Re: linux-next: build failure after merge of the tip tree

[Ojaswin Mujoo]

On Wed, Mar 11, 2026 at 12:00:20AM +0000, Mark Brown wrote:
> On Tue, Mar 10, 2026 at 06:28:30PM +0000, Mark Brown wrote:
> > Hi all,
> > 
> > After merging the tip tree, today's linux-next started crashing running
> > arm64 KUnit like this:
> > 
> > [18:12:16] [PASSED] split unwrit extent to 3 extents and convert 2nd half writ (non-endio, zeroout) (highlevel)
> > [18:12:16] =============== [PASSED] test_split_convert ================
> > [18:12:16] ================ [PASSED] ext4_extents_test ================
> > [18:12:16] ============== ext4_mballoc_test (7 subtests) ==============
> > Command '['qemu-system-aarch64', '-nodefaults', '-m', '1024', '-kernel', '/tmp/next/arm64_kunit/arch/arm64/boot/Image.gz', '-append', 'kunit.enable=1 console=ttyAMA0 kunit_shutdown=reboot', '-no-reboot', '-nographic', '-accel', 'kvm', '-accel', 'hvf', '-accel', 'tcg', '-serial', 'stdio', '-machine', 'virt', '-cpu', 'max']' timed out after 300 seconds
> > 
> > I didn't figure out what the source of the issue was, I merged the tip
> > tree from 20260309 instead.
> 
> I tried to leave a bisect running but it got confused because a lot of
> the branches are based on v7.0-rc1 which has a separate bug that causes
> KUnit to lock up so the results are nonsense.  I did confirm an issue
> with just tip/master.  My KUnit command line running on current Debian
> stable is:
> 
>     ./tools/testing/kunit/kunit.py run --alltests --arch arm64 --cross_compile=aarch64-linux-gnu-
> 
> and I also tried:
> 
>     ./tools/testing/kunit/kunit.py run --alltests --arch x86_64 --cross_compile=x86_64-linux-gnu-
> 
> and got:
> 
> [23:51:03] [PASSED] split unwrit extent to 3 extents and convert 2nd half writ (non-endio, zeroout) (highlevel)
> [23:51:03] =============== [PASSED] test_split_convert ================
> [23:51:03] ================ [PASSED] ext4_extents_test ================
> [23:51:03] ============== ext4_mballoc_test (7 subtests) ==============
> [23:51:03] ================= test_new_blocks_simple  ==================
> [23:51:03] [FAILED] block_bits=10 cluster_bits=3 blocks_per_group=8192 group_count=4 desc_size=64
> 
> before the deadlock which looks awfully similar, though it does make a
> bit more progress and gives a crash:
> 
> [23:55:51] [ERROR] Test: test_new_blocks_simple: missing subtest result line!
> [23:55:51] BUG: unable to handle page fault for address: 00000008bcf958c7
> [23:55:51] #PF: supervisor write access in kernel mode
> [23:55:51] #PF: error_code(0x0002) - not-present page
> 
> ...
> 
> [23:58:45] Call Trace:
> [23:58:45] <TASK>
> [23:58:45] ext4_mb_release (??:?)
> [23:58:45] mbt_kunit_exit (mballoc.c:?)
> [23:58:45] ? __pfx_kunit_generic_run_threadfn_adapter (try-catch.c:?)
> [23:58:45] kunit_try_run_case_cleanup (test.c:?)
> [23:58:45] kunit_generic_run_threadfn_adapter (try-catch.c:?)
> [23:58:45] kthread (kthread.c:?)
> [23:58:45] ? __pfx_kthread (kthread.c:?)
> [23:58:45] ret_from_fork (??:?)
> [23:58:45] ? __pfx_kthread (kthread.c:?)
> [23:58:45] ret_from_fork_asm (??:?)
> 
> plus subsequent things that look like aftereffects of the same thing.

Hi Mark, I'm unable to hit this bug with the 20260310 branch [1]. Am I
on the wrong branch? I'm using the following command on x86 machine:

 tools/testing/kunit/kunit.py run --alltests --arch x86_64

Also IIUC there's another deadlock issue you are
facing other than the one here? Would you be able to try out this
patchset [2], it fixes some of the cleanup related functions which might
fix this issue.

[1] https://git.kernel.org/pub/scm/linux/kernel/git/next/linux-next.git/tag/?h=next-20260310
[2] https://lore.kernel.org/linux-ext4/20260310130412.3156753-1-yebin@xxxxxxxxxxxxxxx/T/#t

Regards,
ojaswin