sfrench/cifs-2.6.git
2 years agobtrfs: use chunk_root in find_free_extent_update_loop
Josef Bacik [Fri, 5 Nov 2021 20:45:38 +0000 (16:45 -0400)]
btrfs: use chunk_root in find_free_extent_update_loop

We're only using this to start the transaction with to possibly allocate
a chunk.  It doesn't really matter which root to use, but with extent
tree v2 we'll need a bytenr to look up a extent root which makes the
usage of the extent_root awkward here.  Simply change it to the
chunk_root.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: make remove_extent_backref pass the root
Josef Bacik [Fri, 5 Nov 2021 20:45:37 +0000 (16:45 -0400)]
btrfs: make remove_extent_backref pass the root

With extent tree v2 we'll have a different extent root based on where
the bytenr is located, so adjust the remove_extent_backref() helper and
it's helpers to pass the extent_root around.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: add a btrfs_block_group_root() helper
Josef Bacik [Fri, 5 Nov 2021 20:45:36 +0000 (16:45 -0400)]
btrfs: add a btrfs_block_group_root() helper

With extent tree v2 we will have a separate root to hold the block group
items.  Add a btrfs_block_group_root() that will return the appropriate
root given the flags of the fs, and convert all functions that need to
modify block group items to use the helper.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove BUG_ON(!eie) in find_parent_nodes
Josef Bacik [Fri, 5 Nov 2021 20:45:35 +0000 (16:45 -0400)]
btrfs: remove BUG_ON(!eie) in find_parent_nodes

If we're looking for leafs that point to a data extent we want to record
the extent items that point at our bytenr.  At this point we have the
reference and we know for a fact that this leaf should have a reference
to our bytenr.  However if there's some sort of corruption we may not
find any references to our leaf, and thus could end up with eie == NULL.
Replace this BUG_ON() with an ASSERT() and then return -EUCLEAN for the
mortals.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove BUG_ON() in find_parent_nodes()
Josef Bacik [Fri, 5 Nov 2021 20:45:34 +0000 (16:45 -0400)]
btrfs: remove BUG_ON() in find_parent_nodes()

We search for an extent entry with .offset = -1, which shouldn't be a
thing, but corruption happens.  Add an ASSERT() for the developers,
return -EUCLEAN for mortals.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove SANITY_TESTS check form find_parent_nodes
Josef Bacik [Fri, 5 Nov 2021 20:45:33 +0000 (16:45 -0400)]
btrfs: remove SANITY_TESTS check form find_parent_nodes

We define __TRANS_DUMMY always, so this extra ifdef stuff is not needed.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: move comment in find_parent_nodes()
Josef Bacik [Fri, 5 Nov 2021 20:45:32 +0000 (16:45 -0400)]
btrfs: move comment in find_parent_nodes()

This comment was much closer to the related code when it was originally
added, but has slowly migrated north far from its ancestral lands.  Move
it back down with its people.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: pass the root to add_keyed_refs
Josef Bacik [Fri, 5 Nov 2021 20:45:31 +0000 (16:45 -0400)]
btrfs: pass the root to add_keyed_refs

We pass in the path, but use btrfs_next_item() using the root we
searched with.  Pass the root down to add_keyed_refs() instead of the
fs_info so we can continue to use the same root we searched with.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove trans_handle->root
Josef Bacik [Fri, 5 Nov 2021 20:45:30 +0000 (16:45 -0400)]
btrfs: remove trans_handle->root

Nobody is using this anymore, remove it.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: pass fs_info to trace_btrfs_transaction_commit
Josef Bacik [Fri, 5 Nov 2021 20:45:29 +0000 (16:45 -0400)]
btrfs: pass fs_info to trace_btrfs_transaction_commit

The root on the trans->root can be anything, and generally we're
committing from the transaction kthread so it's usually the tree_root.
Change this to just take an fs_info, and to maintain compatibility
simply put the ROOT_TREE_OBJECTID as the root objectid for the
tracepoint.  This will allow use to remove trans->root.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: rework async transaction committing
Josef Bacik [Fri, 5 Nov 2021 20:45:28 +0000 (16:45 -0400)]
btrfs: rework async transaction committing

Currently we do this awful thing where we get another ref on a trans
handle, async off that handle and commit the transaction from that work.
Because we do this we have to mess with current->journal_info and the
freeze counting stuff.

We already have an async thing to kick for the transaction commit, the
transaction kthread.  Replace this work struct with a flag on the
fs_info to tell the kthread to go ahead and commit even if it's before
our timeout.  Then we can drastically simplify the async transaction
commit path.

Note: this can be simplified and functionality based on the pending
operation COMMIT.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
[ add note ]
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove unused BTRFS_FS_BARRIER flag
Josef Bacik [Fri, 5 Nov 2021 20:45:27 +0000 (16:45 -0400)]
btrfs: remove unused BTRFS_FS_BARRIER flag

This is no longer used, the -o nobarrier is handled by
BTRFS_MOUNT_NOBARRIER.  Remove the flag.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: eliminate if in main loop in tree_search_offset
Nikolay Borisov [Tue, 23 Nov 2021 07:23:42 +0000 (09:23 +0200)]
btrfs: eliminate if in main loop in tree_search_offset

Reshuffle the code inside the first loop of tree_search_offset so that
one if() is eliminated and the becomes more linear.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: don't check stripe length if the profile is not stripe based
Qu Wenruo [Fri, 19 Nov 2021 06:19:33 +0000 (14:19 +0800)]
btrfs: don't check stripe length if the profile is not stripe based

[BUG]
When debugging calc_bio_boundaries(), I found that even for RAID1
metadata, we're following stripe length to calculate stripe boundary.

  # mkfs.btrfs -m raid1 -d raid1 /dev/test/scratch[12]
  # mount /dev/test/scratch /mnt/btrfs
  # xfs_io -f -c "pwrite 0 64K" /mnt/btrfs/file
  # umount

Above very basic operations will make calc_bio_boundaries() to report
the following result:

  submit_extent_page: r/i=1/1 file_offset=22036480 len_to_stripe_boundary=49152
  submit_extent_page: r/i=1/1 file_offset=30474240 len_to_stripe_boundary=65536
  ...
  submit_extent_page: r/i=1/1 file_offset=30523392 len_to_stripe_boundary=16384
  submit_extent_page: r/i=1/1 file_offset=30457856 len_to_stripe_boundary=16384
  submit_extent_page: r/i=5/257 file_offset=0 len_to_stripe_boundary=65536
  submit_extent_page: r/i=5/257 file_offset=65536 len_to_stripe_boundary=65536
  submit_extent_page: r/i=1/1 file_offset=30490624 len_to_stripe_boundary=49152
  submit_extent_page: r/i=1/1 file_offset=30507008 len_to_stripe_boundary=32768

Where "r/i" is the rootid and inode, 1/1 means they metadata.
The remaining names match the member used in kernel.

Even all data/metadata are using RAID1, we're still following stripe
length.

[CAUSE]
This behavior is caused by a wrong condition in btrfs_get_io_geometry():

if (map->type & BTRFS_BLOCK_GROUP_PROFILE_MASK) {
/* Fill using stripe_len */
len = min_t(u64, em->len - offset, max_len);
} else {
len = em->len - offset;
}

This means, only for SINGLE we will not follow stripe_len.

However for profiles like RAID1*, DUP, they don't need to bother
stripe_len.

This can lead to unnecessary bio split for RAID1*/DUP profiles, and can
even be a blockage for future zoned RAID support.

[FIX]
Introduce one single-use macro, BTRFS_BLOCK_GROUP_STRIPE_MASK, and
change the condition to only calculate the length using stripe length
for stripe based profiles.

Reviewed-by: Johannes Thumshirn <johannes.thumshirn@wdc.com>
Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: get next entry in tree_search_offset before doing checks
Nikolay Borisov [Mon, 22 Nov 2021 15:16:46 +0000 (17:16 +0200)]
btrfs: get next entry in tree_search_offset before doing checks

This is a small optimisation since the currently 'entry' is already
checked in the if () {} else if {} construct above the loop. In essence
the first iteration of the final while loop is redundant. To eliminate
this extra check simply get the next entry at the beginning of the loop.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: add self test for bytes_index free space cache
Josef Bacik [Thu, 18 Nov 2021 21:33:16 +0000 (16:33 -0500)]
btrfs: add self test for bytes_index free space cache

I noticed a few corner cases when looking at my bytes_index patch for
obvious bugs, so add a bunch of tests to validate proper behavior of the
bytes_index tree.  A couple of basic tests to make sure it puts things
in the correct order, and then more complicated tests to make sure it
re-arranges bitmap entries properly and does the right thing when we try
to make allocations.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: index free space entries on size
Josef Bacik [Thu, 18 Nov 2021 21:33:15 +0000 (16:33 -0500)]
btrfs: index free space entries on size

Currently we index free space on offset only, because usually we have a
hint from the allocator that we want to honor for locality reasons.
However if we fail to use this hint we have to go back to a brute force
search through the free space entries to find a large enough extent.

With sufficiently fragmented free space this becomes quite expensive, as
we have to linearly search all of the free space entries to find if we
have a part that's long enough.

To fix this add a cached rb tree to index based on free space entry
bytes.  This will allow us to quickly look up the largest chunk in the
free space tree for this block group, and stop searching once we've
found an entry that is too small to satisfy our allocation.  We simply
choose to use this tree if we're searching from the beginning of the
block group, as we know we do not care about locality at that point.

I wrote an allocator test that creates a 10TiB ram backed null block
device and then fallocates random files until the file system is full.
I think go through and delete all of the odd files.  Then I spawn 8
threads that fallocate 64MiB files (1/2 our extent size cap) until the
file system is full again.  I use bcc's funclatency to measure the
latency of find_free_extent.  The baseline results are

     nsecs               : count     distribution
         0 -> 1          : 0        |                                        |
         2 -> 3          : 0        |                                        |
         4 -> 7          : 0        |                                        |
         8 -> 15         : 0        |                                        |
        16 -> 31         : 0        |                                        |
        32 -> 63         : 0        |                                        |
        64 -> 127        : 0        |                                        |
       128 -> 255        : 0        |                                        |
       256 -> 511        : 10356    |****                                    |
       512 -> 1023       : 58242    |*************************               |
      1024 -> 2047       : 74418    |********************************        |
      2048 -> 4095       : 90393    |****************************************|
      4096 -> 8191       : 79119    |***********************************     |
      8192 -> 16383      : 35614    |***************                         |
     16384 -> 32767      : 13418    |*****                                   |
     32768 -> 65535      : 12811    |*****                                   |
     65536 -> 131071     : 17090    |*******                                 |
    131072 -> 262143     : 26465    |***********                             |
    262144 -> 524287     : 40179    |*****************                       |
    524288 -> 1048575    : 55469    |************************                |
   1048576 -> 2097151    : 48807    |*********************                   |
   2097152 -> 4194303    : 26744    |***********                             |
   4194304 -> 8388607    : 35351    |***************                         |
   8388608 -> 16777215   : 13918    |******                                  |
  16777216 -> 33554431   : 21       |                                        |

avg = 908079 nsecs, total: 580889071441 nsecs, count: 639690

And the patch results are

     nsecs               : count     distribution
         0 -> 1          : 0        |                                        |
         2 -> 3          : 0        |                                        |
         4 -> 7          : 0        |                                        |
         8 -> 15         : 0        |                                        |
        16 -> 31         : 0        |                                        |
        32 -> 63         : 0        |                                        |
        64 -> 127        : 0        |                                        |
       128 -> 255        : 0        |                                        |
       256 -> 511        : 6883     |**                                      |
       512 -> 1023       : 54346    |*********************                   |
      1024 -> 2047       : 79170    |********************************        |
      2048 -> 4095       : 98890    |****************************************|
      4096 -> 8191       : 81911    |*********************************       |
      8192 -> 16383      : 27075    |**********                              |
     16384 -> 32767      : 14668    |*****                                   |
     32768 -> 65535      : 13251    |*****                                   |
     65536 -> 131071     : 15340    |******                                  |
    131072 -> 262143     : 26715    |**********                              |
    262144 -> 524287     : 43274    |*****************                       |
    524288 -> 1048575    : 53870    |*********************                   |
   1048576 -> 2097151    : 55368    |**********************                  |
   2097152 -> 4194303    : 41036    |****************                        |
   4194304 -> 8388607    : 24927    |**********                              |
   8388608 -> 16777215   : 33       |                                        |
  16777216 -> 33554431   : 9        |                                        |

avg = 623599 nsecs, total: 397259314759 nsecs, count: 637042

There's a little variation in the amount of calls done because of timing
of the threads with metadata requirements, but the avg, total, and
count's are relatively consistent between runs (usually within 2-5% of
each other).  As you can see here we have around a 30% decrease in
average latency with a 30% decrease in overall time spent in
find_free_extent.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: only use ->max_extent_size if it is set in the bitmap
Josef Bacik [Thu, 18 Nov 2021 21:33:14 +0000 (16:33 -0500)]
btrfs: only use ->max_extent_size if it is set in the bitmap

While adding self tests for my space index change I was hitting a
problem where the space indexed tree wasn't returning the expected
->max_extent_size.  This is because we will skip searching any entry
that doesn't have ->bytes >= the amount of bytes we want.  However we'll
still set the max_extent_size based on that entry.  The problem is if we
don't search the bitmap we won't have ->max_extent_size set properly, so
we can't really trust it.

