[openib-general] [PATCH v3 1/7] IB/core - Add DMA mapping functions to allow device drivers to interpose

Ralph Campbell ralphc at pathscale.com
Wed Dec 6 10:31:59 PST 2006


This version of the patch adds ib_dma_alloc_coherent()
and ib_dma_free_coherent() to the list of wrapped DMA
functions.  The earlier V2 patches are the same since
this addition doesn't affect them.


The QLogic InfiniPath HCAs use programmed I/O instead of HW DMA.
This patch allows a verbs device driver to interpose on DMA mapping
function calls in order to avoid relying on bus_to_virt() and
phys_to_virt() to undo the mappings created by dma_map_single(),
dma_map_sg(), etc.

From: Ralph Campbell <ralph.campbell at qlogic.com>

diff -r c76ed2f1387b include/rdma/ib_verbs.h
--- a/include/rdma/ib_verbs.h	Wed Nov 29 13:28:14 2006 +0800
+++ b/include/rdma/ib_verbs.h	Tue Dec 05 15:50:07 2006 -0800
@@ -43,6 +43,8 @@
 
 #include <linux/types.h>
 #include <linux/device.h>
+#include <linux/mm.h>
+#include <linux/dma-mapping.h>
 
 #include <asm/atomic.h>
 #include <asm/scatterlist.h>
@@ -846,6 +848,49 @@ struct ib_cache {
 	struct ib_pkey_cache  **pkey_cache;
 	struct ib_gid_cache   **gid_cache;
 	u8                     *lmc_cache;
+};
+
+struct ib_dma_mapping_ops {
+	int		(*mapping_error)(struct ib_device *dev,
+					 u64 dma_addr);
+	u64		(*map_single)(struct ib_device *dev,
+				      void *ptr, size_t size,
+				      enum dma_data_direction direction);
+	void		(*unmap_single)(struct ib_device *dev,
+					u64 addr, size_t size,
+					enum dma_data_direction direction);
+	u64		(*map_page)(struct ib_device *dev,
+				    struct page *page, unsigned long offset,
+				    size_t size,
+				    enum dma_data_direction direction);
+	void		(*unmap_page)(struct ib_device *dev,
+				      u64 addr, size_t size,
+				      enum dma_data_direction direction);
+	int		(*map_sg)(struct ib_device *dev,
+				  struct scatterlist *sg, int nents,
+				  enum dma_data_direction direction);
+	void		(*unmap_sg)(struct ib_device *dev,
+				    struct scatterlist *sg, int nents,
+				    enum dma_data_direction direction);
+	u64		(*dma_address)(struct ib_device *dev,
+				       struct scatterlist *sg);
+	unsigned int	(*dma_len)(struct ib_device *dev,
+				   struct scatterlist *sg);
+	void		(*sync_single_for_cpu)(struct ib_device *dev,
+					       u64 dma_handle,
+					       size_t size,
+				               enum dma_data_direction dir);
+	void		(*sync_single_for_device)(struct ib_device *dev,
+						  u64 dma_handle,
+						  size_t size,
+						  enum dma_data_direction dir);
+	void		*(*alloc_coherent)(struct ib_device *dev,
+					   size_t size,
+					   u64 *dma_handle,
+					   gfp_t flag);
+	void		(*free_coherent)(struct ib_device *dev,
+					 size_t size, void *cpu_addr,
+					 u64 dma_handle);
 };
 
 struct iw_cm_verbs;
@@ -992,6 +1037,8 @@ struct ib_device {
 						  struct ib_mad *in_mad,
 						  struct ib_mad *out_mad);
 
