Re: [PATCH RFC] xen/swiotlb: avoid arch_sync_dma_* on per-device DMA memory

[Stefano Stabellini <sstabellini@kernel.org>]

[-- Attachment #1: Type: text/plain, Size: 2336 bytes --]

On Fri, 24 Apr 2026, Peng Fan wrote:
> Hi Juergen,
> 
> On Mon, Apr 20, 2026 at 02:19:34PM +0200, Jürgen Groß wrote:
> >On 15.04.26 17:08, Peng Fan (OSS) wrote:
> >> From: Peng Fan <peng.fan@nxp.com>
> >> 
> >> On ARM64, arch_sync_dma_for_{cpu,device}() assumes that the
> >> physical address passed in refers to normal RAM that is part of the
> >> kernel linear(direct) mapping, as it unconditionally derives a CPU
> >> virtual address via phys_to_virt().
> >> 
> >> With Xen swiotlb, devices may use per-device coherent DMA memory,
> >> such as reserved-memory regions described by 'shared-dma-pool',
> >> which are assigned to dev->dma_mem. These regions may be marked
> >> no-map in DT and therefore are not part of the kernel linear map.
> >> In such cases, pfn_valid() still returns true, but phys_to_virt()
> >> is not valid and cache maintenance via arch_sync_dma_* will fault.
> >> 
> >> Prevent this by excluding devices with a private DMA memory pool
> >> (dev->dma_mem) from the arch_sync_dma_* fast path, and always
> >> fall back to xen_dma_sync_* for those devices to avoid invalid
> >> phys_to_virt() conversions for no-map DMA memory while preserving the
> >> existing fast path for normal, linear-mapped RAM.
> >> 
> >> Signed-off-by: Peng Fan <peng.fan@nxp.com>
> >> ---
> >>   drivers/xen/swiotlb-xen.c | 17 +++++++++++++----
> >>   1 file changed, 13 insertions(+), 4 deletions(-)
> >> 
> >> diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c
> >> index 2cbf2b588f5b20cfbf9e83a8339dc22092c9559a..b1445df99d9a8f1d18a83b8c413bada6e5579209 100644
> >> --- a/drivers/xen/swiotlb-xen.c
> >> +++ b/drivers/xen/swiotlb-xen.c
> >> @@ -195,6 +195,11 @@ xen_swiotlb_free_coherent(struct device *dev, size_t size, void *vaddr,
> >>   }
> >>   #endif /* CONFIG_X86 */
> >> +static inline bool dev_has_private_dma_pool(struct device *dev)
> >> +{
> >> +	return dev && dev->dma_mem;
> >> +}
> >> +
> >
> >I don't think this will compile on x86.
> 
> My bad, I only tried the build for ARM64. I use below changes, if it looks
> good.
> 
> #ifdef CONFIG_ARM64
> static inline bool dev_has_private_dma_pool(struct device *dev)
> {
> 	return dev && dev->dma_mem;
> }
> #else
> static inline bool dev_has_private_dma_pool(struct device *dev)
> {
> 	return false;
> }
> #endif

Please use CONFIG_DMA_DECLARE_COHERENT