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/writeback.h>
24 #include <asm/uaccess.h>
25 #include <linux/dlm.h>
26 #include <linux/dlm_plock.h>
42 #include "ops_address.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_diskflags);
162 if (!S_ISDIR(inode->i_mode) && ip->i_diskflags & 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 unsigned int flags = inode->i_flags;
177 flags &= ~(S_SYNC|S_APPEND|S_IMMUTABLE|S_NOATIME|S_DIRSYNC);
178 if (ip->i_diskflags & GFS2_DIF_IMMUTABLE)
179 flags |= S_IMMUTABLE;
180 if (ip->i_diskflags & GFS2_DIF_APPENDONLY)
182 if (ip->i_diskflags & GFS2_DIF_NOATIME)
184 if (ip->i_diskflags & GFS2_DIF_SYNC)
186 inode->i_flags = flags;
189 /* Flags that can be set by user space */
190 #define GFS2_FLAGS_USER_SET (GFS2_DIF_JDATA| \
191 GFS2_DIF_IMMUTABLE| \
192 GFS2_DIF_APPENDONLY| \
196 GFS2_DIF_INHERIT_JDATA)
199 * gfs2_set_flags - set flags on an inode
201 * @flags: The flags to set
202 * @mask: Indicates which flags are valid
205 static int do_gfs2_set_flags(struct file *filp, u32 reqflags, u32 mask)
207 struct inode *inode = filp->f_path.dentry->d_inode;
208 struct gfs2_inode *ip = GFS2_I(inode);
209 struct gfs2_sbd *sdp = GFS2_SB(inode);
210 struct buffer_head *bh;
211 struct gfs2_holder gh;
213 u32 new_flags, flags;
215 error = mnt_want_write(filp->f_path.mnt);
219 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
223 flags = ip->i_diskflags;
224 new_flags = (flags & ~mask) | (reqflags & mask);
225 if ((new_flags ^ flags) == 0)
229 if ((new_flags ^ flags) & ~GFS2_FLAGS_USER_SET)
233 if (IS_IMMUTABLE(inode) && (new_flags & GFS2_DIF_IMMUTABLE))
235 if (IS_APPEND(inode) && (new_flags & GFS2_DIF_APPENDONLY))
237 if (((new_flags ^ flags) & GFS2_DIF_IMMUTABLE) &&
238 !capable(CAP_LINUX_IMMUTABLE))
240 if (!IS_IMMUTABLE(inode)) {
241 error = gfs2_permission(inode, MAY_WRITE);
245 if ((flags ^ new_flags) & GFS2_DIF_JDATA) {
246 if (flags & GFS2_DIF_JDATA)
247 gfs2_log_flush(sdp, ip->i_gl);
248 error = filemap_fdatawrite(inode->i_mapping);
251 error = filemap_fdatawait(inode->i_mapping);
255 error = gfs2_trans_begin(sdp, RES_DINODE, 0);
258 error = gfs2_meta_inode_buffer(ip, &bh);
261 gfs2_trans_add_bh(ip->i_gl, bh, 1);
262 ip->i_diskflags = new_flags;
263 gfs2_dinode_out(ip, bh->b_data);
265 gfs2_set_inode_flags(inode);
266 gfs2_set_aops(inode);
270 gfs2_glock_dq_uninit(&gh);
272 mnt_drop_write(filp->f_path.mnt);
276 static int gfs2_set_flags(struct file *filp, u32 __user *ptr)
278 struct inode *inode = filp->f_path.dentry->d_inode;
279 u32 fsflags, gfsflags;
280 if (get_user(fsflags, ptr))
282 gfsflags = fsflags_cvt(fsflags_to_gfs2, fsflags);
283 if (!S_ISDIR(inode->i_mode)) {
284 if (gfsflags & GFS2_DIF_INHERIT_JDATA)
285 gfsflags ^= (GFS2_DIF_JDATA | GFS2_DIF_INHERIT_JDATA);
286 return do_gfs2_set_flags(filp, gfsflags, ~0);
288 return do_gfs2_set_flags(filp, gfsflags, ~GFS2_DIF_JDATA);
291 static long gfs2_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
294 case FS_IOC_GETFLAGS:
295 return gfs2_get_flags(filp, (u32 __user *)arg);
296 case FS_IOC_SETFLAGS:
297 return gfs2_set_flags(filp, (u32 __user *)arg);
303 * gfs2_allocate_page_backing - Use bmap to allocate blocks
304 * @page: The (locked) page to allocate backing for
306 * We try to allocate all the blocks required for the page in
307 * one go. This might fail for various reasons, so we keep
308 * trying until all the blocks to back this page are allocated.
