dma-mapping updates for Linux 6.8

- reduce area lock contention for non-primary IO TLB pools (Petr Tesarik)
  - don't store redundant offsets in the dma_ranges stuctures
    (Robin Murphy)
  - clear dev->dma_mem when freeing per-device pools (Joakim Zhang)
 -----BEGIN PGP SIGNATURE-----
 
 iQI/BAABCgApFiEEgdbnc3r/njty3Iq9D55TZVIEUYMFAmWcGxQLHGhjaEBsc3Qu
 ZGUACgkQD55TZVIEUYNf6hAAi9wP1ehnIqHCczCxpideyJnE76r+LgjInjudQUqE
 cnpl7E+dO/e/7Trk+L7hIrzi5uz8m9e+DZgL9wUY4h5mvJ+8ELet3Ec62UMVL1g0
 cWSYOtlMdUZn9Oy+qy2TTCa//1HyzcWQdplVwcOqD7zCLO4PavUR18+Vw5eDUpBR
 TE8EB+7P8ta8XSFnsryZS4zI1AhTTjZfh8ZgPdp+niBh7XAqOFNn3WiGK4qvA9o6
 nIjIV6ydBjZYkyYPeDsqszqmZG64mEeGUZhLWmjAyg1/c8so7uFviNfJ05od34js
 aWpmFxrM9Mm4BaBiU3FsSQkMBGCGaD/H2UXjIl1Qayt+pzUfaP7+8UWWo/T7Mj35
 RFKe9xzlPY8rqOszdBBvy6lCWguHXw4d4IFoqOz+YoUaxlV+RAbFOCHtW2BNvtPe
 b1YCr/FKNQ8NxsJWnbcehDtClY461pqBbaDrio3K7eTJgG10biAoWBfPhV+5VEer
 aB14krQcn7v1vXjfLu2huSrPt1ZjXuWVfXA3nO3Mt3VWxZWat82gLkFyt5N6ZfQ4
 juaDMX3Vzlz3VPf4MHFC+yFRx55b/9X26lC1BlSoo4tAknoo746Lvy/PasZarILC
 sGPt+2BVlPQ466zkjky4GtoNof1TMNuPF0Xr/mNCEjxCYUGQcTIdmOvA9y8mu/V+
 7M4=
 =v5F+
 -----END PGP SIGNATURE-----

Merge tag 'dma-mapping-6.8-2024-01-08' of git://git.infradead.org/users/hch/dma-mapping

Pull dma-mapping updates from Christoph Hellwig:

 - reduce area lock contention for non-primary IO TLB pools (Petr
   Tesarik)

 - don't store redundant offsets in the dma_ranges stuctures (Robin
   Murphy)

 - clear dev->dma_mem when freeing per-device pools (Joakim Zhang)

* tag 'dma-mapping-6.8-2024-01-08' of git://git.infradead.org/users/hch/dma-mapping:
  dma-mapping: clear dev->dma_mem to NULL after freeing it
  swiotlb: reduce area lock contention for non-primary IO TLB pools
  dma-mapping: don't store redundant offsets

drivers/acpi/scan.c

@@ -1532,7 +1532,6 @@ int acpi_dma_get_range(struct device *dev, const struct bus_dma_region **map)
 			r->cpu_start = rentry->res->start;
 			r->dma_start = rentry->res->start - rentry->offset;
 			r->size = resource_size(rentry->res);
-			r->offset = rentry->offset;
 			r++;
 		}
 	}

drivers/of/address.c

@@ -955,7 +955,6 @@ int of_dma_get_range(struct device_node *np, const struct bus_dma_region **map)
 		r->cpu_start = range.cpu_addr;
 		r->dma_start = range.bus_addr;
 		r->size = range.size;
-		r->offset = range.cpu_addr - range.bus_addr;
 		r++;
 	}
 out:

include/linux/dma-direct.h

@@ -21,7 +21,6 @@ struct bus_dma_region {
 	phys_addr_t	cpu_start;
 	dma_addr_t	dma_start;
 	u64		size;
-	u64		offset;
 };
 
 static inline dma_addr_t translate_phys_to_dma(struct device *dev,
@@ -29,9 +28,12 @@ static inline dma_addr_t translate_phys_to_dma(struct device *dev,
 {
 	const struct bus_dma_region *m;
 
