLE
)
180 flags
|= S_IMMUTABLE
;
181 if (di
->di_flags
& GFS2_DIF_APPENDONLY
)
183 if (di
->di_flags
& GFS2_DIF_NOATIME
)
185 if (di
->di_flags
& GFS2_DIF_SYNC
)
187 inode
->i_flags
= flags
;
190 /* Flags that can be set by user space */
191 #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \
192 GFS2_DIF_IMMUTABLE| \
193 GFS2_DIF_APPENDONLY| \
197 GFS2_DIF_INHERIT_JDATA)
200 * gfs2_set_flags - set flags on an inode
202 * @flags: The flags to set
203 * @mask: Indicates which flags are valid
206 static int do_gfs2_set_flags(struct file
*filp
, u32 reqflags
, u32 mask
)
208 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
209 struct gfs2_inode
*ip
= GFS2_I(inode
);
210 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
211 struct buffer_head
*bh
;
212 struct gfs2_holder gh
;
214 u32 new_flags
, flags
;
216 error
= mnt_want_write(filp
->f_path
.mnt
);
220 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
224 flags
= ip
->i_di
.di_flags
;
225 new_flags
= (flags
& ~mask
) | (reqflags
& mask
);
226 if ((new_flags
^ flags
) == 0)
230 if ((new_flags
^ flags
) & ~GFS2_FLAGS_USER_SET
)
234 if (IS_IMMUTABLE(inode
) && (new_flags
& GFS2_DIF_IMMUTABLE
))
236 if (IS_APPEND(inode
) && (new_flags
& GFS2_DIF_APPENDONLY
))
238 if (((new_flags
^ flags
) & GFS2_DIF_IMMUTABLE
) &&
239 !capable(CAP_LINUX_IMMUTABLE
))
241 if (!IS_IMMUTABLE(inode
)) {
242 error
= gfs2_permission(inode
, MAY_WRITE
);
246 if ((flags
^ new_flags
) & GFS2_DIF_JDATA
) {
247 if (flags
& GFS2_DIF_JDATA
)
248 gfs2_log_flush(sdp
, ip
->i_gl
);
249 error
= filemap_fdatawrite(inode
->i_mapping
);
252 error
= filemap_fdatawait(inode
->i_mapping
);
256 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
259 error
= gfs2_meta_inode_buffer(ip
, &bh
);
262 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
263 ip
->i_di
.di_flags
= new_flags
;
264 gfs2_dinode_out(ip
, bh
->b_data
);
266 gfs2_set_inode_flags(inode
);
267 gfs2_set_aops(inode
);
271 gfs2_glock_dq_uninit(&gh
);
273 mnt_drop_write(filp
->f_path
.mnt
);
277 static int gfs2_set_flags(struct file
*filp
, u32 __user
*ptr
)
279 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
280 u32 fsflags
, gfsflags
;
281 if (get_user(fsflags
, ptr
))
283 gfsflags
= fsflags_cvt(fsflags_to_gfs2
, fsflags
);
284 if (!S_ISDIR(inode
->i_mode
)) {
285 if (gfsflags
& GFS2_DIF_INHERIT_JDATA
)
286 gfsflags
^= (GFS2_DIF_JDATA
| GFS2_DIF_INHERIT_JDATA
);
287 return do_gfs2_set_flags(filp
, gfsflags
, ~0);
289 return do_gfs2_set_flags(filp
, gfsflags
, ~GFS2_DIF_JDATA
);
292 static long gfs2_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
295 case FS_IOC_GETFLAGS
:
296 return gfs2_get_flags(filp
, (u32 __user
*)arg
);
297 case FS_IOC_SETFLAGS
:
298 return gfs2_set_flags(filp
, (u32 __user
*)arg
);
304 * gfs2_allocate_page_backing - Use bmap to allocate blocks
305 * @page: The (locked) page to allocate backing for
307 * We try to allocate all the blocks required for the page in
308 * one go. This might fail for various reasons, so we keep
309 * trying until all the blocks to back this page are allocated.
310 * If some of the blocks are already allocated, thats ok too.
313 static int gfs2_allocate_page_backing(struct page
*page
)
315 struct inode
*inode
= page
->mapping
->host
;
316 struct buffer_head bh
;
317 unsigned long size
= PAGE_CACHE_SIZE
;
318 u64 lblock
= page
->index
<< (PAGE_CACHE_SHIFT
- inode
->i_blkbits
);
323 gfs2_block_map(inode
, lblock
, &bh
, 1);
324 if (!buffer_mapped(&bh
))
327 lblock
+= (bh
.b_size
>> inode
->i_blkbits
);
333 * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable
334 * @vma: The virtual memory area
335 * @page: The page which is about to become writable
337 * When the page becomes writable, we need to ensure that we have
338 * blocks allocated on disk to back that page.
