Large streaming reads make for large bios, which means each entry on the
list async work queues represents a large amount of data. IO
congestion throttling on the device was kicking in before the async
worker threads decided a single thread was busy and needed some help.
The end result was that a streaming read would result in a single CPU
running at 100% instead of balancing the work off to other CPUs.
This patch also changes the pre-IO checksum lookup done by reads to
work on a per-bio basis instead of a per-page. This results in many
extra btree lookups on large streaming reads. Doing the checksum lookup
right before bio submit allows us to reuse searches while processing
adjacent offsets.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The memory reclaiming issue happens when snapshot exists. In that
case, some cache entries may not be used during old snapshot dropping,
so they will remain in the cache until umount.
The patch adds a field to struct btrfs_leaf_ref to record create time. Besides,
the patch makes all dead roots of a given snapshot linked together in order of
create time. After a old snapshot was completely dropped, we check the dead
root list and remove all cache entries created before the oldest dead root in
the list.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
It was incorrectly clearing the up to date flag on the buffer even
when the buffer properly verified.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
A large reference cache is directly related to a lot of work pending
for the cleaner thread. This throttles back new operations based on
the size of the reference cache so the cleaner thread will be able to keep
up.
Overall, this actually makes the FS faster because the cleaner thread will
be more likely to find things in cache.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This changes the reference cache to make a single cache per root
instead of one cache per transaction, and to key by the byte number
of the disk block instead of the keys inside.
This makes it much less likely to have cache misses if a snapshot
or something has an extra reference on a higher node or a leaf while
the first transaction that added the leaf into the cache is dropping.
Some throttling is added to functions that free blocks heavily so they
wait for old transactions to drop.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Much of the IO done while dropping snapshots is done looking up
leaves in the filesystem trees to see if they point to any extents and
to drop the references on any extents found.
This creates a cache so that IO isn't required.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Before setting an extent to delalloc, the code needs to wait for
pending ordered extents.
Also, the relocation code needs to wait for ordered IO before scanning
the block group again. This is because the extents are not removed
until the IO for the new extents is finished
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Checksum items are not inserted into the tree until all of the io from a
given extent is complete. This means one dirty page from an extent may
be written, freed, and then read again before the entire extent is on disk
and the checksum item is inserted.
The checksums themselves are stored in the ordered extent so they can
be inserted in bulk when IO is complete. On read, if a checksum item isn't
found, the ordered extents were being searched for a checksum record.
This all worked most of the time, but the checksum insertion code tries
to reduce the number of tree operations by pre-inserting checksum items
based on i_size and a few other factors. This means the read code might
find a checksum item that hasn't yet really been filled in.
This commit changes things to check the ordered extents first and only
dive into the btree if nothing was found. This removes the need for
extra locking and is more reliable.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Before, extent buffers were a temporary object, meant to map a number of pages
at once and collect operations on them.
But, a few extra fields have crept in, and they are also the best place to
store a per-tree block lock field as well. This commit puts the extent
buffers into an rbtree, and ensures a single extent buffer for each
tree block.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
btrfs_commit_transaction has to loop waiting for any writers in the
transaction to finish before it can proceed. btrfs_start_transaction
should be polite and not join a transaction that is in the process
of being finished off.
There are a few places that can't wait, basically the ones doing IO that
might be needed to finish the transaction. For them, btrfs_join_transaction
is added.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Higher layers sometimes call set_page_dirty without asking the filesystem
to help. This causes many problems for the data=ordered and cow code.
This commit detects pages that haven't been properly setup for IO and
kicks off an async helper to deal with them.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The old data=ordered code would force commit to wait until
all the data extents from the transaction were fully on disk. This
introduced large latencies into the commit and stalled new writers
in the transaction for a long time.
The new code changes the way data allocations and extents work:
* When delayed allocation is filled, data extents are reserved, and
the extent bit EXTENT_ORDERED is set on the entire range of the extent.
A struct btrfs_ordered_extent is allocated an inserted into a per-inode
rbtree to track the pending extents.
* As each page is written EXTENT_ORDERED is cleared on the bytes corresponding
to that page.
* When all of the bytes corresponding to a single struct btrfs_ordered_extent
are written, The previously reserved extent is inserted into the FS
btree and into the extent allocation trees. The checksums for the file
data are also updated.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The btree defragger wasn't making forward progress because the new key wasn't
being saved by the btrfs_search_forward function.
This also disables the automatic btree defrag, it wasn't scaling well to
huge filesystems. The auto-defrag needs to be done differently.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This creates one kthread for commits and one kthread for
deleting old snapshots. All the work queues are removed.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The existing throttle mechanism was often not sufficient to prevent
new writers from coming in and making a given transaction run forever.
