In fixing the readdir+pagefault deadlock I accidentally introduced a
stale entry regression in readdir.  If we get close to full for the
temporary buffer, and then skip a few delayed deletions, and then try to
add another entry that won't fit, we will emit the entries we found and
retry.  Unfortunately we delete entries from our del_list as we find
them, assuming we won't need them.  However our pos will be with
whatever our last entry was, which could be before the delayed deletions
we skipped, so the next search will add the deleted entries back into
our readdir buffer.  So instead don't delete entries we find in our
del_list so we can make sure we always find our delayed deletions.  This
is a slight perf hit for readdir with lots of pending deletions, but
hopefully this isn't a common occurrence.  If it is we can revist this
and optimize it.

cc: stable@vger.kernel.org
Fixes: 23b5ec74

 ("btrfs: fix readdir deadlock with pagefault")
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

refcounts have a generic implementation and an asm optimized one.  The
generic version has extra debugging to make sure that once a refcount
goes to zero, refcount_inc won't increase it.

The btrfs delayed inode code wasn't expecting this, and we're tripping
over the warnings when the generic refcounts are used.  We ended up with
this race:

Process A                                         Process B
                                                  btrfs_get_delayed_node()
						  spin_lock(root->inode_lock)
						  radix_tree_lookup()
__btrfs_release_delayed_node()
refcount_dec_and_test(&delayed_node->refs)
our refcount is now zero
						  refcount_add(2) <---
						  warning here, refcount
                                                  unchanged

spin_lock(root->inode_lock)
radix_tree_delete()

With the generic refcounts, we actually warn again when process B above
tries to release his refcount because refcount_add() turned into a
no-op.

We saw this in production on older kernels without the asm optimized
refcounts.

The fix used here is to use refcount_inc_not_zero() to detect when the
object is in the middle of being freed and return NULL.  This is almost
always the right answer anyway, since we usually end up pitching the
delayed_node if it didn't have fresh data in it.

This also changes __btrfs_release_delayed_node() to remove the extra
check for zero refcounts before radix tree deletion.
btrfs_get_delayed_node() was the only path that was allowing refcounts
to go from zero to one.

Fixes: 6de5f18e

 ("btrfs: fix refcount_t usage when deleting btrfs_delayed_node")
CC: <stable@vger.kernel.org> # 4.12+
Signed-off-by: Chris Mason <clm@fb.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Commit e0ae9994 ("btrfs: preallocate device flush bio") reworked
the way the flush bio is allocated and used. Concretely it allocates
the bio in __alloc_device and then re-uses it multiple times with a
very simple endio routine that just calls complete() without consuming
a reference. Allocated bios by default come with a ref count of 1,
which is then consumed by the endio routine (or not, in which case they
should be bio_put by the caller). The way the impleementation works now
is that the flush bio has a refcount of 2 and we only ever bio_put it
once, leaving it to hang indefinitely. Fix this by removing the extra
bio_get in __alloc_device.

Fixes: e0ae9994

 ("btrfs: preallocate device flush bio")
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

The name char array passed to btrfs_search_path_in_tree is of size
BTRFS_INO_LOOKUP_PATH_MAX (4080). So the actual accessible char indexes
are in the range of [0, 4079]. Currently the code uses the define but this
represents an off-by-one.

Implications:

Size of btrfs_ioctl_ino_lookup_args is 4096, so the new byte will be
written to extra space, not some padding that could be provided by the
allocator.

btrfs-progs store the arguments on stack, but kernel does own copy of
the ioctl buffer and the off-by-one overwrite does not affect userspace,
but the ending 0 might be lost.

Kernel ioctl buffer is allocated dynamically so we're overwriting
somebody else's memory, and the ioctl is privileged if args.objectid is
not 256. Which is in most cases, but resolving a subvolume stored in
another directory will trigger that path.

Before this patch the buffer was one byte larger, but then the -1 was
not added.

Fixes: ac8e9819

 ("Btrfs: add search and inode lookup ioctls")
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ added implications ]
Signed-off-by: David Sterba <dsterba@suse.com>

I was seeing disk flushes still happening when I mounted a Btrfs
filesystem with nobarrier for testing. This is because we use FUA to
write out the first super block, and on devices without FUA support, the
block layer translates FUA to a flush. Even on devices supporting true
FUA, using FUA when we asked for no barriers is surprising.

