From 85fcb57c983f423180ba6ec5d0034242da05cc54 Mon Sep 17 00:00:00 2001 From: Juergen Gross Date: Mon, 10 Feb 2025 08:43:39 +0100 Subject: [PATCH 1/3] xen/swiotlb: relax alignment requirements When mapping a buffer for DMA via .map_page or .map_sg DMA operations, there is no need to check the machine frames to be aligned according to the mapped areas size. All what is needed in these cases is that the buffer is contiguous at machine level. So carve out the alignment check from range_straddles_page_boundary() and move it to a helper called by xen_swiotlb_alloc_coherent() and xen_swiotlb_free_coherent() directly. Fixes: 9f40ec84a797 ("xen/swiotlb: add alignment check for dma buffers") Reported-by: Jan Vejvalka Tested-by: Jan Vejvalka Signed-off-by: Juergen Gross Reviewed-by: Stefano Stabellini Signed-off-by: Juergen Gross --- drivers/xen/swiotlb-xen.c | 20 ++++++++++++-------- 1 file changed, 12 insertions(+), 8 deletions(-) diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c index a337edcf8faf..26c62e0d34e9 100644 --- a/drivers/xen/swiotlb-xen.c +++ b/drivers/xen/swiotlb-xen.c @@ -74,19 +74,21 @@ static inline phys_addr_t xen_dma_to_phys(struct device *dev, return xen_bus_to_phys(dev, dma_to_phys(dev, dma_addr)); } +static inline bool range_requires_alignment(phys_addr_t p, size_t size) +{ + phys_addr_t algn = 1ULL << (get_order(size) + PAGE_SHIFT); + phys_addr_t bus_addr = pfn_to_bfn(XEN_PFN_DOWN(p)) << XEN_PAGE_SHIFT; + + return IS_ALIGNED(p, algn) && !IS_ALIGNED(bus_addr, algn); +} + 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); - phys_addr_t algn = 1ULL << (get_order(size) + PAGE_SHIFT); next_bfn = pfn_to_bfn(xen_pfn); - /* If buffer is physically aligned, ensure DMA alignment. */ - if (IS_ALIGNED(p, algn) && - !IS_ALIGNED((phys_addr_t)next_bfn << XEN_PAGE_SHIFT, algn)) - return 1; - for (i = 1; i < nr_pages; i++) if (pfn_to_bfn(++xen_pfn) != ++next_bfn) return 1; @@ -156,7 +158,8 @@ xen_swiotlb_alloc_coherent(struct device *dev, size_t size, *dma_handle = xen_phys_to_dma(dev, phys); if (*dma_handle + size - 1 > dma_mask || - range_straddles_page_boundary(phys, size)) { + range_straddles_page_boundary(phys, size) || + range_requires_alignment(phys, size)) { if (xen_create_contiguous_region(phys, order, fls64(dma_mask), dma_handle) != 0) goto out_free_pages; @@ -182,7 +185,8 @@ xen_swiotlb_free_coherent(struct device *dev, size_t size, void *vaddr, size = ALIGN(size, XEN_PAGE_SIZE); if (WARN_ON_ONCE(dma_handle + size - 1 > dev->coherent_dma_mask) || - WARN_ON_ONCE(range_straddles_page_boundary(phys, size))) + WARN_ON_ONCE(range_straddles_page_boundary(phys, size) || + range_requires_alignment(phys, size))) return; if (TestClearPageXenRemapped(virt_to_page(vaddr))) From e93ec87286bd1fd30b7389e7a387cfb259f297e3 Mon Sep 17 00:00:00 2001 From: Juergen Gross Date: Tue, 11 Feb 2025 11:16:28 +0100 Subject: [PATCH 2/3] x86/xen: allow larger contiguous memory regions in PV guests Today a PV guest (including dom0) can create 2MB contiguous memory regions for DMA buffers at max. This has led to problems at least with the megaraid_sas driver, which wants to allocate a 2.3MB DMA buffer. The limiting factor is the frame array used to do the hypercall for making the memory contiguous, which has 512 entries and is just a static array in mmu_pv.c. In order to not waste memory for non-PV guests, put the initial frame array into .init.data section and dynamically allocate an array from the .init_after_bootmem hook of PV guests. In case a contiguous memory area larger than the initially supported 2MB is requested, allocate a larger buffer for the frame list. Note that such an allocation is tried only after memory management has been initialized properly, which is tested via a flag being set in the .init_after_bootmem hook. Fixes: 9f40ec84a797 ("xen/swiotlb: add alignment check for dma buffers") Signed-off-by: Juergen Gross Tested-by: Alan Robinson Reviewed-by: Jan Beulich Signed-off-by: Juergen Gross --- arch/x86/xen/mmu_pv.c | 71 +++++++++++++++++++++++++++++++++++++------ 1 file changed, 62 insertions(+), 9 deletions(-) diff --git a/arch/x86/xen/mmu_pv.c