Re: [PATCH 4/4] drm/msm/a6xx: Capture gmu log in devcoredump

[Akhil P Oommen]

On 11/23/2021 12:36 AM, Rob Clark wrote:
> On Mon, Nov 22, 2021 at 10:26 AM Rob Clark <robdclark@xxxxxxxxx> wrote:
> > On Thu, Nov 18, 2021 at 2:21 AM Akhil P Oommen <akhilpo@xxxxxxxxxxxxxx> wrote:
> > > Capture gmu log in coredump to enhance debugging.
> > >
> > > Signed-off-by: Akhil P Oommen <akhilpo@xxxxxxxxxxxxxx>
> > > ---
> > >
> > >   drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c | 41 +++++++++++++++++++++++++++++
> > >   drivers/gpu/drm/msm/adreno/adreno_gpu.c     |  2 +-
> > >   drivers/gpu/drm/msm/adreno/adreno_gpu.h     |  2 ++
> > >   3 files changed, 44 insertions(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
> > > index 7501849..9fa3fa6 100644
> > > --- a/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
> > > +++ b/drivers/gpu/drm/msm/adreno/a6xx_gpu_state.c
> > > @@ -42,6 +42,8 @@ struct a6xx_gpu_state {
> > >          struct a6xx_gpu_state_obj *cx_debugbus;
> > >          int nr_cx_debugbus;
> > >
> > > +       struct msm_gpu_state_bo *gmu_log;
> > > +
> > >          struct list_head objs;
> > >   };
> > >
> > > @@ -800,6 +802,30 @@ static void a6xx_get_gmu_registers(struct msm_gpu *gpu,
> > >                  &a6xx_state->gmu_registers[2], false);
> > >   }
> > >
> > > +static void a6xx_get_gmu_log(struct msm_gpu *gpu,
> > > +               struct a6xx_gpu_state *a6xx_state)
> > > +{
> > > +       struct adreno_gpu *adreno_gpu = to_adreno_gpu(gpu);
> > > +       struct a6xx_gpu *a6xx_gpu = to_a6xx_gpu(adreno_gpu);
> > > +       struct a6xx_gmu *gmu = &a6xx_gpu->gmu;
> > > +       struct msm_gpu_state_bo *gmu_log;
> > > +
> > > +       gmu_log = state_kcalloc(a6xx_state,
> > > +               1, sizeof(*a6xx_state->gmu_log));
> > > +       if (!gmu_log)
> > > +               return;
> > > +
> > > +       gmu_log->iova = gmu->log.iova;
> > > +       gmu_log->size = gmu->log.size;
> > > +       gmu_log->data = kvzalloc(gmu_log->size, GFP_KERNEL);
> > > +       if (!gmu_log->data)
> > > +               return;
> > > +
> > > +       memcpy(gmu_log->data, gmu->log.virt, gmu->log.size);
> > > +
> > > +       a6xx_state->gmu_log = gmu_log;
> > > +}
> > > +
> > >   #define A6XX_GBIF_REGLIST_SIZE   1
> > >   static void a6xx_get_registers(struct msm_gpu *gpu,
> > >                  struct a6xx_gpu_state *a6xx_state,
> > > @@ -937,6 +963,8 @@ struct msm_gpu_state *a6xx_gpu_state_get(struct msm_gpu *gpu)
> > >
> > >          a6xx_get_gmu_registers(gpu, a6xx_state);
> > >
> > > +       a6xx_get_gmu_log(gpu, a6xx_state);
> > > +
> > >          /* If GX isn't on the rest of the data isn't going to be accessible */
> > >          if (!a6xx_gmu_gx_is_on(&a6xx_gpu->gmu))
> > >                  return &a6xx_state->base;
> > > @@ -978,6 +1006,9 @@ static void a6xx_gpu_state_destroy(struct kref *kref)
> > >          struct a6xx_gpu_state *a6xx_state = container_of(state,
> > >                          struct a6xx_gpu_state, base);
> > >
> > > +       if (a6xx_state->gmu_log && a6xx_state->gmu_log->data)
> > > +               kvfree(a6xx_state->gmu_log->data);
> > > +
> > >          list_for_each_entry_safe(obj, tmp, &a6xx_state->objs, node)
> > >                  kfree(obj);
> > >
> > > @@ -1191,6 +1222,16 @@ void a6xx_show(struct msm_gpu *gpu, struct msm_gpu_state *state,
> > >
> > >          adreno_show(gpu, state, p);
> > >
> > > +       drm_puts(p, "gmu-log:\n");
> > > +       if (a6xx_state->gmu_log) {
> > > +               struct msm_gpu_state_bo *gmu_log = a6xx_state->gmu_log;
> > > +
> > > +               drm_printf(p, "    iova: 0x%016llx\n", gmu_log->iova);
> > > +               drm_printf(p, "    size: %d\n", gmu_log->size);
> >
> > fwiw, that wants to be:
> >
> >   +               drm_printf(p, "    size: %zu\n", gmu_log->size);
> >
> > with that fixed, r-b
>
> Hmm, actually, I seem to be getting an empty log.. is special gmu fw,
> or non-fused device needed for this to work?
>
> BR,
> -R

No, there is no special fw. I tested this on 7c3 and it worked for me. 
a618/a630 has an old version of gmu firmware which is pretty different 
from the newer ones. Let me check.

-Akhil.

> > BR,
> > -R
> >
> > > +               adreno_show_object(p, &gmu_log->data, gmu_log->size,
> > > +                               &gmu_log->encoded);
> > > +       }
> > > +
> > >          drm_puts(p, "registers:\n");
> > >          for (i = 0; i < a6xx_state->nr_registers; i++) {
> > >                  struct a6xx_gpu_state_obj *obj = &a6xx_state->registers[i];
> > > diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> > > index 7486652..7d1ff20 100644
> > > --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> > > +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
> > > @@ -630,7 +630,7 @@ static char *adreno_gpu_ascii85_encode(u32 *src, size_t len)
> > >   }
> > >
> > >   /* len is expected to be in bytes */
> > > -static void adreno_show_object(struct drm_printer *p, void **ptr, int len,
> > > +void adreno_show_object(struct drm_printer *p, void **ptr, int len,
> > >                  bool *encoded)
> > >   {
> > >          if (!*ptr || !len)
> > > diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
> > > index 225c277..6762308 100644
> > > --- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
> > > +++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
> > > @@ -306,6 +306,8 @@ void adreno_gpu_state_destroy(struct msm_gpu_state *state);
> > >
> > >   int adreno_gpu_state_get(struct msm_gpu *gpu, struct msm_gpu_state *state);
> > >   int adreno_gpu_state_put(struct msm_gpu_state *state);
> > > +void adreno_show_object(struct drm_printer *p, void **ptr, int len,
> > > +               bool *encoded);
> > >
> > >   /*
> > >    * Common helper function to initialize the default address space for arm-smmu
> > > --
> > > QUALCOMM INDIA, on behalf of Qualcomm Innovation Center, Inc. is a member
> > > of Code Aurora Forum, hosted by The Linux Foundation.