[PATCH v10 06/24] iommu/dma: Factor out a iommu_dma_map_swiotlb helper
Leon Romanovsky
leon at kernel.org
Mon Apr 28 23:18:49 PDT 2025
On Tue, Apr 29, 2025 at 01:58:06PM +0800, Baolu Lu wrote:
> On 4/29/25 13:53, Leon Romanovsky wrote:
> > On Tue, Apr 29, 2025 at 12:58:18PM +0800, Baolu Lu wrote:
> > > On 4/28/25 17:22, Leon Romanovsky wrote:
> > > > From: Christoph Hellwig<hch at lst.de>
> > > >
> > > > Split the iommu logic from iommu_dma_map_page into a separate helper.
> > > > This not only keeps the code neatly separated, but will also allow for
> > > > reuse in another caller.
> > > >
> > > > Signed-off-by: Christoph Hellwig<hch at lst.de>
> > > > Tested-by: Jens Axboe<axboe at kernel.dk>
> > > > Reviewed-by: Luis Chamberlain<mcgrof at kernel.org>
> > > > Signed-off-by: Leon Romanovsky<leonro at nvidia.com>
> > >
> > > Reviewed-by: Lu Baolu <baolu.lu at linux.intel.com>
> > >
> > > with a nit below ...
> > >
> > > > ---
> > > > drivers/iommu/dma-iommu.c | 73 ++++++++++++++++++++++-----------------
> > > > 1 file changed, 41 insertions(+), 32 deletions(-)
> > > >
> > > > diff --git a/drivers/iommu/dma-iommu.c b/drivers/iommu/dma-iommu.c
> > > > index d3211a8d755e..d7684024c439 100644
> > > > --- a/drivers/iommu/dma-iommu.c
> > > > +++ b/drivers/iommu/dma-iommu.c
> > > > @@ -1138,6 +1138,43 @@ void iommu_dma_sync_sg_for_device(struct device *dev, struct scatterlist *sgl,
> > > > arch_sync_dma_for_device(sg_phys(sg), sg->length, dir);
> > > > }
> > > > +static phys_addr_t iommu_dma_map_swiotlb(struct device *dev, phys_addr_t phys,
> > > > + size_t size, enum dma_data_direction dir, unsigned long attrs)
> > > > +{
> > > > + struct iommu_domain *domain = iommu_get_dma_domain(dev);
> > > > + struct iova_domain *iovad = &domain->iova_cookie->iovad;
> > > > +
> > > > + if (!is_swiotlb_active(dev)) {
> > > > + dev_warn_once(dev, "DMA bounce buffers are inactive, unable to map unaligned transaction.\n");
> > > > + return (phys_addr_t)DMA_MAPPING_ERROR;
> > > > + }
> > > > +
> > > > + trace_swiotlb_bounced(dev, phys, size);
> > > > +
> > > > + phys = swiotlb_tbl_map_single(dev, phys, size, iova_mask(iovad), dir,
> > > > + attrs);
> > > > +
> > > > + /*
> > > > + * Untrusted devices should not see padding areas with random leftover
> > > > + * kernel data, so zero the pre- and post-padding.
> > > > + * swiotlb_tbl_map_single() has initialized the bounce buffer proper to
> > > > + * the contents of the original memory buffer.
> > > > + */
> > > > + if (phys != (phys_addr_t)DMA_MAPPING_ERROR && dev_is_untrusted(dev)) {
> > > > + size_t start, virt = (size_t)phys_to_virt(phys);
> > > > +
> > > > + /* Pre-padding */
> > > > + start = iova_align_down(iovad, virt);
> > > > + memset((void *)start, 0, virt - start);
> > > > +
> > > > + /* Post-padding */
> > > > + start = virt + size;
> > > > + memset((void *)start, 0, iova_align(iovad, start) - start);
> > > > + }
> > > > +
> > > > + return phys;
> > > > +}
> > > > +
> > > > dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> > > > unsigned long offset, size_t size, enum dma_data_direction dir,
> > > > unsigned long attrs)
> > > > @@ -1151,42 +1188,14 @@ dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> > > > dma_addr_t iova, dma_mask = dma_get_mask(dev);
> > > > /*
> > > > - * If both the physical buffer start address and size are
> > > > - * page aligned, we don't need to use a bounce page.
