Re: [RFC] ext3 freeze feature ver 0.2
Takashi Sato wrote: o Elevate XFS ioctl numbers (XFS_IOC_FREEZE and XFS_IOC_THAW) to the VFS As Andreas Dilger and Christoph Hellwig advised me, I have elevated them to include/linux/fs.h as below. #define FIFREEZE_IOWR('X', 119, int) #define FITHAW _IOWR('X', 120, int) The ioctl numbers used by XFS applications don't need to be changed. But my following ioctl for the freeze needs the parameter as the timeout period. So if XFS applications don't want the timeout feature as the current implementation, the parameter needs to be changed 1 (level?) into 0. So, existing xfs applications calling the xfs ioctl now will behave differently, right? We can only keep the same ioctl number if the calling semantics are the same. Keeping the same number but changing the semantics is harmful, IMHO -Eric - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature ver 0.2
On Feb 26, 2008 08:39 -0800, Eric Sandeen wrote: Takashi Sato wrote: o Elevate XFS ioctl numbers (XFS_IOC_FREEZE and XFS_IOC_THAW) to the VFS As Andreas Dilger and Christoph Hellwig advised me, I have elevated them to include/linux/fs.h as below. #define FIFREEZE_IOWR('X', 119, int) #define FITHAW _IOWR('X', 120, int) The ioctl numbers used by XFS applications don't need to be changed. But my following ioctl for the freeze needs the parameter as the timeout period. So if XFS applications don't want the timeout feature as the current implementation, the parameter needs to be changed 1 (level?) into 0. So, existing xfs applications calling the xfs ioctl now will behave differently, right? We can only keep the same ioctl number if the calling semantics are the same. Keeping the same number but changing the semantics is harmful, IMHO Do we know what this parameter was supposed to mean? We could special case 1 if needed to keep compatibility (documenting this clearly), either making it == 0, or some very long timeout (1h or whatever). A relatively minor wart I think. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, #defineFS_IOC_GETFLAGS _IOR('f', 1, long) #defineFS_IOC_SETFLAGS _IOW('f', 2, long) as generic vfs ioctls. These ioctls started out as EXT2_IOC_SETFLAGS/EXT2_IOC_GETFLAGS but they were generically useful, other filesystems picked them up, and they were elevated to the vfs. Thank you for good information. I will elevate XFS_IOC_FREEZE and XFS_IOC_THAW to the VFS. And xfs_freeze calls XFS_IOC_FREEZE with a magic number 1, but what is 1? Looks like it's called level but it's probably a holdover, it doesn't look like it's used. I see. Instead, I'd like the sec to timeout on freeze API in order to thaw the filesystem automatically. It can prevent a filesystem from staying frozen forever. (Because a freezer may cause a deadlock by accessing the frozen filesystem.) I'm still not very comfortable with the timeout; if you un-freeze on a timer, how do you know that the work for which you needed the fileystem frozen is complete? How would you know if your snapshot was good if there's a possibility that the fs unfroze while it was being taken? My following freeze ioctl never perform the timeout when 0 is specified as timeval. So, existent applications which don't expect the timeout can stay frozen with 0. int ioctl(int fd, int FIFREEZE, long *timeval); fd:file descriptor of mountpoint FIFREEZE:request cord for freeze timeval:timeout period (second) And how about adding the new ioctl to reset the timeval like below? (Dmitri proposed this idea before.) int ioctl(int fd, int FIFREEZE_RESET_TIMEOUT, long *timeval); fd:file descriptor of mountpoint FIFREEZE_RESET_TIMEOUT:request cord for reset of timeout period timeval:new timeout period This is useful for the application to set the timeval more accurately. For example, the freezer resets the timeval to 10 seconds every 5 seconds. In this approach, even if the freezer causes a deadlock by accessing the frozen filesystem, it will be solved by the timeout in 10 seconds and the freezer can recognize that at the next reset of timeval. Any comments are very welcome. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Feb 15, 2008 at 08:51:15PM +0900, Takashi Sato wrote: So XFS_IOC_FREEZE and XFS_IOC_THAW cannot be lifted to generic code simply. I think we should create new generic numbers for freeze and thaw Actually we've lifted specific ioctls to the generic layer before all the time in drivers. That's the only way to make functionality that was specific to a single driver (or in this case filesystem) generic. If the numbering issues confuses you make sure to add a big comment describing it And xfs_freeze calls XFS_IOC_FREEZE with a magic number 1, but what is 1? As Eric said it's ignored. Instead, I'd like the sec to timeout on freeze API in order to thaw the filesystem automatically. It can prevent a filesystem from staying frozen forever. (Because a freezer may cause a deadlock by accessing the frozen filesystem.) Timeout based locking is generally a horrible idea, there's a reason we don't have any primitives for that in the kernel :) - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, Christoph Hellwig wrote: On Fri, Feb 08, 2008 at 08:26:57AM -0500, Andreas Dilger wrote: You may as well make the common ioctl the same as the XFS version, both by number and parameters, so that applications which already understand the XFS ioctl will work on other filesystems. Yes. In facy you should be able to lift the implementations of XFS_IOC_FREEZE and XFS_IOC_THAW to generic code, there's nothing XFS-specific in there. According to Documentation/ioctl-number.txt, XFS_IOC_XXXs (_IOWR('X', aa, bb)) are defined for XFS like below. From Documentation/ioctl-number.txt: CodeSeq#Include FileComments : : 'X' all linux/xfs_fs.h So XFS_IOC_FREEZE and XFS_IOC_THAW cannot be lifted to generic code simply. I think we should create new generic numbers for freeze and thaw like FIBMAP as followings. linux/fs.h: #define FIFREEZE _IO(0x00,3) #define FITHAW _IO(0x00,4) And xfs_freeze calls XFS_IOC_FREEZE with a magic number 1, but what is 1? Instead, I'd like the sec to timeout on freeze API in order to thaw the filesystem automatically. It can prevent a filesystem from staying frozen forever. (Because a freezer may cause a deadlock by accessing the frozen filesystem.) Any comments are very welcome. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Takashi Sato wrote: Hi, Christoph Hellwig wrote: On Fri, Feb 08, 2008 at 08:26:57AM -0500, Andreas Dilger wrote: You may as well make the common ioctl the same as the XFS version, both by number and parameters, so that applications which already understand the XFS ioctl will work on other filesystems. Yes. In facy you should be able to lift the implementations of XFS_IOC_FREEZE and XFS_IOC_THAW to generic code, there's nothing XFS-specific in there. According to Documentation/ioctl-number.txt, XFS_IOC_XXXs (_IOWR('X', aa, bb)) are defined for XFS like below. From Documentation/ioctl-number.txt: CodeSeq#Include FileComments : : 'X' all linux/xfs_fs.h It also says: 'f' 00-1F linux/ext2_fs.h and yet include/linux.h has: #define FS_IOC_GETFLAGS _IOR('f', 1, long) #define FS_IOC_SETFLAGS _IOW('f', 2, long) as generic vfs ioctls. These ioctls started out as EXT2_IOC_SETFLAGS/EXT2_IOC_GETFLAGS but they were generically useful, other filesystems picked them up, and they were elevated to the vfs. So XFS_IOC_FREEZE and XFS_IOC_THAW cannot be lifted to generic code simply. It would be a simple matter of changing the documentation, I think. I think we should create new generic numbers for freeze and thaw like FIBMAP as followings. linux/fs.h: #define FIFREEZE _IO(0x00,3) #define FITHAW _IO(0x00,4) And xfs_freeze calls XFS_IOC_FREEZE with a magic number 1, but what is 1? Looks like it's called level but it's probably a holdover, it doesn't look like it's used. Instead, I'd like the sec to timeout on freeze API in order to thaw the filesystem automatically. It can prevent a filesystem from staying frozen forever. (Because a freezer may cause a deadlock by accessing the frozen filesystem.) I'm still not very comfortable with the timeout; if you un-freeze on a timer, how do you know that the work for which you needed the fileystem frozen is complete? How would you know if your snapshot was good if there's a possibility that the fs unfroze while it was being taken? Thanks, -Eric Any comments are very welcome. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, P.S. Oh yeah, it should be noted that freezing at the filesystem layer does *not* guarantee that changes to the block device aren't happening via mmap()'ed files. The LVM needs to freeze writes the block device level if it wants to guarantee a completely stable snapshot image. So the proposed patch doens't quite give you those guarantees, if that was the intended goal. I don't think a mmap()'ed file is written to a block device while a filesystem is frozen. pdflush starts the writing procedure of the mmap()'ed file's data and calls ext3_ordered_writepage. ext3_ordered_writepage calls ext3_journal_start to get the journal handle. As a result, the process waits for unfreeze in start_this_handle. pdflush :: ext3_ordered_writepage ext3_journal_start ext3_journal_start_sb journal_start start_this_handle --- wait here I actually tried freezing the filesystem after updating the mmap()'ed file's data. But, the writing to the block device didn't happen. (It happened right after unfreeze.) I don't think the freeze feature on the block device level is needed because the writing for the mmap()'ed file is suspended on the frozen filesystem. Any comments are very welcome. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, Ted wrote: And I do agree that we probably should just implement this in filesystem independent way, in which case all of the filesystems that support this already have super_operations functions write_super_lockfs() and unlockfs(). So if this is done using a new system call, there should be no filesystem-specific changes needed, and all filesystems which support those super_operations method functions would be able to provide this functionality to the new system call. OK I would like to implement the freeze feature on VFS as the filesystem independent ioctl so that it can be available on filesystems that have already had write_super_lockfs() and unlockfs(). The usage for the freeze ioctl is the following. int ioctl(int fd, int FIFREEZE, long *timeval); fd:file descriptor of mountpoint FIFREEZE:request cord for freeze timeval:timeout period (second) And the unfreeze ioctl is the following. int ioctl(int fd, int FITHAW, NULL); fd:file descriptor of mountpoint FITHAW:Request cord for unfreeze I think we need the timeout feature which thaws the filesystem after lapse of specified time for a fail-safe in case the freezer accesses the frozen filesystem and causes a deadlock. I intend to implement the timeout feature on VFS. (This is realized by registering the delayed work which calls thaw_bdev() to the delayed work queue.) Any comments are very welcome. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Feb 08, 2008 at 08:26:57AM -0500, Andreas Dilger wrote: You may as well make the common ioctl the same as the XFS version, both by number and parameters, so that applications which already understand the XFS ioctl will work on other filesystems. Yes. In facy you should be able to lift the implementations of XFS_IOC_FREEZE and XFS_IOC_THAW to generic code, there's nothing XFS-specific in there. - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Feb 08, 2008 19:48 +0900, Takashi Sato wrote: OK I would like to implement the freeze feature on VFS as the filesystem independent ioctl so that it can be available on filesystems that have already had write_super_lockfs() and unlockfs(). The usage for the freeze ioctl is the following. int ioctl(int fd, int FIFREEZE, long *timeval); fd:file descriptor of mountpoint FIFREEZE:request cord for freeze timeval:timeout period (second) And the unfreeze ioctl is the following. int ioctl(int fd, int FITHAW, NULL); fd:file descriptor of mountpoint FITHAW:Request cord for unfreeze You may as well make the common ioctl the same as the XFS version, both by number and parameters, so that applications which already understand the XFS ioctl will work on other filesystems. Cheers, Andreas -- Andreas Dilger Sr. Staff Engineer, Lustre Group Sun Microsystems of Canada, Inc. - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, What you *could* do is to start putting processes to sleep if they attempt to write to the frozen filesystem, and then detect the deadlock case where the process holding the file descriptor used to freeze the filesystem gets frozen because it attempted to write to the filesystem --- at which point it gets some kind of signal (which defaults to killing the process), and the filesystem is unfrozen and as part of the unfreeze you wake up all of the processes that were put to sleep for touching the frozen filesystem. The other approach would be to say, oh well, the freeze ioctl is inherently dangerous, and root is allowed to himself in the foot, so who cares. :-) Currently the XFS freezer doesn't solve a deadlock automatically and we rely on administrators for ensuring that the freezer will not access the filesystem. And even if the wrong freezer causes a deadlock, it can be solved by other unfreeze process(unfreeze command). So I don't think the freezer itself needs to solve the deadlock. I think the timeout is effective for a unexpected deadlock and the timeout extending feature is very useful as Dmitri proposed. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Sato-san, What you *could* do is to start putting processes to sleep if they attempt to write to the frozen filesystem, and then detect the deadlock case where the process holding the file descriptor used to freeze the filesystem gets frozen because it attempted to write to the filesystem --- at which point it gets some kind of signal (which defaults to killing the process), and the filesystem is unfrozen and as part of the unfreeze you wake up all of the processes that were put to sleep for touching the frozen filesystem. I don't think close() usually writes to journal and the deadlock occurs. Is there the special case which close() writes to journal in case of getting signal? I am afraid that Ted-san is concerning about the fact that the freeze program can touch the target filesystem by mistake (rather than just close() and its journal behavior) # cd /mnt # freeze /mnt ./logfile This is more unclear to admins than rm -rf / case. So we need to implement some bail-out mechanism as he pointed out, such that if kernel noticed that the freezer is trying to touch the target filesystem, kernel would kill the freezer and automatically unfreeze the filesystem again. Regards, - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, What you *could* do is to start putting processes to sleep if they attempt to write to the frozen filesystem, and then detect the deadlock case where the process holding the file descriptor used to freeze the filesystem gets frozen because it attempted to write to the filesystem --- at which point it gets some kind of signal (which defaults to killing the process), and the filesystem is unfrozen and as part of the unfreeze you wake up all of the processes that were put to sleep for touching the frozen filesystem. I don't think close() usually writes to journal and the deadlock occurs. Is there the special case which close() writes to journal in case of getting signal? Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, Thank you for your comments. That's inherently unsafe - you can have multiple unfreezes running in parallel which seriously screws with the bdev semaphore count that is used to lock the device due to doing multiple up()s for every down. Your timeout thingy guarantee that at some point you will get multiple up()s occuring due to the timer firing racing with a thaw ioctl. If this interface is to be more widely exported, then it needs a complete revamp of the bdev is locked while it is frozen so that there is no chance of a double up() ever occuring on the bd_mount_sem due to racing thaws. My patch has the race condition as you said. I will fix it. Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Jan 25, 2008 at 09:42:30PM +0900, Takashi Sato wrote: Hi, I am also wondering whether we should have system call(s) for these: On Jan 25, 2008 12:59 PM, Takashi Sato [EMAIL PROTECTED] wrote: + case EXT3_IOC_FREEZE: { + case EXT3_IOC_THAW: { And just convert XFS to use them too? I think it is reasonable to implement it as the generic system call, as you said. Does XFS folks think so? Given that XFS has implemented the ioctls for such a long time it might make more sense to simply move the ioctl implementation to fs/ioctl.c so it applies to all filesystem. No need to add a new syscall when the equivalent-functionality ioctls have to be supported forever anyway. - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
[RFC] ext3 freeze feature
Hi, Currently, ext3 doesn't have the freeze feature which suspends write requests. So, we cannot get a backup which keeps the filesystem's consistency with the storage device's features (snapshot, replication) while it is mounted. In many case, a commercial filesystems (e.g. VxFS) has the freeze feature and it would be used to get the consistent backup. So I am planning on implementing the ioctl of the freeze feature for ext3. I think we can get the consistent backup with the following steps. 1. Freeze the filesystem with ioctl. 2. Separate the replication volume or get the snapshot with the storage device's feature. 3. Unfreeze the filesystem with ioctl. 4. Get the backup from the separated replication volume or the snapshot. The usage of the ioctl is as below. int ioctl(int fd, int cmd, long *timeval) fd: The file descriptor of the mountpoint. cmd: EXT3_IOC_FREEZE for the freeze or EXT3_IOC_THAW for the unfreeze. timeval: The timeout value expressed in seconds. If it's 0, the timeout isn't set. Return value: 0 if the operation succeeds. Otherwise, -1. I have made sure that write requests were suspended with the experimental patch for this feature and attached it in this mail. The points of the implementation are followings. - Add calls of the freeze function (freeze_bdev) and the unfreeze function (thaw_bdev) in ext3_ioctl(). - ext3_freeze_timeout() which calls the unfreeze function (thaw_bdev) is registered to the delayed work queue to unfreeze the filesystem automatically after the lapse of the specified time. Any comments are very welcome. Signed-off-by: Takashi Sato [EMAIL PROTECTED] --- diff -uprN -X linux-2.6.24-rc8/Documentation/dontdiff linux-2.6.24-rc8/fs/ext3/ioctl.c linux-2.6.24-rc8-freeze/fs/ext3/ioctl.c --- linux-2.6.24-rc8/fs/ext3/ioctl.c2008-01-16 13:22:48.0 +0900 +++ linux-2.6.24-rc8-freeze/fs/ext3/ioctl.c 2008-01-22 18:20:33.0 +0900 @@ -254,6 +254,42 @@ flags_err: return err; } + case EXT3_IOC_FREEZE: { + long timeout_sec; + long timeout_msec; + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + if (inode-i_sb-s_frozen != SB_UNFROZEN) + return -EINVAL; + /* arg(sec) to tick value */ + get_user(timeout_sec, (long __user *) arg); + timeout_msec = timeout_sec * 1000; + if (timeout_msec 0) + return -EINVAL; + + /* Freeze */ + freeze_bdev(inode-i_sb-s_bdev); + + /* set up unfreeze timer */ + if (timeout_msec 0) + ext3_add_freeze_timeout(EXT3_SB(inode-i_sb), + timeout_msec); + return 0; + } + case EXT3_IOC_THAW: { + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + if (inode-i_sb-s_frozen == SB_UNFROZEN) + return -EINVAL; + + /* delete unfreeze timer */ + ext3_del_freeze_timeout(EXT3_SB(inode-i_sb)); + + /* Unfreeze */ + thaw_bdev(inode-i_sb-s_bdev, inode-i_sb); + + return 0; + } default: return -ENOTTY; diff -uprN -X linux-2.6.24-rc8/Documentation/dontdiff linux-2.6.24-rc8/fs/ext3/super.c linux-2.6.24-rc8-freeze/fs/ext3/super.c --- linux-2.6.24-rc8/fs/ext3/super.c2008-01-16 13:22:48.0 +0900 +++ linux-2.6.24-rc8-freeze/fs/ext3/super.c 2008-01-22 18:20:33.0 +0900 @@ -63,6 +63,7 @@ static int ext3_statfs (struct dentry * static void ext3_unlockfs(struct super_block *sb); static void ext3_write_super (struct super_block * sb); static void ext3_write_super_lockfs(struct super_block *sb); +static void ext3_freeze_timeout(struct work_struct *work); /* * Wrappers for journal_start/end. @@ -323,6 +324,44 @@ void ext3_update_dynamic_rev(struct supe } /* + * ext3_add_freeze_timeout - Add timeout for ext3 freeze. + * + * @sbi: ext3 super block + * @timeout_msec : timeout period + * + * Add the delayed work for ext3 freeze timeout + * to the delayed work queue. + */ +void ext3_add_freeze_timeout(struct ext3_sb_info *sbi, + long timeout_msec) +{ + s64 timeout_jiffies = msecs_to_jiffies(timeout_msec); + + /* +* setup freeze timeout function +*/ + INIT_DELAYED_WORK(sbi-s_freeze_timeout, ext3_freeze_timeout); + + /* set delayed work queue */ + cancel_delayed_work(sbi-s_freeze_timeout); + schedule_delayed_work(sbi-s_freeze_timeout, timeout_jiffies); +} + +/* + * ext3_del_freeze_timeout - Delete timeout for ext3 freeze. + * + * @sbi: ext3 super block + * + * Delete the delayed work for ext3 freeze timeout + * from the delayed work queue. + */ +void ext3_del_freeze_timeout(struct
Re: [RFC] ext3 freeze feature
Hi, diff -uprN -X linux-2.6.24-rc8/Documentation/dontdiff linux-2.6.24-rc8/include/linux/ext3_fs_sb.h linux-2.6.24-rc8-freeze/include/linux/ext3_fs_sb.h --- linux-2.6.24-rc8/include/linux/ext3_fs_sb.h 2008-01-16 13:22:48.0 +0900 +++ linux-2.6.24-rc8-freeze/include/linux/ext3_fs_sb.h 2008-01-22 18:20:33.0 +0900 @@ -81,6 +81,8 @@ struct ext3_sb_info { char *s_qf_names[MAXQUOTAS];/* Names of quota files with journalled quota */ int s_jquota_fmt; /* Format of quota to use */ #endif + /* Delayed work for freeze */ + struct delayed_work s_freeze_timeout; Why not put this in struct super_block? Then you don't need this +/** + * get_super_block - get super_block + * @s_fs_info : filesystem dependent information + * (super_block.s_fs_info) + * + * Get super_block which holds s_fs_info from super_blocks. + * get_super_block() returns a pointer of super block or + * %NULL if it have failed. + */ +struct super_block *get_super_block(void *s_fs_info) +{ And these can be put to generic code: /* + * ext3_add_freeze_timeout - Add timeout for ext3 freeze. + * + * @sbi: ext3 super block + * @timeout_msec : timeout period + * + * Add the delayed work for ext3 freeze timeout + * to the delayed work queue. + */ +void ext3_add_freeze_timeout(struct ext3_sb_info *sbi, + long timeout_msec) +{ + s64 timeout_jiffies = msecs_to_jiffies(timeout_msec); + + /* +* setup freeze timeout function +*/ + INIT_DELAYED_WORK(sbi-s_freeze_timeout, ext3_freeze_timeout); + + /* set delayed work queue */ + cancel_delayed_work(sbi-s_freeze_timeout); + schedule_delayed_work(sbi-s_freeze_timeout, timeout_jiffies); +} + +/* + * ext3_del_freeze_timeout - Delete timeout for ext3 freeze. + * + * @sbi: ext3 super block + * + * Delete the delayed work for ext3 freeze timeout + * from the delayed work queue. + */ +void ext3_del_freeze_timeout(struct ext3_sb_info *sbi) +{ + if (delayed_work_pending(sbi-s_freeze_timeout)) + cancel_delayed_work(sbi-s_freeze_timeout); +} +/* + * ext3_freeze_timeout - Thaw the filesystem. + * + * @work : work queue (delayed_work.work) + * + * Called by the delayed work when elapsing the timeout period. + * Thaw the filesystem. + */ +static void ext3_freeze_timeout(struct