341 static int gfs2_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
343 struct inode
*inode
= vma
->vm_file
->f_path
.dentry
->d_inode
;
344 struct gfs2_inode
*ip
= GFS2_I(inode
);
345 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
346 unsigned long last_index
;
347 u64 pos
= page
->index
<< (PAGE_CACHE_SIZE
- inode
->i_blkbits
);
348 unsigned int data_blocks
, ind_blocks
, rblocks
;
349 int alloc_required
= 0;
350 struct gfs2_holder gh
;
351 struct gfs2_alloc
*al
;
354 gfs2_holder_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
355 ret
= gfs2_glock_nq(&gh
);
359 set_bit(GIF_SW_PAGED
, &ip
->i_flags
);
360 gfs2_write_calc_reserv(ip
, PAGE_CACHE_SIZE
, &data_blocks
, &ind_blocks
);
361 ret
= gfs2_write_alloc_required(ip
, pos
, PAGE_CACHE_SIZE
, &alloc_required
);
362 if (ret
|| !alloc_required
)
365 al
= gfs2_alloc_get(ip
);
369 ret
= gfs2_quota_lock_check(ip
);
372 al
->al_requested
= data_blocks
+ ind_blocks
;
373 ret
= gfs2_inplace_reserve(ip
);
375 goto out_quota_unlock
;
377 rblocks
= RES_DINODE
+ ind_blocks
;
378 if (gfs2_is_jdata(ip
))
379 rblocks
+= data_blocks
? data_blocks
: 1;
380 if (ind_blocks
|| data_blocks
)
381 rblocks
+= RES_STATFS
+ RES_QUOTA
;
382 ret
= gfs2_trans_begin(sdp
, rblocks
, 0);
388 last_index
= ip
->i_inode
.i_size
>> PAGE_CACHE_SHIFT
;
389 if (page
->index
> last_index
)
390 goto out_unlock_page
;
392 if (!PageUptodate(page
) || page
->mapping
!= ip
->i_inode
.i_mapping
)
393 goto out_unlock_page
;
394 if (gfs2_is_stuffed(ip
)) {
395 ret
= gfs2_unstuff_dinode(ip
, page
);
397 goto out_unlock_page
;
399 ret
= gfs2_allocate_page_backing(page
);
405 gfs2_inplace_release(ip
);
407 gfs2_quota_unlock(ip
);
413 gfs2_holder_uninit(&gh
);
417 static struct vm_operations_struct gfs2_vm_ops
= {
418 .fault
= filemap_fault
,
419 .page_mkwrite
= gfs2_page_mkwrite
,
425 * @file: The file to map
426 * @vma: The VMA which described the mapping
428 * Returns: 0 or error code
431 static int gfs2_mmap(struct file
*file
, struct vm_area_struct
*vma
)
433 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
434 struct gfs2_holder i_gh
;
437 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, 0, &i_gh
);
438 error
= gfs2_glock_nq(&i_gh
);
440 gfs2_holder_uninit(&i_gh
);
444 vma
->vm_ops
= &gfs2_vm_ops
;
446 gfs2_glock_dq_uninit(&i_gh
);
452 * gfs2_open - open a file
453 * @inode: the inode to open
454 * @file: the struct file for this opening
459 static int gfs2_open(struct inode
*inode
, struct file
*file
)
461 struct gfs2_inode
*ip
= GFS2_I(inode
);
462 struct gfs2_holder i_gh
;
463 struct gfs2_file
*fp
;
466 fp
= kzalloc(sizeof(struct gfs2_file
), GFP_KERNEL
);
470 mutex_init(&fp
->f_fl_mutex
);
472 gfs2_assert_warn(GFS2_SB(inode
), !file
->private_data
);
473 file
->private_data
= fp
;
475 if (S_ISREG(ip
->i_inode
.i_mode
)) {
476 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, LM_FLAG_ANY
,
481 if (!(file
->f_flags
& O_LARGEFILE
) &&
482 ip
->i_di
.di_size
> MAX_NON_LFS
) {
487 gfs2_glock_dq_uninit(&i_gh
);
493 gfs2_glock_dq_uninit(&i_gh
);
495 file
->private_data
= NULL
;
501 * gfs2_close - called to close a struct file
502 * @inode: the inode the struct file belongs to
503 * @file: the struct file being closed
508 static int gfs2_close(struct inode
*inode
, struct file
*file
)
510 struct gfs2_sbd
*sdp
= inode
->i_sb
->s_fs_info
;
511 struct gfs2_file
*fp
;
513 fp
= file
->private_data
;
514 file
->private_data
= NULL
;
516 if (gfs2_assert_warn(sdp
, fp
))
525 * gfs2_fsync - sync the dirty data for a file (across the cluster)
526 * @file: the file that points to the dentry (we ignore this)
527 * @dentry: the dentry that points to the inode to sync
529 * The VFS will flush "normal" data for us. We only need to worry
530 * about metadata here. For journaled data, we just do a log flush
531 * as we can't avoid it. Otherwise we can just bale out if datasync
532 * is set. For stuffed inodes we must flush the log in order to
533 * ensure that all data is on disk.
