APPLIED[J]: [PATCH 0/1][SRU][Jammy] Fix UBSAN warning for amdgpu yellow carp

Andrea Righi andrea.righi at canonical.com
Mon Jan 17 15:00:06 UTC 2022


On Mon, Jan 17, 2022 at 10:18:43PM +0800, You-Sheng Yang wrote:
> BugLink: https://bugs.launchpad.net/bugs/1956497
> 
> [Impact]
> 
> An UBSAN error is dumped once when unpluging a type-c cable (at DP Alt
> Mode) from a AMD Yellow Carp equipped platform. No side effects found
> so far, and subsequent hotplugs won't have this error again.
> 
> drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c:6084:84
> [ 38.099550] load of value 32 is not a valid value for type '_Bool'
> [ 38.099552] CPU: 9 PID: 1451 Comm: Xorg Not tainted
> 5.15.0-2013-generic #13~20.04.1+pre14+lp1953008.1
> [ 38.099554] Hardware name: AMD LilacKD-RMB/LilacKD-RMB, BIOS
> RRL0080C 09/15/2021
> [ 38.099555] Call Trace:
> [ 38.099556] <TASK>
> [ 38.099558] dump_stack_lvl+0x4a/0x5f
> [ 38.099563] dump_stack+0x10/0x12
> [ 38.099564] ubsan_epilogue+0x9/0x45
> [ 38.099565] __ubsan_handle_load_invalid_value+0x70/0x80
> [ 38.099568] create_stream_for_sink.cold+0x5d/0xbb [amdgpu]
> ...
> 
> [Fix]
> 
> Commit 63ad5371cd1e ("drm/amd/display: explicitly set is_dsc_supported
> to false before use") in linux-next.
> 
> [Test Case]
> 
> 1. Connect an external monitor via USB Type-C able to target platform
> 2. After external monitor turned on and screen was extended, unplug
>    Type-C cable
> 3. check if a similar ubsan warning is dumpped to dmesg.
> 
> [Where problems could occur]
> 
> No. This initialize a variable before being referenced.
> 
> [Other Info]
> 
> While UBSAN is only turned on on 5.15 and newer, only Jammy is nominated
> for this fix.

Applied to jammy/linux, thanks!

-Andrea



More information about the kernel-team mailing list