2 * Copyright (c) 2000-2005 Silicon Graphics, Inc. All Rights Reserved.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms of version 2 of the GNU General Public License as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it would be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
12 * Further, this software is distributed without any warranty that it is
13 * free of the rightful claim of any third person regarding infringement
14 * or the like. Any license provided herein, whether implied or
15 * otherwise, applies only to this software file. Patent licenses, if
16 * any, provided herein do not apply to combinations of this program with
17 * other software, or any other product whatsoever.
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write the Free Software Foundation, Inc., 59
21 * Temple Place - Suite 330, Boston MA 02111-1307, USA.
23 * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
24 * Mountain View, CA 94043, or:
28 * For further information regarding this notice, see:
30 * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
38 #include "xfs_trans.h"
42 #include "xfs_alloc.h"
43 #include "xfs_dmapi.h"
44 #include "xfs_quota.h"
45 #include "xfs_mount.h"
46 #include "xfs_alloc_btree.h"
47 #include "xfs_bmap_btree.h"
48 #include "xfs_ialloc_btree.h"
49 #include "xfs_btree.h"
50 #include "xfs_ialloc.h"
51 #include "xfs_attr_sf.h"
52 #include "xfs_dir_sf.h"
53 #include "xfs_dir2_sf.h"
54 #include "xfs_dinode.h"
55 #include "xfs_inode.h"
58 #include "xfs_rtalloc.h"
59 #include "xfs_error.h"
60 #include "xfs_itable.h"
66 #include "xfs_buf_item.h"
67 #include "xfs_utils.h"
68 #include "xfs_version.h"
70 #include <linux/namei.h>
71 #include <linux/init.h>
72 #include <linux/mount.h>
73 #include <linux/writeback.h>
75 STATIC struct quotactl_ops linvfs_qops;
76 STATIC struct super_operations linvfs_sops;
77 STATIC kmem_zone_t *linvfs_inode_zone;
79 STATIC struct xfs_mount_args *
81 struct super_block *sb)
83 struct xfs_mount_args *args;
85 args = kmem_zalloc(sizeof(struct xfs_mount_args), KM_SLEEP);
86 args->logbufs = args->logbufsize = -1;
87 strncpy(args->fsname, sb->s_id, MAXNAMELEN);
89 /* Copy the already-parsed mount(2) flags we're interested in */
90 if (sb->s_flags & MS_NOATIME)
91 args->flags |= XFSMNT_NOATIME;
92 if (sb->s_flags & MS_DIRSYNC)
93 args->flags |= XFSMNT_DIRSYNC;
94 if (sb->s_flags & MS_SYNCHRONOUS)
95 args->flags |= XFSMNT_WSYNC;
97 /* Default to 32 bit inodes on Linux all the time */
98 args->flags |= XFSMNT_32BITINODES;
105 unsigned int blockshift)
107 unsigned int pagefactor = 1;
108 unsigned int bitshift = BITS_PER_LONG - 1;
110 /* Figure out maximum filesize, on Linux this can depend on
111 * the filesystem blocksize (on 32 bit platforms).
112 * __block_prepare_write does this in an [unsigned] long...
113 * page->index << (PAGE_CACHE_SHIFT - bbits)
114 * So, for page sized blocks (4K on 32 bit platforms),
115 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
116 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
117 * but for smaller blocksizes it is less (bbits = log2 bsize).
118 * Note1: get_block_t takes a long (implicit cast from above)
119 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
120 * can optionally convert the [unsigned] long from above into
121 * an [unsigned] long long.