work_struct *work) +{ + struct ext3_sb_info *sbi = container_of(work, + struct ext3_sb_info, + s_freeze_timeout.work); + struct super_block *sb = get_super_block(sbi); + + BUG_ON(sb == NULL); + + if (sb-s_frozen != SB_UNFROZEN) + thaw_bdev(sb-s_bdev, sb); +} + I am also wondering whether we should have system call(s) for these: On Jan 25, 2008 12:59 PM, Takashi Sato [EMAIL PROTECTED] wrote: + case EXT3_IOC_FREEZE: { + case EXT3_IOC_THAW: { And just convert XFS to use them too? Pekka - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Hi, I am also wondering whether we should have system call(s) for these: On Jan 25, 2008 12:59 PM, Takashi Sato [EMAIL PROTECTED] wrote: + case EXT3_IOC_FREEZE: { + case EXT3_IOC_THAW: { And just convert XFS to use them too? I think it is reasonable to implement it as the generic system call, as you said. Does XFS folks think so? Cheers, Takashi - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On 19:59 Fri 25 Jan , Takashi Sato wrote: Hi, Currently, ext3 doesn't have the freeze feature which suspends write requests. So, we cannot get a backup which keeps the filesystem's consistency with the storage device's features (snapshot, replication) while it is mounted. In many case, a commercial filesystems (e.g. VxFS) has the freeze feature and it would be used to get the consistent backup. First of all Linux already have at least one open-source(dm-snap), and several commercial snapshot solutions. In fact dm-snaps it not perfect: a) bit map loading is not supported (this is useful for freezing only used blocks) which causing significant slowdown even for new writes b) non patched dm-snap code has significant performance slowdown for all rewrite requests. c) IMHO memory footprint is too big. BUT, it works well for most file-systems. So I am planning on implementing the ioctl of the freeze feature for ext3. I think we can get the consistent backup with the following steps. 1. Freeze the filesystem with ioctl. So you plan to do it from userspace.. well good luck with it :) 2. Separate the replication volume or get the snapshot with the storage device's feature. 3. Unfreeze the filesystem with ioctl. You have to realize what delay between 1-3 stages have to be minimal. for example dm-snap perform it only for explicit journal flushing. From my experience if delay is more than 4-5 seconds whole system becomes unstable. BTW: you have to always remember that while locking ext3 via freeze_bdev sb-ext3_write_super_lockfs() will be called wich implemented as simple journal lock. This means what some bio-s still may reach original device even after file system was locked (i've observed this in real life situation). 4. Get the backup from the separated replication volume or the snapshot. The usage of the ioctl is as below. int ioctl(int fd, int cmd, long *timeval) fd: The file descriptor of the mountpoint. cmd: EXT3_IOC_FREEZE for the freeze or EXT3_IOC_THAW for the unfreeze. timeval: The timeout value expressed in seconds. If it's 0, the timeout isn't set. Return value: 0 if the operation succeeds. Otherwise, -1. I have made sure that write requests were suspended with the experimental patch for this feature and attached it in this mail. The points of the implementation are followings. - Add calls of the freeze function (freeze_bdev) and the unfreeze function (thaw_bdev) in ext3_ioctl(). - ext3_freeze_timeout() which calls the unfreeze function (thaw_bdev) is registered to the delayed work queue to unfreeze the filesystem automatically after the lapse of the specified time. Any comments are very welcome. Signed-off-by: Takashi Sato [EMAIL PROTECTED] --- diff -uprN -X linux-2.6.24-rc8/Documentation/dontdiff linux-2.6.24-rc8/fs/ext3/ioctl.c linux-2.6.24-rc8-freeze/fs/ext3/ioctl.c --- linux-2.6.24-rc8/fs/ext3/ioctl.c 2008-01-16 13:22:48.0 +0900 +++ linux-2.6.24-rc8-freeze/fs/ext3/ioctl.c 2008-01-22 18:20:33.0 +0900 @@ -254,6 +254,42 @@ flags_err: return err; } + case EXT3_IOC_FREEZE: { + long timeout_sec; + long timeout_msec; + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + if (inode-i_sb-s_frozen != SB_UNFROZEN) + return -EINVAL WOW timeout extending is not supported !? So you wanna say what caller have to set timer to the maximal possible timeout from the very beginning. IMHO it is better to use heart-beat timer approach, for example: each second caller extend it's timeout for two seconds. in this approach even after caller was killed by any reason, it's timeout will be expired in two seconds. if (inode-i_sb-s_frozen == SB_FROZEN) /* extending timeout */ .. + /* arg(sec) to tick value */ + get_user(timeout_sec, (long __user *) arg); + timeout_msec = timeout_sec * 1000; + if (timeout_msec 0) + return -EINVAL; + + /* Freeze */ + freeze_bdev(inode-i_sb-s_bdev); + + /* set up unfreeze timer */ + if (timeout_msec 0) + ext3_add_freeze_timeout(EXT3_SB(inode-i_sb), + timeout_msec); + return 0; + } + case EXT3_IOC_THAW: { + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + if (inode-i_sb-s_frozen == SB_UNFROZEN) + return -EINVAL; + + /* delete unfreeze timer */ + ext3_del_freeze_timeout(EXT3_SB(inode-i_sb)); + + /* Unfreeze */ + thaw_bdev(inode-i_sb-s_bdev, inode-i_sb); + + return 0; + } default: return -ENOTTY; diff -uprN -X
