Rename the field commit_used in struct btrfs_block_group to last_used,
for clarity and consistency with the similar fields we're about to add.
It's not obvious that commit_flags means "flags as of the last commit"
rather than "flags related to a commit".
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is the following potential problem with the remap tree and delayed refs:
* Remapped extent freed in a delayed ref, which removes an entry from the
remap tree
* Remap tree now small enough to fit in a single leaf
* Corruption as we now have a level-0 block with a level-1 metadata item
in the extent tree
One solution to this would be to rework the remap tree code so that it operates
via delayed refs. But as we're hoping to remove cow-only metadata items in the
future anyway, change things so that the remap tree doesn't have any entries in
the extent tree. This also has the benefit of reducing write amplification.
We also make it so that the clear_cache mount option is a no-op, as with the
extent tree v2, as the free-space tree can no longer be recreated from the
extent tree.
Finally disable relocating the remap tree itself, which is added back in
a later patch. As it is we would get corruption as the traditional
relocation method walks the extent tree, and we're removing its metadata
items.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Signed-off-by: David Sterba <dsterba@suse.com>
No new allocations can be done from block groups that have the REMAPPED
flag set, so there's no value in their having entries in the free-space
tree.
Prevent a search through the free-space tree being scheduled for such a
block group, and prevent any additions to the in-memory free-space tree.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Signed-off-by: David Sterba <dsterba@suse.com>
When a chunk has been fully remapped, we are going to set its
num_stripes to 0, as it will no longer represent a physical location on
disk.
Change tree-checker to allow for this, and fix read_one_chunk() to avoid
a divide by zero.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add a new METADATA_REMAP chunk type, which is a metadata chunk that holds the
remap tree.
This is needed for bootstrapping purposes: the remap tree can't itself
be remapped, and must be relocated the existing way, by COWing every
leaf. The remap tree can't go in the SYSTEM chunk as space there is
limited, because a copy of the chunk item gets placed in the superblock.
The changes in fs/btrfs/volumes.h are because we're adding a new block
group type bit after the profile bits, and so can no longer rely on the
const_ilog2 trick.
The sizing to 32MB per chunk, matching the SYSTEM chunk, is an estimate
here, we can adjust it later if it proves to be too big or too small.
This works out to be ~500,000 remap items, which for a 4KB block size
covers ~2GB of remapped data in the worst case and ~500TB in the best case.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Add an incompat flag for the new remap-tree feature, and the constants
and definitions needed to support it.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Mark Harmstone <mark@harmstone.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We have currently three places that compute how much available space a
block group has. Add a helper function for this and use it in those
places.
Reviewed-by: Boris Burkov <boris@bur.io>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We don't expect to get errors unless we have a corrupted fs, bad RAM or a
bug. So tag the error handling as unlikely.
This slightly reduces the module's text size on x86_64 using gcc 14.2.0-19
from Debian.
Before this change:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1939458 172512 15592 2127562 2076ca fs/btrfs/btrfs.ko
After this change:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1939398 172512 15592 2127502 20768e fs/btrfs/btrfs.ko
Reviewed-by: Boris Burkov <boris@bur.io>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is no need for an else branch to deal with an unexpected delayed ref
type. We can just change the previous branch to deal with this by checking
if the ref type is not BTRFS_EXTENT_OWNER_REF_KEY, since that branch is
useless as it only sets 'ret' to zero when it's already zero. So merge the
two branches.
Reviewed-by: Boris Burkov <boris@bur.io>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There is no need to BUG(), we can just return an error and log an error
message.
Reviewed-by: Boris Burkov <boris@bur.io>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
btrfs_verify_dev_extents() iterates through the entire device tree
during mount to verify dev extents against chunks. Since this function
scans the whole tree, READA_FORWARD_ALWAYS is more appropriate than
READA_FORWARD.
While the device tree is typically small (a few hundred KB even for
multi-TB filesystems), using the correct readahead mode for full-tree
iteration is more consistent with the intended usage.
Signed-off-by: robbieko <robbieko@synology.com>
Signed-off-by: jinbaohong <jinbaohong@synology.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are two main causes of holes inside btrfs_device:
- The single bytes member of last_flush_error
Not only it's a single byte member, but we never really care about the
exact error number.
