Re: [PATCH 09/17] drm/radeon: use common fence implementation for fences

From: Maarten Lankhorst <maarten.lankhorst@canonical.com>
To: "Christian König" <christian.koenig@amd.com>,
	"Christian König" <deathsimple@vodafone.de>,
	"Dave Airlie" <airlied@gmail.com>,
	"Thomas Hellstrom" <thellstrom@vmware.com>,
	nouveau <nouveau@lists.freedesktop.org>,
	LKML <linux-kernel@vger.kernel.org>,
	dri-devel <dri-devel@lists.freedesktop.org>,
	"Ben Skeggs" <bskeggs@redhat.com>,
	"Deucher, Alexander" <alexander.deucher@amd.com>
Subject: Re: [PATCH 09/17] drm/radeon: use common fence implementation for fences
Date: Tue, 22 Jul 2014 16:47:16 +0200	[thread overview]
Message-ID: <53CE7974.2010209@canonical.com> (raw)
In-Reply-To: <53CE77B4.6020801@amd.com>

op 22-07-14 16:39, Christian König schreef:
> Am 22.07.2014 16:27, schrieb Maarten Lankhorst:
>> op 22-07-14 16:24, Christian König schreef:
>>>> No, you really shouldn't be doing much in the check anyway, it's meant to be a lightweight check. If you're not ready yet because of a lockup simply return not signaled yet.
>>> It's not only the lockup case from radeon I have in mind here. For userspace queues it might be necessary to call copy_from_user to figure out if a fence is signaled or not.
>>>
>>> Returning false all the time is probably not a good idea either.
>> Having userspace implement a fence sounds like an awful idea, why would you want to do that?
>
> Marketing moves in mysterious ways. Don't ask me, but that the direction it currently moves with userspace queues and IOMMU etc...
>
>> A fence could be exported to userspace, but that would only mean it can wait for it to be signaled with an interface like poll..
>
> Yeah agree totally, but the point for the fence interface is that I can't predict what's necessary to check if a fence is signaled or not on future hardware.
>
> For the currently available radeon hardware I can say that reading a value from a kernel page is pretty much all you need. But for older hardware that was reading from a register which might become very tricky if the hardware is power off or currently inside a reset cycle.
>
> Because off this I would avoid any such interface if it's not absolutely required by some use case, and currently I don't see this requirement because the functionality you want to archive could be implemented without this.
Oh? I've already done that in radeon_fence, there is no way enable_signaling will fiddle with hardware registers during a reset cycle.
I've also made sure that __radeon_fence_is_signaled grabs exclusive_lock in read mode before touching any hw state.

Older hardware also doesn't implement optimus, so I think power off is not much of a worry for them, if you could point me at the checking done for that I could make sure that this is the case.

~Maarten