Re: [RFC] ext3 freeze feature
On Sat, Jan 26, 2008 at 04:35:26PM +1100, David Chinner wrote: On Fri, Jan 25, 2008 at 07:59:38PM +0900, Takashi Sato wrote: The points of the implementation are followings. - Add calls of the freeze function (freeze_bdev) and the unfreeze function (thaw_bdev) in ext3_ioctl(). - ext3_freeze_timeout() which calls the unfreeze function (thaw_bdev) is registered to the delayed work queue to unfreeze the filesystem automatically after the lapse of the specified time. Seems like pointless complexity to me - what happens if a timeout occurs while the filsystem is still freezing? It's not uncommon for a freeze to take minutes if memory is full of dirty data that needs to be flushed out, esp. if dm-snap is doing COWs for every write issued Sorry, ignore this bit - I just realised the timer is set up after the freeze has occurred Still, that makes it potentially dangerous to whatever is being done while the filesystem is frozen Cheers, Dave. -- Dave Chinner Principal Engineer SGI Australian Software Group - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Jan 25, 2008 at 07:59:38PM +0900, Takashi Sato wrote: The points of the implementation are followings. - Add calls of the freeze function (freeze_bdev) and the unfreeze function (thaw_bdev) in ext3_ioctl(). - ext3_freeze_timeout() which calls the unfreeze function (thaw_bdev) is registered to the delayed work queue to unfreeze the filesystem automatically after the lapse of the specified time. Seems like pointless complexity to me - what happens if a timeout occurs while the filsystem is still freezing? It's not uncommon for a freeze to take minutes if memory is full of dirty data that needs to be flushed out, esp. if dm-snap is doing COWs for every write issued + case EXT3_IOC_FREEZE: { + if (inode-i_sb-s_frozen != SB_UNFROZEN) + return -EINVAL; + freeze_bdev(inode-i_sb-s_bdev); + case EXT3_IOC_THAW: { + if (!capable(CAP_SYS_ADMIN)) + return -EPERM; + if (inode-i_sb-s_frozen == SB_UNFROZEN) + return -EINVAL; . + /* Unfreeze */ + thaw_bdev(inode-i_sb-s_bdev, inode-i_sb); That's inherently unsafe - you can have multiple unfreezes running in parallel which seriously screws with the bdev semaphore count that is used to lock the device due to doing multiple up()s for every down. Your timeout thingy guarantee that at some point you will get multiple up()s occuring due to the timer firing racing with a thaw ioctl. If this interface is to be more widely exported, then it needs a complete revamp of the bdev is locked while it is frozen so that there is no chance of a double up() ever occuring on the bd_mount_sem due to racing thaws. Cheers, Dave. -- Dave Chinner Principal Engineer SGI Australian Software Group - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Jan 25, 2008 at 09:42:30PM +0900, Takashi Sato wrote: I am also wondering whether we should have system call(s) for these: On Jan 25, 2008 12:59 PM, Takashi Sato [EMAIL PROTECTED] wrote: + case EXT3_IOC_FREEZE: { + case EXT3_IOC_THAW: { And just convert XFS to use them too? I think it is reasonable to implement it as the generic system call, as you said. Does XFS folks think so? Sure. Note that we can't immediately remove the XFS ioctls otherwise we'd break userspace utilities that use them Cheers, Dave. -- Dave Chinner Principal Engineer SGI Australian Software Group - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Jan 25, 2008 at 10:34:25AM -0600, Eric Sandeen wrote: But it was this concern which is why ext3 never exported freeze functionality to userspace, even though other commercial filesystems do support this. It wasn't that it wasn't considered, but the concern about whether or not it was sufficiently safe to make available. What's the safety concern; that the admin will forget to unfreeze? That the admin would manage to deadlock him/herself and wedge up the whole system... I'm also not sure I see the point of the timeout in the original patch; either you are done snapshotting and ready to unfreeze, or you're not; 1, or 2, or 3 seconds doesn't really matter. When you're done, you're done, and you can only unfreeze then. Shouldn't this be done programmatically, and not with some pre-determined timeout? This is only a guess, but I suspect it was a fail-safe in case the admin did manage to deadlock him/herself. I would think a better approach would be to make the filesystem unfreeze if the file descriptor that was used to freeze the filesystem is closed, and then have explicit deadlock detection that kills the process doing the freeze, at which point the filesystem unlocks and the system can recover. - Ted - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