Fixes: 387125fc

 ("Btrfs: fix barrier flushes")
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

If btrfs_del_root fails in btrfs_drop_snapshot, we'll pick up the
error but then return 0 anyway due to mixing err and ret.

Fixes: 79787eaa

 ("btrfs: replace many BUG_ONs with proper error handling")
Cc: <stable@vger.kernel.org> # v3.4+
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Since commit fb235dc0 (btrfs: qgroup: Move half of the qgroup
accounting time out of commit trans) the assumption that
btrfs_add_delayed_{data,tree}_ref can only return 0 or -ENOMEM has
been false.  The qgroup operations call into btrfs_search_slot
and friends and can now return the full spectrum of error codes.

Fortunately, the fix here is easy since update_ref_for_cow failing
is already handled so we just need to bail early with the error
code.

Fixes: fb235dc0

 (btrfs: qgroup: Move half of the qgroup accounting ...)
Cc: <stable@vger.kernel.org> # v4.11+
Signed-off-by: Jeff Mahoney <jeffm@suse.com>
Reviewed-by: Edmund Nadolski <enadolski@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Commit c6887cd1 ("Btrfs: don't do nocow check unless we have to")
changed the behavior of __btrfs_buffered_write() so that it first tries
to get a data space reservation, and then skips the relatively expensive
nocow check if the reservation succeeded.

If we have quotas enabled, the data space reservation also includes a
quota reservation.  But in the rewrite case, the space has already been
accounted for in qgroups.  So btrfs_check_data_free_space() increases
the quota reservation, but it never gets decreased when the data
actually gets written and overwrites the pre-existing data.  So we're
left with both the qgroup and qgroup reservation accounting for the same
space.

This commit adds the missing btrfs_qgroup_free_data() call in the case
of BTRFS_ORDERED_PREALLOC extents.

Fixes: c6887cd1

 ("Btrfs: don't do nocow check unless we have to")
Signed-off-by: Justin Maggard <jmaggard@netgear.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Under some circumstances, an incremental send operation can issue wrong
paths for unlink commands related to files that have multiple hard links
and some (or all) of those links were renamed between the parent and send
snapshots. Consider the following example:

Parent snapshot

 .                                                      (ino 256)
 |---- a/                                               (ino 257)
 |     |---- b/                                         (ino 259)
 |     |     |---- c/                                   (ino 260)
 |     |     |---- f2                                   (ino 261)
 |     |
 |     |---- f2l1                                       (ino 261)
 |
 |---- d/                                               (ino 262)
       |---- f1l1_2                                     (ino 258)
       |---- f2l2                                       (ino 261)
       |---- f1_2                                       (ino 258)

Send snapshot

 .                                                      (ino 256)
 |---- a/                                               (ino 257)
 |     |---- f2l1/                                      (ino 263)
 |             |---- b2/                                (ino 259)
 |                   |---- c/                           (ino 260)
 |                   |     |---- d3                     (ino 262)
 |                   |           |---- f1l1_2           (ino 258)
 |                   |           |---- f2l2_2           (ino 261)
 |                   |           |---- f1_2             (ino 258)
 |                   |
 |                   |---- f2                           (ino 261)
 |                   |---- f1l2                         (ino 258)
 |
 |---- d                                                (ino 261)

When computing the incremental send stream the following steps happen:

1) When processing inode 261, a rename operation is issued that renames
   inode 262, which currently as a path of "d", to an orphan name of
   "o262-7-0". This is done because in the send snapshot, inode 261 has
   of its hard links with a path of "d" as well.

2) Two link operations are issued that create the new hard links for
   inode 261, whose names are "d" and "f2l2_2", at paths "/" and
   "o262-7-0/" respectively.

3) Still while processing inode 261, unlink operations are issued to
   remove the old hard links of inode 261, with names "f2l1" and "f2l2",
   at paths "a/" and "d/". However path "d/" does not correspond anymore
   to the directory inode 262 but corresponds instead to a hard link of
   inode 261 (link command issued in the previous step). This makes the
   receiver fail with a ENOTDIR error when attempting the unlink
   operation.

The problem happens because before sending the unlink operation, we failed
to detect that inode 262 was one of ancestors for inode 261 in the parent
snapshot, and therefore we didn't recompute the path for inode 262 before
issuing the unlink operation for the link named "f2l2" of inode 262. The
detection failed because the function "is_ancestor()" only follows the
first hard link it finds for an inode instead of all of its hard links
(as it was originally created for being used with directories only, for
which only one hard link exists). So fix this by making "is_ancestor()"
follow all hard links of the input inode.