b/arch/x86/xen/mmu_pv.c index 2c70cd35e72c..d078de2c952b 100644 --- a/arch/x86/xen/mmu_pv.c +++ b/arch/x86/xen/mmu_pv.c @@ -111,6 +111,51 @@ static pud_t level3_user_vsyscall[PTRS_PER_PUD] __page_aligned_bss; */ static DEFINE_SPINLOCK(xen_reservation_lock); +/* Protected by xen_reservation_lock. */ +#define MIN_CONTIG_ORDER 9 /* 2MB */ +static unsigned int discontig_frames_order = MIN_CONTIG_ORDER; +static unsigned long discontig_frames_early[1UL << MIN_CONTIG_ORDER] __initdata; +static unsigned long *discontig_frames __refdata = discontig_frames_early; +static bool discontig_frames_dyn; + +static int alloc_discontig_frames(unsigned int order) +{ + unsigned long *new_array, *old_array; + unsigned int old_order; + unsigned long flags; + + BUG_ON(order < MIN_CONTIG_ORDER); + BUILD_BUG_ON(sizeof(discontig_frames_early) != PAGE_SIZE); + + new_array = (unsigned long *)__get_free_pages(GFP_KERNEL, + order - MIN_CONTIG_ORDER); + if (!new_array) + return -ENOMEM; + + spin_lock_irqsave(&xen_reservation_lock, flags); + + old_order = discontig_frames_order; + + if (order > discontig_frames_order || !discontig_frames_dyn) { + if (!discontig_frames_dyn) + old_array = NULL; + else + old_array = discontig_frames; + + discontig_frames = new_array; + discontig_frames_order = order; + discontig_frames_dyn = true; + } else { + old_array = new_array; + } + + spin_unlock_irqrestore(&xen_reservation_lock, flags); + + free_pages((unsigned long)old_array, old_order - MIN_CONTIG_ORDER); + + return 0; +} + /* * Note about cr3 (pagetable base) values: * @@ -814,6 +859,9 @@ static void __init xen_after_bootmem(void) SetPagePinned(virt_to_page(level3_user_vsyscall)); #endif xen_pgd_walk(&init_mm, xen_mark_pinned, FIXADDR_TOP); + + if (alloc_discontig_frames(MIN_CONTIG_ORDER)) + BUG(); } static void xen_unpin_page(struct mm_struct *mm, struct page *page, @@ -2203,10 +2251,6 @@ void __init xen_init_mmu_ops(void) memset(dummy_mapping, 0xff, PAGE_SIZE); } -/* Protected by xen_reservation_lock. */ -#define MAX_CONTIG_ORDER 9 /* 2MB */ -static unsigned long discontig_frames[1< MAX_CONTIG_ORDER)) - return -ENOMEM; + if (unlikely(order > discontig_frames_order)) { + if (!discontig_frames_dyn) + return -ENOMEM; + + if (alloc_discontig_frames(order)) + return -ENOMEM; + } memset((void *) vstart, 0, PAGE_SIZE << order); spin_lock_irqsave(&xen_reservation_lock, flags); + in_frames = discontig_frames; + /* 1. Zap current PTEs, remembering MFNs. */ xen_zap_pfn_range(vstart, order, in_frames, NULL); @@ -2358,12 +2409,12 @@ int xen_create_contiguous_region(phys_addr_t pstart, unsigned int order, void xen_destroy_contiguous_region(phys_addr_t pstart, unsigned int order) { - unsigned long *out_frames = discontig_frames, in_frame; + unsigned long *out_frames, in_frame; unsigned long flags; int success; unsigned long vstart; - if (unlikely(order > MAX_CONTIG_ORDER)) + if (unlikely(order > discontig_frames_order)) return; vstart = (unsigned long)phys_to_virt(pstart); @@ -2371,6 +2422,8 @@ void xen_destroy_contiguous_region(phys_addr_t pstart, unsigned int order) spin_lock_irqsave(&xen_reservation_lock, flags); + out_frames = discontig_frames; + /* 1. Find start MFN of contiguous extent. */ in_frame = virt_to_mfn((void *)vstart); From 75ad02318af2e4ae669e26a79f001bd5e1f97472 Mon Sep 17 00:00:00 2001 From: Jan Beulich Date: Wed, 12 Feb 2025 16:14:38 +0100 Subject: [PATCH 3/3] Xen/swiotlb: mark xen_swiotlb_fixup() __init It's sole user (pci_xen_swiotlb_init()) is __init, too. Signed-off-by: Jan Beulich Reviewed-by: Stefano Stabellini Message-ID: Signed-off-by: Juergen Gross --- drivers/xen/swiotlb-xen.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/drivers/xen/swiotlb-xen.c b/drivers/xen/swiotlb-xen.c index 26c62e0d34e9..1f65795cf5d7 100644 --- a/drivers/xen/swiotlb-xen.c +++ b/drivers/xen/swiotlb-xen.c @@ -113,7 +113,7 @@ static struct io_tlb_pool *xen_swiotlb_find_pool(struct device *dev, } #ifdef CONFIG_X86 -int xen_swiotlb_fixup(void *buf, unsigned long nslabs) +int __init xen_swiotlb_fixup(void *buf, unsigned long nslabs) { int rc; unsigned int order = get_order(IO_TLB_SEGSIZE << IO_TLB_SHIFT);