124 #if BITS_PER_LONG == 32
125 # if defined(CONFIG_LBD)
126 ASSERT(sizeof(sector_t) == 8);
127 pagefactor = PAGE_CACHE_SIZE;
128 bitshift = BITS_PER_LONG;
130 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
134 return (((__uint64_t)pagefactor) << bitshift) - 1;
137 STATIC __inline__ void
141 switch (inode->i_mode & S_IFMT) {
143 inode->i_op = &linvfs_file_inode_operations;
144 inode->i_fop = &linvfs_file_operations;
145 inode->i_mapping->a_ops = &linvfs_aops;
148 inode->i_op = &linvfs_dir_inode_operations;
149 inode->i_fop = &linvfs_dir_operations;
152 inode->i_op = &linvfs_symlink_inode_operations;
154 inode->i_mapping->a_ops = &linvfs_aops;
157 inode->i_op = &linvfs_file_inode_operations;
158 init_special_inode(inode, inode->i_mode, inode->i_rdev);
163 STATIC __inline__ void
164 xfs_revalidate_inode(
169 struct inode *inode = LINVFS_GET_IP(vp);
171 inode->i_mode = ip->i_d.di_mode;
172 inode->i_nlink = ip->i_d.di_nlink;
173 inode->i_uid = ip->i_d.di_uid;
174 inode->i_gid = ip->i_d.di_gid;
176 switch (inode->i_mode & S_IFMT) {
180 MKDEV(sysv_major(ip->i_df.if_u2.if_rdev) & 0x1ff,
181 sysv_minor(ip->i_df.if_u2.if_rdev));
188 inode->i_blksize = PAGE_CACHE_SIZE;
189 inode->i_generation = ip->i_d.di_gen;
190 i_size_write(inode, ip->i_d.di_size);
192 XFS_FSB_TO_BB(mp, ip->i_d.di_nblocks + ip->i_delayed_blks);
193 inode->i_atime.tv_sec = ip->i_d.di_atime.t_sec;
194 inode->i_atime.tv_nsec = ip->i_d.di_atime.t_nsec;
195 inode->i_mtime.tv_sec = ip->i_d.di_mtime.t_sec;
196 inode->i_mtime.tv_nsec = ip->i_d.di_mtime.t_nsec;
197 inode->i_ctime.tv_sec = ip->i_d.di_ctime.t_sec;
198 inode->i_ctime.tv_nsec = ip->i_d.di_ctime.t_nsec;
199 if (ip->i_d.di_flags & XFS_DIFLAG_IMMUTABLE)
200 inode->i_flags |= S_IMMUTABLE;
202 inode->i_flags &= ~S_IMMUTABLE;
203 if (ip->i_d.di_flags & XFS_DIFLAG_APPEND)
204 inode->i_flags |= S_APPEND;
206 inode->i_flags &= ~S_APPEND;
207 if (ip->i_d.di_flags & XFS_DIFLAG_SYNC)
208 inode->i_flags |= S_SYNC;
210 inode->i_flags &= ~S_SYNC;
211 if (ip->i_d.di_flags & XFS_DIFLAG_NOATIME)
212 inode->i_flags |= S_NOATIME;
214 inode->i_flags &= ~S_NOATIME;
215 vp->v_flag &= ~VMODIFIED;
219 xfs_initialize_vnode(
222 bhv_desc_t *inode_bhv,
225 xfs_inode_t *ip = XFS_BHVTOI(inode_bhv);
226 struct inode *inode = LINVFS_GET_IP(vp);
228 if (!inode_bhv->bd_vobj) {
229 vp->v_vfsp = bhvtovfs(bdp);
230 bhv_desc_init(inode_bhv, ip, vp, &xfs_vnodeops);
231 bhv_insert(VN_BHV_HEAD(vp), inode_bhv);
235 * We need to set the ops vectors, and unlock the inode, but if
236 * we have been called during the new inode create process, it is
237 * too early to fill in the Linux inode. We will get called a
238 * second time once the inode is properly set up, and then we can
241 if (ip->i_d.di_mode != 0 && unlock && (inode->i_state & I_NEW)) {
242 xfs_revalidate_inode(XFS_BHVTOM(bdp), vp, ip);
243 xfs_set_inodeops(inode);
245 ip->i_flags &= ~XFS_INEW;
248 unlock_new_inode(inode);
256 struct block_device **bdevp)
260 *bdevp = open_bdev_excl(name, 0, mp);