+	struct ib_dma_mapping_ops   *dma_ops;
+
 	struct module               *owner;
 	struct class_device          class_dev;
 	struct kobject               ports_parent;
@@ -1395,8 +1442,214 @@ static inline int ib_req_ncomp_notif(str
  *   usable for DMA.
  * @pd: The protection domain associated with the memory region.
  * @mr_access_flags: Specifies the memory access rights.
+ *
+ * Note that the ib_dma_*() functions defined below must be used 
+ * to create/destroy addresses used with the Lkey or Rkey returned
+ * by ib_get_dma_mr().
  */
 struct ib_mr *ib_get_dma_mr(struct ib_pd *pd, int mr_access_flags);
+
+/**
+ * ib_dma_mapping_error - check a DMA addr for error
+ * @dev: The device for which the dma_addr was created
+ * @dma_addr: The DMA address to check
+ */
+static inline int ib_dma_mapping_error(struct ib_device *dev, u64 dma_addr)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->mapping_error(dev, dma_addr) :
+		dma_mapping_error(dma_addr);
+}
+
+/**
+ * ib_dma_map_single - Map a kernel virtual address to DMA address
+ * @dev: The device for which the dma_addr is to be created
+ * @cpu_addr: The kernel virtual address
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline u64 ib_dma_map_single(struct ib_device *dev,
+				    void *cpu_addr, size_t size,
+				    enum dma_data_direction direction)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->map_single(dev, cpu_addr, size, direction) :
+		dma_map_single(dev->dma_device, cpu_addr, size, direction);
+}
+
+/**
+ * ib_dma_unmap_single - Destroy a mapping created by ib_dma_map_single()
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline void ib_dma_unmap_single(struct ib_device *dev,
+				       u64 addr, size_t size,
+				       enum dma_data_direction direction)
+{
+	dev->dma_ops ?
+		dev->dma_ops->unmap_single(dev, addr, size, direction) :
+		dma_unmap_single(dev->dma_device, addr, size, direction);
+}
+
+/**
+ * ib_dma_map_page - Map a physical page to DMA address
+ * @dev: The device for which the dma_addr is to be created
+ * @page: The page to be mapped
+ * @offset: The offset within the page
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline u64 ib_dma_map_page(struct ib_device *dev,
+				  struct page *page,
+				  unsigned long offset,
+				  size_t size,
+					 enum dma_data_direction direction)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->map_page(dev, page, offset, size, direction) :
+		dma_map_page(dev->dma_device, page, offset, size, direction);
+}
+
+/**
+ * ib_dma_unmap_page - Destroy a mapping created by ib_dma_map_page()
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @direction: The direction of the DMA
+ */
+static inline void ib_dma_unmap_page(struct ib_device *dev,
+				     u64 addr, size_t size,
+				     enum dma_data_direction direction)
+{
+	dev->dma_ops ?
+		dev->dma_ops->unmap_page(dev, addr, size, direction) :
+		dma_unmap_page(dev->dma_device, addr, size, direction);
+}
+
+/**
+ * ib_dma_map_sg - Map a scatter/gather list to DMA addresses
+ * @dev: The device for which the DMA addresses are to be created
+ * @sg: The array of scatter/gather entries
+ * @nents: The number of scatter/gather entries
+ * @direction: The direction of the DMA
+ */
+static inline int ib_dma_map_sg(struct ib_device *dev,
+				struct scatterlist *sg, int nents,
+				enum dma_data_direction direction)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->map_sg(dev, sg, nents, direction) :
+		dma_map_sg(dev->dma_device, sg, nents, direction);
+}
