On 8.01.2024 16:19, Greg KH wrote: > On Mon, Jan 08, 2024 at 08:33:00PM +0530, Amit Pundir wrote: >> On Mon, 8 Jan 2024 at 19:42, Greg KH <gregkh@xxxxxxxxxxxxxxxxxxx> wrote: >>> >>> On Sun, Jan 07, 2024 at 09:27:02PM +0530, Amit Pundir wrote: >>>> This reverts commit 3637f6bdfe2ccd53c493836b6e43c9a73e4513b3 which is >>>> commit bfc7db1cb94ad664546d70212699f8cc6c539e8c upstream. >>>> >>>> This resulted in boot regression on RB5 (sm8250), causing the device >>>> to hard crash into USB crash dump mode everytime. >>>> >>>> Signed-off-by: Amit Pundir <amit.pundir@xxxxxxxxxx> >>> >>> Any link to that report? Is this also an issue in 6.7 and/or 6.6.y? >> >> Here is a fresh RB5 crash report running AOSP with upstream v6.1.71 >> https://lkft.validation.linaro.org/scheduler/job/7151629#L4239 >> >> I do not see this crash on v6.7. > > So does that mean we are instead missing something here for this tree? Yes, however I'm not sure anybody is keen on tracking that down, as (in short) for the platform to work correctly (.sync_state on interconnect not crashing the thing into oblivion is one of the signs), a lot of things need to be in place. And some developers never validated that properly.. Removing .sync_state from the SoC interconnect driver translates into "keep the power flowing on all data buses", which helps avoid crashes that are mainly caused by unclocked accesses and alike. Konrad
