Add logic to consolidate TRBs for Synopsys xHC

This commit adds the documentation for sgl-trb-cache-size-quirk, and snps,sgl-trb-cache-size-quirk property. These when set enables the quirk for XHCI driver for consolidation of sg list into a temporary buffer when small buffer sizes are scattered over the sg list not making up to MPS or total transfer size within TRB cache size with Synopsys xHC.

Signed-off-by Tejas Joglekar <[email protected]>

---
Changes in v2:
- Renamed the property

 Documentation/devicetree/bindings/usb/dwc3.txt     | 4 ++++
 Documentation/devicetree/bindings/usb/usb-xhci.txt | 3 +++
 2 files changed, 7 insertions(+)

Documentation/devicetree/bindings/usb/dwc3.txt +4 -0

index 9946ff9ba735..6d0418ee4dbd 100644
@@ -104,6 +104,10 @@ Optional properties:
 			this and tx-thr-num-pkt-prd to a valid, non-zero value
 			1-16 (DWC_usb31 programming guide section 1.2.3) to
 			enable periodic ESS TX threshold.
+ - snps,sgl-trb-cache-size-quirk: enable sg list consolidation - host mode
+			only. Set to use SG buffers of at least MPS size
+			by consolidating smaller SG buffers list into a
+			single buffer.
 
  - <DEPRECATED> tx-fifo-resize: determines if the FIFO *has* to be reallocated.
  - snps,incr-burst-type-adjustment: Value for INCR burst type of GSBUSCFG0

Documentation/devicetree/bindings/usb/usb-xhci.txt +0 -0

index 3f378951d624..14d900474894 100644

Documentation/devicetree/bindings/usb/usb-xhci.txt +0 -0

 additionally the properties from usb-hcd.yaml (in the current directory) are
 supported.

Signature · 1 line

2.11.0

Quoted 29 lines · Tejas Joglekar

This commit adds the documentation for sgl-trb-cache-size-quirk, and snps,sgl-trb-cache-size-quirk property. These when set enables the quirk for XHCI driver for consolidation of sg list into a temporary buffer when small buffer sizes are scattered over the sg list not making up to MPS or total transfer size within TRB cache size with Synopsys xHC.

Signed-off-by Tejas Joglekar <[email protected]>

---
Changes in v2:
- Renamed the property

 Documentation/devicetree/bindings/usb/dwc3.txt     | 4 ++++
 Documentation/devicetree/bindings/usb/usb-xhci.txt | 3 +++
 2 files changed, 7 insertions(+)

Documentation/devicetree/bindings/usb/dwc3.txt +4 -0

index 9946ff9ba735..6d0418ee4dbd 100644
@@ -104,6 +104,10 @@ Optional properties:
 			this and tx-thr-num-pkt-prd to a valid, non-zero value
 			1-16 (DWC_usb31 programming guide section 1.2.3) to
 			enable periodic ESS TX threshold.
+ - snps,sgl-trb-cache-size-quirk: enable sg list consolidation - host mode
+			only. Set to use SG buffers of at least MPS size
+			by consolidating smaller SG buffers list into a
+			single buffer.
 

Documentation/devicetree/bindings/usb/usb-xhci.txt Optional properties: · line 48

Quoted 13 lines

Documentation/devicetree/bindings/usb/dwc3.txt +0 -0

  - <DEPRECATED> tx-fifo-resize: determines if the FIFO *has* to be reallocated.
  - snps,incr-burst-type-adjustment: Value for INCR burst type of GSBUSCFG0

Documentation/devicetree/bindings/usb/usb-xhci.txt +3 -0

index 3f378951d624..14d900474894 100644
@@ -43,6 +43,9 @@ Optional properties:
   - quirk-broken-port-ped: set if the controller has broken port disable mechanism
   - imod-interval-ns: default interrupt moderation interval is 5000ns
   - phys : see usb-hcd.yaml in the current directory
+  - sgl-trb-cache-size-quirk: set if you need to consolidate sg list into a
+    temporary buffer when small SG buffer sizes does not make upto MPS
+    size or total transfer size across the TRB cache size.

Why do we have 2 different names?

Quoted 6 lines

Documentation/devicetree/bindings/usb/usb-xhci.txt +0 -0

 additionally the properties from usb-hcd.yaml (in the current directory) are
 supported.

