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"
44 * gfs2_llseek - seek to a location in a file
47 * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END)
49 * SEEK_END requires the glock for the file because it references the
52 * Returns: The new offset, or errno
55 static loff_t
gfs2_llseek(struct file
*file
, loff_t offset
, int origin
)
57 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
58 struct gfs2_holder i_gh
;
62 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, LM_FLAG_ANY
,
65 error
= generic_file_llseek_unlocked(file
, offset
, origin
);
66 gfs2_glock_dq_uninit(&i_gh
);
69 error
= generic_file_llseek_unlocked(file
, offset
, origin
);
75 * gfs2_readdir - Read directory entries from a directory
76 * @file: The directory to read from
77 * @dirent: Buffer for dirents
78 * @filldir: Function used to do the copying
83 static int gfs2_readdir(struct file
*file
, void *dirent
, filldir_t filldir
)
85 struct inode
*dir
= file
->f_mapping
->host
;
86 struct gfs2_inode
*dip
= GFS2_I(dir
);
87 struct gfs2_holder d_gh
;
88 u64 offset
= file
->f_pos
;
91 gfs2_holder_init(dip
->i_gl
, LM_ST_SHARED
, 0, &d_gh
);
92 error
= gfs2_glock_nq(&d_gh
);
94 gfs2_holder_uninit(&d_gh
);
98 error
= gfs2_dir_read(dir
, &offset
, dirent
, filldir
);
100 gfs2_glock_dq_uninit(&d_gh
);
102 file
->f_pos
= offset
;
109 * @table: A table of 32 u32 flags
110 * @val: a 32 bit value to convert
112 * This function can be used to convert between fsflags values and
113 * GFS2's own flags values.
115 * Returns: the converted flags
117 static u32
fsflags_cvt(const u32
*table
, u32 val
)
129 static const u32 fsflags_to_gfs2
[32] = {
131 [4] = GFS2_DIF_IMMUTABLE
,
132 [5] = GFS2_DIF_APPENDONLY
,
133 [7] = GFS2_DIF_NOATIME
,
134 [12] = GFS2_DIF_EXHASH
,
135 [14] = GFS2_DIF_INHERIT_JDATA
,
138 static const u32 gfs2_to_fsflags
[32] = {
139 [gfs2fl_Sync
] = FS_SYNC_FL
,
140 [gfs2fl_Immutable
] = FS_IMMUTABLE_FL
,
141 [gfs2fl_AppendOnly
] = FS_APPEND_FL
,
142 [gfs2fl_NoAtime
] = FS_NOATIME_FL
,
143 [gfs2fl_ExHash
] = FS_INDEX_FL
,
144 [gfs2fl_InheritJdata
] = FS_JOURNAL_DATA_FL
,
147 static int gfs2_get_flags(struct file
*filp
, u32 __user
*ptr
)
149 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
150 struct gfs2_inode
*ip
= GFS2_I(inode
);
151 struct gfs2_holder gh
;
155 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, 0, &gh
);
156 error
= gfs2_glock_nq(&gh
);
160 fsflags
= fsflags_cvt(gfs2_to_fsflags
, ip
->i_diskflags
);
161 if (!S_ISDIR(inode
->i_mode
) && ip
->i_diskflags
& GFS2_DIF_JDATA
)
162 fsflags
|= FS_JOURNAL_DATA_FL
;
163 if (put_user(fsflags
, ptr
))
167 gfs2_holder_uninit(&gh
);
171 void gfs2_set_inode_flags(struct inode
*inode
)
173 struct gfs2_inode
*ip
= GFS2_I(inode
);
174 unsigned int flags
= inode
->i_flags
;
176 flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
177 if (ip
->i_diskflags
& GFS2_DIF_IMMUTABLE
)
178 flags
|= S_IMMUTABLE
;
179 if (ip
->i_diskflags
& GFS2_DIF_APPENDONLY
)
181 if (ip
->i_diskflags
& GFS2_DIF_NOATIME
)
183 if (ip
->i_diskflags
& GFS2_DIF_SYNC
)
185 inode
->i_flags
= flags
;
188 /* Flags that can be set by user space */
189 #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \
190 GFS2_DIF_IMMUTABLE| \
191 GFS2_DIF_APPENDONLY| \
195 GFS2_DIF_INHERIT_JDATA)
198 * gfs2_set_flags - set flags on an inode
200 * @flags: The flags to set