This doesn't really result in a problem per-se, it can just result in us
not finding contiguous area that may exist.  Fix the max_extent_size
helper to return ->bytes if ->max_extent_size isn't set, and add a big
comment explaining why we're doing this.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove unnecessary @nr_written parameters
Qu Wenruo [Fri, 12 Nov 2021 05:33:14 +0000 (13:33 +0800)]
btrfs: remove unnecessary @nr_written parameters

We use @nr_written to record how many pages have been started by
btrfs_run_delalloc_range().

Currently there are only two cases that would populate @nr_written:

- Inline extent creation
- Compressed write

But both cases will also set @page_started to one.

In fact, in writepage_delalloc() we have the following code, showing
that @nr_written is really only utilized for above two cases:

/* did the fill delalloc function already unlock and start
 * the IO?
 */
if (page_started) {
/*
 * we've unlocked the page, so we can't update
 * the mapping's writeback index, just update
 * nr_to_write.
 */
wbc->nr_to_write -= nr_written;
return 1;
}

But for such cases, writepage_delalloc() will return 1, and exit
__extent_writepage() without going through __extent_writepage_io().

Thus this means, inside __extent_writepage_io(), we always get
@nr_written as 0.

So this patch is going to remove the unnecessary parameter from the
following functions:

- writepage_delalloc()

  As @nr_written passed in is always the initial value 0.

  Although inside that function, we still need a local @nr_written
  to update wbc->nr_to_write.

- __extent_writepage_io()

  As explained above, @nr_written passed in can only be 0.

  This also means we can remove one update_nr_written() call.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: change root to fs_info for btrfs_reserve_metadata_bytes
Josef Bacik [Tue, 9 Nov 2021 15:12:07 +0000 (10:12 -0500)]
btrfs: change root to fs_info for btrfs_reserve_metadata_bytes

We used to need the root for btrfs_reserve_metadata_bytes to check the
orphan cleanup state, but we no longer need that, we simply need the
fs_info.  Change btrfs_reserve_metadata_bytes() to use the fs_info, and
change both btrfs_block_rsv_refill() and btrfs_block_rsv_add() to do the
same as they simply call btrfs_reserve_metadata_bytes() and then
manipulate the block_rsv that is being used.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: get rid of root->orphan_cleanup_state
Josef Bacik [Tue, 9 Nov 2021 15:12:06 +0000 (10:12 -0500)]
btrfs: get rid of root->orphan_cleanup_state

Now that we don't care about the stage of the orphan_cleanup_state,
simply replace it with a bit on ->state to make sure we don't call the
orphan cleanup every time we wander into this root.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove global rsv stealing logic for orphan cleanup
Josef Bacik [Tue, 9 Nov 2021 15:12:05 +0000 (10:12 -0500)]
btrfs: remove global rsv stealing logic for orphan cleanup

This is very old code before we were stealing from the global reserve
during evict.  We have proper ways to steal from the global reserve
while we're evicting, so rip out this code as it's no longer necessary.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: make BTRFS_RESERVE_FLUSH_EVICT use the global rsv stealing code
Josef Bacik [Tue, 9 Nov 2021 15:12:04 +0000 (10:12 -0500)]
btrfs: make BTRFS_RESERVE_FLUSH_EVICT use the global rsv stealing code

I forgot to convert this over when I introduced the global reserve
stealing code to the space flushing code.  Evict was simply trying to
make its reservation and then if it failed it would steal from the
global rsv, which is racey because it's outside of the normal ticketing
code.

Fix this by setting ticket->steal if we are BTRFS_RESERVE_FLUSH_EVICT,
and then make the priority flushing path do the steal for us.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: check ticket->steal in steal_from_global_block_rsv
Josef Bacik [Tue, 9 Nov 2021 15:12:03 +0000 (10:12 -0500)]
btrfs: check ticket->steal in steal_from_global_block_rsv

We're going to use this helper in the priority flushing loop, move this
check into the helper to simplify the logic.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: check for priority ticket granting before flushing
Josef Bacik [Tue, 9 Nov 2021 15:12:02 +0000 (10:12 -0500)]
btrfs: check for priority ticket granting before flushing

Since we're dropping locks before we enter the priority flushing loops
we could have had our ticket granted before we got the space_info->lock.
So add this check to avoid doing some extra flushing in the priority
flushing cases.

The case in priority_reclaim_metadata_space is an optimization.  Think
we came in to reserve, we didn't have the space, we added our ticket to
the list.  But at the same time somebody was waiting on the space_info
lock to add space and do btrfs_try_granting_ticket(), so we drop the
lock, get satisfied, come in to do our loop, and we have been
satisfied.

This is the priority reclaim path, so to_reclaim could be !0 still
because we may have only satisfied the priority tickets and still left
non priority tickets on the list.  We would then have to_reclaim but
->bytes == 0.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
[ add note about the optimization ]
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: handle priority ticket failures in their respective helpers
Josef Bacik [Tue, 9 Nov 2021 15:12:01 +0000 (10:12 -0500)]
btrfs: handle priority ticket failures in their respective helpers

Currently the error case for the priority tickets is handled where we
deal with all of the tickets, priority and non-priority.  This is OK in
general, but it makes for some awkward locking.  We take and drop the
space_info->lock back to back because of these different types of
tickets.

Rework the code to handle priority ticket failures in their respective
helpers.  This allows us to be less wonky with our space_info->lock
usage, and means that the main handler simply has to check
ticket->error, as the ticket is guaranteed to be off any list and
completely handled by the time it exits one of the handlers.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: zoned: cache reported zone during mount
Naohiro Aota [Thu, 11 Nov 2021 05:14:38 +0000 (14:14 +0900)]
btrfs: zoned: cache reported zone during mount

When mounting a device, we are reporting the zones twice: once for
checking the zone attributes in btrfs_get_dev_zone_info and once for
loading block groups' zone info in
btrfs_load_block_group_zone_info(). With a lot of block groups, that
leads to a lot of REPORT ZONE commands and slows down the mount
process.

This patch introduces a zone info cache in struct
btrfs_zoned_device_info. The cache is populated while in
btrfs_get_dev_zone_info() and used for
btrfs_load_block_group_zone_info() to reduce the number of REPORT ZONE
commands. The zone cache is then released after loading the block
groups, as it will not be much effective during the run time.

Benchmark: Mount an HDD with 57,007 block groups
Before patch: 171.368 seconds
After patch: 64.064 seconds

While it still takes a minute due to the slowness of loading all the
block groups, the patch reduces the mount time by 1/3.

Link: https://lore.kernel.org/linux-btrfs/CAHQ7scUiLtcTqZOMMY5kbWUBOhGRwKo6J6wYPT5WY+C=cD49nQ@mail.gmail.com/
Fixes: 5b316468983d ("btrfs: get zone information of zoned block devices")
CC: stable@vger.kernel.org
Signed-off-by: Naohiro Aota <naohiro.aota@wdc.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove unused parameter fs_devices from btrfs_init_workqueues
Su Yue [Wed, 10 Nov 2021 06:42:17 +0000 (14:42 +0800)]
btrfs: remove unused parameter fs_devices from btrfs_init_workqueues

Since commit ba8a9d079543 ("Btrfs: delete the entire async bio submission
framework") removed submit workqueues, the parameter fs_devices is not used
anymore.

Remove it, no functional changes.

Reviewed-by: Anand Jain <anand.jain@oracle.com>
Signed-off-by: Su Yue <l@damenly.su>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: reduce the scope of the tree log mutex during transaction commit
Filipe Manana [Wed, 10 Nov 2021 10:05:21 +0000 (10:05 +0000)]
btrfs: reduce the scope of the tree log mutex during transaction commit

In the transaction commit path we are acquiring the tree log mutex too
early and we have a stale comment because:

1) It mentions a function named btrfs_commit_tree_roots(), which does not
   exists anymore, it was the old name of commit_cowonly_roots(), renamed
   a very long time ago by commit 5d4f98a28c7d33 ("Btrfs: Mixed back
   reference  (FORWARD ROLLING FORMAT CHANGE)"));

2) It mentions that we need to acquire the tree log mutex at that point
   to ensure we have no running log writers. That is not correct anymore,
   for many years at least, since we are guaranteed that we do not have
   any log writers at that point simply because we have set the state of
   the transaction to TRANS_STATE_COMMIT_DOING and have waited for all
   writers to complete - meaning no one can log until we change the state
   of the transaction to TRANS_STATE_UNBLOCKED. Any attempts to join the
   transaction or start a new one will block until we do that state
   transition;

3) The comment mentions a "trans mutex" which doesn't exists since 2011,
   commit a4abeea41adf ("Btrfs: kill trans_mutex") removed it;

4) The current use of the tree log mutex is to ensure proper serialization
   of super block writes - if someone started a new transaction and uses it
   for logging, it will wait for the previous transaction to write its
   super block before writing the super block when attempting to sync the
   log.

So acquire the tree log mutex only when it's absolutely needed, before
setting the transaction state to TRANS_STATE_UNBLOCKED, fix and move the
stale comment, add some assertions and new comments where appropriate.

Also, this has no effect on concurrency or performance, since the new
start of the critical section is still when the transaction is in the
state TRANS_STATE_COMMIT_DOING.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: consolidate device_list_mutex in prepare_sprout to its parent
Anand Jain [Tue, 9 Nov 2021 09:51:58 +0000 (17:51 +0800)]
btrfs: consolidate device_list_mutex in prepare_sprout to its parent

btrfs_prepare_sprout() splices seed devices into its own struct fs_devices,
so that its parent function btrfs_init_new_device() can add the new sprout
device to fs_info->fs_devices.

Both btrfs_prepare_sprout() and btrfs_init_new_device() need
device_list_mutex. But they are holding it separately, thus create a
small race window. Close it and hold device_list_mutex across both
functions btrfs_init_new_device() and btrfs_prepare_sprout().

Split btrfs_prepare_sprout() into btrfs_init_sprout() and
btrfs_setup_sprout(). This split is essential because device_list_mutex
must not be held for allocations in btrfs_init_sprout() but must be held
for btrfs_setup_sprout(). So now a common device_list_mutex can be used
between btrfs_init_new_device() and btrfs_setup_sprout().

