platform/kernel/linux-starfive.git
15 months agoext4: move setting of trimmed bit into ext4_try_to_trim_range()
Jan Kara [Wed, 13 Sep 2023 15:04:54 +0000 (17:04 +0200)]
ext4: move setting of trimmed bit into ext4_try_to_trim_range()

Currently we set the group's trimmed bit in ext4_trim_all_free() based
on return value of ext4_try_to_trim_range(). However when we will want
to abort trimming because of suspend attempt, we want to return success
from ext4_try_to_trim_range() but not set the trimmed bit. Instead
implementing awkward propagation of this information, just move setting
of trimmed bit into ext4_try_to_trim_range() when the whole group is
trimmed.

Cc: stable@kernel.org
Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230913150504.9054-1-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agojbd2: Fix memory leak in journal_init_common()
Li Zetao [Mon, 11 Sep 2023 02:51:38 +0000 (10:51 +0800)]
jbd2: Fix memory leak in journal_init_common()

There is a memory leak reported by kmemleak:

  unreferenced object 0xff11000105903b80 (size 64):
    comm "mount", pid 3382, jiffies 4295032021 (age 27.826s)
    hex dump (first 32 bytes):
      04 00 00 00 00 00 00 00 01 00 00 00 00 00 00 00  ................
      ff ff ff ff 00 00 00 00 00 00 00 00 00 00 00 00  ................
    backtrace:
      [<ffffffffae86ac40>] __kmalloc_node+0x50/0x160
      [<ffffffffaf2486d8>] crypto_alloc_tfmmem.isra.0+0x38/0x110
      [<ffffffffaf2498e5>] crypto_create_tfm_node+0x85/0x2f0
      [<ffffffffaf24a92c>] crypto_alloc_tfm_node+0xfc/0x210
      [<ffffffffaedde777>] journal_init_common+0x727/0x1ad0
      [<ffffffffaede1715>] jbd2_journal_init_inode+0x2b5/0x500
      [<ffffffffaed786b5>] ext4_load_and_init_journal+0x255/0x2440
      [<ffffffffaed8b423>] ext4_fill_super+0x8823/0xa330
      ...

The root cause was traced to an error handing path in journal_init_common()
when malloc memory failed in register_shrinker(). The checksum driver is
used to reference to checksum algorithm via cryptoapi and the user should
release the memory when the driver is no longer needed or the journal
initialization failed.

Fix it by calling crypto_free_shash() on the "err_cleanup" error handing
path in journal_init_common().

Fixes: c30713084ba5 ("jbd2: move load_superblock() into journal_init_common()")
Signed-off-by: Li Zetao <lizetao1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230911025138.983101-1-lizetao1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agojbd2: Remove page size assumptions
Ritesh Harjani (IBM) [Thu, 7 Sep 2023 13:47:32 +0000 (09:47 -0400)]
jbd2: Remove page size assumptions

jbd2_alloc() allocates a buffer from slab when the block size is smaller
than PAGE_SIZE, and slab may be using a compound page.  Before commit
8147c4c4546f, we set b_page to the precise page containing the buffer
and this code worked well.  Now we set b_page to the head page of the
allocation, so we can no longer use offset_in_page().  While we could
do a 1:1 replacement with offset_in_folio(), use the more idiomatic
bh_offset() and the folio APIs to map the buffer.

This isn't enough to support a b_size larger than PAGE_SIZE on HIGHMEM
machines, but this is good enough to fix the actual bug we're seeing.

Fixes: 8147c4c4546f ("jbd2: use a folio in jbd2_journal_write_metadata_buffer()")
Reported-by: Zorro Lang <zlang@kernel.org>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
[converted to be more folio]
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
15 months agobuffer: Make bh_offset() work for compound pages
Matthew Wilcox (Oracle) [Thu, 7 Sep 2023 13:40:07 +0000 (09:40 -0400)]
buffer: Make bh_offset() work for compound pages

If the buffer pointed to by the buffer_head is part of a compound page,
bh_offset() assumes that b_page is the precise page that contains
the data.  A recent change to jbd2 inadvertently violated that assumption.

By using page_size(), we support both b_page being set to the head page
(as page_size() will return the size of the entire folio) and the precise
page (as it will return PAGE_SIZE for a tail page).

Fixes: 8147c4c4546f ("jbd2: use a folio in jbd2_journal_write_metadata_buffer()")
Reported-by: Zorro Lang <zlang@kernel.org>
Tested-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Signed-off-by: Matthew Wilcox (Oracle) <willy@infradead.org>
15 months agoext4: fix slab-use-after-free in ext4_es_insert_extent()
Baokun Li [Tue, 15 Aug 2023 07:08:08 +0000 (15:08 +0800)]
ext4: fix slab-use-after-free in ext4_es_insert_extent()

Yikebaer reported an issue:
==================================================================
BUG: KASAN: slab-use-after-free in ext4_es_insert_extent+0xc68/0xcb0
fs/ext4/extents_status.c:894
Read of size 4 at addr ffff888112ecc1a4 by task syz-executor/8438

CPU: 1 PID: 8438 Comm: syz-executor Not tainted 6.5.0-rc5 #1
Call Trace:
 [...]
 kasan_report+0xba/0xf0 mm/kasan/report.c:588
 ext4_es_insert_extent+0xc68/0xcb0 fs/ext4/extents_status.c:894
 ext4_map_blocks+0x92a/0x16f0 fs/ext4/inode.c:680
 ext4_alloc_file_blocks.isra.0+0x2df/0xb70 fs/ext4/extents.c:4462
 ext4_zero_range fs/ext4/extents.c:4622 [inline]
 ext4_fallocate+0x251c/0x3ce0 fs/ext4/extents.c:4721
 [...]

Allocated by task 8438:
 [...]
 kmem_cache_zalloc include/linux/slab.h:693 [inline]
 __es_alloc_extent fs/ext4/extents_status.c:469 [inline]
 ext4_es_insert_extent+0x672/0xcb0 fs/ext4/extents_status.c:873
 ext4_map_blocks+0x92a/0x16f0 fs/ext4/inode.c:680
 ext4_alloc_file_blocks.isra.0+0x2df/0xb70 fs/ext4/extents.c:4462
 ext4_zero_range fs/ext4/extents.c:4622 [inline]
 ext4_fallocate+0x251c/0x3ce0 fs/ext4/extents.c:4721
 [...]

Freed by task 8438:
 [...]
 kmem_cache_free+0xec/0x490 mm/slub.c:3823
 ext4_es_try_to_merge_right fs/ext4/extents_status.c:593 [inline]
 __es_insert_extent+0x9f4/0x1440 fs/ext4/extents_status.c:802
 ext4_es_insert_extent+0x2ca/0xcb0 fs/ext4/extents_status.c:882
 ext4_map_blocks+0x92a/0x16f0 fs/ext4/inode.c:680
 ext4_alloc_file_blocks.isra.0+0x2df/0xb70 fs/ext4/extents.c:4462
 ext4_zero_range fs/ext4/extents.c:4622 [inline]
 ext4_fallocate+0x251c/0x3ce0 fs/ext4/extents.c:4721
 [...]
==================================================================

The flow of issue triggering is as follows:
1. remove es
      raw es               es  removed  es1
|-------------------| -> |----|.......|------|

2. insert es
  es   insert   es1      merge with es  es1     merge with es and free es1
|----|.......|------| -> |------------|------| -> |-------------------|

es merges with newes, then merges with es1, frees es1, then determines
if es1->es_len is 0 and triggers a UAF.

The code flow is as follows:
ext4_es_insert_extent
  es1 = __es_alloc_extent(true);
  es2 = __es_alloc_extent(true);
  __es_remove_extent(inode, lblk, end, NULL, es1)
    __es_insert_extent(inode, &newes, es1) ---> insert es1 to es tree
  __es_insert_extent(inode, &newes, es2)
    ext4_es_try_to_merge_right
      ext4_es_free_extent(inode, es1) --->  es1 is freed
  if (es1 && !es1->es_len)
    // Trigger UAF by determining if es1 is used.

We determine whether es1 or es2 is used immediately after calling
__es_remove_extent() or __es_insert_extent() to avoid triggering a
UAF if es1 or es2 is freed.

Reported-by: Yikebaer Aizezi <yikebaer61@gmail.com>
Closes: https://lore.kernel.org/lkml/CALcu4raD4h9coiyEBL4Bm0zjDwxC2CyPiTwsP3zFuhot6y9Beg@mail.gmail.com
Fixes: 2a69c450083d ("ext4: using nofail preallocation in ext4_es_insert_extent()")
Cc: stable@kernel.org
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230815070808.3377171-1-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agolibfs: remove redundant checks of s_encoding
Eric Biggers [Mon, 14 Aug 2023 18:29:03 +0000 (11:29 -0700)]
libfs: remove redundant checks of s_encoding

Now that neither ext4 nor f2fs allows inodes with the casefold flag to
be instantiated when unsupported, it's unnecessary to repeatedly check
for support later on during random filesystem operations.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20230814182903.37267-4-ebiggers@kernel.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: remove redundant checks of s_encoding
Eric Biggers [Mon, 14 Aug 2023 18:29:02 +0000 (11:29 -0700)]
ext4: remove redundant checks of s_encoding

Now that ext4 does not allow inodes with the casefold flag to be
instantiated when unsupported, it's unnecessary to repeatedly check for
support later on during random filesystem operations.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20230814182903.37267-3-ebiggers@kernel.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: reject casefold inode flag without casefold feature
Eric Biggers [Mon, 14 Aug 2023 18:29:01 +0000 (11:29 -0700)]
ext4: reject casefold inode flag without casefold feature

It is invalid for the casefold inode flag to be set without the casefold
superblock feature flag also being set.  e2fsck already considers this
case to be invalid and handles it by offering to clear the casefold flag
on the inode.  __ext4_iget() also already considered this to be invalid,
sort of, but it only got so far as logging an error message; it didn't
actually reject the inode.  Make it reject the inode so that other code
doesn't have to handle this case.  This matches what f2fs does.

Note: we could check 's_encoding != NULL' instead of
ext4_has_feature_casefold().  This would make the check robust against
the casefold feature being enabled by userspace writing to the page
cache of the mounted block device.  However, it's unsolvable in general
for filesystems to be robust against concurrent writes to the page cache
of the mounted block device.  Though this very particular scenario
involving the casefold feature is solvable, we should not pretend that
we can support this model, so let's just check the casefold feature.
tune2fs already forbids enabling casefold on a mounted filesystem.

Signed-off-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20230814182903.37267-2-ebiggers@kernel.org
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: use LIST_HEAD() to initialize the list_head in mballoc.c
Ruan Jinjie [Sat, 12 Aug 2023 07:18:39 +0000 (15:18 +0800)]
ext4: use LIST_HEAD() to initialize the list_head in mballoc.c

Use LIST_HEAD() to initialize the list_head instead of open-coding it.

Signed-off-by: Ruan Jinjie <ruanjinjie@huawei.com>
Link: https://lore.kernel.org/r/20230812071839.3481909-1-ruanjinjie@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: do not mark inode dirty every time when appending using delalloc
Liu Song [Thu, 10 Aug 2023 15:43:33 +0000 (23:43 +0800)]
ext4: do not mark inode dirty every time when appending using delalloc

In the delalloc append write scenario, if inode's i_size is extended due
to buffer write, there are delalloc writes pending in the range up to
i_size, and no need to touch i_disksize since writeback will push
i_disksize up to i_size eventually. Offers significant performance
improvement in high-frequency append write scenarios.

I conducted tests in my 32-core environment by launching 32 concurrent
threads to append write to the same file. Each write operation had a
length of 1024 bytes and was repeated 100000 times. Without using this
patch, the test was completed in 7705 ms. However, with this patch, the
test was completed in 5066 ms, resulting in a performance improvement of
34%.

Moreover, in test scenarios of Kafka version 2.6.2, using packet size of
2K, with this patch resulted in a 10% performance improvement.

Signed-off-by: Liu Song <liusong@linux.alibaba.com>
Suggested-by: Jan Kara <jack@suse.cz>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230810154333.84921-1-liusong@linux.alibaba.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: rename s_error_work to s_sb_upd_work
Theodore Ts'o [Wed, 23 Aug 2023 03:43:38 +0000 (23:43 -0400)]
ext4: rename s_error_work to s_sb_upd_work

The most common use that s_error_work will get scheduled is now the
periodic update of the superblock.  So rename it to s_sb_upd_work.