201 * @mask: Indicates which flags are valid
204 static int do_gfs2_set_flags(struct file
*filp
, u32 reqflags
, u32 mask
)
206 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
207 struct gfs2_inode
*ip
= GFS2_I(inode
);
208 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
209 struct buffer_head
*bh
;
210 struct gfs2_holder gh
;
212 u32 new_flags
, flags
;
214 error
= mnt_want_write(filp
->f_path
.mnt
);
218 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
222 flags
= ip
->i_diskflags
;
223 new_flags
= (flags
& ~mask
) | (reqflags
& mask
);
224 if ((new_flags
^ flags
) == 0)
228 if ((new_flags
^ flags
) & ~GFS2_FLAGS_USER_SET
)
232 if (IS_IMMUTABLE(inode
) && (new_flags
& GFS2_DIF_IMMUTABLE
))
234 if (IS_APPEND(inode
) && (new_flags
& GFS2_DIF_APPENDONLY
))
236 if (((new_flags
^ flags
) & GFS2_DIF_IMMUTABLE
) &&
237 !capable(CAP_LINUX_IMMUTABLE
))
239 if (!IS_IMMUTABLE(inode
)) {
240 error
= gfs2_permission(inode
, MAY_WRITE
);
244 if ((flags
^ new_flags
) & GFS2_DIF_JDATA
) {
245 if (flags
& GFS2_DIF_JDATA
)
246 gfs2_log_flush(sdp
, ip
->i_gl
);
247 error
= filemap_fdatawrite(inode
->i_mapping
);
250 error
= filemap_fdatawait(inode
->i_mapping
);
254 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
257 error
= gfs2_meta_inode_buffer(ip
, &bh
);
260 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
261 ip
->i_diskflags
= new_flags
;
262 gfs2_dinode_out(ip
, bh
->b_data
);
264 gfs2_set_inode_flags(inode
);
265 gfs2_set_aops(inode
);
269 gfs2_glock_dq_uninit(&gh
);
271 mnt_drop_write(filp
->f_path
.mnt
);
275 static int gfs2_set_flags(struct file
*filp
, u32 __user
*ptr
)
277 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
278 u32 fsflags
, gfsflags
;
279 if (get_user(fsflags
, ptr
))
281 gfsflags
= fsflags_cvt(fsflags_to_gfs2
, fsflags
);
282 if (!S_ISDIR(inode
->i_mode
)) {
283 if (gfsflags
& GFS2_DIF_INHERIT_JDATA
)
284 gfsflags
^= (GFS2_DIF_JDATA
| GFS2_DIF_INHERIT_JDATA
);
285 return do_gfs2_set_flags(filp
, gfsflags
, ~0);
287 return do_gfs2_set_flags(filp
, gfsflags
, ~GFS2_DIF_JDATA
);
290 static long gfs2_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
293 case FS_IOC_GETFLAGS
:
294 return gfs2_get_flags(filp
, (u32 __user
*)arg
);
295 case FS_IOC_SETFLAGS
:
296 return gfs2_set_flags(filp
, (u32 __user
*)arg
);
302 * gfs2_allocate_page_backing - Use bmap to allocate blocks
303 * @page: The (locked) page to allocate backing for
305 * We try to allocate all the blocks required for the page in
306 * one go. This might fail for various reasons, so we keep
307 * trying until all the blocks to back this page are allocated.
308 * If some of the blocks are already allocated, thats ok too.
311 static int gfs2_allocate_page_backing(struct page
*page
)
313 struct inode
*inode
= page
->mapping
->host
;
314 struct buffer_head bh
;
315 unsigned long size
= PAGE_CACHE_SIZE
;
316 u64 lblock
= page
->index
<< (PAGE_CACHE_SHIFT
- inode
->i_blkbits
);
321 gfs2_block_map(inode
, lblock
, &bh
, 1);
322 if (!buffer_mapped(&bh
))
325 lblock
+= (bh
.b_size
>> inode
->i_blkbits
);
331 * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable
332 * @vma: The virtual memory area
333 * @page: The page which is about to become writable
335 * When the page becomes writable, we need to ensure that we have
336 * blocks allocated on disk to back that page.
