openwrt/staging/blogic.git
5 years agobtrfs: tests: return errors from extent map test case 2
David Sterba [Fri, 15 Mar 2019 17:41:06 +0000 (18:41 +0100)]
btrfs: tests: return errors from extent map test case 2

Replace asserts with error messages and return errors.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: return errors from extent map test case 1
David Sterba [Fri, 15 Mar 2019 17:41:06 +0000 (18:41 +0100)]
btrfs: tests: return errors from extent map test case 1

Replace asserts with error messages and return errors.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: return errors from extent map tests
David Sterba [Fri, 15 Mar 2019 17:06:16 +0000 (18:06 +0100)]
btrfs: tests: return errors from extent map tests

The individual testcases for extent maps do not return an error on
allocation failures. This is not a big problem as the allocation don't
fail in general but there are functional tests handled with ASSERTS.
This makes tests dependent on them and it's not reliable.

This patch adds the allocation failure handling and allows for the
conversion of the asserts to proper error handling and reporting.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: properly initialize fs_info of extent buffer
David Sterba [Fri, 15 Mar 2019 16:42:07 +0000 (17:42 +0100)]
btrfs: tests: properly initialize fs_info of extent buffer

The fs_info is supposed to be valid, even though it's not used right
now and the test does not crash.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: use standard error message after block group allocation failure
David Sterba [Fri, 15 Mar 2019 16:28:46 +0000 (17:28 +0100)]
btrfs: tests: use standard error message after block group allocation failure

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: use standard error message after inode allocation failure
David Sterba [Fri, 15 Mar 2019 16:28:46 +0000 (17:28 +0100)]
btrfs: tests: use standard error message after inode allocation failure

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: use standard error message after path allocation failure
David Sterba [Fri, 15 Mar 2019 16:28:46 +0000 (17:28 +0100)]
btrfs: tests: use standard error message after path allocation failure

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: use standard error message after extent buffer allocation failure
David Sterba [Fri, 15 Mar 2019 16:28:46 +0000 (17:28 +0100)]
btrfs: tests: use standard error message after extent buffer allocation failure

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: use standard error message after root allocation failure
David Sterba [Fri, 15 Mar 2019 16:28:46 +0000 (17:28 +0100)]
btrfs: tests: use standard error message after root allocation failure

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: use standard error message after fs_info allocation failure
David Sterba [Fri, 15 Mar 2019 16:28:46 +0000 (17:28 +0100)]
btrfs: tests: use standard error message after fs_info allocation failure

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: add table of most common errors
David Sterba [Fri, 15 Mar 2019 16:23:30 +0000 (17:23 +0100)]
btrfs: tests: add table of most common errors

Allocation of main objects like fs_info or extent buffers is in each
test so let's simplify and unify the error messages to a table and add a
convenience helper.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: print file:line for error messages
David Sterba [Fri, 15 Mar 2019 16:03:55 +0000 (17:03 +0100)]
btrfs: tests: print file:line for error messages

For better diagnostics print the file name and line to locate the
errors. Sample output:

[    9.052924] BTRFS: selftest: fs/btrfs/tests/extent-io-tests.c:283 offset bits do not match

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: don't leak fs_info in extent_io bitmap tests
David Sterba [Fri, 15 Mar 2019 15:46:55 +0000 (16:46 +0100)]
btrfs: tests: don't leak fs_info in extent_io bitmap tests

The fs_info is not freed at the end of the function and leaks. The
function is called twice so there can be up to 2x sizeof(struct
btrfs_fs_info) of leaked memory.  Fortunatelly this affects only testing
builds, the size could be 16k with several debugging features enabled.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tests: handle fs_info allocation failure in extent_io tests
David Sterba [Fri, 15 Mar 2019 15:43:11 +0000 (16:43 +0100)]
btrfs: tests: handle fs_info allocation failure in extent_io tests

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: disk-io: Show the timing of corrupted tree block explicitly
Qu Wenruo [Wed, 20 Mar 2019 06:27:40 +0000 (14:27 +0800)]
btrfs: disk-io: Show the timing of corrupted tree block explicitly

Just add one extra line to show when the corruption is detected.
Currently only read time detection is possible.

The planned distinguish line would be:

  read time:
    <detailed report>
    block=XXXXX read time tree block corruption detected

  write time:
    <detailed report>
    block=XXXXX write time tree block corruption detected

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: fix panic during relocation after ENOSPC before writeback happens
Josef Bacik [Mon, 25 Feb 2019 16:14:45 +0000 (11:14 -0500)]
btrfs: fix panic during relocation after ENOSPC before writeback happens

We've been seeing the following sporadically throughout our fleet

panic: kernel BUG at fs/btrfs/relocation.c:4584!
netversion: 5.0-0
Backtrace:
 #0 [ffffc90003adb880] machine_kexec at ffffffff81041da8
 #1 [ffffc90003adb8c8] __crash_kexec at ffffffff8110396c
 #2 [ffffc90003adb988] crash_kexec at ffffffff811048ad
 #3 [ffffc90003adb9a0] oops_end at ffffffff8101c19a
 #4 [ffffc90003adb9c0] do_trap at ffffffff81019114
 #5 [ffffc90003adba00] do_error_trap at ffffffff810195d0
 #6 [ffffc90003adbab0] invalid_op at ffffffff81a00a9b
    [exception RIP: btrfs_reloc_cow_block+692]
    RIP: ffffffff8143b614  RSP: ffffc90003adbb68  RFLAGS: 00010246
    RAX: fffffffffffffff7  RBX: ffff8806b9c32000  RCX: ffff8806aad00690
    RDX: ffff880850b295e0  RSI: ffff8806b9c32000  RDI: ffff88084f205bd0
    RBP: ffff880849415000   R8: ffffc90003adbbe0   R9: ffff88085ac90000
    R10: ffff8805f7369140  R11: 0000000000000000  R12: ffff880850b295e0
    R13: ffff88084f205bd0  R14: 0000000000000000  R15: 0000000000000000
    ORIG_RAX: ffffffffffffffff  CS: 0010  SS: 0018
 #7 [ffffc90003adbbb0] __btrfs_cow_block at ffffffff813bf1cd
 #8 [ffffc90003adbc28] btrfs_cow_block at ffffffff813bf4b3
 #9 [ffffc90003adbc78] btrfs_search_slot at ffffffff813c2e6c

The way relocation moves data extents is by creating a reloc inode and
preallocating extents in this inode and then copying the data into these
preallocated extents.  Once we've done this for all of our extents,
we'll write out these dirty pages, which marks the extent written, and
goes into btrfs_reloc_cow_block().  From here we get our current
reloc_control, which _should_ match the reloc_control for the current
block group we're relocating.

However if we get an ENOSPC in this path at some point we'll bail out,
never initiating writeback on this inode.  Not a huge deal, unless we
happen to be doing relocation on a different block group, and this block
group is now rc->stage == UPDATE_DATA_PTRS.  This trips the BUG_ON() in
btrfs_reloc_cow_block(), because we expect to be done modifying the data
inode.  We are in fact done modifying the metadata for the data inode
we're currently using, but not the one from the failed block group, and
thus we BUG_ON().

(This happens when writeback finishes for extents from the previous
group, when we are at btrfs_finish_ordered_io() which updates the data
reloc tree (inode item, drops/adds extent items, etc).)

Fix this by writing out the reloc data inode always, and then breaking
out of the loop after that point to keep from tripping this BUG_ON()
later.

Signed-off-by: Josef Bacik <josef@toxicpanda.com>
Reviewed-by: Filipe Manana <fdmanana@suse.com>
[ add note from Filipe ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Use less confusing condition for uptodate parameter to btrfs_writepage_endio_f...
Nikolay Borisov [Wed, 20 Mar 2019 19:53:16 +0000 (21:53 +0200)]
btrfs: Use less confusing condition for uptodate parameter to btrfs_writepage_endio_finish_ordered

The uptodate parameter of btrfs_writepage_endio_finish_ordered is used
to signal whether an error has occured while writing the given page.
0 signals an error, which is propagated to callees and 1 signifies
success. In end_compressed_bio_write the ->bi_status is checked and
based on it either BLK_STS_OK (0) or BLK_STS_NOTSUPP (1) are used. While
from functional point of view this is ok it's a for the poor reader of
the code, since the block layer values are conflated with the semantics
of the parameter.

Just use plain 0 or 1. No functional changes.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: Handle errors better in extent_writepages()
Qu Wenruo [Wed, 20 Mar 2019 06:27:48 +0000 (14:27 +0800)]
btrfs: extent_io: Handle errors better in extent_writepages()

We can only get <=0 from extent_write_cache_pages, add an ASSERT() for
it just in case.

Then instead of submitting the write bio even if we got some error,
check the return value first.
If we have already hit some error, just clean up the corrupted or
half-baked bio, and return error.

If there is no error so far, then call flush_write_bio() and return the
result.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: add proper error handling to lock_extent_buffer_for_io()
Qu Wenruo [Wed, 20 Mar 2019 06:27:46 +0000 (14:27 +0800)]
btrfs: extent_io: add proper error handling to lock_extent_buffer_for_io()

This function needs some extra checks on locked pages and eb.  For error
handling we need to unlock locked pages and the eb.

There is a rare >0 return value branch, where all pages get locked
while write bio is not flushed.

Thankfully it's handled by the only caller, btree_write_cache_pages(),
as later write_one_eb() call will trigger submit_one_bio().  So there
shouldn't be any problem.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: Handle errors better in extent_write_locked_range()
Qu Wenruo [Wed, 20 Mar 2019 06:27:45 +0000 (14:27 +0800)]
btrfs: extent_io: Handle errors better in extent_write_locked_range()

We can only get @ret <= 0.  Add an ASSERT() for it just in case.

Then, instead of submitting the write bio even we got some error, check
the return value first.

If we have already hit some error, just clean up the corrupted or
half-baked bio, and return error.

If there is no error so far, then call flush_write_bio() and return the
result.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: Kill dead condition in extent_write_cache_pages()
Qu Wenruo [Wed, 20 Mar 2019 06:27:44 +0000 (14:27 +0800)]
btrfs: extent_io: Kill dead condition in extent_write_cache_pages()

Since __extent_writepage() will no longer return >0 value,
(ret == AOP_WRITEPAGE_ACTIVATE) will never be true.