- The @devt member
Which is placed between two u64 members.
Shrink the size of btrfs_device by:
- Use a single bit flag for flush error
Use BTRFS_DEV_STATE_FLUSH_FAILED so that we no longer need that
dedicated member.
- Move @devt to the hole after dev_stat_values[]
This reduces the size of btrfs_device from 528 to exact 512 bytes for
x86_64.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Make the comment more detailed about why we need to flush delalloc and
wait for ordered extent completion before attempting to invalidate the
page cache.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The offload csum mode was introduced to allow developers to compare the
performance of generating checksum for data writes at different timings:
- During btrfs_submit_chunk()
This is the most common one, if any of the following condition is met
we go this path:
* The csum is fast
For now it's CRC32C and xxhash.
* It's a synchronous write
* Zoned
- Delay the checksum generation to a workqueue
However since commit dd57c78aec ("btrfs: introduce
btrfs_bio::async_csum") we no longer need to bother any of them.
As if it's an experimental build, async checksum generation at the
background will be faster anyway.
And if not an experimental build, we won't even have the offload csum
mode support.
Considering the async csum will be the new default, let's remove the
offload csum mode code.
There will be no impact to end users, and offload csum mode is still
under experimental features.
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are two tests in btrfs_fs_closing() but checking the
BTRFS_FS_CLOSING_DONE bit is done only in one place
load_extent_tree_free(). As this is an inline we can reduce size of the
generated code. The types can be also changed to bool as this becomes a
simple condition.
text data bss dec hex filename
1674006 146704 15560 1836270 1c04ee pre/btrfs.ko
1673772 146704 15560 1836036 1c0404 post/btrfs.ko
DELTA: -234
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently for an inode that needs compression, even if there is a delalloc
range that is single fs block sized and can not be inlined, we will
still go through the compression path.
Then inside compress_file_range(), we have one extra check to reject
single block sized range, and fall back to regular uncompressed write.
This rejection is in fact a little too late, we have already allocated
memory to async_chunk, delayed the submission, just to fallback to the
same uncompressed write.
Change the behavior to reject such cases earlier at
inode_need_compress(), so for such single block sized range we won't
even bother trying to go through compress path.
And since the inline small block check is inside inode_need_compress()
and compress_file_range() also calls that function, we no longer need a
dedicate check inside compress_file_range().
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The function add_block_group_free_space() was renamed
btrfs_add_block_group_free_space() by commit 6fc5ef7829 ("btrfs:
add btrfs prefix to free space tree exported functions"). Update
the comment accordingly.
Do some reorganization of the next few lines to keep the comment
within 80 characters.
Signed-off-by: Julia Lawall <Julia.Lawall@inria.fr>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[BUG]
Before btrfs-progs v6.16.1 release, mkfs.btrfs can leave free space tree
entries for deleted chunks:
# mkfs.btrfs -f -O fst $dev
# btrfs ins dump-tree -t chunk $dev
btrfs-progs v6.16
chunk tree
leaf 22036480 items 4 free space 15781 generation 8 owner CHUNK_TREE
leaf 22036480 flags 0x1(WRITTEN) backref revision 1
item 0 key (DEV_ITEMS DEV_ITEM 1) itemoff 16185 itemsize 98
item 1 key (FIRST_CHUNK_TREE CHUNK_ITEM 13631488) itemoff 16105 itemsize 80
^^^ The first chunk is at 13631488
item 2 key (FIRST_CHUNK_TREE CHUNK_ITEM 22020096) itemoff 15993 itemsize 112
item 3 key (FIRST_CHUNK_TREE CHUNK_ITEM 30408704) itemoff 15881 itemsize 112
# btrfs ins dump-tree -t free-space-tree $dev
btrfs-progs v6.16
free space tree key (FREE_SPACE_TREE ROOT_ITEM 0)
leaf 30556160 items 13 free space 15918 generation 8 owner FREE_SPACE_TREE
leaf 30556160 flags 0x1(WRITTEN) backref revision 1
item 0 key (1048576 FREE_SPACE_INFO 4194304) itemoff 16275 itemsize 8
free space info extent count 1 flags 0
item 1 key (1048576 FREE_SPACE_EXTENT 4194304) itemoff 16275 itemsize 0
free space extent
item 2 key (5242880 FREE_SPACE_INFO 8388608) itemoff 16267 itemsize 8
free space info extent count 1 flags 0
item 3 key (5242880 FREE_SPACE_EXTENT 8388608) itemoff 16267 itemsize 0
free space extent
^^^ Above 4 items are all before the first chunk.
