Message-ID: <CACSVV027yBWLKaghuBnnGPg84W6RvrO=O+55ROx-SOeqTMbDGA@mail.gmail.com>
Date: Mon, 8 Sep 2025 12:13:57 -0700
From: Rob Clark <rob.clark@....qualcomm.com>
To: Ryan Eatmon <reatmon@...com>
Cc: Rob Clark <robdclark@...il.com>,
        Dmitry Baryshkov <dmitry.baryshkov@....qualcomm.com>,
        Viswanath Kraleti <viswanath.kraleti@....qualcomm.com>,
        Abhinav Kumar <quic_abhinavk@...cinc.com>, Sean Paul <sean@...rly.run>,
        Marijn Suijten <marijn.suijten@...ainline.org>,
        David Airlie <airlied@...il.com>, Simona Vetter <simona@...ll.ch>,
        linux-arm-msm@...r.kernel.org, dri-devel@...ts.freedesktop.org,
        freedreno@...ts.freedesktop.org, linux-kernel@...r.kernel.org,
        Bruce Ashfield <bruce.ashfield@...il.com>
Subject: Re: [PATCH v2] drivers: gpu: drm: msm: registers: improve reproducibility

On Mon, Sep 8, 2025 at 10:59 AM Ryan Eatmon <reatmon@...com> wrote:
>
>
>
> On 9/8/2025 9:19 AM, Rob Clark wrote:
> > On Mon, Sep 8, 2025 at 6:39 AM Ryan Eatmon <reatmon@...com> wrote:
> >>
> >>
> >>
> >> On 9/6/2025 6:24 PM, Rob Clark wrote:
> >>> On Sat, May 24, 2025 at 10:15 AM Dmitry Baryshkov
> >>> <dmitry.baryshkov@....qualcomm.com> wrote:
> >>>>
> >>>> On Sat, May 24, 2025 at 09:25:37PM +0530, Viswanath Kraleti wrote:
> >>>>> From: Ryan Eatmon <reatmon@...com>
> >>>>>
> >>>>> The files generated by gen_header.py capture the source path to the
> >>>>> input files and the date.  While that can be informative, it varies
> >>>>> based on where and when the kernel was built as the full path is
> >>>>> captured.
> >>>>>
> >>>>> Since all of the files that this tool is run on is under the drivers
> >>>>> directory, this modifies the application to strip all of the path before
> >>>>> drivers.  Additionally it prints <stripped> instead of the date.
> >>>>>
> >>>>> Signed-off-by: Ryan Eatmon <reatmon@...com>
> >>>>> Signed-off-by: Bruce Ashfield <bruce.ashfield@...il.com>
> >>>>> Signed-off-by: Viswanath Kraleti <viswanath.kraleti@....qualcomm.com>
> >>>>> ---
> >>>>> The files generated by gen_header.py include the source path to the
> >>>>> input files and the build date. While this information can be useful,
> >>>>> it inadvertently exposes build system configuration details in the
> >>>>> binaries. This hinders binary reproducibility, as the output will
> >>>>> vary if the build environment changes.
> >>>>>
> >>>>> This change was originally submitted to the linux-yocto-dev kernel [1]
> >>>>> to address binary reproducibility QA errors. However, the fix is generic
> >>>>> enough to be applicable to the mainline kernel and would benefit other
> >>>>> distributions as well. So proposing it here for broader inclusion.
> >>>>>
> >>>>> [1] https://git.yoctoproject.org/linux-yocto-dev/commit/?id=f36faf0f9f8d8f5b4c43a68e5c6bd83a62253140
> >>>>> ---
> >>>>> Changes in v2:
> >>>>> - Corrected author id
> >>>>> - Link to v1: https://lore.kernel.org/r/20250523-binrep-v1-1-c3a446518847@oss.qualcomm.com
> >>>>> ---
> >>>>>    drivers/gpu/drm/msm/registers/gen_header.py | 8 +++++---
> >>>>>    1 file changed, 5 insertions(+), 3 deletions(-)
> >>>>>
> >>>>
> >>>> Acked-by: Dmitry Baryshkov <dmitry.baryshkov@....qualcomm.com>
> >>>>
> >>>> Rob, WDYT?
> >>>
> >>> I'm revisiting this one, in the context of trying to re-sync
> >>> gen_header.py with mesa.. but it is only changing the contents of
> >>> comments, so it's not quite clear to me how this ends up mattering for
> >>> binary reproducibility.
> >>
> >> The reason it matters is that for Yocto, the generated header file is
> >> identified as a file that needs to be installed into the sysroot.  All
> >> files going into the sysroot are checked to make sure they do not
> >> contain dates and/or paths to the build directory contained within.
> >> Since this is a generated header file that is included in the sysroot we
> >> needed to strip out the path and date.
> >>
> >> The idea for the reproducible builds are that the same files on a
> >> different a machine at a different time should produce 100% identical
> >> files.  Including paths and dates violates that tenet.
> >>
> >> Hope that helps explain why we needed this.  So long as the
> >> gen_header.py is being called to generate header files then we need to
> >> maintain the reproducible aspect.
> >>
> >
> > My plan is (was?) to just replace the entire comment header with simply:
> >
> >    /* Autogenerated file, DO NOT EDIT manually! */
> >
> > That said, I'm not entirely sure why these files should get installed
> > into the sysroot?  I'm not super hands-on familiar with Yocto, so
> > maybe there is a good reason.. but if there is, maybe the plan to
> > remove the license/etc from the comment header isn't such a good idea
> > after all?
>
> The generated header files would be part of a linux-headers package that
> would be needed to build other packages as part of the distro.  And so
> the header files are all checked against the rules.  A linux-headers
> type package is common for distros to have available.
>

These headers should only be used to build the kernel, they are not in
include/uapi and as such should not be used for building any other
userspace package (or out of tree kernel module, for that matter).

BR,
-R

>
> > BR,
> > -R
> >
> >>
> >>> That said, since the generated files are no longer checked in to mesa
> >>> or the kernel, we could probably just drop all of this if it mattered.
> >>>
> >>> BR,
> >>> -R
> >>
> >> --
> >> Ryan Eatmon                reatmon@...com
> >> -----------------------------------------
> >> Texas Instruments, Inc.  -  LCPD  -  MGTS
> >>
> >>
>
> --
> Ryan Eatmon                reatmon@...com
> -----------------------------------------
> Texas Instruments, Inc.  -  LCPD  -  MGTS
>

Powered by blists - more mailing lists