Signature · 1 line

2.11.0

This commit adds the documentation for sgl-trb-cache-size-quirk, and snps,sgl-trb-cache-size-quirk property. These when set enables the quirk for XHCI driver for consolidation of sg list into a temporary buffer when small buffer sizes are scattered over the sg list not making up to MPS or total transfer size within TRB cache size with Synopsys xHC.

Signed-off-by Tejas Joglekar <[email protected]>

---
Changes in v2:
- Renamed the property

 Documentation/devicetree/bindings/usb/dwc3.txt     | 4 ++++
 Documentation/devicetree/bindings/usb/usb-xhci.txt | 3 +++
 2 files changed, 7 insertions(+)

Documentation/devicetree/bindings/usb/dwc3.txt +4 -0

index 9946ff9ba735..6d0418ee4dbd 100644
@@ -104,6 +104,10 @@ Optional properties:
 			this and tx-thr-num-pkt-prd to a valid, non-zero value
 			1-16 (DWC_usb31 programming guide section 1.2.3) to
 			enable periodic ESS TX threshold.
+ - snps,sgl-trb-cache-size-quirk: enable sg list consolidation - host mode
+			only. Set to use SG buffers of at least MPS size
+			by consolidating smaller SG buffers list into a
+			single buffer.
 
  - <DEPRECATED> tx-fifo-resize: determines if the FIFO *has* to be reallocated.
  - snps,incr-burst-type-adjustment: Value for INCR burst type of GSBUSCFG0

Documentation/devicetree/bindings/usb/usb-xhci.txt +3 -0

index 3f378951d624..14d900474894 100644
@@ -43,6 +43,9 @@ Optional properties:
   - quirk-broken-port-ped: set if the controller has broken port disable mechanism
   - imod-interval-ns: default interrupt moderation interval is 5000ns
   - phys : see usb-hcd.yaml in the current directory
+  - sgl-trb-cache-size-quirk: set if you need to consolidate sg list into a
+    temporary buffer when small SG buffer sizes does not make upto MPS
+    size or total transfer size across the TRB cache size.
 
 additionally the properties from usb-hcd.yaml (in the current directory) are
 supported.

Signature · 1 line

2.11.0

Quoted 7 lines · Tejas Joglekar

This commit adds the documentation for sgl-trb-cache-size-quirk, and snps,sgl-trb-cache-size-quirk property. These when set enables the quirk for XHCI driver for consolidation of sg list into a temporary buffer when small buffer sizes are scattered over the sg list not making up to MPS or total transfer size within TRB cache size with Synopsys xHC.

Quoted 1 line

Signed-off-by Tejas Joglekar <[email protected]>

The author and S-o-b emails don't match.

Quoted 35 lines

---
Changes in v2:
- Renamed the property

 Documentation/devicetree/bindings/usb/dwc3.txt     | 4 ++++
 Documentation/devicetree/bindings/usb/usb-xhci.txt | 3 +++
 2 files changed, 7 insertions(+)

Documentation/devicetree/bindings/usb/dwc3.txt +4 -0

index 9946ff9ba735..6d0418ee4dbd 100644
@@ -104,6 +104,10 @@ Optional properties:
 			this and tx-thr-num-pkt-prd to a valid, non-zero value
 			1-16 (DWC_usb31 programming guide section 1.2.3) to
 			enable periodic ESS TX threshold.
+ - snps,sgl-trb-cache-size-quirk: enable sg list consolidation - host mode
+			only. Set to use SG buffers of at least MPS size
+			by consolidating smaller SG buffers list into a
+			single buffer.
 
  - <DEPRECATED> tx-fifo-resize: determines if the FIFO *has* to be reallocated.
  - snps,incr-burst-type-adjustment: Value for INCR burst type of GSBUSCFG0

Documentation/devicetree/bindings/usb/usb-xhci.txt +3 -0

index 3f378951d624..14d900474894 100644
@@ -43,6 +43,9 @@ Optional properties:
   - quirk-broken-port-ped: set if the controller has broken port disable mechanism
   - imod-interval-ns: default interrupt moderation interval is 5000ns
   - phys : see usb-hcd.yaml in the current directory
+  - sgl-trb-cache-size-quirk: set if you need to consolidate sg list into a
+    temporary buffer when small SG buffer sizes does not make upto MPS
+    size or total transfer size across the TRB cache size.
 

