linux.git/fs/block_dev.c, branch v2.6.20

[PATCH] revert blockdev direct io back to 2.6.19 version

2007-02-03T19:26:06+00:00

Andrew Vasquez is reporting as-iosched oopses and a 65% throughput
slowdown due to the recent special-casing of direct-io against
blockdevs.  We don't know why either of these things are occurring.

The patch minimally reverts us back to the 2.6.19 code for a 2.6.20
release.

Cc: Andrew Vasquez 
Cc: Ken Chen 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] fix blk_direct_IO bio preparation

2007-01-23T15:52:06+00:00

For large size DIO that needs multiple bio, one full page worth of data was
lost at the boundary of bio's maximum sector or segment limits.  After a
bio is full and got submitted.  The outer while (nbytes) { ...  } loop will
allocate a new bio and just march on to index into next page.  It just
forgets about the page that bio_add_page() rejected when previous bio is
full.  Fix it by put the rejected page back to pvec so we pick it up again
for the next bio.

Signed-off-by: Ken Chen 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] blockdev direct_io: fix signedness bug

2007-01-23T15:52:05+00:00

size_t is unsigned.  IO errors aren't getting through.

Cc: "Chen, Kenneth W" 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] Revert bd_mount_mutex back to a semaphore

2007-01-12T02:18:21+00:00

Revert bd_mount_mutex back to a semaphore so that xfs_freeze -f /mnt/newtest;
xfs_freeze -u /mnt/newtest works safely and doesn't produce lockdep warnings.

(XFS unlocks the semaphore from a different task, by design.  The mutex
code warns about this)

Signed-off-by: Dave Chinner 
Cc: Ingo Molnar 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] optimize o_direct on block devices

2006-12-13T17:05:50+00:00

Implement block device specific .direct_IO method instead of going through
generic direct_io_worker for block device.

direct_io_worker() is fairly complex because it needs to handle O_DIRECT on
file system, where it needs to perform block allocation, hole detection,
extents file on write, and tons of other corner cases.  The end result is
that it takes tons of CPU time to submit an I/O.

For block device, the block allocation is much simpler and a tight triple
loop can be written to iterate each iovec and each page within the iovec in
order to construct/prepare bio structure and then subsequently submit it to
the block layer.  This significantly speeds up O_D on block device.

[akpm@osdl.org: small speedup]
Signed-off-by: Ken Chen 
Cc: Christoph Hellwig 
Cc: Zach Brown 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] VFS: change struct file to use struct path

2006-12-08T16:28:41+00:00

This patch changes struct file to use struct path instead of having
independent pointers to struct dentry and struct vfsmount, and converts all
users of f_{dentry,vfsmnt} in fs/ to use f_path.{dentry,mnt}.

Additionally, it adds two #define's to make the transition easier for users of
the f_dentry and f_vfsmnt.

Signed-off-by: Josef "Jeff" Sipek 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] bdev: fix ->bd_part_count leak

2006-12-08T16:28:39+00:00

Don't leak a ->bd_part_count when the partition open fails with -ENXIO.

Signed-off-by: Peter Zijlstra 
Acked-by: Neil Brown 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] lockdep: use mutex_lock_nested for bd_mutex to avoid lockdep warning

2006-12-08T16:28:39+00:00

Now that the nesting in blkdev_{get,put} is simpler, adding mutex_lock_nested
is trivial.

Cc: Ingo Molnar 
Acked-by: Peter Zijlstra 
Signed-off-by: Neil Brown 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] lockdep: simplify some aspects of bd_mutex nesting

2006-12-08T16:28:39+00:00

When we open (actually blkdev_get) a partition we need to also open (get) the
whole device that holds the partition.  The involves some limited recursion.
This patch tries to simplify some aspects of this.

As well as opening the whole device, we need to increment ->bd_part_count when
a partition is opened (this is used by rescan_partitions to avoid a rescan if
any partition is active, as that would be confusing).

The main change this patch makes is to move the inc/dec of bd_part_count into
blkdev_{get,put} for the whole rather than doing it in blkdev_{get,put} for
the partition.

More specifically, we introduce __blkdev_get and __blkdev_put which do exactly
what blkdev_{get,put} did, only with an extra "for_part" argument
(blkget_{get,put} then call the __ version with a '0' for the extra argument).

If for_part is 1, then the blkdev is being get(put) because a partition is
being opened(closed) for the first(last) time, and so bd_part_count should be
updated (on success).  The particular advantage of pushing this function down
is that the bd_mutex lock (which is needed to update bd_part_count) is already
held at the lower level.

Note that this slightly changes the semantics of bd_part_count.  Instead of
updating it whenever a partition is opened or released, it is now only updated
on the first open or last release.  This is an adequate semantic as it is only
ever tested for "== 0".

Having introduced these functions we remove the current bd_part_count updates
from do_open (which is really the body of blkdev_get) and call
__blkdev_get(...  1).  Similarly in blkget_put we remove the old bd_part_count
updates and call __blkget_put(..., 1).  This call is moved to the end of
__blkdev_put to avoid nested locks of bd_mutex.

Finally the mutex_lock on whole->bd_mutex in do_open can be removed.  It was
only really needed to protect bd_part_count, and that is now managed (and
protected) within the recursive call.

The observation that bd_part_count is central to the locking issues, and the
modifications to create __blkdev_put are from Peter Zijlstra.

Cc: Ingo Molnar 
Acked-by: Peter Zijlstra 
Signed-off-by: Neil Brown 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds

[PATCH] lockdep: remove lock_key approach to managing nested bd_mutex locks

2006-12-08T16:28:38+00:00

The extra call to get_gendisk is not good.  It causes a ->probe and possible
module load before it is really appropriate to do this.

Cc: Ingo Molnar 
Acked-by: Peter Zijlstra 
Signed-off-by: Neil Brown 
Signed-off-by: Andrew Morton 
Signed-off-by: Linus Torvalds