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>
19 #include <linux/gfs2_ondisk.h>
20 #include <linux/ext2_fs.h>
21 #include <linux/crc32.h>
22 #include <linux/lm_interface.h>
23 #include <linux/writeback.h>
24 #include <asm/uaccess.h>
40 #include "ops_address.h"
41 #include "ops_inode.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, GL_ATIME, &d_gh);
92 error = gfs2_glock_nq_atime(&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,
136 [20] = GFS2_DIF_INHERIT_DIRECTIO,
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_InheritDirectio] = FS_DIRECTIO_FL,
146 [gfs2fl_InheritJdata] = FS_JOURNAL_DATA_FL,
149 static int gfs2_get_flags(struct file *filp, u32 __user *ptr)
151 struct inode *inode = filp->f_path.dentry->d_inode;
152 struct gfs2_inode *ip = GFS2_I(inode);
153 struct gfs2_holder gh;
157 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &gh);
158 error = gfs2_glock_nq_atime(&gh);
162 fsflags = fsflags_cvt(gfs2_to_fsflags, ip->i_di.di_flags);
163 if (!S_ISDIR(inode->i_mode)) {
164 if (ip->i_di.di_flags & GFS2_DIF_JDATA)
165 fsflags |= FS_JOURNAL_DATA_FL;
166 if (ip->i_di.di_flags & GFS2_DIF_DIRECTIO)
167 fsflags |= FS_DIRECTIO_FL;
169 if (put_user(fsflags, ptr))
173 gfs2_holder_uninit(&gh);
177 void gfs2_set_inode_flags(struct inode *inode)
179 struct gfs2_inode *ip = GFS2_I(inode);
180 struct gfs2_dinode_host *di = &ip->i_di;
181 unsigned int flags = inode->i_flags;
183 flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
184 if (di->di_flags & GFS2_DIF_IMMUTABLE)
185 flags |= S_IMMUTABLE;
186 if (di->di_flags & GFS2_DIF_APPENDONLY)
188 if (di->di_flags & GFS2_DIF_NOATIME)
190 if (di->di_flags & GFS2_DIF_SYNC)
192 inode->i_flags = flags;
195 /* Flags that can be set by user space */
196 #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \
198 GFS2_DIF_IMMUTABLE| \
199 GFS2_DIF_APPENDONLY| \
203 GFS2_DIF_INHERIT_DIRECTIO| \
204 GFS2_DIF_INHERIT_JDATA)
207 * gfs2_set_flags - set flags on an inode
209 * @flags: The flags to set
210 * @mask: Indicates which flags are valid
213 static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask)
215 struct inode *inode = filp->f_path.dentry->d_inode;
216 struct gfs2_inode *ip = GFS2_I(inode);
217 struct gfs2_sbd *sdp = GFS2_SB(inode);
218 struct buffer_head *bh;
219 struct gfs2_holder gh;
221 u32 new_flags, flags;
223 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
227 flags = ip->i_di.di_flags;
228 new_flags = (flags & ~mask) | (reqflags & mask);
229 if ((new_flags ^ flags) == 0)
233 if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET)
237 if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE))
239 if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY))
241 if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) &&
242 !capable(CAP_LINUX_IMMUTABLE))
244 if (!IS_IMMUTABLE(inode)) {
245 error = permission(inode, MAY_WRITE, NULL);
249 if ((flags ^ new_flags) & GFS2_DIF_JDATA) {
250 if (flags & GFS2_DIF_JDATA)
251 gfs2_log_flush(sdp, ip->i_gl);
252 error = filemap_fdatawrite(inode->i_mapping);
255 error = filemap_fdatawait(inode->i_mapping);
259 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
262 error = gfs2_meta_inode_buffer(ip, &bh);
265 gfs2_trans_add_bh(ip->i_gl, bh, 1);
266 ip->i_di.di_flags = new_flags;
267 gfs2_dinode_out(ip, bh->b_data);
269 gfs2_set_inode_flags(inode);
270 gfs2_set_aops(inode);
274 gfs2_glock_dq_uninit(&gh);
