Hi All,
The previous attempt to address the issue tackled by this series, commit
44cc89f76464 ("PM: runtime: Update device status before letting suppliers
suspend") was incorrect, because it introduced a rather nasty race condition
into __rpm_callback(), so let's revert it (patch [1/2]).
Instead, let's avoid suspending the suppliers immediately after dropping the
PM-runtime references to them and do that later, when the status of the
consumer has changed to RPM_SUSPENDED.
Please see the patch changelogs for details.
Elaine, please test this series on the system where you saw the original
problem.
Thanks!
On Thursday, March 18, 2021 7:06:54 PM CET Rafael J. Wysocki wrote:
> Hi All,
>
> The previous attempt to address the issue tackled by this series, commit
> 44cc89f76464 ("PM: runtime: Update device status before letting suppliers
> suspend") was incorrect, because it introduced a rather nasty race condition
> into __rpm_callback(), so let's revert it (patch [1/2]).
>
> Instead, let's avoid suspending the suppliers immediately after dropping the
> PM-runtime references to them and do that later, when the status of the
> consumer has changed to RPM_SUSPENDED.
>
> Please see the patch changelogs for details.
>
> Elaine, please test this series on the system where you saw the original
> problem.
The above still applies.
The v2 is sent because of some updates in the second patch, plus the tags from Ulf.
Thanks!