Hi Krzysztof,
On 2026-11-04 at 11:10:59 +0000, Krzysztof Karas wrote:
> Move away from generic pr prints to GT aware gt_* versions,
> which tag each message with device and the tile it came from.
>
> i915_perf.c now uses drm_* prints, as it reports on a device
> wide perf stream with no GT to attribute the message to.
>
> Ensure that 80 character line limit is respected wherever
> possible.
>
> Signed-off-by: Krzysztof Karas <[email protected]>
> Acked-by: Jani Nikula <[email protected]>
> ---
LGTM with one note:
<snip to relevant file>
> diff --git a/drivers/gpu/drm/i915/selftests/i915_vma.c
> b/drivers/gpu/drm/i915/selftests/i915_vma.c
> index fc068dab27f4..aa2527314ab0 100644
> --- a/drivers/gpu/drm/i915/selftests/i915_vma.c
> +++ b/drivers/gpu/drm/i915/selftests/i915_vma.c
> @@ -28,6 +28,8 @@
> #include "gem/i915_gem_internal.h"
> #include "gem/selftests/mock_context.h"
<snip to relevant function>
> goto out_object;
> }
> @@ -1080,7 +1143,8 @@ static int igt_vma_remapped_gtt(void *arg)
>
> val = ioread32(&map[offset /
> sizeof(*map)]);
> if (val != exp) {
> - pr_err("%s VMA write test
> failed, expected 0x%x, found 0x%x\n",
> + gt_err(to_gt(i915),
Note, not nit, but I think it's worth saying: technically this might run
on multi-gt platforms (so non-mock) where to_gt() might not be always
correct; however in this case the VMA comes from
i915_gem_object_ggtt_pin() which uses hardcoded to_gt()->ggtt anyway [1], so
this is functionally correct.
Again, not a blocker, you just can't assume mock device here so I'm
explicitly stating the due dilligence check for this.
[1]
https://elixir.bootlin.com/linux/v7.3.1/source/drivers/gpu/drm/i915/i915_gem.c#L909
All in all:
Reviewed-by: Krzysztof Niemiec <[email protected]>
Thanks
Krzysztof