A test case for fstests follows soon.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

[BUG]
If we run btrfs with CONFIG_BTRFS_FS_RUN_SANITY_TESTS=y, it will
instantly cause kernel panic like:

------
...
assertion failed: 0, file: fs/btrfs/disk-io.c, line: 3853
...
Call Trace:
 btrfs_mark_buffer_dirty+0x187/0x1f0 [btrfs]
 setup_items_for_insert+0x385/0x650 [btrfs]
 __btrfs_drop_extents+0x129a/0x1870 [btrfs]
...
-----

[Cause]
Btrfs will call btrfs_check_leaf() in btrfs_mark_buffer_dirty() to check
if the leaf is valid with CONFIG_BTRFS_FS_RUN_SANITY_TESTS=y.

However quite some btrfs_mark_buffer_dirty() callers(*) don't really
initialize its item data but only initialize its item pointers, leaving
item data uninitialized.

This makes tree-checker catch uninitialized data as error, causing
such panic.

*: These callers include but not limited to
setup_items_for_insert()
btrfs_split_item()
btrfs_expand_item()

[Fix]
Add a new parameter @check_item_data to btrfs_check_leaf().
With @check_item_data set to false, item data check will be skipped and
fallback to old btrfs_check_leaf() behavior.

So we can still get early warning if we screw up item pointers, and
avoid false panic.

Cc: Filipe Manana <fdmanana@gmail.com>
Reported-by: Lakshmipathi.G <lakshmipathi.g@gmail.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Xfstests btrfs/146 revealed this corruption,

[   58.138831] Buffer I/O error on dev dm-0, logical block 2621424, async page read
[   58.151233] BTRFS error (device sdf): bdev /dev/mapper/error-test errs: wr 1, rd 0, flush 0, corrupt 0, gen 0
[   58.152403] list_add corruption. prev->next should be next (ffff88005e6775d8), but was ffffc9000189be88. (prev=ffffc9000189be88).
[   58.153518] ------------[ cut here ]------------
[   58.153892] WARNING: CPU: 1 PID: 1287 at lib/list_debug.c:31 __list_add_valid+0x169/0x1f0
...
[   58.157379] RIP: 0010:__list_add_valid+0x169/0x1f0
...
[   58.161956] Call Trace:
[   58.162264]  btrfs_log_inode_parent+0x5bd/0xfb0 [btrfs]
[   58.163583]  btrfs_log_dentry_safe+0x60/0x80 [btrfs]
[   58.164003]  btrfs_sync_file+0x4c2/0x6f0 [btrfs]
[   58.164393]  vfs_fsync_range+0x5f/0xd0
[   58.164898]  do_fsync+0x5a/0x90
[   58.165170]  SyS_fsync+0x10/0x20
[   58.165395]  entry_SYSCALL_64_fastpath+0x1f/0xbe
...

It turns out that we could record btrfs_log_ctx:io_err in
log_one_extents when IO fails, but make log_one_extents() return '0'
instead of -EIO, so the IO error is not acknowledged by the callers,
i.e.  btrfs_log_inode_parent(), which would remove btrfs_log_ctx:list
from list head 'root->log_ctxs'.  Since btrfs_log_ctx is allocated
from stack memory, it'd get freed with a object alive on the
list. then a future list_add will throw the above warning.

This returns the correct error in the above case.

Jeff also reported this while testing against his fsync error
patch set[1].

[1]: https://www.spinics.net/lists/linux-btrfs/msg65308.html
"btrfs list corruption and soft lockups while testing writeback error handling"

Fixes: 8407f553

 ("Btrfs: fix data corruption after fast fsync and writeback error")
Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

