Re: [PATCH v5 1/3] drm/shmem: add support for per object caching flags.

From: "Thomas Hellström (VMware)" <thomas_os@shipmail.org>
To: Gerd Hoffmann <kraxel@redhat.com>
Cc: Guillaume.Gardet@arm.com, David Airlie <airlied@linux.ie>,
	open list <linux-kernel@vger.kernel.org>,
	dri-devel@lists.freedesktop.org, gurchetansingh@chromium.org,
	yuq825@gmail.com, tzimmermann@suse.de
Subject: Re: [PATCH v5 1/3] drm/shmem: add support for per object caching flags.
Date: Thu, 27 Feb 2020 14:49:25 +0100	[thread overview]
Message-ID: <b6f21041-d114-96c5-8f86-13eafd102c5d@shipmail.org> (raw)
In-Reply-To: <78eb099e-020f-91d1-672e-15176bf12cd4@shipmail.org>

On 2/27/20 2:44 PM, Thomas Hellström (VMware) wrote:
> Hi,
>
> On 2/27/20 2:21 PM, Gerd Hoffmann wrote:
>>    Hi,
>>
>>>> So I'd like to push patches 1+2 to -fixes and sort everything else 
>>>> later
>>>> in -next.  OK?
>>> OK with me.
>> Done.
>>
>>>> [ context: why shmem helpers use pgprot_writecombine + 
>>>> pgprot_decrypted?
>>>>             we get conflicting mappings because of that, linear kernel
>>>>             map vs. gem object vmap/mmap ]
>>> Do we have any idea what drivers are actually using
>>> write-combine and decrypted?
>> drivers/gpu/drm# find -name Kconfig* -print | xargs grep -l 
>> DRM_GEM_SHMEM_HELPER
>> ./lima/Kconfig
>> ./tiny/Kconfig
>> ./cirrus/Kconfig
>> ./Kconfig
>> ./panfrost/Kconfig
>> ./udl/Kconfig
>> ./v3d/Kconfig
>> ./virtio/Kconfig
>>
>> virtio needs cached.
>> cirrus+udl should be ok with cached too.
>>
>> Not clue about the others (lima, tiny, panfrost, v3d).  Maybe they use
>> write-combine just because this is what they got by default from
>> drm_gem_mmap_obj().  Maybe they actually need that.  Trying to Cc:
>> maintainters (and drop stable@).
>>
>> On decrypted: I guess that is only relevant for virtual machines, i.e.
>> virtio-gpu and cirrus?
>>
>> virtio-gpu needs it, otherwise the host can't show the virtual display.
>> cirrus bounces everything via blits to vram, so it should be ok without
>> decrypted.  I guess that implies we should make decrypted configurable.
>
> Decrypted here is clearly incorrect and violates the SEV spec, 
> regardless of a config option.
> The only correct way is currently to use dma_alloc_coherent() and 
> mmap_coherent() to allocate decrypted memory and then use the 
> pgprot_decrypted flag.
>
> Since the same page is aliased with two physical addresses (one 
> encrypted and one decrypted) switching memory from one to the other 
> needs extensive handling even to flush stale vmaps...
>
> So if virtio-gpu needs it for some bos, it should move away from shmem 
> for those bos.

(But this is of course up to the virtio-gpu and drm maintainers), but 
IMO, again, pgprot_decrypted() shouldn't be part of generic helpers.

/Thomas

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/dri-devel