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: switch seeding_dev in init_new_device to bool
Anand Jain [Tue, 21 Sep 2021 04:33:23 +0000 (12:33 +0800)]
btrfs: switch seeding_dev in init_new_device to bool

Declare int seeding_dev as a bool. Also, move its declaration a line
below to adjust packing.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: send: remove unused type parameter to iterate_inode_ref_t
Omar Sandoval [Fri, 5 Nov 2021 00:00:13 +0000 (17:00 -0700)]
btrfs: send: remove unused type parameter to iterate_inode_ref_t

Again, I don't think this was ever used since iterate_dir_item() is only
used for xattrs. No functional change.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: send: remove unused found_type parameter to lookup_dir_item_inode()
Omar Sandoval [Fri, 5 Nov 2021 00:00:12 +0000 (17:00 -0700)]
btrfs: send: remove unused found_type parameter to lookup_dir_item_inode()

As far as I can tell, this was never used. No functional change.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Omar Sandoval <osandov@fb.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: rename btrfs_item_end_nr to btrfs_item_data_end
Josef Bacik [Thu, 21 Oct 2021 18:58:37 +0000 (14:58 -0400)]
btrfs: rename btrfs_item_end_nr to btrfs_item_data_end

The name btrfs_item_end_nr() is a bit of a misnomer, as it's actually
the offset of the end of the data the item points to.  In fact all of
the helpers that we use btrfs_item_end_nr() use data in their name, like
BTRFS_LEAF_DATA_SIZE() and leaf_data().  Rename to btrfs_item_data_end()
to make it clear what this helper is giving us.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove the btrfs_item_end() helper
Josef Bacik [Thu, 21 Oct 2021 18:58:36 +0000 (14:58 -0400)]
btrfs: remove the btrfs_item_end() helper

We're only using btrfs_item_end() from btrfs_item_end_nr(), so this can
be collapsed.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: drop the _nr from the item helpers
Josef Bacik [Thu, 21 Oct 2021 18:58:35 +0000 (14:58 -0400)]
btrfs: drop the _nr from the item helpers

Now that all call sites are using the slot number to modify item values,
rename the SETGET helpers to raw_item_*(), and then rework the _nr()
helpers to be the btrfs_item_*() btrfs_set_item_*() helpers, and then
rename all of the callers to the new helpers.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: introduce item_nr token variant helpers
Josef Bacik [Thu, 21 Oct 2021 18:58:34 +0000 (14:58 -0400)]
btrfs: introduce item_nr token variant helpers

The last remaining place where we have the pattern of

item = btrfs_item_nr(slot)
<do something with the item>

are the token helpers.  Handle this by introducing token helpers that
will do the btrfs_item_nr() work inside of the helper itself, and then
convert all users of the btrfs_item token helpers to the new _nr()
variants.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: make btrfs_file_extent_inline_item_len take a slot
Josef Bacik [Thu, 21 Oct 2021 18:58:33 +0000 (14:58 -0400)]
btrfs: make btrfs_file_extent_inline_item_len take a slot

Instead of getting the btrfs_item for this, simply pass in the slot of
the item and then use the btrfs_item_size_nr() helper inside of
btrfs_file_extent_inline_item_len().

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: add btrfs_set_item_*_nr() helpers
Josef Bacik [Thu, 21 Oct 2021 18:58:32 +0000 (14:58 -0400)]
btrfs: add btrfs_set_item_*_nr() helpers

We have the pattern of

item = btrfs_item_nr(slot);
btrfs_set_item_*(leaf, item);

in a bunch of places in our code.  Fix this by adding
btrfs_set_item_*_nr() helpers which will do the appropriate work, and
replace those calls with

btrfs_set_item_*_nr(leaf, slot);

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: use btrfs_item_size_nr/btrfs_item_offset_nr everywhere
Josef Bacik [Thu, 21 Oct 2021 18:58:31 +0000 (14:58 -0400)]
btrfs: use btrfs_item_size_nr/btrfs_item_offset_nr everywhere

We have this pattern in a lot of places

item = btrfs_item_nr(slot);
btrfs_item_size(leaf, item);

when we could simply use

btrfs_item_size(leaf, slot);

Fix all callers of btrfs_item_size() and btrfs_item_offset() to use the
_nr variation of the helpers.

Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove no longer needed logic for replaying directory deletes
Filipe Manana [Mon, 25 Oct 2021 16:31:54 +0000 (17:31 +0100)]
btrfs: remove no longer needed logic for replaying directory deletes

Now that we log only dir index keys when logging a directory, we no longer
need to deal with dir item keys in the log replay code for replaying
directory deletes. This is also true for the case when we replay a log
tree created by a kernel that still logs dir items.

So remove the remaining code of the replay of directory deletes algorithm
that deals with dir item keys.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: only copy dir index keys when logging a directory
Filipe Manana [Mon, 25 Oct 2021 16:31:53 +0000 (17:31 +0100)]
btrfs: only copy dir index keys when logging a directory

Currently, when logging a directory, we copy both dir items and dir index
items from the fs/subvolume tree to the log tree. Both items have exactly
the same data (same struct btrfs_dir_item), the difference lies in the key
values, where a dir index key contains the index number of a directory
entry while the dir item key does not, as it's used for doing fast lookups
of an entry by name, while the former is used for sorting entries when
listing a directory.

We can exploit that and log only the dir index items, since they contain
all the information needed to correctly add, replace and delete directory
entries when replaying a log tree. Logging only the dir index items is
also backward and forward compatible: an unpatched kernel (without this
change) can correctly replay a log tree generated by a patched kernel
(with this patch), and a patched kernel can correctly replay a log tree
generated by an unpatched kernel.

The backward compatibility is ensured because:

1) For inserting a new dentry: a dentry is only inserted when we find a
   new dir index key - we can only insert if we know the dir index offset,
   which is encoded in the dir index key's offset;

2) For deleting dentries: during log replay, before adding or replacing
   dentries, we first replay dentry deletions. Whenever we find a dir item
   key or a dir index key in the subvolume/fs tree that is not logged in
   a range for which the log tree is authoritative, we do the unlink of
   the dentry, which removes both the existing dir item key and the dir
   index key. Therefore logging just dir index keys is enough to ensure
   dentry deletions are correctly replayed;

3) For dentry replacements: they work when we log only dir index keys
   and this is mostly due to a combination of 1) and 2). If we replace a
   dentry with name "foobar" to point from inode A to inode B, then we
   know the dir index key for the new dentry is different from the old
   one, as it has an index number (key offset) larger than the old one.
   This results in replaying a deletion, through replay_dir_deletes(),
   that causes the old dentry to be removed, both the dir item key and
   the dir index key, as mentioned at 2). Then when processing the new
   dir index key, we add the new dentry, adding both a new dir item key
   and a new index key pointing to inode B, as stated in 1).

The forward compatibility, the ability for a patched kernel to replay a
log created by an older, unpatched kernel, comes from the changes required
for making sure we are able to replay a log that only contains dir index
keys - we simply ignore every dir item key we find.

So modify directory logging to log only dir index items, and modify the
log replay process to ignore dir item keys, from log trees created by an
unpatched kernel, and process only with dir index keys. This reduces the
amount of logged metadata by about half, and therefore the time spent
logging or fsyncing large directories (less CPU time and less IO).

The following test script was used to measure this change:

   #!/bin/bash

   DEV=/dev/nvme0n1
   MNT=/mnt/nvme0n1

   NUM_NEW_FILES=1000000
   NUM_FILE_DELETES=10000

   mkfs.btrfs -f $DEV
   mount -o ssd $DEV $MNT

   mkdir $MNT/testdir

   for ((i = 1; i <= $NUM_NEW_FILES; i++)); do
           echo -n > $MNT/testdir/file_$i
   done

   start=$(date +%s%N)
   xfs_io -c "fsync" $MNT/testdir
   end=$(date +%s%N)

   dur=$(( (end - start) / 1000000 ))
   echo "dir fsync took $dur ms after adding $NUM_NEW_FILES files"

   # sync to force transaction commit and wipeout the log.
   sync

   del_inc=$(( $NUM_NEW_FILES / $NUM_FILE_DELETES ))
   for ((i = 1; i <= $NUM_NEW_FILES; i += $del_inc)); do
           rm -f $MNT/testdir/file_$i
   done

   start=$(date +%s%N)
   xfs_io -c "fsync" $MNT/testdir
   end=$(date +%s%N)

   dur=$(( (end - start) / 1000000 ))
   echo "dir fsync took $dur ms after deleting $NUM_FILE_DELETES files"
   echo

   umount $MNT

The tests were run on a physical machine, with a non-debug kernel (Debian's
default kernel config), for different values of $NUM_NEW_FILES and
$NUM_FILE_DELETES, and the results were the following:

** Before patch, NUM_NEW_FILES = 1 000 000, NUM_DELETE_FILES = 10 000 **

dir fsync took 8412 ms after adding 1000000 files
dir fsync took 500 ms after deleting 10000 files

** After patch, NUM_NEW_FILES = 1 000 000, NUM_DELETE_FILES = 10 000 **

dir fsync took 4252 ms after adding 1000000 files   (-49.5%)
dir fsync took 269 ms after deleting 10000 files    (-46.2%)

** Before patch, NUM_NEW_FILES = 100 000, NUM_DELETE_FILES = 1 000 **

dir fsync took 745 ms after adding 100000 files
dir fsync took 59 ms after deleting 1000 files

** After patch, NUM_NEW_FILES = 100 000, NUM_DELETE_FILES = 1 000 **

dir fsync took 404 ms after adding 100000 files   (-45.8%)
dir fsync took 31 ms after deleting 1000 files    (-47.5%)

** Before patch, NUM_NEW_FILES = 10 000, NUM_DELETE_FILES = 1 000 **

dir fsync took 67 ms after adding 10000 files
dir fsync took 9 ms after deleting 1000 files

** After patch, NUM_NEW_FILES = 10 000, NUM_DELETE_FILES = 1 000 **

dir fsync took 36 ms after adding 10000 files   (-46.3%)
dir fsync took 5 ms after deleting 1000 files   (-44.4%)

** Before patch, NUM_NEW_FILES = 1 000, NUM_DELETE_FILES = 100 **

dir fsync took 9 ms after adding 1000 files
dir fsync took 4 ms after deleting 100 files

** After patch, NUM_NEW_FILES = 1 000, NUM_DELETE_FILES = 100 **

dir fsync took 7 ms after adding 1000 files     (-22.2%)
dir fsync took 3 ms after deleting 100 files    (-25.0%)

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: remove spurious unlock/lock of unused_bgs_lock
Nikolay Borisov [Thu, 14 Oct 2021 07:03:11 +0000 (10:03 +0300)]
btrfs: remove spurious unlock/lock of unused_bgs_lock

Since both unused block groups and reclaim bgs lists are protected by
unused_bgs_lock then free them in the same critical section without
doing an extra unlock/lock pair.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: fix deadlock between quota enable and other quota operations
Filipe Manana [Wed, 27 Oct 2021 17:30:25 +0000 (18:30 +0100)]
btrfs: fix deadlock between quota enable and other quota operations

When enabling quotas, we attempt to commit a transaction while holding the
mutex fs_info->qgroup_ioctl_lock. This can result on a deadlock with other
quota operations such as:

- qgroup creation and deletion, ioctl BTRFS_IOC_QGROUP_CREATE;

- adding and removing qgroup relations, ioctl BTRFS_IOC_QGROUP_ASSIGN.

This is because these operations join a transaction and after that they
attempt to lock the mutex fs_info->qgroup_ioctl_lock. Acquiring that mutex
after joining or starting a transaction is a pattern followed everywhere
in qgroups, so the quota enablement operation is the one at fault here,
and should not commit a transaction while holding that mutex.

Fix this by making the transaction commit while not holding the mutex.
We are safe from two concurrent tasks trying to enable quotas because
we are serialized by the rw semaphore fs_info->subvol_sem at
btrfs_ioctl_quota_ctl(), which is the only call site for enabling
quotas.