[BUG]
Kernel panic when mounting with "-o compress" mount option.
KASAN will report like:
------
==================================================================
BUG: KASAN: wild-memory-access in strncmp+0x31/0xc0
Read of size 1 at addr d86735fce994f800 by task mount/662
...
Call Trace:
 dump_stack+0xe3/0x175
 kasan_report+0x163/0x370
 __asan_load1+0x47/0x50
 strncmp+0x31/0xc0
 btrfs_compress_str2level+0x20/0x70 [btrfs]
 btrfs_parse_options+0xff4/0x1870 [btrfs]
 open_ctree+0x2679/0x49f0 [btrfs]
 btrfs_mount+0x1b7f/0x1d30 [btrfs]
 mount_fs+0x49/0x190
 vfs_kern_mount.part.29+0xba/0x280
 vfs_kern_mount+0x13/0x20
 btrfs_mount+0x31e/0x1d30 [btrfs]
 mount_fs+0x49/0x190
 vfs_kern_mount.part.29+0xba/0x280
 do_mount+0xaad/0x1a00
 SyS_mount+0x98/0xe0
 entry_SYSCALL_64_fastpath+0x1f/0xbe
------

[Cause]
For 'compress' and 'compress_force' options, its token doesn't expect
any parameter so its args[0] contains uninitialized data.
Accessing args[0] will cause above wild memory access.

[Fix]
For Opt_compress and Opt_compress_force, set compression level to
the default.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ set the default in advance ]
Signed-off-by: David Sterba <dsterba@suse.com>

If we fail to prepare our pages for whatever reason (out of memory in
our case) we need to make sure to drop the block_group->data_rwsem,
otherwise hilarity ensues.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Reviewed-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ add label and use existing unlocking code ]
Signed-off-by: David Sterba <dsterba@suse.com>

We discovered a box that had double allocations, and suspected the space
cache may be to blame.  While auditing the write out path I noticed that
if we've already setup the space cache we will just carry on.  This
means that any error we hit after cache_save_setup before we go to
actually write the cache out we won't reset the inode generation, so
whatever was already written will be considered correct, except it'll be
stale.  Fix this by _always_ resetting the generation on the block group
inode, this way we only ever have valid or invalid cache.

With this patch I was no longer able to reproduce cache corruption with
dm-log-writes and my bpf error injection tool.

Cc: stable@vger.kernel.org
Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

The patch from commit a7e3b975 ("Btrfs: fix reported number of inode
blocks") introduced a regression where if we do a buffered write starting
at position equal to or greater than the file's size and then stat(2) the
file before writeback is triggered, the number of used blocks does not
change (unless there's a prealloc/unwritten extent). Example:

  $ xfs_io -f -c "pwrite -S 0xab 0 64K" foobar
  $ du -h foobar
  0	foobar
  $ sync
  $ du -h foobar
  64K	foobar

The first version of that patch didn't had this regression and the second
version, which was the one committed, was made only to address some
performance regression detected by the intel test robots using fs_mark.

This fixes the regression by setting the new delaloc bit in the range, and
doing it at btrfs_dirty_pages() while setting the regular dealloc bit as
well, so that this way we set both bits at once avoiding navigation of the
inode's io tree twice. Doing it at btrfs_dirty_pages() is also the most
meaninful place, as we should set the new dellaloc bit when if we set the
delalloc bit, which happens only if we copied bytes into the pages at
__btrfs_buffered_write().

This was making some of LTP's du tests fail, which can be quickly run
using a command line like the following:

  $ ./runltp -q -p -l /ltp.log -f commands -s du -d /mnt

Fixes: a7e3b975

 ("Btrfs: fix reported number of inode blocks")
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Move the definition of the function btrfs_find_new_delalloc_bytes() closer
to the function btrfs_dirty_pages(), because in a future commit it will be
used exclusively by btrfs_dirty_pages(). This just moves the function's
definition, with no functional changes at all.

Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

If a file's DIR_ITEM key is invalid (due to memory errors) and gets
written to disk, a future lookup_path can end up with kernel panic due
to BUG_ON().

This gets rid of the BUG_ON(), meanwhile output the corrupted key and
return ENOENT if it's invalid.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reported-by: Guillaume Bouchard <bouchard@mercs-eng.com>
Signed-off-by: David Sterba <dsterba@suse.com>

The dev_alloc_list list could be protected by various mutexes,
depending on the context. The list tracks devices that can take part of
allocating new chunks, so the closest mutex is chunk_mutex. Adding a new
device from inside the ADD_DEV ioctl will need device_list_mutex and
registering a new device from the ioctl needs uuid_mutex.

All mutexes naturally guarantee exclusivity against the same context.
The device ownership can move between the contexts and the exclusivity
is guaranteed by other means, eg. during the mount with the uuid_mutex.

There's no RCU involved for dev_alloc_list.

