[igt-dev] [PATCH v2] tests/syncobj_eventfd: new test

[Simon Ser]

On Wednesday, July 12th, 2023 at 09:37, Lionel Landwerlin <lionel.g.landwerlin at intel.com> wrote:

> These tests look correct to me.

Thanks for having a look!

> Are you not planning on testing/supporting
> DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT and DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL ?

DRM_SYNCOBJ_WAIT_FLAGS_WAIT_FOR_SUBMIT is implied: the IOCTL will never
fail because a fence hasn't materialized. There is no use-case for this
flag: if user-space wants this behavior they can try to extract a
sync_file (which will fail if it hasn't materialized yet) and then poll
that sync_file.

DRM_SYNCOBJ_WAIT_FLAGS_WAIT_ALL doesn't make sense for this IOCTL
because only a single syncobj is passed in.

Simon