When this deadlock happens, it produces a trace like the following:

  INFO: task syz-executor:25604 blocked for more than 143 seconds.
  Not tainted 5.15.0-rc6 #4
  "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  task:syz-executor state:D stack:24800 pid:25604 ppid: 24873 flags:0x00004004
  Call Trace:
  context_switch kernel/sched/core.c:4940 [inline]
  __schedule+0xcd9/0x2530 kernel/sched/core.c:6287
  schedule+0xd3/0x270 kernel/sched/core.c:6366
  btrfs_commit_transaction+0x994/0x2e90 fs/btrfs/transaction.c:2201
  btrfs_quota_enable+0x95c/0x1790 fs/btrfs/qgroup.c:1120
  btrfs_ioctl_quota_ctl fs/btrfs/ioctl.c:4229 [inline]
  btrfs_ioctl+0x637e/0x7b70 fs/btrfs/ioctl.c:5010
  vfs_ioctl fs/ioctl.c:51 [inline]
  __do_sys_ioctl fs/ioctl.c:874 [inline]
  __se_sys_ioctl fs/ioctl.c:860 [inline]
  __x64_sys_ioctl+0x193/0x200 fs/ioctl.c:860
  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
  do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
  entry_SYSCALL_64_after_hwframe+0x44/0xae
  RIP: 0033:0x7f86920b2c4d
  RSP: 002b:00007f868f61ac58 EFLAGS: 00000246 ORIG_RAX: 0000000000000010
  RAX: ffffffffffffffda RBX: 00007f86921d90a0 RCX: 00007f86920b2c4d
  RDX: 0000000020005e40 RSI: 00000000c0109428 RDI: 0000000000000008
  RBP: 00007f869212bd80 R08: 0000000000000000 R09: 0000000000000000
  R10: 0000000000000000 R11: 0000000000000246 R12: 00007f86921d90a0
  R13: 00007fff6d233e4f R14: 00007fff6d233ff0 R15: 00007f868f61adc0
  INFO: task syz-executor:25628 blocked for more than 143 seconds.
  Not tainted 5.15.0-rc6 #4
  "echo 0 > /proc/sys/kernel/hung_task_timeout_secs" disables this message.
  task:syz-executor state:D stack:29080 pid:25628 ppid: 24873 flags:0x00004004
  Call Trace:
  context_switch kernel/sched/core.c:4940 [inline]
  __schedule+0xcd9/0x2530 kernel/sched/core.c:6287
  schedule+0xd3/0x270 kernel/sched/core.c:6366
  schedule_preempt_disabled+0xf/0x20 kernel/sched/core.c:6425
  __mutex_lock_common kernel/locking/mutex.c:669 [inline]
  __mutex_lock+0xc96/0x1680 kernel/locking/mutex.c:729
  btrfs_remove_qgroup+0xb7/0x7d0 fs/btrfs/qgroup.c:1548
  btrfs_ioctl_qgroup_create fs/btrfs/ioctl.c:4333 [inline]
  btrfs_ioctl+0x683c/0x7b70 fs/btrfs/ioctl.c:5014
  vfs_ioctl fs/ioctl.c:51 [inline]
  __do_sys_ioctl fs/ioctl.c:874 [inline]
  __se_sys_ioctl fs/ioctl.c:860 [inline]
  __x64_sys_ioctl+0x193/0x200 fs/ioctl.c:860
  do_syscall_x64 arch/x86/entry/common.c:50 [inline]
  do_syscall_64+0x35/0xb0 arch/x86/entry/common.c:80
  entry_SYSCALL_64_after_hwframe+0x44/0xae

Reported-by: Hao Sun <sunhao.th@gmail.com>
Link: https://lore.kernel.org/linux-btrfs/CACkBjsZQF19bQ1C6=yetF3BvL10OSORpFUcWXTP6HErshDB4dQ@mail.gmail.com/
Fixes: 340f1aa27f36 ("btrfs: qgroups: Move transaction management inside btrfs_quota_enable/disable")
CC: stable@vger.kernel.org # 4.19
Reviewed-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agobtrfs: fix ENOSPC failure when attempting direct IO write into NOCOW range
Filipe Manana [Thu, 28 Oct 2021 15:03:41 +0000 (16:03 +0100)]
btrfs: fix ENOSPC failure when attempting direct IO write into NOCOW range

When doing a direct IO write against a file range that either has
preallocated extents in that range or has regular extents and the file
has the NOCOW attribute set, the write fails with -ENOSPC when all of
the following conditions are met:

1) There are no data blocks groups with enough free space matching
   the size of the write;

2) There's not enough unallocated space for allocating a new data block
   group;

3) The extents in the target file range are not shared, neither through
   snapshots nor through reflinks.

This is wrong because a NOCOW write can be done in such case, and in fact
it's possible to do it using a buffered IO write, since when failing to
allocate data space, the buffered IO path checks if a NOCOW write is
possible.

The failure in direct IO write path comes from the fact that early on,
at btrfs_dio_iomap_begin(), we try to allocate data space for the write
and if it that fails we return the error and stop - we never check if we
can do NOCOW. But later, at btrfs_get_blocks_direct_write(), we check
if we can do a NOCOW write into the range, or a subset of the range, and
then release the previously reserved data space.

Fix this by doing the data reservation only if needed, when we must COW,
at btrfs_get_blocks_direct_write() instead of doing it at
btrfs_dio_iomap_begin(). This also simplifies a bit the logic and removes
the inneficiency of doing unnecessary data reservations.

The following example test script reproduces the problem:

  $ cat dio-nocow-enospc.sh
  #!/bin/bash

  DEV=/dev/sdj
  MNT=/mnt/sdj

  # Use a small fixed size (1G) filesystem so that it's quick to fill
  # it up.
  # Make sure the mixed block groups feature is not enabled because we
  # later want to not have more space available for allocating data
  # extents but still have enough metadata space free for the file writes.
  mkfs.btrfs -f -b $((1024 * 1024 * 1024)) -O ^mixed-bg $DEV
  mount $DEV $MNT

  # Create our test file with the NOCOW attribute set.
  touch $MNT/foobar
  chattr +C $MNT/foobar

  # Now fill in all unallocated space with data for our test file.
  # This will allocate a data block group that will be full and leave
  # no (or a very small amount of) unallocated space in the device, so
  # that it will not be possible to allocate a new block group later.
  echo
  echo "Creating test file with initial data..."
  xfs_io -c "pwrite -S 0xab -b 1M 0 900M" $MNT/foobar

  # Now try a direct IO write against file range [0, 10M[.
  # This should succeed since this is a NOCOW file and an extent for the
  # range was previously allocated.
  echo
  echo "Trying direct IO write over allocated space..."
  xfs_io -d -c "pwrite -S 0xcd -b 10M 0 10M" $MNT/foobar

  umount $MNT

When running the test:

  $ ./dio-nocow-enospc.sh
  (...)

  Creating test file with initial data...
  wrote 943718400/943718400 bytes at offset 0
  900 MiB, 900 ops; 0:00:01.43 (625.526 MiB/sec and 625.5265 ops/sec)

  Trying direct IO write over allocated space...
  pwrite: No space left on device

A test case for fstests will follow, testing both this direct IO write
scenario as well as the buffered IO write scenario to make it less likely
to get future regressions on the buffered IO case.

Reviewed-by: Josef Bacik <josef@toxicpanda.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
2 years agoLinux 5.16-rc8 v5.16-rc8
Linus Torvalds [Sun, 2 Jan 2022 22:23:25 +0000 (14:23 -0800)]
Linux 5.16-rc8

2 years agoMerge tag 'perf-tools-fixes-for-v5.16-2022-01-02' of git://git.kernel.org/pub/scm...
Linus Torvalds [Sun, 2 Jan 2022 22:09:03 +0000 (14:09 -0800)]
Merge tag 'perf-tools-fixes-for-v5.16-2022-01-02' of git://git./linux/kernel/git/acme/linux

Pull perf tools fixes from Arnaldo Carvalho de Melo:

 - Fix TUI exit screen refresh race condition in 'perf top'.

 - Fix parsing of Intel PT VM time correlation arguments.

 - Honour CPU filtering command line request of a script's switch events
   in 'perf script'.

 - Fix printing of switch events in Intel PT python script.

 - Fix duplicate alias events list printing in 'perf list', noticed on
   heterogeneous arm64 systems.

 - Fix return value of ids__new(), users expect NULL for failure, not
   ERR_PTR(-ENOMEM).

* tag 'perf-tools-fixes-for-v5.16-2022-01-02' of git://git.kernel.org/pub/scm/linux/kernel/git/acme/linux:
  perf top: Fix TUI exit screen refresh race condition
  perf pmu: Fix alias events list
  perf scripts python: intel-pt-events.py: Fix printing of switch events
  perf script: Fix CPU filtering of a script's switch events
  perf intel-pt: Fix parsing of VM time correlation arguments
  perf expr: Fix return value of ids__new()

2 years agoMerge branch 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa...
Linus Torvalds [Sun, 2 Jan 2022 18:36:09 +0000 (10:36 -0800)]
Merge branch 'i2c/for-current' of git://git./linux/kernel/git/wsa/linux

Pull i2c fixes from Wolfram Sang:
 "Better input validation for compat ioctls and a documentation bugfix
  for 5.16"

* 'i2c/for-current' of git://git.kernel.org/pub/scm/linux/kernel/git/wsa/linux:
  Docs: Fixes link to I2C specification
  i2c: validate user data in compat ioctl

2 years agoMerge tag 'x86_urgent_for_v5.16_rc8' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sun, 2 Jan 2022 17:02:54 +0000 (09:02 -0800)]
Merge tag 'x86_urgent_for_v5.16_rc8' of git://git./linux/kernel/git/tip/tip

Pull x86 fix from Borislav Petkov:

 - Use the proper CONFIG symbol in a preprocessor check.

* tag 'x86_urgent_for_v5.16_rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/build: Use the proper name CONFIG_FW_LOADER

2 years agoperf top: Fix TUI exit screen refresh race condition
yaowenbin [Wed, 29 Dec 2021 08:55:19 +0000 (16:55 +0800)]
perf top: Fix TUI exit screen refresh race condition

When the following command is executed several times, a coredump file is
generated.

$ timeout -k 9 5 perf top -e task-clock
*******
*******
*******
0.01%  [kernel]                  [k] __do_softirq
0.01%  libpthread-2.28.so        [.] __pthread_mutex_lock
0.01%  [kernel]                  [k] __ll_sc_atomic64_sub_return
double free or corruption (!prev) perf top --sort comm,dso
timeout: the monitored command dumped core

When we terminate "perf top" using sending signal method,
SLsmg_reset_smg() called. SLsmg_reset_smg() resets the SLsmg screen
management routines by freeing all memory allocated while it was active.

However SLsmg_reinit_smg() maybe be called by another thread.

SLsmg_reinit_smg() will free the same memory accessed by
SLsmg_reset_smg(), thus it results in a double free.

SLsmg_reinit_smg() is called already protected by ui__lock, so we fix
the problem by adding pthread_mutex_trylock of ui__lock when calling
SLsmg_reset_smg().

Signed-off-by: Wenyu Liu <liuwenyu7@huawei.com>
Tested-by: Arnaldo Carvalho de Melo <acme@redhat.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: wuxu.wu@huawei.com
Link: http://lore.kernel.org/lkml/a91e3943-7ddc-f5c0-a7f5-360f073c20e6@huawei.com
Signed-off-by: Hewenliang <hewenliang4@huawei.com>
Signed-off-by: yaowenbin <yaowenbin1@huawei.com>
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf pmu: Fix alias events list
John Garry [Tue, 21 Dec 2021 16:11:30 +0000 (00:11 +0800)]
perf pmu: Fix alias events list

Commit 0e0ae8742207c3b4 ("perf list: Display hybrid PMU events with cpu
type") changes the event list for uncore PMUs or arm64 heterogeneous CPU
systems, such that duplicate aliases are incorrectly listed per PMU
(which they should not be), like:

  # perf list
  ...
  unc_cbo_cache_lookup.any_es
  [Unit: uncore_cbox L3 Lookup any request that access cache and found
  line in E or S-state]
  unc_cbo_cache_lookup.any_es
  [Unit: uncore_cbox L3 Lookup any request that access cache and found
  line in E or S-state]
  unc_cbo_cache_lookup.any_i
  [Unit: uncore_cbox L3 Lookup any request that access cache and found
  line in I-state]
  unc_cbo_cache_lookup.any_i
  [Unit: uncore_cbox L3 Lookup any request that access cache and found
  line in I-state]
  ...

Notice how the events are listed twice.

The named commit changed how we remove duplicate events, in that events
for different PMUs are not treated as duplicates. I suppose this is to
handle how "Each hybrid pmu event has been assigned with a pmu name".

Fix PMU alias listing by restoring behaviour to remove duplicates for
non-hybrid PMUs.