Signed-off-by: David Sterba <dsterba@suse.com>

This fixes potential bio leaks, in several error paths. Unfortunatelly
the device structure freeing is opencoded in many places and I missed
them when introducing the flush_bio.

Most of the time, devices get freed through call_rcu(..., free_device),
so it at least it's not that easy to hit the leak, but it's still
possible through the path that frees stale devices.

Fixes: e0ae9994

 ("btrfs: preallocate device flush bio")
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: David Sterba <dsterba@suse.com>

btrfs_rm_dev_item calls several function under an active transaction,
however it fails to abort it if an error happens. Fix this by adding
explicit btrfs_abort_transaction/btrfs_end_transaction calls.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Compression code path has only flaged bios with REQ_OP_WRITE no matter
where the bios come from, but it could be a sync write if fsync starts
this writeback or a normal writeback write if wb kthread starts a
periodic writeback.

It breaks the rule that sync writes and writeback writes need to be
differentiated from each other, because from the POV of block layer,
all bios need to be recognized by these flags in order to do some
management, e.g. throttlling.

This passes writeback_control to compression write path so that it can
send bios with proper flags to block layer.

Signed-off-by: Liu Bo <bo.li.liu@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Fix bug of commit 74d46992

 ("block: replace bi_bdev with a gendisk
pointer and partitions index").

bio_dev(bio) is used to find the dev state in function
__btrfsic_submit_bio. But when dev_state is added to the hashtable, it
is using dev_t of block_device.

bio_dev(bio) returns a dev_t of part0 which is different from dev_t in
block_device(bd_dev). bd_dev in block_device represents the exact
partition.

block_device.bd_dev =
	bio->bi_partno (same as block_device.bd_partno) + bio_dev(bio).

When adding a dev_state into hashtable, we use the exact partition dev_t.
So when looking it up, it should also use the exact partition dev_t.

Reproducer of this bug:

Use MOUNT_OPTIONS="-o check_int" and run btrfs/001 in fstests.
Then there will be WARNING like below.

WARNING:
btrfs: attempt to write superblock which references block M @29523968 (sda7     /1111654400/2) which is never written!

Signed-off-by: Gu JinXiang <gujx@cn.fujitsu.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Byte distribution check in heuristic will filter edge data cases and
some time fail to classify input data.

Let's fix that by adding Shannon entropy calculation, that will cover
classification of most other data types.

As Shannon entropy needs log2 with some precision to work, let's use
ilog2(N) and for increased precision, by do ilog2(pow(N, 4)).

Shannon entropy has been slightly changed to avoid signed numbers and
division.

The calculation is direct by the formula, successor of precalculated
table or chains of if-else.

The accuracy errors of ilog2 are compensated by

@ENTROPY_LVL_ACEPTABLE 70 -> 65
@ENTROPY_LVL_HIGH      85 -> 80

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update comments ]
Signed-off-by: David Sterba <dsterba@suse.com>

Calculate byte core set for data sample:
- sort buckets' numbers in decreasing order
- count how many values cover 90% of the sample

If the core set size is low (<=25%), data are easily compressible.
If the core set size is high (>=80%), data are not compressible.

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update comments ]
Signed-off-by: David Sterba <dsterba@suse.com>

Calculate byte set size for data sample:
- calculate how many unique bytes have been in the sample
- for all bytes count > 0, check if we're still in the low count range
  (~25%), such data are easily compressible, otherwise furhter analysis
  is needed

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ update comments ]
Signed-off-by: David Sterba <dsterba@suse.com>

Walk over data sample and use memcmp to detect repeated patterns, like
zeros, but a bit more general.

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ minor coding style fixes ]
Signed-off-by: David Sterba <dsterba@suse.com>

Copy sample data from the input data range to sample buffer then
calculate byte value count for that sample into bucket.

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
[ minor comment updates ]
Signed-off-by: David Sterba <dsterba@suse.com>

Add basic defines and structures for data sampling.

Added macros:
 - For future sampling algo
 - For bucket size

Heuristic workspace:
 - Add bucket for storing byte type counters
 - Add sample array for storing partial copy of input data range
 - Add counter for store current sample size to workspace

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ minor coding style fixes, comments updated ]
Signed-off-by: David Sterba <dsterba@suse.com>

Compression heuristic itself is not a compression type, as current
infrastructure provides workspaces for several compression types, it's
difficult to just add heuristic workspace.

