- 25 Sep, 2008 40 commits
-
-
Chris Mason authored
The current code waits for the count of async bio submits to get below a given threshold if it is too high right after adding the latest bio to the work queue. This isn't optimal because the caller may have sequential adjacent bios pending they are waiting to send down the pipe. This changeset requires the caller to wait on the async bio count, and changes the async checksumming submits to wait for async bios any time they self throttle. The end result is much higher sequential throughput. Signed-off-by:Chris Mason <chris.mason@oracle.com>
-
Chris Mason authored
Signed-off-by: -
David Woodhouse authored
Date: Tue, 19 Aug 2008 22:33:04 +0100 Signed-off-by:
David Woodhouse <David.Woodhouse@intel.com> Signed-off-by:
-
David Woodhouse authored
Date: Tue, 19 Aug 2008 22:20:17 +0100 btrfs_lookup_fs_root() only finds subvol roots which have already been seen and put into the cache. For btrfs_get_dentry() we actually have to go to the medium -- so use btrfs_read_fs_root_no_name() instead. In btrfs_get_parent(), notice when we've hit the root of the subvolume and return the real root instead. Signed-off-by:
-
David Woodhouse authored
Date: Tue, 19 Aug 2008 19:21:57 +0100 Using a 64-bit hash as the readdir cookie is just asking for trouble. And gets it, when we try to export the file system by NFS. Signed-off-by:
-
David Woodhouse authored
Date: Tue, 19 Aug 2008 16:49:35 +0100 This disappeared when I removed the special case for '.' in btrfs_lookup() Signed-off-by:
-
David Woodhouse authored
Date: Mon, 18 Aug 2008 22:50:22 +0100 Signed-off-by:
-
David Woodhouse authored
Date: Mon, 18 Aug 2008 13:10:20 +0100 This means that subvolumes get a different fsid, and NFS exporting them works properly. Signed-off-by:
-
David Woodhouse authored
Date: Mon, 18 Aug 2008 12:01:52 +0100 Signed-off-by:
-
David Woodhouse authored
Date: Sun, 17 Aug 2008 17:12:56 +0100 Signed-off-by:
-
David Woodhouse authored
Date: Sun, 17 Aug 2008 17:08:36 +0100 Signed-off-by:
-
David Woodhouse authored
Date: Sun, 17 Aug 2008 15:14:48 +0100 We never get asked by the VFS to lookup either of them, and we can handle the readdir() case a lot more simply, too. Signed-off-by:
-
David Woodhouse authored
Date: Tue, 12 Aug 2008 14:13:26 +0100 Signed-off-by:
-
David Woodhouse authored
Date: Wed, 6 Aug 2008 19:42:33 +0100 Signed-off-by:
-
Balaji Rao authored
Date: Mon, 21 Jul 2008 02:01:56 +0530 Here's an implementation of NFS support for btrfs. It relies on the fixes which are going in to 2.6.28 for the NFS readdir/lookup deadlock. This uses the btrfs_iget helper introduced previously. [dwmw2: Tidy up a little, switch to d_obtain_alias() w/compat routine, change fh_type, store parent's root object ID where needed, fix some get_parent() and fs_to_dentry() bugs] Signed-off-by:
Balaji Rao <balajirrao@gmail.com> Signed-off-by:
-
Balaji Rao authored
Date: Mon, 21 Jul 2008 02:01:04 +0530 This patch introduces a btrfs_iget helper to be used in NFS support. Signed-off-by:
-
Chris Mason authored
Before, the btrfs bdi congestion function was used to test for too many async bios. This keeps that check to throttle pdflush, but also adds a check while queuing bios. Signed-off-by: -
Chris Mason authored
This optimization had been removed because I thought it was triggering csum errors. The real cause of the errors was elsewhere, and so this optimization is back. Signed-off-by: -
Chris Mason authored
add_extent_mapping was allowing the insertion of overlapping extents. This never used to happen because it only inserted the extents from disk and those were never overlapping. But, with the data=ordered code, the disk and memory representations of the file are not the same. add_extent_mapping needs to ensure a new extent does not overlap before it inserts. Signed-off-by: -
David Woodhouse authored
These ended up freeing objects while they were still using them. Under guidance from Chris, just rip out the 'clever' bits and do things the simple way. Signed-off-by:
-
Chris Mason authored
This takes the csum mutex deeper in the call chain and releases it more often. Signed-off-by: -
Chris Mason authored