Quoted 5 lines

Documentation/devicetree/bindings/usb/usb-xhci.txt +0 -0

 additionally the properties from usb-hcd.yaml (in the current directory) are
 supported.

Signature · 1 line

2.11.0

The Synopsys xHC has an internal TRB cache of size TRB_CACHE_SIZE for each endpoint. The default value for TRB_CACHE_SIZE is 16 for SS and 8 for HS. The controller loads and updates the TRB cache from the transfer ring in system memory whenever the driver issues a start transfer or update transfer command.

For chained TRBs, the Synopsys xHC requires that the total amount of bytes for all TRBs loaded in the TRB cache be greater than or equal to 1 MPS. Or the chain ends within the TRB cache (with a last TRB).

If this requirement is not met, the controller will not be able to send or receive a packet and it will hang causing a driver timeout and error.

This can be a problem if a class driver queues SG requests with many small-buffer entries. The XHCI driver will create a chained TRB for each entry which may trigger this issue.

This patch adds logic to the XHCI driver to detect and prevent this from happening.

For every (TRB_CACHE_SIZE - 2), we check the total buffer size of the SG list and if the last window of (TRB_CACHE_SIZE - 2) SG list length and we don't make up at least 1 MPS, we create a temporary buffer to consolidate full SG list into the buffer.

We check at (TRB_CACHE_SIZE - 2) window because it is possible that there would be a link and/or event data TRB that take up to 2 of the cache entries.

We discovered this issue with devices on other platforms but have not yet come across any device that triggers this on Linux. But it could be a real problem now or in the future. All it takes is N number of small chained TRBs. And other instances of the Synopsys IP may have smaller values for the TRB_CACHE_SIZE which would exacerbate the problem.

Signed-off-by Tejas Joglekar <[email protected]>

---
Changes in v2:
- Removed redundant debug messages
- Modified logic to remove unnecessary changes in hcd.c
- Rename the quirk

 drivers/usb/host/xhci-ring.c |   2 +-
 drivers/usb/host/xhci.c      | 125 +++++++++++++++++++++++++++++++++++++++++++
 drivers/usb/host/xhci.h      |   4 ++
 3 files changed, 130 insertions(+), 1 deletion(-)

drivers/usb/host/xhci-ring.c +1 -1

index a78787bb5133..2fad9474912a 100644
@@ -3291,7 +3291,7 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	full_len = urb->transfer_buffer_length;
 	/* If we have scatter/gather list, we use it. */
-	if (urb->num_sgs) {
+	if (urb->num_sgs && !(urb->transfer_flags & URB_DMA_MAP_SINGLE)) {
 		num_sgs = urb->num_mapped_sgs;
 		sg = urb->sg;
 		addr = (u64) sg_dma_address(sg);

drivers/usb/host/xhci.c +125 -0

index fe38275363e0..15f06bc6b1ad 100644
@@ -1256,6 +1256,106 @@ EXPORT_SYMBOL_GPL(xhci_resume);
 