Just refactor the code to support compression/heuristic workspaces with
maximum code sharing and minimum changes in it.

Signed-off-by: Timofey Titovets <nefelim4ag@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ coding style fixes ]
Signed-off-by: David Sterba <dsterba@suse.com>

Since we do a delalloc reserve in btrfs_truncate_block we can deadlock
with freeze.  If somebody else is trying to allocate metadata for this
inode and it gets stuck in start_delalloc_inodes because of freeze we
will deadlock.  Be safe and move this outside of a trans handle.  This
also has a side-effect of making sure that we're not leaving stale data
behind in the other_encoding or encryption case.  Not an issue now since
nobody uses it, but it would be a problem in the future.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

We're holding the sb_start_intwrite lock at this point, and doing async
filemap_flush of the inodes will result in a deadlock if we freeze the
fs during this operation.  This is because we could do a
btrfs_join_transaction() in the thread we are waiting on which would
block at sb_start_intwrite, and thus deadlock.  Using
writeback_inodes_sb() side steps the problem by not introducing all of
these extra locking dependencies.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

If we get a significant amount of delayed refs for a single block (think
modifying multiple snapshots) we can end up spending an ungodly amount
of time looping through all of the entries trying to see if they can be
merged.  This is because we only add them to a list, so we have O(2n)
for every ref head.  This doesn't make any sense as we likely have refs
for different roots, and so they cannot be merged.  Tracking in a tree
will allow us to break as soon as we hit an entry that doesn't match,
making our worst case O(n).

With this we can also merge entries more easily.  Before we had to hope
that matching refs were on the ends of our list, but with the tree we
can search down to exact matches and merge them at insert time.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Instead of open-coding the delayed ref comparisons, add a helper to do
the comparisons generically and use that everywhere.  We compare
sequence numbers last for following patches.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Make it more consistent, we want the inserted ref to be compared against
what's already in there.  This will make the order go from lowest seq ->
highest seq, which will make us more likely to make forward progress if
there's a seqlock currently held.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

The way we handle delalloc metadata reservations has gotten
progressively more complicated over the years.  There is so much cruft
and weirdness around keeping the reserved count and outstanding counters
consistent and handling the error cases that it's impossible to
understand.

Fix this by making the delalloc block rsv per-inode.  This way we can
calculate the actual size of the outstanding metadata reservations every
time we make a change, and then reserve the delta based on that amount.
This greatly simplifies the code everywhere, and makes the error
handling in btrfs_delalloc_reserve_metadata far less terrifying.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

This is handy for tracing problems with modifying the outstanding
extents counters.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Right now we do a lot of weird hoops around outstanding_extents in order
to keep the extent count consistent.  This is because we logically
transfer the outstanding_extent count from the initial reservation
through the set_delalloc_bits.  This makes it pretty difficult to get a
handle on how and when we need to mess with outstanding_extents.

Fix this by revamping the rules of how we deal with outstanding_extents.
Now instead everybody that is holding on to a delalloc extent is
required to increase the outstanding extents count for itself.  This
means we'll have something like this

btrfs_delalloc_reserve_metadata	- outstanding_extents = 1
 btrfs_set_extent_delalloc	- outstanding_extents = 2
btrfs_release_delalloc_extents	- outstanding_extents = 1

for an initial file write.  Now take the append write where we extend an
existing delalloc range but still under the maximum extent size

btrfs_delalloc_reserve_metadata - outstanding_extents = 2
  btrfs_set_extent_delalloc
    btrfs_set_bit_hook		- outstanding_extents = 3
    btrfs_merge_extent_hook	- outstanding_extents = 2
btrfs_delalloc_release_extents	- outstanding_extnets = 1

In order to make the ordered extent transition we of course must now
make ordered extents carry their own outstanding_extent reservation, so
for cow_file_range we end up with

btrfs_add_ordered_extent	- outstanding_extents = 2
clear_extent_bit		- outstanding_extents = 1
btrfs_remove_ordered_extent	- outstanding_extents = 0

This makes all manipulations of outstanding_extents much more explicit.
Every successful call to btrfs_delalloc_reserve_metadata _must_ now be
combined with btrfs_release_delalloc_extents, even in the error case, as
that is the only function that actually modifies the
outstanding_extents counter.

