Thanks for the report! However, this has been reported earlier during the v2 review [1]. This is also why a fix have been provided [2] earlier today. Notice that shape of link->exit() found here is shared by other Intel boards e.g.: SOF ones. In general, the initial discussion regarding card->remove() revealed some 'probe vs remove' problems within the framework.
It's rather difficult to follow these changes and error reports buried in email report sent on a Sunday of a three-day week-end for me. I also had additional errors not reported,
[ 36.125113] kernel: rt286 i2c-INT343A:00: ASoC: unknown pin HV [ 36.125128] kernel: rt286 i2c-INT343A:00: ASoC: unknown pin VREF [ 36.125130] kernel: rt286 i2c-INT343A:00: ASoC: unknown pin LDO1 [ 36.125921] kernel: rt286 i2c-INT343A:00: ASoC: DAPM unknown pin LDO1
it's unclear to me why a dailink change in a machine driver would cause such codec-side issues.
If the changes in this 17-patch series need to be tied to a framework fix, you have to make the dependencies explicit and better yet provide a self-contained patch series that does not introduce a temporary regression, or introduce the framework change first and clearly describe the dependency in a longer Broadwell-specific patchset. This is an 8-yr old device, it shouldn't be that hard.
The last part is not helpful in solving the problem.
This reply comments 00/17 whereas in fact you are speaking solely about 16/17. Because of that I'm suggesting: leave that patch (the 16/17 one) out when merging. It will be send later once link->exit() issue is dealt with. All other patches are independent of either of changes.
That's fine with me. It wasn't self-explanatory from this cover letter or your earlier answer that this patch 16 can be dropped for now. If that patch is omitted, feel free to add my
Tested-by: Pierre-Louis Bossart pierre-louis.bossart@linux.intel.com
Simultaneously the link->exit() fix, which is the fruit of this discussion, is still valid and can be send as standalone patch - what is already the case [1].
That's fine as well. What I was arguing on is the relationship between patchsets and dependencies, what you are suggesting is perfectly acceptable.