item 4 key (13631488 FREE_SPACE_INFO 8388608) itemoff 16259 itemsize 8
free space info extent count 1 flags 0
item 5 key (13631488 FREE_SPACE_EXTENT 8388608) itemoff 16259 itemsize 0
free space extent
...
This can trigger btrfs check errors.
[CAUSE]
It's a bug in free space tree implementation of btrfs-progs, which
doesn't delete involved fst entries for the to-be-deleted chunk/block
group.
[ENHANCEMENT]
The mostly common fix is to clear the space cache and rebuild it, but
that requires a ro->rw remount which may not be possible for rootfs,
and also relies on users to use "clear_cache" mount option manually.
Here introduce a kernel fix for it, which will delete any entries that
is before the first block group automatically at the first RW mount.
For filesystems without such problem, the overhead is just a single tree
search and no modification to the free space tree, thus the overhead
should be minimal.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This function already dereferences 'inode' multiple times earlier,
making the additional NULL check at line 840 redundant since the
function would have crashed already if inode were NULL.
After commit 81cea6cd70 ("btrfs: remove btrfs_bio::fs_info by
extracting it from btrfs_bio::inode"), the btrfs_bio::inode field is
mandatory for all btrfs_bio allocations and is guaranteed to be
non-NULL.
Simplify the condition for allocating dummy checksums for zoned
NODATASUM data by removing the unnecessary 'inode &&' check.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Zhen Ni <zhen.ni@easystack.cn>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's no point in committing the transaction if we failed to insert the
balance item, since we haven't done anything else after we started/joined
the transaction. Also stop using two variables for tracking the return
value and use only 'ret'.
Reviewed-by: Daniel Vacek <neelx@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Instead of surrounding every caller of btrfs_is_shutdown() with unlikely,
move the unlikely into the helper itself, like we do in other places in
btrfs and is common in the kernel outside btrfs too. Also make the fs_info
argument of btrfs_is_shutdown() const.
On a x86_84 box using gcc 14.2.0-19 from Debian, this resulted in a slight
reduction of the module's text size.
Before:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1939044 172568 15592 2127204 207564 fs/btrfs/btrfs.ko
After:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1938876 172568 15592 2127036 2074bc fs/btrfs/btrfs.ko
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Errors are unexpected during the transaction commit path, and when they
happen we abort the transaction (by calling cleanup_transaction() under
the label 'cleanup_transaction' in btrfs_commit_transaction()). So mark
every error check in the transaction commit path as unlikely, to hint the
compiler so that it can possibly generate better code, and make it clear
for a reader about being unexpected.
On a x86_84 box using gcc 14.2.0-19 from Debian, this resulted in a slight
reduction of the module's text size.
Before:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1939476 172568 15592 2127636 207714 fs/btrfs/btrfs.ko
After:
$ size fs/btrfs/btrfs.ko
text data bss dec hex filename
1939044 172568 15592 2127204 207564 fs/btrfs/btrfs.ko
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The return statement after btrfs_backref_panic() is unreachable since
btrfs_backref_panic() calls BUG() which never returns. Remove the
return to unify it with the other calls to btrfs_backref_panic().
Signed-off-by: Zhen Ni <zhen.ni@easystack.cn>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently inside the main loop of cow_file_range(), we do the following
sequence:
- Reserve an extent
- Lock the IO tree range
- Create an IO extent map
- Create an ordered extent
Every step will need extra steps to do cleanup in the following order:
- Drop the newly created extent map
- Unlock extent range and cleanup the involved folios
- Free the reserved extent
However currently the error handling is done inconsistently:
- Extent map drop is handled in a dedicated tag
Out of the main loop, make it much harder to track.
