Re: not ok 1 - single_pixel_source_buffer: The buggy address belongs to the physical page

[David Gow]

On Tue, Oct 18, 2022 at 3:54 PM Javier Martinez Canillas
<javierm@xxxxxxxxxx> wrote:
>
> [adding a few folks to Cc list that might help with this issue]
>
> Hello Naresh,
>
> Thanks a lot for your report.
>
> On 10/18/22 08:40, Naresh Kamboju wrote:
> > Following kunit tests started failing on Linux mainline.
> >  - drm_format_helper_test — FAIL
> >  - drm_test_fb_xrgb8888_to_xrgb2101010 — FAIL
> >  - single_pixel_source_buffer — FAIL
> >
> > Good: v6.0-3015-g2bca25eaeba6
> > Bad: v6.0-5118-g833477fce7a1
> >
>
> Could you please let me know how you are running this? I tried to reproduce it
> on v6.1-rc1 with the following command but all tests passed:
>
> ./tools/testing/kunit/kunit.py run --kunitconfig=drivers/gpu/drm/tests/.kunitconfig --arch=x86_64
> [09:41:53] Configuring KUnit Kernel ...
> [09:41:53] Building KUnit Kernel ...
> Populating config with:
> $ make ARCH=x86_64 O=.kunit olddefconfig
> Building with:
> $ make ARCH=x86_64 O=.kunit --jobs=8
> [09:45:51] Starting KUnit Kernel (1/1)...
> [09:45:51] ============================================================
> Running tests with:
> $ qemu-system-x86_64 -nodefaults -m 1024 -kernel .kunit/arch/x86/boot/bzImage -append 'kunit.enable=1 console=ttyS0 kunit_shutdown=reboot' -no-reboot -nographic -serial stdio
> ...
> [09:47:40] Testing complete. Ran 195 tests: passed: 195
> [09:47:40] Elapsed time: 347.817s total, 0.003s configuring, 238.009s building, 109.771s running

I can reproduce this with:
./tools/testing/kunit/kunit.py run --kunitconfig drivers/gpu/drm/tests
--kconfig_add CONFIG_KASAN=y --kconfig_add CONFIG_KASAN_VMALLOC=y
--kconfig_add CONFIG_KASAN_KUNIT_TEST=y
drm_format_helper_test.*xrgb2101010

(The issue shows up only with KASAN enabled, and it looks like there's
a bug whereby KASAN failures don't trigger test failures unless
CONFIG_KASAN_KUNIT_TEST=y)

It looks like the issue is probably with the conversion back to le32 here:
https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/gpu/drm/tests/drm_format_helper_test.c#n441

That second call to le32buf_to_cpu() should probably take dst_size (or
rather, dst_size / sizeof(u32)) rather than using TEST_BUF_SIZE, which
is the maximum possible size of the buffer, not the actual size. That
fixes it here for me, though a proper fix would avoid the division.

Cheers,
-- David

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature