[Bug 94125] [SNA HSW bisected] git commit "sna: Skip sync if we have an idle coherent bo" causes corruption

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=94125

--- Comment #6 from Chris Wilson <chris at chris-wilson.co.uk> ---
(In reply to Matti Hämäläinen from comment #5)
> Pulled and now testing with added asserts. The issue is not presenting (with
> 1h of testing), nor am I hitting the asserts - so it is working, but that
> kinda raises the question whether I _should_ hit the asserts at some point?
> E.g. how to hit the situation the original changeset was meant to address?

The original regressing patch was a micro-optimisation to avoid a syscall and 
walking a potentially long list of requests in the kernel (trying to keep that
list alive to reduce future work).

Any time we read back from a pixmap, I expect (at least the second time) to hit
the short-circuit so given a workload where you saw a failure, I would expect
it still be taking the short-circuit, just less often (having avoided the bug).

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20160213/fb51b9d4/attachment-0001.html>