- The extent unlock and folios cleanup is done separately
The extent is unlocked through btrfs_unlock_extent(), then
extent_clear_unlock_delalloc() again in a dedicated tag.
Meanwhile all other callsites (compression/encoded/nocow) all just
call extent_clear_unlock_delalloc() to handle unlock and folio clean
up in one go.
- Reserved extent freeing is handled in a dedicated tag
Out of the main loop, make it much harder to track.
- Error handling of btrfs_reloc_clone_csums() is relying out-of-loop
tags
This is due to the special requirement to finish ordered extents to
handle the metadata reserved space.
Enhance the error handling and align the behavior by:
- Introduce a dedicated cow_one_range() helper
Which do the reserve/lock/allocation in the helper.
And also handle the errors inside the helper.
No more dedicated tags out of the main loop.
- Use a single extent_clear_unlock_delalloc() to unlock and cleanup
folios
- Move the btrfs_reloc_clone_csums() error handling into the new helper
Thankfully it's not that complex compared to other cases.
And since we're here, also reduce the width of the following local
variables to u32:
- cur_alloc_size
- min_alloc_size
Each allocation won't go beyond 128M, thus u32 is more than enough.
- blocksize
The maximum is 64K, no need for u64.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Move space_info_flag_to_str() to space-info.h and as it now isn't static
to space-info.c any more prefix it with 'btrfs_'.
This way it can be re-used in other places.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's no need to call btrfs_handle_fs_error() as we are inside a
transaction and if we get an error we jump to the 'scrub_continue' label
and end up calling cleanup_transaction(), which aborts the transaction.
This is odd given that we have a transaction handle and that in the
transaction commit path any error makes us abort the transaction and
it's the only place that calls btrfs_handle_fs_error().
Remove the btrfs_handle_fs_error() call and replace it with an error
message so that if it happens we know what went wrong during the
transaction commit. Also annotate the condition in the if statement
with 'unlikely' since this is not expected to happen.
We've been wanting to remove btrfs_handle_fs_error(), so this removes
one user that does not even needs it.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's no need to call btrfs_handle_fs_error() as we are inside a
transaction and we propagate the error returned from
btrfs_write_and_wait_transaction() to the caller and it ends going up the
call chain to btrfs_commit_transaction() (returned by the call to
create_pending_snapshots()), where we jump to the 'unlock_reloc' label
and end up calling cleanup_transaction(), which aborts the transaction.
This is odd given that we have a transaction handle and that in the
transaction commit path any error makes us abort the transaction and,
besides another place inside btrfs_commit_transaction(), it's the only
place that calls btrfs_handle_fs_error().
Remove the btrfs_handle_fs_error() call and replace it with an error
message so that if it happens we know what went wrong during the
transaction commit. Also annotate the condition in the if statement
with 'unlikely' since this is not expected to happen.
We've been wanting to remove btrfs_handle_fs_error(), so this removes
one user that does not even need it.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In btrfs_find_orphan_roots() we don't need to call btrfs_handle_fs_error()
if we fail to delete the orphan item for the current root. This is because
we haven't done anything yet regarding the current root and previous
iterations of the loop dealt with other roots, so there's nothing we need
to undo. Instead log an error message and return the error to the caller,
which will result either in a mount failure or remount failure (the only
contexts it's called from).
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In btrfs_find_orphan_roots() we don't need to call btrfs_handle_fs_error()
if we fail to join a transaction. This is because we haven't done anything
yet regarding the current root and previous iterations of the loop dealt
with other roots, so there's nothing we need to undo. Instead log an error
message and return the error to the caller, which will result either in
a mount failure or remount failure (the only contexts it's called from).
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's no need to release the path in the if branch used when the root
does not exists since we released the path before the call to
btrfs_get_fs_root(). So remove that redundant btrfs_release_path() call.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We use both 'ret' and 'err' which is a pattern that generates confusion
and resulted in subtle bugs in the past. Remove 'err' and use only 'ret'.
Also move simplify the error flow by directly returning from the function
instead of breaking of the loop, since there are no resources to cleanup
after the loop.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's no point in committing the transaction if we failed to delete the
item, since we haven't done anything before. Also stop using two variables
for tracking the return value and use only 'ret'.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
We mention that the reserved data space is page size aligned but that's
not true anymore, as it's sector size aligned instead.
In commit 0bb067ca64 ("btrfs: fix the qgroup data free range for inline
data extents") we updated the amount passed to btrfs_qgroup_free_data()
from page size to sector size, but forgot to update the comment.
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There's no need for an on stack key to define the root's key as we have
already defined the key in the root itself. So remove the stack variable
and use the key in the root.
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Re-flow prepare allocation zoned to make it a bit more readable by
returning early and removing unnecessary indentations.
This patch does not change any functionality.
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This is done by:
- Shrink the size of btrfs_bio::mirror_num
From 32 bits unsigned int to u16.
Normally btrfs mirror number is either 0 (all profiles), 1 (all
profiles), 2 (DUP/RAID1/RAID10/RAID5), 3 (RAID1C3) or 4 (RAID1C4).
But for RAID6 the mirror number can go as large as the number of
devices of that chunk.
Currently the limit for number of devices for a data chunk is
BTRFS_MAX_DEVS(), which is around 500 for the default 16K nodesize.
And if going the max 64K nodesize, we can have a little over 2000
devices for a chunk.
Although I'd argue it's way overkilled, we don't reject such cases yet
thus u8 is not going to cut it, and have to use u16 (max out at 64K).
- Use bit fields for boolean members
Although it's not always safe for racy call sites, those members are
safe.
* csum_search_commit_root
* is_scrub
Those two are set immediately after bbio allocation and no more
writes after allocation, thus they are very safe.
* async_csum
* can_use_append
Those two are set for each split range, and after that there is no
writes into those two members in different threads, thus they are
also safe.
And there are spaces for 4 more bits before increasing the size of
btrfs_bio again, which should be future proof enough.
- Reorder the structure members
Now we always put the largest member first (after the huge 120 bytes
union), making it easier to fill any holes.
This reduce the size of btrfs_bio by 8 bytes, from 312 bytes to 304 bytes.
Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
The minimum gcc version is 8 since 118c40b7b5 ("kbuild: require
gcc-8 and binutils-2.30"), the workaround for missing __VA_OPT__ support
is not needed.
Signed-off-by: David Sterba <dsterba@suse.com>
There's code in _btrfs_printk() to parse the message level from the
input string so we can augment the message with the level description
for better visibility in the logs.
The parsing code has evolved over time, see commits:
- 40f7828b36 ("btrfs: better handle btrfs_printk() defaults")
- 262c5e86fe ("printk/btrfs: handle more message headers")
- 533574c6bc ("btrfs: use printk_get_level and printk_skip_level, add __printf, fix fallout")
- 4da3511342 ("btrfs: add varargs to btrfs_error")
As we are using the specific level helpers everywhere we can simply pass
the message level so we don't have to parse it. The proper printk()
message header is created as KERN_SOH + "level".
Signed-off-by: David Sterba <dsterba@suse.com>
The printk() can be compiled out depending on CONFIG_PRINTK, this is
reflected in our helpers. The indirection is provided by btrfs_printk()
used in the ratelimited and RCU wrapper macros.
Drop the btrfs_printk() helper and define the ratelimit and RCU helpers
directly when CONFIG_PRINTK is undefined. This will allow further
changes to the _btrfs_printk() interface (which is internal), any
message in other code should use the level-specific helpers.
Signed-off-by: David Sterba <dsterba@suse.com>
struct btrfs_block_group used to be called struct
btrfs_block_group_cache but got renamed to btrfs_block_group with
commit 32da5386d9 ("btrfs: rename btrfs_block_group_cache").
Rename btrfs_create_block_group_cache() to btrfs_create_block_group() to
reflect that change.
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
In many places we have pattern:
ret = ...;
return ret;
This can be simplified to a direct return, removing 'ret' if not
otherwise needed. The places in self tests are not converted so we can
add more test cases without changing surrounding code
(extent-map-tests.c:test_case_4()).
