Re: [PATCH] xen/swiotlb: add alignment check for dma buffers

[Jan Beulich <jbeulich@suse.com>]

On 13.09.2024 16:56, Juergen Gross wrote:
> --- a/drivers/xen/swiotlb-xen.c
> +++ b/drivers/xen/swiotlb-xen.c
> @@ -78,9 +78,15 @@ static inline int range_straddles_page_boundary(phys_addr_t p, size_t size)
>  {
>  	unsigned long next_bfn, xen_pfn = XEN_PFN_DOWN(p);
>  	unsigned int i, nr_pages = XEN_PFN_UP(xen_offset_in_page(p) + size);
> +	unsigned int order = get_order(size);
>  
>  	next_bfn = pfn_to_bfn(xen_pfn);
>  
> +	/* If buffer is physically aligned, ensure DMA alignment. */
> +	if (IS_ALIGNED(p, 1UL << (order + PAGE_SHIFT)) &&

Why this check? xen_swiotlb_alloc_coherent() guarantees it, while
xen_swiotlb_free_coherent() only checks properties of the original
allocation. And for xen_swiotlb_map_page() this looks actively
wrong to me, in case that function was called with offset non-zero.

Unrelated to this, but in related code: xen_swiotlb_alloc_coherent()
can't validly use XEN_PAGE_SHIFT, can it (in the way it does at
least)? If XEN_PAGE_SHIFT < PAGE_SHIFT, and with order calculated
from the latter, the used size will be too small afaict.

Jan