[PATCH] process orphan list if device transitions to readonly
I have a patch in -mm now to skip orphan inode processing on a read-only device (where IO may fail if issued), but Stephen points out that if the device ever transitions back to readwrite, and the filesystem is remounted as rewrite, we should process the orphan inode list at that point. Today, I'm not sure how we can easily get into this situation - for example, lvm apparently cannot transform a RO snapshot into RW. But it is at least possible to do this. For example I tested by: create orphan list crash mark block device RO via BLKROSET mount fs RO mark block device RW via BLKROSET remount fs RW so it's within the realm of possibility, certainly. here's what I came up with; I'm not very pleased with it though, but I need some way to let ext3_orphan_del know that the sb is already locked and it's safe to do this orphan processing w/o re-taking it, so I added a state flag... It's worked in my testing. Comments or commit would be appreciated. Thanks, -Eric Signed-off-by: Eric Sandeen [EMAIL PROTECTED] Index: linux-2.6.19/fs/ext3/namei.c === --- linux-2.6.19.orig/fs/ext3/namei.c +++ linux-2.6.19/fs/ext3/namei.c @@ -1942,15 +1942,15 @@ int ext3_orphan_del(handle_t *handle, st struct ext3_iloc iloc; int err = 0; - lock_super(inode-i_sb); - if (list_empty(ei-i_orphan)) { - unlock_super(inode-i_sb); - return 0; - } + sbi = EXT3_SB(inode-i_sb); + if (!(sbi-s_mount_state EXT3_REMOUNTING_FS)) + lock_super(inode-i_sb); + + if (list_empty(ei-i_orphan)) + goto out; ino_next = NEXT_ORPHAN(inode); prev = ei-i_orphan.prev; - sbi = EXT3_SB(inode-i_sb); jbd_debug(4, remove inode %lu from orphan list\n, inode-i_ino); @@ -1996,7 +1996,8 @@ int ext3_orphan_del(handle_t *handle, st out_err: ext3_std_error(inode-i_sb, err); out: - unlock_super(inode-i_sb); + if (!(sbi-s_mount_state EXT3_REMOUNTING_FS)) + unlock_super(inode-i_sb); return err; out_brelse: Index: linux-2.6.19/fs/ext3/super.c === --- linux-2.6.19.orig/fs/ext3/super.c +++ linux-2.6.19/fs/ext3/super.c @@ -2358,6 +2358,10 @@ static int ext3_remount (struct super_bl goto restore_opts; if (!ext3_setup_super (sb, es, 0)) sb-s_flags = ~MS_RDONLY; + EXT3_SB(sb)-s_mount_state |= (EXT3_ORPHAN_FS|EXT3_REMOUNTING_FS); + ext3_orphan_cleanup(sb, es); + EXT3_SB(sb)-s_mount_state = ~(EXT3_ORPHAN_FS|EXT3_REMOUNTING_FS); + } } #ifdef CONFIG_QUOTA Index: linux-2.6.19/include/linux/ext3_fs.h === --- linux-2.6.19.orig/include/linux/ext3_fs.h +++ linux-2.6.19/include/linux/ext3_fs.h @@ -356,6 +356,7 @@ struct ext3_inode { #defineEXT3_VALID_FS 0x0001 /* Unmounted cleanly */ #defineEXT3_ERROR_FS 0x0002 /* Errors detected */ #defineEXT3_ORPHAN_FS 0x0004 /* Orphans being recovered */ +#defineEXT3_REMOUNTING_FS 0x0008 /* Filesystem remounting, sb locked */ /* * Mount flags - To unsubscribe from this list: send the line unsubscribe linux-ext4 in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
[RFC][PATCH] process orphan list if device transitions to readonly
I have a patch in -mm now to skip orphan inode processing on a read-only device (where IO may fail if issued), but Stephen points out that if the device ever transitions back to readwrite, and the filesystem is remounted as rewrite, we should process the orphan inode list at that point. Today, I'm not sure how we can easily get into this situation - for example, lvm apparently cannot transform a RO snapshot into RW. But it is at least possible to do this. For example I tested by: create orphan list crash mark block device RO via BLKROSET mount fs RO mark block device RW via BLKROSET remount fs RW so it's within the realm of possibility, certainly. here's what I came up with; I'm not very pleased with it though, but I need some way to let ext3_orphan_del know that the sb is already locked and it's safe to do this orphan processing w/o re-taking it, so I added a state flag... Comments? Thanks, -Eric Signed-off-by: Eric Sandeen [EMAIL PROTECTED] Index: linux-2.6.19/fs/ext3/namei.c === --- linux-2.6.19.orig/fs/ext3/namei.c +++ linux-2.6.19/fs/ext3/namei.c @@ -1942,15 +1942,15 @@ int ext3_orphan_del(handle_t *handle, st struct ext3_iloc iloc; int err = 0; - lock_super(inode-i_sb); - if (list_empty(ei-i_orphan)) { - unlock_super(inode-i_sb); - return 0; - } + sbi = EXT3_SB(inode-i_sb); + if (!(sbi-s_mount_state EXT3_REMOUNTING_FS)) + lock_super(inode-i_sb); + + if (list_empty(ei-i_orphan)) + goto out; ino_next = NEXT_ORPHAN(inode); prev = ei-i_orphan.prev; - sbi = EXT3_SB(inode-i_sb); jbd_debug(4, remove inode %lu from orphan list\n, inode-i_ino); @@ -1996,7 +1996,8 @@ int ext3_orphan_del(handle_t *handle, st out_err: ext3_std_error(inode-i_sb, err); out: - unlock_super(inode-i_sb); + if (!(sbi-s_mount_state EXT3_REMOUNTING_FS)) + unlock_super(inode-i_sb); return err; out_brelse: Index: linux-2.6.19/fs/ext3/super.c === --- linux-2.6.19.orig/fs/ext3/super.c +++ linux-2.6.19/fs/ext3/super.c @@ -2358,6 +2358,10 @@ static int ext3_remount (struct super_bl goto restore_opts; if (!ext3_setup_super (sb, es, 0)) sb-s_flags = ~MS_RDONLY; + EXT3_SB(sb)-s_mount_state |= (EXT3_ORPHAN_FS|EXT3_REMOUNTING_FS); + ext3_orphan_cleanup(sb, es); + EXT3_SB(sb)-s_mount_state = ~(EXT3_ORPHAN_FS|EXT3_REMOUNTING_FS); + } } #ifdef CONFIG_QUOTA Index: linux-2.6.19/include/linux/ext3_fs.h === --- linux-2.6.19.orig/include/linux/ext3_fs.h +++ linux-2.6.19/include/linux/ext3_fs.h @@ -356,6 +356,7 @@ struct ext3_inode { #defineEXT3_VALID_FS 0x0001 /* Unmounted cleanly */ #defineEXT3_ERROR_FS 0x0002 /* Errors detected */ #defineEXT3_ORPHAN_FS 0x0004 /* Orphans being recovered */ +#defineEXT3_REMOUNTING_FS 0x0008 /* Filesystem remounting, sb locked */ /* * Mount flags - To unsubscribe from this list: send the line unsubscribe linux-ext4 in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html