Fixes: 0e0ae8742207c3b4 ("perf list: Display hybrid PMU events with cpu type")
Signed-off-by: John Garry <john.garry@huawei.com>
Tested-by: Zhengjun Xing <zhengjun.xing@linux.intel.com>
Cc: Alexander Shishkin <alexander.shishkin@linux.intel.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Kan Liang <kan.liang@linux.intel.com>
Cc: Mark Rutland <mark.rutland@arm.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Link: https://lore.kernel.org/r/1640103090-140490-1-git-send-email-john.garry@huawei.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 1 Jan 2022 18:21:49 +0000 (10:21 -0800)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Two small fixups for spaceball joystick driver and appletouch touchpad
  driver"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: spaceball - fix parsing of movement data packets
  Input: appletouch - initialize work before device registration

2 years agomm: vmscan: reduce throttling due to a failure to make progress -fix
Mel Gorman [Fri, 31 Dec 2021 21:10:09 +0000 (13:10 -0800)]
mm: vmscan: reduce throttling due to a failure to make progress -fix

Hugh Dickins reported the following

My tmpfs swapping load (tweaked to use huge pages more heavily
than in real life) is far from being a realistic load: but it was
notably slowed down by your throttling mods in 5.16-rc, and this
patch makes it well again - thanks.

But: it very quickly hit NULL pointer until I changed that last
line to

        if (first_pgdat)
                consider_reclaim_throttle(first_pgdat, sc);

The likely issue is that huge pages are a major component of the test
workload.  When this is the case, first_pgdat may never get set if
compaction is ready to continue due to this check

        if (IS_ENABLED(CONFIG_COMPACTION) &&
            sc->order > PAGE_ALLOC_COSTLY_ORDER &&
            compaction_ready(zone, sc)) {
                sc->compaction_ready = true;
                continue;
        }

If this was true for every zone in the zonelist, first_pgdat would never
get set resulting in a NULL pointer exception.

Link: https://lkml.kernel.org/r/20211209095453.GM3366@techsingularity.net
Fixes: 1b4e3f26f9f75 ("mm: vmscan: Reduce throttling due to a failure to make progress")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Hugh Dickins <hughd@google.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Vlastimil Babka <vbabka@suse.cz>
Cc: Rik van Riel <riel@surriel.com>
Cc: Mike Galbraith <efault@gmx.de>
Cc: Darrick J. Wong <djwong@kernel.org>
Cc: Shakeel Butt <shakeelb@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agomm: vmscan: Reduce throttling due to a failure to make progress
Mel Gorman [Thu, 2 Dec 2021 15:06:14 +0000 (15:06 +0000)]
mm: vmscan: Reduce throttling due to a failure to make progress

Mike Galbraith, Alexey Avramov and Darrick Wong all reported similar
problems due to reclaim throttling for excessive lengths of time.  In
Alexey's case, a memory hog that should go OOM quickly stalls for
several minutes before stalling.  In Mike and Darrick's cases, a small
memcg environment stalled excessively even though the system had enough
memory overall.

Commit 69392a403f49 ("mm/vmscan: throttle reclaim when no progress is
being made") introduced the problem although commit a19594ca4a8b
("mm/vmscan: increase the timeout if page reclaim is not making
progress") made it worse.  Systems at or near an OOM state that cannot
be recovered must reach OOM quickly and memcg should kill tasks if a
memcg is near OOM.

To address this, only stall for the first zone in the zonelist, reduce
the timeout to 1 tick for VMSCAN_THROTTLE_NOPROGRESS and only stall if
the scan control nr_reclaimed is 0, kswapd is still active and there
were excessive pages pending for writeback.  If kswapd has stopped
reclaiming due to excessive failures, do not stall at all so that OOM
triggers relatively quickly.  Similarly, if an LRU is simply congested,
only lightly throttle similar to NOPROGRESS.

Alexey's original case was the most straight forward

for i in {1..3}; do tail /dev/zero; done

On vanilla 5.16-rc1, this test stalled heavily, after the patch the test
completes in a few seconds similar to 5.15.

Alexey's second test case added watching a youtube video while tail runs
10 times.  On 5.15, playback only jitters slightly, 5.16-rc1 stalls a
lot with lots of frames missing and numerous audio glitches.  With this
patch applies, the video plays similarly to 5.15.

[lkp@intel.com: Fix W=1 build warning]

Link: https://lore.kernel.org/r/99e779783d6c7fce96448a3402061b9dc1b3b602.camel@gmx.de
Link: https://lore.kernel.org/r/20211124011954.7cab9bb4@mail.inbox.lv
Link: https://lore.kernel.org/r/20211022144651.19914-1-mgorman@techsingularity.net
Link: https://lore.kernel.org/r/20211202150614.22440-1-mgorman@techsingularity.net
Link: https://linux-regtracking.leemhuis.info/regzbot/regression/20211124011954.7cab9bb4@mail.inbox.lv/
Reported-and-tested-by: Alexey Avramov <hakavlad@inbox.lv>
Reported-and-tested-by: Mike Galbraith <efault@gmx.de>
Reported-and-tested-by: Darrick J. Wong <djwong@kernel.org>
Reported-by: kernel test robot <lkp@intel.com>
Acked-by: Hugh Dickins <hughd@google.com>
Tracked-by: Thorsten Leemhuis <regressions@leemhuis.info>
Fixes: 69392a403f49 ("mm/vmscan: throttle reclaim when no progress is being made")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Fri, 31 Dec 2021 17:28:48 +0000 (09:28 -0800)]
Merge branch 'akpm' (patches from Andrew)

Merge misc mm fixes from Andrew Morton:
 "2 patches.

  Subsystems affected by this patch series: mm (userfaultfd and damon)"

* akpm:
  mm/damon/dbgfs: fix 'struct pid' leaks in 'dbgfs_target_ids_write()'
  userfaultfd/selftests: fix hugetlb area allocations

2 years agoMerge tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi
Linus Torvalds [Fri, 31 Dec 2021 17:22:25 +0000 (09:22 -0800)]
Merge tag 'scsi-fixes' of git://git./linux/kernel/git/jejb/scsi

Pull SCSI fixes from James Bottomley:
 "Three fixes, all in drivers. The lpfc one doesn't look exploitable,
  but nasty things could happen in string operations if mybuf ends up
  with an on stack unterminated string"

* tag 'scsi-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/jejb/scsi:
  scsi: vmw_pvscsi: Set residual data length conditionally
  scsi: libiscsi: Fix UAF in iscsi_conn_get_param()/iscsi_conn_teardown()
  scsi: lpfc: Terminate string in lpfc_debugfs_nvmeio_trc_write()

2 years agomm/damon/dbgfs: fix 'struct pid' leaks in 'dbgfs_target_ids_write()'
SeongJae Park [Fri, 31 Dec 2021 04:12:34 +0000 (20:12 -0800)]
mm/damon/dbgfs: fix 'struct pid' leaks in 'dbgfs_target_ids_write()'

DAMON debugfs interface increases the reference counts of 'struct pid's
for targets from the 'target_ids' file write callback
('dbgfs_target_ids_write()'), but decreases the counts only in DAMON
monitoring termination callback ('dbgfs_before_terminate()').

Therefore, when 'target_ids' file is repeatedly written without DAMON
monitoring start/termination, the reference count is not decreased and
therefore memory for the 'struct pid' cannot be freed.  This commit
fixes this issue by decreasing the reference counts when 'target_ids' is
written.

Link: https://lkml.kernel.org/r/20211229124029.23348-1-sj@kernel.org
Fixes: 4bc05954d007 ("mm/damon: implement a debugfs-based user space interface")
Signed-off-by: SeongJae Park <sj@kernel.org>
Cc: <stable@vger.kernel.org> [5.15+]
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agouserfaultfd/selftests: fix hugetlb area allocations
Mike Kravetz [Fri, 31 Dec 2021 04:12:31 +0000 (20:12 -0800)]
userfaultfd/selftests: fix hugetlb area allocations

Currently, userfaultfd selftest for hugetlb as run from run_vmtests.sh
or any environment where there are 'just enough' hugetlb pages will
always fail with:

  testing events (fork, remap, remove):
ERROR: UFFDIO_COPY error: -12 (errno=12, line=616)

The ENOMEM error code implies there are not enough hugetlb pages.
However, there are free hugetlb pages but they are all reserved.  There
is a basic problem with the way the test allocates hugetlb pages which
has existed since the test was originally written.

Due to the way 'cleanup' was done between different phases of the test,
this issue was masked until recently.  The issue was uncovered by commit
8ba6e8640844 ("userfaultfd/selftests: reinitialize test context in each
test").

For the hugetlb test, src and dst areas are allocated as PRIVATE
mappings of a hugetlb file.  This means that at mmap time, pages are
reserved for the src and dst areas.  At the start of event testing (and
other tests) the src area is populated which results in allocation of
huge pages to fill the area and consumption of reserves associated with
the area.  Then, a child is forked to fault in the dst area.  Note that
the dst area was allocated in the parent and hence the parent owns the
reserves associated with the mapping.  The child has normal access to
the dst area, but can not use the reserves created/owned by the parent.
Thus, if there are no other huge pages available allocation of a page
for the dst by the child will fail.

Fix by not creating reserves for the dst area.  In this way the child
can use free (non-reserved) pages.

Also, MAP_PRIVATE of a file only makes sense if you are interested in
the contents of the file before making a COW copy.  The test does not do
this.  So, just use MAP_ANONYMOUS | MAP_HUGETLB to create an anonymous
hugetlb mapping.  There is no need to create a hugetlb file in the
non-shared case.

Link: https://lkml.kernel.org/r/20211217172919.7861-1-mike.kravetz@oracle.com
Signed-off-by: Mike Kravetz <mike.kravetz@oracle.com>
Cc: Axel Rasmussen <axelrasmussen@google.com>
Cc: Peter Xu <peterx@redhat.com>
Cc: Andrea Arcangeli <aarcange@redhat.com>
Cc: Mina Almasry <almasrymina@google.com>
Cc: Shuah Khan <shuah@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoDocs: Fixes link to I2C specification
Deep Majumder [Fri, 19 Nov 2021 06:14:01 +0000 (11:44 +0530)]
Docs: Fixes link to I2C specification

The link to the I2C specification is broken. Although
"https://www.nxp.com" hosts Rev 7 (2021) of this specification, it is
behind a login-wall. Thus, an additional link has been added (which
doesn't require a login) and the NXP official docs link has been
updated.

Signed-off-by: Deep Majumder <deep@fastmail.in>
[wsa: minor updates to text and commit message]
Signed-off-by: Wolfram Sang <wsa@kernel.org>
2 years agoi2c: validate user data in compat ioctl
Pavel Skripkin [Thu, 30 Dec 2021 22:47:50 +0000 (01:47 +0300)]
i2c: validate user data in compat ioctl

Wrong user data may cause warning in i2c_transfer(), ex: zero msgs.
Userspace should not be able to trigger warnings, so this patch adds
validation checks for user data in compact ioctl to prevent reported
warnings

Reported-and-tested-by: syzbot+e417648b303855b91d8a@syzkaller.appspotmail.com
Fixes: 7d5cb45655f2 ("i2c compat ioctls: move to ->compat_ioctl()")
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Signed-off-by: Wolfram Sang <wsa@kernel.org>
2 years agoInput: spaceball - fix parsing of movement data packets
Leo L. Schwab [Fri, 31 Dec 2021 05:05:00 +0000 (21:05 -0800)]
Input: spaceball - fix parsing of movement data packets

The spaceball.c module was not properly parsing the movement reports
coming from the device.  The code read axis data as signed 16-bit
little-endian values starting at offset 2.

In fact, axis data in Spaceball movement reports are signed 16-bit
big-endian values starting at offset 3.  This was determined first by
visually inspecting the data packets, and later verified by consulting:
http://spacemice.org/pdf/SpaceBall_2003-3003_Protocol.pdf

If this ever worked properly, it was in the time before Git...

Signed-off-by: Leo L. Schwab <ewhac@ewhac.org>
Link: https://lore.kernel.org/r/20211221101630.1146385-1-ewhac@ewhac.org
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
2 years agoInput: appletouch - initialize work before device registration
Pavel Skripkin [Fri, 31 Dec 2021 04:57:46 +0000 (20:57 -0800)]
Input: appletouch - initialize work before device registration

Syzbot has reported warning in __flush_work(). This warning is caused by
work->func == NULL, which means missing work initialization.

