mbox series

[SRU,O,0/4] drm/xe: Fix LNL gpu getting wedged after idling

Message ID 20241030145114.243657-1-tjaalton@ubuntu.com
Headers show
Series drm/xe: Fix LNL gpu getting wedged after idling | expand

Message

Timo Aaltonen Oct. 30, 2024, 2:51 p.m. UTC
BugLink: https://bugs.launchpad.net/bugs/2085944

[Impact]

Leaving a system with Lunar Lake idle for an hour or so will end up with the display driver getting 
wedged requiring a reboot.

This is fixed upstream by this commit:

commit 22ef43c78647dd37b0dafe2182b8650b99dbbe59
Author: Badal Nilawar <badal.nilawar@intel.com>
Date: Thu Oct 17 16:44:10 2024 +0530

    drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout

but in order to apply cleanly, we also need a set of three patches which are already in v6.11.5

[Test case]

Boot a patched kernel on LNL, leave it idle for a couple of hours/overnight and check that it's
still working normally.

Badal Nilawar (1):
  drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout

Matthew Auld (3):
  drm/xe/guc_submit: fix xa_store() error checking
  drm/xe/ct: prevent UAF in send_recv()
  drm/xe/ct: fix xa_store() error checking

 drivers/gpu/drm/xe/xe_guc_ct.c     | 62 +++++++++++++++++++++---------
 drivers/gpu/drm/xe/xe_guc_submit.c |  9 ++---
 2 files changed, 47 insertions(+), 24 deletions(-)

Comments

Guoqing Jiang Oct. 30, 2024, 4 p.m. UTC | #1
Hi,

On 10/30/24 15:51, Timo Aaltonen wrote:
> BugLink: https://bugs.launchpad.net/bugs/2085944
>
> [Impact]
>
> Leaving a system with Lunar Lake idle for an hour or so will end up with the display driver getting
> wedged requiring a reboot.
>
> This is fixed upstream by this commit:
>
> commit 22ef43c78647dd37b0dafe2182b8650b99dbbe59
> Author: Badal Nilawar <badal.nilawar@intel.com>
> Date: Thu Oct 17 16:44:10 2024 +0530
>
>      drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout
>
> but in order to apply cleanly, we also need a set of three patches which are already in v6.11.5
>
> [Test case]
>
> Boot a patched kernel on LNL, leave it idle for a couple of hours/overnight and check that it's
> still working normally.
>
> Badal Nilawar (1):
>    drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout
>
> Matthew Auld (3):
>    drm/xe/guc_submit: fix xa_store() error checking
>    drm/xe/ct: prevent UAF in send_recv()
>    drm/xe/ct: fix xa_store() error checking

Looks the above three missed one original SOB

Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>

and relevant cherry-pick info.

42465603a310 drm/xe/guc_submit: fix xa_store() error checking
e863781abe4f drm/xe/ct: fix xa_store() error checking
db7f92af6261 drm/xe/ct: prevent UAF in send_recv()

Thanks,
Guoqing

>   drivers/gpu/drm/xe/xe_guc_ct.c     | 62 +++++++++++++++++++++---------
>   drivers/gpu/drm/xe/xe_guc_submit.c |  9 ++---
>   2 files changed, 47 insertions(+), 24 deletions(-)
>
Timo Aaltonen Oct. 30, 2024, 4:10 p.m. UTC | #2
Guoqing Jiang kirjoitti 30.10.2024 klo 18.00:
> Hi,
> 
> On 10/30/24 15:51, Timo Aaltonen wrote:
>> BugLink: https://bugs.launchpad.net/bugs/2085944
>>
>> [Impact]
>>
>> Leaving a system with Lunar Lake idle for an hour or so will end up 
>> with the display driver getting
>> wedged requiring a reboot.
>>
>> This is fixed upstream by this commit:
>>
>> commit 22ef43c78647dd37b0dafe2182b8650b99dbbe59
>> Author: Badal Nilawar <badal.nilawar@intel.com>
>> Date: Thu Oct 17 16:44:10 2024 +0530
>>
>>      drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout
>>
>> but in order to apply cleanly, we also need a set of three patches 
>> which are already in v6.11.5
>>
>> [Test case]
>>
>> Boot a patched kernel on LNL, leave it idle for a couple of hours/ 
>> overnight and check that it's
>> still working normally.
>>
>> Badal Nilawar (1):
>>    drm/xe/guc/ct: Flush g2h worker in case of g2h response timeout
>>
>> Matthew Auld (3):
>>    drm/xe/guc_submit: fix xa_store() error checking
>>    drm/xe/ct: prevent UAF in send_recv()
>>    drm/xe/ct: fix xa_store() error checking
> 
> Looks the above three missed one original SOB
> 
> Signed-off-by: Lucas De Marchi <lucas.demarchi@intel.com>
> 
> and relevant cherry-pick info.

I used the linux.git/master as the source, not stable. Confusingly, the 
fourth one has cherry-pick info there as well.