309 * If some of the blocks are already allocated, thats ok too.
312 static int gfs2_allocate_page_backing(struct page *page)
314 struct inode *inode = page->mapping->host;
315 struct buffer_head bh;
316 unsigned long size = PAGE_CACHE_SIZE;
317 u64 lblock = page->index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
322 gfs2_block_map(inode, lblock, &bh, 1);
323 if (!buffer_mapped(&bh))
326 lblock += (bh.b_size >> inode->i_blkbits);
332 * gfs2_page_mkwrite - Make a shared, mmap()ed, page writable
333 * @vma: The virtual memory area
334 * @page: The page which is about to become writable
336 * When the page becomes writable, we need to ensure that we have
337 * blocks allocated on disk to back that page.
340 static int gfs2_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf)
342 struct page *page = vmf->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_SHIFT;
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(GLF_DIRTY, &ip->i_gl->gl_flags);
360 set_bit(GIF_SW_PAGED, &ip->i_flags);
362 ret = gfs2_write_alloc_required(ip, pos, PAGE_CACHE_SIZE, &alloc_required);
363 if (ret || !alloc_required)
366 al = gfs2_alloc_get(ip);
370 ret = gfs2_quota_lock_check(ip);
373 gfs2_write_calc_reserv(ip, PAGE_CACHE_SIZE, &data_blocks, &ind_blocks);
374 al->al_requested = data_blocks + ind_blocks;
375 ret = gfs2_inplace_reserve(ip);
377 goto out_quota_unlock;
379 rblocks = RES_DINODE + ind_blocks;
380 if (gfs2_is_jdata(ip))
381 rblocks += data_blocks ? data_blocks : 1;
382 if (ind_blocks || data_blocks)
383 rblocks += RES_STATFS + RES_QUOTA;
384 ret = gfs2_trans_begin(sdp, rblocks, 0);
390 last_index = ip->i_inode.i_size >> PAGE_CACHE_SHIFT;
391 if (page->index > last_index)
392 goto out_unlock_page;
394 if (!PageUptodate(page) || page->mapping != ip->i_inode.i_mapping)
395 goto out_unlock_page;
396 if (gfs2_is_stuffed(ip)) {
397 ret = gfs2_unstuff_dinode(ip, page);
399 goto out_unlock_page;
401 ret = gfs2_allocate_page_backing(page);
407 gfs2_inplace_release(ip);
409 gfs2_quota_unlock(ip);
415 gfs2_holder_uninit(&gh);
419 ret = VM_FAULT_SIGBUS;
423 static struct vm_operations_struct gfs2_vm_ops = {
424 .fault = filemap_fault,
425 .page_mkwrite = gfs2_page_mkwrite,
431 * @file: The file to map
432 * @vma: The VMA which described the mapping
434 * Returns: 0 or error code
437 static int gfs2_mmap(struct file *file, struct vm_area_struct *vma)
439 struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
440 struct gfs2_holder i_gh;
443 gfs2_holder_init(ip->i_gl, LM_ST_SHARED, 0, &i_gh);
444 error = gfs2_glock_nq(&i_gh);
446 gfs2_holder_uninit(&i_gh);
450 vma->vm_ops = &gfs2_vm_ops;
452 gfs2_glock_dq_uninit(&i_gh);
458 * gfs2_open - open a file
459 * @inode: the inode to open
460 * @file: the struct file for this opening
465 static int gfs2_open(struct inode *inode, struct file *file)
467 struct gfs2_inode *ip = GFS2_I(inode);
468 struct gfs2_holder i_gh;
469 struct gfs2_file *fp;
472 fp = kzalloc(sizeof(struct gfs2_file), GFP_KERNEL);
476 mutex_init(&fp->f_fl_mutex);
