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/sched.h>
11 #include <linux/slab.h>
12 #include <linux/spinlock.h>
13 #include <linux/completion.h>
14 #include <linux/buffer_head.h>
15 #include <linux/pagemap.h>
16 #include <linux/uio.h>
17 #include <linux/blkdev.h>
19 #include <linux/smp_lock.h>
21 #include <linux/gfs2_ondisk.h>
22 #include <linux/ext2_fs.h>
23 #include <linux/crc32.h>
24 #include <linux/lm_interface.h>
25 #include <linux/writeback.h>
26 #include <asm/uaccess.h>
46 /* For regular, non-NFS */
48 struct gfs2_sbd
*fdr_sbd
;
51 filldir_t fdr_filldir
;
56 * Most fields left uninitialised to catch anybody who tries to
57 * use them. f_flags set to prevent file_accessed() from touching
58 * any other part of this. Its use is purely as a flag so that we
59 * know (in readpage()) whether or not do to locking.
61 struct file gfs2_internal_file_sentinel
= {
62 .f_flags
= O_NOATIME
|O_RDONLY
,
65 static int gfs2_read_actor(read_descriptor_t
*desc
, struct page
*page
,
66 unsigned long offset
, unsigned long size
)
69 unsigned long count
= desc
->count
;
75 memcpy(desc
->arg
.data
, kaddr
+ offset
, size
);
78 desc
->count
= count
- size
;
79 desc
->written
+= size
;
80 desc
->arg
.buf
+= size
;
84 int gfs2_internal_read(struct gfs2_inode
*ip
, struct file_ra_state
*ra_state
,
85 char *buf
, loff_t
*pos
, unsigned size
)
87 struct inode
*inode
= &ip
->i_inode
;
88 read_descriptor_t desc
;
93 do_generic_mapping_read(inode
->i_mapping
, ra_state
,
94 &gfs2_internal_file_sentinel
, pos
, &desc
,
96 return desc
.written
? desc
.written
: desc
.error
;
100 * gfs2_llseek - seek to a location in a file
102 * @offset: the offset
103 * @origin: Where to seek from (SEEK_SET, SEEK_CUR, or SEEK_END)
105 * SEEK_END requires the glock for the file because it references the
108 * Returns: The new offset, or errno
111 static loff_t
gfs2_llseek(struct file
*file
, loff_t offset
, int origin
)
113 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
114 struct gfs2_holder i_gh
;
118 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, LM_FLAG_ANY
,
121 error
= remote_llseek(file
, offset
, origin
);
122 gfs2_glock_dq_uninit(&i_gh
);
125 error
= remote_llseek(file
, offset
, origin
);
131 * filldir_func - Report a directory entry to the caller of gfs2_dir_read()
132 * @opaque: opaque data used by the function
133 * @name: the name of the directory entry
134 * @length: the length of the name
135 * @offset: the entry's offset in the directory
136 * @inum: the inode number the entry points to
137 * @type: the type of inode the entry points to
139 * Returns: 0 on success, 1 if buffer full
142 static int filldir_func(void *opaque
, const char *name
, unsigned int length
,
143 u64 offset
, struct gfs2_inum_host
*inum
,
146 struct filldir_reg
*fdr
= (struct filldir_reg
*)opaque
;
147 struct gfs2_sbd
*sdp
= fdr
->fdr_sbd
;
150 error
= fdr
->fdr_filldir(fdr
->fdr_opaque
, name
, length
, offset
,
151 inum
->no_addr
, type
);
155 if (fdr
->fdr_prefetch
&& !(length
== 1 && *name
== '.')) {
156 gfs2_glock_prefetch_num(sdp
, inum
->no_addr
, &gfs2_inode_glops
,
157 LM_ST_SHARED
, LM_FLAG_TRY
| LM_FLAG_ANY
);
158 gfs2_glock_prefetch_num(sdp
, inum
->no_addr
, &gfs2_iopen_glops
,
159 LM_ST_SHARED
, LM_FLAG_TRY
);
166 * gfs2_readdir - Read directory entries from a directory
167 * @file: The directory to read from
168 * @dirent: Buffer for dirents
169 * @filldir: Function used to do the copying
174 static int gfs2_readdir(struct file
*file
, void *dirent
, filldir_t filldir
)
176 struct inode
*dir
= file
->f_mapping
->host
;
177 struct gfs2_inode
*dip
= GFS2_I(dir
);
178 struct filldir_reg fdr
;
179 struct gfs2_holder d_gh
;
180 u64 offset
= file
->f_pos
;
183 fdr
.fdr_sbd
= GFS2_SB(dir
);
184 fdr
.fdr_prefetch
= 1;
185 fdr
.fdr_filldir
= filldir
;
186 fdr
.fdr_opaque
= dirent
;
188 gfs2_holder_init(dip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &d_gh
);
189 error
= gfs2_glock_nq_atime(&d_gh
);
191 gfs2_holder_uninit(&d_gh
);
195 error
= gfs2_dir_read(dir
, &offset
, &fdr
, filldir_func
);
197 gfs2_glock_dq_uninit(&d_gh
);
199 file
->f_pos
= offset
;
206 * @table: A table of 32 u32 flags
207 * @val: a 32 bit value to convert
209 * This function can be used to convert between fsflags values and
210 * GFS2's own flags values.