-	for (m = dev->dma_range_map; m->size; m++)
-		if (paddr >= m->cpu_start && paddr - m->cpu_start < m->size)
-			return (dma_addr_t)paddr - m->offset;
+	for (m = dev->dma_range_map; m->size; m++) {
+		u64 offset = paddr - m->cpu_start;
+
+		if (paddr >= m->cpu_start && offset < m->size)
+			return m->dma_start + offset;
+	}
 
 	/* make sure dma_capable fails when no translation is available */
 	return DMA_MAPPING_ERROR;
@@ -42,9 +44,12 @@ static inline phys_addr_t translate_dma_to_phys(struct device *dev,
 {
 	const struct bus_dma_region *m;
 
-	for (m = dev->dma_range_map; m->size; m++)
-		if (dma_addr >= m->dma_start && dma_addr - m->dma_start < m->size)
-			return (phys_addr_t)dma_addr + m->offset;
+	for (m = dev->dma_range_map; m->size; m++) {
+		u64 offset = dma_addr - m->dma_start;
+
+		if (dma_addr >= m->dma_start && offset < m->size)
+			return m->cpu_start + offset;
+	}
 
 	return (phys_addr_t)-1;
 }

kernel/dma/coherent.c

@@ -132,8 +132,10 @@ int dma_declare_coherent_memory(struct device *dev, phys_addr_t phys_addr,
 
 void dma_release_coherent_memory(struct device *dev)
 {
-	if (dev)
+	if (dev) {
 		_dma_release_coherent_memory(dev->dma_mem);
+		dev->dma_mem = NULL;
+	}
 }
 
 static void *__dma_alloc_from_coherent(struct device *dev,

kernel/dma/direct.c

@@ -677,7 +677,6 @@ int dma_direct_set_offset(struct device *dev, phys_addr_t cpu_start,
 		return -ENOMEM;
 	map[0].cpu_start = cpu_start;
 	map[0].dma_start = dma_start;
-	map[0].offset = offset;
 	map[0].size = size;
 	dev->dma_range_map = map;
 	return 0;

kernel/dma/swiotlb.c

@@ -957,7 +957,7 @@ static void dec_used(struct io_tlb_mem *mem, unsigned int nslots)
 #endif /* CONFIG_DEBUG_FS */
 
 /**
- * swiotlb_area_find_slots() - search for slots in one IO TLB memory area
+ * swiotlb_search_pool_area() - search one memory area in one pool
  * @dev:	Device which maps the buffer.
  * @pool:	Memory pool to be searched.
  * @area_index:	Index of the IO TLB memory area to be searched.
@@ -972,7 +972,7 @@ static void dec_used(struct io_tlb_mem *mem, unsigned int nslots)
  *
  * Return: Index of the first allocated slot, or -1 on error.
  */
-static int swiotlb_area_find_slots(struct device *dev, struct io_tlb_pool *pool,
+static int swiotlb_search_pool_area(struct device *dev, struct io_tlb_pool *pool,
 		int area_index, phys_addr_t orig_addr, size_t alloc_size,
 		unsigned int alloc_align_mask)
 {
@@ -1066,41 +1066,50 @@ static int swiotlb_area_find_slots(struct device *dev, struct io_tlb_pool *pool,
 	return slot_index;
 }
 
+#ifdef CONFIG_SWIOTLB_DYNAMIC
+
 /**
- * swiotlb_pool_find_slots() - search for slots in one memory pool
+ * swiotlb_search_area() - search one memory area in all pools
  * @dev:	Device which maps the buffer.
- * @pool:	Memory pool to be searched.
+ * @start_cpu:	Start CPU number.
+ * @cpu_offset:	Offset from @start_cpu.
  * @orig_addr:	Original (non-bounced) IO buffer address.
  * @alloc_size: Total requested size of the bounce buffer,
  *		including initial alignment padding.
  * @alloc_align_mask:	Required alignment of the allocated buffer.
+ * @retpool:	Used memory pool, updated on return.
  *
- * Search through one memory pool to find a sequence of slots that match the
+ * Search one memory area in all pools for a sequence of slots that match the
  * allocation constraints.
  *
  * Return: Index of the first allocated slot, or -1 on error.
  */
-static int swiotlb_pool_find_slots(struct device *dev, struct io_tlb_pool *pool,
-		phys_addr_t orig_addr, size_t alloc_size,
-		unsigned int alloc_align_mask)
+static int swiotlb_search_area(struct device *dev, int start_cpu,
+		int cpu_offset, phys_addr_t orig_addr, size_t alloc_size,
+		unsigned int alloc_align_mask, struct io_tlb_pool **retpool)
 {
-	int start = raw_smp_processor_id() & (pool->nareas - 1);
-	int i = start, index;
+	struct io_tlb_mem *mem = dev->dma_io_tlb_mem;
+	struct io_tlb_pool *pool;
+	int area_index;
+	int index = -1;
 