This adds an explicit wait at the end of most operations so they will
allow the current transaction to close.
There is no wait inside file_write, inode updates, or cow filling, all which
have different deadlock possibilities.
This is a temporary measure until better asynchronous commit support is
added. This code leads to stalls as it waits for data=ordered
writeback, and it really needs to be fixed.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
One lock per btree block can make for significant congestion if everyone
has to wait for IO at the high levels of the btree. This drops
locks held by a path when doing reads during a tree search.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Extent alloctions are still protected by a large alloc_mutex.
Objectid allocations are covered by a objectid mutex
Other btree operations are protected by a lock on individual btree nodes
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The allocation trees and the chunk trees are serialized via their own
dedicated mutexes. This means allocation location is still not very
fine grained.
The main FS btree is protected by locks on each block in the btree. Locks
are taken top / down, and as processing finishes on a given level of the
tree, the lock is released after locking the lower level.
The end result of a search is now a path where only the lowest level
is locked. Releasing or freeing the path drops any locks held.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
If a bio submission is after a lock holder waiting for the bio
on the work queue, it is possible to deadlock. Move the bios
into their own pool.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
mount -o thread_pool_size changes the default, which is
min(num_cpus + 2, 8). Larger thread pools would make more sense on
very large disk arrays.
This mount option controls the max size of each thread pool. There
are multiple thread pools, so the total worker count will be larger
than the mount option.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Btrfs has been using workqueues to spread the checksumming load across
other CPUs in the system. But, workqueues only schedule work on the
same CPU that queued the work, giving them a limited benefit for systems with
higher CPU counts.
This code adds a generic facility to schedule work with pools of kthreads,
and changes the bio submission code to queue bios up. The queueing is
important to make sure large numbers of procs on the system don't
turn streaming workloads into random workloads by sending IO down
concurrently.
The end result of all of this is much higher performance (and CPU usage) when
doing checksumming on large machines. Two worker pools are created,
one for writes and one for endio processing. The two could deadlock if
we tried to service both from a single pool.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Also adds lots of comments to describe what's going on here.
Signed-off-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This avoids IO stalls and poorly ordered IO from inline writers mixing in
with the async submission queue
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The async submit workqueue was absorbing too many requests, leading to long
stalls where the async submitters were stalling.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Devices can change after the scan ioctls are done, and btrfs_open_devices
needs to be able to verify them as they are opened and used by the FS.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
When duplicate copies exist, writes are allowed to fail to one of those
copies. This changeset includes a few changes that allow the FS to
continue even when some IOs fail.
It also adds verification of the parent generation number for btree blocks.
This generation is stored in the pointer to a block, and it ensures
that missed writes to are detected.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This required a few structural changes to the code that manages bdev pointers:
The VFS super block now gets an anon-bdev instead of a pointer to the
lowest bdev. This allows us to avoid swapping the super block bdev pointer
around at run time.
The code to read in the super block no longer goes through the extent
buffer interface. Things got ugly keeping the mapping constant.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
2.6.18 seems to get caught in an infinite loop when
cancel_rearming_delayed_workqueue is called more than once, so this switches
to cancel_delayed_work, which is arguably more correct.
Also, balance_dirty_pages can run into problems with 2.6.18 based kernels
because it doesn't have the per-bdi dirty limits. This avoids calling
balance_dirty_pages on the btree inode unless there is actually something
to balance, which is a good optimization in general.
Finally there's a compile fix for ordered-data.h
Signed-off-by: Chris Mason <chris.mason@oracle.com>
btrfs_invalidatepage is not allowed to leave pages around on the lru.
Any such pages will trigger an oops later on because the VM will see
page->private and assume it is a buffer head.
This also forces extra flushes of the async work queues before
dropping all the pages on the btree inode during unmount. Left over
items on the work queues are one possible cause of busy state ranges
during truncate_inode_pages.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The btree inode should only have a single extent_map in the cache,
it doesn't make sense to ever drop it.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This allows checksumming to happen in parallel among many cpus, and
keeps us from bogging down pdflush with the checksumming code.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Block headers now store the chunk tree uuid
Chunk items records the device uuid for each stripes
Device extent items record better back refs to the chunk tree
Block groups record better back refs to the chunk tree
The chunk tree format has also changed. The objectid of BTRFS_CHUNK_ITEM_KEY
used to be the logical offset of the chunk. Now it is a chunk tree id,
with the logical offset being stored in the offset field of the key.