212 * Returns: the converted flags
214 static u32
fsflags_cvt(const u32
*table
, u32 val
)
226 static const u32 fsflags_to_gfs2
[32] = {
228 [4] = GFS2_DIF_IMMUTABLE
,
229 [5] = GFS2_DIF_APPENDONLY
,
230 [7] = GFS2_DIF_NOATIME
,
231 [12] = GFS2_DIF_EXHASH
,
232 [14] = GFS2_DIF_JDATA
,
233 [20] = GFS2_DIF_DIRECTIO
,
236 static const u32 gfs2_to_fsflags
[32] = {
237 [gfs2fl_Sync
] = FS_SYNC_FL
,
238 [gfs2fl_Immutable
] = FS_IMMUTABLE_FL
,
239 [gfs2fl_AppendOnly
] = FS_APPEND_FL
,
240 [gfs2fl_NoAtime
] = FS_NOATIME_FL
,
241 [gfs2fl_ExHash
] = FS_INDEX_FL
,
242 [gfs2fl_Jdata
] = FS_JOURNAL_DATA_FL
,
243 [gfs2fl_Directio
] = FS_DIRECTIO_FL
,
244 [gfs2fl_InheritDirectio
] = FS_DIRECTIO_FL
,
245 [gfs2fl_InheritJdata
] = FS_JOURNAL_DATA_FL
,
248 static int gfs2_get_flags(struct file
*filp
, u32 __user
*ptr
)
250 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
251 struct gfs2_inode
*ip
= GFS2_I(inode
);
252 struct gfs2_holder gh
;
256 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &gh
);
257 error
= gfs2_glock_nq_atime(&gh
);
261 fsflags
= fsflags_cvt(gfs2_to_fsflags
, ip
->i_di
.di_flags
);
262 if (put_user(fsflags
, ptr
))
265 gfs2_glock_dq_m(1, &gh
);
266 gfs2_holder_uninit(&gh
);
270 void gfs2_set_inode_flags(struct inode
*inode
)
272 struct gfs2_inode
*ip
= GFS2_I(inode
);
273 struct gfs2_dinode_host
*di
= &ip
->i_di
;
274 unsigned int flags
= inode
->i_flags
;
276 flags
&= ~(S_SYNC
|S_APPEND
|S_IMMUTABLE
|S_NOATIME
|S_DIRSYNC
);
277 if (di
->di_flags
& GFS2_DIF_IMMUTABLE
)
278 flags
|= S_IMMUTABLE
;
279 if (di
->di_flags
& GFS2_DIF_APPENDONLY
)
281 if (di
->di_flags
& GFS2_DIF_NOATIME
)
283 if (di
->di_flags
& GFS2_DIF_SYNC
)
285 inode
->i_flags
= flags
;
288 /* Flags that can be set by user space */
289 #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \
291 GFS2_DIF_IMMUTABLE| \
292 GFS2_DIF_APPENDONLY| \
296 GFS2_DIF_INHERIT_DIRECTIO| \
297 GFS2_DIF_INHERIT_JDATA)
300 * gfs2_set_flags - set flags on an inode
302 * @flags: The flags to set
303 * @mask: Indicates which flags are valid
306 static int do_gfs2_set_flags(struct file
*filp
, u32 reqflags
, u32 mask
)
308 struct inode
*inode
= filp
->f_path
.dentry
->d_inode
;
309 struct gfs2_inode
*ip
= GFS2_I(inode
);
310 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
311 struct buffer_head
*bh
;
312 struct gfs2_holder gh
;
314 u32 new_flags
, flags
;
316 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
320 flags
= ip
->i_di
.di_flags
;
321 new_flags
= (flags
& ~mask
) | (reqflags
& mask
);
322 if ((new_flags
^ flags
) == 0)
325 if (S_ISDIR(inode
->i_mode
)) {
326 if ((new_flags
^ flags
) & GFS2_DIF_JDATA
)
327 new_flags
^= (GFS2_DIF_JDATA
|GFS2_DIF_INHERIT_JDATA
);