Also rename the function flush_stashed_error_work() to
update_super_work().

Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: add periodic superblock update check
Vitaliy Kuznetsov [Thu, 10 Aug 2023 14:38:52 +0000 (18:38 +0400)]
ext4: add periodic superblock update check

This patch introduces a mechanism to periodically check and update
the superblock within the ext4 file system. The main purpose of this
patch is to keep the disk superblock up to date. The update will be
performed if more than one hour has passed since the last update, and
if more than 16MB of data have been written to disk.

This check and update is performed within the ext4_journal_commit_callback
function, ensuring that the superblock is written while the disk is
active, rather than based on a timer that may trigger during disk idle
periods.

Discussion https://www.spinics.net/lists/linux-ext4/msg85865.html

Signed-off-by: Vitaliy Kuznetsov <vk.en.mail@gmail.com>
Link: https://lore.kernel.org/r/20230810143852.40228-1-vk.en.mail@gmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: drop dio overwrite only flag and associated warning
Brian Foster [Thu, 10 Aug 2023 16:55:59 +0000 (12:55 -0400)]
ext4: drop dio overwrite only flag and associated warning

The commit referenced below opened up concurrent unaligned dio under
shared locking for pure overwrites. In doing so, it enabled use of
the IOMAP_DIO_OVERWRITE_ONLY flag and added a warning on unexpected
-EAGAIN returns as an extra precaution, since ext4 does not retry
writes in such cases. The flag itself is advisory in this case since
ext4 checks for unaligned I/Os and uses appropriate locking up
front, rather than on a retry in response to -EAGAIN.

As it turns out, the warning check is susceptible to false positives
because there are scenarios where -EAGAIN can be expected from lower
layers without necessarily having IOCB_NOWAIT set on the iocb. For
example, one instance of the warning has been seen where io_uring
sets IOCB_HIPRI, which in turn results in REQ_POLLED|REQ_NOWAIT on
the bio. This results in -EAGAIN if the block layer is unable to
allocate a request, etc. [Note that there is an outstanding patch to
untangle REQ_POLLED and REQ_NOWAIT such that the latter relies on
IOCB_NOWAIT, which would also address this instance of the warning.]

Another instance of the warning has been reproduced by syzbot. A dio
write is interrupted down in __get_user_pages_locked() waiting on
the mm lock and returns -EAGAIN up the stack. If the iomap dio
iteration layer has made no progress on the write to this point,
-EAGAIN returns up to the filesystem and triggers the warning.

This use of the overwrite flag in ext4 is precautionary and
half-baked. I.e., ext4 doesn't actually implement overwrite checking
in the iomap callbacks when the flag is set, so the only extra
verification it provides are i_size checks in the generic iomap dio
layer. Combined with the tendency for false positives, the added
verification is not worth the extra trouble. Remove the flag,
associated warning, and update the comments to document when
concurrent unaligned dio writes are allowed and why said flag is not
used.

Cc: stable@kernel.org
Reported-by: syzbot+5050ad0fb47527b1808a@syzkaller.appspotmail.com
Reported-by: Pengfei Xu <pengfei.xu@intel.com>
Fixes: 310ee0902b8d ("ext4: allow concurrent unaligned dio overwrites")
Signed-off-by: Brian Foster <bfoster@redhat.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230810165559.946222-1-bfoster@redhat.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: add correct group descriptors and reserved GDT blocks to system zone
Wang Jianjian [Wed, 2 Aug 2023 16:28:39 +0000 (00:28 +0800)]
ext4: add correct group descriptors and reserved GDT blocks to system zone

When setup_system_zone, flex_bg is not initialized so it is always 1.
Use a new helper function, ext4_num_base_meta_blocks() which does not
depend on sbi->s_log_groups_per_flex being initialized.

[ Squashed two patches in the Link URL's below together into a single
  commit, which is simpler to review/understand.  Also fix checkpatch
  warnings. --TYT ]

Cc: stable@kernel.org
Signed-off-by: Wang Jianjian <wangjianjian0@foxmail.com>
Link: https://lore.kernel.org/r/tencent_21AF0D446A9916ED5C51492CC6C9A0A77B05@qq.com
Link: https://lore.kernel.org/r/tencent_D744D1450CC169AEA77FCF0A64719909ED05@qq.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: remove unused function declaration
Cai Xinchen [Wed, 2 Aug 2023 03:00:25 +0000 (03:00 +0000)]
ext4: remove unused function declaration

These functions do not have its function implementation.
So those function declaration is useless. Remove these

Signed-off-by: Cai Xinchen <caixinchen1@huawei.com>
Link: https://lore.kernel.org/r/20230802030025.173148-1-caixinchen1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: mballoc: avoid garbage value from err
Su Hui [Tue, 25 Jul 2023 04:33:11 +0000 (12:33 +0800)]
ext4: mballoc: avoid garbage value from err

clang's static analysis warning: fs/ext4/mballoc.c
line 4178, column 6, Branch condition evaluates to a garbage value.

err is uninitialized and will be judged when 'len <= 0' or
it first enters the loop while the condition "!ext4_sb_block_valid()"
is true. Although this can't make problems now, it's better to
correct it.

Signed-off-by: Su Hui <suhui@nfschina.com>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Link: https://lore.kernel.org/r/20230725043310.1227621-1-suhui@nfschina.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: use sbi instead of EXT4_SB(sb) in ext4_mb_new_blocks_simple()
Lu Hongfei [Fri, 7 Jul 2023 11:59:07 +0000 (19:59 +0800)]
ext4: use sbi instead of EXT4_SB(sb) in ext4_mb_new_blocks_simple()

Signed-off-by: Lu Hongfei <luhongfei@vivo.com>
Link: https://lore.kernel.org/r/20230707115907.26637-1-luhongfei@vivo.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: change the type of blocksize in ext4_mb_init_cache()
Lu Hongfei [Fri, 7 Jul 2023 10:55:16 +0000 (18:55 +0800)]
ext4: change the type of blocksize in ext4_mb_init_cache()

The return value type of i_blocksize() is 'unsigned int', so the
type of blocksize has been modified from 'int' to 'unsigned int'
to ensure data type consistency.

Signed-off-by: Lu Hongfei <luhongfei@vivo.com>
Link: https://lore.kernel.org/r/20230707105516.9156-1-luhongfei@vivo.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: fix unttached inode after power cut with orphan file feature enabled
Zhihao Cheng [Wed, 28 Jun 2023 13:20:11 +0000 (21:20 +0800)]
ext4: fix unttached inode after power cut with orphan file feature enabled

Running generic/475(filesystem consistent tests after power cut) could
easily trigger unattached inode error while doing fsck:
  Unattached zero-length inode 39405.  Clear? no

  Unattached inode 39405
  Connect to /lost+found? no

Above inconsistence is caused by following process:
       P1                       P2
ext4_create
 inode = ext4_new_inode_start_handle  // itable records nlink=1
 ext4_add_nondir
   err = ext4_add_entry  // ENOSPC
    ext4_append
     ext4_bread
      ext4_getblk
       ext4_map_blocks // returns ENOSPC
   drop_nlink(inode) // won't be updated into disk inode
   ext4_orphan_add(handle, inode)
    ext4_orphan_file_add
 ext4_journal_stop(handle)
      jbd2_journal_commit_transaction // commit success
              >> power cut <<
ext4_fill_super
 ext4_load_and_init_journal   // itable records nlink=1
 ext4_orphan_cleanup
  ext4_process_orphan
   if (inode->i_nlink)        // true, inode won't be deleted

Then, allocated inode will be reserved on disk and corresponds to no
dentries, so e2fsck reports 'unattached inode' problem.

The problem won't happen if orphan file feature is disabled, because
ext4_orphan_add() will update disk inode in orphan list mode. There
are several places not updating disk inode while putting inode into
orphan area, such as ext4_add_nondir(), ext4_symlink() and whiteout
in ext4_rename(). Fix it by updating inode into disk in all error
branches of these places.

Link: https://bugzilla.kernel.org/show_bug.cgi?id=217605
Fixes: 02f310fcf47f ("ext4: Speedup ext4 orphan inode handling")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230628132011.650383-1-chengzhihao1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agojbd2: correct the end of the journal recovery scan range
Zhang Yi [Mon, 26 Jun 2023 07:33:22 +0000 (15:33 +0800)]
jbd2: correct the end of the journal recovery scan range

We got a filesystem inconsistency issue below while running generic/475
I/O failure pressure test with fast_commit feature enabled.

 Symlink /p3/d3/d1c/d6c/dd6/dce/l101 (inode #132605) is invalid.

If fast_commit feature is enabled, a special fast_commit journal area is
appended to the end of the normal journal area. The journal->j_last
point to the first unused block behind the normal journal area instead
of the whole log area, and the journal->j_fc_last point to the first
unused block behind the fast_commit journal area. While doing journal
recovery, do_one_pass(PASS_SCAN) should first scan the normal journal
area and turn around to the first block once it meet journal->j_last,
but the wrap() macro misuse the journal->j_fc_last, so the recovering
could not read the next magic block (commit block perhaps) and would end
early mistakenly and missing tN and every transaction after it in the
following example. Finally, it could lead to filesystem inconsistency.

 | normal journal area                             | fast commit area |
 +-------------------------------------------------+------------------+
 | tN(rere) | tN+1 |~| tN-x |...| tN-1 | tN(front) |       ....       |
 +-------------------------------------------------+------------------+
                     /                             /                  /
                start               journal->j_last journal->j_fc_last

This patch fix it by use the correct ending journal->j_last.

Fixes: 5b849b5f96b4 ("jbd2: fast commit recovery path")
Cc: stable@kernel.org
Reported-by: Theodore Ts'o <tytso@mit.edu>
Link: https://lore.kernel.org/linux-ext4/20230613043120.GB1584772@mit.edu/
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230626073322.3956567-1-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
15 months agoext4: ext4_get_{dev}_journal return proper error value
Zhang Yi [Fri, 11 Aug 2023 06:36:10 +0000 (14:36 +0800)]
ext4: ext4_get_{dev}_journal return proper error value

ext4_get_journal() and ext4_get_dev_journal() return NULL if they failed
to init journal, making them return proper error value instead, also
rename them to ext4_open_{inode,dev}_journal().

[ Folded fix to ext4_calculate_overhead() to check for an ERR_PTR
  instead of NULL. ]

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-13-yi.zhang@huaweicloud.com
Reported-by: syzbot+b3123e6d9842e526de39@syzkaller.appspotmail.com
Link: https://lore.kernel.org/r/20230826011029.2023140-1-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: cleanup ext4_get_dev_journal() and ext4_get_journal()
Zhang Yi [Fri, 11 Aug 2023 06:36:09 +0000 (14:36 +0800)]
ext4: cleanup ext4_get_dev_journal() and ext4_get_journal()

Factor out a new helper form ext4_get_dev_journal() to get external
journal bdev and check validation of this device, drop ext4_blkdev_get()
helper, and also remove duplicate check of journal feature. It makes
ext4_get_dev_journal() more clear than before.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-12-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: jbd2_journal_init_{dev,inode} return proper error return value
Zhang Yi [Fri, 11 Aug 2023 06:36:08 +0000 (14:36 +0800)]
jbd2: jbd2_journal_init_{dev,inode} return proper error return value

Current jbd2_journal_init_{dev,inode} return NULL if some error
happens, make them to pass out proper error return value.