This allows a single chunk tree to record multiple logical address spaces,
upping the number of bytes indexed by a chunk tree from 2^64 to
2^128.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This includes fixing a missing spinlock init call that caused oops on mount
for most kernels other than 2.6.25.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Fix for a endianess BUG when using btrfs v0.13 with kernels older than 2.6.23
Problem:
Has of v0.13, btrfs-progs is using crc32c.c equivalent to the one found on
linux-2.6.23/lib/libcrc32c.c Since crc32c_le() changed in linux-2.6.23, when
running btrfs v0.13 with older kernels we have a missmatch between the versions
of crc32c_le() from btrfs-progs and libcrc32c in the kernel. This missmatch
causes a bug when using btrfs on big endian machines.
Solution:
btrfs_crc32c() macro that when compiling for kernels older than 2.6.23, does
endianess conversion to parameters and return value of crc32c().
This endianess conversion nullifies the differences in implementation
of crc32c_le().
If kernel 2.6.23 or better, it calls crc32c().
Signed-off-by: Miguel Sousa Filipe <miguel.filipe@gmail.com>
---
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Before it was done by the bio end_io routine, the work queue code is able
to scale much better with faster IO subsystems.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Before, metadata checksumming was done by the callers of read_tree_block,
which would set EXTENT_CSUM bits in the extent tree to show that a given
range of pages was already checksummed and didn't need to be verified
again.
But, those bits could go away via try_to_releasepage, and the end
result was bogus checksum failures on pages that never left the cache.
The new code validates checksums when the page is read. It is a little
tricky because metadata blocks can span pages and a single read may
end up going via multiple bios.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Checksums were only verified by btrfs_read_tree_block, which meant the
functions to probe the page cache for blocks were not validating checksums.
Normally this is fine because the buffers will only be in cache if they
have already been validated.
But, there is a window while the buffer is being read from disk where
it could be up to date in the cache but not yet verified. This patch
makes sure all buffers go through checksum verification before they
are used.
This is safer, and it prevents modification of buffers before they go
through the csum code.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This allows detection of blocks that have already been written in the
running transaction so they can be recowed instead of modified again.
It is step one in trusting the transid field of the block pointers.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Reduce CPU time searching for free blocks by optimizing find_first_extent_bit
Fix find_free_extent to make better use of the last_alloc hint. Before it
was often finding blocks just before the hint.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The end_bio routines are changed to take a pointer to the extent state
struct, and the state tree is walked in order to set/clear appropriate
bits as IO completes. This greatly reduces the number of rbtree searches
done by the end_bio handlers, and reduces lock contention.
The extent_io releasepage function is changed to avoid expensive searches
for locked state.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
There is now extent_map for mapping offsets in the file to disk and
extent_io for state tracking, IO submission and extent_bufers.
The new extent_map code shifts from [start,end] pairs to [start,len], and
pushes the locking out into the caller. This allows a few performance
optimizations and is easier to use.
A number of extent_map usage bugs were fixed, mostly with failing
to remove extent_map entries when changing the file.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This forces file data extents down the disk along with the metadata that
references them. The current implementation is fairly simple, and just
writes out all of the dirty pages in an inode before the commit.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This is intended to prevent accidentally filling the drive. A determined
user can still make things oops.
It includes some accounting of the current bytes under delayed allocation,
but this will change as things get optimized
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This patch adds a helper function 'update_pinned_extents' to
extent-tree.c. The usage of the helper function is similar to
'update_block_group', the last parameter of the function indicates
pin vs unpin.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
btrfs_btree_balance_dirty is changed to pass the number of pages dirtied
for more accurate dirty throttling. This lets the VM make better decisions
about when to force some writeback.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Dead roots are trees left over after a crash, and they were either in the
process of being removed or were waiting to be removed when the box crashed.
Before, a search of the entire tree of root pointers was done on mount
looking for dead roots. Now, the search is done the first time we load
a root.
This makes mount faster when there are a large number of snapshots, and it
enables the block accounting code to properly update the block counts on
the latest root as old versions of the root are reaped after a crash.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
File data checksums are only done during writepage, so we have to make sure
all pages are written when the snapshot is taken. This also adds some
locking so that new writes don't race in and add new dirty pages.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
This adds two types of btree defrag, a run time form that tries to
defrag recently allocated blocks in the btree when they are still in ram,
and an ioctl that forces defrag of all btree blocks.
File data blocks are not defragged yet, but this can make a huge difference
in sequential btree reads.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Almost none of the files including module.h need to do so,
remove them.
Include sched.h in extent-tree.c to silence a warning about cond_resched()
being undeclared.
Signed-off-by: Zach Brown <zach.brown@oracle.com>
Signed-off-by: Chris Mason <chris.mason@oracle.com>
The super block written during commit was not consistent with the state of
the trees. This change adds an in-memory copy of the super so that we can
make sure to write out consistent data during a commit.
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Remove implicit commit in del_item and insert_item
Add implicit commit to close()
Add commit op in random-test
Signed-off-by: Chris Mason <chris.mason@oracle.com>
Chris Mason