[PATCH] drm/syncobj: return meaningful value to user space

[Chris Wilson]

Quoting Chunming Zhou (2019-07-18 14:15:49)
> 
> 在 2019/7/18 21:10, Chris Wilson 写道:
> > Quoting Chunming Zhou (2019-07-18 14:04:09)
> >> 在 2019/7/18 19:18, Chris Wilson 写道:
> >>> Quoting Chunming Zhou (2019-07-18 12:13:39)
> >>>> if WAIT_FOR_SUBMIT isn't set and in the meanwhile no underlying fence on syncobj,
> >>>> then return non-block error code to user sapce.
> >>> Block device required?
> >> Yes, if WAIT_FOR_SUBMIT is set, then that will block device.
> > No, the error message is that it requires a "block device", not that it
> > will block the device, which is EWOULDBLOCK.
> 
> OK, I got your mean.
> 
> Any other possile value suggestted?

ENXIO -- for the non-existent "address" along the timeline.
EOPNOTSUPP -- also makes sense, but we've started to use that for
interface not supported, so would be a bit inconsistent across drm.

Or ENOTBLK, being very clear in the commit message how it doesn't reflect
the original meaning (as would be given by strerror()) and why the
seemingly more natural EWOULDBLOCK doesn't work for drm in this case,
and what use case that needs to distinguish this particular case (i.e.
why EINVAL isn't good enough).
-Chris