This may happen, since input_dev->close() calls
cancel_work_sync(&dev->work), but dev->work initalization happens _after_
input_register_device() call.

So this patch moves dev->work initialization before registering input
device

Fixes: 5a6eb676d3bc ("Input: appletouch - improve powersaving for Geyser3 devices")
Reported-and-tested-by: syzbot+b88c5eae27386b252bbd@syzkaller.appspotmail.com
Signed-off-by: Pavel Skripkin <paskripkin@gmail.com>
Link: https://lore.kernel.org/r/20211230141151.17300-1-paskripkin@gmail.com
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
2 years agoMerge tag 'drm-fixes-2021-12-31' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 31 Dec 2021 02:25:43 +0000 (18:25 -0800)]
Merge tag 'drm-fixes-2021-12-31' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "This is a bit bigger than I'd like, however it has two weeks of amdgpu
  fixes in it, since they missed last week, which was very small.

  The nouveau regression is probably the biggest fix in here, and it
  needs to go into 5.15 as well, two i915 fixes, and then a scattering
  of amdgpu fixes. The biggest fix in there is for a fencing NULL
  pointer dereference, the rest are pretty minor.

  For the misc team, I've pulled the two misc fixes manually since I'm
  not sure what is happening at this time of year!

  The amdgpu maintainers have the outstanding runpm regression to fix
  still, they are just working through the last bits of it now.

  Summary:

  nouveau:
   - fencing regression fix

  i915:
   - Fix possible uninitialized variable
   - Fix composite fence seqno icrement on each fence creation

  amdgpu:
   - Fencing fix
   - XGMI fix
   - VCN regression fix
   - IP discovery regression fixes
   - Fix runpm documentation
   - Suspend/resume fixes
   - Yellow Carp display fixes
   - MCLK power management fix
   - dma-buf fix"

* tag 'drm-fixes-2021-12-31' of git://anongit.freedesktop.org/drm/drm:
  drm/amd/display: Changed pipe split policy to allow for multi-display pipe split
  drm/amd/display: Fix USB4 null pointer dereference in update_psp_stream_config
  drm/amd/display: Set optimize_pwr_state for DCN31
  drm/amd/display: Send s0i2_rdy in stream_count == 0 optimization
  drm/amd/display: Added power down for DCN10
  drm/amd/display: fix B0 TMDS deepcolor no dislay issue
  drm/amdgpu: no DC support for headless chips
  drm/amdgpu: put SMU into proper state on runpm suspending for BOCO capable platform
  drm/amdgpu: always reset the asic in suspend (v2)
  drm/amd/pm: skip setting gfx cgpg in the s0ix suspend-resume
  drm/i915: Increment composite fence seqno
  drm/i915: Fix possible uninitialized variable in parallel extension
  drm/amdgpu: fix runpm documentation
  drm/nouveau: wait for the exclusive fence after the shared ones v2
  drm/amdgpu: add support for IP discovery gc_info table v2
  drm/amdgpu: When the VCN(1.0) block is suspended, powergating is explicitly enabled
  drm/amd/pm: Fix xgmi link control on aldebaran
  drm/amdgpu: introduce new amdgpu_fence object to indicate the job embedded fence
  drm/amdgpu: fix dropped backing store handling in amdgpu_dma_buf_move_notify

2 years agoMerge branch 'drm-misc-fixes' of ssh://git.freedesktop.org/git/drm/drm-misc into...
Dave Airlie [Fri, 31 Dec 2021 01:40:29 +0000 (11:40 +1000)]
Merge branch 'drm-misc-fixes' of ssh://git.freedesktop.org/git/drm/drm-misc into drm-fixes

This merges two fixes that haven't been sent to me yet, but I wanted to get in.

One amdgpu fix, but one nouveau regression fixer.

Signed-off-by: Dave Airlie <airlied@redhat.com>
2 years agofs/mount_setattr: always cleanup mount_kattr
Christian Brauner [Thu, 30 Dec 2021 19:23:09 +0000 (20:23 +0100)]
fs/mount_setattr: always cleanup mount_kattr

Make sure that finish_mount_kattr() is called after mount_kattr was
succesfully built in both the success and failure case to prevent
leaking any references we took when we built it.  We returned early if
path lookup failed thereby risking to leak an additional reference we
took when building mount_kattr when an idmapped mount was requested.

Cc: linux-fsdevel@vger.kernel.org
Cc: stable@vger.kernel.org
Fixes: 9caccd41541a ("fs: introduce MOUNT_ATTR_IDMAP")
Signed-off-by: Christian Brauner <christian.brauner@ubuntu.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
2 years agoMerge tag 'net-5.16-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net
Linus Torvalds [Thu, 30 Dec 2021 19:12:12 +0000 (11:12 -0800)]
Merge tag 'net-5.16-rc8' of git://git./linux/kernel/git/netdev/net

Pull networking fixes from Jakub Kicinski:
 "Including fixes from.. Santa?

  No regressions on our radar at this point. The igc problem fixed here
  was the last one I was tracking but it was broken in previous
  releases, anyway. Mostly driver fixes and a couple of largish SMC
  fixes.

  Current release - regressions:

   - xsk: initialise xskb free_list_node, fixup for a -rc7 fix

  Current release - new code bugs:

   - mlx5: handful of minor fixes:

   - use first online CPU instead of hard coded CPU

   - fix some error handling paths in 'mlx5e_tc_add_fdb_flow()'

   - fix skb memory leak when TC classifier action offloads are disabled

   - fix memory leak with rules with internal OvS port

  Previous releases - regressions:

   - igc: do not enable crosstimestamping for i225-V models

  Previous releases - always broken:

   - udp: use datalen to cap ipv6 udp max gso segments

   - fix use-after-free in tw_timer_handler due to early free of stats

   - smc: fix kernel panic caused by race of smc_sock

   - smc: don't send CDC/LLC message if link not ready, avoid timeouts

   - sctp: use call_rcu to free endpoint, avoid UAF in sock diag

   - bridge: mcast: add and enforce query interval minimum

   - usb: pegasus: do not drop long Ethernet frames

   - mlx5e: fix ICOSQ recovery flow for XSK

   - nfc: uapi: use kernel size_t to fix user-space builds"

* tag 'net-5.16-rc8' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (47 commits)
  fsl/fman: Fix missing put_device() call in fman_port_probe
  selftests: net: using ping6 for IPv6 in udpgro_fwd.sh
  Documentation: fix outdated interpretation of ip_no_pmtu_disc
  net/ncsi: check for error return from call to nla_put_u32
  net: bridge: mcast: fix br_multicast_ctx_vlan_global_disabled helper
  net: fix use-after-free in tw_timer_handler
  selftests: net: Fix a typo in udpgro_fwd.sh
  selftests/net: udpgso_bench_tx: fix dst ip argument
  net: bridge: mcast: add and enforce startup query interval minimum
  net: bridge: mcast: add and enforce query interval minimum
  ipv6: raw: check passed optlen before reading
  xsk: Initialise xskb free_list_node
  net/mlx5e: Fix wrong features assignment in case of error
  net/mlx5e: TC, Fix memory leak with rules with internal port
  ionic: Initialize the 'lif->dbid_inuse' bitmap
  igc: Fix TX timestamp support for non-MSI-X platforms
  igc: Do not enable crosstimestamping for i225-V models
  net/smc: fix kernel panic caused by race of smc_sock
  net/smc: don't send CDC/LLC message if link not ready
  NFC: st21nfca: Fix memory leak in device probe and remove
  ...

2 years agoMerge tag 'char-misc-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh...
Linus Torvalds [Thu, 30 Dec 2021 17:52:32 +0000 (09:52 -0800)]
Merge tag 'char-misc-5.16' of git://git./linux/kernel/git/gregkh/char-misc

Pull char/misc fixes from Greg KH:
 "Here are two misc driver fixes for 5.16-final:

   - binder accounting fix to resolve reported problem

   - nitro_enclaves fix for mmap assert warning output

  Both of these have been for over a week with no reported issues"

* tag 'char-misc-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/char-misc:
  nitro_enclaves: Use get_user_pages_unlocked() call to handle mmap assert
  binder: fix async_free_space accounting for empty parcels

2 years agoMerge tag 'usb-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb
Linus Torvalds [Thu, 30 Dec 2021 17:49:54 +0000 (09:49 -0800)]
Merge tag 'usb-5.16' of git://git./linux/kernel/git/gregkh/usb

Pull USB fixes from Greg KH:
 "Here are some small USB driver fixes for 5.16 to resolve some reported
  problems:

   - mtu3 driver fixes

   - typec ucsi driver fix

   - xhci driver quirk added

   - usb gadget f_fs fix for reported crash

  All of these have been in linux-next for a while with no reported
  problems"

* tag 'usb-5.16' of git://git.kernel.org/pub/scm/linux/kernel/git/gregkh/usb:
  usb: typec: ucsi: Only check the contract if there is a connection
  xhci: Fresco FL1100 controller should not have BROKEN_MSI quirk set.
  usb: mtu3: set interval of FS intr and isoc endpoint
  usb: mtu3: fix list_head check warning
  usb: mtu3: add memory barrier before set GPD's HWO
  usb: mtu3: fix interval value for intr and isoc
  usb: gadget: f_fs: Clear ffs_eventfd in ffs_data_clear.

2 years agofsl/fman: Fix missing put_device() call in fman_port_probe
Miaoqian Lin [Thu, 30 Dec 2021 12:26:27 +0000 (12:26 +0000)]
fsl/fman: Fix missing put_device() call in fman_port_probe

The reference taken by 'of_find_device_by_node()' must be released when
not needed anymore.
Add the corresponding 'put_device()' in the and error handling paths.

Fixes: 18a6c85fcc78 ("fsl/fman: Add FMan Port Support")
Signed-off-by: Miaoqian Lin <linmq006@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoselftests: net: using ping6 for IPv6 in udpgro_fwd.sh
Jianguo Wu [Thu, 30 Dec 2021 10:40:29 +0000 (18:40 +0800)]
selftests: net: using ping6 for IPv6 in udpgro_fwd.sh

udpgro_fwd.sh output following message:
  ping: 2001:db8:1::100: Address family for hostname not supported

Using ping6 when pinging IPv6 addresses.

Fixes: a062260a9d5f ("selftests: net: add UDP GRO forwarding self-tests")
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoDocumentation: fix outdated interpretation of ip_no_pmtu_disc
xu xin [Thu, 30 Dec 2021 03:28:56 +0000 (03:28 +0000)]
Documentation: fix outdated interpretation of ip_no_pmtu_disc

The updating way of pmtu has changed, but documentation is still in the
old way. So this patch updates the interpretation of ip_no_pmtu_disc and
min_pmtu.

See commit 28d35bcdd3925 ("net: ipv4: don't let PMTU updates increase
route MTU")

Reported-by: Zeal Robot <zealci@zte.com.cn>
Signed-off-by: xu xin <xu.xin16@zte.com.cn>
Signed-off-by: David S. Miller <davem@davemloft.net>
2 years agoMerge tag 'amd-drm-fixes-5.16-2021-12-29' of https://gitlab.freedesktop.org/agd5f...
Dave Airlie [Thu, 30 Dec 2021 03:55:47 +0000 (13:55 +1000)]
Merge tag 'amd-drm-fixes-5.16-2021-12-29' of https://gitlab.freedesktop.org/agd5f/linux into drm-fixes

amd-drm-fixes-5.16-2021-12-29:

amdgpu:
- Fencing fix
- XGMI fix
- VCN regression fix
- IP discovery regression fixes
- Fix runpm documentation
- Suspend/resume fixes
- Yellow Carp display fixes
- MCLK power management fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexander.deucher@amd.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20211229155129.5789-1-alexander.deucher@amd.com
2 years agoMerge tag 'mlx5-fixes-2021-12-28' of git://git.kernel.org/pub/scm/linux/kernel/git...
Jakub Kicinski [Thu, 30 Dec 2021 02:19:01 +0000 (18:19 -0800)]
Merge tag 'mlx5-fixes-2021-12-28' of git://git./linux/kernel/git/saeed/linux

Saeed Mahameed says:

====================
mlx5 fixes 2021-12-28

This series provides bug fixes to mlx5 driver.