[ Fix from Yang Yingliang folded in. ]

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-11-yi.zhang@huaweicloud.com
Link: https://lore.kernel.org/r/20230822030018.644419-1-yangyingliang@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: drop useless error tag in jbd2_journal_wipe()
Zhang Yi [Fri, 11 Aug 2023 06:36:07 +0000 (14:36 +0800)]
jbd2: drop useless error tag in jbd2_journal_wipe()

no_recovery is redundant, just drop it.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-10-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: cleanup journal_init_common()
Zhang Yi [Fri, 11 Aug 2023 06:36:06 +0000 (14:36 +0800)]
jbd2: cleanup journal_init_common()

Adjust the initialization sequence and error handle of journal_t, moving
load superblock to the begin, and classify others initialization.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-9-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: add fast_commit space check
Zhang Yi [Fri, 11 Aug 2023 06:36:05 +0000 (14:36 +0800)]
jbd2: add fast_commit space check

If JBD2_FEATURE_INCOMPAT_FAST_COMMIT bit is set, it means the journal
have fast commit records need to recover, so the fast commit size
should not be too large, and the leftover normal journal size should
never less than JBD2_MIN_JOURNAL_BLOCKS. If it happens, the
journal->j_last is likely to be wrong and will probably lead to
incorrect journal recovery. So add a check into the
journal_check_superblock(), and drop the pointless check when
initializing the fastcommit parameters.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-8-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: cleanup load_superblock()
Zhang Yi [Fri, 11 Aug 2023 06:36:04 +0000 (14:36 +0800)]
jbd2: cleanup load_superblock()

Rename load_superblock() to journal_load_superblock(), move getting and
reading superblock from journal_init_common() and
journal_get_superblock() to this function, and also rename
journal_get_superblock() to journal_check_superblock(), make it a pure
check helper to check superblock validity from disk.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-7-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: open code jbd2_verify_csum_type() helper
Zhang Yi [Fri, 11 Aug 2023 06:36:03 +0000 (14:36 +0800)]
jbd2: open code jbd2_verify_csum_type() helper

jbd2_verify_csum_type() helper check checksum type in the superblock for
v2 or v3 checksum feature, it always return true if these features are
not enabled, and it has only one user, so open code it is more clear.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-6-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: checking valid features early in journal_get_superblock()
Zhang Yi [Fri, 11 Aug 2023 06:36:02 +0000 (14:36 +0800)]
jbd2: checking valid features early in journal_get_superblock()

journal_get_superblock() is used to check validity of the jounal
supberblock, so move the features checks from jbd2_journal_load() to
journal_get_superblock().

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-5-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: don't load superblock in jbd2_journal_check_used_features()
Zhang Yi [Fri, 11 Aug 2023 06:36:01 +0000 (14:36 +0800)]
jbd2: don't load superblock in jbd2_journal_check_used_features()

Since load_superblock() has been moved to journal_init_common(), the
in-memory superblock structure is initialized and contains valid data
once the file system has a journal_t object, so it's safe to access it,
let's drop the call to journal_get_superblock() from
jbd2_journal_check_used_features() and also drop the setting/clearing of
the veirfy bit of the superblock buffer.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-4-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: move load_superblock() into journal_init_common()
Zhang Yi [Fri, 11 Aug 2023 06:36:00 +0000 (14:36 +0800)]
jbd2: move load_superblock() into journal_init_common()

Move the call to load_superblock() from jbd2_journal_load() and
jbd2_journal_wipe() early into journal_init_common(), the journal
superblock gets read and the in-memory journal_t structure gets
initialised after calling jbd2_journal_init_{dev,inode}, it's safe to
do following initialization according to it.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-3-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: move load_superblock() dependent functions
Zhang Yi [Fri, 11 Aug 2023 06:35:59 +0000 (14:35 +0800)]
jbd2: move load_superblock() dependent functions

Move load_superblock() declaration and the functions it calls before
journal_init_common(). This is a preparation for moving a call to
load_superblock() from jbd2_journal_load() and jbd2_journal_wipe() to
journal_init_common(). No functional changes.

Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230811063610.2980059-2-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: don't use CR_BEST_AVAIL_LEN for non-regular files
Ritesh Harjani [Sun, 16 Jul 2023 14:03:34 +0000 (19:33 +0530)]
ext4: don't use CR_BEST_AVAIL_LEN for non-regular files

Using CR_BEST_AVAIL_LEN only make sense for regular files, as for
non-regular files we never normalize the allocation request length i.e.
goal len is same as original length (ac_g_ex.fe_len == ac_o_ex.fe_len).

Hence there is no scope of trimming the goal length to make it
satisfy original request len. Thus this patch avoids using
CR_BEST_AVAIL_LEN criteria for non-regular files request.

Cc: stable@kernel.org
Fixes: 33122aa930f1 ("ext4: Add allocation criteria 1.5 (CR1_5)")
Reported-by: Eric Whitney <enwlinux@gmail.com>
Signed-off-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Tested-by: Eric Whitney <enwlinux@gmail.com>
Link: https://lore.kernel.org/r/2a694c748ff8b8c4b416995a24f06f07b55047a8.1689516047.git.ritesh.list@gmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: remove unused t_handle_lock
Wang Jianjian [Wed, 2 Aug 2023 14:45:34 +0000 (22:45 +0800)]
jbd2: remove unused t_handle_lock

Since commit f7f497cb7024 ("jbd2: kill t_handle_lock
transaction spinlock"), this lock has been no use.

Fixes: f7f497cb7024 ("jbd2: kill t_handle_lock transaction spinlock")
Signed-off-by: Wang Jianjian <wangjianjian0@foxmail.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/tencent_8477CBE568348A1862C64E393D587B342008@qq.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: fix memory leaks in ext4_fname_{setup_filename,prepare_lookup}
Luís Henriques [Thu, 3 Aug 2023 09:17:13 +0000 (10:17 +0100)]
ext4: fix memory leaks in ext4_fname_{setup_filename,prepare_lookup}

If the filename casefolding fails, we'll be leaking memory from the
fscrypt_name struct, namely from the 'crypto_buf.name' member.

Make sure we free it in the error path on both ext4_fname_setup_filename()
and ext4_fname_prepare_lookup() functions.

Cc: stable@kernel.org
Fixes: 1ae98e295fa2 ("ext4: optimize match for casefolded encrypted dirs")
Signed-off-by: Luís Henriques <lhenriques@suse.de>
Reviewed-by: Eric Biggers <ebiggers@google.com>
Link: https://lore.kernel.org/r/20230803091713.13239-1-lhenriques@suse.de
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: remove unused function '__cp_buffer_busy'
Yang Li [Fri, 14 Jul 2023 02:55:28 +0000 (10:55 +0800)]
jbd2: remove unused function '__cp_buffer_busy'

The code calling function '__cp_buffer_busy' has been removed, so the
function should also be removed.
silence the warning:
fs/jbd2/checkpoint.c:48:20: warning: unused function '__cp_buffer_busy'

Reported-by: Abaci Robot <abaci@linux.alibaba.com>
Closes: https://bugzilla.openanolis.cn/show_bug.cgi?id=5518
Signed-off-by: Yang Li <yang.lee@linux.alibaba.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230714025528.564988-4-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: check 'jh->b_transaction' before removing it from checkpoint
Zhihao Cheng [Fri, 14 Jul 2023 02:55:27 +0000 (10:55 +0800)]
jbd2: check 'jh->b_transaction' before removing it from checkpoint

Following process will corrupt ext4 image:
Step 1:
jbd2_journal_commit_transaction
 __jbd2_journal_insert_checkpoint(jh, commit_transaction)
 // Put jh into trans1->t_checkpoint_list
 journal->j_checkpoint_transactions = commit_transaction
 // Put trans1 into journal->j_checkpoint_transactions

Step 2:
do_get_write_access
 test_clear_buffer_dirty(bh) // clear buffer dirty,set jbd dirty
 __jbd2_journal_file_buffer(jh, transaction) // jh belongs to trans2

Step 3:
drop_cache
 journal_shrink_one_cp_list
  jbd2_journal_try_remove_checkpoint
   if (!trylock_buffer(bh))  // lock bh, true
   if (buffer_dirty(bh))     // buffer is not dirty
   __jbd2_journal_remove_checkpoint(jh)
   // remove jh from trans1->t_checkpoint_list

Step 4:
jbd2_log_do_checkpoint
 trans1 = journal->j_checkpoint_transactions
 // jh is not in trans1->t_checkpoint_list
 jbd2_cleanup_journal_tail(journal)  // trans1 is done

Step 5: Power cut, trans2 is not committed, jh is lost in next mounting.

Fix it by checking 'jh->b_transaction' before remove it from checkpoint.

Cc: stable@kernel.org
Fixes: 46f881b5b175 ("jbd2: fix a race when checking checkpoint buffer busy")
Signed-off-by: Zhihao Cheng <chengzhihao1@huawei.com>
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230714025528.564988-3-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agojbd2: fix checkpoint cleanup performance regression
Zhang Yi [Fri, 14 Jul 2023 02:55:26 +0000 (10:55 +0800)]
jbd2: fix checkpoint cleanup performance regression

journal_clean_one_cp_list() has been merged into
journal_shrink_one_cp_list(), but do chekpoint buffer cleanup from the
committing process is just a best effort, it should stop scan once it
meet a busy buffer, or else it will cause a lot of invalid buffer scan
and checks. We catch a performance regression when doing fs_mark tests
below.

Test cmd:
 ./fs_mark  -d  scratch  -s  1024  -n  10000  -t  1  -D  100  -N  100

Before merging checkpoint buffer cleanup:
 FSUse%        Count         Size    Files/sec     App Overhead
     95        10000         1024       8304.9            49033

After merging checkpoint buffer cleanup:
 FSUse%        Count         Size    Files/sec     App Overhead
     95        10000         1024       7649.0            50012
 FSUse%        Count         Size    Files/sec     App Overhead
     95        10000         1024       2107.1            50871

After merging checkpoint buffer cleanup, the total loop count in
journal_shrink_one_cp_list() could be up to 6,261,600+ (50,000+ ~
100,000+ in general), most of them are invalid. This patch fix it
through passing 'shrink_type' into journal_shrink_one_cp_list() and add
a new 'SHRINK_BUSY_STOP' to indicate it should stop once meet a busy
buffer. After fix, the loop count descending back to 10,000+.

After this fix:
 FSUse%        Count         Size    Files/sec     App Overhead
     95        10000         1024       8558.4            49109

Cc: stable@kernel.org
Fixes: b98dba273a0e ("jbd2: remove journal_clean_one_cp_list()")
Signed-off-by: Zhang Yi <yi.zhang@huawei.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230714025528.564988-2-yi.zhang@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: correct some stale comment of criteria
Kemeng Shi [Tue, 1 Aug 2023 14:32:04 +0000 (22:32 +0800)]
ext4: correct some stale comment of criteria

We named criteria with CR_XXX, correct stale comment to criteria with
raw number.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-11-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: return found group directly in ext4_mb_choose_next_group_best_avail
Kemeng Shi [Tue, 1 Aug 2023 14:32:03 +0000 (22:32 +0800)]
ext4: return found group directly in ext4_mb_choose_next_group_best_avail

Return good group when it's found in loop to remove futher check if good
group is found after loop.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-10-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: return found group directly in ext4_mb_choose_next_group_goal_fast
Kemeng Shi [Tue, 1 Aug 2023 14:32:02 +0000 (22:32 +0800)]
ext4: return found group directly in ext4_mb_choose_next_group_goal_fast

Return good group when it's found in loop to remove futher check if good
group is found after loop.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-9-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: remove unused ext4_{set}/{clear}_bit_atomic
Kemeng Shi [Tue, 1 Aug 2023 14:32:01 +0000 (22:32 +0800)]
ext4: remove unused ext4_{set}/{clear}_bit_atomic

Remove ext4_set_bit_atomic and ext4_clear_bit_atomic which are defined but not
used.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-8-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: replace the traditional ternary conditional operator with with max()/min()
Kemeng Shi [Tue, 1 Aug 2023 14:32:00 +0000 (22:32 +0800)]
ext4: replace the traditional ternary conditional operator with with max()/min()

Replace the traditional ternary conditional operator with with max()/min()

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-7-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: remove unnecessary return for void function
Kemeng Shi [Tue, 1 Aug 2023 14:31:59 +0000 (22:31 +0800)]
ext4: remove unnecessary return for void function

The return at end of void function is unnecessary, just remove it.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-6-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: use is_power_of_2 helper in ext4_mb_regular_allocator
Kemeng Shi [Tue, 1 Aug 2023 14:31:58 +0000 (22:31 +0800)]
ext4: use is_power_of_2 helper in ext4_mb_regular_allocator

Use intuitive is_power_of_2 helper in ext4_mb_regular_allocator.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-5-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: return found group directly in ext4_mb_choose_next_group_p2_aligned
Kemeng Shi [Tue, 1 Aug 2023 14:31:57 +0000 (22:31 +0800)]
ext4: return found group directly in ext4_mb_choose_next_group_p2_aligned

Return good group when it's found in loop to remove unnecessary NULL
initialization of grp and futher check if good group is found after loop.

Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-4-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: avoid potential data overflow in next_linear_group
Kemeng Shi [Tue, 1 Aug 2023 14:31:56 +0000 (22:31 +0800)]
ext4: avoid potential data overflow in next_linear_group

ngroups is ext4_group_t (unsigned int) while next_linear_group treat it
in int. If ngroups is bigger than max number described by int, it will
be treat as a negative number. Then "return group + 1 >= ngroups ? 0 :
group + 1;" may keep returning 0.
Switch int to ext4_group_t in next_linear_group to fix the overflow.

Fixes: 196e402adf2e ("ext4: improve cr 0 / cr 1 group scanning")
Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-3-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: correct grp validation in ext4_mb_good_group
Kemeng Shi [Tue, 1 Aug 2023 14:31:55 +0000 (22:31 +0800)]
ext4: correct grp validation in ext4_mb_good_group

Group corruption check will access memory of grp and will trigger kernel
crash if grp is NULL. So do NULL check before corruption check.

Fixes: 5354b2af3406 ("ext4: allow ext4_get_group_info() to fail")
Signed-off-by: Kemeng Shi <shikemeng@huaweicloud.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230801143204.2284343-2-shikemeng@huaweicloud.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: replace CR_FAST macro with inline function for readability
Ojaswin Mujoo [Fri, 30 Jun 2023 08:59:27 +0000 (14:29 +0530)]
ext4: replace CR_FAST macro with inline function for readability

Replace CR_FAST with ext4_mb_cr_expensive() inline function for better
readability. This function returns true if the criteria is one of the
expensive/slower ones where lots of disk IO/prefetching is acceptable.

No functional changes are intended in this patch.

Signed-off-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Reviewed-by: Jan Kara <jack@suse.cz>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230630085927.140137-1-ojaswin@linux.ibm.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: replace read-only check for shutdown check in mmp code
Jan Kara [Fri, 16 Jun 2023 16:50:57 +0000 (18:50 +0200)]
ext4: replace read-only check for shutdown check in mmp code

The multi-mount protection kthread checks for read-only filesystem and
aborts in that case. The remount code actually handles stopping of the
kthread on remount so the only purpose of the check is in case of
emergency remount read-only. Replace the check for read-only filesystem
with a check for shutdown filesystem as running MMP on such is risky
anyway and it makes ordering of things during remount simpler.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-11-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: drop read-only check from ext4_force_commit()
Jan Kara [Fri, 16 Jun 2023 16:50:56 +0000 (18:50 +0200)]
ext4: drop read-only check from ext4_force_commit()

JBD2 code will quickly return without doing anything when there's
nothing to commit so there's no point in the read-only check in
ext4_force_commit(). Just drop it.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-10-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: drop read-only check in ext4_write_inode()
Jan Kara [Fri, 16 Jun 2023 16:50:55 +0000 (18:50 +0200)]
ext4: drop read-only check in ext4_write_inode()

We should not have dirty inodes on read-only filesystem. Also silently
bailing without writing anything would be a problem when we enable
quotas during remount while the filesystem is read-only. So drop the
read-only check.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-9-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: drop read-only check in ext4_init_inode_table()
Jan Kara [Fri, 16 Jun 2023 16:50:54 +0000 (18:50 +0200)]
ext4: drop read-only check in ext4_init_inode_table()

We better should not be initializing inode tables on read-only
filesystem. The following transaction start will warn us and make the
function bail anyway so drop the pointless check.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-8-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: warn on read-only filesystem in ext4_journal_check_start()
Jan Kara [Fri, 16 Jun 2023 16:50:53 +0000 (18:50 +0200)]
ext4: warn on read-only filesystem in ext4_journal_check_start()

Now that filesystem abort marks the filesystem as shutdown, we shouldn't
be ever hitting the sb_rdonly() check in ext4_journal_check_start().
Since this is a suitable place for catching all sorts of programming
errors, convert the check to WARN_ON instead of dropping it.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-7-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: avoid starting transaction on read-only fs in ext4_quota_off()
Jan Kara [Fri, 16 Jun 2023 16:50:52 +0000 (18:50 +0200)]
ext4: avoid starting transaction on read-only fs in ext4_quota_off()

When the filesystem gets first remounted read-only and then unmounted,
ext4_quota_off() will try to start a transaction (and fail) on read-only
filesystem to cleanup inode flags for legacy quota files. Just bail
before trying to start a transaction instead since that is going to
issue a warning.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-6-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: drop EXT4_MF_FS_ABORTED flag
Jan Kara [Fri, 16 Jun 2023 16:50:51 +0000 (18:50 +0200)]
ext4: drop EXT4_MF_FS_ABORTED flag

EXT4_MF_FS_ABORTED flag has practically the same intent as
EXT4_FLAGS_SHUTDOWN flag. The shutdown flag is checked in many more
places than the aborted flag which is mostly the historical artifact
where we were relying on SB_RDONLY checks instead of the aborted flag
checks. There are only three places - ext4_sync_file(),
__ext4_remount(), and mballoc debug code - which check aborted flag and
not shutdown flag and this is arguably a bug. Avoid these
inconsistencies by removing EXT4_MF_FS_ABORTED flag and using
EXT4_FLAGS_SHUTDOWN everywhere.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-5-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: make 'abort' mount option handling standard
Jan Kara [Fri, 16 Jun 2023 16:50:50 +0000 (18:50 +0200)]
ext4: make 'abort' mount option handling standard

'abort' mount option is the only mount option that has special handling
and sets a bit in sbi->s_mount_flags. There is not strong reason for
that so just simplify the code and make 'abort' set a bit in
sbi->s_mount_opt2 as any other mount option. This simplifies the code
and will allow us to drop EXT4_MF_FS_ABORTED completely in the following
patch.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-4-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: make ext4_forced_shutdown() take struct super_block
Jan Kara [Fri, 16 Jun 2023 16:50:49 +0000 (18:50 +0200)]
ext4: make ext4_forced_shutdown() take struct super_block

Currently ext4_forced_shutdown() takes struct ext4_sb_info but most
callers need to get it from struct super_block anyway. So just pass in
struct super_block to save all callers from some boilerplate code. No
functional changes.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-3-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: use sb_rdonly() helper for checking read-only flag
Jan Kara [Fri, 16 Jun 2023 16:50:48 +0000 (18:50 +0200)]
ext4: use sb_rdonly() helper for checking read-only flag

sb_rdonly() helper instead of directly checking sb->s_flags.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-2-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: remove pointless sb_rdonly() checks from freezing code
Jan Kara [Fri, 16 Jun 2023 16:50:47 +0000 (18:50 +0200)]
ext4: remove pointless sb_rdonly() checks from freezing code

ext4_freeze() and ext4_unfreeze() checks for sb_rdonly(). However this
check is pointless as VFS already checks for read-only filesystem before
calling filesystem specific methods. Remove the pointless checks.

Signed-off-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230616165109.21695-1-jack@suse.cz
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agofs: jbd2: fix an incorrect warn log
Guoqing Cai [Thu, 13 Apr 2023 09:57:39 +0000 (17:57 +0800)]
fs: jbd2: fix an incorrect warn log

In jbd2_journal_load(), when journal_reset fails, it prints an incorrect
warn log.

Fix this by changing the goto statement to return statement.

Also, return actual error code from jbd2_journal_recover() and journal_reset().

Signed-off-by: Guoqing Cai <u202112087@hust.edu.cn>
Reviewed-by: Jan Kara <jack@suse.cz>
Link: https://lore.kernel.org/r/20230413095740.2222066-1-u202112087@hust.edu.cn
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: avoid overlapping preallocations due to overflow
Baokun Li [Mon, 24 Jul 2023 12:10:59 +0000 (20:10 +0800)]
ext4: avoid overlapping preallocations due to overflow

Let's say we want to allocate 2 blocks starting from 4294966386, after
predicting the file size, start is aligned to 4294965248, len is changed
to 2048, then end = start + size = 0x100000000. Since end is of
type ext4_lblk_t, i.e. uint, end is truncated to 0.

This causes (pa->pa_lstart >= end) to always hold when checking if the
current extent to be allocated crosses already preallocated blocks, so the
resulting ac_g_ex may cross already preallocated blocks. Hence we convert
the end type to loff_t and use pa_logical_end() to avoid overflow.

Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230724121059.11834-4-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: fix BUG in ext4_mb_new_inode_pa() due to overflow
Baokun Li [Mon, 24 Jul 2023 12:10:58 +0000 (20:10 +0800)]
ext4: fix BUG in ext4_mb_new_inode_pa() due to overflow

When we calculate the end position of ext4_free_extent, this position may
be exactly where ext4_lblk_t (i.e. uint) overflows. For example, if
ac_g_ex.fe_logical is 4294965248 and ac_orig_goal_len is 2048, then the
computed end is 0x100000000, which is 0. If ac->ac_o_ex.fe_logical is not
the first case of adjusting the best extent, that is, new_bex_end > 0, the
following BUG_ON will be triggered:

=========================================================
kernel BUG at fs/ext4/mballoc.c:5116!
invalid opcode: 0000 [#1] PREEMPT SMP PTI
CPU: 3 PID: 673 Comm: xfs_io Tainted: G E 6.5.0-rc1+ #279
RIP: 0010:ext4_mb_new_inode_pa+0xc5/0x430
Call Trace:
 <TASK>
 ext4_mb_use_best_found+0x203/0x2f0
 ext4_mb_try_best_found+0x163/0x240
 ext4_mb_regular_allocator+0x158/0x1550
 ext4_mb_new_blocks+0x86a/0xe10
 ext4_ext_map_blocks+0xb0c/0x13a0
 ext4_map_blocks+0x2cd/0x8f0
 ext4_iomap_begin+0x27b/0x400
 iomap_iter+0x222/0x3d0
 __iomap_dio_rw+0x243/0xcb0
 iomap_dio_rw+0x16/0x80
=========================================================

A simple reproducer demonstrating the problem:

mkfs.ext4 -F /dev/sda -b 4096 100M
mount /dev/sda /tmp/test
fallocate -l1M /tmp/test/tmp
fallocate -l10M /tmp/test/file
fallocate -i -o 1M -l16777203M /tmp/test/file
fsstress -d /tmp/test -l 0 -n 100000 -p 8 &
sleep 10 && killall -9 fsstress
rm -f /tmp/test/tmp
xfs_io -c "open -ad /tmp/test/file" -c "pwrite -S 0xff 0 8192"

We simply refactor the logic for adjusting the best extent by adding
a temporary ext4_free_extent ex and use extent_logical_end() to avoid
overflow, which also simplifies the code.

Cc: stable@kernel.org # 6.4
Fixes: 93cdf49f6eca ("ext4: Fix best extent lstart adjustment logic in ext4_mb_new_inode_pa()")
Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230724121059.11834-3-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
16 months agoext4: add two helper functions extent_logical_end() and pa_logical_end()
Baokun Li [Mon, 24 Jul 2023 12:10:57 +0000 (20:10 +0800)]
ext4: add two helper functions extent_logical_end() and pa_logical_end()

When we use lstart + len to calculate the end of free extent or prealloc
space, it may exceed the maximum value of 4294967295(0xffffffff) supported
by ext4_lblk_t and cause overflow, which may lead to various problems.

Therefore, we add two helper functions, extent_logical_end() and
pa_logical_end(), to limit the type of end to loff_t, and also convert
lstart to loff_t for calculation to avoid overflow.