 /*-------------------------------------------------------------------------*/
 
+static int xhci_map_temp_buffer(struct usb_hcd *hcd, struct urb *urb)
+{
+	void *temp;
+	int ret = 0;
+	unsigned int len;
+	unsigned int buf_len;
+	enum dma_data_direction dir;
+	struct xhci_hcd *xhci;
+
+	xhci = hcd_to_xhci(hcd);
+	dir = usb_urb_dir_in(urb) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
+	buf_len = urb->transfer_buffer_length;
+
+	temp = kzalloc_node(buf_len, GFP_ATOMIC,
+			    dev_to_node(hcd->self.sysdev));
+
+	if (usb_urb_dir_out(urb))
+		len = sg_pcopy_to_buffer(urb->sg, urb->num_sgs,
+					 temp, buf_len, 0);
+
+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;
+	}
+
+	return ret;
+}
+
+static bool xhci_urb_temp_buffer_required(struct usb_hcd *hcd,
+					  struct urb *urb)
+{
+	bool ret = false;
+	unsigned int i;
+	unsigned int len = 0;
+	unsigned int buf_len;
+	unsigned int trb_size;
+	unsigned int max_pkt;
+	struct scatterlist *sg;
+	struct scatterlist *tail_sg;
+
+	sg = urb->sg;
+	tail_sg = urb->sg;
+	buf_len = urb->transfer_buffer_length;
+	max_pkt = usb_endpoint_maxp(&urb->ep->desc);
+
+	if (!urb->num_sgs)
+		return ret;
+
+	if (urb->dev->speed >= USB_SPEED_SUPER)
+		trb_size = TRB_CACHE_SIZE_SS;
+	else
+		trb_size = TRB_CACHE_SIZE_HS;
+
+	if (urb->transfer_buffer_length != 0 &&
+	    !(urb->transfer_flags & URB_NO_TRANSFER_DMA_MAP)) {
+		for_each_sg(urb->sg, sg, urb->num_sgs, i) {
+			len = len + sg->length;
+			if (i > trb_size - 2) {
+				len = len - tail_sg->length;
+				if (len < max_pkt) {
+					ret = true;
+					break;
+				}
+
+				tail_sg = sg_next(tail_sg);
+			}
+		}
+	}
+	return ret;
+}
+
+static void xhci_unmap_temp_buf(struct urb *urb)
+{
+	struct scatterlist *sg;
+	unsigned int len;
+	unsigned int buf_len;
+
+	sg = urb->sg;
+	buf_len = urb->transfer_buffer_length;
+
+	if (usb_urb_dir_in(urb)) {
+		len = sg_pcopy_from_buffer(urb->sg, urb->num_sgs,
+					   urb->transfer_buffer,
+					   buf_len,
+					   0);
+	}
+
+	kfree(urb->transfer_buffer);
+	urb->transfer_buffer = NULL;
+}
+
 /*
  * Bypass the DMA mapping if URB is suitable for Immediate Transfer (IDT),
  * we'll copy the actual data into the TRB address register. This is limited to

@@ -1265,12 +1365,36 @@ EXPORT_SYMBOL_GPL(xhci_resume);
 static int xhci_map_urb_for_dma(struct usb_hcd *hcd, struct urb *urb,
 				gfp_t mem_flags)
 {
+	struct xhci_hcd *xhci;
+
+	xhci = hcd_to_xhci(hcd);
+
 	if (xhci_urb_suitable_for_idt(urb))
 		return 0;
 
+	if (xhci->quirks & XHCI_SG_TRB_CACHE_SIZE_QUIRK) {
+		if (xhci_urb_temp_buffer_required(hcd, urb))
+			return xhci_map_temp_buffer(hcd, urb);
+	}
 	return usb_hcd_map_urb_for_dma(hcd, urb, mem_flags);
 }
 
+static void xhci_unmap_urb_for_dma(struct usb_hcd *hcd, struct urb *urb)
+{
+	struct xhci_hcd *xhci;
+	bool unmap_temp_buf = false;
+
+	xhci = hcd_to_xhci(hcd);
+
+	if (urb->num_sgs && (urb->transfer_flags & URB_DMA_MAP_SINGLE))
+		unmap_temp_buf = true;
+
+	usb_hcd_unmap_urb_for_dma(hcd, urb);
+
+	if ((xhci->quirks & XHCI_SG_TRB_CACHE_SIZE_QUIRK) && unmap_temp_buf)
+		xhci_unmap_temp_buf(urb);
+}
+
 /**
  * xhci_get_endpoint_index - Used for passing endpoint bitmasks between the core and
  * HCDs.  Find the index for an endpoint given its descriptor.  Use the return

@@ -5315,6 +5439,7 @@ static const struct hc_driver xhci_hc_driver = {
 	 * managing i/o requests and associated device resources
 	 */
 	.map_urb_for_dma =      xhci_map_urb_for_dma,
+	.unmap_urb_for_dma =    xhci_unmap_urb_for_dma,
 	.urb_enqueue =		xhci_urb_enqueue,
 	.urb_dequeue =		xhci_urb_dequeue,
 	.alloc_dev =		xhci_alloc_dev,

drivers/usb/host/xhci.h +0 -0

index 4db825459b40..14600214188f 100644

The Synopsys xHC has an internal TRB cache of size TRB_CACHE_SIZE for each endpoint. The default value for TRB_CACHE_SIZE is 16 for SS and 8 for HS. The controller loads and updates the TRB cache from the transfer ring in system memory whenever the driver issues a start transfer or update transfer command.