261 if (IS_ERR(*bdevp)) {
262 error = PTR_ERR(*bdevp);
263 printk("XFS: Invalid device [%s], error=%d\n", name, error);
271 struct block_device *bdev)
274 close_bdev_excl(bdev);
278 STATIC struct inode *
280 struct super_block *sb)
284 vp = (vnode_t *)kmem_cache_alloc(linvfs_inode_zone,
285 kmem_flags_convert(KM_SLEEP));
288 return LINVFS_GET_IP(vp);
292 linvfs_destroy_inode(
295 kmem_cache_free(linvfs_inode_zone, LINVFS_GET_VP(inode));
301 kmem_cache_t *cachep,
304 vnode_t *vp = (vnode_t *)data;
306 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
307 SLAB_CTOR_CONSTRUCTOR)
308 inode_init_once(LINVFS_GET_IP(vp));
312 init_inodecache( void )
314 linvfs_inode_zone = kmem_cache_create("linvfs_icache",
315 sizeof(vnode_t), 0, SLAB_RECLAIM_ACCOUNT,
317 if (linvfs_inode_zone == NULL)
323 destroy_inodecache( void )
325 if (kmem_cache_destroy(linvfs_inode_zone))
326 printk(KERN_WARNING "%s: cache still in use!\n", __FUNCTION__);
330 * Attempt to flush the inode, this will actually fail
331 * if the inode is pinned, but we dirty the inode again
332 * at the point when it is unpinned after a log write,
333 * since this is when the inode itself becomes flushable.
340 vnode_t *vp = LINVFS_GET_VP(inode);
341 int error = 0, flags = FLUSH_INODE;
344 vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address);
347 VOP_IFLUSH(vp, flags, error);
348 if (error == EAGAIN) {
350 VOP_IFLUSH(vp, flags | FLUSH_LOG, error);
363 vnode_t *vp = LINVFS_GET_VP(inode);
367 vn_trace_entry(vp, __FUNCTION__, (inst_t *)__return_address);
369 * Do all our cleanup, and remove this vnode.
377 * Enqueue a work item to be picked up by the vfs xfssyncd thread.
378 * Doing this has two advantages:
379 * - It saves on stack space, which is tight in certain situations
380 * - It can be used (with care) as a mechanism to avoid deadlocks.
381 * Flushing while allocating in a full filesystem requires both.
384 xfs_syncd_queue_work(
387 void (*syncer)(vfs_t *, void *))
389 vfs_sync_work_t *work;
391 work = kmem_alloc(sizeof(struct vfs_sync_work), KM_SLEEP);
392 INIT_LIST_HEAD(&work->w_list);
393 work->w_syncer = syncer;
396 spin_lock(&vfs->vfs_sync_lock);
397 list_add_tail(&work->w_list, &vfs->vfs_sync_list);
398 spin_unlock(&vfs->vfs_sync_lock);
399 wake_up_process(vfs->vfs_sync_task);
403 * Flush delayed allocate data, attempting to free up reserved space
404 * from existing allocations. At this point a new allocation attempt
405 * has failed with ENOSPC and we are in the process of scratching our
406 * heads, looking about for more room...
409 xfs_flush_inode_work(
413 filemap_flush(((struct inode *)inode)->i_mapping);
414 iput((struct inode *)inode);
421 struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip));
422 struct vfs *vfs = XFS_MTOVFS(ip->i_mount);
425 xfs_syncd_queue_work(vfs, inode, xfs_flush_inode_work);
430 * This is the "bigger hammer" version of xfs_flush_inode_work...
431 * (IOW, "If at first you don't succeed, use a Bigger Hammer").