535 * The call to write_inode_now() is there to write back metadata and
536 * the inode itself. It does also try and write the data, but thats
537 * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
543 static int gfs2_fsync(struct file
*file
, struct dentry
*dentry
, int datasync
)
545 struct inode
*inode
= dentry
->d_inode
;
546 int sync_state
= inode
->i_state
& (I_DIRTY_SYNC
|I_DIRTY_DATASYNC
);
549 if (gfs2_is_jdata(GFS2_I(inode
))) {
550 gfs2_log_flush(GFS2_SB(inode
), GFS2_I(inode
)->i_gl
);
554 if (sync_state
!= 0) {
556 ret
= write_inode_now(inode
, 0);
558 if (gfs2_is_stuffed(GFS2_I(inode
)))
559 gfs2_log_flush(GFS2_SB(inode
), GFS2_I(inode
)->i_gl
);
566 * gfs2_setlease - acquire/release a file lease
567 * @file: the file pointer
574 static int gfs2_setlease(struct file
*file
, long arg
, struct file_lock
**fl
)
576 struct gfs2_sbd
*sdp
= GFS2_SB(file
->f_mapping
->host
);
579 * We don't currently have a way to enforce a lease across the whole
580 * cluster; until we do, disable leases (by just returning -EINVAL),
581 * unless the administrator has requested purely local locking.
583 if (!sdp
->sd_args
.ar_localflocks
)
585 return generic_setlease(file
, arg
, fl
);
588 static int gfs2_lm_plock_get(struct gfs2_sbd
*sdp
, struct lm_lockname
*name
,
589 struct file
*file
, struct file_lock
*fl
)
592 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
593 error
= sdp
->sd_lockstruct
.ls_ops
->lm_plock_get(
594 sdp
->sd_lockstruct
.ls_lockspace
, name
, file
, fl
);
598 static int gfs2_lm_plock(struct gfs2_sbd
*sdp
, struct lm_lockname
*name
,
599 struct file
*file
, int cmd
, struct file_lock
*fl
)
602 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
603 error
= sdp
->sd_lockstruct
.ls_ops
->lm_plock(
604 sdp
->sd_lockstruct
.ls_lockspace
, name
, file
, cmd
, fl
);
608 static int gfs2_lm_punlock(struct gfs2_sbd
*sdp
, struct lm_lockname
*name
,
609 struct file
*file
, struct file_lock
*fl
)
612 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
613 error
= sdp
->sd_lockstruct
.ls_ops
->lm_punlock(
614 sdp
->sd_lockstruct
.ls_lockspace
, name
, file
, fl
);
619 * gfs2_lock - acquire/release a posix lock on a file
620 * @file: the file pointer
621 * @cmd: either modify or retrieve lock state, possibly wait
622 * @fl: type and range of lock
627 static int gfs2_lock(struct file
*file
, int cmd
, struct file_lock
*fl
)
629 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
630 struct gfs2_sbd
*sdp
= GFS2_SB(file
->f_mapping
->host
);
631 struct lm_lockname name
=
632 { .ln_number
= ip
->i_no_addr
,
633 .ln_type
= LM_TYPE_PLOCK
};
635 if (!(fl
->fl_flags
& FL_POSIX
))
637 if (__mandatory_lock(&ip
->i_inode
))
640 if (cmd
== F_CANCELLK
) {
643 fl
->fl_type
= F_UNLCK
;
646 return gfs2_lm_plock_get(sdp
, &name
, file
, fl
);
647 else if (fl
->fl_type
== F_UNLCK
)
648 return gfs2_lm_punlock(sdp
, &name
, file
, fl
);
650 return gfs2_lm_plock(sdp
, &name
, file
, cmd
, fl
);
653 static int do_flock(struct file