+
+/**
+ * ib_dma_unmap_sg - Unmap a scatter/gather list of DMA addresses
+ * @dev: The device for which the DMA addresses were created
+ * @sg: The array of scatter/gather entries
+ * @nents: The number of scatter/gather entries
+ * @direction: The direction of the DMA
+ */
+static inline void ib_dma_unmap_sg(struct ib_device *dev,
+				   struct scatterlist *sg, int nents,
+				   enum dma_data_direction direction)
+{
+	dev->dma_ops ?
+		dev->dma_ops->unmap_sg(dev, sg, nents, direction) :
+		dma_unmap_sg(dev->dma_device, sg, nents, direction);
+}
+
+/**
+ * ib_sg_dma_address - Return the DMA address from a scatter/gather entry
+ * @dev: The device for which the DMA addresses were created
+ * @sg: The scatter/gather entry
+ */
+static inline u64 ib_sg_dma_address(struct ib_device *dev,
+				    struct scatterlist *sg)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->dma_address(dev, sg) : sg_dma_address(sg);
+}
+
+/**
+ * ib_sg_dma_len - Return the DMA length from a scatter/gather entry
+ * @dev: The device for which the DMA addresses were created
+ * @sg: The scatter/gather entry
+ */
+static inline unsigned int ib_sg_dma_len(struct ib_device *dev,
+					 struct scatterlist *sg)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->dma_len(dev, sg) : sg_dma_len(sg);
+}
+
+/**
+ * ib_dma_sync_single_for_cpu - Prepare DMA region to be accessed by CPU
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @dir: The direction of the DMA
+ */
+static inline void ib_dma_sync_single_for_cpu(struct ib_device *dev,
+					      u64 addr,
+					      size_t size,
+					      enum dma_data_direction dir)
+{
+	dev->dma_ops ?
+		dev->dma_ops->sync_single_for_cpu(dev, addr, size, dir) :
+		dma_sync_single_for_cpu(dev->dma_device, addr, size, dir);
+}
+
+/**
+ * ib_dma_sync_single_for_device - Prepare DMA region to be accessed by device
+ * @dev: The device for which the DMA address was created
+ * @addr: The DMA address
+ * @size: The size of the region in bytes
+ * @dir: The direction of the DMA
+ */
+static inline void ib_dma_sync_single_for_device(struct ib_device *dev,
+						 u64 addr,
+						 size_t size,
+						 enum dma_data_direction dir)
+{
+	dev->dma_ops ?
+		dev->dma_ops->sync_single_for_device(dev, addr, size, dir) :
+		dma_sync_single_for_device(dev->dma_device, addr, size, dir);
+}
+
+/**
+ * ib_dma_alloc_coherent - Allocate memory and map it for DMA
+ * @dev: The device for which the DMA address is requested
+ * @size: The size of the region to allocate in bytes
+ * @dma_handle: A pointer for returning the DMA address of the region
+ * @flag: memory allocator flags
+ */
+static inline void *ib_dma_alloc_coherent(struct ib_device *dev,
+					   size_t size,
+					   u64 *dma_handle,
+					   gfp_t flag)
+{
+	return dev->dma_ops ?
+		dev->dma_ops->alloc_coherent(dev, size, dma_handle, flag) :
+		dma_alloc_coherent(dev->dma_device, size, dma_handle, flag);
+}
+
+/**
+ * ib_dma_free_coherent - Free memory allocated by ib_dma_alloc_coherent()
+ * @dev: The device for which the DMA addresses were allocated
+ * @size: The size of the region
+ * @cpu_addr: the address returned by ib_dma_alloc_coherent()
+ * @dma_handle: the DMA address returned by ib_dma_alloc_coherent()
+ */
+static inline void ib_dma_free_coherent(struct ib_device *dev,
+					size_t size, void *cpu_addr,
+					u64 dma_handle)
+{
+	dev->dma_ops ?
+		dev->dma_ops->free_coherent(dev, size, cpu_addr, dma_handle) :
+		dma_free_coherent(dev->dma_device, size, cpu_addr, dma_handle);
+}
 
