[Intel-gfx] [PATCH v8 03/20] drm/i915: Modify error handler for per engine hang recovery

[Chris Wilson]

Quoting Michel Thierry (2017-06-02 21:38:56)
> One thing I forgot to ask, what should I do with the error/reset 
> uevents? As it is, we will only tell userspace in case of global reset.

My first thought is that we don't care to send the uevent for anything
less than a global reset. The original use case was for abrtd style
automatic bug reporters, but afaik no one is harvesting those error
reports any more. I think we could drop it and no one would be the
wiser. (It's not an ABI break if no one notices.) On the other hand, we
do use the uevent to catch bugs (unexpected hangs) in igt. For igt, I
could just disable per-engine resets for the duration of the those tests.
Those tests are not about testing the reset and so missing coverage of
per-engine reset is not an issue.

Since this is an i915-specific feature, I think the actual listeners to
this uevent are going to be rare, tending to 0. I hope.
-Chris