434 xfs_flush_device_work(
438 sync_blockdev(vfs->vfs_super->s_bdev);
439 iput((struct inode *)inode);
446 struct inode *inode = LINVFS_GET_IP(XFS_ITOV(ip));
447 struct vfs *vfs = XFS_MTOVFS(ip->i_mount);
450 xfs_syncd_queue_work(vfs, inode, xfs_flush_device_work);
452 xfs_log_force(ip->i_mount, (xfs_lsn_t)0, XFS_LOG_FORCE|XFS_LOG_SYNC);
455 #define SYNCD_FLAGS (SYNC_FSDATA|SYNC_BDFLUSH|SYNC_ATTR)
463 if (!(vfsp->vfs_flag & VFS_RDONLY))
464 VFS_SYNC(vfsp, SYNCD_FLAGS, NULL, error);
465 vfsp->vfs_sync_seq++;
467 wake_up(&vfsp->vfs_wait_single_sync_task);
475 vfs_t *vfsp = (vfs_t *) arg;
476 struct list_head tmp;
477 struct vfs_sync_work *work, *n;
479 daemonize("xfssyncd");
481 vfsp->vfs_sync_work.w_vfs = vfsp;
482 vfsp->vfs_sync_work.w_syncer = vfs_sync_worker;
483 vfsp->vfs_sync_task = current;
485 wake_up(&vfsp->vfs_wait_sync_task);
487 INIT_LIST_HEAD(&tmp);
488 timeleft = (xfs_syncd_centisecs * HZ) / 100;
490 set_current_state(TASK_INTERRUPTIBLE);
491 timeleft = schedule_timeout(timeleft);
494 if (vfsp->vfs_flag & VFS_UMOUNT)
497 spin_lock(&vfsp->vfs_sync_lock);
499 * We can get woken by laptop mode, to do a sync -
500 * that's the (only!) case where the list would be
501 * empty with time remaining.
503 if (!timeleft || list_empty(&vfsp->vfs_sync_list)) {
505 timeleft = (xfs_syncd_centisecs * HZ) / 100;
506 INIT_LIST_HEAD(&vfsp->vfs_sync_work.w_list);
507 list_add_tail(&vfsp->vfs_sync_work.w_list,
508 &vfsp->vfs_sync_list);
510 list_for_each_entry_safe(work, n, &vfsp->vfs_sync_list, w_list)
511 list_move(&work->w_list, &tmp);
512 spin_unlock(&vfsp->vfs_sync_lock);
514 list_for_each_entry_safe(work, n, &tmp, w_list) {
515 (*work->w_syncer)(vfsp, work->w_data);
516 list_del(&work->w_list);
517 if (work == &vfsp->vfs_sync_work)
519 kmem_free(work, sizeof(struct vfs_sync_work));
523 vfsp->vfs_sync_task = NULL;
525 wake_up(&vfsp->vfs_wait_sync_task);
536 pid = kernel_thread(xfssyncd, (void *) vfsp,
537 CLONE_VM | CLONE_FS | CLONE_FILES);
540 wait_event(vfsp->vfs_wait_sync_task, vfsp->vfs_sync_task);
548 vfsp->vfs_flag |= VFS_UMOUNT;
551 wake_up_process(vfsp->vfs_sync_task);
552 wait_event(vfsp->vfs_wait_sync_task, !vfsp->vfs_sync_task);
557 struct super_block *sb)
559 vfs_t *vfsp = LINVFS_GET_VFS(sb);
562 linvfs_stop_syncd(vfsp);
563 VFS_SYNC(vfsp, SYNC_ATTR|SYNC_DELWRI, NULL, error);
565 VFS_UNMOUNT(vfsp, 0, NULL, error);
567 printk("XFS unmount got error %d\n", error);
568 printk("%s: vfsp/0x%p left dangling!\n", __FUNCTION__, vfsp);
572 vfs_deallocate(vfsp);
577 struct super_block *sb)
579 vfs_t *vfsp = LINVFS_GET_VFS(sb);
582 if (sb->s_flags & MS_RDONLY) {
583 sb->s_dirt = 0; /* paranoia */
586 /* Push the log and superblock a little */
587 VFS_SYNC(vfsp, SYNC_FSDATA, NULL, error);
593 struct super_block *sb,
596 vfs_t *vfsp = LINVFS_GET_VFS(sb);
598 int flags = SYNC_FSDATA;
600 if (unlikely(sb->s_frozen == SB_FREEZE_WRITE))
601 flags = SYNC_QUIESCE;
603 flags = SYNC_FSDATA | (wait ? SYNC_WAIT : 0);
605 VFS_SYNC(vfsp, flags, NULL, error);
608 if (unlikely(laptop_mode)) {
609 int prev_sync_seq = vfsp->vfs_sync_seq;
612 * The disk must be active because we're syncing.
