Re: [PATCH v2 4/6] iommu/vt-d: Remove 1:1 mappings from identity domain

[Baolu Lu]

On 12/11/23 11:58 AM, Tian, Kevin wrote:
> > > there are two cases above which mandates IDENTITY. Have you confirmed
> > > that those platforms support hardware passthrough? otherwise this
> > change
> > > is broken.
> > Those two cases should be hardware quirks for SoC-integrated devices. It
> > makes no reason that a quirk requires IOMMU passthrough translation, but
> > the hardware doesn't support it.
> >
> > If, unfortunately, those quirks turn out to be workarounds for a poorly
> > designed device driver, we should remove those quirks and request the
> > device driver to utilize the DMA ownership framework to achieve the same
> > functionality within the driver itself.
> if that is the case you should fix the drivers first before breaking them.
>
> But at a glance looks those two quirks are just fine.
>
> For Azalia sound device the problem is that BIOS enables a dedicated
> DMAR for it but allocates zero TLB entries to cause deadlock. This
> implies a hw passthrough mode otherwise it's still broken.

Yes. It's safe for Azalia sound device.

> For GFX it's a workaround added since day one. there is even still
> an option CONFIG_INTEL_IOMMU_BROKEN_GFX_WA available. But
> now its meaning is really disabling IOMMU instead of using identity.
>
> sounds like IDENTMAP_GFX can be fully removed now:
>
> #ifdef CONFIG_INTEL_IOMMU_BROKEN_GFX_WA
> 	dmar_map_gfx = 0;
> #endif

We should already remove the workaround for the 2.6 kernel. :-)

It's default "n". Therefore, if any gfx driver still needs this
workaround, there should already be a bug report.

> 	if (!dmar_map_gfx)
> 		iommu_identity_mapping |= IDENTMAP_GFX;

So with above cleaned up, we have no need to worry about drivers that
are not capable of handling remapped dma address any more.

Did I miss anything?

Best regards,
baolu