Kill that dead branch.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: Handle errors better in btree_write_cache_pages()
Qu Wenruo [Wed, 20 Mar 2019 06:27:43 +0000 (14:27 +0800)]
btrfs: extent_io: Handle errors better in btree_write_cache_pages()

In btree_write_cache_pages(), we can only get @ret <= 0.
Add an ASSERT() for it just in case.

Then instead of submitting the write bio even we got some error, check
the return value first.
If we have already hit some error, just clean up the corrupted or
half-baked bio, and return error.

If there is no error so far, then call flush_write_bio() and return the
result.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: Handle errors better in extent_write_full_page()
Qu Wenruo [Wed, 20 Mar 2019 06:27:42 +0000 (14:27 +0800)]
btrfs: extent_io: Handle errors better in extent_write_full_page()

Since now flush_write_bio() could return error, kill the BUG_ON() first.
Then don't call flush_write_bio() unconditionally, instead we check the
return value from __extent_writepage() first.

If __extent_writepage() fails, we do cleanup, and return error without
submitting the possible corrupted or half-baked bio.

If __extent_writepage() successes, then we call flush_write_bio() and
return the result.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: extent_io: Move the BUG_ON() in flush_write_bio() one level up
Qu Wenruo [Wed, 20 Mar 2019 06:27:41 +0000 (14:27 +0800)]
btrfs: extent_io: Move the BUG_ON() in flush_write_bio() one level up

We have a BUG_ON() in flush_write_bio() to handle the return value of
submit_one_bio().

Move the BUG_ON() one level up to all its callers.

This patch will introduce temporary variable, @flush_ret to keep code
change minimal in this patch. That variable will be cleaned up when
enhancing the error handling later.

Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Always output error message when key/level verification fails
Qu Wenruo [Wed, 20 Mar 2019 06:27:39 +0000 (14:27 +0800)]
btrfs: Always output error message when key/level verification fails

We have internal report of strange transaction abort due to EUCLEAN
without any error message.

Since error message inside verify_level_key() is only enabled for
CONFIG_BTRFS_DEBUG, the error message won't be printed on most builds.

This patch will make the error message mandatory, so when problem
happens we know what's causing the problem.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Check the first key and level for cached extent buffer
Qu Wenruo [Tue, 12 Mar 2019 09:10:40 +0000 (17:10 +0800)]
btrfs: Check the first key and level for cached extent buffer

[BUG]
When reading a file from a fuzzed image, kernel can panic like:

  BTRFS warning (device loop0): csum failed root 5 ino 270 off 0 csum 0x98f94189 expected csum 0x00000000 mirror 1
  assertion failed: !memcmp_extent_buffer(b, &disk_key, offsetof(struct btrfs_leaf, items[0].key), sizeof(disk_key)), file: fs/btrfs/ctree.c, line: 2544
  ------------[ cut here ]------------
  kernel BUG at fs/btrfs/ctree.h:3500!
  invalid opcode: 0000 [#1] PREEMPT SMP NOPTI
  RIP: 0010:btrfs_search_slot.cold.24+0x61/0x63 [btrfs]
  Call Trace:
   btrfs_lookup_csum+0x52/0x150 [btrfs]
   __btrfs_lookup_bio_sums+0x209/0x640 [btrfs]
   btrfs_submit_bio_hook+0x103/0x170 [btrfs]
   submit_one_bio+0x59/0x80 [btrfs]
   extent_read_full_page+0x58/0x80 [btrfs]
   generic_file_read_iter+0x2f6/0x9d0
   __vfs_read+0x14d/0x1a0
   vfs_read+0x8d/0x140
   ksys_read+0x52/0xc0
   do_syscall_64+0x60/0x210
   entry_SYSCALL_64_after_hwframe+0x49/0xbe

[CAUSE]
The fuzzed image has a corrupted leaf whose first key doesn't match its
parent:

  checksum tree key (CSUM_TREE ROOT_ITEM 0)
  node 29741056 level 1 items 14 free 107 generation 19 owner CSUM_TREE
  fs uuid 3381d111-94a3-4ac7-8f39-611bbbdab7e6
  chunk uuid 9af1c3c7-2af5-488b-8553-530bd515f14c
   ...
          key (EXTENT_CSUM EXTENT_CSUM 79691776) block 29761536 gen 19

  leaf 29761536 items 1 free space 1726 generation 19 owner CSUM_TREE
  leaf 29761536 flags 0x1(WRITTEN) backref revision 1
  fs uuid 3381d111-94a3-4ac7-8f39-611bbbdab7e6
  chunk uuid 9af1c3c7-2af5-488b-8553-530bd515f14c
          item 0 key (EXTENT_CSUM EXTENT_CSUM 8798638964736) itemoff 1751 itemsize 2244
                  range start 8798638964736 end 8798641262592 length 2297856

When reading the above tree block, we have extent_buffer->refs = 2 in
the context:

- initial one from __alloc_extent_buffer()
  alloc_extent_buffer()
  |- __alloc_extent_buffer()
     |- atomic_set(&eb->refs, 1)

- one being added to fs_info->buffer_radix
  alloc_extent_buffer()
  |- check_buffer_tree_ref()
     |- atomic_inc(&eb->refs)

So if even we call free_extent_buffer() in read_tree_block or other
similar situation, we only decrease the refs by 1, it doesn't reach 0
and won't be freed right now.

The staled eb and its corrupted content will still be kept cached.

Furthermore, we have several extra cases where we either don't do first
key check or the check is not proper for all callers:

- scrub
  We just don't have first key in this context.

- shared tree block
  One tree block can be shared by several snapshot/subvolume trees.
  In that case, the first key check for one subvolume doesn't apply to
  another.

So for the above reasons, a corrupted extent buffer can sneak into the
buffer cache.

[FIX]
Call verify_level_key in read_block_for_search to do another
verification. For that purpose the function is exported.

Due to above reasons, although we can free corrupted extent buffer from
cache, we still need the check in read_block_for_search(), for scrub and
shared tree blocks.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=202755
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202757
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202759
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202761
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202767
Link: https://bugzilla.kernel.org/show_bug.cgi?id=202769
Reported-by: Yoon Jungyeon <jungyeon@gatech.edu>
CC: stable@vger.kernel.org # 4.19+
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Correctly free extent buffer in case btree_read_extent_buffer_pages fails
Nikolay Borisov [Thu, 14 Mar 2019 07:52:35 +0000 (09:52 +0200)]
btrfs: Correctly free extent buffer in case btree_read_extent_buffer_pages fails

If a an eb fails to be read for whatever reason - it's corrupted on disk
and parent transid/key validations fail or IO for eb pages fail then
this buffer must be removed from the buffer cache. Currently the code
calls free_extent_buffer if an error occurs. Unfortunately this doesn't
achieve the desired behavior since btrfs_find_create_tree_block returns
with eb->refs == 2.

On the other hand free_extent_buffer will only decrement the refs once
leaving it added to the buffer cache radix tree.  This enables later
code to look up the buffer from the cache and utilize it potentially
leading to a crash.

The correct way to free the buffer is call free_extent_buffer_stale.
This function will correctly call atomic_dec explicitly for the buffer
and subsequently call release_extent_buffer which will decrement the
final reference thus correctly remove the invalid buffer from buffer
cache. This change affects only newly allocated buffers since they have
eb->refs == 2.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=202755
Reported-by: Jungyeon <jungyeon@gatech.edu>
CC: stable@vger.kernel.org # 4.4+
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Make btrfs_(set|clear)_header_flag return void
Qu Wenruo [Tue, 19 Mar 2019 06:04:17 +0000 (14:04 +0800)]
btrfs: Make btrfs_(set|clear)_header_flag return void

From the introduction of btrfs_(set|clear)_header_flag, there is no
usage of its return value.  So just make it return void.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: reloc: Fix NULL pointer dereference due to expanded reloc_root lifespan
Qu Wenruo [Mon, 18 Mar 2019 02:48:19 +0000 (10:48 +0800)]
btrfs: reloc: Fix NULL pointer dereference due to expanded reloc_root lifespan

Commit d2311e698578 ("btrfs: relocation: Delay reloc tree deletion after
merge_reloc_roots()") expands the life span of root->reloc_root.

This breaks certain checs of fs_info->reloc_ctl.  Before that commit, if
we have a root with valid reloc_root, then it's ensured to have
fs_info->reloc_ctl.

But now since reloc_root doesn't always mean a valid fs_info->reloc_ctl,
such check is unreliable and can cause the following NULL pointer
dereference:

  BUG: unable to handle kernel NULL pointer dereference at 00000000000005c1
  IP: btrfs_reloc_pre_snapshot+0x20/0x50 [btrfs]
  PGD 0 P4D 0
  Oops: 0000 [#1] SMP PTI
  CPU: 0 PID: 10379 Comm: snapperd Not tainted
  Call Trace:
   create_pending_snapshot+0xd7/0xfc0 [btrfs]
   create_pending_snapshots+0x8e/0xb0 [btrfs]
   btrfs_commit_transaction+0x2ac/0x8f0 [btrfs]
   btrfs_mksubvol+0x561/0x570 [btrfs]
   btrfs_ioctl_snap_create_transid+0x189/0x190 [btrfs]
   btrfs_ioctl_snap_create_v2+0x102/0x150 [btrfs]
   btrfs_ioctl+0x5c9/0x1e60 [btrfs]
   do_vfs_ioctl+0x90/0x5f0
   SyS_ioctl+0x74/0x80
   do_syscall_64+0x7b/0x150
   entry_SYSCALL_64_after_hwframe+0x3d/0xa2
  RIP: 0033:0x7fd7cdab8467

Fix it by explicitly checking fs_info->reloc_ctl other than using the
implied root->reloc_root.

Fixes: d2311e698578 ("btrfs: relocation: Delay reloc tree deletion after merge_reloc_roots")
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Remove unused -EIO assignment in end_bio_extent_readpage
Nikolay Borisov [Mon, 18 Mar 2019 15:45:18 +0000 (17:45 +0200)]
btrfs: Remove unused -EIO assignment in end_bio_extent_readpage

In case we hit the error case for a metadata buffer in
end_bio_extent_readpage then 'ret' won't really be checked before it's
written again to. This means the -EIO in this case will never be
checked, just remove it.

