On Mon, Jun 22, 2020 at 4:12 AM Michal Soltys <msoltyspl@xxxxxxxxx> wrote:
>
> On 6/20/20 2:14 AM, Song Liu wrote:
> > On Fri, Jun 19, 2020 at 4:35 AM Michal Soltys <msoltyspl@xxxxxxxxx> wrote:
> >>
> >> On 6/17/20 7:11 PM, Song Liu wrote:
> >>>>
> >>>>> 1. There are two pr_debug() calls in handle_stripe():
> >>>>> pr_debug("handling stripe %llu, state=%#lx cnt=%d, "
> >>>>> pr_debug("locked=%d uptodate=%d to_read=%d"
> >>>>>
> >>>>> Did you enable all of them? Or only the first one?
> >>>>
> >>>> I enabled all of them (I think), to be precise:
> >>>>
> >>>> echo -n 'func handle_stripe +p' >/sys/kernel/debug/dynamic_debug/control
> >>>>
> >>>> Haven't seen any `locked` lines though.
> >>>
> >>> That's a little weird, and probably explains why we stuck. Could you
> >>> please try the attached patch?
> >>>
> >>> Thanks,
> >>> Song
> >>>
> >>
> >> I've started assembly with the above patch, the output can be inspected
> >> from this file:
> >>
> >> https://ufile.io/yx4bbcb4
> >>
> >> This is ~5mb packed dmesg from start of the boot to the relevant parts,
> >> unpacks to ~150mb file.
> >
> > Thanks for the trace. Looks like we may have some issues with
> > MD_SB_CHANGE_PENDING.
> > Could you please try the attached patch?
>
> Should I run this along with pr_debugs from the previous patch enabled ?
We don't need those pr_debug() here.
Thanks,
Song