328 if ((new_flags
^ flags
) & GFS2_DIF_DIRECTIO
)
329 new_flags
^= (GFS2_DIF_DIRECTIO
|GFS2_DIF_INHERIT_DIRECTIO
);
333 if ((new_flags
^ flags
) & ~GFS2_FLAGS_USER_SET
)
337 if (IS_IMMUTABLE(inode
) && (new_flags
& GFS2_DIF_IMMUTABLE
))
339 if (IS_APPEND(inode
) && (new_flags
& GFS2_DIF_APPENDONLY
))
341 if (((new_flags
^ flags
) & GFS2_DIF_IMMUTABLE
) &&
342 !capable(CAP_LINUX_IMMUTABLE
))
344 if (!IS_IMMUTABLE(inode
)) {
345 error
= permission(inode
, MAY_WRITE
, NULL
);
350 error
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
353 error
= gfs2_meta_inode_buffer(ip
, &bh
);
356 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
357 ip
->i_di
.di_flags
= new_flags
;
358 gfs2_dinode_out(ip
, bh
->b_data
);
360 gfs2_set_inode_flags(inode
);
364 gfs2_glock_dq_uninit(&gh
);
368 static int gfs2_set_flags(struct file
*filp
, u32 __user
*ptr
)
370 u32 fsflags
, gfsflags
;
371 if (get_user(fsflags
, ptr
))
373 gfsflags
= fsflags_cvt(fsflags_to_gfs2
, fsflags
);
374 return do_gfs2_set_flags(filp
, gfsflags
, ~0);
377 static long gfs2_ioctl(struct file
*filp
, unsigned int cmd
, unsigned long arg
)
380 case FS_IOC_GETFLAGS
:
381 return gfs2_get_flags(filp
, (u32 __user
*)arg
);
382 case FS_IOC_SETFLAGS
:
383 return gfs2_set_flags(filp
, (u32 __user
*)arg
);
391 * @file: The file to map
392 * @vma: The VMA which described the mapping
394 * Returns: 0 or error code
397 static int gfs2_mmap(struct file
*file
, struct vm_area_struct
*vma
)
399 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
400 struct gfs2_holder i_gh
;
403 gfs2_holder_init(ip
->i_gl
, LM_ST_SHARED
, GL_ATIME
, &i_gh
);
404 error
= gfs2_glock_nq_atime(&i_gh
);
406 gfs2_holder_uninit(&i_gh
);
410 /* This is VM_MAYWRITE instead of VM_WRITE because a call
411 to mprotect() can turn on VM_WRITE later. */
413 if ((vma
->vm_flags
& (VM_MAYSHARE
| VM_MAYWRITE
)) ==
414 (VM_MAYSHARE
| VM_MAYWRITE
))
415 vma
->vm_ops
= &gfs2_vm_ops_sharewrite
;
417 vma
->vm_ops
= &gfs2_vm_ops_private
;
419 gfs2_glock_dq_uninit(&i_gh
);
425 * gfs2_open - open a file
426 * @inode: the inode to open
427 * @file: the struct file for this opening
432 static int gfs2_open(struct inode
*inode
, struct file
*file
)
434 struct gfs2_inode
*ip
= GFS2_I(inode
);
435 struct gfs2_holder i_gh
;
436 struct gfs2_file
*fp
;
439 fp
= kzalloc(sizeof(struct gfs2_file
), GFP_KERNEL
);
443 mutex_init(&fp
->f_fl_mutex
);
445 gfs2_assert_warn(GFS2_SB(inode
), !file
->private_data
);
446 file
->private_data
= fp
;
448 if (S_ISREG(ip
->i_inode
.i_mode
)) {
449 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, LM_FLAG_ANY
,
454 if (!(file
->f_flags
& O_LARGEFILE
) &&
455 ip
->i_di
.di_size
> MAX_NON_LFS
) {