339 static int gfs2_page_mkwrite(struct vm_area_struct
*vma
, struct page
*page
)
341 struct inode
*inode
= vma
->vm_file
->f_path
.dentry
->d_inode
;
342 struct gfs2_inode
*ip
= GFS2_I(inode
);
343 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
344 unsigned long last_index
;
345 u64 pos
= page
->index
<< PAGE_CACHE_SHIFT
;
346 unsigned int data_blocks
, ind_blocks
, rblocks
;
347 int alloc_required
= 0;
348 struct gfs2_holder gh
;
349 struct gfs2_alloc
*al
;
352 gfs2_holder_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
353 ret
= gfs2_glock_nq(&gh
);
357 set_bit(GIF_SW_PAGED
, &ip
->i_flags
);
358 ret
= gfs2_write_alloc_required(ip
, pos
, PAGE_CACHE_SIZE
, &alloc_required
);
359 if (ret
|| !alloc_required
)
362 al
= gfs2_alloc_get(ip
);
366 ret
= gfs2_quota_lock_check(ip
);
369 gfs2_write_calc_reserv(ip
, PAGE_CACHE_SIZE
, &data_blocks
, &ind_blocks
);
370 al
->al_requested
= data_blocks
+ ind_blocks
;
371 ret
= gfs2_inplace_reserve(ip
);
373 goto out_quota_unlock
;
375 rblocks
= RES_DINODE
+ ind_blocks
;
376 if (gfs2_is_jdata(ip
))
377 rblocks
+= data_blocks
? data_blocks
: 1;
378 if (ind_blocks
|| data_blocks
)
379 rblocks
+= RES_STATFS
+ RES_QUOTA
;
380 ret
= gfs2_trans_begin(sdp
, rblocks
, 0);
386 last_index
= ip
->i_inode
.i_size
>> PAGE_CACHE_SHIFT
;
387 if (page
->index
> last_index
)
388 goto out_unlock_page
;
390 if (!PageUptodate(page
) || page
->mapping
!= ip
->i_inode
.i_mapping
)
391 goto out_unlock_page
;
392 if (gfs2_is_stuffed(ip
)) {
393 ret
= gfs2_unstuff_dinode(ip
, page
);
395 goto out_unlock_page
;
397 ret
= gfs2_allocate_page_backing(page
);
403 gfs2_inplace_release(ip
);
405 gfs2_quota_unlock(ip
);
411 gfs2_holder_uninit(&gh
);
415 static struct vm_operations_struct gfs2_vm_ops
= {
416 .fault
= filemap_fault
,
417 .page_mkwrite
= gfs2_page_mkwrite
,
423 * @file: The file to map
424 * @vma: The VMA which described the mapping
426 * Returns: 0 or error code
429 static int gfs2_mmap(struct file
*file
, struct vm_area_struct
*vma
)
431 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
432 struct gfs2_holder i_gh
;
435 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, 0, &i_gh
);
436 error
= gfs2_glock_nq(&i_gh
);
438 gfs2_holder_uninit(&i_gh
);
442 vma
->vm_ops
= &gfs2_vm_ops
;
444 gfs2_glock_dq_uninit(&i_gh
);
450 * gfs2_open - open a file
451 * @inode: the inode to open
452 * @file: the struct file for this opening
457 static int gfs2_open(struct inode
*inode
, struct file
*file
)
459 struct gfs2_inode
*ip
= GFS2_I(inode
);
460 struct gfs2_holder i_gh
;
461 struct gfs2_file
*fp
;
464 fp
= kzalloc(sizeof(struct gfs2_file
), GFP_KERNEL
);
468 mutex_init(&fp
->f_fl_mutex
);
470 gfs2_assert_warn(GFS2_SB(inode
), !file
->private_data
);
471 file
->private_data
= fp
;
473 if (S_ISREG(ip
->i_inode
.i_mode
)) {
474 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, LM_FLAG_ANY
,
479 if (!(file
->f_flags
& O_LARGEFILE
) &&
480 ip
->i_disksize
> MAX_NON_LFS
) {
485 gfs2_glock_dq_uninit(&i_gh
);
491 gfs2_glock_dq_uninit(&i_gh
);
493 file
->private_data
= NULL
;
499 * gfs2_close - called to close a struct file
500 * @inode: the inode the struct file belongs to
501 * @file: the struct file being closed
506 static int gfs2_close(struct inode
*inode
, struct file
*file
)
508 struct gfs2_sbd
*sdp
= inode
->i_sb
->s_fs_info
;
509 struct gfs2_file
*fp
;
511 fp
= file
->private_data
;
512 file
->private_data
= NULL
;
514 if (gfs2_assert_warn(sdp
, fp
))
523 * gfs2_fsync - sync the dirty data for a file (across the cluster)
524 * @file: the file that points to the dentry (we ignore this)
525 * @dentry: the dentry that points to the inode to sync
527 * The VFS will flush "normal" data for us. We only need to worry
528 * about metadata here. For journaled data, we just do a log flush
529 * as we can't avoid it. Otherwise we can just bale out if datasync
530 * is set. For stuffed inodes we must flush the log in order to
531 * ensure that all data is on disk.