For chained TRBs, the Synopsys xHC requires that the total amount of bytes for all TRBs loaded in the TRB cache be greater than or equal to 1 MPS. Or the chain ends within the TRB cache (with a last TRB).

If this requirement is not met, the controller will not be able to send or receive a packet and it will hang causing a driver timeout and error.

This can be a problem if a class driver queues SG requests with many small-buffer entries. The XHCI driver will create a chained TRB for each entry which may trigger this issue.

This patch adds logic to the XHCI driver to detect and prevent this from happening.

For every (TRB_CACHE_SIZE - 2), we check the total buffer size of the SG list and if the last window of (TRB_CACHE_SIZE - 2) SG list length and we don't make up at least 1 MPS, we create a temporary buffer to consolidate full SG list into the buffer.

We check at (TRB_CACHE_SIZE - 2) window because it is possible that there would be a link and/or event data TRB that take up to 2 of the cache entries.

We discovered this issue with devices on other platforms but have not yet come across any device that triggers this on Linux. But it could be a real problem now or in the future. All it takes is N number of small chained TRBs. And other instances of the Synopsys IP may have smaller values for the TRB_CACHE_SIZE which would exacerbate the problem.

Signed-off-by Tejas Joglekar <[email protected]>

---
Changes in v2:
- Removed redundant debug messages
- Modified logic to remove unnecessary changes in hcd.c
- Rename the quirk

 drivers/usb/host/xhci-ring.c |   2 +-
 drivers/usb/host/xhci.c      | 125 +++++++++++++++++++++++++++++++++++++++++++
 drivers/usb/host/xhci.h      |   4 ++
 3 files changed, 130 insertions(+), 1 deletion(-)

drivers/usb/host/xhci-ring.c +1 -1

index a78787bb5133..2fad9474912a 100644
@@ -3291,7 +3291,7 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	full_len = urb->transfer_buffer_length;
 	/* If we have scatter/gather list, we use it. */
-	if (urb->num_sgs) {
+	if (urb->num_sgs && !(urb->transfer_flags & URB_DMA_MAP_SINGLE)) {
 		num_sgs = urb->num_mapped_sgs;
 		sg = urb->sg;
 		addr = (u64) sg_dma_address(sg);

drivers/usb/host/xhci.c +22 -0

index fe38275363e0..15f06bc6b1ad 100644
@@ -1256,6 +1256,106 @@ EXPORT_SYMBOL_GPL(xhci_resume);
 
 /*-------------------------------------------------------------------------*/
 
+static int xhci_map_temp_buffer(struct usb_hcd *hcd, struct urb *urb)
+{
+	void *temp;
+	int ret = 0;
+	unsigned int len;
+	unsigned int buf_len;
+	enum dma_data_direction dir;
+	struct xhci_hcd *xhci;
+
+	xhci = hcd_to_xhci(hcd);
+	dir = usb_urb_dir_in(urb) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
+	buf_len = urb->transfer_buffer_length;
+
+	temp = kzalloc_node(buf_len, GFP_ATOMIC,
+			    dev_to_node(hcd->self.sysdev));
+
+	if (usb_urb_dir_out(urb))
+		len = sg_pcopy_to_buffer(urb->sg, urb->num_sgs,
+					 temp, buf_len, 0);
+
+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,

drivers/usb/host/xhci.c +56 -0

+	}
+
+	return ret;
+}
+
+static bool xhci_urb_temp_buffer_required(struct usb_hcd *hcd,
+					  struct urb *urb)
+{
+	bool ret = false;
+	unsigned int i;
+	unsigned int len = 0;
+	unsigned int buf_len;
+	unsigned int trb_size;
+	unsigned int max_pkt;
+	struct scatterlist *sg;
+	struct scatterlist *tail_sg;
+
+	sg = urb->sg;
+	tail_sg = urb->sg;
+	buf_len = urb->transfer_buffer_length;
+	max_pkt = usb_endpoint_maxp(&urb->ep->desc);
+
+	if (!urb->num_sgs)
+		return ret;
+
+	if (urb->dev->speed >= USB_SPEED_SUPER)
+		trb_size = TRB_CACHE_SIZE_SS;
+	else
+		trb_size = TRB_CACHE_SIZE_HS;
+
+	if (urb->transfer_buffer_length != 0 &&
+	    !(urb->transfer_flags & URB_NO_TRANSFER_DMA_MAP)) {
+		for_each_sg(urb->sg, sg, urb->num_sgs, i) {
+			len = len + sg->length;
+			if (i > trb_size - 2) {
+				len = len - tail_sg->length;
+				if (len < max_pkt) {
+					ret = true;
+					break;
+				}
+
+				tail_sg = sg_next(tail_sg);
+			}
+		}
+	}
+	return ret;
+}
+
+static void xhci_unmap_temp_buf(struct urb *urb)
+{
+	struct scatterlist *sg;
+	unsigned int len;
+	unsigned int buf_len;
+
+	sg = urb->sg;
+	buf_len = urb->transfer_buffer_length;