-	do {
-		index = swiotlb_area_find_slots(dev, pool, i, orig_addr,
-						alloc_size, alloc_align_mask);
-		if (index >= 0)
-			return index;
-		if (++i >= pool->nareas)
-			i = 0;
-	} while (i != start);
-
-	return -1;
+	rcu_read_lock();
+	list_for_each_entry_rcu(pool, &mem->pools, node) {
+		if (cpu_offset >= pool->nareas)
+			continue;
+		area_index = (start_cpu + cpu_offset) & (pool->nareas - 1);
+		index = swiotlb_search_pool_area(dev, pool, area_index,
+						 orig_addr, alloc_size,
+						 alloc_align_mask);
+		if (index >= 0) {
+			*retpool = pool;
+			break;
+		}
+	}
+	rcu_read_unlock();
+	return index;
 }
 
-#ifdef CONFIG_SWIOTLB_DYNAMIC
-
 /**
  * swiotlb_find_slots() - search for slots in the whole swiotlb
  * @dev:	Device which maps the buffer.
@@ -1124,18 +1133,17 @@ static int swiotlb_find_slots(struct device *dev, phys_addr_t orig_addr,
 	unsigned long nslabs;
 	unsigned long flags;
 	u64 phys_limit;
+	int cpu, i;
 	int index;
 
-	rcu_read_lock();
-	list_for_each_entry_rcu(pool, &mem->pools, node) {
-		index = swiotlb_pool_find_slots(dev, pool, orig_addr,
-						alloc_size, alloc_align_mask);
-		if (index >= 0) {
-			rcu_read_unlock();
+	cpu = raw_smp_processor_id();
+	for (i = 0; i < default_nareas; ++i) {
+		index = swiotlb_search_area(dev, cpu, i, orig_addr, alloc_size,
+					    alloc_align_mask, &pool);
+		if (index >= 0)
 			goto found;
-		}
 	}
-	rcu_read_unlock();
+
 	if (!mem->can_grow)
 		return -1;
 
@@ -1148,8 +1156,8 @@ static int swiotlb_find_slots(struct device *dev, phys_addr_t orig_addr,
 	if (!pool)
 		return -1;
 
-	index = swiotlb_pool_find_slots(dev, pool, orig_addr,
-					alloc_size, alloc_align_mask);
+	index = swiotlb_search_pool_area(dev, pool, 0, orig_addr,
+					 alloc_size, alloc_align_mask);
 	if (index < 0) {
 		swiotlb_dyn_free(&pool->rcu);
 		return -1;
@@ -1192,9 +1200,21 @@ static int swiotlb_find_slots(struct device *dev, phys_addr_t orig_addr,
 		size_t alloc_size, unsigned int alloc_align_mask,
 		struct io_tlb_pool **retpool)
 {
-	*retpool = &dev->dma_io_tlb_mem->defpool;
-	return swiotlb_pool_find_slots(dev, *retpool,
-				       orig_addr, alloc_size, alloc_align_mask);
+	struct io_tlb_pool *pool;
+	int start, i;
+	int index;
+
+	*retpool = pool = &dev->dma_io_tlb_mem->defpool;
+	i = start = raw_smp_processor_id() & (pool->nareas - 1);
+	do {
+		index = swiotlb_search_pool_area(dev, pool, i, orig_addr,
+						 alloc_size, alloc_align_mask);
+		if (index >= 0)
+			return index;
+		if (++i >= pool->nareas)
+			i = 0;
+	} while (i != start);
+	return -1;
 }
 
 #endif /* CONFIG_SWIOTLB_DYNAMIC */