Fixes-coverity-id: 1442513
Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Exploit the fact that pages passed to extent_readpages are always contiguous
Nikolay Borisov [Mon, 11 Mar 2019 07:55:38 +0000 (09:55 +0200)]
btrfs: Exploit the fact that pages passed to extent_readpages are always contiguous

Currently extent_readpages (called from btrfs_readpages) will always
call __extent_readpages which tries to create contiguous range of pages
and call __do_contiguous_readpages when such contiguous range is
created.

It turns out this is unnecessary due to the fact that generic MM code
always calls filesystem's ->readpages callback (btrfs_readpages in
this case) with already contiguous pages. Armed with this knowledge it's
possible to simplify extent_readpages by eliminating the call to
__extent_readpages and directly calling contiguous_readpages.

The only edge case that needs to be handled is when
add_to_page_cache_lru fails. This is easy as all that is needed is to
submit whatever is the number of pages successfully added to the lru.
This can happen when the page is already in the range, so it does not
need to be read again, and we can't do anything else in case of other
errors.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: switch extent_buffer::lock_nested to bool
David Sterba [Fri, 24 Aug 2018 14:31:17 +0000 (16:31 +0200)]
btrfs: switch extent_buffer::lock_nested to bool

The member is tracking simple status of the lock, we can use bool for
that and make some room for further space reduction in the structure.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: use assertion helpers for extent buffer write lock counters
David Sterba [Fri, 24 Aug 2018 14:24:26 +0000 (16:24 +0200)]
btrfs: use assertion helpers for extent buffer write lock counters

Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::write_locks become unused and can be
moved to the appropriate section, saving a few bytes.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: add assertion helpers for extent buffer write lock counters
David Sterba [Fri, 24 Aug 2018 14:20:02 +0000 (16:20 +0200)]
btrfs: add assertion helpers for extent buffer write lock counters

The write_locks are a simple counter to track locking balance and used
to assert tree locks.  Add helpers to make it conditionally work only in
DEBUG builds.  Will be used in followup patches.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: use assertion helpers for extent buffer read lock counters
David Sterba [Fri, 24 Aug 2018 14:15:51 +0000 (16:15 +0200)]
btrfs: use assertion helpers for extent buffer read lock counters

Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::read_locks become unused and can be
moved to the appropriate section, saving a few bytes.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: add assertion helpers for extent buffer read lock counters
David Sterba [Fri, 24 Aug 2018 14:13:41 +0000 (16:13 +0200)]
btrfs: add assertion helpers for extent buffer read lock counters

The read_locks are a simple counter to track locking balance and used to
assert tree locks.  Add helpers to make it conditionally work only in
DEBUG builds.  Will be used in followup patches.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: use assertion helpers for spinning readers
David Sterba [Fri, 24 Aug 2018 13:57:38 +0000 (15:57 +0200)]
btrfs: use assertion helpers for spinning readers

Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::spining_readers become unused and can be
moved to the appropriate section, saving a few bytes.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: add assertion helpers for spinning readers
David Sterba [Fri, 24 Aug 2018 13:53:42 +0000 (15:53 +0200)]
btrfs: add assertion helpers for spinning readers

Add helpers for conditional DEBUG build to assert that the extent buffer
spinning_readers constraints are met. Will be used in followup patches.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: use assertion helpers for spinning writers
David Sterba [Fri, 24 Aug 2018 12:56:28 +0000 (14:56 +0200)]
btrfs: use assertion helpers for spinning writers

Use the helpers where open coded. On non-debug builds, the warnings will
not trigger and extent_buffer::spining_writers become unused and can be
moved to the appropriate section, saving a few bytes.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: add assertion helpers for spinning writers
David Sterba [Fri, 24 Aug 2018 12:45:20 +0000 (14:45 +0200)]
btrfs: add assertion helpers for spinning writers

Add helpers for conditional DEBUG build to assert that the extent buffer
spinning_writers constraints are met. Will be used in followup patches.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Remove EXTENT_IOBITS
Nikolay Borisov [Thu, 14 Mar 2019 13:28:31 +0000 (15:28 +0200)]
btrfs: Remove EXTENT_IOBITS

This flag just became synonymous to EXTENT_LOCKED, so just remove it and
used EXTENT_LOCKED directly. No functional changes.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Remove EXTENT_WRITEBACK
Nikolay Borisov [Thu, 14 Mar 2019 13:28:30 +0000 (15:28 +0200)]
btrfs: Remove EXTENT_WRITEBACK

This flag was introduced in a52d9a8033c4 ("Btrfs: Extent based page
cache code.") and subsequently it's usage effectively was removed by
1edbb734b4e0 ("Btrfs: reduce CPU usage in the extent_state tree") and
f2a97a9dbd86 ("btrfs: remove all unused functions"). Just remove it,
no functional changes.

Signed-off-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Turn an 'else if' into an 'else' in btrfs_uuid_tree_add
Nathan Chancellor [Thu, 7 Mar 2019 16:35:15 +0000 (09:35 -0700)]
btrfs: Turn an 'else if' into an 'else' in btrfs_uuid_tree_add

When building with -Wsometimes-uninitialized, Clang warns:

fs/btrfs/uuid-tree.c:129:13: warning: variable 'eb' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]
fs/btrfs/uuid-tree.c:129:13: warning: variable 'offset' is used uninitialized whenever 'if' condition is false [-Wsometimes-uninitialized]

Clang can't tell that all cases are covered with this final else if.
Just turn it into an else so that it is clear.

Link: https://github.com/ClangBuiltLinux/linux/issues/385
Suggested-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Nathan Chancellor <natechancellor@gmail.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: refactor btrfs_set_prop and add btrfs_set_prop_trans
Anand Jain [Fri, 1 Mar 2019 04:34:50 +0000 (12:34 +0800)]
btrfs: refactor btrfs_set_prop and add btrfs_set_prop_trans

btrfs_set_prop() takes transaction pointer as the first argument,
however in ioctl.c for the purpose of setting the compression property,
we call btrfs_set_prop() with NULL transaction pointer. Down in
the call chain  btrfs_setxattr() starts transaction to update the
attribute and also to update the inode.

So for clarity, create btrfs_set_prop_trans() with no transaction
pointer as argument, in preparation to start transaction here instead of
doing it down the call chain at btrfs_setxattr().

Also now the btrfs_set_prop() is a static function.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: rename fs_info argument to fs_private
Anand Jain [Fri, 1 Mar 2019 04:34:49 +0000 (12:34 +0800)]
btrfs: rename fs_info argument to fs_private

fs_info is commonly used to represent struct fs_info *, rename
to fs_private to avoid confusion.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: drop redundant forward declaration in props.c
Anand Jain [Fri, 1 Mar 2019 04:34:48 +0000 (12:34 +0800)]
btrfs: drop redundant forward declaration in props.c

Drop forward declaration of the functions:

- prop_compression_validate
- prop_compression_apply
- prop_compression_extract

No functional changes.

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>
5 years agobtrfs: merge _btrfs_set_prop helpers
Anand Jain [Fri, 1 Mar 2019 04:34:47 +0000 (12:34 +0800)]
btrfs: merge _btrfs_set_prop helpers

btrfs_set_prop() is a redirect to __btrfs_set_prop() with the
transaction handle equal to NULL.  __btrfs_set_prop() in turn passes
this to do_setxattr() which then transaction is actually created.

Instead merge  __btrfs_set_prop() to btrfs_set_prop(), and update the
caller with NULL argument.

Signed-off-by: Anand Jain <anand.jain@oracle.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: reduce kmap_atomic time for checksumming
Johannes Thumshirn [Thu, 7 Mar 2019 16:14:00 +0000 (17:14 +0100)]
btrfs: reduce kmap_atomic time for checksumming

Since commit c40a3d38aff4 ("Btrfs: Compute and look up csums based on
sectorsized blocks") we do a kmap_atomic() on the contents of a bvec.
The code before c40a3d38aff4 had the kmap region just around the
checksumming too.

kmap_atomic() in turn does a preempt_disable() and pagefault_disable(),
so we shouldn't map the data for too long. Reduce the time the bvec's
page is mapped to when we actually need it.

Performance wise it doesn't seem to make a huge difference with a 2 vcpu VM
on a /dev/zram device:

       vanilla      patched      delta
write  17.4MiB/s    17.8MiB/s +0.4MiB/s (+2%)
read   40.6MiB/s    41.5MiB/s   +0.9MiB/s (+2%)

The following fio job profile was used in the comparision:

[global]
ioengine=libaio
direct=1
sync=1
norandommap
time_based
runtime=10m
size=100m
group_reporting
numjobs=2

[test]
filename=/mnt/test/fio
rw=randrw
rwmixread=70

Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: tracepoints: Add trace events for extent_io_tree
Qu Wenruo [Fri, 1 Mar 2019 02:48:00 +0000 (10:48 +0800)]
btrfs: tracepoints: Add trace events for extent_io_tree

Although btrfs heavily relies on extent_io_tree, we don't really have
any good trace events for them.

This patch will add the folowing trace events:
- trace_btrfs_set_extent_bit()
- trace_btrfs_clear_extent_bit()
- trace_btrfs_convert_extent_bit()

Since selftests could create temporary extent_io_tree without fs_info,
modify TP_fast_assign_fsid() to accept NULL as fs_info.  NULL fs_info
will lead to all zero fsid.