Signed-off-by: David Sterba <dsterba@suse.com>
In prepare_one_folio(), ret is initialized to 0 at declaration,
and in an error path we assign ret = 0 before jumping to the
again label to retry the operation. However, ret is immediately
overwritten by ret = set_folio_extent_mapped(folio) after the
again label.
Both assignments are never observed by any code path,
therefore they can be safely removed.
Signed-off-by: Massimiliano Pellizzer <mpellizzer.dev@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Inside extent_io.c, there are several simple call sites doing things
like:
for_each_set_bit(bit, bitmap, bitmap_size) {
/* handle one fs block */
}
The workload includes:
- set_bit()
Inside extent_writepage_io().
This can be replaced with a bitmap_set().
- btrfs_folio_set_lock()
- btrfs_mark_ordered_io_finished()
Inside writepage_delalloc().
Instead of calling it multiple times, we can pass a range into the
function with one call.
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Currently submit_one_sector() has only one failure path from
btrfs_get_extent().
However the error handling is split into two parts, one inside
submit_one_sector(), which clears the dirty flag and finishes the
writeback for the fs block.
The other part is to submit any remaining bio inside bio_ctrl and mark
the ordered extent finished for the fs block.
There is no special reason that we must split the error handling, let's
just concentrate all the error handling into submit_one_sector().
Reviewed-by: Boris Burkov <boris@bur.io>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
[CORNER CASE]
If we have the following file extents layout, btrfs_get_extent() can
return a smaller hole during read, and cause unnecessary extra tree
searches:
item 6 key (257 EXTENT_DATA 0) itemoff 15810 itemsize 53
generation 9 type 1 (regular)
extent data disk byte 13631488 nr 4096
extent data offset 0 nr 4096 ram 4096
extent compression 0 (none)
item 7 key (257 EXTENT_DATA 32768) itemoff 15757 itemsize 53
generation 9 type 1 (regular)
extent data disk byte 13635584 nr 4096
extent data offset 0 nr 4096 ram 4096
extent compression 0 (none)
In above case, range [0, 4K) and [32K, 36K) are regular extents, and
there is a hole in range [4K, 32K), and the fs has "no-holes" feature,
meaning the hole will not have a file extent item.
[INEFFICIENCY]
Assume the system has 4K page size, and we're doing readahead for range
[4K, 32K), no large folio yet.
btrfs_readahead() for range [4K, 32K)
|- btrfs_do_readpage() for folio 4K
| |- get_extent_map() for range [4K, 8K)
| |- btrfs_get_extent() for range [4K, 8K)
| We hit item 6, then for the next item 7.
| At this stage we know range [4K, 32K) is a hole.
| But our search range is only [4K, 8K), not reaching 32K, thus
| we go into not_found: tag, returning a hole em for [4K, 8K).
|
|- btrfs_do_readpage() for folio 8K
| |- get_extent_map() for range [8K, 12K)
| |- btrfs_get_extent() for range [8K, 12K)
| We hit the same item 6, and then item 7.
| But still we goto not_found tag, inserting a new hole em,
| which will be merged with previous one.
|
| [ Repeat the same btrfs_get_extent() calls until the end ]
So we're calling btrfs_get_extent() again and again, just for a
different part of the same hole range [4K, 32K).
[ENHANCEMENT]
Make btrfs_do_readpage() to search for a larger extent map if readahead
is involved.
For btrfs_readahead() we have bio_ctrl::ractl set, and lock extents for
the whole readahead range.
If we find bio_ctrl::ractl is set, we can use that end range as extent
map search end, this allows btrfs_get_extent() to return a much larger
hole, thus reduce the need to call btrfs_get_extent() again and again.
btrfs_readahead() for range [4K, 32K)
|- btrfs_do_readpage() for folio 4K
| |- get_extent_map() for range [4K, 32K)
| |- btrfs_get_extent() for range [4K, 32K)
| We hit item 6, then for the next item 7.
| At this stage we know range [4K, 32K) is a hole.
| So the hole em for range [4K, 32K) is returned.
|
|- btrfs_do_readpage() for folio 8K
| |- get_extent_map() for range [8K, 32K)
| The cached hole em range [4K, 32K) covers the range,
| and reuse that em.
|
| [ Repeat the same btrfs_get_extent() calls until the end ]
Now we only call btrfs_get_extent() once for the whole range [4K, 32K),
other than the old 8 times.
