On Fri, 2026-08-28 at 01:38 +0200, Andi Shyti wrote:
> Hi Krzysztof,
>
> ...
>
> > @@ -400,11 +400,10 @@ static int intel_mmio_bases_check(void *arg)
> > u32 base = info->mmio_bases[j].base;
> >
> > if (ver >= prev) {
> > - pr_err("%s(%s, class:%d, instance:%d): mmio
> > base for graphics ver %u is before the one for ver %u\n",
> > - __func__,
> > - intel_engine_class_repr(info->class),
> > - info->class, info->instance,
> > - prev, ver);
> > + drm_err(NULL,
> > + "%s(%s, class:%d, instance:%d): mmio
> > base for graphics ver %u is before the one for ver %u\n",
> > + __func__,
> > intel_engine_class_repr(info->class),
> > + info->class, info->instance, prev, ver);
>
> yah! drm_err(NULL, ...) is quite pointless. If there is no way
> for retrieving the drm device, then let's just stick to pr_*,
> this is their role.
While I said in one of my comments to 3/5 that I'm not sure if that makes
sense, maybe it does. With drm_*(NULL, ...), log messages are still
tagged with [drm] at least. So as a minimum, in line with my comment to
1/5, I think commit message of 1/5 that introduces only drm_*(NULL, ...)
could tell us that we want that [drm] tag even without context, unless we
don't want it, then prefer to keep pr_*().
Thanks,
Janusz
>
> Thanks,
> Andi
>
> > return -EINVAL;
> > }