533 * The call to write_inode_now() is there to write back metadata and
534 * the inode itself. It does also try and write the data, but thats
535 * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
541 static int gfs2_fsync(struct file
*file
, struct dentry
*dentry
, int datasync
)
543 struct inode
*inode
= dentry
->d_inode
;
544 int sync_state
= inode
->i_state
& (I_DIRTY_SYNC
|I_DIRTY_DATASYNC
);
547 if (gfs2_is_jdata(GFS2_I(inode
))) {
548 gfs2_log_flush(GFS2_SB(inode
), GFS2_I(inode
)->i_gl
);
552 if (sync_state
!= 0) {
554 ret
= write_inode_now(inode
, 0);
556 if (gfs2_is_stuffed(GFS2_I(inode
)))
557 gfs2_log_flush(GFS2_SB(inode
), GFS2_I(inode
)->i_gl
);
564 * gfs2_setlease - acquire/release a file lease
565 * @file: the file pointer
572 static int gfs2_setlease(struct file
*file
, long arg
, struct file_lock
**fl
)
574 struct gfs2_sbd
*sdp
= GFS2_SB(file
->f_mapping
->host
);
577 * We don't currently have a way to enforce a lease across the whole
578 * cluster; until we do, disable leases (by just returning -EINVAL),
579 * unless the administrator has requested purely local locking.
581 if (!sdp
->sd_args
.ar_localflocks
)
583 return generic_setlease(file
, arg
, fl
);
586 static int gfs2_lm_plock_get(struct gfs2_sbd
*sdp
, struct lm_lockname
*name
,
587 struct file
*file
, struct file_lock
*fl
)
590 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
591 error
= sdp
->sd_lockstruct
.ls_ops
->lm_plock_get(
592 sdp
->sd_lockstruct
.ls_lockspace
, name
, file
, fl
);
596 static int gfs2_lm_plock(struct gfs2_sbd
*sdp
, struct lm_lockname
*name
,
597 struct file
*file
, int cmd
, struct file_lock
*fl
)
600 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
601 error
= sdp
->sd_lockstruct
.ls_ops
->lm_plock(
602 sdp
->sd_lockstruct
.ls_lockspace
, name
, file
, cmd
, fl
);
606 static int gfs2_lm_punlock(struct gfs2_sbd
*sdp
, struct lm_lockname
*name
,
607 struct file
*file
, struct file_lock
*fl
)
610 if (likely(!test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
)))
611 error
= sdp
->sd_lockstruct
.ls_ops
->lm_punlock(
612 sdp
->sd_lockstruct
.ls_lockspace
, name
, file
, fl
);
617 * gfs2_lock - acquire/release a posix lock on a file
618 * @file: the file pointer
619 * @cmd: either modify or retrieve lock state, possibly wait
620 * @fl: type and range of lock
625 static int gfs2_lock(struct file
*file
, int cmd
, struct file_lock
*fl
)
627 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
628 struct gfs2_sbd
*sdp
= GFS2_SB(file
->f_mapping
->host
);
629 struct lm_lockname name
=
630 { .ln_number
= ip
->i_no_addr
,
631 .ln_type
= LM_TYPE_PLOCK
};
633 if (!(fl
->fl_flags
& FL_POSIX
))
635 if (__mandatory_lock(&ip
->i_inode
))
638 if (cmd
== F_CANCELLK
) {
641 fl
->fl_type
= F_UNLCK
;
644 return gfs2_lm_plock_get(sdp
, &name
, file
, fl
);
645 else if (fl
->fl_type
== F_UNLCK
)
646 return gfs2_lm_punlock(sdp
, &name
, file
, fl
);
648 return gfs2_lm_plock(sdp
, &name
, file
, cmd
, fl
);
