From 8ebc423238341b52912c7295b045a32477b33f09 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Tue, 7 Apr 2009 04:19:49 +0200 Subject: reiserfs: kill-the-BKL This patch is an attempt to remove the Bkl based locking scheme from reiserfs and is intended. It is a bit inspired from an old attempt by Peter Zijlstra: http://lkml.indiana.edu/hypermail/linux/kernel/0704.2/2174.html The bkl is heavily used in this filesystem to prevent from concurrent write accesses on the filesystem. Reiserfs makes a deep use of the specific properties of the Bkl: - It can be acqquired recursively by a same task - It is released on the schedule() calls and reacquired when schedule() returns The two properties above are a roadmap for the reiserfs write locking so it's very hard to simply replace it with a common mutex. - We need a recursive-able locking unless we want to restructure several blocks of the code. - We need to identify the sites where the bkl was implictly relaxed (schedule, wait, sync, etc...) so that we can in turn release and reacquire our new lock explicitly. Such implicit releases of the lock are often required to let other resources producer/consumer do their job or we can suffer unexpected starvations or deadlocks. So the new lock that replaces the bkl here is a per superblock mutex with a specific property: it can be acquired recursively by a same task, like the bkl. For such purpose, we integrate a lock owner and a lock depth field on the superblock information structure. The first axis on this patch is to turn reiserfs_write_(un)lock() function into a wrapper to manage this mutex. Also some explicit calls to lock_kernel() have been converted to reiserfs_write_lock() helpers. The second axis is to find the important blocking sites (schedule...(), wait_on_buffer(), sync_dirty_buffer(), etc...) and then apply an explicit release of the write lock on these locations before blocking. Then we can safely wait for those who can give us resources or those who need some. Typically this is a fight between the current writer, the reiserfs workqueue (aka the async commiter) and the pdflush threads. The third axis is a consequence of the second. The write lock is usually on top of a lock dependency chain which can include the journal lock, the flush lock or the commit lock. So it's dangerous to release and trying to reacquire the write lock while we still hold other locks. This is fine with the bkl: T1 T2 lock_kernel() mutex_lock(A) unlock_kernel() // do something lock_kernel() mutex_lock(A) -> already locked by T1 schedule() (and then unlock_kernel()) lock_kernel() mutex_unlock(A) .... This is not fine with a mutex: T1 T2 mutex_lock(write) mutex_lock(A) mutex_unlock(write) // do something mutex_lock(write) mutex_lock(A) -> already locked by T1 schedule() mutex_lock(write) -> already locked by T2 deadlock The solution in this patch is to provide a helper which releases the write lock and sleep a bit if we can't lock a mutex that depend on it. It's another simulation of the bkl behaviour. The last axis is to locate the fs callbacks that are called with the bkl held, according to Documentation/filesystem/Locking. Those are: - reiserfs_remount - reiserfs_fill_super - reiserfs_put_super Reiserfs didn't need to explicitly lock because of the context of these callbacks. But now we must take care of that with the new locking. After this patch, reiserfs suffers from a slight performance regression (for now). On UP, a high volume write with dd reports an average of 27 MB/s instead of 30 MB/s without the patch applied. Signed-off-by: Frederic Weisbecker Reviewed-by: Ingo Molnar Cc: Jeff Mahoney Cc: Peter Zijlstra Cc: Bron Gondwana Cc: Andrew Morton Cc: Linus Torvalds Cc: Alexander Viro LKML-Reference: <1239070789-13354-1-git-send-email-fweisbec@gmail.com> Signed-off-by: Ingo Molnar --- fs/reiserfs/ioctl.