Merge tag 'dma-mapping-6.6-2023-09-30' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping fixes from Christoph Hellwig:

 - fix the narea calculation in swiotlb initialization (Ross Lagerwall)

 - fix the check whether a device has used swiotlb (Petr Tesarik)

* tag 'dma-mapping-6.6-2023-09-30' of git://git.infradead.org/users/hch/dma-mapping:
  swiotlb: fix the check whether a device has used software IO TLB
  swiotlb: use the calculated number of areas

include/linux/swiotlb.h

@@ -172,14 +172,23 @@ static inline bool is_swiotlb_buffer(struct device *dev, phys_addr_t paddr)
 	if (!mem)
 		return false;
 
-	if (IS_ENABLED(CONFIG_SWIOTLB_DYNAMIC)) {
-		/* Pairs with smp_wmb() in swiotlb_find_slots() and
-		 * swiotlb_dyn_alloc(), which modify the RCU lists.
-		 */
-		smp_rmb();
-		return swiotlb_find_pool(dev, paddr);
-	}
+#ifdef CONFIG_SWIOTLB_DYNAMIC
+	/*
+	 * All SWIOTLB buffer addresses must have been returned by
+	 * swiotlb_tbl_map_single() and passed to a device driver.
+	 * If a SWIOTLB address is checked on another CPU, then it was
+	 * presumably loaded by the device driver from an unspecified private
+	 * data structure. Make sure that this load is ordered before reading
+	 * dev->dma_uses_io_tlb here and mem->pools in swiotlb_find_pool().
+	 *
+	 * This barrier pairs with smp_mb() in swiotlb_find_slots().
+	 */
+	smp_rmb();
+	return READ_ONCE(dev->dma_uses_io_tlb) &&
+		swiotlb_find_pool(dev, paddr);
+#else
 	return paddr >= mem->defpool.start && paddr < mem->defpool.end;
+#endif
 }
 
 static inline bool is_swiotlb_force_bounce(struct device *dev)

kernel/dma/swiotlb.c

@@ -399,14 +399,13 @@ void __init swiotlb_init_remap(bool addressing_limit, unsigned int flags,
 	}
 
 	mem->areas = memblock_alloc(array_size(sizeof(struct io_tlb_area),
-		default_nareas), SMP_CACHE_BYTES);
+		nareas), SMP_CACHE_BYTES);
 	if (!mem->areas) {
 		pr_warn("%s: Failed to allocate mem->areas.\n", __func__);
 		return;
 	}
 
-	swiotlb_init_io_tlb_pool(mem, __pa(tlb), nslabs, false,
-				 default_nareas);
+	swiotlb_init_io_tlb_pool(mem, __pa(tlb), nslabs, false, nareas);
 	add_mem_pool(&io_tlb_default_mem, mem);
 
 	if (flags & SWIOTLB_VERBOSE)
@@ -729,9 +728,6 @@ static void swiotlb_dyn_alloc(struct work_struct *work)
 	}
 
 	add_mem_pool(mem, pool);
-
-	/* Pairs with smp_rmb() in is_swiotlb_buffer(). */
-	smp_wmb();
 }
 
 /**
@@ -1152,9 +1148,26 @@ static int swiotlb_find_slots(struct device *dev, phys_addr_t orig_addr,
 	spin_unlock_irqrestore(&dev->dma_io_tlb_lock, flags);
 
 found:
-	dev->dma_uses_io_tlb = true;
-	/* Pairs with smp_rmb() in is_swiotlb_buffer() */
-	smp_wmb();
+	WRITE_ONCE(dev->dma_uses_io_tlb, true);
+
+	/*
+	 * The general barrier orders reads and writes against a presumed store
+	 * of the SWIOTLB buffer address by a device driver (to a driver private
+	 * data structure). It serves two purposes.
+	 *
+	 * First, the store to dev->dma_uses_io_tlb must be ordered before the
+	 * presumed store. This guarantees that the returned buffer address
+	 * cannot be passed to another CPU before updating dev->dma_uses_io_tlb.
+	 *
+	 * Second, the load from mem->pools must be ordered before the same
+	 * presumed store. This guarantees that the returned buffer address
+	 * cannot be observed by another CPU before an update of the RCU list
+	 * that was made by swiotlb_dyn_alloc() on a third CPU (cf. multicopy
+	 * atomicity).
+	 *
+	 * See also the comment in is_swiotlb_buffer().
+	 */
+	smp_mb();
 
 	*retpool = pool;
 	return index;