Quoted 90 lines · Tejas Joglekar

The Synopsys xHC has an internal TRB cache of size TRB_CACHE_SIZE for each endpoint. The default value for TRB_CACHE_SIZE is 16 for SS and 8 for HS. The controller loads and updates the TRB cache from the transfer ring in system memory whenever the driver issues a start transfer or update transfer command.

For chained TRBs, the Synopsys xHC requires that the total amount of bytes for all TRBs loaded in the TRB cache be greater than or equal to 1 MPS. Or the chain ends within the TRB cache (with a last TRB).

If this requirement is not met, the controller will not be able to send or receive a packet and it will hang causing a driver timeout and error.

This can be a problem if a class driver queues SG requests with many small-buffer entries. The XHCI driver will create a chained TRB for each entry which may trigger this issue.

This patch adds logic to the XHCI driver to detect and prevent this from happening.

For every (TRB_CACHE_SIZE - 2), we check the total buffer size of the SG list and if the last window of (TRB_CACHE_SIZE - 2) SG list length and we don't make up at least 1 MPS, we create a temporary buffer to consolidate full SG list into the buffer.

We check at (TRB_CACHE_SIZE - 2) window because it is possible that there would be a link and/or event data TRB that take up to 2 of the cache entries.

We discovered this issue with devices on other platforms but have not yet come across any device that triggers this on Linux. But it could be a real problem now or in the future. All it takes is N number of small chained TRBs. And other instances of the Synopsys IP may have smaller values for the TRB_CACHE_SIZE which would exacerbate the problem.

Signed-off-by Tejas Joglekar <[email protected]>

---
Changes in v2:
- Removed redundant debug messages
- Modified logic to remove unnecessary changes in hcd.c
- Rename the quirk

 drivers/usb/host/xhci-ring.c |   2 +-
 drivers/usb/host/xhci.c      | 125 +++++++++++++++++++++++++++++++++++++++++++
 drivers/usb/host/xhci.h      |   4 ++
 3 files changed, 130 insertions(+), 1 deletion(-)

drivers/usb/host/xhci-ring.c +1 -1

index a78787bb5133..2fad9474912a 100644
@@ -3291,7 +3291,7 @@ int xhci_queue_bulk_tx(struct xhci_hcd *xhci, gfp_t mem_flags,
 
 	full_len = urb->transfer_buffer_length;
 	/* If we have scatter/gather list, we use it. */
-	if (urb->num_sgs) {
+	if (urb->num_sgs && !(urb->transfer_flags & URB_DMA_MAP_SINGLE)) {
 		num_sgs = urb->num_mapped_sgs;
 		sg = urb->sg;
 		addr = (u64) sg_dma_address(sg);

drivers/usb/host/xhci.c +22 -0

index fe38275363e0..15f06bc6b1ad 100644
@@ -1256,6 +1256,106 @@ EXPORT_SYMBOL_GPL(xhci_resume);
 
 /*-------------------------------------------------------------------------*/
 
+static int xhci_map_temp_buffer(struct usb_hcd *hcd, struct urb *urb)
+{
+	void *temp;
+	int ret = 0;
+	unsigned int len;
+	unsigned int buf_len;
+	enum dma_data_direction dir;
+	struct xhci_hcd *xhci;
+
+	xhci = hcd_to_xhci(hcd);
+	dir = usb_urb_dir_in(urb) ? DMA_FROM_DEVICE : DMA_TO_DEVICE;
+	buf_len = urb->transfer_buffer_length;
+
+	temp = kzalloc_node(buf_len, GFP_ATOMIC,
+			    dev_to_node(hcd->self.sysdev));
+
+	if (usb_urb_dir_out(urb))
+		len = sg_pcopy_to_buffer(urb->sg, urb->num_sgs,
+					 temp, buf_len, 0);
+
+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,

drivers/usb/host/xhci.c EXPORT_SYMBOL_GPL(xhci_resume); · line 1290

Quoted 10 lines

drivers/usb/host/xhci.c +10 -0

+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

If Greg or Alan don't object then it's fine for me as well.