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-) (limited to 'fs/reiserfs/ioctl.c') diff --git a/fs/reiserfs/ioctl.c b/fs/reiserfs/ioctl.c index 0ccc3fdda7b..5e40b0cd4c3 100644 --- a/fs/reiserfs/ioctl.c +++ b/fs/reiserfs/ioctl.c @@ -141,9 +141,11 @@ long reiserfs_compat_ioctl(struct file *file, unsigned int cmd, default: return -ENOIOCTLCMD; } - lock_kernel(); + + reiserfs_write_lock(inode->i_sb); ret = reiserfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg)); - unlock_kernel(); + reiserfs_write_unlock(inode->i_sb); + return ret; } #endif -- cgit v1.2.3-70-g09d2 From ac78a07893d24d95ff5f39d0433c25210f224f07 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 14 Oct 2009 23:08:43 +0200 Subject: kill-the-bkl/reiserfs: always lock the ioctl path Reiserfs uses the ioctl callback for its file operations, which means that its ioctl path is still locked by the bkl, this was synchronizing with the rest of the filsystem operations. We have changed that by locking it with the new reiserfs lock but we do that only from the compat_ioctl callback. Fix that by locking reiserfs_ioctl() everytime. Signed-off-by: Frederic Weisbecker Cc: Jeff Mahoney Cc: Chris Mason Cc: Ingo Molnar Cc: Alexander Beregalov Cc: Laurent Riffard Cc: Thomas Gleixner --- fs/reiserfs/ioctl.c | 66 +++++++++++++++++++++++++++++++---------------------- 1 file changed, 39 insertions(+), 27 deletions(-) (limited to 'fs/reiserfs/ioctl.c') diff --git a/fs/reiserfs/ioctl.c b/fs/reiserfs/ioctl.c index 5e40b0cd4c3..e30e8be0917 100644 --- a/fs/reiserfs/ioctl.c +++ b/fs/reiserfs/ioctl.c @@ -13,44 +13,52 @@ #include /* -** reiserfs_ioctl - handler for ioctl for inode -** supported commands: -** 1) REISERFS_IOC_UNPACK - try to unpack tail from direct item into indirect -** and prevent packing file (argument arg has to be non-zero) -** 2) REISERFS_IOC_[GS]ETFLAGS, REISERFS_IOC_[GS]ETVERSION -** 3) That's all for a while ... -*/ + * reiserfs_ioctl - handler for ioctl for inode + * supported commands: + * 1) REISERFS_IOC_UNPACK - try to unpack tail from direct item into indirect + * and prevent packing file (argument arg has to be non-zero) + * 2) REISERFS_IOC_[GS]ETFLAGS, REISERFS_IOC_[GS]ETVERSION + * 3) That's all for a while ... + */ int reiserfs_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, unsigned long arg) { unsigned int flags; int err = 0; + reiserfs_write_lock(inode->i_sb); + switch (cmd) { case REISERFS_IOC_UNPACK: if (S_ISREG(inode->i_mode)) { if (arg) - return reiserfs_unpack(inode, filp); - else - return 0; + err = reiserfs_unpack(inode, filp); } else - return -ENOTTY; - /* following two cases are taken from fs/ext2/ioctl.c by Remy - Card (card@masi.ibp.fr) */ + err = -ENOTTY; + break; + /* + * following two cases are taken from fs/ext2/ioctl.c by Remy + * Card (card@masi.ibp.fr) + */ case REISERFS_IOC_GETFLAGS: - if (!reiserfs_attrs(inode->i_sb)) - return -ENOTTY; + if (!reiserfs_attrs(inode->i_sb)) { + err = -ENOTTY; + break; + } flags = REISERFS_I(inode)->i_attrs; i_attrs_to_sd_attrs(inode, (__u16 *) & flags); - return put_user(flags, (int __user *)arg); + err = put_user(flags, (int __user *)arg); + break; case REISERFS_IOC_SETFLAGS:{ - if (!reiserfs_attrs(inode->i_sb)) - return -ENOTTY; + if (!reiserfs_attrs(inode->i_sb)) { + err = -ENOTTY; + break; + } err = mnt_want_write(filp->f_path.mnt); if (err) - return err; + break; if (!is_owner_or_cap(inode)) { err = -EPERM; @@ -90,16 +98,18 @@ int reiserfs_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, mark_inode_dirty(inode); setflags_out: mnt_drop_write(filp->f_path.mnt); - return err; + break; } case REISERFS_IOC_GETVERSION: - return put_user(inode->i_generation, (int __user *)arg); + err = put_user(inode->i_generation, (int __user *)arg); + break; case REISERFS_IOC_SETVERSION: if (!is_owner_or_cap(inode)) - return -EPERM; + err = -EPERM; + break; err = mnt_want_write(filp->f_path.mnt); if (err) - return err; + break; if (get_user(inode->i_generation, (int __user *)arg)) { err = -EFAULT; goto setversion_out; @@ -108,10 +118,14 @@ setflags_out: mark_inode_dirty(inode); setversion_out: mnt_drop_write(filp->f_path.mnt); - return err; + break; default: - return -ENOTTY; + err = -ENOTTY; } + + reiserfs_write_unlock(inode->i_sb); + + return err; } #ifdef CONFIG_COMPAT @@ -142,9 +156,7 @@ long reiserfs_compat_ioctl(struct file *file, unsigned int cmd, return -ENOIOCTLCMD; } - reiserfs_write_lock(inode->i_sb); ret = reiserfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg)); - reiserfs_write_unlock(inode->i_sb); return ret; } -- cgit v1.2.3-70-g09d2 From 205cb37b89ab37db553907e5ac17962eec561804 Mon Sep 17 00:00:00 2001 From: Frederic Weisbecker Date: Wed, 14 Oct 2009 23:22:17 +0200 Subject: kill-the-bkl/reiserfs: definitely drop the bkl from reiserfs_ioctl() The reiserfs ioctl path doesn't need the big kernel lock anymore , now that the filesystem synchronizes through its own lock. We can then turn reiserfs_ioctl() into an unlocked_ioctl callback. Signed-off-by: Frederic Weisbecker Cc: Jeff Mahoney Cc: Chris Mason Cc: Ingo Molnar Cc: Alexander Beregalov Cc: Laurent Riffard Cc: Thomas Gleixner --- fs/reiserfs/dir.c | 2 +- fs/reiserfs/file.c | 2 +- fs/reiserfs/ioctl.c | 11 +++-------- include/linux/reiserfs_fs.h | 3 +-- 4 files changed, 6 insertions(+), 12 deletions(-) (limited to 'fs/reiserfs/ioctl.c') diff --git a/fs/reiserfs/dir.c b/fs/reiserfs/dir.c index 17f31ad379c..c094f58c744 100644 --- a/fs/reiserfs/dir.c +++ b/fs/reiserfs/dir.c @@ -20,7 +20,7 @@ const struct file_operations reiserfs_dir_operations = { .read = generic_read_dir, .readdir = reiserfs_readdir, .fsync = reiserfs_dir_fsync, - .ioctl = reiserfs_ioctl, + .unlocked_ioctl = reiserfs_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = reiserfs_compat_ioctl, #endif diff --git a/fs/reiserfs/file.c b/fs/reiserfs/file.c index 9f436668b7f..da2dba082e2 100644 --- a/fs/reiserfs/file.c +++ b/fs/reiserfs/file.c @@ -284,7 +284,7 @@ static ssize_t reiserfs_file_write(struct file *file, /* the file we are going t const struct file_operations reiserfs_file_operations = { .read = do_sync_read, .write = reiserfs_file_write, - .ioctl = reiserfs_ioctl, + .unlocked_ioctl = reiserfs_ioctl, #ifdef CONFIG_COMPAT .compat_ioctl = reiserfs_compat_ioctl, #endif diff --git a/fs/reiserfs/ioctl.c b/fs/reiserfs/ioctl.c index e30e8be0917..ace77451ceb 100644 --- a/fs/reiserfs/ioctl.c +++ b/fs/reiserfs/ioctl.c @@ -20,9 +20,9 @@ * 2) REISERFS_IOC_[GS]ETFLAGS, REISERFS_IOC_[GS]ETVERSION * 3) That's all for a while ... */ -int reiserfs_ioctl(struct inode *inode, struct file *filp, unsigned int cmd, - unsigned long arg) +long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg) { + struct inode *inode = filp->f_path.dentry->d_inode; unsigned int flags; int err = 0; @@ -132,9 +132,6 @@ setversion_out: long reiserfs_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) { - struct inode *inode = file->f_path.dentry->d_inode; - int ret; - /* These are just misnamed, they actually get/put from/to user an int */ switch (cmd) { case REISERFS_IOC32_UNPACK: @@ -156,9 +153,7 @@ long reiserfs_compat_ioctl(struct file *file, unsigned int cmd, return -ENOIOCTLCMD; } - ret = reiserfs_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg)); - - return ret; + return reiserfs_ioctl(file, cmd, (unsigned long) compat_ptr(arg)); } #endif diff --git a/include/linux/reiserfs_fs.h b/include/linux/reiserfs_fs.h index a498d9266d8..a05b4a20768 100644 --- a/include/linux/reiserfs_fs.h +++ b/include/linux/reiserfs_fs.h @@ -2314,8 +2314,7 @@ __u32 r5_hash(const signed char *msg, int len); #define SPARE_SPACE 500 /* prototypes from ioctl.c */ -int reiserfs_ioctl(struct inode *inode, struct file *filp, - unsigned int cmd, unsigned long arg); +long reiserfs_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); long reiserfs_compat_ioctl(struct file *filp, unsigned int cmd, unsigned long arg); int reiserfs_unpack(struct inode *inode, struct file *filp); -- cgit v1.2.3-70-g09d2