Theodore Tso wrote: The other approach would be to say, oh well, the freeze ioctl is inherently dangerous, and root is allowed to himself in the foot, so who cares. :-) I tend to agree. Either you need your fs frozen, or not, and if you do, be prepared for the consequences. But it was this concern which is why ext3 never exported freeze functionality to userspace, even though other commercial filesystems do support this. It wasn't that it wasn't considered, but the concern about whether or not it was sufficiently safe to make available. What's the safety concern; that the admin will forget to unfreeze? And I do agree that we probably should just implement this in filesystem independent way, in which case all of the filesystems that support this already have super_operations functions write_super_lockfs() and unlockfs(). That's what I was thinking; can't the path to freeze_bdev just be elevated out of dm-ioctl.c to fs/ioctl.c and exposed, such that any filesystem which implements .write_super_lockfs can be frozen? This is essentially what the xfs_freeze userspace does via xfs_ioctl/XFS_IOC_FREEZE - which, AFAIK, isn't used much now that the lvm hooks are in place. I'm also not sure I see the point of the timeout in the original patch; either you are done snapshotting and ready to unfreeze, or you're not; 1, or 2, or 3 seconds doesn't really matter. When you're done, you're done, and you can only unfreeze then. Shouldn't this be done programmatically, and not with some pre-determined timeout? -Eric - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: [RFC] ext3 freeze feature
On Fri, Jan 25, 2008 at 03:18:51PM +0300, Dmitri Monakhov wrote: First of all Linux already have at least one open-source(dm-snap), and several commercial snapshot solutions. Yes, but it requires that the filesystem be stored under LVM. Unlike what EVMS v1 allowed us to do, we can't currently take a snapshot of a bare block device. This patch could potentially be useful for systems which aren't using LVM, however You have to realize what delay between 1-3 stages have to be minimal. for example dm-snap perform it only for explicit journal flushing. From my experience if delay is more than 4-5 seconds whole system becomes unstable. That's the problem. You can't afford to freeze for very long. What you *could* do is to start putting processes to sleep if they attempt to write to the frozen filesystem, and then detect the deadlock case where the process holding the file descriptor used to freeze the filesystem gets frozen because it attempted to write to the filesystem --- at which point it gets some kind of signal (which defaults to killing the process), and the filesystem is unfrozen and as part of the unfreeze you wake up all of the processes that were put to sleep for touching the frozen filesystem. The other approach would be to say, oh well, the freeze ioctl is inherently dangerous, and root is allowed to himself in the foot, so who cares. :-) But it was this concern which is why ext3 never exported freeze functionality to userspace, even though other commercial filesystems do support this. It wasn't that it wasn't considered, but the concern about whether or not it was sufficiently safe to make available. And I do agree that we probably should just implement this in filesystem independent way, in which case all of the filesystems that support this already have super_operations functions write_super_lockfs() and unlockfs(). So if this is done using a new system call, there should be no filesystem-specific changes needed, and all filesystems which support those super_operations method functions would be able to provide this functionality to the new system call. - Ted P.S. Oh yeah, it should be noted that freezing at the filesystem layer does *not* guarantee that changes to the block device aren't happening via mmap()'ed files. The LVM needs to freeze writes the block device level if it wants to guarantee a completely stable snapshot image. So the proposed patch doens't quite give you those guarantees, if that was the intended goal. - To unsubscribe from this list: send the line unsubscribe linux-fsdevel in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html