Signed-off-by: Baokun Li <libaokun1@huawei.com>
Reviewed-by: Ritesh Harjani (IBM) <ritesh.list@gmail.com>
Link: https://lore.kernel.org/r/20230724121059.11834-2-libaokun1@huawei.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
17 months agoLinux 6.5-rc3
Linus Torvalds [Sun, 23 Jul 2023 22:24:10 +0000 (15:24 -0700)]
Linux 6.5-rc3

17 months agoMerge tag 'trace-v6.5-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace...
Linus Torvalds [Sun, 23 Jul 2023 22:19:14 +0000 (15:19 -0700)]
Merge tag 'trace-v6.5-rc2' of git://git./linux/kernel/git/trace/linux-trace

Pull tracing fixes from Steven Rostedt:

 - Swapping the ring buffer for snapshotting (for things like irqsoff)
   can crash if the ring buffer is being resized. Disable swapping when
   this happens. The missed swap will be reported to the tracer

 - Report error if the histogram fails to be created due to an error in
   adding a histogram variable, in event_hist_trigger_parse()

 - Remove unused declaration of tracing_map_set_field_descr()

* tag 'trace-v6.5-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
  tracing/histograms: Return an error if we fail to add histogram to hist_vars list
  ring-buffer: Do not swap cpu_buffer during resize process
  tracing: Remove unused extern declaration tracing_map_set_field_descr()

17 months agoMerge tag 'kbuild-fixes-v6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/masahi...
Linus Torvalds [Sun, 23 Jul 2023 21:55:41 +0000 (14:55 -0700)]
Merge tag 'kbuild-fixes-v6.5' of git://git./linux/kernel/git/masahiroy/linux-kbuild

Pull Kbuild fixes from Masahiro Yamada:

 - Fix stale help text in gconfig

 - Support *.S files in compile_commands.json

 - Flatten KBUILD_CFLAGS

 - Fix external module builds with Rust so that temporary files are
   created in the modules directories instead of the kernel tree

* tag 'kbuild-fixes-v6.5' of git://git.kernel.org/pub/scm/linux/kernel/git/masahiroy/linux-kbuild:
  kbuild: rust: avoid creating temporary files
  kbuild: flatten KBUILD_CFLAGS
  gen_compile_commands: add assembly files to compilation database
  kconfig: gconfig: correct program name in help text
  kconfig: gconfig: drop the Show Debug Info help text

17 months agokbuild: rust: avoid creating temporary files
Miguel Ojeda [Sun, 23 Jul 2023 14:21:28 +0000 (16:21 +0200)]
kbuild: rust: avoid creating temporary files

`rustc` outputs by default the temporary files (i.e. the ones saved
by `-Csave-temps`, such as `*.rcgu*` files) in the current working
directory when `-o` and `--out-dir` are not given (even if
`--emit=x=path` is given, i.e. it does not use those for temporaries).

Since out-of-tree modules are compiled from the `linux` tree,
`rustc` then tries to create them there, which may not be accessible.

Thus pass `--out-dir` explicitly, even if it is just for the temporary
files.

Similarly, do so for Rust host programs too.

Reported-by: Raphael Nestler <raphael.nestler@gmail.com>
Closes: https://github.com/Rust-for-Linux/linux/issues/1015
Reported-by: Andrea Righi <andrea.righi@canonical.com>
Tested-by: Raphael Nestler <raphael.nestler@gmail.com> # non-hostprogs
Tested-by: Andrea Righi <andrea.righi@canonical.com> # non-hostprogs
Fixes: 295d8398c67e ("kbuild: specify output names separately for each emission type from rustc")
Cc: stable@vger.kernel.org
Signed-off-by: Miguel Ojeda <ojeda@kernel.org>
Tested-by: Martin Rodriguez Reboredo <yakoyoku@gmail.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
17 months agoMerge tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm
Linus Torvalds [Sun, 23 Jul 2023 17:44:38 +0000 (10:44 -0700)]
Merge tag 'for-linus' of git://git./virt/kvm/kvm

Pull kvm fixes from Paolo Bonzini:
 "ARM:

   - Avoid pKVM finalization if KVM initialization fails

   - Add missing BTI instructions in the hypervisor, fixing an early
     boot failure on BTI systems

   - Handle MMU notifiers correctly for non hugepage-aligned memslots

   - Work around a bug in the architecture where hypervisor timer
     controls have UNKNOWN behavior under nested virt

   - Disable preemption in kvm_arch_hardware_enable(), fixing a kernel
     BUG in cpu hotplug resulting from per-CPU accessor sanity checking

   - Make WFI emulation on GICv4 systems robust w.r.t. preemption,
     consistently requesting a doorbell interrupt on vcpu_put()

   - Uphold RES0 sysreg behavior when emulating older PMU versions

   - Avoid macro expansion when initializing PMU register names,
     ensuring the tracepoints pretty-print the sysreg

  s390:

   - Two fixes for asynchronous destroy

  x86 fixes will come early next week"

* tag 'for-linus' of git://git.kernel.org/pub/scm/virt/kvm/kvm:
  KVM: s390: pv: fix index value of replaced ASCE
  KVM: s390: pv: simplify shutdown and fix race
  KVM: arm64: Fix the name of sys_reg_desc related to PMU
  KVM: arm64: Correctly handle RES0 bits PMEVTYPER<n>_EL0.evtCount
  KVM: arm64: vgic-v4: Make the doorbell request robust w.r.t preemption
  KVM: arm64: Add missing BTI instructions
  KVM: arm64: Correctly handle page aging notifiers for unaligned memslot
  KVM: arm64: Disable preemption in kvm_arch_hardware_enable()
  KVM: arm64: Handle kvm_arm_init failure correctly in finalize_pkvm
  KVM: arm64: timers: Use CNTHCTL_EL2 when setting non-CNTKCTL_EL1 bits

17 months agoMerge tag 'ext4_for_linus-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sun, 23 Jul 2023 17:21:49 +0000 (10:21 -0700)]
Merge tag 'ext4_for_linus-6.5-rc3' of git://git./linux/kernel/git/tytso/ext4

Pull ext4 fixes from Ted Ts'o:
 "Bug and regression fixes for 6.5-rc3 for ext4's mballoc and jbd2's
  checkpoint code"

* tag 'ext4_for_linus-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4:
  ext4: fix rbtree traversal bug in ext4_mb_use_preallocated
  ext4: fix off by one issue in ext4_mb_choose_next_group_best_avail()
  ext4: correct inline offset when handling xattrs in inode body
  jbd2: remove __journal_try_to_free_buffer()
  jbd2: fix a race when checking checkpoint buffer busy
  jbd2: Fix wrongly judgement for buffer head removing while doing checkpoint
  jbd2: remove journal_clean_one_cp_list()
  jbd2: remove t_checkpoint_io_list
  jbd2: recheck chechpointing non-dirty buffer

17 months agoMerge tag '6.5-rc2-smb3-client-fixes-ver2' of git://git.samba.org/sfrench/cifs-2.6
Linus Torvalds [Sun, 23 Jul 2023 17:16:44 +0000 (10:16 -0700)]
Merge tag '6.5-rc2-smb3-client-fixes-ver2' of git://git.samba.org/sfrench/cifs-2.6

Pull smb client fix from Steve French:
 "Add minor debugging improvement.

  The change improves ability to read a network trace to debug problems
  on encrypted connections which are very common (e.g. using wireshark
  or tcpdump).

  That works today with tools like 'smbinfo keys /mnt/file' but requires
  passing in a filename on the mount (see e.g. [1]), but it often makes
  more sense to just pass in the mount point path (ie a directory not a
  filename).

  So this fix was needed to debug some types of problems (an obvious
  example is on an encrypted connection failing operations on an empty
  share or with no files in the root of the directory) - so you can
  simply pass in the 'smbinfo keys <mntpoint>' and get the information
  that wireshark needs"

Link: https://wiki.samba.org/index.php/Wireshark_Decryption
* tag '6.5-rc2-smb3-client-fixes-ver2' of git://git.samba.org/sfrench/cifs-2.6:
  cifs: update internal module version number for cifs.ko
  cifs: allow dumping keys for directories too

17 months agoMerge tag 'kvm-s390-master-6.5-1' of https://git.kernel.org/pub/scm/linux/kernel...
Paolo Bonzini [Sun, 23 Jul 2023 16:50:30 +0000 (12:50 -0400)]
Merge tag 'kvm-s390-master-6.5-1' of https://git./linux/kernel/git/kvms390/linux into HEAD

Two fixes for asynchronous destroy

17 months agoMerge tag 'kvmarm-fixes-6.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/kvmar...
Paolo Bonzini [Sun, 23 Jul 2023 16:50:14 +0000 (12:50 -0400)]
Merge tag 'kvmarm-fixes-6.5-1' of git://git./linux/kernel/git/kvmarm/kvmarm into HEAD

KVM/arm64 fixes for 6.5, part #1

 - Avoid pKVM finalization if KVM initialization fails

 - Add missing BTI instructions in the hypervisor, fixing an early boot
   failure on BTI systems

 - Handle MMU notifiers correctly for non hugepage-aligned memslots

 - Work around a bug in the architecture where hypervisor timer controls
   have UNKNOWN behavior under nested virt.

 - Disable preemption in kvm_arch_hardware_enable(), fixing a kernel BUG
   in cpu hotplug resulting from per-CPU accessor sanity checking.

 - Make WFI emulation on GICv4 systems robust w.r.t. preemption,
   consistently requesting a doorbell interrupt on vcpu_put()

 - Uphold RES0 sysreg behavior when emulating older PMU versions

 - Avoid macro expansion when initializing PMU register names, ensuring
   the tracepoints pretty-print the sysreg.

17 months agotracing/histograms: Return an error if we fail to add histogram to hist_vars list
Mohamed Khalfella [Fri, 14 Jul 2023 20:33:41 +0000 (20:33 +0000)]
tracing/histograms: Return an error if we fail to add histogram to hist_vars list

Commit 6018b585e8c6 ("tracing/histograms: Add histograms to hist_vars if
they have referenced variables") added a check to fail histogram creation
if save_hist_vars() failed to add histogram to hist_vars list. But the
commit failed to set ret to failed return code before jumping to
unregister histogram, fix it.

Link: https://lore.kernel.org/linux-trace-kernel/20230714203341.51396-1-mkhalfella@purestorage.com
Cc: stable@vger.kernel.org
Fixes: 6018b585e8c6 ("tracing/histograms: Add histograms to hist_vars if they have referenced variables")
Signed-off-by: Mohamed Khalfella <mkhalfella@purestorage.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
17 months agoring-buffer: Do not swap cpu_buffer during resize process
Chen Lin [Wed, 19 Jul 2023 07:58:47 +0000 (15:58 +0800)]
ring-buffer: Do not swap cpu_buffer during resize process

When ring_buffer_swap_cpu was called during resize process,
the cpu buffer was swapped in the middle, resulting in incorrect state.
Continuing to run in the wrong state will result in oops.

This issue can be easily reproduced using the following two scripts:
/tmp # cat test1.sh
//#! /bin/sh
for i in `seq 0 100000`
do
         echo 2000 > /sys/kernel/debug/tracing/buffer_size_kb
         sleep 0.5
         echo 5000 > /sys/kernel/debug/tracing/buffer_size_kb
         sleep 0.5
done
/tmp # cat test2.sh
//#! /bin/sh
for i in `seq 0 100000`
do
        echo irqsoff > /sys/kernel/debug/tracing/current_tracer
        sleep 1
        echo nop > /sys/kernel/debug/tracing/current_tracer
        sleep 1
done
/tmp # ./test1.sh &
/tmp # ./test2.sh &

A typical oops log is as follows, sometimes with other different oops logs.