The output would be:
  btrfs_set_extent_bit: <FDID>: io_tree=INODE_IO ino=1 root=1 start=22036480 len=4096 set_bits=LOCKED
  btrfs_set_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22040576 len=4096 set_bits=LOCKED
  btrfs_set_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22044672 len=4096 set_bits=LOCKED
  btrfs_set_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22048768 len=4096 set_bits=LOCKED
  btrfs_clear_extent_bit: <FSID>: io_tree=INODE_IO ino=1 root=1 start=22036480 len=16384 clear_bits=LOCKED
  ^^^ Extent buffer 22036480 read from disk, the locking progress

  btrfs_set_extent_bit: <FSID>: io_tree=TRANS_DIRTY_PAGES ino=1 root=1 start=30425088 len=16384 set_bits=DIRTY
  btrfs_set_extent_bit: <FSID>: io_tree=TRANS_DIRTY_PAGES ino=1 root=1 start=30441472 len=16384 set_bits=DIRTY
  ^^^ 2 new tree blocks allocated in one transaction

  btrfs_set_extent_bit: <FSID>: io_tree=FREED_EXTENTS0 ino=0 root=0 start=30523392 len=16384 set_bits=DIRTY
  btrfs_set_extent_bit: <FSID>: io_tree=FREED_EXTENTS0 ino=0 root=0 start=30556160 len=16384 set_bits=DIRTY
  ^^^ 2 old tree blocks get pinned down

There is one point which need attention:
1) Those trace events can be pretty heavy:
   The following workload would generate over 400 trace events.

mkfs.btrfs -f $dev
start_trace
mount $dev $mnt -o enospc_debug
sync
touch $mnt/file1
touch $mnt/file2
touch $mnt/file3
xfs_io -f -c "pwrite 0 16k" $mnt/file4
umount $mnt
end_trace

   It's not recommended to use them in real world environment.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
[ rename enums ]
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Introduce extent_io_tree::owner to distinguish different io_trees
Qu Wenruo [Fri, 1 Mar 2019 02:47:59 +0000 (10:47 +0800)]
btrfs: Introduce extent_io_tree::owner to distinguish different io_trees

Btrfs has the following different extent_io_trees used:

- fs_info::free_extents[2]
- btrfs_inode::io_tree - for both normal inodes and the btree inode
- btrfs_inode::io_failure_tree
- btrfs_transaction::dirty_pages
- btrfs_root::dirty_log_pages

If we want to trace changes in those trees, it will be pretty hard to
distinguish them.

Instead of using hard-to-read pointer address, this patch will introduce
a new member extent_io_tree::owner to track the owner.

This modification needs all the callers of extent_io_tree_init() to
accept a new parameter @owner.

This patch provides the basis for later trace events.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: switch extent_io_tree::track_uptodate to bool
David Sterba [Mon, 11 Mar 2019 14:58:30 +0000 (15:58 +0100)]
btrfs: switch extent_io_tree::track_uptodate to bool

This patch is split from the following one "btrfs: Introduce
extent_io_tree::owner to distinguish different io_trees" from Qu, so the
different changes are not mixed together.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Introduce fs_info to extent_io_tree
Qu Wenruo [Fri, 1 Mar 2019 02:47:58 +0000 (10:47 +0800)]
btrfs: Introduce fs_info to extent_io_tree

This patch will add a new member fs_info to extent_io_tree.

This provides the basis for later trace events to distinguish the output
between different btrfs filesystems. While this increases the size of
the structure, we want to know the source of the trace events and
passing the fs_info as an argument to all contexts is not possible.

The selftests are now allowed to set it to NULL as they don't use the
tracepoints.

Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agoBtrfs: remove no longer used 'sync' member from transaction handle
Filipe Manana [Mon, 11 Mar 2019 13:10:56 +0000 (13:10 +0000)]
Btrfs: remove no longer used 'sync' member from transaction handle

Commit db2462a6ad3d ("btrfs: don't run delayed refs in the end transaction
logic") removed its last use, so now it does absolutely nothing, therefore
remove it.

Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: zstd: remove indirect calls for local functions
Dennis Zhou [Wed, 27 Feb 2019 21:21:28 +0000 (16:21 -0500)]
btrfs: zstd: remove indirect calls for local functions

While calling functions inside zstd, we don't need to use the
indirection provided by the workspace_manager. Forward declarations are
added to maintain the function order of btrfs_compress_op.

Signed-off-by: Dennis Zhou <dennis@kernel.org>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: scrub: return EAGAIN when fs is closing
David Sterba [Mon, 25 Feb 2019 18:57:41 +0000 (19:57 +0100)]
btrfs: scrub: return EAGAIN when fs is closing

The error code used here is wrong as it's not invalid to try to start
scrub when umount has begun.  Returning EAGAIN is more user friendly as
it's recoverable.

Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Initialize inode::i_mapping once in btrfs_symlink
Goldwyn Rodrigues [Mon, 25 Feb 2019 19:07:43 +0000 (13:07 -0600)]
btrfs: Initialize inode::i_mapping once in btrfs_symlink

inode->i_op is initialized multiple times. Perform it once. This was
left by 4779cc04248d ("Btrfs: get rid of btrfs_symlink_aops").

Signed-off-by: Goldwyn Rodrigues <rgoldwyn@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: Don't panic when we can't find a root key
Qu Wenruo [Tue, 26 Feb 2019 08:33:56 +0000 (16:33 +0800)]
btrfs: Don't panic when we can't find a root key

When we failed to find a root key in btrfs_update_root(), we just panic.

That's definitely not cool, fix it by outputting an unique error
message, aborting current transaction and return -EUCLEAN. This should
not normally happen as the root has been used by the callers in some
way.

Reviewed-by: Filipe Manana <fdmanana@suse.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: warn if extent buffer mapping crosses a page boundary in csum_tree_block
Johannes Thumshirn [Mon, 25 Feb 2019 13:24:16 +0000 (14:24 +0100)]
btrfs: warn if extent buffer mapping crosses a page boundary in csum_tree_block

Since commit d2e174d5d3ee ("btrfs: document extent mapping assumptions in
checksum") we have a comment in place why map_private_extent_buffer()
can't return 1 in the csum_tree_block() case.

Make this a bit more explicit and WARN_ON() in case this this assumption
breaks.

Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agobtrfs: factor our read/write stage off csum_tree_block into its callers
Johannes Thumshirn [Mon, 25 Feb 2019 13:24:15 +0000 (14:24 +0100)]
btrfs: factor our read/write stage off csum_tree_block into its callers

Currently csum_tree_block() does two things, first it as it's name
suggests it calculates the checksum for a tree-block. But it also writes
this checksum to disk or reads an extent_buffer from disk and compares the
checksum with the calculated checksum, depending on the verify argument.

Furthermore one of the two callers passes in '1' for the verify argument,
the other one passes in '0'.

For clarity and less layering violations, factor out the second stage in
csum_tree_block()'s callers.

Suggested-by: Nikolay Borisov <nborisov@suse.com>
Reviewed-by: Qu Wenruo <wqu@suse.com>
Reviewed-by: Nikolay Borisov <nborisov@suse.com>
Signed-off-by: Johannes Thumshirn <jthumshirn@suse.de>
Reviewed-by: David Sterba <dsterba@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
5 years agoLinux 5.1-rc7
Linus Torvalds [Mon, 29 Apr 2019 00:04:13 +0000 (17:04 -0700)]
Linux 5.1-rc7

5 years agoMerge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm
Linus Torvalds [Sun, 28 Apr 2019 17:50:57 +0000 (10:50 -0700)]
Merge tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm

Pull ARM fixes from Russell King:
 "A small number of ARM fixes

   - Fix function tracer and unwinder dependencies so that we don't end
     up building kernels that will crash

   - Fix ARMv7M nommu initialisation (missing register initialisation)

   - Fix EFI decompressor entry (ensuring barrier instructions are
     enabled prior to use)"

* tag 'for-linus' of git://git.armlinux.org.uk/~rmk/linux-arm:
  ARM: 8857/1: efi: enable CP15 DMB instructions before cleaning the cache
  ARM: 8856/1: NOMMU: Fix CCR register faulty initialization when MPU is disabled
  ARM: fix function graph tracer and unwinder dependencies

5 years agoMerge tag 'powerpc-5.1-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sun, 28 Apr 2019 17:43:15 +0000 (10:43 -0700)]
Merge tag 'powerpc-5.1-6' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:
 "A one-liner to make our Radix MMU support depend on HUGETLB_PAGE. We
  use some of the hugetlb inlines (eg. pud_huge()) when operating on the
  linear mapping and if they're compiled into empty wrappers we can
  corrupt memory.

  Then two fixes to our VFIO IOMMU code. The first is not a regression
  but fixes the locking to avoid a user-triggerable deadlock.

  The second does fix a regression since rc1, and depends on the first
  fix. It makes it possible to run guests with large amounts of memory
  again (~256GB).

  Thanks to Alexey Kardashevskiy"

* tag 'powerpc-5.1-6' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  powerpc/mm_iommu: Allow pinning large regions
  powerpc/mm_iommu: Fix potential deadlock
  powerpc/mm/radix: Make Radix require HUGETLB_PAGE

5 years agoMerge tag 'for-linus-20190428' of git://git.kernel.dk/linux-block
Linus Torvalds [Sun, 28 Apr 2019 17:06:32 +0000 (10:06 -0700)]
Merge tag 'for-linus-20190428' of git://git.kernel.dk/linux-block

Pull block fixes from Jens Axboe:
 "A set of io_uring fixes that should go into this release. In
  particular, this contains:

   - The mutex lock vs ctx ref count fix (me)

   - Removal of a dead variable (me)

   - Two race fixes (Stefan)

   - Ring head/tail condition fix for poll full SQ detection (Stefan)"

* tag 'for-linus-20190428' of git://git.kernel.dk/linux-block:
  io_uring: remove 'state' argument from io_{read,write} path
  io_uring: fix poll full SQ detection
  io_uring: fix race condition when sq threads goes sleeping
  io_uring: fix race condition reading SQ entries
  io_uring: fail io_uring_register(2) on a dying io_uring instance

5 years agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma
Linus Torvalds [Sun, 28 Apr 2019 17:00:45 +0000 (10:00 -0700)]
Merge tag 'for-linus' of git://git./linux/kernel/git/rdma/rdma

Pull rdma fixes from Jason Gunthorpe:
 "One core bug fix and a few driver ones

   - FRWR memory registration for hfi1/qib didn't work with with some
     iovas causing a NFSoRDMA failure regression due to a fix in the NFS
     side

   - A command flow error in mlx5 allowed user space to send a corrupt
     command (and also smash the kernel stack we've since learned)

   - Fix a regression and some bugs with device hot unplug that was
     discovered while reviewing Andrea's patches