460 /* Listen to the Direct I/O flag */
462 if (ip
->i_di
.di_flags
& GFS2_DIF_DIRECTIO
)
463 file
->f_flags
|= O_DIRECT
;
465 gfs2_glock_dq_uninit(&i_gh
);
471 gfs2_glock_dq_uninit(&i_gh
);
473 file
->private_data
= NULL
;
479 * gfs2_close - called to close a struct file
480 * @inode: the inode the struct file belongs to
481 * @file: the struct file being closed
486 static int gfs2_close(struct inode
*inode
, struct file
*file
)
488 struct gfs2_sbd
*sdp
= inode
->i_sb
->s_fs_info
;
489 struct gfs2_file
*fp
;
491 fp
= file
->private_data
;
492 file
->private_data
= NULL
;
494 if (gfs2_assert_warn(sdp
, fp
))
503 * gfs2_fsync - sync the dirty data for a file (across the cluster)
504 * @file: the file that points to the dentry (we ignore this)
505 * @dentry: the dentry that points to the inode to sync
507 * The VFS will flush "normal" data for us. We only need to worry
508 * about metadata here. For journaled data, we just do a log flush
509 * as we can't avoid it. Otherwise we can just bale out if datasync
510 * is set. For stuffed inodes we must flush the log in order to
511 * ensure that all data is on disk.
513 * The call to write_inode_now() is there to write back metadata and
514 * the inode itself. It does also try and write the data, but thats
515 * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
521 static int gfs2_fsync(struct file
*file
, struct dentry
*dentry
, int datasync
)
523 struct inode
*inode
= dentry
->d_inode
;
524 int sync_state
= inode
->i_state
& (I_DIRTY_SYNC
|I_DIRTY_DATASYNC
);
527 if (gfs2_is_jdata(GFS2_I(inode
))) {
528 gfs2_log_flush(GFS2_SB(inode
), GFS2_I(inode
)->i_gl
);
532 if (sync_state
!= 0) {
534 ret
= write_inode_now(inode
, 0);
536 if (gfs2_is_stuffed(GFS2_I(inode
)))
537 gfs2_log_flush(GFS2_SB(inode
), GFS2_I(inode
)->i_gl
);
544 * gfs2_lock - acquire/release a posix lock on a file
545 * @file: the file pointer
546 * @cmd: either modify or retrieve lock state, possibly wait
547 * @fl: type and range of lock
552 static int gfs2_lock(struct file
*file
, int cmd
, struct file_lock
*fl
)
554 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
555 struct gfs2_sbd
*sdp
= GFS2_SB(file
->f_mapping
->host
);
556 struct lm_lockname name
=
557 { .ln_number
= ip
->i_num
.no_addr
,
558 .ln_type
= LM_TYPE_PLOCK
};
560 if (!(fl
->fl_flags
& FL_POSIX
))
562 if ((ip
->i_inode
.i_mode
& (S_ISGID
| S_IXGRP
)) == S_ISGID
)
565 if (sdp
->sd_args
.ar_localflocks
) {
567 struct file_lock tmp
;
569 ret
= posix_test_lock(file
, fl
, &tmp
);
570 fl
->fl_type
= F_UNLCK
;
572 memcpy(fl
, &tmp
, sizeof(struct file_lock
));
575 return posix_lock_file_wait(file
, fl
);
580 return gfs2_lm_plock_get(sdp
, &name
, file
, fl
);
581 else if (fl