Quoted 60 lines

drivers/usb/host/xhci.c +60 -0

+	}
+
+	return ret;
+}
+
+static bool xhci_urb_temp_buffer_required(struct usb_hcd *hcd,
+					  struct urb *urb)
+{
+	bool ret = false;
+	unsigned int i;
+	unsigned int len = 0;
+	unsigned int buf_len;
+	unsigned int trb_size;
+	unsigned int max_pkt;
+	struct scatterlist *sg;
+	struct scatterlist *tail_sg;
+
+	sg = urb->sg;
+	tail_sg = urb->sg;
+	buf_len = urb->transfer_buffer_length;
+	max_pkt = usb_endpoint_maxp(&urb->ep->desc);
+
+	if (!urb->num_sgs)
+		return ret;
+
+	if (urb->dev->speed >= USB_SPEED_SUPER)
+		trb_size = TRB_CACHE_SIZE_SS;
+	else
+		trb_size = TRB_CACHE_SIZE_HS;
+
+	if (urb->transfer_buffer_length != 0 &&
+	    !(urb->transfer_flags & URB_NO_TRANSFER_DMA_MAP)) {
+		for_each_sg(urb->sg, sg, urb->num_sgs, i) {
+			len = len + sg->length;
+			if (i > trb_size - 2) {
+				len = len - tail_sg->length;
+				if (len < max_pkt) {
+					ret = true;
+					break;
+				}
+
+				tail_sg = sg_next(tail_sg);
+			}
+		}
+	}
+	return ret;
+}
+
+static void xhci_unmap_temp_buf(struct urb *urb)
+{
+	struct scatterlist *sg;
+	unsigned int len;
+	unsigned int buf_len;
+
+	sg = urb->sg;
+	buf_len = urb->transfer_buffer_length;
+
+	if (usb_urb_dir_in(urb)) {
+		len = sg_pcopy_from_buffer(urb->sg, urb->num_sgs,
+					   urb->transfer_buffer,

drivers/usb/host/xhci.c EXPORT_SYMBOL_GPL(xhci_resume); · line 1356

Quoted 6 lines

drivers/usb/host/xhci.c +6 -0

+					   buf_len,
+					   0);
+	}
+
+	kfree(urb->transfer_buffer);
+	urb->transfer_buffer = NULL;

clear URB_DMA_MAP_SINGLE from urb->transfer_flags?

Everything else looks good do me.
-Mathias

Quoted 16 lines

drivers/usb/host/xhci.c EXPORT_SYMBOL_GPL(xhci_resume); · line 1290

Quoted 12 lines

drivers/usb/host/xhci.c +12 -0

+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

Quoted 4 lines

If Greg or Alan don't object then it's fine for me as well.

@Greg/Alan do you suggest any change for the flag here?

Quoted 20 lines · Tejas Joglekar

Quoted 20 lines

drivers/usb/host/xhci.c EXPORT_SYMBOL_GPL(xhci_resume); · line 1290

Quoted 12 lines

drivers/usb/host/xhci.c +12 -0

+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

If Greg or Alan don't object then it's fine for me as well.

Quoted 1 line

@Greg/Alan do you suggest any change for the flag here?

This requires care, because the core will already have set other flags for this URB. I don't remember the details and I don't have time to check them now. But it wouldn't be at all surprising if the core doesn't expect an URB both to use SG and to use DMA_MAP_SINGLE.

Alan Stern

Hi,

Quoted 23 lines · Alan Stern

Quoted 21 lines · Tejas Joglekar

Quoted 16 lines

drivers/usb/host/xhci.c EXPORT_SYMBOL_GPL(xhci_resume); · line 1290

Quoted 12 lines

drivers/usb/host/xhci.c +12 -0

+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

Quoted 4 lines

If Greg or Alan don't object then it's fine for me as well.