651 static int do_flock(struct file
*file
, int cmd
, struct file_lock
*fl
)
653 struct gfs2_file
*fp
= file
->private_data
;
654 struct gfs2_holder
*fl_gh
= &fp
->f_fl_gh
;
655 struct gfs2_inode
*ip
= GFS2_I(file
->f_path
.dentry
->d_inode
);
656 struct gfs2_glock
*gl
;
661 state
= (fl
->fl_type
== F_WRLCK
) ? LM_ST_EXCLUSIVE
: LM_ST_SHARED
;
662 flags
= (IS_SETLKW(cmd
) ? 0 : LM_FLAG_TRY
) | GL_EXACT
| GL_NOCACHE
;
664 mutex_lock(&fp
->f_fl_mutex
);
668 if (fl_gh
->gh_state
== state
)
670 flock_lock_file_wait(file
,
671 &(struct file_lock
){.fl_type
= F_UNLCK
});
672 gfs2_glock_dq_wait(fl_gh
);
673 gfs2_holder_reinit(state
, flags
, fl_gh
);
675 error
= gfs2_glock_get(GFS2_SB(&ip
->i_inode
), ip
->i_no_addr
,
676 &gfs2_flock_glops
, CREATE
, &gl
);
679 gfs2_holder_init(gl
, state
, flags
, fl_gh
);
682 error
= gfs2_glock_nq(fl_gh
);
684 gfs2_holder_uninit(fl_gh
);
685 if (error
== GLR_TRYFAILED
)
688 error
= flock_lock_file_wait(file
, fl
);
689 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
693 mutex_unlock(&fp
->f_fl_mutex
);
697 static void do_unflock(struct file
*file
, struct file_lock
*fl
)
699 struct gfs2_file
*fp
= file
->private_data
;
700 struct gfs2_holder
*fl_gh
= &fp
->f_fl_gh
;
702 mutex_lock(&fp
->f_fl_mutex
);
703 flock_lock_file_wait(file
, fl
);
705 gfs2_glock_dq_uninit(fl_gh
);
706 mutex_unlock(&fp
->f_fl_mutex
);
710 * gfs2_flock - acquire/release a flock lock on a file
711 * @file: the file pointer
712 * @cmd: either modify or retrieve lock state, possibly wait
713 * @fl: type and range of lock
718 static int gfs2_flock(struct file
*file
, int cmd
, struct file_lock
*fl
)
720 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
722 if (!(fl
->fl_flags
& FL_FLOCK
))
724 if (__mandatory_lock(&ip
->i_inode
))
727 if (fl
->fl_type
== F_UNLCK
) {
728 do_unflock(file
, fl
);
731 return do_flock(file
, cmd
, fl
);
735 const struct file_operations gfs2_file_fops
= {
736 .llseek
= gfs2_llseek
,
737 .read
= do_sync_read
,
738 .aio_read
= generic_file_aio_read
,
739 .write
= do_sync_write
,
740 .aio_write
= generic_file_aio_write
,
741 .unlocked_ioctl
= gfs2_ioctl
,
744 .release
= gfs2_close
,
748 .splice_read
= generic_file_splice_read
,
749 .splice_write
= generic_file_splice_write
,
750 .setlease
= gfs2_setlease
,
753 const struct file_operations gfs2_dir_fops
= {
754 .readdir
= gfs2_readdir
,
755 .unlocked_ioctl
= gfs2_ioctl
,
757 .release
= gfs2_close
,
763 const struct file_operations gfs2_file_fops_nolock
= {
764 .llseek
= gfs2_llseek
,
765 .read
= do_sync_read
,
766 .aio_read
= generic_file_aio_read
,
767 .write
= do_sync_write
,
768 .aio_write
= generic_file_aio_write
,
769 .unlocked_ioctl
= gfs2_ioctl
,
772 .release
= gfs2_close
,
774 .splice_read
= generic_file_splice_read
,
775 .splice_write
= generic_file_splice_write
,
776 .setlease
= gfs2_setlease
,
779 const struct file_operations gfs2_dir_fops_nolock
= {
780 .readdir
= gfs2_readdir
,
781 .unlocked_ioctl
= gfs2_ioctl
,
783 .release
= gfs2_close
,