[  231.711293] WARNING: CPU: 0 PID: 9 at kernel/trace/ring_buffer.c:2026 rb_update_pages+0x378/0x3f8
[  231.713375] Modules linked in:
[  231.714735] CPU: 0 PID: 9 Comm: kworker/0:1 Tainted: G        W          6.5.0-rc1-00276-g20edcec23f92 #15
[  231.716750] Hardware name: linux,dummy-virt (DT)
[  231.718152] Workqueue: events update_pages_handler
[  231.719714] pstate: 60000005 (nZCv daif -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[  231.721171] pc : rb_update_pages+0x378/0x3f8
[  231.722212] lr : rb_update_pages+0x25c/0x3f8
[  231.723248] sp : ffff800082b9bd50
[  231.724169] x29: ffff800082b9bd50 x28: ffff8000825f7000 x27: 0000000000000000
[  231.726102] x26: 0000000000000001 x25: fffffffffffff010 x24: 0000000000000ff0
[  231.728122] x23: ffff0000c3a0b600 x22: ffff0000c3a0b5c0 x21: fffffffffffffe0a
[  231.730203] x20: ffff0000c3a0b600 x19: ffff0000c0102400 x18: 0000000000000000
[  231.732329] x17: 0000000000000000 x16: 0000000000000000 x15: 0000ffffe7aa8510
[  231.734212] x14: 0000000000000000 x13: 0000000000000000 x12: 0000000000000002
[  231.736291] x11: ffff8000826998a8 x10: ffff800082b9baf0 x9 : ffff800081137558
[  231.738195] x8 : fffffc00030e82c8 x7 : 0000000000000000 x6 : 0000000000000001
[  231.740192] x5 : ffff0000ffbafe00 x4 : 0000000000000000 x3 : 0000000000000000
[  231.742118] x2 : 00000000000006aa x1 : 0000000000000001 x0 : ffff0000c0007208
[  231.744196] Call trace:
[  231.744892]  rb_update_pages+0x378/0x3f8
[  231.745893]  update_pages_handler+0x1c/0x38
[  231.746893]  process_one_work+0x1f0/0x468
[  231.747852]  worker_thread+0x54/0x410
[  231.748737]  kthread+0x124/0x138
[  231.749549]  ret_from_fork+0x10/0x20
[  231.750434] ---[ end trace 0000000000000000 ]---
[  233.720486] Unable to handle kernel NULL pointer dereference at virtual address 0000000000000000
[  233.721696] Mem abort info:
[  233.721935]   ESR = 0x0000000096000004
[  233.722283]   EC = 0x25: DABT (current EL), IL = 32 bits
[  233.722596]   SET = 0, FnV = 0
[  233.722805]   EA = 0, S1PTW = 0
[  233.723026]   FSC = 0x04: level 0 translation fault
[  233.723458] Data abort info:
[  233.723734]   ISV = 0, ISS = 0x00000004, ISS2 = 0x00000000
[  233.724176]   CM = 0, WnR = 0, TnD = 0, TagAccess = 0
[  233.724589]   GCS = 0, Overlay = 0, DirtyBit = 0, Xs = 0
[  233.725075] user pgtable: 4k pages, 48-bit VAs, pgdp=0000000104943000
[  233.725592] [0000000000000000] pgd=0000000000000000, p4d=0000000000000000
[  233.726231] Internal error: Oops: 0000000096000004 [#1] PREEMPT SMP
[  233.726720] Modules linked in:
[  233.727007] CPU: 0 PID: 9 Comm: kworker/0:1 Tainted: G        W          6.5.0-rc1-00276-g20edcec23f92 #15
[  233.727777] Hardware name: linux,dummy-virt (DT)
[  233.728225] Workqueue: events update_pages_handler
[  233.728655] pstate: 200000c5 (nzCv daIF -PAN -UAO -TCO -DIT -SSBS BTYPE=--)
[  233.729054] pc : rb_update_pages+0x1a8/0x3f8
[  233.729334] lr : rb_update_pages+0x154/0x3f8
[  233.729592] sp : ffff800082b9bd50
[  233.729792] x29: ffff800082b9bd50 x28: ffff8000825f7000 x27: 0000000000000000
[  233.730220] x26: 0000000000000000 x25: ffff800082a8b840 x24: ffff0000c0102418
[  233.730653] x23: 0000000000000000 x22: fffffc000304c880 x21: 0000000000000003
[  233.731105] x20: 00000000000001f4 x19: ffff0000c0102400 x18: ffff800082fcbc58
[  233.731727] x17: 0000000000000000 x16: 0000000000000001 x15: 0000000000000001
[  233.732282] x14: ffff8000825fe0c8 x13: 0000000000000001 x12: 0000000000000000
[  233.732709] x11: ffff8000826998a8 x10: 0000000000000ae0 x9 : ffff8000801b760c
[  233.733148] x8 : fefefefefefefeff x7 : 0000000000000018 x6 : ffff0000c03298c0
[  233.733553] x5 : 0000000000000002 x4 : 0000000000000000 x3 : 0000000000000000
[  233.733972] x2 : ffff0000c3a0b600 x1 : 0000000000000000 x0 : 0000000000000000
[  233.734418] Call trace:
[  233.734593]  rb_update_pages+0x1a8/0x3f8
[  233.734853]  update_pages_handler+0x1c/0x38
[  233.735148]  process_one_work+0x1f0/0x468
[  233.735525]  worker_thread+0x54/0x410
[  233.735852]  kthread+0x124/0x138
[  233.736064]  ret_from_fork+0x10/0x20
[  233.736387] Code: 92400000 910006b5 aa000021 aa0303f7 (f9400060)
[  233.736959] ---[ end trace 0000000000000000 ]---

After analysis, the seq of the error is as follows [1-5]:

int ring_buffer_resize(struct trace_buffer *buffer, unsigned long size,
int cpu_id)
{
for_each_buffer_cpu(buffer, cpu) {
cpu_buffer = buffer->buffers[cpu];
//1. get cpu_buffer, aka cpu_buffer(A)
...
...
schedule_work_on(cpu,
 &cpu_buffer->update_pages_work);
//2. 'update_pages_work' is queue on 'cpu', cpu_buffer(A) is passed to
// update_pages_handler, do the update process, set 'update_done' in
// complete(&cpu_buffer->update_done) and to wakeup resize process.
//---->
//3. Just at this moment, ring_buffer_swap_cpu is triggered,
//cpu_buffer(A) be swaped to cpu_buffer(B), the max_buffer.
//ring_buffer_swap_cpu is called as the 'Call trace' below.

Call trace:
 dump_backtrace+0x0/0x2f8
 show_stack+0x18/0x28
 dump_stack+0x12c/0x188
 ring_buffer_swap_cpu+0x2f8/0x328
 update_max_tr_single+0x180/0x210
 check_critical_timing+0x2b4/0x2c8
 tracer_hardirqs_on+0x1c0/0x200
 trace_hardirqs_on+0xec/0x378
 el0_svc_common+0x64/0x260
 do_el0_svc+0x90/0xf8
 el0_svc+0x20/0x30
 el0_sync_handler+0xb0/0xb8
 el0_sync+0x180/0x1c0
//<----

/* wait for all the updates to complete */
for_each_buffer_cpu(buffer, cpu) {
cpu_buffer = buffer->buffers[cpu];
//4. get cpu_buffer, cpu_buffer(B) is used in the following process,
//the state of cpu_buffer(A) and cpu_buffer(B) is totally wrong.
//for example, cpu_buffer(A)->update_done will leave be set 1, and will
//not 'wait_for_completion' at the next resize round.
  if (!cpu_buffer->nr_pages_to_update)
continue;

if (cpu_online(cpu))
wait_for_completion(&cpu_buffer->update_done);
cpu_buffer->nr_pages_to_update = 0;
}
...
}
//5. the state of cpu_buffer(A) and cpu_buffer(B) is totally wrong,
//Continuing to run in the wrong state, then oops occurs.

Link: https://lore.kernel.org/linux-trace-kernel/202307191558478409990@zte.com.cn
Signed-off-by: Chen Lin <chen.lin5@zte.com.cn>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
17 months agotracing: Remove unused extern declaration tracing_map_set_field_descr()
YueHaibing [Sat, 22 Jul 2023 03:21:23 +0000 (11:21 +0800)]
tracing: Remove unused extern declaration tracing_map_set_field_descr()

Since commit 08d43a5fa063 ("tracing: Add lock-free tracing_map"),
this is never used, so can be removed.

Link: https://lore.kernel.org/linux-trace-kernel/20230722032123.24664-1-yuehaibing@huawei.com
Cc: <mhiramat@kernel.org>
Signed-off-by: YueHaibing <yuehaibing@huawei.com>
Signed-off-by: Steven Rostedt (Google) <rostedt@goodmis.org>
17 months agokbuild: flatten KBUILD_CFLAGS
Alexey Dobriyan [Thu, 13 Jul 2023 18:52:28 +0000 (21:52 +0300)]
kbuild: flatten KBUILD_CFLAGS

Make it slightly easier to see which compiler options are added and
removed (and not worry about column limit too!).

Signed-off-by: Alexey Dobriyan <adobriyan@gmail.com>
Reviewed-by: Nicolas Schier <n.schier@avm.de>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
17 months agogen_compile_commands: add assembly files to compilation database
Benjamin Gray [Wed, 19 Jul 2023 03:19:12 +0000 (13:19 +1000)]
gen_compile_commands: add assembly files to compilation database

Like C source files, tooling can find it useful to have the assembly
source file compilation recorded.

The .S extension appears to used across all architectures.

Signed-off-by: Benjamin Gray <bgray@linux.ibm.com>
Reviewed-by: Fangrui Song <maskray@google.com>
Reviewed-by: Nathan Chancellor <nathan@kernel.org>
Reviewed-by: Nick Desaulniers <ndesaulniers@google.com>
Signed-off-by: Masahiro Yamada <masahiroy@kernel.org>
17 months agoext4: fix rbtree traversal bug in ext4_mb_use_preallocated
Ojaswin Mujoo [Sat, 22 Jul 2023 17:15:24 +0000 (22:45 +0530)]
ext4: fix rbtree traversal bug in ext4_mb_use_preallocated

During allocations, while looking for preallocations(PA) in the per
inode rbtree, we can't do a direct traversal of the tree because
ext4_mb_discard_group_preallocation() can paralelly mark the pa deleted
and that can cause direct traversal to skip some entries. This was
leading to a BUG_ON() being hit [1] when we missed a PA that could satisfy
our request and ultimately tried to create a new PA that would overlap
with the missed one.

To makes sure we handle that case while still keeping the performance of
the rbtree, we make use of the fact that the only pa that could possibly
overlap the original goal start is the one that satisfies the below
conditions:

  1. It must have it's logical start immediately to the left of
  (ie less than) original logical start.

  2. It must not be deleted

To find this pa we use the following traversal method:

1. Descend into the rbtree normally to find the immediate neighboring
PA. Here we keep descending irrespective of if the PA is deleted or if
it overlaps with our request etc. The goal is to find an immediately
adjacent PA.

2. If the found PA is on right of original goal, use rb_prev() to find
the left adjacent PA.

3. Check if this PA is deleted and keep moving left with rb_prev() until
a non deleted PA is found.

4. This is the PA we are looking for. Now we can check if it can satisfy
the original request and proceed accordingly.

This approach also takes care of having deleted PAs in the tree.

(While we are at it, also fix a possible overflow bug in calculating the
end of a PA)

[1] https://lore.kernel.org/linux-ext4/CA+G9fYv2FRpLqBZf34ZinR8bU2_ZRAUOjKAD3+tKRFaEQHtt8Q@mail.gmail.com/

Cc: stable@kernel.org # 6.4
Fixes: 3872778664e3 ("ext4: Use rbtrees to manage PAs instead of inode i_prealloc_list")
Signed-off-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Reported-by: Naresh Kamboju <naresh.kamboju@linaro.org>
Reviewed-by: Ritesh Harjani (IBM) ritesh.list@gmail.com
Tested-by: Ritesh Harjani (IBM) ritesh.list@gmail.com
Link: https://lore.kernel.org/r/edd2efda6a83e6343c5ace9deea44813e71dbe20.1690045963.git.ojaswin@linux.ibm.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
17 months agoext4: fix off by one issue in ext4_mb_choose_next_group_best_avail()
Ojaswin Mujoo [Fri, 9 Jun 2023 10:34:03 +0000 (16:04 +0530)]
ext4: fix off by one issue in ext4_mb_choose_next_group_best_avail()

In ext4_mb_choose_next_group_best_avail(), we want the start order to be
1 less than goal length and the min_order to be, at max, 1 more than the
original length. This commit fixes an off by one issue that arose due to
the fact that 1 << fls(n) > (n).

After all the processing:

order = 1 order below goal len
min_order = maximum of the three:-
             - order - trim_order
             - 1 order below B2C(s_stripe)
             - 1 order above original len

Cc: stable@kernel.org
Fixes: 33122aa930 ("ext4: Add allocation criteria 1.5 (CR1_5)")
Signed-off-by: Ojaswin Mujoo <ojaswin@linux.ibm.com>
Link: https://lore.kernel.org/r/20230609103403.112807-1-ojaswin@linux.ibm.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
17 months agoext4: correct inline offset when handling xattrs in inode body
Eric Whitney [Mon, 22 May 2023 18:15:20 +0000 (14:15 -0400)]
ext4: correct inline offset when handling xattrs in inode body

When run on a file system where the inline_data feature has been
enabled, xfstests generic/269, generic/270, and generic/476 cause ext4
to emit error messages indicating that inline directory entries are
corrupted.  This occurs because the inline offset used to locate
inline directory entries in the inode body is not updated when an
xattr in that shared region is deleted and the region is shifted in
memory to recover the space it occupied.  If the deleted xattr precedes
the system.data attribute, which points to the inline directory entries,
that attribute will be moved further up in the region.  The inline
offset continues to point to whatever is located in system.data's former
location, with unfortunate effects when used to access directory entries
or (presumably) inline data in the inode body.

