Re: ASoC: soc-pcm: Don't zero TDM masks in __soc_pcm_open() breaks SOF Audio in Lenovo laptops
On 9/12/22 16:18, Mark Brown wrote:
On Fri, Dec 09, 2022 at 05:38:54PM +0200, Péter Ujfalusi wrote:
On 09/12/2022 16:42, Pierre-Louis Bossart wrote:
Agree with the analysis, so what would be the least bad recommendation? a) revert the "don't zero TDM masks" patch b) backport the change to use set_stream()?
I would vote for b) unless other platforms report regressions.
I don't really care either way given the backport policy.
Neither do I. If you want to minimize risk, revert the patch in older backports.
Lessons learned: Don't hijack a data member to pass something different from what it is intended to hold. Don't depend on a bug. Don't assume all code is using a data member for what is supposed to be in that member.
On Fri, 2022-12-09 at 16:26 +0000, Richard Fitzgerald wrote:
On 9/12/22 16:18, Mark Brown wrote:
On Fri, Dec 09, 2022 at 05:38:54PM +0200, Péter Ujfalusi wrote:
On 09/12/2022 16:42, Pierre-Louis Bossart wrote:
Agree with the analysis, so what would be the least bad recommendation? a) revert the "don't zero TDM masks" patch b) backport the change to use set_stream()?
I would vote for b) unless other platforms report regressions.
I don't really care either way given the backport policy.
Neither do I. If you want to minimize risk, revert the patch in older backports.
Lessons learned: Don't hijack a data member to pass something different from what it is intended to hold. Don't depend on a bug. Don't assume all code is using a data member for what is supposed to be in that member.
Did you reach a conclusion w.r.t what the should be?
Jocke
On 12/13/22 10:57, Joakim Tjernlund wrote:
On Fri, 2022-12-09 at 16:26 +0000, Richard Fitzgerald wrote:
On 9/12/22 16:18, Mark Brown wrote:
On Fri, Dec 09, 2022 at 05:38:54PM +0200, Péter Ujfalusi wrote:
On 09/12/2022 16:42, Pierre-Louis Bossart wrote:
Agree with the analysis, so what would be the least bad recommendation? a) revert the "don't zero TDM masks" patch b) backport the change to use set_stream()?
I would vote for b) unless other platforms report regressions.
I don't really care either way given the backport policy.
Neither do I. If you want to minimize risk, revert the patch in older backports.
Lessons learned: Don't hijack a data member to pass something different from what it is intended to hold. Don't depend on a bug. Don't assume all code is using a data member for what is supposed to be in that member.
Did you reach a conclusion w.r.t what the should be?
I read the consensus is for a backport of the 'set_stream' stuff. You're welcome to submit a patch.
participants (3)
-
Joakim Tjernlund
-
Pierre-Louis Bossart
-
Richard Fitzgerald