478 gfs2_assert_warn(GFS2_SB(inode), !file->private_data);
479 file->private_data = fp;
481 if (S_ISREG(ip->i_inode.i_mode)) {
482 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, LM_FLAG_ANY,
487 if (!(file->f_flags & O_LARGEFILE) &&
488 ip->i_disksize > MAX_NON_LFS) {
493 gfs2_glock_dq_uninit(&i_gh);
499 gfs2_glock_dq_uninit(&i_gh);
501 file->private_data = NULL;
507 * gfs2_close - called to close a struct file
508 * @inode: the inode the struct file belongs to
509 * @file: the struct file being closed
514 static int gfs2_close(struct inode *inode, struct file *file)
516 struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
517 struct gfs2_file *fp;
519 fp = file->private_data;
520 file->private_data = NULL;
522 if (gfs2_assert_warn(sdp, fp))
531 * gfs2_fsync - sync the dirty data for a file (across the cluster)
532 * @file: the file that points to the dentry (we ignore this)
533 * @dentry: the dentry that points to the inode to sync
535 * The VFS will flush "normal" data for us. We only need to worry
536 * about metadata here. For journaled data, we just do a log flush
537 * as we can't avoid it. Otherwise we can just bale out if datasync
538 * is set. For stuffed inodes we must flush the log in order to
539 * ensure that all data is on disk.
541 * The call to write_inode_now() is there to write back metadata and
542 * the inode itself. It does also try and write the data, but thats
543 * (hopefully) a no-op due to the VFS having already called filemap_fdatawrite()
549 static int gfs2_fsync(struct file *file, struct dentry *dentry, int datasync)
551 struct inode *inode = dentry->d_inode;
552 int sync_state = inode->i_state & (I_DIRTY_SYNC|I_DIRTY_DATASYNC);
555 if (gfs2_is_jdata(GFS2_I(inode))) {
556 gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
560 if (sync_state != 0) {
562 ret = write_inode_now(inode, 0);
564 if (gfs2_is_stuffed(GFS2_I(inode)))
565 gfs2_log_flush(GFS2_SB(inode), GFS2_I(inode)->i_gl);
571 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
574 * gfs2_setlease - acquire/release a file lease
575 * @file: the file pointer
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.
586 static int gfs2_setlease(struct file *file, long arg, struct file_lock **fl)
592 * gfs2_lock - acquire/release a posix lock on a file
593 * @file: the file pointer
594 * @cmd: either modify or retrieve lock state, possibly wait
595 * @fl: type and range of lock
600 static int gfs2_lock(struct file *file, int cmd, struct file_lock *fl)
602 struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
603 struct gfs2_sbd *sdp = GFS2_SB(file->f_mapping->host);
604 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
606 if (!(fl->fl_flags & FL_POSIX))
608 if (__mandatory_lock(&ip->i_inode))
611 if (cmd == F_CANCELLK) {
614 fl->fl_type = F_UNLCK;
616 if (unlikely(test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
619 return dlm_posix_get(ls->ls_dlm, ip->i_no_addr, file, fl);
620 else if (fl->fl_type == F_UNLCK)
621 return dlm_posix_unlock(ls->ls_dlm, ip->i_no_addr, file, fl);
