Re: [PATCH v2] dma-mapping: treat dev->bus_dma_mask as a DMA limit

[Nathan Chancellor]

On Thu, Nov 21, 2019 at 10:26:44AM +0100, Nicolas Saenz Julienne wrote:
> Using a mask to represent bus DMA constraints has a set of limitations.
> The biggest one being it can only hold a power of two (minus one). The
> DMA mapping code is already aware of this and treats dev->bus_dma_mask
> as a limit. This quirk is already used by some architectures although
> still rare.
> 
> With the introduction of the Raspberry Pi 4 we've found a new contender
> for the use of bus DMA limits, as its PCIe bus can only address the
> lower 3GB of memory (of a total of 4GB). This is impossible to represent
> with a mask. To make things worse the device-tree code rounds non power
> of two bus DMA limits to the next power of two, which is unacceptable in
> this case.
> 
> In the light of this, rename dev->bus_dma_mask to dev->bus_dma_limit all
> over the tree and treat it as such. Note that dev->bus_dma_limit should
> contain the higher accesible DMA address.
> 
> Signed-off-by: Nicolas Saenz Julienne <nsaenzjulienne@xxxxxxx>
<snip>
> diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> index 041066f3ec99..0cc702a70a96 100644
> --- a/drivers/iommu/dma-iommu.c
> +++ b/drivers/iommu/dma-iommu.c
> @@ -421,8 +421,7 @@ static dma_addr_t iommu_dma_alloc_iova(struct iommu_domain *domain,
>  	if (iova_len < (1 << (IOVA_RANGE_CACHE_MAX_SIZE - 1)))
>  		iova_len = roundup_pow_of_two(iova_len);
>  
> -	if (dev->bus_dma_mask)
> -		dma_limit &= dev->bus_dma_mask;
> +	dma_limit = min_not_zero(dma_limit, dev->bus_dma_limit);
>  
>  	if (domain->geometry.force_aperture)
>  		dma_limit = min(dma_limit, domain->geometry.aperture_end);

Just as an FYI, this introduces a warning on arm32 allyesconfig for me:

In file included from ../include/linux/list.h:9,
                 from ../include/linux/kobject.h:19,
                 from ../include/linux/of.h:17,
                 from ../include/linux/irqdomain.h:35,
                 from ../include/linux/acpi.h:13,
                 from ../include/linux/acpi_iort.h:10,
                 from ../drivers/iommu/dma-iommu.c:11:
../drivers/iommu/dma-iommu.c: In function 'iommu_dma_alloc_iova':
../include/linux/kernel.h:851:29: warning: comparison of distinct pointer types lacks a cast
  851 |   (!!(sizeof((typeof(x) *)1 == (typeof(y) *)1)))
      |                             ^~
../include/linux/kernel.h:865:4: note: in expansion of macro '__typecheck'
  865 |   (__typecheck(x, y) && __no_side_effects(x, y))
      |    ^~~~~~~~~~~
../include/linux/kernel.h:875:24: note: in expansion of macro '__safe_cmp'
  875 |  __builtin_choose_expr(__safe_cmp(x, y), \
      |                        ^~~~~~~~~~
../include/linux/kernel.h:884:19: note: in expansion of macro '__careful_cmp'
  884 | #define min(x, y) __careful_cmp(x, y, <)
      |                   ^~~~~~~~~~~~~
../include/linux/kernel.h:917:39: note: in expansion of macro 'min'
  917 |  __x == 0 ? __y : ((__y == 0) ? __x : min(__x, __y)); })
      |                                       ^~~
../drivers/iommu/dma-iommu.c:424:14: note: in expansion of macro 'min_not_zero'
  424 |  dma_limit = min_not_zero(dma_limit, dev->bus_dma_limit);
      |              ^~~~~~~~~~~~

Cheers,
Nathan