Such change will reduce the overhead of reading large holes a little.
For current experimental build (with larger folios) on aarch64, there
will be a tiny but consistent ~1% improvement reading a large hole file:
Reading a 1GiB sparse file (all hole) using xfs_io, with 64K block
size, the result is the time needed to read the whole file, reported
from xfs_io.
32 runs, experimental build (with large folios).
64K page size, 4K fs block size.
- Avg before: 0.20823 s
- Avg after: 0.20635 s
- Diff: -0.9%
Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
There are already several bugs with on-stack btrfs_path involved, even
it is already a little safer than btrfs_path pointers (only leaks the
extent buffers, not the btrfs_path structure itself)
- Patch "btrfs: make sure extent and csum paths are always released in
scrub_raid56_parity_stripe()"
- Patch "btrfs: fix a potential path leak in print_data_reloc_error()"
Thus there is a real need to apply auto release for those on-stack paths.
Introduces a new macro, BTRFS_PATH_AUTO_RELEASE() which defines one
on-stack btrfs_path structure, initialize it all to 0, then call
btrfs_release_path() on it when exiting the scope.
This applies to current 3 on-stack path usages:
- defrag_get_extent() in defrag.c
- print_data_reloc_error() in inode.c
There is a special case where we want to release the path early before
the time consuming iterate_extent_inodes() call, thus that manual
early release is kept as is, with an extra comment added.
- scrub_radi56_parity_stripe() in scrub.c
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Previously direct IO was disabled if the fs block size was larger than
the page size, the reasons are:
- Iomap direct IO can split the range ignoring the fs block alignment
Which could trigger the bio size check from btrfs_submit_bio().
- The buffer is only ensured to be contiguous in user space memory
The underlying physical memory is not ensured to be contiguous, and
that can cause problems for the checksum generation/verification and
RAID56 handling.
However the above problems are solved by the following upstream commits:
- 001397f5ef ("iomap: add IOMAP_DIO_FSBLOCK_ALIGNED flag")
Which added an extra flag that can be utilized by the fs to ensure
the bio submitted by iomap is always aligned to fs block size.
- ec20799064 ("btrfs: enable encoded read/write/send for bs > ps cases")
- 8870dbeedc ("btrfs: raid56: enable bs > ps support")
Which makes btrfs to handle bios that are not backed by large folios
but still are aligned to fs block size.
As the commits have been merged we can enable direct IO support for
bs > ps cases.
Reviewed-by: Neal Gompa <neal@gompa.dev>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
Make btrfs use the library APIs instead of crypto_shash, for all
checksum computations. This has many benefits:
- Allows future checksum types, e.g. XXH3 or CRC64, to be more easily
supported. Only a library API will be needed, not crypto_shash too.
- Eliminates the overhead of the generic crypto layer, including an
indirect call for every function call and other API overhead. A
microbenchmark of btrfs_check_read_bio() with crc32c checksums shows a
speedup from 658 cycles to 608 cycles per 4096-byte block.
- Decreases the stack usage of btrfs by reducing the size of checksum
contexts from 384 bytes to 240 bytes, and by eliminating the need for
some functions to declare a checksum context at all.
- Increases reliability. The library functions always succeed and
return void. In contrast, crypto_shash can fail and return errors.
Also, the library functions are guaranteed to be available when btrfs
is loaded; there's no longer any need to use module softdeps to try to
work around the crypto modules sometimes not being loaded.
- Fixes a bug where blake2b checksums didn't work on kernels booted with
fips=1. Since btrfs checksums are for integrity only, it's fine for
them to use non-FIPS-approved algorithms.
Note that with having to handle 4 algorithms instead of just 1-2, this
commit does result in a slightly positive diffstat. That being said,
this wouldn't have been the case if btrfs had actually checked for
errors from crypto_shash, which technically it should have been doing.
Reviewed-by: Ard Biesheuvel <ardb@kernel.org>
Reviewed-by: Neal Gompa <neal@gompa.dev>
Signed-off-by: Eric Biggers <ebiggers@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
