Thread (67 messages) 67 messages, 6 authors, 10d ago

Re: [PATCH v4 04/13] dma: swiotlb: track pool encryption state and honor DMA_ATTR_CC_SHARED

From: Aneesh Kumar K.V <aneesh.kumar@kernel.org>
Date: 2026-05-21 17:21:11
Also in: linux-arm-kernel, linux-coco, linux-iommu, linux-s390, lkml

Mostafa Saleh [off-list ref] writes:
On Tue, May 12, 2026 at 10:05 AM Aneesh Kumar K.V (Arm)
[off-list ref] wrote:
quoted
@@ -1411,6 +1436,16 @@ phys_addr_t swiotlb_tbl_map_single(struct device *dev, phys_addr_t orig_addr,
        if (cc_platform_has(CC_ATTR_MEM_ENCRYPT))
                pr_warn_once("Memory encryption is active and system is using DMA bounce buffers\n");

+       /*
+        * if we are trying to swiotlb map a decrypted paddr or the paddr is encrypted
+        * but the device is forcing decryption, use decrypted io_tlb_mem
+        */
+       if ((attrs & DMA_ATTR_CC_SHARED) || force_dma_unencrypted(dev))
I don't think swiotlb needs to know about force_dma_unencrypted(), the
dma/direct caller should have all the information to pass the
appropriate flags.

Thanks.
Mostafa
quoted
+               require_decrypted = true;
+
+       if (require_decrypted != mem->unencrypted)
+               return (phys_addr_t)DMA_MAPPING_ERROR;
+
Based on other email threads, this is now updated to
@@ -1372,9 +1417,19 @@ static unsigned long mem_used(struct io_tlb_mem *mem)
  *			any pre- or post-padding for alignment
  * @alloc_align_mask:	Required start and end alignment of the allocated buffer
  * @dir:		DMA direction
- * @attrs:		Optional DMA attributes for the map operation
+ * @attrs:		Optional DMA attributes for the map operation, updated
+ *			to match the selected SWIOTLB pool
  *
  * Find and allocate a suitable sequence of IO TLB slots for the request.
+ * The device's SWIOTLB pool must match the device's current DMA encryption
+ * requirements. If the device requires decrypted DMA, bouncing is done through
+ * an unencrypted pool and the mapping is marked shared. If the device can DMA
+ * to encrypted memory, bouncing is done through an encrypted pool even when the
+ * original DMA address was unencrypted. Enabling encrypted DMA for a device is
+ * therefore expected to update its default io_tlb_mem to an encrypted pool, so
+ * later bounce mappings for both encrypted and decrypted original memory use
+ * that encrypted pool.
+ *
  * The allocated space starts at an alignment specified by alloc_align_mask,
  * and the size of the allocated space is rounded up so that the total amount
  * of allocated space is a multiple of (alloc_align_mask + 1). If
@@ -1411,6 +1466,16 @@ phys_addr_t swiotlb_tbl_map_single(struct device *dev, phys_addr_t orig_addr,
 	if (cc_platform_has(CC_ATTR_MEM_ENCRYPT))
 		pr_warn_once("Memory encryption is active and system is using DMA bounce buffers\n");
 
+	/* swiotlb pool is incorrect for this device */
+	if (unlikely(mem->unencrypted != force_dma_unencrypted(dev)))
+		return (phys_addr_t)DMA_MAPPING_ERROR;
+
+	/* Force attrs to match the kind of memory in the pool */
+	if (mem->unencrypted)
+		*attrs |= DMA_ATTR_CC_SHARED;
+	else
+		*attrs &= ~DMA_ATTR_CC_SHARED;
+
 	/*
 	 * The default swiotlb memory pool is allocated with PAGE_SIZE
 	 * alignment. If a mapping is requested with larger alignment,
@@ -1608,8 +1673,11 @@ dma_addr_t swiotlb_map(struct device *dev, phys_addr_t paddr, size_t size,
 	if (swiotlb_addr == (phys_addr_t)DMA_MAPPING_ERROR)
 		return DMA_MAPPING_ERROR;
 
-	/* Ensure that the address returned is DMA'ble */
-	dma_addr = phys_to_dma_unencrypted(dev, swiotlb_addr);
+	if (attrs & DMA_ATTR_CC_SHARED)
+		dma_addr = phys_to_dma_unencrypted(dev, swiotlb_addr);
+	else
+		dma_addr = phys_to_dma_encrypted(dev, swiotlb_addr);
+
 	if (unlikely(!dma_capable(dev, dma_addr, size, true))) {
 		__swiotlb_tbl_unmap_single(dev, swiotlb_addr, size, dir,
 			attrs | DMA_ATTR_SKIP_CPU_SYNC,
@@ -1773,7 +1841,7 @@ static inline void swiotlb_create_debugfs_files(struct io_tlb_mem *mem,
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help