*file
, int cmd
, struct file_lock
*fl
)
655 struct gfs2_file
*fp
= file
->private_data
;
656 struct gfs2_holder
*fl_gh
= &fp
->f_fl_gh
;
657 struct gfs2_inode
*ip
= GFS2_I(file
->f_path
.dentry
->d_inode
);
658 struct gfs2_glock
*gl
;
663 state
= (fl
->fl_type
== F_WRLCK
) ? LM_ST_EXCLUSIVE
: LM_ST_SHARED
;
664 flags
= (IS_SETLKW(cmd
) ? 0 : LM_FLAG_TRY
) | GL_EXACT
| GL_NOCACHE
;
666 mutex_lock(&fp
->f_fl_mutex
);
670 if (fl_gh
->gh_state
== state
)
672 flock_lock_file_wait(file
,
673 &(struct file_lock
){.fl_type
= F_UNLCK
});
674 gfs2_glock_dq_wait(fl_gh
);
675 gfs2_holder_reinit(state
, flags
, fl_gh
);
677 error
= gfs2_glock_get(GFS2_SB(&ip
->i_inode
), ip
->i_no_addr
,
678 &gfs2_flock_glops
, CREATE
, &gl
);
681 gfs2_holder_init(gl
, state
, flags
, fl_gh
);
684 error
= gfs2_glock_nq(fl_gh
);
686 gfs2_holder_uninit(fl_gh
);
687 if (error
== GLR_TRYFAILED
)
690 error
= flock_lock_file_wait(file
, fl
);
691 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
695 mutex_unlock(&fp
->f_fl_mutex
);
699 static void do_unflock(struct file
*file
, struct file_lock
*fl
)
701 struct gfs2_file
*fp
= file
->private_data
;
702 struct gfs2_holder
*fl_gh
= &fp
->f_fl_gh
;
704 mutex_lock(&fp
->f_fl_mutex
);
705 flock_lock_file_wait(file
, fl
);
707 gfs2_glock_dq_uninit(fl_gh
);
708 mutex_unlock(&fp
->f_fl_mutex
);
712 * gfs2_flock - acquire/release a flock lock on a file
713 * @file: the file pointer
714 * @cmd: either modify or retrieve lock state, possibly wait
715 * @fl: type and range of lock
720 static int gfs2_flock(struct file
*file
, int cmd
, struct file_lock
*fl
)
722 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
724 if (!(fl
->fl_flags
& FL_FLOCK
))
726 if (__mandatory_lock(&ip
->i_inode
))
729 if (fl
->fl_type
== F_UNLCK
) {
730 do_unflock(file
, fl
);
733 return do_flock(file
, cmd
, fl
);
737 const struct file_operations gfs2_file_fops
= {
738 .llseek
= gfs2_llseek
,
739 .read
= do_sync_read
,
740 .aio_read
= generic_file_aio_read
,
741 .write
= do_sync_write
,
742 .aio_write
= generic_file_aio_write
,
743 .unlocked_ioctl
= gfs2_ioctl
,
746 .release
= gfs2_close
,
750 .splice_read
= generic_file_splice_read
,
751 .splice_write
= generic_file_splice_write
,
752 .setlease
= gfs2_setlease
,
755 const struct file_operations gfs2_dir_fops
= {
756 .readdir
= gfs2_readdir
,
757 .unlocked_ioctl
= gfs2_ioctl
,
759 .release
= gfs2_close
,
765 const struct file_operations gfs2_file_fops_nolock
= {
766 .llseek
= gfs2_llseek
,
767 .read
= do_sync_read
,
768 .aio_read
= generic_file_aio_read
,
769 .write
= do_sync_write
,
770 .aio_write
= generic_file_aio_write
,
771 .unlocked_ioctl
= gfs2_ioctl
,
774 .release
= gfs2_close
,
776 .splice_read
= generic_file_splice_read
,
777 .splice_write
= generic_file_splice_write
,
778 .setlease
= gfs2_setlease
,
781 const struct file_operations gfs2_dir_fops_nolock
= {
782 .readdir
= gfs2_readdir
,
783 .unlocked_ioctl
= gfs2_ioctl
,
785 .release
= gfs2_close
,