On Tue, Oct 9, 2018 at 8:02 AM Benjamin Herrenschmidt
<b...@kernel.crashing.org> wrote:
>
> On Wed, 2018-10-03 at 16:10 -0700, Alexander Duyck wrote:
> > > -        * Because 32-bit DMA masks are so common we expect every 
> > > architecture
> > > -        * to be able to satisfy them - either by not supporting more 
> > > physical
> > > -        * memory, or by providing a ZONE_DMA32.  If neither is the case, 
> > > the
> > > -        * architecture needs to use an IOMMU instead of the direct 
> > > mapping.
> > > -        */
> > > -       if (mask < phys_to_dma(dev, DMA_BIT_MASK(32)))
> > > +       u64 min_mask;
> > > +
> > > +       if (IS_ENABLED(CONFIG_ZONE_DMA))
> > > +               min_mask = DMA_BIT_MASK(ARCH_ZONE_DMA_BITS);
> > > +       else
> > > +               min_mask = DMA_BIT_MASK(32);
> > > +
> > > +       min_mask = min_t(u64, min_mask, (max_pfn - 1) << PAGE_SHIFT);
> > > +
> > > +       if (mask >= phys_to_dma(dev, min_mask))
> > >                  return 0;
> > > -#endif
> > >          return 1;
> > >   }
> >
> > So I believe I have run into the same issue that Guenter reported. On
> > an x86_64 system w/ Intel IOMMU. I wasn't able to complete boot and
> > all probe attempts for various devices were failing with -EIO errors.
> >
> > I believe the last mask check should be "if (mask < phys_to_dma(dev,
> > min_mask))" not a ">=" check.
>
> Right, that test is backwards. I needed to change it here too (powermac
> with the rest of the powerpc series).
>
> Cheers,
> Ben.
>
>

Hi, I'm working on a MIPS64 soc with PCIe root complex on it, and it
appears that this series of patches are causing all PCI drivers that
request 64bit mask to fail with -5.
It's broken in 4.19. However, I just checked, it working on master.
We may need to backport a couple of patches to 4.19. I'm not sure
though which patches should be backported as there were at least 10
patches resolving this dma_direct area recently.
Christoph, Robin.
Can we ask Greg to backport all these changes ? What do you think ?

Thanks,
Ramon.

Reply via email to