[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <916a5783c2111c601d69a3665d4fee9c51239158.camel@intel.com>
Date: Thu, 20 Dec 2018 08:50:34 +0000
From: "Yang, Bin" <bin.yang@...el.com>
To: "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"Saarinen, Jani" <jani.saarinen@...el.com>,
"Wang, Zhi A" <zhi.a.wang@...el.com>,
"intel-gfx@...ts.freedesktop.org" <intel-gfx@...ts.freedesktop.org>,
"zhenyuw@...ux.intel.com" <zhenyuw@...ux.intel.com>,
"daniel@...ll.ch" <daniel@...ll.ch>,
"dri-devel@...ts.freedesktop.org" <dri-devel@...ts.freedesktop.org>,
"Du, Alek" <alek.du@...el.com>,
"ping.a.gao@...el.com" <ping.a.gao@...el.com>,
"airlied@...ux.ie" <airlied@...ux.ie>,
"joonas.lahtinen@...ux.intel.com" <joonas.lahtinen@...ux.intel.com>,
"chris@...is-wilson.co.uk" <chris@...is-wilson.co.uk>,
"jani.nikula@...ux.intel.com" <jani.nikula@...ux.intel.com>,
"Vivi, Rodrigo" <rodrigo.vivi@...el.com>
Subject: Re: [PATCH] drm/i915: Fix i915_gem_wait_for_idle oops due to
active_requests check
On Thu, 2018-12-20 at 08:35 +0000, Chris Wilson wrote:
> Quoting Bin Yang (2018-12-20 08:01:35)
> > Normally, i915_request_alloc() and i915_request_add() will be called
> > in sequence with drm.struct_mutex locked. But in
> > intel_vgpu_create_workload(), it will pre-allocate the request and
> > call i915_request_add() in the workload thread for performance
> > optimization. The above issue will be triggered.
>
> That's your bug. It's not normally, it's a strict requirement that the
> struct_mutex (request generation mutex) be held over the course of
> generating the request.
> -Chris
This code is introduced by below patch. Add original patch owners to
discuss this issue.
commit d0302e74003bf1f0fc41c06948b745204c4704ea
Author: Ping Gao <ping.a.gao@...el.com>
Date: Thu Jun 29 12:22:43 2017 +0800
drm/i915/gvt: Audit and shadow workload during ELSP writing
Let the workload audit and shadow ahead of vGPU scheduling, that
will eliminate GPU idle time and improve performance for multi-VM.
The performance of Heaven running simultaneously in 3VMs has
improved 20% after this patch.
v2:Remove condition current->vgpu==vgpu when shadow during ELSP
writing.
Signed-off-by: Ping Gao <ping.a.gao@...el.com>
Reviewed-by: Zhi Wang <zhi.a.wang@...el.com>
Signed-off-by: Zhenyu Wang <zhenyuw@...ux.intel.com>
Powered by blists - more mailing lists