613 * We schedule xfssyncd now (now that the disk is
614 * active) instead of later (when it might not be).
616 wake_up_process(vfsp->vfs_sync_task);
618 * We have to wait for the sync iteration to complete.
619 * If we don't, the disk activity caused by the sync
620 * will come after the sync is completed, and that
621 * triggers another sync from laptop mode.
623 wait_event(vfsp->vfs_wait_single_sync_task,
624 vfsp->vfs_sync_seq != prev_sync_seq);
632 struct super_block *sb,
633 struct kstatfs *statp)
635 vfs_t *vfsp = LINVFS_GET_VFS(sb);
638 VFS_STATVFS(vfsp, statp, NULL, error);
644 struct super_block *sb,
648 vfs_t *vfsp = LINVFS_GET_VFS(sb);
649 struct xfs_mount_args *args = xfs_args_allocate(sb);
652 VFS_PARSEARGS(vfsp, options, args, 1, error);
654 VFS_MNTUPDATE(vfsp, flags, args, error);
655 kmem_free(args, sizeof(*args));
661 struct super_block *sb)
663 VFS_FREEZE(LINVFS_GET_VFS(sb));
669 struct vfsmount *mnt)
671 struct vfs *vfsp = LINVFS_GET_VFS(mnt->mnt_sb);
674 VFS_SHOWARGS(vfsp, m, error);
680 struct super_block *sb,
681 struct fs_quota_stat *fqs)
683 struct vfs *vfsp = LINVFS_GET_VFS(sb);
686 VFS_QUOTACTL(vfsp, Q_XGETQSTAT, 0, (caddr_t)fqs, error);
692 struct super_block *sb,
696 struct vfs *vfsp = LINVFS_GET_VFS(sb);
699 VFS_QUOTACTL(vfsp, op, 0, (caddr_t)&flags, error);
705 struct super_block *sb,
708 struct fs_disk_quota *fdq)
710 struct vfs *vfsp = LINVFS_GET_VFS(sb);
713 getmode = (type == USRQUOTA) ? Q_XGETQUOTA :
714 ((type == GRPQUOTA) ? Q_XGETGQUOTA : Q_XGETPQUOTA);
715 VFS_QUOTACTL(vfsp, getmode, id, (caddr_t)fdq, error);
721 struct super_block *sb,
724 struct fs_disk_quota *fdq)
726 struct vfs *vfsp = LINVFS_GET_VFS(sb);
729 setmode = (type == USRQUOTA) ? Q_XSETQLIM :
730 ((type == GRPQUOTA) ? Q_XSETGQLIM : Q_XSETPQLIM);
731 VFS_QUOTACTL(vfsp, setmode, id, (caddr_t)fdq, error);
737 struct super_block *sb,
742 struct vfs *vfsp = vfs_allocate();
743 struct xfs_mount_args *args = xfs_args_allocate(sb);
744 struct kstatfs statvfs;
747 vfsp->vfs_super = sb;
748 LINVFS_SET_VFS(sb, vfsp);
749 if (sb->s_flags & MS_RDONLY)
750 vfsp->vfs_flag |= VFS_RDONLY;
751 bhv_insert_all_vfsops(vfsp);
753 VFS_PARSEARGS(vfsp, (char *)data, args, 0, error);
755 bhv_remove_all_vfsops(vfsp, 1);
759 sb_min_blocksize(sb, BBSIZE);
760 #ifdef CONFIG_XFS_EXPORT
761 sb->s_export_op = &linvfs_export_ops;
763 sb->s_qcop = &linvfs_qops;
764 sb->s_op = &linvfs_sops;
