Re: Feedback on Qualcomm's minidump (debug) solution for end user device crash

[Mukesh Ojha]

On 8/9/2023 1:19 PM, Kukjin Kim wrote:
> > 2023. 8. 8. 오전 12:08, Brian Masney <bmasney@xxxxxxxxxx> 작성:
> >
> > On Mon, Aug 07, 2023 at 06:01:27PM +0530, Mukesh Ojha wrote:
> > > > On 7/30/2023 5:14 PM, Krzysztof Kozlowski wrote:
> > > > On 24/07/2023 18:59, Brian Masney wrote:
> > > > > + linux-arm-kernel list
> > > > >
> > > > > On Thu, Jul 20, 2023 at 08:32:24PM +0530, Mukesh Ojha wrote:
> > > > > > Hi Samsung/MTK/Any other SOC vendors,
> > > > > >
> > > > > > This is to bring to your notice that, we (Qualcomm) are working on
> > > > > > upstreaming our minidump solution which is to address the problem of
> > > > > > debugging on field device crashes where collecting entire ddr dump
> > > > > > would not be feasible and collecting minimal data from the ddr would
> > > > > > help in debug direction or even help in root causing issue.
> > > > > >
> > > > > > We have recently posted v4 version here [1]
> > > > > >
> > > > > > Based on comments[2], community is more worried about, if each SOC
> > > > > > vendor come up with their own dumping method today or in future and
> > > > > > whether it can have a common solution to a similar problem faced by
> > > > > > other SOC vendor.
> > > > > >
> > > > > > We wanted to take your feedback if you also encounter a similar problem
> > > > > > or maintain something similar solution in downstream which can be
> > > > > > upstreamed. This will help us in a way to have a common solution in
> > > > > > upstream.
> > > > > >
> > > > > > [1]
> > > > > > https://lore.kernel.org/lkml/10dd2ead-758a-89f0-cda4-70ae927269eb@xxxxxxxxxxx/
> > > > > >
> > > > > > [2]
> > > > > > https://lore.kernel.org/lkml/CAL_JsqLO9yey2-4FcWsaGxijiS6hGL0SH9VoMuiyei-u9=Cv=w@xxxxxxxxxxxxxx/
> > > > >
> > > > > Adding the main ARM list to solicit feedback from other silicon
> > > > > manufacturers.
> > > > >
> > > > > The cover sheet on the v4 patch set is available at:
> > > > > https://lore.kernel.org/lkml/1687955688-20809-1-git-send-email-quic_mojha@xxxxxxxxxxx/
> > > >
> > > > I doubt anyone follows the lists, so at least Cc some maintainers.
> > > >
> > > > +Cc Alim, Kukjin, Vignesh, Nishanth, Matthias.
> > >
> > > Thanks @Krzysztof/@Brian for extending the list.
> >
> > Hi Mukesh,
> >
> > Since no one has responded yet: I suspect your best bet to land the
> > minidump functionality upstream is to refactor it to use the pstore
> > functionality that Rob suggested:
> >
> > https://lore.kernel.org/lkml/CAL_JsqK7MHR09U5h01=Gf1ZLeDVCgZdN-W1hQRH3AX+E94_uUg@xxxxxxxxxxxxxx/
> >
> > Brian
> Hi all,
>
> Sorry for the late response and thanks for the asking.
>
> In Samsung side, we’re checking about that internally as well. I’d like to know whether the minidump upstreaming is considered to be used in other chipset or some logic of that can be used. In addition, if Samsung wants, own the way upstreaming can be acceptable. It doesn’t mean we have a plan at this moment though.

Thanks for the response @Kukjin

It is not something considered already instead it is what Qualcomm has
support for its SOC in the downstream and we want to upstream the linux
drivers and Minidump will need boot firmware support to be able to be
used by other SOC vendors.

So, the reason of seeking feedback on exactly the problem statement
to debug crash from field devices where complete ddr dump is not
feasible instead collect minimal information available to debug
crashes and also about if SOC vendors have any solution for this
issue which is common and upstream-able and can be leveraged.

So, if the problem statement is valid and now it is the time to
share and get conclusion on some common solution, Feel free to
check and comment on our minidump patches.

https://lore.kernel.org/lkml/10dd2ead-758a-89f0-cda4-70ae927269eb@xxxxxxxxxxx/

-Mukesh

> Thanks,
> Kukjin Kim <kgene(at)kernel.org>