> > > > + * If both the physical buffer start address and size are page aligned,
> > > > + * we don't need to use a bounce page.
> > > > */
> > > > if (dev_use_swiotlb(dev, size, dir) &&
> > > > iova_offset(iovad, phys | size)) {
> > > > - if (!is_swiotlb_active(dev)) {
> > >
> > > ... Is it better to move this check into the helper? Simply no-op if a
> > > bounce page is not needed:
> > >
> > > if (!dev_use_swiotlb(dev, size, dir) ||
> > > !iova_offset(iovad, phys | size))
> > > return phys;
> >
> > Am I missing something? iommu_dma_map_page() has more code after this
> > check, so it is not correct to return immediately:
> >
> > 1189 dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> > 1190 unsigned long offset, size_t size, enum dma_data_direction dir,
> > 1191 unsigned long attrs)
> > 1192 {
> >
> > <...>
> >
> > 1201 /*
> > 1202 * If both the physical buffer start address and size are page aligned,
> > 1203 * we don't need to use a bounce page.
> > 1204 */
> > 1205 if (dev_use_swiotlb(dev, size, dir) &&
> > 1206 iova_unaligned(iovad, phys, size)) {
> > 1207 phys = iommu_dma_map_swiotlb(dev, phys, size, dir, attrs);
> > 1208 if (phys == (phys_addr_t)DMA_MAPPING_ERROR)
> > 1209 return DMA_MAPPING_ERROR;
> > 1210 }
> > 1211
> > 1212 if (!coherent && !(attrs & DMA_ATTR_SKIP_CPU_SYNC))
> > 1213 arch_sync_dma_for_device(phys, size, dir);
> > 1214
> > 1215 iova = __iommu_dma_map(dev, phys, size, prot, dma_mask);
> > 1216 if (iova == DMA_MAPPING_ERROR)
> > 1217 swiotlb_tbl_unmap_single(dev, phys, size, dir, attrs);
> > 1218 return iova;
> > 1219 }
>
> static phys_addr_t iommu_dma_map_swiotlb(struct device *dev, phys_addr_t
> phys,
> size_t size, enum dma_data_direction dir, unsigned long attrs)
> {
> <...>
> /*
> * If both the physical buffer start address and size are page aligned,
> * we don't need to use a bounce page.
> */
> if (!dev_use_swiotlb(dev, size, dir) ||
> !iova_offset(iovad, phys | size))
> return phys;
> <...>
> }
>
> Then,
>
> dma_addr_t iommu_dma_map_page(struct device *dev, struct page *page,
> unsigned long offset, size_t size, enum dma_data_direction dir,
> unsigned long attrs)
> {
> <...>
> phys = iommu_dma_map_swiotlb(dev, phys, size, dir, attrs);
> if (phys == (phys_addr_t)DMA_MAPPING_ERROR)
> return DMA_MAPPING_ERROR;
> <...>
> }
Such change will cause to extra function call for everyone who doesn't
use SWIOTLB (RDMA, HMM e.t.c).
In addition, iommu_dma_map_swiotlb() is called through
dma_iova_link ->
iommu_dma_iova_link_swiotlb ->
iommu_dma_iova_bounce_and_link() ->
iommu_dma_map_swiotlb()
and dma_iova_link() has this "if (dev_use_swiotlb(dev, size, dir) && iova_unaligned(iovad, phys, size))"
very early at call stack.
So, in dma_iova_link() we will find ourselves with same check twice.
Thanks
>
> Thanks,
> baolu
>
More information about the Linux-nvme
mailing list