The drawback to this is now we are much more likely to have transient
cases where outstanding_extents is much larger than it actually should
be.  This could happen before as we manipulated the delalloc bits, but
now it happens basically at every write.  This may put more pressure on
the ENOSPC flushing code, but I think making this code simpler is worth
the cost.  I have another change coming to mitigate this side-effect
somewhat.

I also added trace points for the counter manipulation.  These were used
by a bpf script I wrote to help track down leak issues.

Signed-off-by: Josef Bacik <jbacik@fb.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Build-server workloads have hundreds of references per file after dedup.
Multiply by a few snapshots and we quickly exhaust the limit of 2730
references per extent that can fit into a 64K buffer.

Raise the limit to 16M to be consistent with other btrfs ioctls
(e.g. TREE_SEARCH_V2, FILE_EXTENT_SAME).

To minimize surprising userspace behavior, apply this change only to
the LOGICAL_INO_V2 ioctl.

Signed-off-by: Zygo Blaxell <ce3g8jdj@umail.furryterror.org>
Reviewed-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Tested-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Signed-off-by: David Sterba <dsterba@suse.com>

Now that check_extent_in_eb()'s extent offset filter can be turned off,
we need a way to do it from userspace.

Add a 'flags' field to the btrfs_logical_ino_args structure to disable
extent offset filtering, taking the place of one of the existing
reserved[] fields.

Previous versions of LOGICAL_INO neglected to check whether any of the
reserved fields have non-zero values.  Assigning meaning to those fields
now may change the behavior of existing programs that left these fields
uninitialized.  The lack of a zero check also means that new programs
have no way to know whether the kernel is honoring the flags field.

To avoid these problems, define a new ioctl LOGICAL_INO_V2.  We can
use the same argument layout as LOGICAL_INO, but shorten the reserved[]
array by one element and turn it into the 'flags' field.  The V2 ioctl
explicitly checks that reserved fields and unsupported flag bits are zero
so that userspace can negotiate future feature bits as they are defined.

Since the memory layouts of the two ioctls' arguments are compatible,
there is no need for a separate function for logical_to_ino_v2 (contrast
with tree_search_v2 vs tree_search where the layout and code are quite
different).  A version parameter and an 'if' statement will suffice.

Now that we have a flags field in logical_ino_args, add a flag
BTRFS_LOGICAL_INO_ARGS_IGNORE_OFFSET to get the behavior we want,
and pass it down the stack to iterate_inodes_from_logical.

Motivation and background, copied from the patchset cover letter:

Suppose we have a file with one extent:

    root@tester:~# zcat /usr/share/doc/cpio/changelog.gz > /test/a
    root@tester:~# sync

Split the extent by overwriting it in the middle:

    root@tester:~# cat /dev/urandom | dd bs=4k seek=2 skip=2 count=1 conv=notrunc of=/test/a

We should now have 3 extent refs to 2 extents, with one block unreachable.
The extent tree looks like:

    root@tester:~# btrfs-debug-tree /dev/vdc -t 2
    [...]
            item 9 key (1103101952 EXTENT_ITEM 73728) itemoff 15942 itemsize 53
                    extent refs 2 gen 29 flags DATA
                    extent data backref root 5 objectid 261 offset 0 count 2
    [...]
            item 11 key (1103175680 EXTENT_ITEM 4096) itemoff 15865 itemsize 53
                    extent refs 1 gen 30 flags DATA
                    extent data backref root 5 objectid 261 offset 8192 count 1
    [...]

and the ref tree looks like:

    root@tester:~# btrfs-debug-tree /dev/vdc -t 5
    [...]
            item 6 key (261 EXTENT_DATA 0) itemoff 15825 itemsize 53
                    extent data disk byte 1103101952 nr 73728
                    extent data offset 0 nr 8192 ram 73728
                    extent compression(none)
            item 7 key (261 EXTENT_DATA 8192) itemoff 15772 itemsize 53
                    extent data disk byte 1103175680 nr 4096
                    extent data offset 0 nr 4096 ram 4096
                    extent compression(none)
            item 8 key (261 EXTENT_DATA 12288) itemoff 15719 itemsize 53
                    extent data disk byte 1103101952 nr 73728
                    extent data offset 12288 nr 61440 ram 73728
                    extent compression(none)
    [...]

There are two references to the same extent with different, non-overlapping
byte offsets:

    [------------------72K extent at 1103101952----------------------]
    [--8K----------------|--4K unreachable----|--60K-----------------]
    ^                                         ^
    |                                         |
    [--8K ref offset 0--][--4K ref offset 0--][--60K ref offset 12K--]
                         |
                         v
                         [-----4K extent-----] at 1103175680

We want to find all of the references to extent bytenr 1103101952.