* tag 'mlx5-fixes-2021-12-28' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
  net/mlx5e: Fix wrong features assignment in case of error
  net/mlx5e: TC, Fix memory leak with rules with internal port
====================

Link: https://lore.kernel.org/r/20211229065352.30178-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 'drm-intel-fixes-2021-12-29' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Thu, 30 Dec 2021 02:12:40 +0000 (12:12 +1000)]
Merge tag 'drm-intel-fixes-2021-12-29' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

drm/i915 fixes for v5.16:
- Fix possible uninitialized variable
- Fix composite fence seqno icrement on each fence creation

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Jani Nikula <jani.nikula@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/87h7ark5r5.fsf@intel.com
2 years agonet/ncsi: check for error return from call to nla_put_u32
Jiasheng Jiang [Wed, 29 Dec 2021 03:21:18 +0000 (11:21 +0800)]
net/ncsi: check for error return from call to nla_put_u32

As we can see from the comment of the nla_put() that it could return
-EMSGSIZE if the tailroom of the skb is insufficient.
Therefore, it should be better to check the return value of the
nla_put_u32 and return the error code if error accurs.
Also, there are many other functions have the same problem, and if this
patch is correct, I will commit a new version to fix all.

Fixes: 955dc68cb9b2 ("net/ncsi: Add generic netlink family")
Signed-off-by: Jiasheng Jiang <jiasheng@iscas.ac.cn>
Link: https://lore.kernel.org/r/20211229032118.1706294-1-jiasheng@iscas.ac.cn
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: bridge: mcast: fix br_multicast_ctx_vlan_global_disabled helper
Nikolay Aleksandrov [Tue, 28 Dec 2021 15:31:42 +0000 (17:31 +0200)]
net: bridge: mcast: fix br_multicast_ctx_vlan_global_disabled helper

We need to first check if the context is a vlan one, then we need to
check the global bridge multicast vlan snooping flag, and finally the
vlan's multicast flag, otherwise we will unnecessarily enable vlan mcast
processing (e.g. querier timers).

Fixes: 7b54aaaf53cb ("net: bridge: multicast: add vlan state initialization and control")
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Link: https://lore.kernel.org/r/20211228153142.536969-1-nikolay@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: fix use-after-free in tw_timer_handler
Muchun Song [Tue, 28 Dec 2021 10:41:45 +0000 (18:41 +0800)]
net: fix use-after-free in tw_timer_handler

A real world panic issue was found as follow in Linux 5.4.

    BUG: unable to handle page fault for address: ffffde49a863de28
    PGD 7e6fe62067 P4D 7e6fe62067 PUD 7e6fe63067 PMD f51e064067 PTE 0
    RIP: 0010:tw_timer_handler+0x20/0x40
    Call Trace:
     <IRQ>
     call_timer_fn+0x2b/0x120
     run_timer_softirq+0x1ef/0x450
     __do_softirq+0x10d/0x2b8
     irq_exit+0xc7/0xd0
     smp_apic_timer_interrupt+0x68/0x120
     apic_timer_interrupt+0xf/0x20

This issue was also reported since 2017 in the thread [1],
unfortunately, the issue was still can be reproduced after fixing
DCCP.

The ipv4_mib_exit_net is called before tcp_sk_exit_batch when a net
namespace is destroyed since tcp_sk_ops is registered befrore
ipv4_mib_ops, which means tcp_sk_ops is in the front of ipv4_mib_ops
in the list of pernet_list. There will be a use-after-free on
net->mib.net_statistics in tw_timer_handler after ipv4_mib_exit_net
if there are some inflight time-wait timers.

This bug is not introduced by commit f2bf415cfed7 ("mib: add net to
NET_ADD_STATS_BH") since the net_statistics is a global variable
instead of dynamic allocation and freeing. Actually, commit
61a7e26028b9 ("mib: put net statistics on struct net") introduces
the bug since it put net statistics on struct net and free it when
net namespace is destroyed.

Moving init_ipv4_mibs() to the front of tcp_init() to fix this bug
and replace pr_crit() with panic() since continuing is meaningless
when init_ipv4_mibs() fails.

[1] https://groups.google.com/g/syzkaller/c/p1tn-_Kc6l4/m/smuL_FMAAgAJ?pli=1

Fixes: 61a7e26028b9 ("mib: put net statistics on struct net")
Signed-off-by: Muchun Song <songmuchun@bytedance.com>
Cc: Cong Wang <cong.wang@bytedance.com>
Cc: Fam Zheng <fam.zheng@bytedance.com>
Cc: <stable@vger.kernel.org>
Link: https://lore.kernel.org/r/20211228104145.9426-1-songmuchun@bytedance.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests: net: Fix a typo in udpgro_fwd.sh
Jianguo Wu [Wed, 29 Dec 2021 07:27:30 +0000 (15:27 +0800)]
selftests: net: Fix a typo in udpgro_fwd.sh

$rvs -> $rcv

Fixes: a062260a9d5f ("selftests: net: add UDP GRO forwarding self-tests")
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
Link: https://lore.kernel.org/r/d247d7c8-a03a-0abf-3c71-4006a051d133@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoselftests/net: udpgso_bench_tx: fix dst ip argument
wujianguo [Wed, 29 Dec 2021 10:58:10 +0000 (18:58 +0800)]
selftests/net: udpgso_bench_tx: fix dst ip argument

udpgso_bench_tx call setup_sockaddr() for dest address before
parsing all arguments, if we specify "-p ${dst_port}" after "-D ${dst_ip}",
then ${dst_port} will be ignored, and using default cfg_port 8000.

This will cause test case "multiple GRO socks" failed in udpgro.sh.

Setup sockaddr after parsing all arguments.

Fixes: 3a687bef148d ("selftests: udp gso benchmark")
Signed-off-by: Jianguo Wu <wujianguo@chinatelecom.cn>
Reviewed-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/ff620d9f-5b52-06ab-5286-44b945453002@163.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agox86/build: Use the proper name CONFIG_FW_LOADER
Lukas Bulwahn [Wed, 29 Dec 2021 11:15:53 +0000 (12:15 +0100)]
x86/build: Use the proper name CONFIG_FW_LOADER

Commit in Fixes intends to add the expression regex only when FW_LOADER
is enabled - not FW_LOADER_BUILTIN. Latter is a leftover from a previous
patchset and not a valid config item.

So, adjust the condition to the actual name of the config.

  [ bp: Cleanup commit message. ]

Fixes: c8dcf655ec81 ("x86/build: Tuck away built-in firmware under FW_LOADER")
Signed-off-by: Lukas Bulwahn <lukas.bulwahn@gmail.com>
Signed-off-by: Borislav Petkov <bp@suse.de>
Reviewed-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Link: https://lore.kernel.org/r/20211229111553.5846-1-lukas.bulwahn@gmail.com
2 years agoMerge branch 'net-bridge-mcast-add-and-enforce-query-interval-minimum'
Jakub Kicinski [Wed, 29 Dec 2021 20:59:43 +0000 (12:59 -0800)]
Merge branch 'net-bridge-mcast-add-and-enforce-query-interval-minimum'

Nikolay Aleksandrov says:

====================
net: bridge: mcast: add and enforce query interval minimum

This set adds and enforces 1 second minimum value for bridge multicast
query and startup query intervals in order to avoid rearming the timers
too often which could lock and crash the host. I doubt anyone is using
such low values or anything lower than 1 second, so it seems like a good
minimum. In order to be compatible if the value is lower then it is
overwritten and a log message is emitted, since we can't return an error
at this point.

Eric, I looked for the syzbot reports in its dashboard but couldn't find
them so I've added you as the reporter.

I've prepared a global bridge igmp rate limiting patch but wasn't
sure if it's ok for -net. It adds a static limit of 32k packets per
second, I plan to send it for net-next with added drop counters for
each bridge so it can be easily debugged.

Original report can be seen at:
https://lore.kernel.org/netdev/e8b9ce41-57b9-b6e2-a46a-ff9c791cf0ba@gmail.com/
====================

Link: https://lore.kernel.org/r/20211227172116.320768-1-nikolay@nvidia.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: bridge: mcast: add and enforce startup query interval minimum
Nikolay Aleksandrov [Mon, 27 Dec 2021 17:21:16 +0000 (19:21 +0200)]
net: bridge: mcast: add and enforce startup query interval minimum

As reported[1] if startup query interval is set too low in combination with
large number of startup queries and we have multiple bridges or even a
single bridge with multiple querier vlans configured we can crash the
machine. Add a 1 second minimum which must be enforced by overwriting the
value if set lower (i.e. without returning an error) to avoid breaking
user-space. If that happens a log message is emitted to let the admin know
that the startup interval has been set to the minimum. It doesn't make
sense to make the startup interval lower than the normal query interval
so use the same value of 1 second. The issue has been present since these
intervals could be user-controlled.

[1] https://lore.kernel.org/netdev/e8b9ce41-57b9-b6e2-a46a-ff9c791cf0ba@gmail.com/

Fixes: d902eee43f19 ("bridge: Add multicast count/interval sysfs entries")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet: bridge: mcast: add and enforce query interval minimum
Nikolay Aleksandrov [Mon, 27 Dec 2021 17:21:15 +0000 (19:21 +0200)]
net: bridge: mcast: add and enforce query interval minimum

As reported[1] if query interval is set too low and we have multiple
bridges or even a single bridge with multiple querier vlans configured
we can crash the machine. Add a 1 second minimum which must be enforced
by overwriting the value if set lower (i.e. without returning an error) to
avoid breaking user-space. If that happens a log message is emitted to let
the administrator know that the interval has been set to the minimum.
The issue has been present since these intervals could be user-controlled.

[1] https://lore.kernel.org/netdev/e8b9ce41-57b9-b6e2-a46a-ff9c791cf0ba@gmail.com/

Fixes: d902eee43f19 ("bridge: Add multicast count/interval sysfs entries")
Reported-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: Nikolay Aleksandrov <nikolay@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoipv6: raw: check passed optlen before reading
Tamir Duberstein [Wed, 29 Dec 2021 20:09:47 +0000 (15:09 -0500)]
ipv6: raw: check passed optlen before reading

Add a check that the user-provided option is at least as long as the
number of bytes we intend to read. Before this patch we would blindly
read sizeof(int) bytes even in cases where the user passed
optlen<sizeof(int), which would potentially read garbage or fault.

Discovered by new tests in https://github.com/google/gvisor/pull/6957 .

The original get_user call predates history in the git repo.

Signed-off-by: Tamir Duberstein <tamird@gmail.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
Link: https://lore.kernel.org/r/20211229200947.2862255-1-willemdebruijn.kernel@gmail.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoMerge tag 's390-5.16-6' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Wed, 29 Dec 2021 18:07:20 +0000 (10:07 -0800)]
Merge tag 's390-5.16-6' of git://git./linux/kernel/git/s390/linux

Pull s390 fix from Heiko Carstens:

 - fix s390 mcount regex typo in recordmcount.pl

* tag 's390-5.16-6' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  recordmcount.pl: fix typo in s390 mcount regex

2 years agoxsk: Initialise xskb free_list_node
Ciara Loftus [Mon, 20 Dec 2021 15:52:50 +0000 (15:52 +0000)]
xsk: Initialise xskb free_list_node

This commit initialises the xskb's free_list_node when the xskb is
allocated. This prevents a potential false negative returned from a call
to list_empty for that node, such as the one introduced in commit
199d983bc015 ("xsk: Fix crash on double free in buffer pool")

In my environment this issue caused packets to not be received by
the xdpsock application if the traffic was running prior to application
launch. This happened when the first batch of packets failed the xskmap
lookup and XDP_PASS was returned from the bpf program. This action is
handled in the i40e zc driver (and others) by allocating an skbuff,
freeing the xdp_buff and adding the associated xskb to the
xsk_buff_pool's free_list if it hadn't been added already. Without this
fix, the xskb is not added to the free_list because the check to determine
if it was added already returns an invalid positive result. Later, this
caused allocation errors in the driver and the failure to receive packets.