766 VFS_MOUNT(vfsp, args, NULL, error);
768 bhv_remove_all_vfsops(vfsp, 1);
772 VFS_STATVFS(vfsp, &statvfs, NULL, error);
777 sb->s_magic = statvfs.f_type;
778 sb->s_blocksize = statvfs.f_bsize;
779 sb->s_blocksize_bits = ffs(statvfs.f_bsize) - 1;
780 sb->s_maxbytes = xfs_max_file_offset(sb->s_blocksize_bits);
782 set_posix_acl_flag(sb);
784 VFS_ROOT(vfsp, &rootvp, error);
788 sb->s_root = d_alloc_root(LINVFS_GET_IP(rootvp));
793 if (is_bad_inode(sb->s_root->d_inode)) {
797 if ((error = linvfs_start_syncd(vfsp)))
799 vn_trace_exit(rootvp, __FUNCTION__, (inst_t *)__return_address);
801 kmem_free(args, sizeof(*args));
813 VFS_UNMOUNT(vfsp, 0, NULL, error2);
816 vfs_deallocate(vfsp);
817 kmem_free(args, sizeof(*args));
821 STATIC struct super_block *
823 struct file_system_type *fs_type,
825 const char *dev_name,
828 return get_sb_bdev(fs_type, flags, dev_name, data, linvfs_fill_super);
831 STATIC struct super_operations linvfs_sops = {
832 .alloc_inode = linvfs_alloc_inode,
833 .destroy_inode = linvfs_destroy_inode,
834 .write_inode = linvfs_write_inode,
835 .clear_inode = linvfs_clear_inode,
836 .put_super = linvfs_put_super,
837 .write_super = linvfs_write_super,
838 .sync_fs = linvfs_sync_super,
839 .write_super_lockfs = linvfs_freeze_fs,
840 .statfs = linvfs_statfs,
841 .remount_fs = linvfs_remount,
842 .show_options = linvfs_show_options,
845 STATIC struct quotactl_ops linvfs_qops = {
846 .get_xstate = linvfs_getxstate,
847 .set_xstate = linvfs_setxstate,
848 .get_xquota = linvfs_getxquota,
849 .set_xquota = linvfs_setxquota,
852 STATIC struct file_system_type xfs_fs_type = {
853 .owner = THIS_MODULE,
855 .get_sb = linvfs_get_sb,
856 .kill_sb = kill_block_super,
857 .fs_flags = FS_REQUIRES_DEV,
866 static char message[] __initdata = KERN_INFO \
867 XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled\n";
872 xfs_physmem = si.totalram;
876 error = init_inodecache();
878 goto undo_inodecache;
880 error = pagebuf_init();
889 error = register_filesystem(&xfs_fs_type);
892 XFS_DM_INIT(&xfs_fs_type);
899 destroy_inodecache();
909 XFS_DM_EXIT(&xfs_fs_type);
910 unregister_filesystem(&xfs_fs_type);
913 destroy_inodecache();
917 module_init(init_xfs_fs);
918 module_exit(exit_xfs_fs);
920 MODULE_AUTHOR("Silicon Graphics, Inc.");
921 MODULE_DESCRIPTION(XFS_VERSION_STRING " with " XFS_BUILD_OPTIONS " enabled");
922 MODULE_LICENSE("GPL");