278 static int gfs2_set_flags(struct file *filp, u32 __user *ptr)
280 struct inode *inode = filp->f_path.dentry->d_inode;
281 u32 fsflags, gfsflags;
282 if (get_user(fsflags, ptr))
284 gfsflags = fsflags_cvt(fsflags_to_gfs2, fsflags);
285 if (!S_ISDIR(inode->i_mode)) {
286 if (gfsflags & GFS2_DIF_INHERIT_JDATA)
287 gfsflags ^= (GFS2_DIF_JDATA | GFS2_DIF_INHERIT_JDATA);
288 if (gfsflags & GFS2_DIF_INHERIT_DIRECTIO)
289 gfsflags ^= (GFS2_DIF_DIRECTIO | GFS2_DIF_INHERIT_DIRECTIO);
290 return do_gfs2_set_flags(filp, gfsflags, ~0);
292 return do_gfs2_set_flags(filp, gfsflags, ~GFS2_DIF_JDATA);
295 static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
298 case FS_IOC_GETFLAGS:
299 return gfs2_get_flags(filp, (u32 __user *)arg);
300 case FS_IOC_SETFLAGS:
301 return gfs2_set_flags(filp, (u32 __user *)arg);
307 * gfs2_allocate_page_backing - Use bmap to allocate blocks
308 * @page: The (locked) page to allocate backing for
310 * We try to allocate all the blocks required for the page in
311 * one go. This might fail for various reasons, so we keep
312 * trying until all the blocks to back this page are allocated.
313 * If some of the blocks are already allocated, thats ok too.
316 static int gfs2_allocate_page_backing(struct page *page)
318 struct inode *inode = page->mapping->host;
319 struct buffer_head bh;
320 unsigned long size = PAGE_CACHE_SIZE;
321 u64 lblock = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
326 gfs2_block_map(inode, lblock, &bh, 1);
327 if (!buffer_mapped(&bh))
330 lblock += (bh.b_size >> inode->i_blkbits);
336 * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable
337 * @vma: The virtual memory area
338 * @page: The page which is about to become writable
340 * When the page becomes writable, we need to ensure that we have
341 * blocks allocated on disk to back that page.
344 static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct page *page)
346 struct inode *inode = vma->vm_file->f_path.dentry->d_inode;
347 struct gfs2_inode *ip = GFS2_I(inode);
348 struct gfs2_sbd *sdp = GFS2_SB(inode);
349 unsigned long last_index;
350 u64 pos = page->index << (PAGE_CACHE_SIZE - inode->i_blkbits);
351 unsigned int data_blocks, ind_blocks, rblocks;
352 int alloc_required = 0;
353 struct gfs2_holder gh;
354 struct gfs2_alloc *al;
357 gfs2_holder_init(ip->i_gl, LM_ST_EXCLUSIVE, GL_ATIME, &gh);
358 ret = gfs2_glock_nq_atime(&gh);
362 set_bit(GIF_SW_PAGED, &ip->i_flags);
363 gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks);
364 ret = gfs2_write_alloc_required(ip, pos, PAGE_CACHE_SIZE, &alloc_required);
365 if (ret || !alloc_required)
368 al = gfs2_alloc_get(ip);
372 ret = gfs2_quota_lock_check(ip);
375 al->al_requested = data_blocks + ind_blocks;
376 ret = gfs2_inplace_reserve(ip);
378 goto out_quota_unlock;
380 rblocks = RES_DINODE + ind_blocks;
381 if (gfs2_is_jdata(ip))
382 rblocks += data_blocks ? data_blocks : 1;
383 if (ind_blocks || data_blocks)
384 rblocks += RES_STATFS + RES_QUOTA;
385 ret = gfs2_trans_begin(sdp, rblocks, 0);
391 last_index = ip->i_inode.i_size >> PAGE_CACHE_SHIFT;
392 if (page->index > last_index)
393 goto out_unlock_page;
395 if (!PageUptodate(page) || page->mapping != ip->i_inode.i_mapping)
396 goto out_unlock_page;
397 if (gfs2_is_stuffed(ip)) {
398 ret = gfs2_unstuff_dinode(ip, page);
400 goto out_unlock_page;
402 ret = gfs2_allocate_page_backing(page);
408 gfs2_inplace_release(ip);
410 gfs2_quota_unlock(ip);