Cc: stable@kernel.org
Signed-off-by: Eric Whitney <enwlinux@gmail.com>
Link: https://lore.kernel.org/r/20230522181520.1570360-1-enwlinux@gmail.com
Signed-off-by: Theodore Ts'o <tytso@mit.edu>
17 months agoMerge tag 'powerpc-6.5-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc...
Linus Torvalds [Sun, 23 Jul 2023 02:32:00 +0000 (19:32 -0700)]
Merge tag 'powerpc-6.5-4' of git://git./linux/kernel/git/powerpc/linux

Pull powerpc fixes from Michael Ellerman:

 - Reinstate support for little endian ELFv1 binaries, which it turns
   out still exist in the wild.

 - Revert a change which used asm goto for WARN_ON/__WARN_FLAGS, as it
   lead to dead code generation and seemed to trigger compiler bugs in
   some edge cases.

 - Fix a deadlock in the pseries VAS code, between live migration and
   the driver's mmap handler.

 - Disable KCOV instrumentation in the powerpc KASAN code.

Thanks to Andrew Donnellan, Benjamin Gray, Christophe Leroy, Haren
Myneni, Russell Currey, and Uwe Kleine-König.

* tag 'powerpc-6.5-4' of git://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux:
  Revert "powerpc/64s: Remove support for ELFv1 little endian userspace"
  powerpc/kasan: Disable KCOV in KASAN code
  powerpc/512x: lpbfifo: Convert to platform remove callback returning void
  powerpc/crypto: Add gitignore for generated P10 AES/GCM .S files
  Revert "powerpc/bug: Provide better flexibility to WARN_ON/__WARN_FLAGS() with asm goto"
  powerpc/pseries/vas: Hold mmap_mutex after mmap lock during window close

17 months agocifs: update internal module version number for cifs.ko
Steve French [Thu, 20 Jul 2023 13:30:32 +0000 (08:30 -0500)]
cifs: update internal module version number for cifs.ko

From 2.43 to 2.44

Signed-off-by: Steve French <stfrench@microsoft.com>
17 months agocifs: allow dumping keys for directories too
Shyam Prasad N [Fri, 16 Jun 2023 10:37:46 +0000 (10:37 +0000)]
cifs: allow dumping keys for directories too

Dumping the enc/dec keys is a session wide operation.
And it should not matter if the ioctl was run on
a regular file or a directory.

Currently, we obtain the tcon pointer from the
cifs file handle. But since there's no dir open call
in cifs, this is not populated for dirs.

This change allows dumping of session keys using ioctl
even for directories. To do this, we'll now get the
tcon pointer from the superblock, and not from the file
handle.

Signed-off-by: Shyam Prasad N <sprasad@microsoft.com>
Signed-off-by: Steve French <stfrench@microsoft.com>
17 months agoMerge tag 's390-6.5-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux
Linus Torvalds [Sat, 22 Jul 2023 18:24:03 +0000 (11:24 -0700)]
Merge tag 's390-6.5-3' of git://git./linux/kernel/git/s390/linux

Pull s390 fixes from Heiko Carstens:

 - Fix per vma lock fault handling: add missing !(fault & VM_FAULT_ERROR)
   check to fault handler to prevent error handling for return values
   that don't indicate an error

 - Use kfree_sensitive() instead of kfree() in paes crypto code to clear
   memory that may contain keys before freeing it

 - Fix reply buffer size calculation for CCA replies in zcrypt device
   driver

* tag 's390-6.5-3' of git://git.kernel.org/pub/scm/linux/kernel/git/s390/linux:
  s390/zcrypt: fix reply buffer calculations for CCA replies
  s390/crypto: use kfree_sensitive() instead of kfree()
  s390/mm: fix per vma lock fault handling

17 months agoMerge tag 'block-6.5-2023-07-21' of git://git.kernel.dk/linux
Linus Torvalds [Sat, 22 Jul 2023 18:05:15 +0000 (11:05 -0700)]
Merge tag 'block-6.5-2023-07-21' of git://git.kernel.dk/linux

Pull block fixes from Jens Axboe:

 - Fix for loop regressions (Mauricio)

 - Fix a potential stall with batched wakeups in sbitmap (David)

 - Fix for stall with recursive plug flushes (Ross)

 - Skip accounting of empty requests for blk-iocost (Chengming)

 - Remove a dead field in struct blk_mq_hw_ctx (Chengming)

* tag 'block-6.5-2023-07-21' of git://git.kernel.dk/linux:
  loop: do not enforce max_loop hard limit by (new) default
  loop: deprecate autoloading callback loop_probe()
  sbitmap: fix batching wakeup
  blk-iocost: skip empty flush bio in iocost
  blk-mq: delete dead struct blk_mq_hw_ctx->queued field
  blk-mq: Fix stall due to recursive flush plug

17 months agoMerge tag 'io_uring-6.5-2023-07-21' of git://git.kernel.dk/linux
Linus Torvalds [Sat, 22 Jul 2023 17:46:30 +0000 (10:46 -0700)]
Merge tag 'io_uring-6.5-2023-07-21' of git://git.kernel.dk/linux

Pull io_uring fixes from Jens Axboe:

 - Fix for io-wq not always honoring REQ_F_NOWAIT, if it was set and
   punted directly (eg via DRAIN) (me)

 - Capability check fix (Ondrej)

 - Regression fix for the mmap changes that went into 6.4, which
   apparently broke IA64 (Helge)

* tag 'io_uring-6.5-2023-07-21' of git://git.kernel.dk/linux:
  ia64: mmap: Consider pgoff when searching for free mapping
  io_uring: Fix io_uring mmap() by using architecture-provided get_unmapped_area()
  io_uring: treat -EAGAIN for REQ_F_NOWAIT as final for io-wq
  io_uring: don't audit the capability check in io_uring_create()

17 months agoMerge tag 'devicetree-fixes-for-6.5-1' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 22 Jul 2023 17:28:22 +0000 (10:28 -0700)]
Merge tag 'devicetree-fixes-for-6.5-1' of git://git./linux/kernel/git/robh/linux

Pull devicetree fixes from Rob Herring:

 - Fix moortec,mr75203 schema usage of 'multipleOf' keyword

 - Fix regression in systems depending on "of-display" device name

 - Build fix for s390 with CONFIG_PCI=n and OF_EARLY_FLATTREE=y

 - Drop two obsolete serial .txt bindings

* tag 'devicetree-fixes-for-6.5-1' of git://git.kernel.org/pub/scm/linux/kernel/git/robh/linux:
  dt-bindings: serial: Remove obsolete nxp,lpc1850-uart.txt
  dt-bindings: serial: Remove obsolete cavium-uart.txt
  dt-bindings: hwmon: moortec,mr75203: fix multipleOf for coefficients
  of: Preserve "of-display" device name for compatibility
  of: make OF_EARLY_FLATTREE depend on HAS_IOMEM

17 months agoMerge tag 'regmap-fix-v6.5-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
Linus Torvalds [Sat, 22 Jul 2023 17:20:56 +0000 (10:20 -0700)]
Merge tag 'regmap-fix-v6.5-rc2' of git://git./linux/kernel/git/broonie/regmap

Pull regmap fixes from Mark Brown:
 "Three fixes here:

   - The issues with accounting for register and padding length on raw
     buses turn out to be quite widespread in custom buses.

     In order to avoid disturbing anything drop the initial fixes and
     fall back to a point fix in the SMBus code where the issue was
     originally noticed, a more substantial refactoring of the API which
     ensures that all buses make the same assumptions will follow.

   - The generic regcache code had been forcing on async I/O which did
     not work with the new maple tree sync code when used with SPI.

     Since that was mainly for the rbtree cache and the assumptions
     about hardware that drove the choice are probably not true any more
     fix this by pushing the enablement of async down into the rbtree
     code.

     This probably also makes cache syncs for systems faster though it's
     not the point.

   - The test code was triggering use of the rbtree and maple tree
     caches with dynamic allocation of nodes since all the testing is
     with RAM backed caches with no I/O performance issues.

     Just disable the locking in the tests to avoid triggering warnings
     when allocation debugging is turned on, it's not really what's
     being tested"

* tag 'regmap-fix-v6.5-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git/broonie/regmap:
  regmap: Disable locking for RBTREE and MAPLE unit tests
  regcache: Push async I/O request down into the rbtree cache
  regmap: Account for register length in SMBus I/O limits
  regmap: Drop initial version of maximum transfer length fixes

17 months agoMerge tag 'gpio-fixes-for-v6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel...
Linus Torvalds [Sat, 22 Jul 2023 17:14:04 +0000 (10:14 -0700)]
Merge tag 'gpio-fixes-for-v6.5-rc3' of git://git./linux/kernel/git/brgl/linux

Pull gpio fixes from Bartosz Golaszewski:

 - fix initial value handling for output-only pins in gpio-tps68470

 - fix two resource leaks in gpio-mvebu

* tag 'gpio-fixes-for-v6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/brgl/linux:
  gpio: mvebu: fix irq domain leak
  gpio: mvebu: Make use of devm_pwmchip_add
  gpio: tps68470: Make tps68470_gpio_output() always set the initial value

17 months agodt-bindings: serial: Remove obsolete nxp,lpc1850-uart.txt
Rob Herring [Fri, 7 Jul 2023 22:16:06 +0000 (16:16 -0600)]
dt-bindings: serial: Remove obsolete nxp,lpc1850-uart.txt

nxp,lpc1850-uart.txt binding is already covered by 8250.yaml, so remove
it.

Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230707221607.1064888-1-robh@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
17 months agodt-bindings: serial: Remove obsolete cavium-uart.txt
Rob Herring [Fri, 7 Jul 2023 22:16:02 +0000 (16:16 -0600)]
dt-bindings: serial: Remove obsolete cavium-uart.txt

cavium-uart.txt binding is already covered by 8250.yaml, so remove it.

Reviewed-by: Conor Dooley <conor.dooley@microchip.com>
Link: https://lore.kernel.org/r/20230707221602.1063972-1-robh@kernel.org
Signed-off-by: Rob Herring <robh@kernel.org>
17 months agoloop: do not enforce max_loop hard limit by (new) default
Mauricio Faria de Oliveira [Thu, 20 Jul 2023 14:30:33 +0000 (11:30 -0300)]
loop: do not enforce max_loop hard limit by (new) default

Problem:

The max_loop parameter is used for 2 different purposes:

1) initial number of loop devices to pre-create on init
2) maximum number of loop devices to add on access/open()

Historically, its default value (zero) caused 1) to create non-zero
number of devices (CONFIG_BLK_DEV_LOOP_MIN_COUNT), and no hard limit on
2) to add devices with autoloading.

However, the default value changed in commit 85c50197716c ("loop: Fix
the max_loop commandline argument treatment when it is set to 0") to
CONFIG_BLK_DEV_LOOP_MIN_COUNT, for max_loop=0 not to pre-create devices.

That does improve 1), but unfortunately it breaks 2), as the default
behavior changed from no-limit to hard-limit.

Example:

For example, this userspace code broke for N >= CONFIG, if the user
relied on the default value 0 for max_loop:

    mknod("/dev/loopN");
    open("/dev/loopN");  // now fails with ENXIO

Though affected users may "fix" it with (loop.)max_loop=0, this means to
require a kernel parameter change on stable kernel update (that commit
Fixes: an old commit in stable).
Solution:

The original semantics for the default value in 2) can be applied if the
parameter is not set (ie, default behavior).

This still keeps the intended function in 1) and 2) if set, and that
commit's intended improvement in 1) if max_loop=0.