 /**
  * ib_reg_phys_mr - Prepares a virtually addressed memory region for use
diff -r c76ed2f1387b drivers/infiniband/core/mad.c
--- a/drivers/infiniband/core/mad.c	Wed Nov 29 13:28:14 2006 +0800
+++ b/drivers/infiniband/core/mad.c	Wed Nov 29 13:54:36 2006 -0800
@@ -999,16 +999,16 @@ int ib_send_mad(struct ib_mad_send_wr_pr
 
 	mad_agent = mad_send_wr->send_buf.mad_agent;
 	sge = mad_send_wr->sg_list;
-	sge[0].addr = dma_map_single(mad_agent->device->dma_device,
-				     mad_send_wr->send_buf.mad,
-				     sge[0].length,
-				     DMA_TO_DEVICE);
+	sge[0].addr = ib_dma_map_single(mad_agent->device,
+					mad_send_wr->send_buf.mad,
+					sge[0].length,
+					DMA_TO_DEVICE);
 	pci_unmap_addr_set(mad_send_wr, header_mapping, sge[0].addr);
 
-	sge[1].addr = dma_map_single(mad_agent->device->dma_device,
-				     ib_get_payload(mad_send_wr),
-				     sge[1].length,
-				     DMA_TO_DEVICE);
+	sge[1].addr = ib_dma_map_single(mad_agent->device,
+					ib_get_payload(mad_send_wr),
+					sge[1].length,
+					DMA_TO_DEVICE);
 	pci_unmap_addr_set(mad_send_wr, payload_mapping, sge[1].addr);
 
 	spin_lock_irqsave(&qp_info->send_queue.lock, flags);
@@ -1027,12 +1027,14 @@ int ib_send_mad(struct ib_mad_send_wr_pr
 	}
 	spin_unlock_irqrestore(&qp_info->send_queue.lock, flags);
 	if (ret) {
-		dma_unmap_single(mad_agent->device->dma_device,
-				 pci_unmap_addr(mad_send_wr, header_mapping),
-				 sge[0].length, DMA_TO_DEVICE);
-		dma_unmap_single(mad_agent->device->dma_device,
-				 pci_unmap_addr(mad_send_wr, payload_mapping),
-				 sge[1].length, DMA_TO_DEVICE);
+		ib_dma_unmap_single(mad_agent->device,
+				    pci_unmap_addr(mad_send_wr,
+						   header_mapping),
+				    sge[0].length, DMA_TO_DEVICE);
+		ib_dma_unmap_single(mad_agent->device,
+				    pci_unmap_addr(mad_send_wr,
+						   payload_mapping),
+				    sge[1].length, DMA_TO_DEVICE);
 	}
 	return ret;
 }
@@ -1851,11 +1853,11 @@ static void ib_mad_recv_done_handler(str
 	mad_priv_hdr = container_of(mad_list, struct ib_mad_private_header,
 				    mad_list);
 	recv = container_of(mad_priv_hdr, struct ib_mad_private, header);
-	dma_unmap_single(port_priv->device->dma_device,
-			 pci_unmap_addr(&recv->header, mapping),
-			 sizeof(struct ib_mad_private) -
-			 sizeof(struct ib_mad_private_header),
-			 DMA_FROM_DEVICE);
+	ib_dma_unmap_single(port_priv->device,
+			    pci_unmap_addr(&recv->header, mapping),
+			    sizeof(struct ib_mad_private) -
+			      sizeof(struct ib_mad_private_header),
+			    DMA_FROM_DEVICE);
 
 	/* Setup MAD receive work completion from "normal" work completion */
 	recv->header.wc = *wc;
@@ -2081,12 +2083,12 @@ static void ib_mad_send_done_handler(str
 	qp_info = send_queue->qp_info;
 