623 return dlm_posix_lock(ls->ls_dlm, ip->i_no_addr, file, cmd, fl);
626 static int do_flock(struct file *file, int cmd, struct file_lock *fl)
628 struct gfs2_file *fp = file->private_data;
629 struct gfs2_holder *fl_gh = &fp->f_fl_gh;
630 struct gfs2_inode *ip = GFS2_I(file->f_path.dentry->d_inode);
631 struct gfs2_glock *gl;
636 state = (fl->fl_type == F_WRLCK) ? LM_ST_EXCLUSIVE : LM_ST_SHARED;
637 flags = (IS_SETLKW(cmd) ? 0 : LM_FLAG_TRY) | GL_EXACT | GL_NOCACHE;
639 mutex_lock(&fp->f_fl_mutex);
643 if (fl_gh->gh_state == state)
645 flock_lock_file_wait(file,
646 &(struct file_lock){.fl_type = F_UNLCK});
647 gfs2_glock_dq_wait(fl_gh);
648 gfs2_holder_reinit(state, flags, fl_gh);
650 error = gfs2_glock_get(GFS2_SB(&ip->i_inode), ip->i_no_addr,
651 &gfs2_flock_glops, CREATE, &gl);
654 gfs2_holder_init(gl, state, flags, fl_gh);
657 error = gfs2_glock_nq(fl_gh);
659 gfs2_holder_uninit(fl_gh);
660 if (error == GLR_TRYFAILED)
663 error = flock_lock_file_wait(file, fl);
664 gfs2_assert_warn(GFS2_SB(&ip->i_inode), !error);
668 mutex_unlock(&fp->f_fl_mutex);
672 static void do_unflock(struct file *file, struct file_lock *fl)
674 struct gfs2_file *fp = file->private_data;
675 struct gfs2_holder *fl_gh = &fp->f_fl_gh;
677 mutex_lock(&fp->f_fl_mutex);
678 flock_lock_file_wait(file, fl);
680 gfs2_glock_dq_uninit(fl_gh);
681 mutex_unlock(&fp->f_fl_mutex);
685 * gfs2_flock - acquire/release a flock lock on a file
686 * @file: the file pointer
687 * @cmd: either modify or retrieve lock state, possibly wait
688 * @fl: type and range of lock
693 static int gfs2_flock(struct file *file, int cmd, struct file_lock *fl)
695 struct gfs2_inode *ip = GFS2_I(file->f_mapping->host);
697 if (!(fl->fl_flags & FL_FLOCK))
699 if (__mandatory_lock(&ip->i_inode))
702 if (fl->fl_type == F_UNLCK) {
703 do_unflock(file, fl);
706 return do_flock(file, cmd, fl);
710 const struct file_operations gfs2_file_fops = {
711 .llseek = gfs2_llseek,
712 .read = do_sync_read,
713 .aio_read = generic_file_aio_read,
714 .write = do_sync_write,
715 .aio_write = generic_file_aio_write,
716 .unlocked_ioctl = gfs2_ioctl,
719 .release = gfs2_close,
723 .splice_read = generic_file_splice_read,
724 .splice_write = generic_file_splice_write,
725 .setlease = gfs2_setlease,
728 const struct file_operations gfs2_dir_fops = {
729 .readdir = gfs2_readdir,
730 .unlocked_ioctl = gfs2_ioctl,
732 .release = gfs2_close,
738 #endif /* CONFIG_GFS2_FS_LOCKING_DLM */
740 const struct file_operations gfs2_file_fops_nolock = {
741 .llseek = gfs2_llseek,
742 .read = do_sync_read,
743 .aio_read = generic_file_aio_read,
744 .write = do_sync_write,
745 .aio_write = generic_file_aio_write,
746 .unlocked_ioctl = gfs2_ioctl,
749 .release = gfs2_close,
751 .splice_read = generic_file_splice_read,
752 .splice_write = generic_file_splice_write,
753 .setlease = generic_setlease,
756 const struct file_operations gfs2_dir_fops_nolock = {
757 .readdir = gfs2_readdir,
758 .unlocked_ioctl = gfs2_ioctl,
760 .release = gfs2_close,