Before 85c50197716c:
  - default:     1) CONFIG devices   2) no limit
  - max_loop=0:  1) CONFIG devices   2) no limit
  - max_loop=X:  1) X devices        2) X limit

After 85c50197716c:
  - default:     1) CONFIG devices   2) CONFIG limit (*)
  - max_loop=0:  1) 0 devices (*)    2) no limit
  - max_loop=X:  1) X devices        2) X limit

This commit:
  - default:     1) CONFIG devices   2) no limit (*)
  - max_loop=0:  1) 0 devices        2) no limit
  - max_loop=X:  1) X devices        2) X limit

Future:

The issue/regression from that commit only affects code under the
CONFIG_BLOCK_LEGACY_AUTOLOAD deprecation guard, thus the fix too is
contained under it.

Once that deprecated functionality/code is removed, the purpose 2) of
max_loop (hard limit) is no longer in use, so the module parameter
description can be changed then.

Tests:

Linux 6.4-rc7
CONFIG_BLK_DEV_LOOP_MIN_COUNT=8
CONFIG_BLOCK_LEGACY_AUTOLOAD=y

- default (original)

# ls -1 /dev/loop*
/dev/loop-control
/dev/loop0
...
/dev/loop7

# ./test-loop
open: /dev/loop8: No such device or address

- default (patched)

# ls -1 /dev/loop*
/dev/loop-control
/dev/loop0
...
/dev/loop7

# ./test-loop
#

- max_loop=0 (original & patched):

# ls -1 /dev/loop*
/dev/loop-control

# ./test-loop
#

- max_loop=8 (original & patched):

# ls -1 /dev/loop*
/dev/loop-control
/dev/loop0
...
/dev/loop7

# ./test-loop
open: /dev/loop8: No such device or address

- max_loop=0 (patched; CONFIG_BLOCK_LEGACY_AUTOLOAD is not set)

# ls -1 /dev/loop*
/dev/loop-control

# ./test-loop
open: /dev/loop8: No such device or address

Fixes: 85c50197716c ("loop: Fix the max_loop commandline argument treatment when it is set to 0")
Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20230720143033.841001-3-mfo@canonical.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
17 months agoloop: deprecate autoloading callback loop_probe()
Mauricio Faria de Oliveira [Thu, 20 Jul 2023 14:30:32 +0000 (11:30 -0300)]
loop: deprecate autoloading callback loop_probe()

The 'probe' callback in __register_blkdev() is only used under the
CONFIG_BLOCK_LEGACY_AUTOLOAD deprecation guard.

The loop_probe() function is only used for that callback, so guard it
too, accordingly.

See commit fbdee71bb5d8 ("block: deprecate autoloading based on dev_t").

Signed-off-by: Mauricio Faria de Oliveira <mfo@canonical.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Link: https://lore.kernel.org/r/20230720143033.841001-2-mfo@canonical.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
17 months agosbitmap: fix batching wakeup
David Jeffery [Fri, 21 Jul 2023 09:57:15 +0000 (17:57 +0800)]
sbitmap: fix batching wakeup

Current code supposes that it is enough to provide forward progress by
just waking up one wait queue after one completion batch is done.

Unfortunately this way isn't enough, cause waiter can be added to wait
queue just after it is woken up.

Follows one example(64 depth, wake_batch is 8)

1) all 64 tags are active

2) in each wait queue, there is only one single waiter

3) each time one completion batch(8 completions) wakes up just one
   waiter in each wait queue, then immediately one new sleeper is added
   to this wait queue

4) after 64 completions, 8 waiters are wakeup, and there are still 8
   waiters in each wait queue

5) after another 8 active tags are completed, only one waiter can be
   wakeup, and the other 7 can't be waken up anymore.

Turns out it isn't easy to fix this problem, so simply wakeup enough
waiters for single batch.

Cc: Kemeng Shi <shikemeng@huaweicloud.com>
Cc: Chengming Zhou <zhouchengming@bytedance.com>
Cc: Jan Kara <jack@suse.cz>
Signed-off-by: David Jeffery <djeffery@redhat.com>
Signed-off-by: Ming Lei <ming.lei@redhat.com>
Reviewed-by: Gabriel Krisman Bertazi <krisman@suse.de>
Reviewed-by: Keith Busch <kbusch@kernel.org>
Link: https://lore.kernel.org/r/20230721095715.232728-1-ming.lei@redhat.com
Signed-off-by: Jens Axboe <axboe@kernel.dk>
17 months agoMerge tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux
Linus Torvalds [Fri, 21 Jul 2023 17:24:21 +0000 (10:24 -0700)]
Merge tag 'arm64-fixes' of git://git./linux/kernel/git/arm64/linux

Pull arm64 fixes from Will Deacon:
 "I've picked up a handful of arm64 fixes while Catalin's been away, so
  here they are. Below is the usual summary, but we have basically have
  two cleanups, a fix for an SME crash and a fix for hibernation:

   - Fix saving of SME state after SVE vector length is changed

   - Fix sparse warnings for missing vDSO function prototypes

   - Fix hibernation resume path when kfence is enabled

   - Fix field names for the HFGxTR_EL2 register"

* tag 'arm64-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/arm64/linux:
  arm64/fpsimd: Ensure SME storage is allocated after SVE VL changes
  arm64: vdso: Clear common make C=2 warnings
  arm64: mm: Make hibernation aware of KFENCE
  arm64: Fix HFGxTR_EL2 field naming

17 months agoMerge tag 'pm-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm
Linus Torvalds [Fri, 21 Jul 2023 17:16:20 +0000 (10:16 -0700)]
Merge tag 'pm-6.5-rc3' of git://git./linux/kernel/git/rafael/linux-pm

Pull power management fixes from Rafael Wysocki:
 "Revert three recent intel_idle commits that introduced a functional
  issue, included a coding mistake and have been questioned at the
  design level"

* tag 'pm-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/rafael/linux-pm:
  Revert "intel_idle: Add support for using intel_idle in a VM guest using just hlt"
  Revert "intel_idle: Add a "Long HLT" C1 state for the VM guest mode"
  Revert "intel_idle: Add __init annotation to matchup_vm_state_with_baremetal()"

17 months agoMerge tag 'sound-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai...
Linus Torvalds [Fri, 21 Jul 2023 17:10:18 +0000 (10:10 -0700)]
Merge tag 'sound-6.5-rc3' of git://git./linux/kernel/git/tiwai/sound

Pull sound fixes from Takashi Iwai:
 "A pile of fixes that have been gathered since the previous pull. Most
  of changes are device-specific, and nothing looks too scary.

   - A memory leak fix in ALSA sequencer code in 6.5-rc

   - Many fixes for ASoC Qualcomm CODEC drivers, covering SoundWire
     probe problems

   - A series of ASoC AMD fixes

   - A few fixes and cleanups of selftest stuff

   - HD-audio codec fixes and quirks for Clevo, HP, Lenovo, Dell"

* tag 'sound-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/tiwai/sound: (52 commits)
  ALSA: hda/realtek: Add support for DELL Oasis 13/14/16 laptops
  ALSA: hda/realtek: Fix generic fixup definition for cs35l41 amp
  ALSA: hda/realtek: Enable Mute LED on HP Laptop 15s-eq2xxx
  selftests: ALSA: Add test-pcmtest-driver to .gitignore
  ALSA: hda/realtek: Add quirk for Clevo NS70AU
  ASoC: fsl_sai: Disable bit clock with transmitter
  ALSA: seq: Fix memory leak at error path in snd_seq_create_port()
  ASoC: SOF: ipc3-dtrace: uninitialized data in dfsentry_trace_filter_write()
  ASoC: cs42l51: fix driver to properly autoload with automatic module loading
  MAINTAINERS: Redo addition of ssm3515 to APPLE SOUND
  ASoC: rt5640: Fix the issue of speaker noise
  ALSA: hda/realtek - remove 3k pull low procedure
  selftests: ALSA: Fix fclose on an already fclosed file pointer
  ALSA: pcmtest: Don't use static storage to track per device data
  ALSA: pcmtest: Convert to platform remove callback returning void
  ASoC: dt-bindings: audio-graph-card2: Drop incomplete example
  ASoC: dt-bindings: Update maintainer email id
  ASoC: amd: ps: Fix extraneous error messages
  ASoC: fsl_sai: Revert "ASoC: fsl_sai: Enable MCTL_MCLK_EN bit for master mode"
  ASoC: codecs: SND_SOC_WCD934X should select REGMAP_IRQ
  ...

17 months agoMerge tag 'fbdev-for-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller...
Linus Torvalds [Fri, 21 Jul 2023 17:00:09 +0000 (10:00 -0700)]
Merge tag 'fbdev-for-6.5-rc3' of git://git./linux/kernel/git/deller/linux-fbdev

Pull fbdev fixes and cleanups from Helge Deller:
 "Just the usual bunch of code cleanups in various drivers, this time
  mostly in vgacon and imxfb:

   - Code cleanup in vgacon (Jiri Slaby)

   - Explicitly include correct DT includes (Rob Herring)

   - imxfb code cleanup (Yangtao Li, Martin Kaiser)

   - kyrofb: make arrays const and smaller (Colin Ian King)

   - ep93xx-fb: return value check fix (Yuanjun Gong)

   - au1200fb: add missing IRQ check (Zhang Shurong)"

* tag 'fbdev-for-6.5-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/deller/linux-fbdev:
  fbdev: Explicitly include correct DT includes
  fbdev: ep93xx-fb: fix return value check in ep93xxfb_probe
  fbdev: au1200fb: Fix missing IRQ check in au1200fb_drv_probe
  fbdev: kyro: make some const read-only arrays static and reduce type size
  fbcon: remove unused display (p) from fbcon_redraw()
  sticon: make sticon_set_def_font() void and remove op parameter
  vgacon: cache vc_cell_height in vgacon_cursor()
  vgacon: let vgacon_doresize() return void
  vgacon: remove unused xpos from vgacon_set_cursor_size()
  vgacon: remove unneeded forward declarations
  vgacon: switch vgacon_scrolldelta() and vgacon_restore_screen()
  fbdev: imxfb: remove unneeded labels
  fbdev: imxfb: Convert to devm_platform_ioremap_resource()
  fbdev: imxfb: Convert to devm_kmalloc_array()
  fbdev: imxfb: Removed unneeded release_mem_region
  fbdev: imxfb: switch to DEFINE_SIMPLE_DEV_PM_OPS
  fbdev: imxfb: warn about invalid left/right margin

17 months agodrm/atomic: Fix potential use-after-free in nonblocking commits
Daniel Vetter [Fri, 21 Jul 2023 13:58:38 +0000 (15:58 +0200)]
drm/atomic: Fix potential use-after-free in nonblocking commits

This requires a bit of background.  Properly done a modeset driver's
unload/remove sequence should be

drm_dev_unplug();
drm_atomic_helper_shutdown();
drm_dev_put();

The trouble is that the drm_dev_unplugged() checks are by design racy,
they do not synchronize against all outstanding ioctl.  This is because
those ioctl could block forever (both for modeset and for driver
specific ioctls), leading to deadlocks in hotunplug.  Instead the code
sections that touch the hardware need to be annotated with
drm_dev_enter/exit, to avoid accessing hardware resources after the
unload/remove has finished.

To avoid use-after-free issues all the involved userspace visible
objects are supposed to hold a reference on the underlying drm_device,
like drm_file does.

The issue now is that we missed one, the atomic modeset ioctl can be run
in a nonblocking fashion, and in that case it cannot rely on the implied
drm_device reference provided by the ioctl calling context.  This can
result in a use-after-free if an nonblocking atomic commit is carefully
raced against a driver unload.

Fix this by unconditionally grabbing a drm_device reference for any
drm_atomic_state structures.  Strictly speaking this isn't required for
blocking commits and TEST_ONLY calls, but it's the simpler approach.

Thanks to shanzhulig for the initial idea of grabbing an unconditional
reference, I just added comments, a condensed commit message and fixed a
minor potential issue in where exactly we drop the final reference.

Reported-by: shanzhulig <shanzhulig@gmail.com>
Suggested-by: shanzhulig <shanzhulig@gmail.com>
Reviewed-by: Maxime Ripard <mripard@kernel.org>
Cc: Maarten Lankhorst <maarten.lankhorst@linux.intel.com>
Cc: Thomas Zimmermann <tzimmermann@suse.de>
Cc: David Airlie <airlied@gmail.com>
Cc: stable@kernel.org
Signed-off-by: Daniel Vetter <daniel.vetter@intel.com>
Signed-off-by: Daniel Vetter <daniel.vetter@ffwll.ch>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>