 retry:
-	dma_unmap_single(mad_send_wr->send_buf.mad_agent->device->dma_device,
-			 pci_unmap_addr(mad_send_wr, header_mapping),
-			 mad_send_wr->sg_list[0].length, DMA_TO_DEVICE);
-	dma_unmap_single(mad_send_wr->send_buf.mad_agent->device->dma_device,
-			 pci_unmap_addr(mad_send_wr, payload_mapping),
-			 mad_send_wr->sg_list[1].length, DMA_TO_DEVICE);
+	ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device,
+			    pci_unmap_addr(mad_send_wr, header_mapping),
+			    mad_send_wr->sg_list[0].length, DMA_TO_DEVICE);
+	ib_dma_unmap_single(mad_send_wr->send_buf.mad_agent->device,
+			    pci_unmap_addr(mad_send_wr, payload_mapping),
+			    mad_send_wr->sg_list[1].length, DMA_TO_DEVICE);
 	queued_send_wr = NULL;
 	spin_lock_irqsave(&send_queue->lock, flags);
 	list_del(&mad_list->list);
@@ -2527,12 +2529,11 @@ static int ib_mad_post_receive_mads(stru
 				break;
 			}
 		}
-		sg_list.addr = dma_map_single(qp_info->port_priv->
-					        device->dma_device,
-					      &mad_priv->grh,
-					      sizeof *mad_priv -
-					        sizeof mad_priv->header,
-					      DMA_FROM_DEVICE);
+		sg_list.addr = ib_dma_map_single(qp_info->port_priv->device,
+						 &mad_priv->grh,
+						 sizeof *mad_priv -
+						   sizeof mad_priv->header,
+						 DMA_FROM_DEVICE);
 		pci_unmap_addr_set(&mad_priv->header, mapping, sg_list.addr);
 		recv_wr.wr_id = (unsigned long)&mad_priv->header.mad_list;
 		mad_priv->header.mad_list.mad_queue = recv_queue;
@@ -2548,12 +2549,12 @@ static int ib_mad_post_receive_mads(stru
 			list_del(&mad_priv->header.mad_list.list);
 			recv_queue->count--;
 			spin_unlock_irqrestore(&recv_queue->lock, flags);
-			dma_unmap_single(qp_info->port_priv->device->dma_device,
-					 pci_unmap_addr(&mad_priv->header,
-							mapping),
-					 sizeof *mad_priv -
-					   sizeof mad_priv->header,
-					 DMA_FROM_DEVICE);
+			ib_dma_unmap_single(qp_info->port_priv->device,
+					    pci_unmap_addr(&mad_priv->header,
+							   mapping),
+					    sizeof *mad_priv -
+					      sizeof mad_priv->header,
+					    DMA_FROM_DEVICE);
 			kmem_cache_free(ib_mad_cache, mad_priv);
 			printk(KERN_ERR PFX "ib_post_recv failed: %d\n", ret);
 			break;
@@ -2585,11 +2586,11 @@ static void cleanup_recv_queue(struct ib
 		/* Remove from posted receive MAD list */
 		list_del(&mad_list->list);
 
-		dma_unmap_single(qp_info->port_priv->device->dma_device,
-				 pci_unmap_addr(&recv->header, mapping),
-				 sizeof(struct ib_mad_private) -
-				 sizeof(struct ib_mad_private_header),
-				 DMA_FROM_DEVICE);
+		ib_dma_unmap_single(qp_info->port_priv->device,
+				    pci_unmap_addr(&recv->header, mapping),
+				    sizeof(struct ib_mad_private) -
+				      sizeof(struct ib_mad_private_header),
+				    DMA_FROM_DEVICE);
 		kmem_cache_free(ib_mad_cache, recv);
 	}
 
diff -r c76ed2f1387b drivers/infiniband/core/uverbs_mem.c
--- a/drivers/infiniband/core/uverbs_mem.c	Wed Nov 29 13:28:14 2006 +0800
+++ b/drivers/infiniband/core/uverbs_mem.c	Wed Nov 29 13:54:36 2006 -0800
@@ -52,8 +52,8 @@ static void __ib_umem_release(struct ib_
 	int i;
 
 	list_for_each_entry_safe(chunk, tmp, &umem->chunk_list, list) {
-		dma_unmap_sg(dev->dma_device, chunk->page_list,
-			     chunk->nents, DMA_BIDIRECTIONAL);
+		ib_dma_unmap_sg(dev, chunk->page_list,
+				chunk->nents, DMA_BIDIRECTIONAL);
 		for (i = 0; i < chunk->nents; ++i) {
 			if (umem->writable && dirty)
 				set_page_dirty_lock(chunk->page_list[i].page);
@@ -136,10 +136,10 @@ int ib_umem_get(struct ib_device *dev, s
 				chunk->page_list[i].length = PAGE_SIZE;
 			}
 
-			chunk->nmap = dma_map_sg(dev->dma_device,
-						 &chunk->page_list[0],
-						 chunk->nents,
-						 DMA_BIDIRECTIONAL);
+			chunk->nmap = ib_dma_map_sg(dev,
+						    &chunk->page_list[0],
+						    chunk->nents,
+						    DMA_BIDIRECTIONAL);
 			if (chunk->nmap <= 0) {
 				for (i = 0; i < chunk->nents; ++i)
 					put_page(chunk->page_list[i].page);






More information about the general mailing list