Before this change, btrfs would use a bdi congestion function to make sure there weren't too many pending async checksum work items. This change makes the process creating async work items wait instead, leading to fewer congestion returns from the bdi. This improves pdflush background_writeout scanning. Signed-off-by: -
Chris Mason authored
Signed-off-by: -
Chris Mason authored
After writing out all the remaining btree blocks in the transaction, the commit code would use filemap_fdatawait to make sure it was all on disk. This means it would wait for blocks written by other procs as well. The new code walks the list of blocks for this transaction again and waits only for those required by this transaction. Signed-off-by: -
Chris Mason authored
Signed-off-by: -
Chris Mason authored
Signed-off-by: -
Chris Mason authored
The writeback_index field is used by write_cache_pages to pick up where writeback on a given inode left off. But, it is never set to a sane value, so writeback can often start at a random offset in the file. Kernels 2.6.28 and higher will have this fixed, but for everyone else, we also fill in the value in btrfs. Signed-off-by: -
David Woodhouse authored
Add backwards compatibility in compat.h Signed-off-by:
-
Eric Sandeen authored
Newer RHEL5 kernels define both ClearPageFSMisc and ClearPageChecked, so test for both before redefining. Signed-off-by:
Eric Sandeen <sandeen@redhat.com> --- Signed-off-by:
-
Chris Mason authored
Signed-off-by: -
Chris Mason authored
Signed-off-by: -
Yan Zheng authored
Signed-off-by: -
Chris Mason authored
rename and link don't always have a lock on the source inode, and our use of a per-inode index variable was racy. This changes things to store the index in a local variable instead. Signed-off-by: -
Chris Mason authored
The multi-bio code is responsible for duplicating blocks in raid1 and single spindle duplication. It has counters to make sure all of the locations for a given extent are properly written before io completion is returned to the higher layers. But, it didn't always complete the same bio it was given, sometimes a clone was completed instead. This lead to problems with the async work queues because they saved a pointer to the bio in a struct off bi_private. The fix is to remember the original bio and only complete that one. Signed-off-by: -
Yan Zheng authored
This patch updates the file clone ioctl for the tree locking and new data ordered code. --- Signed-off-by: -
Yan Zheng authored
This trivial patch contains two locking fixes and a off by one fix. --- Signed-off-by: -
Chris Mason authored
Signed-off-by: -
Chris Mason authored
Signed-off-by: -
Chris Mason authored
Far from the perfect fix, but these structs are small. TODO for the next release. The block group cache structs are referenced in many different places, and it isn't safe to just free them while resizing. A real fix will be a larger change to the allocator so that it doesn't have to carry about the block group cache structs to find good places to search for free blocks. Signed-off-by: -
Sage Weil authored
Commit 597:466b27332893 (btrfs_start_transaction: wait for commits in progress) breaks the transaction start/stop ioctls by making btrfs_start_transaction conditionally wait for the next transaction to start. If an application artificially is holding a transaction open, things deadlock. This workaround maintains a count of open ioctl-initiated transactions in fs_info, and avoids wait_current_trans() if any are currently open (in start_transaction() and btrfs_throttle()). The start transaction ioctl uses a new btrfs_start_ioctl_transaction() that _does_ call wait_current_trans(), effectively pushing the join/wait decision to the outer ioctl-initiated transaction. This more or less neuters btrfs_throttle() when ioctl-initiated transactions are in use, but that seems like a pretty fundamental consequence of wrapping lots of write()'s in a transaction. Btrfs has no way to tell if the application considers a given operation as part of it's transaction. Obviously, if the transaction start/stop ioctls aren't being used, there is no effect on current behavior. Signed-off-by:
Sage Weil <sage@newdream.net> --- ctree.h | 1 + ioctl.c | 12 +++++++++++- transaction.c | 18 +++++++++++++----- transaction.h | 2 ++ 4 files changed, 27 insertions(+), 6 deletions(-) Signed-off-by:
-