   - hns has a failure if the user asks for certain QP configurations"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma:
  RDMA/hns: Bugfix for mapping user db
  RDMA/ucontext: Fix regression with disassociate
  RDMA/mlx5: Use rdma_user_map_io for mapping BAR pages
  RDMA/mlx5: Do not allow the user to write to the clock page
  IB/mlx5: Fix scatter to CQE in DCT QP creation
  IB/rdmavt: Fix frwr memory registration

5 years agoMerge tag 'dmaengine-fix-5.1-rc7' of git://git.infradead.org/users/vkoul/slave-dma
Linus Torvalds [Sun, 28 Apr 2019 16:45:18 +0000 (09:45 -0700)]
Merge tag 'dmaengine-fix-5.1-rc7' of git://git.infradead.org/users/vkoul/slave-dma

Pull dmaengine fixes from Vinod Koul:

 - fix for wrong register use in mediatek driver

 - fix in sh driver for glitch is tx_status and treating 0 a valid
   residue for cyclic

 - fix in bcm driver for using right memory allocation flag

* tag 'dmaengine-fix-5.1-rc7' of git://git.infradead.org/users/vkoul/slave-dma:
  dmaengine: mediatek-cqdma: fix wrong register usage in mtk_cqdma_start
  dmaengine: sh: rcar-dmac: Fix glitch in dmaengine_tx_status
  dmaengine: sh: rcar-dmac: With cyclic DMA residue 0 is valid
  dmaengine: bcm2835: Avoid GFP_KERNEL in device_prep_slave_sg

5 years agoMerge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input
Linus Torvalds [Sat, 27 Apr 2019 23:27:02 +0000 (16:27 -0700)]
Merge branch 'for-linus' of git://git./linux/kernel/git/dtor/input

Pull input fixes from Dmitry Torokhov:
 "Just a couple of fixups for Synaptics RMI4 driver and allowing
  snvs_pwrkey to be selected on more boards"

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dtor/input:
  Input: synaptics-rmi4 - write config register values to the right offset
  Input: synaptics-rmi4 - fix possible double free
  Input: snvs_pwrkey - make it depend on ARCH_MXC

5 years agoMerge branch 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 Apr 2019 17:21:29 +0000 (10:21 -0700)]
Merge branch 'x86-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull x86 fixes from Ingo Molnar:

 - Fix an early boot crash in the RSDP parsing code by effectively
   turning off the parsing call - we ran out of time but want to fix the
   regression. The more involved fix is being worked on.

 - Fix a crash that can trigger in the kmemlek code.

* 'x86-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  x86/mm: Fix a crash with kmemleak_scan()
  x86/boot: Disable RSDP parsing temporarily

5 years agoMerge branch 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 Apr 2019 17:18:40 +0000 (10:18 -0700)]
Merge branch 'sched-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull scheduler fix from Ingo Molnar:
 "Fix a division by zero bug that can trigger in the NUMA placement
  code"

* 'sched-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched/numa: Fix a possible divide-by-zero

5 years agoMerge branch 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 27 Apr 2019 16:41:14 +0000 (09:41 -0700)]
Merge branch 'perf-urgent-for-linus' of git://git./linux/kernel/git/tip/tip

Pull perf fix from Ingo Molnar:
 "A cstate event enumeration fix for Kaby/Coffee Lake CPUs"

* 'perf-urgent-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  perf/x86/intel: Update KBL Package C-state events to also include PC8/PC9/PC10 counters

5 years agoslip: make slhc_free() silently accept an error pointer
Linus Torvalds [Thu, 25 Apr 2019 23:13:58 +0000 (16:13 -0700)]
slip: make slhc_free() silently accept an error pointer

This way, slhc_free() accepts what slhc_init() returns, whether that is
an error or not.

In particular, the pattern in sl_alloc_bufs() is

        slcomp = slhc_init(16, 16);
        ...
        slhc_free(slcomp);

for the error handling path, and rather than complicate that code, just
make it ok to always free what was returned by the init function.

That's what the code used to do before commit 4ab42d78e37a ("ppp, slip:
Validate VJ compression slot parameters completely") when slhc_init()
just returned NULL for the error case, with no actual indication of the
details of the error.

Reported-by: syzbot+45474c076a4927533d2e@syzkaller.appspotmail.com
Fixes: 4ab42d78e37a ("ppp, slip: Validate VJ compression slot parameters completely")
Acked-by: Ben Hutchings <ben@decadent.org.uk>
Cc: David Miller <davem@davemloft.net>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agoMerge branch 'akpm' (patches from Andrew)
Linus Torvalds [Sat, 27 Apr 2019 01:15:33 +0000 (18:15 -0700)]
Merge branch 'akpm' (patches from Andrew)

Merge misc fixes from Andrew Morton:
 "9 fixes"

* emailed patches from Andrew Morton <akpm@linux-foundation.org>:
  fs/proc/proc_sysctl.c: Fix a NULL pointer dereference
  mm/page_alloc.c: fix never set ALLOC_NOFRAGMENT flag
  mm/page_alloc.c: avoid potential NULL pointer dereference
  mm, page_alloc: always use a captured page regardless of compaction result
  mm: do not boost watermarks to avoid fragmentation for the DISCONTIG memory model
  lib/test_vmalloc.c: do not create cpumask_t variable on stack
  lib/Kconfig.debug: fix build error without CONFIG_BLOCK
  zram: pass down the bvec we need to read into in the work struct
  mm/memory_hotplug.c: drop memory device reference after find_memory_block()

5 years agoInput: synaptics-rmi4 - write config register values to the right offset
Lucas Stach [Sat, 27 Apr 2019 00:22:01 +0000 (17:22 -0700)]
Input: synaptics-rmi4 - write config register values to the right offset

Currently any changed config register values don't take effect, as the
function to write them back is called with the wrong register offset.

Fixes: ff8f83708b3e (Input: synaptics-rmi4 - add support for 2D
                     sensors and F11)
Signed-off-by: Lucas Stach <l.stach@pengutronix.de>
Reviewed-by: Philipp Zabel <p.zabel@pengutronix.de>
Cc: stable@vger.kernel.org
Signed-off-by: Dmitry Torokhov <dmitry.torokhov@gmail.com>
5 years agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 26 Apr 2019 18:26:53 +0000 (11:26 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Catalin Marinas:

 - keep the tail of an unaligned initrd reserved

 - adjust ftrace_make_call() to deal with the relative nature of PLTs

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64/module: ftrace: deal with place relative nature of PLTs
  arm64: mm: Ensure tail of unaligned initrd is reserved

5 years agoMerge tag 'trace-v5.1-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt...
Linus Torvalds [Fri, 26 Apr 2019 18:09:55 +0000 (11:09 -0700)]
Merge tag 'trace-v5.1-rc6' of git://git./linux/kernel/git/rostedt/linux-trace

Pull tracing fixes from Steven Rostedt:
 "Three tracing fixes:

   - Use "nosteal" for ring buffer splice pages

   - Memory leak fix in error path of trace_pid_write()

   - Fix preempt_enable_no_resched() (use preempt_enable()) in ring
     buffer code"

* tag 'trace-v5.1-rc6' of git://git.kernel.org/pub/scm/linux/kernel/git/rostedt/linux-trace:
  trace: Fix preempt_enable_no_resched() abuse
  tracing: Fix a memory leak by early error exit in trace_pid_write()
  tracing: Fix buffer_ref pipe ops

5 years agoMerge tag 'gpio-v5.1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux...
Linus Torvalds [Fri, 26 Apr 2019 17:46:22 +0000 (10:46 -0700)]
Merge tag 'gpio-v5.1-3' of git://git./linux/kernel/git/linusw/linux-gpio

Pull GPIO fixes from Linus Walleij:
 "Not much to say about them, regular fixes:

   - Fix a bug on the errorpath of gpiochip_add_data_with_key()

   - IRQ type setting on the spreadtrum GPIO driver"

* tag 'gpio-v5.1-3' of git://git.kernel.org/pub/scm/linux/kernel/git/linusw/linux-gpio:
  gpio: Fix gpiochip_add_data_with_key() error path
  gpio: eic: sprd: Fix incorrect irq type setting for the sync EIC

5 years agoMerge tag 'drm-fixes-2019-04-26' of git://anongit.freedesktop.org/drm/drm
Linus Torvalds [Fri, 26 Apr 2019 17:39:46 +0000 (10:39 -0700)]
Merge tag 'drm-fixes-2019-04-26' of git://anongit.freedesktop.org/drm/drm

Pull drm fixes from Dave Airlie:
 "Regular drm fixes, nothing too outstanding, I'm guessing Easter was
  slowing people down.

  i915:
   - FEC enable fix
   - BXT display lanes fix

  ttm:
   - fix reinit for reloading drivers regression

  imx:
   - DP CSC fix

  sun4i:
   - module unload/load fix

  vc4:
   - memory leak fix
   - compile fix

  dw-hdmi:
   - rockchip scdc overflow fix

  sched:
   - docs fix

  vmwgfx:
   - dma api layering fix"

* tag 'drm-fixes-2019-04-26' of git://anongit.freedesktop.org/drm/drm:
  drm/bridge: dw-hdmi: fix SCDC configuration for ddc-i2c-bus
  drm/vmwgfx: Fix dma API layer violation
  drm/vc4: Fix compilation error reported by kbuild test bot
  drm/sun4i: Unbind components before releasing DRM and memory
  drm/vc4: Fix memory leak during gpu reset.
  drm/sched: Fix description of drm_sched_stop
  drm/imx: don't skip DP channel disable for background plane
  gpu: ipu-v3: dp: fix CSC handling
  drm/ttm: fix re-init of global structures
  drm/sun4i: Fix component unbinding and component master deletion
  drm/sun4i: Set device driver data at bind time for use in unbind
  drm/sun4i: Add missing drm_atomic_helper_shutdown at driver unbind
  drm/i915: Restore correct bxt_ddi_phy_calc_lane_lat_optim_mask() calculation
  drm/i915: Do not enable FEC without DSC
  drm: bridge: dw-hdmi: Fix overflow workaround for Rockchip SoCs

5 years agoMerge tag 'for-5.1-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave...
Linus Torvalds [Fri, 26 Apr 2019 16:46:46 +0000 (09:46 -0700)]
Merge tag 'for-5.1-rc6-tag' of git://git./linux/kernel/git/kdave/linux

Pull btrfs fix from David Sterba:
 "One patch to fix a crash in io submission path, due to memory
  allocation errors.

