Re: [PATCH v2 0/3] serial: qcom-geni: fix lockups

[Johan Hovold]

Hi Doug,

Hope you had a good holiday.

On Mon, Jul 08, 2024 at 04:57:55PM -0700, Doug Anderson wrote:
> On Thu, Jul 4, 2024 at 3:31 AM Greg Kroah-Hartman
> <gregkh@xxxxxxxxxxxxxxxxxxx> wrote:
> >
> > On Thu, Jul 04, 2024 at 12:18:02PM +0200, Johan Hovold wrote:
> > > Since 6.10-rc1, Qualcomm machines with a serial port can easily lock up
> > > hard, for example, when stopping a getty on reboot.
> > >
> > > This was triggered by the kfifo conversion, which turned an existing bug
> > > that caused the driver to print discarded characters after a buffer
> > > flush into a hard lockup.
> > >
> > > This series fixes the regression and a related soft lockup issue that
> > > can be triggered on software flow control and on suspend.
> > >
> > > Doug has posted an alternative series of fixes here that depends on
> > > reworking the driver a fair bit here:
> > >
> > >       https://lore.kernel.org/lkml/20240610222515.3023730-1-dianders@xxxxxxxxxxxx/
> > >
> > > This rework has a significant impact on performance on some platforms,
> > > but fortunately it seems such a rework can be avoided.
> > >
> > > There are further bugs in the console code (e.g. that can lead to lost
> > > characters) that this series does not address, but those can be fixed
> > > separately (and I've started working on that).
> >
> > I'll take these now, thanks!
> 
> Are you going to continue to work on the driver? There are still some
> pretty bad bugs including ones that are affecting Collabora's test
> labs. Unless you want to try to tackle it some other way, I'm going to
> keep pushing for something like my original series to land. I can
> re-post them atop your patches since they've landed. This will regress
> your performance but correctness trumps performance.

Yes, I have a working fix for the console issue that could lead to lost
characters. I need to spend some time on other things this week, but I
intend to follow with fixes for the remaining issues as well.

Johan