[PATCH] drm/amd/powerplay: avoid DPM reenable process on Navi1x ASICs

[Quan, Evan]

> -----Original Message-----
> From: Alex Deucher <alexdeucher at gmail.com>
> Sent: Tuesday, November 12, 2019 1:34 AM
> To: Quan, Evan <Evan.Quan at amd.com>
> Cc: amd-gfx list <amd-gfx at lists.freedesktop.org>
> Subject: Re: [PATCH] drm/amd/powerplay: avoid DPM reenable process on
> Navi1x ASICs
> 
> On Mon, Nov 11, 2019 at 4:25 AM Evan Quan <evan.quan at amd.com> wrote:
> >
> > Otherwise, without RLC reinitialization, the DPM reenablement will
> > fail. That affects the custom pptable uploading.
> >
> > Change-Id: I6fe2ed5ce23f2a5b66f371c0b6d1f924837e5af6
> > Signed-off-by: Evan Quan <evan.quan at amd.com>
> > ---
> >  drivers/gpu/drm/amd/powerplay/amdgpu_smu.c    | 32 +++++++++++++++--
> --
> >  .../gpu/drm/amd/powerplay/inc/amdgpu_smu.h    |  1 +
> >  2 files changed, 26 insertions(+), 7 deletions(-)
> >
> > diff --git a/drivers/gpu/drm/amd/powerplay/amdgpu_smu.c
> > b/drivers/gpu/drm/amd/powerplay/amdgpu_smu.c
> > index 76a4154b3be2..a4d67b30fd72 100644
> > --- a/drivers/gpu/drm/amd/powerplay/amdgpu_smu.c
> > +++ b/drivers/gpu/drm/amd/powerplay/amdgpu_smu.c
> > @@ -1293,10 +1293,25 @@ static int smu_hw_fini(void *handle)
> >                 return ret;
> >         }
> >
> > -       ret = smu_stop_dpms(smu);
> > -       if (ret) {
> > -               pr_warn("Fail to stop Dpms!\n");
> > -               return ret;
> > +       /*
> > +        * For custom pptable uploading, skip the DPM features
> > +        * disable process on Navi1x ASICs.
> > +        *   - As the gfx related features are under control of
> > +        *     RLC on those ASICs. RLC reinitialization will be
> > +        *     needed to reenable them. That will cost much more
> > +        *     efforts.
> > +        *
> > +        *   - SMU firmware can handle the DPM reenablement
> > +        *     properly.
> > +        */
> > +       if (!smu->uploading_custom_pp_table ||
> > +           !((adev->asic_type >= CHIP_NAVI10) &&
> > +             (adev->asic_type <= CHIP_NAVI12))) {
> > +               ret = smu_stop_dpms(smu);
> > +               if (ret) {
> > +                       pr_warn("Fail to stop Dpms!\n");
> > +                       return ret;
> > +               }
> >         }
> >
> >         kfree(table_context->driver_pptable);
> > @@ -1324,13 +1339,16 @@ int smu_reset(struct smu_context *smu)
> >         struct amdgpu_device *adev = smu->adev;
> >         int ret = 0;
> >
> > +       smu->uploading_custom_pp_table = true;
> > +
> 
> Do we need to differentiate between reloading for a pptable update and
> reloading for a gpu reset or suspend/resume or is that already handled?
[Quan, Evan] That was considered(and as I verified suspend/resume was fine).  For gpu reset or suspend/resume, the rlc/gfx will be reinitialized.
So, under those cases, it's OK to perform the dpms disablement.
> Shouldn't we be setting/clearing uploading_custom_pp_table in
> smu_sys_set_pp_table() around the call to smu_reset()?
[Quan, Evan] That's a good point. Will update this in V2.
> 
> Alex
> 
> >         ret = smu_hw_fini(adev);
> >         if (ret)
> > -               return ret;
> > +               goto out;
> >
> >         ret = smu_hw_init(adev);
> > -       if (ret)
> > -               return ret;
> > +
> > +out:
> > +       smu->uploading_custom_pp_table = false;
> >
> >         return ret;
> >  }
> > diff --git a/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h
> > b/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h
> > index 8120e7587585..215841f5fb93 100644
> > --- a/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h
> > +++ b/drivers/gpu/drm/amd/powerplay/inc/amdgpu_smu.h
> > @@ -391,6 +391,7 @@ struct smu_context
> >
> >         uint32_t smc_if_version;
> >
> > +       bool uploading_custom_pp_table;
> >  };
> >
> >  struct i2c_adapter;
> > --
> > 2.24.0
> >
> > _______________________________________________
> > amd-gfx mailing list
> > amd-gfx at lists.freedesktop.org
> > https://lists.freedesktop.org/mailman/listinfo/amd-gfx