  In short, the multipage bio work that landed in 5.1 caused larger bios
  that in turn require larger temporary memory for checksums. The patch
  is a workaround, we're going to rework the allocation so it does not
  require the vmalloc fallback.

  It took a while to identify that it's caused by patches in 5.1 and not
  a patchset that did some changes in error handling in the code. I've
  tested it on various memory/cpu combinations, it could hit OOM but
  does not crash.

  The timestamp of the patch is less than a day due to updates in the
  changelog, tests were running meanwhile"

* tag 'for-5.1-rc6-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/kdave/linux:
  btrfs: Switch memory allocations in async csum calculation path to kvmalloc

5 years agoMerge tag '5.1-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Fri, 26 Apr 2019 16:45:39 +0000 (09:45 -0700)]
Merge tag '5.1-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6

Pull cifs fixes from Steve French:
 "Three small SMB3 fixes (all for stable as well): two leaks and a
  rename bug"

* tag '5.1-rc6-smb3-fixes' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: fix page reference leak with readv/writev
  cifs: do not attempt cifs operation on smb2+ rename error
  cifs: fix memory leak in SMB2_read

5 years agofs/proc/proc_sysctl.c: Fix a NULL pointer dereference
YueHaibing [Fri, 26 Apr 2019 05:24:05 +0000 (22:24 -0700)]
fs/proc/proc_sysctl.c: Fix a NULL pointer dereference

Syzkaller report this:

  sysctl could not get directory: /net//bridge -12
  kasan: CONFIG_KASAN_INLINE enabled
  kasan: GPF could be caused by NULL-ptr deref or user memory access
  general protection fault: 0000 [#1] SMP KASAN PTI
  CPU: 1 PID: 7027 Comm: syz-executor.0 Tainted: G         C        5.1.0-rc3+ #8
  Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.10.2-1ubuntu1 04/01/2014
  RIP: 0010:__write_once_size include/linux/compiler.h:220 [inline]
  RIP: 0010:__rb_change_child include/linux/rbtree_augmented.h:144 [inline]
  RIP: 0010:__rb_erase_augmented include/linux/rbtree_augmented.h:186 [inline]
  RIP: 0010:rb_erase+0x5f4/0x19f0 lib/rbtree.c:459
  Code: 00 0f 85 60 13 00 00 48 89 1a 48 83 c4 18 5b 5d 41 5c 41 5d 41 5e 41 5f c3 48 89 f2 48 b8 00 00 00 00 00 fc ff df 48 c1 ea 03 <80> 3c 02 00 0f 85 75 0c 00 00 4d 85 ed 4c 89 2e 74 ce 4c 89 ea 48
  RSP: 0018:ffff8881bb507778 EFLAGS: 00010206
  RAX: dffffc0000000000 RBX: ffff8881f224b5b8 RCX: ffffffff818f3f6a
  RDX: 000000000000000a RSI: 0000000000000050 RDI: ffff8881f224b568
  RBP: 0000000000000000 R08: ffffed10376a0ef4 R09: ffffed10376a0ef4
  R10: 0000000000000001 R11: ffffed10376a0ef4 R12: ffff8881f224b558
  R13: 0000000000000000 R14: 0000000000000000 R15: 0000000000000000
  FS:  00007f3e7ce13700(0000) GS:ffff8881f7300000(0000) knlGS:0000000000000000
  CS:  0010 DS: 0000 ES: 0000 CR0: 0000000080050033
  CR2: 00007fd60fbe9398 CR3: 00000001cb55c001 CR4: 00000000007606e0
  DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000
  DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400
  PKRU: 55555554
  Call Trace:
   erase_entry fs/proc/proc_sysctl.c:178 [inline]
   erase_header+0xe3/0x160 fs/proc/proc_sysctl.c:207
   start_unregistering fs/proc/proc_sysctl.c:331 [inline]
   drop_sysctl_table+0x558/0x880 fs/proc/proc_sysctl.c:1631
   get_subdir fs/proc/proc_sysctl.c:1022 [inline]
   __register_sysctl_table+0xd65/0x1090 fs/proc/proc_sysctl.c:1335
   br_netfilter_init+0x68/0x1000 [br_netfilter]
   do_one_initcall+0xbc/0x47d init/main.c:901
   do_init_module+0x1b5/0x547 kernel/module.c:3456
   load_module+0x6405/0x8c10 kernel/module.c:3804
   __do_sys_finit_module+0x162/0x190 kernel/module.c:3898
   do_syscall_64+0x9f/0x450 arch/x86/entry/common.c:290
   entry_SYSCALL_64_after_hwframe+0x49/0xbe
  Modules linked in: br_netfilter(+) backlight comedi(C) hid_sensor_hub max3100 ti_ads8688 udc_core fddi snd_mona leds_gpio rc_streamzap mtd pata_netcell nf_log_common rc_winfast udp_tunnel snd_usbmidi_lib snd_usb_toneport snd_usb_line6 snd_rawmidi snd_seq_device snd_hwdep videobuf2_v4l2 videobuf2_common videodev media videobuf2_vmalloc videobuf2_memops rc_gadmei_rm008z 8250_of smm665 hid_tmff hid_saitek hwmon_vid rc_ati_tv_wonder_hd_600 rc_core pata_pdc202xx_old dn_rtmsg as3722 ad714x_i2c ad714x snd_soc_cs4265 hid_kensington panel_ilitek_ili9322 drm drm_panel_orientation_quirks ipack cdc_phonet usbcore phonet hid_jabra hid extcon_arizona can_dev industrialio_triggered_buffer kfifo_buf industrialio adm1031 i2c_mux_ltc4306 i2c_mux ipmi_msghandler mlxsw_core snd_soc_cs35l34 snd_soc_core snd_pcm_dmaengine snd_pcm snd_timer ac97_bus snd_compress snd soundcore gpio_da9055 uio ecdh_generic mdio_thunder of_mdio fixed_phy libphy mdio_cavium iptable_security iptable_raw iptable_mangle
   iptable_nat nf_nat nf_conntrack nf_defrag_ipv6 nf_defrag_ipv4 iptable_filter bpfilter ip6_vti ip_vti ip_gre ipip sit tunnel4 ip_tunnel hsr veth netdevsim vxcan batman_adv cfg80211 rfkill chnl_net caif nlmon dummy team bonding vcan bridge stp llc ip6_gre gre ip6_tunnel tunnel6 tun joydev mousedev ppdev tpm kvm_intel kvm irqbypass crct10dif_pclmul crc32_pclmul crc32c_intel ghash_clmulni_intel aesni_intel ide_pci_generic piix aes_x86_64 crypto_simd cryptd ide_core glue_helper input_leds psmouse intel_agp intel_gtt serio_raw ata_generic i2c_piix4 agpgart pata_acpi parport_pc parport floppy rtc_cmos sch_fq_codel ip_tables x_tables sha1_ssse3 sha1_generic ipv6 [last unloaded: br_netfilter]
  Dumping ftrace buffer:
     (ftrace buffer empty)
  ---[ end trace 68741688d5fbfe85 ]---

commit 23da9588037e ("fs/proc/proc_sysctl.c: fix NULL pointer
dereference in put_links") forgot to handle start_unregistering() case,
while header->parent is NULL, it calls erase_header() and as seen in the
above syzkaller call trace, accessing &header->parent->root will trigger
a NULL pointer dereference.

As that commit explained, there is also no need to call
start_unregistering() if header->parent is NULL.

Link: http://lkml.kernel.org/r/20190409153622.28112-1-yuehaibing@huawei.com
Fixes: 23da9588037e ("fs/proc/proc_sysctl.c: fix NULL pointer dereference in put_links")
Fixes: 0e47c99d7fe25 ("sysctl: Replace root_list with links between sysctl_table_sets")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reported-by: Hulk Robot <hulkci@huawei.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Luis Chamberlain <mcgrof@kernel.org>
Cc: Alexey Dobriyan <adobriyan@gmail.com>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: "Eric W. Biederman" <ebiederm@xmission.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/page_alloc.c: fix never set ALLOC_NOFRAGMENT flag
Andrey Ryabinin [Fri, 26 Apr 2019 05:24:01 +0000 (22:24 -0700)]
mm/page_alloc.c: fix never set ALLOC_NOFRAGMENT flag

Commit 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
removed setting of the ALLOC_NOFRAGMENT flag.  Bring it back.

The runtime effect is that ALLOC_NOFRAGMENT behaviour is restored so
that allocations are spread across local zones to avoid fragmentation
due to mixing pageblocks as long as possible.

Link: http://lkml.kernel.org/r/20190423120806.3503-2-aryabinin@virtuozzo.com
Fixes: 0a79cdad5eb2 ("mm: use alloc_flags to record if kswapd can wake")
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/page_alloc.c: avoid potential NULL pointer dereference
Andrey Ryabinin [Fri, 26 Apr 2019 05:23:58 +0000 (22:23 -0700)]
mm/page_alloc.c: avoid potential NULL pointer dereference

ac.preferred_zoneref->zone passed to alloc_flags_nofragment() can be NULL.
'zone' pointer unconditionally derefernced in alloc_flags_nofragment().
Bail out on NULL zone to avoid potential crash.  Currently we don't see
any crashes only because alloc_flags_nofragment() has another bug which
allows compiler to optimize away all accesses to 'zone'.

Link: http://lkml.kernel.org/r/20190423120806.3503-1-aryabinin@virtuozzo.com
Fixes: 6bb154504f8b ("mm, page_alloc: spread allocations across zones before introducing fragmentation")
Signed-off-by: Andrey Ryabinin <aryabinin@virtuozzo.com>
Acked-by: Mel Gorman <mgorman@techsingularity.net>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm, page_alloc: always use a captured page regardless of compaction result
Mel Gorman [Fri, 26 Apr 2019 05:23:54 +0000 (22:23 -0700)]
mm, page_alloc: always use a captured page regardless of compaction result

During the development of commit 5e1f0f098b46 ("mm, compaction: capture
a page under direct compaction"), a paranoid check was added to ensure
that if a captured page was available after compaction that it was
consistent with the final state of compaction.  The intent was to catch
serious programming bugs such as using a stale page pointer and causing
corruption problems.