416 gfs2_holder_uninit(&gh);
420 static struct vm_operations_struct gfs2_vm_ops = {
421 .fault = filemap_fault,
422 .page_mkwrite = gfs2_page_mkwrite,
428 * @file: The file to map
429 * @vma: The VMA which described the mapping
431 * Returns: 0 or error code
434 static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
436 struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
437 struct gfs2_holder i_gh;
440 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, GL_ATIME, &i_gh);
441 error = gfs2_glock_nq_atime(&i_gh);
443 gfs2_holder_uninit(&i_gh);
447 vma->vm_ops = &gfs2_vm_ops;
449 gfs2_glock_dq_uninit(&i_gh);
455 * gfs2_open - open a file
456 * @inode: the inode to open
457 * @file: the struct file for this opening
462 static int gfs2_open(struct inode *inode, struct file *file)
464 struct gfs2_inode *ip = GFS2_I(inode);
465 struct gfs2_holder i_gh;
466 struct gfs2_file *fp;
469 fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
473 mutex_init(&fp->f_fl_mutex);
475 gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
476 file->private_data = fp;
478 if (S_ISREG(ip->i_inode.i_mode)) {
479 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
484 if (!(file->f_flags & O_LARGEFILE) &&
485 ip->i_di.di_size > MAX_NON_LFS) {
490 /* Listen to the Direct I/O flag */
492 if (ip->i_di.di_flags & GFS2_DIF_DIRECTIO)
493 file->f_flags |= O_DIRECT;
495 gfs2_glock_dq_uninit(&i_gh);
501 gfs2_glock_dq_uninit(&i_gh);
503 file->private_data = NULL;
509 * gfs2_close - called to close a struct file
510 * @inode: the inode the struct file belongs to
511 * @file: the struct file being closed
516 static int gfs2_close(struct inode *inode, struct file *file)
518 struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
519 struct gfs2_file *fp;
521 fp = file->private_data;
522 file->private_data = NULL;
524 if (gfs2_assert_warn(sdp, fp))
533 * gfs2_fsync - sync the dirty data for a file (across the cluster)
534 * @file: the file that points to the dentry (we ignore this)
535 * @dentry: the dentry that points to the inode to sync
537 * The VFS will flush "normal" data for us. We only need to worry
538 * about metadata here. For journaled data, we just do a log flush
539 * as we can't avoid it. Otherwise we can just bale out if datasync
540 * is set. For stuffed inodes we must flush the log in order to
541 * ensure that all data is on disk.
543 * The call to write_inode_now() is there to write back metadata and
544 * the inode itself. It does also try and write the data, but thats
545 * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
551 static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
553 struct inode *inode = dentry->d_inode;
554 int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
557 if (gfs2_is_jdata(GFS2_I(inode))) {
558 gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
562 if (sync_state != 0) {
564 ret = write_inode_now(inode, 0);
566 if (gfs2_is_stuffed(GFS2_I(inode)))
567 gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
574 * gfs2_setlease - acquire/release a file lease
575 * @file: the file pointer
582 static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl)
584 struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host);
587 * We don't currently have a way to enforce a lease across the whole
588 * cluster; until we do, disable leases (by just returning -EINVAL),
589 * unless the administrator has requested purely local locking.