Fixes: 199d983bc015 ("xsk: Fix crash on double free in buffer pool")
Fixes: 2b43470add8c ("xsk: Introduce AF_XDP buffer allocation API")
Signed-off-by: Ciara Loftus <ciara.loftus@intel.com>
Acked-by: Magnus Karlsson <magnus.karlsson@intel.com>
Link: https://lore.kernel.org/r/20211220155250.2746-1-ciara.loftus@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agonet/mlx5e: Fix wrong features assignment in case of error
Gal Pressman [Mon, 29 Nov 2021 09:08:41 +0000 (11:08 +0200)]
net/mlx5e: Fix wrong features assignment in case of error

In case of an error in mlx5e_set_features(), 'netdev->features' must be
updated with the correct state of the device to indicate which features
were updated successfully.
To do that we maintain a copy of 'netdev->features' and update it after
successful feature changes, so we can assign it to back to
'netdev->features' if needed.

However, since not all netdev features are handled by the driver (e.g.
GRO/TSO/etc), some features may not be updated correctly in case of an
error updating another feature.

For example, while requesting to disable TSO (feature which is not
handled by the driver) and enable HW-GRO, if an error occurs during
HW-GRO enable, 'oper_features' will be assigned with 'netdev->features'
and HW-GRO turned off. TSO will remain enabled in such case, which is a
bug.

To solve that, instead of using 'netdev->features' as the baseline of
'oper_features' and changing it on set feature success, use 'features'
instead and update it in case of errors.

Fixes: 75b81ce719b7 ("net/mlx5e: Don't override netdev features field unless in error flow")
Signed-off-by: Gal Pressman <gal@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agonet/mlx5e: TC, Fix memory leak with rules with internal port
Roi Dayan [Wed, 22 Dec 2021 07:20:58 +0000 (09:20 +0200)]
net/mlx5e: TC, Fix memory leak with rules with internal port

Fix a memory leak with decap rule with internal port as destination
device. The driver allocates a modify hdr action but doesn't set
the flow attr modify hdr action which results in skipping releasing
the modify hdr action when releasing the flow.

backtrace:
    [<000000005f8c651c>] krealloc+0x83/0xd0
    [<000000009f59b143>] alloc_mod_hdr_actions+0x156/0x310 [mlx5_core]
    [<000000002257f342>] mlx5e_tc_match_to_reg_set_and_get_id+0x12a/0x360 [mlx5_core]
    [<00000000b44ea75a>] mlx5e_tc_add_fdb_flow+0x962/0x1470 [mlx5_core]
    [<0000000003e384a0>] __mlx5e_add_fdb_flow+0x54c/0xb90 [mlx5_core]
    [<00000000ed8b22b6>] mlx5e_configure_flower+0xe45/0x4af0 [mlx5_core]
    [<00000000024f4ab5>] mlx5e_rep_indr_offload.isra.0+0xfe/0x1b0 [mlx5_core]
    [<000000006c3bb494>] mlx5e_rep_indr_setup_tc_cb+0x90/0x130 [mlx5_core]
    [<00000000d3dac2ea>] tc_setup_cb_add+0x1d2/0x420

Fixes: b16eb3c81fe2 ("net/mlx5: Support internal port as decap route device")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
2 years agoMerge branch '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue
Jakub Kicinski [Wed, 29 Dec 2021 00:19:09 +0000 (16:19 -0800)]
Merge branch '1GbE' of git://git./linux/kernel/git/tnguy/net-queue

Tony Nguyen says:

====================
Intel Wired LAN Driver Updates 2021-12-28

This series contains updates to igc driver only.

Vinicius disables support for crosstimestamp on i225-V as lockups are being
observed.

James McLaughlin fixes Tx timestamping support on non-MSI-X platforms.

* '1GbE' of git://git.kernel.org/pub/scm/linux/kernel/git/tnguy/net-queue:
  igc: Fix TX timestamp support for non-MSI-X platforms
  igc: Do not enable crosstimestamping for i225-V models
====================

Link: https://lore.kernel.org/r/20211228182421.340354-1-anthony.l.nguyen@intel.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agoionic: Initialize the 'lif->dbid_inuse' bitmap
Christophe JAILLET [Sun, 26 Dec 2021 14:06:17 +0000 (15:06 +0100)]
ionic: Initialize the 'lif->dbid_inuse' bitmap

When allocated, this bitmap is not initialized. Only the first bit is set a
few lines below.

Use bitmap_zalloc() to make sure that it is cleared before being used.

Fixes: 6461b446f2a0 ("ionic: Add interrupts and doorbells")
Signed-off-by: Christophe JAILLET <christophe.jaillet@wanadoo.fr>
Signed-off-by: Shannon Nelson <snelson@pensando.io>
Link: https://lore.kernel.org/r/6a478eae0b5e6c63774e1f0ddb1a3f8c38fa8ade.1640527506.git.christophe.jaillet@wanadoo.fr
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
2 years agodrm/amd/display: Changed pipe split policy to allow for multi-display pipe split
Angus Wang [Thu, 9 Dec 2021 22:27:01 +0000 (17:27 -0500)]
drm/amd/display: Changed pipe split policy to allow for multi-display pipe split

[WHY]
Current implementation of pipe split policy prevents pipe split with
multiple displays connected, which caused the MCLK speed to be stuck at
max

[HOW]
Changed the pipe split policies so that pipe split is allowed for
multi-display configurations

Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1522
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1709
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1655
Bug: https://gitlab.freedesktop.org/drm/amd/-/issues/1403

Note this is a backport of this commit from amdgpu drm-next for 5.16.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Aric Cyr <Aric.Cyr@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Angus Wang <angus.wang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Cc: stable@vger.kernel.org
2 years agodrm/amd/display: Fix USB4 null pointer dereference in update_psp_stream_config
Nicholas Kazlauskas [Fri, 17 Dec 2021 19:18:59 +0000 (14:18 -0500)]
drm/amd/display: Fix USB4 null pointer dereference in update_psp_stream_config

[Why]
A porting error on a previous patch left the block of code that
causes the crash from a NULL pointer dereference.

More specifically, we try to access link_enc before it's assigned in
the USB4 case in the following assignment:

config.dio_output_idx = link_enc->transmitter - TRANSMITTER_UNIPHY_A;

[How]
That assignment occurs later depending on the ASIC version. It's only
needed on DCN31 and only after link_enc is already assigned.

Fixes: 986430446c917b ("drm/amd/display: fix a crash on USB4 over C20 PHY")
Reviewed-by: Harry Wentland <harry.wentland@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 years agodrm/amd/display: Set optimize_pwr_state for DCN31
Nicholas Kazlauskas [Thu, 9 Dec 2021 21:05:36 +0000 (16:05 -0500)]
drm/amd/display: Set optimize_pwr_state for DCN31

[Why]
We'll exit optimized power state to do link detection but we won't enter
back into the optimized power state.

This could potentially block s2idle entry depending on the sequencing,
but it also means we're losing some power during the transition period.

[How]
Hook up the handler like DCN21. It was also missed like the
exit_optimized_pwr_state callback.

Fixes: 64b1d0e8d500 ("drm/amd/display: Add DCN3.1 HWSEQ")
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Eric Yang <Eric.Yang2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 years agodrm/amd/display: Send s0i2_rdy in stream_count == 0 optimization
Nicholas Kazlauskas [Thu, 9 Dec 2021 18:53:36 +0000 (13:53 -0500)]
drm/amd/display: Send s0i2_rdy in stream_count == 0 optimization

[Why]
Otherwise SMU won't mark Display as idle when trying to perform s2idle.

[How]
Mark the bit in the dcn31 codepath, doesn't apply to older ASIC.

It needed to be split from phy refclk off to prevent entering s2idle
when PSR was engaged but driver was not ready.

Fixes: 118a33151658 ("drm/amd/display: Add DCN3.1 clock manager support")
Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Eric Yang <Eric.Yang2@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Nicholas Kazlauskas <nicholas.kazlauskas@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 years agodrm/amd/display: Added power down for DCN10
Lai, Derek [Mon, 6 Dec 2021 09:10:59 +0000 (17:10 +0800)]
drm/amd/display: Added power down for DCN10

[Why]
The change of setting a timer callback on boot for 10 seconds is still
working, just lacked power down for DCN10.

[How]
Added power down for DCN10.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Anthony Koo <Anthony.Koo@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Derek Lai <Derek.Lai@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 years agodrm/amd/display: fix B0 TMDS deepcolor no dislay issue
Charlene Liu [Mon, 6 Dec 2021 02:19:30 +0000 (21:19 -0500)]
drm/amd/display: fix B0 TMDS deepcolor no dislay issue

[why]
B0 PHY C map to F, D map to G driver use logic instance, dmub does the
remap. Driver still need use the right PHY instance to access right HW.

[how]
use phyical instance when program PHY register.

[note]
could move resync_control programming to dmub next.

Tested-by: Daniel Wheeler <daniel.wheeler@amd.com>
Reviewed-by: Dmytro Laktyushkin <Dmytro.Laktyushkin@amd.com>
Reviewed-by: Jun Lei <Jun.Lei@amd.com>
Acked-by: Rodrigo Siqueira <Rodrigo.Siqueira@amd.com>
Signed-off-by: Charlene Liu <Charlene.Liu@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
2 years agoMerge tag 'selinux-pr-20211228' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Tue, 28 Dec 2021 21:33:06 +0000 (13:33 -0800)]
Merge tag 'selinux-pr-20211228' of git://git./linux/kernel/git/pcmoore/selinux

Pull selinux fix from Paul Moore:
 "One more small SELinux patch to address an uninitialized stack
  variable"

* tag 'selinux-pr-20211228' of git://git.kernel.org/pub/scm/linux/kernel/git/pcmoore/selinux:
  selinux: initialize proto variable in selinux_ip_postroute_compat()

2 years agoperf scripts python: intel-pt-events.py: Fix printing of switch events
Adrian Hunter [Wed, 15 Dec 2021 08:06:36 +0000 (10:06 +0200)]
perf scripts python: intel-pt-events.py: Fix printing of switch events

The intel-pt-events.py script displays only the last of consecutive switch
statements but that may not be the last switch event for the CPU. Fix by
keeping a dictionary of last context switch keyed by CPU, and make it
possible to see all switch events by adding option --all-switch-events.

Fixes: a92bf335fd82eeee ("perf scripts python: intel-pt-events.py: Add branches to script")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Namhyung Kim <namhyung@kernel.org>
Cc: Riccardo Mancini <rickyman7@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20211215080636.149562-4-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf script: Fix CPU filtering of a script's switch events
Adrian Hunter [Wed, 15 Dec 2021 08:06:35 +0000 (10:06 +0200)]
perf script: Fix CPU filtering of a script's switch events

CPU filtering was not being applied to a script's switch events.

Fixes: 5bf83c29a0ad2e78 ("perf script: Add scripting operation process_switch()")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Riccardo Mancini <rickyman7@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20211215080636.149562-3-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
2 years agoperf intel-pt: Fix parsing of VM time correlation arguments
Adrian Hunter [Wed, 15 Dec 2021 08:06:34 +0000 (10:06 +0200)]
perf intel-pt: Fix parsing of VM time correlation arguments

Parser did not take ':' into account.

Example:

 Before:

  $ perf record -e intel_pt//u uname
  Linux
  [ perf record: Woken up 1 times to write data ]
  [ perf record: Captured and wrote 0.026 MB perf.data ]
  $ perf inject -i perf.data --vm-time-correlation="dry-run 123"
  $ perf inject -i perf.data --vm-time-correlation="dry-run 123:456"
  Failed to parse VM Time Correlation options
  0x620 [0x98]: failed to process type: 70 [Invalid argument]
  $

 After:

  $ perf inject -i perf.data --vm-time-correlation="dry-run 123:456"
  $

Fixes: e3ff42bdebcfeb5f ("perf intel-pt: Parse VM Time Correlation options and set up decoding")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
Acked-by: Namhyung Kim <namhyung@kernel.org>
Cc: Jiri Olsa <jolsa@redhat.com>
Cc: Riccardo Mancini <rickyman7@gmail.com>
Cc: stable@vger.kernel.org
Link: https://lore.kernel.org/r/20211215080636.149562-2-adrian.hunter@intel.com
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>