However, it is possible to get a captured page even if compaction was
unsuccessful if an interrupt triggered and happened to free pages in
interrupt context that got merged into a suitable high-order page.  It's
highly unlikely but Li Wang did report the following warning on s390
occuring when testing OOM handling.  Note that the warning is slightly
edited for clarity.

  WARNING: CPU: 0 PID: 9783 at mm/page_alloc.c:3777 __alloc_pages_direct_compact+0x182/0x190
  Modules linked in: rpcsec_gss_krb5 auth_rpcgss nfsv4 dns_resolver nfs
    lockd grace fscache sunrpc pkey ghash_s390 prng xts aes_s390
    des_s390 des_generic sha512_s390 zcrypt_cex4 zcrypt vmur binfmt_misc
    ip_tables xfs libcrc32c dasd_fba_mod qeth_l2 dasd_eckd_mod dasd_mod
    qeth qdio lcs ctcm ccwgroup fsm dm_mirror dm_region_hash dm_log
    dm_mod
  CPU: 0 PID: 9783 Comm: copy.sh Kdump: loaded Not tainted 5.1.0-rc 5 #1

This patch simply removes the check entirely instead of trying to be
clever about pages freed from interrupt context.  If a serious
programming error was introduced, it is highly likely to be caught by
prep_new_page() instead.

Link: http://lkml.kernel.org/r/20190419085133.GH18914@techsingularity.net
Fixes: 5e1f0f098b46 ("mm, compaction: capture a page under direct compaction")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Li Wang <liwang@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm: do not boost watermarks to avoid fragmentation for the DISCONTIG memory model
Mel Gorman [Fri, 26 Apr 2019 05:23:51 +0000 (22:23 -0700)]
mm: do not boost watermarks to avoid fragmentation for the DISCONTIG memory model

Mikulas Patocka reported that commit 1c30844d2dfe ("mm: reclaim small
amounts of memory when an external fragmentation event occurs") "broke"
memory management on parisc.

The machine is not NUMA but the DISCONTIG model creates three pgdats
even though it's a UMA machine for the following ranges

        0) Start 0x0000000000000000 End 0x000000003fffffff Size   1024 MB
        1) Start 0x0000000100000000 End 0x00000001bfdfffff Size   3070 MB
        2) Start 0x0000004040000000 End 0x00000040ffffffff Size   3072 MB

Mikulas reported:

With the patch 1c30844d2, the kernel will incorrectly reclaim the
first zone when it fills up, ignoring the fact that there are two
completely free zones. Basiscally, it limits cache size to 1GiB.

For example, if I run:
# dd if=/dev/sda of=/dev/null bs=1M count=2048

- with the proper kernel, there should be "Buffers - 2GiB"
when this command finishes. With the patch 1c30844d2, buffers
will consume just 1GiB or slightly more, because the kernel was
incorrectly reclaiming them.

The page allocator and reclaim makes assumptions that pgdats really
represent NUMA nodes and zones represent ranges and makes decisions on
that basis.  Watermark boosting for small pgdats leads to unexpected
results even though this would have behaved reasonably on SPARSEMEM.

DISCONTIG is essentially deprecated and even parisc plans to move to
SPARSEMEM so there is no need to be fancy, this patch simply disables
watermark boosting by default on DISCONTIGMEM.

Link: http://lkml.kernel.org/r/20190419094335.GJ18914@techsingularity.net
Fixes: 1c30844d2dfe ("mm: reclaim small amounts of memory when an external fragmentation event occurs")
Signed-off-by: Mel Gorman <mgorman@techsingularity.net>
Reported-by: Mikulas Patocka <mpatocka@redhat.com>
Tested-by: Mikulas Patocka <mpatocka@redhat.com>
Acked-by: Vlastimil Babka <vbabka@suse.cz>
Cc: James Bottomley <James.Bottomley@hansenpartnership.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agolib/test_vmalloc.c: do not create cpumask_t variable on stack
Uladzislau Rezki (Sony) [Fri, 26 Apr 2019 05:23:47 +0000 (22:23 -0700)]
lib/test_vmalloc.c: do not create cpumask_t variable on stack

On my "Intel(R) Xeon(R) W-2135 CPU @ 3.70GHz" system(12 CPUs) i get the
warning from the compiler about frame size:

   warning: the frame size of 1096 bytes is larger than 1024 bytes [-Wframe-larger-than=]

the size of cpumask_t depends on number of CPUs, therefore just make use
of cpumask_of() in set_cpus_allowed_ptr() as a second argument.

Link: http://lkml.kernel.org/r/20190418193925.9361-1-urezki@gmail.com
Signed-off-by: Uladzislau Rezki (Sony) <urezki@gmail.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Roman Gushchin <guro@fb.com>
Cc: Uladzislau Rezki <urezki@gmail.com>
Cc: Michal Hocko <mhocko@suse.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Thomas Garnier <thgarnie@google.com>
Cc: Oleksiy Avramchenko <oleksiy.avramchenko@sonymobile.com>
Cc: Steven Rostedt <rostedt@goodmis.org>
Cc: Joel Fernandes <joelaf@google.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: Ingo Molnar <mingo@elte.hu>
Cc: Tejun Heo <tj@kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agolib/Kconfig.debug: fix build error without CONFIG_BLOCK
YueHaibing [Fri, 26 Apr 2019 05:23:44 +0000 (22:23 -0700)]
lib/Kconfig.debug: fix build error without CONFIG_BLOCK

If CONFIG_TEST_KMOD is set to M, while CONFIG_BLOCK is not set, XFS and
BTRFS can not be compiled successly.

Link: http://lkml.kernel.org/r/20190410075434.35220-1-yuehaibing@huawei.com
Fixes: d9c6a72d6fa2 ("kmod: add test driver to stress test the module loader")
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Reported-by: Hulk Robot <hulkci@huawei.com>
Reviewed-by: Kees Cook <keescook@chromium.org>
Cc: Masahiro Yamada <yamada.masahiro@socionext.com>
Cc: Petr Mladek <pmladek@suse.com>
Cc: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
Cc: Matthew Wilcox <willy@infradead.org>
Cc: Joe Lawrence <joe.lawrence@redhat.com>
Cc: Robin Murphy <robin.murphy@arm.com>
Cc: Luis Chamberlain <mcgrof@kernel.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agozram: pass down the bvec we need to read into in the work struct
Jérôme Glisse [Fri, 26 Apr 2019 05:23:41 +0000 (22:23 -0700)]
zram: pass down the bvec we need to read into in the work struct

When scheduling work item to read page we need to pass down the proper
bvec struct which points to the page to read into.  Before this patch it
uses a randomly initialized bvec (only if PAGE_SIZE != 4096) which is
wrong.

Note that without this patch on arch/kernel where PAGE_SIZE != 4096
userspace could read random memory through a zram block device (thought
userspace probably would have no control on the address being read).

Link: http://lkml.kernel.org/r/20190408183219.26377-1-jglisse@redhat.com
Signed-off-by: Jérôme Glisse <jglisse@redhat.com>
Reviewed-by: Andrew Morton <akpm@linux-foundation.org>
Reviewed-by: Sergey Senozhatsky <sergey.senozhatsky@gmail.com>
Acked-by: Minchan Kim <minchan@kernel.org>
Cc: Nitin Gupta <ngupta@vflare.org>
Cc: <stable@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agomm/memory_hotplug.c: drop memory device reference after find_memory_block()
David Hildenbrand [Fri, 26 Apr 2019 05:23:37 +0000 (22:23 -0700)]
mm/memory_hotplug.c: drop memory device reference after find_memory_block()

Right now we are using find_memory_block() to get the node id for the
pfn range to online.  We are missing to drop a reference to the memory
block device.  While the device still gets unregistered via
device_unregister(), resulting in no user visible problem, the device is
never released via device_release(), resulting in a memory leak.  Fix
that by properly using a put_device().

Link: http://lkml.kernel.org/r/20190411110955.1430-1-david@redhat.com
Fixes: d0dc12e86b31 ("mm/memory_hotplug: optimize memory hotplug")
Signed-off-by: David Hildenbrand <david@redhat.com>
Reviewed-by: Oscar Salvador <osalvador@suse.de>
Reviewed-by: Wei Yang <richard.weiyang@gmail.com>
Acked-by: Michal Hocko <mhocko@suse.com>
Acked-by: Pankaj Gupta <pagupta@redhat.com>
Cc: David Hildenbrand <david@redhat.com>
Cc: Pavel Tatashin <pasha.tatashin@soleen.com>
Cc: Qian Cai <cai@lca.pw>
Cc: Arun KS <arunks@codeaurora.org>
Cc: Mathieu Malaterre <malat@debian.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
5 years agotrace: Fix preempt_enable_no_resched() abuse
Peter Zijlstra [Tue, 23 Apr 2019 20:03:18 +0000 (22:03 +0200)]
trace: Fix preempt_enable_no_resched() abuse

Unless the very next line is schedule(), or implies it, one must not use
preempt_enable_no_resched(). It can cause a preemption to go missing and
thereby cause arbitrary delays, breaking the PREEMPT=y invariant.

Link: http://lkml.kernel.org/r/20190423200318.GY14281@hirez.programming.kicks-ass.net
Cc: Waiman Long <longman@redhat.com>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Will Deacon <will.deacon@arm.com>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: the arch/x86 maintainers <x86@kernel.org>
Cc: Davidlohr Bueso <dave@stgolabs.net>
Cc: Tim Chen <tim.c.chen@linux.intel.com>
Cc: huang ying <huang.ying.caritas@gmail.com>
Cc: Roman Gushchin <guro@fb.com>
Cc: Alexei Starovoitov <ast@kernel.org>
Cc: Daniel Borkmann <daniel@iogearbox.net>
Cc: stable@vger.kernel.org
Fixes: 2c2d7329d8af ("tracing/ftrace: use preempt_enable_no_resched_notrace in ring_buffer_time_stamp()")
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 years agotracing: Fix a memory leak by early error exit in trace_pid_write()
Wenwen Wang [Sat, 20 Apr 2019 02:22:59 +0000 (21:22 -0500)]
tracing: Fix a memory leak by early error exit in trace_pid_write()

In trace_pid_write(), the buffer for trace parser is allocated through
kmalloc() in trace_parser_get_init(). Later on, after the buffer is used,
it is then freed through kfree() in trace_parser_put(). However, it is
possible that trace_pid_write() is terminated due to unexpected errors,
e.g., ENOMEM. In that case, the allocated buffer will not be freed, which
is a memory leak bug.