591 if (!sdp->sd_args.ar_localflocks)
593 return generic_setlease(file, arg, fl);
596 static int gfs2_lm_plock_get(struct gfs2_sbd *sdp, struct lm_lockname *name,
597 struct file *file, struct file_lock *fl)
600 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
601 error = sdp->sd_lockstruct.ls_ops->lm_plock_get(
602 sdp->sd_lockstruct.ls_lockspace, name, file, fl);
606 static int gfs2_lm_plock(struct gfs2_sbd *sdp, struct lm_lockname *name,
607 struct file *file, int cmd, struct file_lock *fl)
610 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
611 error = sdp->sd_lockstruct.ls_ops->lm_plock(
612 sdp->sd_lockstruct.ls_lockspace, name, file, cmd, fl);
616 static int gfs2_lm_punlock(struct gfs2_sbd *sdp, struct lm_lockname *name,
617 struct file *file, struct file_lock *fl)
620 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
621 error = sdp->sd_lockstruct.ls_ops->lm_punlock(
622 sdp->sd_lockstruct.ls_lockspace, name, file, fl);
627 * gfs2_lock - acquire/release a posix lock on a file
628 * @file: the file pointer
629 * @cmd: either modify or retrieve lock state, possibly wait
630 * @fl: type and range of lock
635 static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
637 struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
638 struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host);
639 struct lm_lockname name =
640 { .ln_number = ip->i_no_addr,
641 .ln_type = LM_TYPE_PLOCK };
643 if (!(fl->fl_flags & FL_POSIX))
645 if (__mandatory_lock(&ip->i_inode))
648 if (cmd == F_CANCELLK) {
651 fl->fl_type = F_UNLCK;
654 return gfs2_lm_plock_get(sdp, &name, file, fl);
655 else if (fl->fl_type == F_UNLCK)
656 return gfs2_lm_punlock(sdp, &name, file, fl);
658 return gfs2_lm_plock(sdp, &name, file, cmd, fl);
661 static int do_flock(struct file *file, int cmd, struct file_lock *fl)
663 struct gfs2_file *fp = file->private_data;
664 struct gfs2_holder *fl_gh = &fp->f_fl_gh;
665 struct gfs2_inode *ip = GFS2_I(file->f_path.dentry->d_inode);
666 struct gfs2_glock *gl;
671 state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED;
672 flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE
675 mutex_lock(&fp->f_fl_mutex);
679 if (fl_gh->gh_state == state)
681 flock_lock_file_wait(file,
682 &(struct file_lock){.fl_type = F_UNLCK});
683 gfs2_glock_dq_wait(fl_gh);
684 gfs2_holder_reinit(state, flags, fl_gh);
686 error = gfs2_glock_get(GFS2_SB(&ip->i_inode),
687 ip->i_no_addr, &gfs2_flock_glops,
691 gfs2_holder_init(gl, state, flags, fl_gh);
694 error = gfs2_glock_nq(fl_gh);
696 gfs2_holder_uninit(fl_gh);
697 if (error == GLR_TRYFAILED)
700 error = flock_lock_file_wait(file, fl);
701 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
705 mutex_unlock(&fp->f_fl_mutex);
709 static void do_unflock(struct file *file, struct file_lock *fl)
711 struct gfs2_file *fp = file->private_data;
712 struct gfs2_holder *fl_gh = &fp->f_fl_gh;
714 mutex_lock(&fp->f_fl_mutex);
715 flock_lock_file_wait(file, fl);
717 gfs2_glock_dq_uninit(fl_gh);
718 mutex_unlock(&fp->f_fl_mutex);
722 * gfs2_flock - acquire/release a flock lock on a file
723 * @file: the file pointer
724 * @cmd: either modify or retrieve lock state, possibly wait
725 * @fl: type and range of lock
730 static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl)
732 struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
734 if (!(fl->fl_flags & FL_FLOCK))
736 if (__mandatory_lock(&ip->i_inode))
739 if (fl->fl_type == F_UNLCK) {
740 do_unflock(file, fl);
743 return do_flock(file, cmd, fl);
747 const struct file_operations gfs2_file_fops = {
748 .llseek = gfs2_llseek,
749 .read = do_sync_read,
750 .aio_read = generic_file_aio_read,
751 .write = do_sync_write,
752 .aio_write = generic_file_aio_write,
753 .unlocked_ioctl = gfs2_ioctl,
756 .release = gfs2_close,
760 .splice_read = generic_file_splice_read,
761 .splice_write = generic_file_splice_write,
762 .setlease = gfs2_setlease,
765 const struct file_operations gfs2_dir_fops = {
766 .readdir = gfs2_readdir,
767 .unlocked_ioctl = gfs2_ioctl,
769 .release = gfs2_close,
775 const struct file_operations gfs2_file_fops_nolock = {
776 .llseek = gfs2_llseek,
777 .read = do_sync_read,
778 .aio_read = generic_file_aio_read,
779 .write = do_sync_write,
780 .aio_write = generic_file_aio_write,
781 .unlocked_ioctl = gfs2_ioctl,
784 .release = gfs2_close,
786 .splice_read = generic_file_splice_read,
787 .splice_write = generic_file_splice_write,
788 .setlease = gfs2_setlease,
791 const struct file_operations gfs2_dir_fops_nolock = {
792 .readdir = gfs2_readdir,
793 .unlocked_ioctl = gfs2_ioctl,
795 .release = gfs2_close,