->fl_type
== F_UNLCK
)
582 return gfs2_lm_punlock(sdp
, &name
, file
, fl
);
584 return gfs2_lm_plock(sdp
, &name
, file
, cmd
, fl
);
587 static int do_flock(struct file
*file
, int cmd
, struct file_lock
*fl
)
589 struct gfs2_file
*fp
= file
->private_data
;
590 struct gfs2_holder
*fl_gh
= &fp
->f_fl_gh
;
591 struct gfs2_inode
*ip
= GFS2_I(file
->f_path
.dentry
->d_inode
);
592 struct gfs2_glock
*gl
;
597 state
= (fl
->fl_type
== F_WRLCK
) ? LM_ST_EXCLUSIVE
: LM_ST_SHARED
;
598 flags
= (IS_SETLKW(cmd
) ? 0 : LM_FLAG_TRY
) | GL_EXACT
| GL_NOCACHE
;
600 mutex_lock(&fp
->f_fl_mutex
);
604 if (fl_gh
->gh_state
== state
)
607 flock_lock_file_wait(file
,
608 &(struct file_lock
){.fl_type
= F_UNLCK
});
609 gfs2_glock_dq_uninit(fl_gh
);
611 error
= gfs2_glock_get(GFS2_SB(&ip
->i_inode
),
612 ip
->i_num
.no_addr
, &gfs2_flock_glops
,
618 gfs2_holder_init(gl
, state
, flags
, fl_gh
);
621 error
= gfs2_glock_nq(fl_gh
);
623 gfs2_holder_uninit(fl_gh
);
624 if (error
== GLR_TRYFAILED
)
627 error
= flock_lock_file_wait(file
, fl
);
628 gfs2_assert_warn(GFS2_SB(&ip
->i_inode
), !error
);
632 mutex_unlock(&fp
->f_fl_mutex
);
636 static void do_unflock(struct file
*file
, struct file_lock
*fl
)
638 struct gfs2_file
*fp
= file
->private_data
;
639 struct gfs2_holder
*fl_gh
= &fp
->f_fl_gh
;
641 mutex_lock(&fp
->f_fl_mutex
);
642 flock_lock_file_wait(file
, fl
);
644 gfs2_glock_dq_uninit(fl_gh
);
645 mutex_unlock(&fp
->f_fl_mutex
);
649 * gfs2_flock - acquire/release a flock lock on a file
650 * @file: the file pointer
651 * @cmd: either modify or retrieve lock state, possibly wait
652 * @fl: type and range of lock
657 static int gfs2_flock(struct file
*file
, int cmd
, struct file_lock
*fl
)
659 struct gfs2_inode
*ip
= GFS2_I(file
->f_mapping
->host
);
660 struct gfs2_sbd
*sdp
= GFS2_SB(file
->f_mapping
->host
);
662 if (!(fl
->fl_flags
& FL_FLOCK
))
664 if ((ip
->i_inode
.i_mode
& (S_ISGID
| S_IXGRP
)) == S_ISGID
)
667 if (sdp
->sd_args
.ar_localflocks
)
668 return flock_lock_file_wait(file
, fl
);
670 if (fl
->fl_type
== F_UNLCK
) {
671 do_unflock(file
, fl
);
674 return do_flock(file
, cmd
, fl
);
678 const struct file_operations gfs2_file_fops
= {
679 .llseek
= gfs2_llseek
,
680 .read
= do_sync_read
,
681 .aio_read
= generic_file_aio_read
,
682 .write
= do_sync_write
,
683 .aio_write
= generic_file_aio_write
,
684 .unlocked_ioctl
= gfs2_ioctl
,
687 .release
= gfs2_close
,
690 .sendfile
= generic_file_sendfile
,
692 .splice_read
= generic_file_splice_read
,
693 .splice_write
= generic_file_splice_write
,
696 const struct file_operations gfs2_dir_fops
= {
697 .readdir
= gfs2_readdir
,
698 .unlocked_ioctl
= gfs2_ioctl
,
700 .release
= gfs2_close
,