[Mesa-dev] [PATCH v2 8/8] panfrost: Add backend targeting the DRM driver

[Rob Herring]

On Fri, Mar 8, 2019 at 3:59 PM Alyssa Rosenzweig <alyssa at rosenzweig.io> wrote:
>
> > +/**
> > + * struct drm_panfrost_wait_bo - ioctl argument for waiting for
> > + * completion of the last DRM_PANFROST_SUBMIT_CL on a BO.
>
> Nit: Should be plain DRM_PANFROST_SUBMIT, there is no CL for us.
>
> > +     __s64 timeout_ns;       /* absolute */
>
> Erm, why is this signed? Semantically, what does a negative timestamp
> mean? Seems suspect. The comment /* absolute */ seems to underscore that
> we really do want an unsigned value, perhaps ascribing a special meaning
> to 0/~0 for "nonblocking" and "block indefinitely" if needed. Of course,
> "(2^64)-1 ns" is essentially indefinite, so the latter need not be a
> special case.

Signed is convention and used internally in the kernel (ktime_t). I
checked that signed is correct with Arnd Bergmann who is leading the
2038 work.

> * It's 585 years, according to a back of the envelope calculation.
>   Panfrost will be obsolete many times over by the time that timeout
>   elapses ;)

I'm pretty sure it will be obsolete in only 240 years too. :)

Rob