Without the patch (and without running btrfs-debug-tree), we have to
do it with 18 LOGICAL_INO calls:

    root@tester:~# btrfs ins log 1103101952 -P /test/
    Using LOGICAL_INO
    inode 261 offset 0 root 5

    root@tester:~# for x in $(seq 0 17); do btrfs ins log $((1103101952 + x * 4096)) -P /test/; done 2>&1 | grep inode
    inode 261 offset 0 root 5
    inode 261 offset 4096 root 5   <- same extent ref as offset 0
                                   (offset 8192 returns empty set, not reachable)
    inode 261 offset 12288 root 5
    inode 261 offset 16384 root 5  \
    inode 261 offset 20480 root 5  |
    inode 261 offset 24576 root 5  |
    inode 261 offset 28672 root 5  |
    inode 261 offset 32768 root 5  |
    inode 261 offset 36864 root 5  \
    inode 261 offset 40960 root 5   > all the same extent ref as offset 12288.
    inode 261 offset 45056 root 5  /  More processing required in userspace
    inode 261 offset 49152 root 5  |  to figure out these are all duplicates.
    inode 261 offset 53248 root 5  |
    inode 261 offset 57344 root 5  |
    inode 261 offset 61440 root 5  |
    inode 261 offset 65536 root 5  |
    inode 261 offset 69632 root 5  /

In the worst case the extents are 128MB long, and we have to do 32768
iterations of the loop to find one 4K extent ref.

With the patch, we just use one call to map all refs to the extent at once:
    root@tester:~# btrfs ins log 1103101952 -P /test/
    Using LOGICAL_INO_V2
    inode 261 offset 0 root 5
    inode 261 offset 12288 root 5

The TREE_SEARCH ioctl allows userspace to retrieve the offset and
extent bytenr fields easily once the root, inode and offset are known.
This is sufficient information to build a complete map of the extent
and all of its references.  Userspace can use this information to make
better choices to dedup or defrag.

Signed-off-by: Zygo Blaxell <ce3g8jdj@umail.furryterror.org>
Reviewed-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
Tested-by: Hans van Kranenburg <hans.van.kranenburg@mendix.com>
[ copy background and motivation from cover letter ]
Signed-off-by: David Sterba <dsterba@suse.com>

The LOGICAL_INO ioctl provides a backward mapping from extent bytenr and
offset (encoded as a single logical address) to a list of extent refs.
LOGICAL_INO complements TREE_SEARCH, which provides the forward mapping
(extent ref -> extent bytenr and offset, or logical address).  These are
useful capabilities for programs that manipulate extents and extent
references from userspace (e.g. dedup and defrag utilities).

When the extents are uncompressed (and not encrypted and not other),
check_extent_in_eb performs filtering of the extent refs to remove any
extent refs which do not contain the same extent offset as the 'logical'
parameter's extent offset.  This prevents LOGICAL_INO from returning
references to more than a single block.

To find the set of extent references to an uncompressed extent from [a, b),
userspace has to run a loop like this pseudocode:

	for (i = a; i < b; ++i)
		extent_ref_set += LOGICAL_INO(i);

At each iteration of the loop (up to 32768 iterations for a 128M extent),
data we are interested in is collected in the kernel, then deleted by
the filter in check_extent_in_eb.

When the extents are compressed (or encrypted or other), the 'logical'
parameter must be an extent bytenr (the 'a' parameter in the loop).
No filtering by extent offset is done (or possible?) so the result is
the complete set of extent refs for the entire extent.  This removes
the need for the loop, since we get all the extent refs in one call.

Add an 'ignore_offset' argument to iterate_inodes_from_logical,
[...several levels of function call graph...], and check_extent_in_eb, so
that we can disable the extent offset filtering for uncompressed extents.
This flag can be set by an improved version of the LOGICAL_INO ioctl to
get either behavior as desired.

There is no functional change in this patch.  The new flag is always
false.

Signed-off-by: Zygo Blaxell <ce3g8jdj@umail.furryterror.org>
Reviewed-by: David Sterba <dsterba@suse.com>
[ minor coding style fixes ]
Signed-off-by: David Sterba <dsterba@suse.com>