To fix this issue, free the allocated buffer when an error is encountered.

Link: http://lkml.kernel.org/r/1555726979-15633-1-git-send-email-wang6495@umn.edu
Fixes: f4d34a87e9c10 ("tracing: Use pid bitmap instead of a pid array for set_event_pid")
Cc: stable@vger.kernel.org
Signed-off-by: Wenwen Wang <wang6495@umn.edu>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 years agotracing: Fix buffer_ref pipe ops
Jann Horn [Thu, 4 Apr 2019 21:59:25 +0000 (23:59 +0200)]
tracing: Fix buffer_ref pipe ops

This fixes multiple issues in buffer_pipe_buf_ops:

 - The ->steal() handler must not return zero unless the pipe buffer has
   the only reference to the page. But generic_pipe_buf_steal() assumes
   that every reference to the pipe is tracked by the page's refcount,
   which isn't true for these buffers - buffer_pipe_buf_get(), which
   duplicates a buffer, doesn't touch the page's refcount.
   Fix it by using generic_pipe_buf_nosteal(), which refuses every
   attempted theft. It should be easy to actually support ->steal, but the
   only current users of pipe_buf_steal() are the virtio console and FUSE,
   and they also only use it as an optimization. So it's probably not worth
   the effort.
 - The ->get() and ->release() handlers can be invoked concurrently on pipe
   buffers backed by the same struct buffer_ref. Make them safe against
   concurrency by using refcount_t.
 - The pointers stored in ->private were only zeroed out when the last
   reference to the buffer_ref was dropped. As far as I know, this
   shouldn't be necessary anyway, but if we do it, let's always do it.

Link: http://lkml.kernel.org/r/20190404215925.253531-1-jannh@google.com
Cc: Ingo Molnar <mingo@redhat.com>
Cc: Masami Hiramatsu <mhiramat@kernel.org>
Cc: Al Viro <viro@zeniv.linux.org.uk>
Cc: stable@vger.kernel.org
Fixes: 73a757e63114d ("ring-buffer: Return reader page back into existing ring buffer")
Signed-off-by: Jann Horn <jannh@google.com>
Signed-off-by: Steven Rostedt (VMware) <rostedt@goodmis.org>
5 years agodmaengine: mediatek-cqdma: fix wrong register usage in mtk_cqdma_start
Shun-Chih Yu [Thu, 25 Apr 2019 03:53:50 +0000 (11:53 +0800)]
dmaengine: mediatek-cqdma: fix wrong register usage in mtk_cqdma_start

This patch fixes wrong register usage in the mtk_cqdma_start. The
destination register should be MTK_CQDMA_DST2 instead.

Fixes: b1f01e48df5a ("dmaengine: mediatek: Add MediaTek Command-Queue DMA controller for MT6765 SoC")
Signed-off-by: Shun-Chih Yu <shun-chih.yu@mediatek.com>
Cc: stable@vger.kernel.org
Signed-off-by: Vinod Koul <vkoul@kernel.org>
5 years agoMerge tag 'imx-drm-fixes-2019-04-25' of git://git.pengutronix.de/pza/linux into drm...
Dave Airlie [Fri, 26 Apr 2019 00:32:57 +0000 (10:32 +1000)]
Merge tag 'imx-drm-fixes-2019-04-25' of git://git.pengutronix.de/pza/linux into drm-fixes

drm/imx: fix DP CSC handling

- Fix the DP color space conversion matrix setup to avoid bugs where
  disabling the overlay plane while both primary and overlay plane are
  routed via the CSC unit would not reconfigure the CSC routing
  properly, leaving the display in a nonworking state, or the CSC
  setting from a previously set mode would be left behind, causing
  wrong colors when reenabling the display in certain configurations.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Philipp Zabel <p.zabel@pengutronix.de>
Link: https://patchwork.freedesktop.org/patch/msgid/1556183136.2271.3.camel@pengutronix.de
5 years agoMerge branch 'vmwgfx-fixes-5.1' of git://people.freedesktop.org/~thomash/linux into...
Dave Airlie [Fri, 26 Apr 2019 00:30:17 +0000 (10:30 +1000)]
Merge branch 'vmwgfx-fixes-5.1' of git://people.freedesktop.org/~thomash/linux into drm-fixes

A single fix for a layer violation requested by Cristoph.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Thomas Hellstrom <thellstrom@vmware.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190425204100.3982-1-thellstrom@vmware.com
5 years agoMerge tag 'drm-misc-fixes-2019-04-25' of git://anongit.freedesktop.org/drm/drm-misc...
Dave Airlie [Fri, 26 Apr 2019 00:29:07 +0000 (10:29 +1000)]
Merge tag 'drm-misc-fixes-2019-04-25' of git://anongit.freedesktop.org/drm/drm-misc into drm-fixes

- sun4i: Fix module loading / unloading
- vc4: Fix a compilation error and memory leak
- dw-hdmi: Fix an overflow on Rockchip and SCDC configuration

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Maxime Ripard <maxime.ripard@bootlin.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190425132739.pngmfiqucqmulxkz@flea
5 years agoMerge branch 'drm-fixes-5.1' of git://people.freedesktop.org/~agd5f/linux into drm...
Dave Airlie [Fri, 26 Apr 2019 00:25:57 +0000 (10:25 +1000)]
Merge branch 'drm-fixes-5.1' of git://people.freedesktop.org/~agd5f/linux into drm-fixes

- ttm regression fix
- sched documentation fix

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Alex Deucher <alexdeucher@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424230120.3423-1-alexander.deucher@amd.com
5 years agoMerge tag 'drm-intel-fixes-2019-04-24' of git://anongit.freedesktop.org/drm/drm-intel...
Dave Airlie [Fri, 26 Apr 2019 00:13:49 +0000 (10:13 +1000)]
Merge tag 'drm-intel-fixes-2019-04-24' of git://anongit.freedesktop.org/drm/drm-intel into drm-fixes

A fix for display lanes calculation for BXT and a protection
to avoid enabling FEC without DSC.

Signed-off-by: Dave Airlie <airlied@redhat.com>
From: Rodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424215359.GA26100@intel.com
5 years agosched/numa: Fix a possible divide-by-zero
Xie XiuQi [Sat, 20 Apr 2019 08:34:16 +0000 (16:34 +0800)]
sched/numa: Fix a possible divide-by-zero

sched_clock_cpu() may not be consistent between CPUs. If a task
migrates to another CPU, then se.exec_start is set to that CPU's
rq_clock_task() by update_stats_curr_start(). Specifically, the new
value might be before the old value due to clock skew.

So then if in numa_get_avg_runtime() the expression:

  'now - p->last_task_numa_placement'

ends up as -1, then the divider '*period + 1' in task_numa_placement()
is 0 and things go bang. Similar to update_curr(), check if time goes
backwards to avoid this.

[ peterz: Wrote new changelog. ]
[ mingo: Tweaked the code comment. ]

Signed-off-by: Xie XiuQi <xiexiuqi@huawei.com>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: cj.chengjian@huawei.com
Cc: <stable@vger.kernel.org>
Link: http://lkml.kernel.org/r/20190425080016.GX11158@hirez.programming.kicks-ass.net
Signed-off-by: Ingo Molnar <mingo@kernel.org>
5 years agoMerge tag 'ceph-for-5.1-rc7' of git://github.com/ceph/ceph-client
Linus Torvalds [Thu, 25 Apr 2019 17:48:50 +0000 (10:48 -0700)]
Merge tag 'ceph-for-5.1-rc7' of git://github.com/ceph/ceph-client

Pull ceph fixes from Ilya Dryomov:
 "dentry name handling fixes from Jeff and a memory leak fix from Zheng.

  Both are old issues, marked for stable"

* tag 'ceph-for-5.1-rc7' of git://github.com/ceph/ceph-client:
  ceph: fix ci->i_head_snapc leak
  ceph: handle the case where a dentry has been renamed on outstanding req
  ceph: ensure d_name stability in ceph_dentry_hash()
  ceph: only use d_name directly when parent is locked

5 years agoMerge branch 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6
Linus Torvalds [Thu, 25 Apr 2019 16:15:03 +0000 (09:15 -0700)]
Merge branch 'linus' of git://git./linux/kernel/git/herbert/crypto-2.6

Pull crypto fixes from Herbert Xu:
 "This fixes a bug in xts and lrw where they may sleep in an atomic
  context"

* 'linus' of git://git.kernel.org/pub/scm/linux/kernel/git/herbert/crypto-2.6:
  crypto: lrw - Fix atomic sleep when walking skcipher
  crypto: xts - Fix atomic sleep when walking skcipher

5 years agoRDMA/hns: Bugfix for mapping user db
Lijun Ou [Tue, 23 Apr 2019 09:30:26 +0000 (17:30 +0800)]
RDMA/hns: Bugfix for mapping user db

When the maximum send wr delivered by the user is zero, the qp does not
have a sq.

When allocating the sq db buffer to store the user sq pi pointer and map
it to the kernel mode, max_send_wr is used as the trigger condition, while
the kernel does not consider the max_send_wr trigger condition when
mapmping db. It will cause sq record doorbell map fail and create qp fail.

The failed print information as follows:

 hns3 0000:7d:00.1: Send cmd: tail - 418, opcode - 0x8504, flag - 0x0011, retval - 0x0000
 hns3 0000:7d:00.1: Send cmd: 0xe59dc000 0x00000000 0x00000000 0x00000000 0x00000116 0x0000ffff
 hns3 0000:7d:00.1: sq record doorbell map failed!
 hns3 0000:7d:00.1: Create RC QP failed

Fixes: 0425e3e6e0c7 ("RDMA/hns: Support flush cqe for hip08 in kernel space")
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Jason Gunthorpe <jgg@mellanox.com>