@Greg/Alan do you suggest any change for the flag here?

Quoted 5 lines

This requires care, because the core will already have set other flags for this URB. I don't remember the details and I don't have time to check them now. But it wouldn't be at all surprising if the core doesn't expect an URB both to use SG and to use DMA_MAP_SINGLE.

When this quirk is enabled and it is required to consolidate SG list to single buffer no other flags are set for the URB.

Only question here is
if it is good to use DMA_MAP_SINGLE flag for the consolidated SG list into a single temporary buffer.

Quoted 30 lines · Tejas Joglekar

Hi,

Quoted 28 lines · Alan Stern

Quoted 16 lines · Tejas Joglekar

Quoted 16 lines

drivers/usb/host/xhci.c EXPORT_SYMBOL_GPL(xhci_resume); · line 1290

Quoted 12 lines

drivers/usb/host/xhci.c +12 -0

+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

Quoted 5 lines

Quoted 4 lines

If Greg or Alan don't object then it's fine for me as well.

@Greg/Alan do you suggest any change for the flag here?

This requires care, because the core will already have set other flags for this URB. I don't remember the details and I don't have time to check them now. But it wouldn't be at all surprising if the core doesn't expect an URB both to use SG and to use DMA_MAP_SINGLE.

Quoted 2 lines

When this quirk is enabled and it is required to consolidate SG list to single buffer no other flags are set for the URB.

I don't believe that. What about URB_DMA_MAP_SG or URB_DMA_MAP_PAGE? The USB core certainly would have set one of them already. Not only would you need to clear that flag, you want also need to undo the mapping.

Quoted 3 lines

Only question here is
if it is good to use DMA_MAP_SINGLE flag for the consolidated SG list into a single temporary buffer.

If you call dma_map_single() then USB_DMA_MAP_SINGLE is the appropriate flag to sest.

Alan Stern

Hi,

Quoted 31 lines · Alan Stern

Quoted 30 lines · Tejas Joglekar

Hi,

Quoted 28 lines · Alan Stern

Quoted 16 lines · Tejas Joglekar

Quoted 16 lines

Review

Quoted 12 lines

(quoted diff) +12 -0

+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

Quoted 5 lines

Quoted 4 lines

If Greg or Alan don't object then it's fine for me as well.

@Greg/Alan do you suggest any change for the flag here?

This requires care, because the core will already have set other flags for this URB. I don't remember the details and I don't have time to check them now. But it wouldn't be at all surprising if the core doesn't expect an URB both to use SG and to use DMA_MAP_SINGLE.

Quoted 8 lines

Quoted 2 lines

When this quirk is enabled and it is required to consolidate SG list to single buffer no other flags are set for the URB.

I don't believe that. What about URB_DMA_MAP_SG or URB_DMA_MAP_PAGE? The USB core certainly would have set one of them already. Not only would you need to clear that flag, you want also need to undo the mapping.

URB_DMA_MAP_SG or URB_DMA_MAP_PAGE flags are set for an URB in core within function usb_hcd_map_urb_for_dma(), with my patch the xhci_map_urb_for_dma() does not call usb_hcd_map_urb_for_dma() when SG list consolidation is required but it calls the newly added function xhci_map_temp_buffer() within that function only dma_map_single() is called to map consolidated SG list.

Quoted 7 lines

Quoted 3 lines

Only question here is
if it is good to use DMA_MAP_SINGLE flag for the consolidated SG list into a single temporary buffer.

If you call dma_map_single() then USB_DMA_MAP_SINGLE is the appropriate flag to sest.

Yes, within xhci_map_temp_buffer(), after consolidating the SG list in a temp buffer I call dma_map_single() to map the temp buffer.

Quoted 2 lines

Alan Stern

Thanks & Regards,
Tejas Joglekar

Quoted 41 lines · Tejas Joglekar

Hi,

Quoted 31 lines · Alan Stern

Quoted 30 lines · Tejas Joglekar

Hi,

Quoted 28 lines · Alan Stern

Quoted 16 lines · Tejas Joglekar

Quoted 16 lines

Review

Quoted 12 lines

(quoted diff) +12 -0

+	urb->transfer_buffer = temp;
+	urb->transfer_dma = dma_map_single(hcd->self.sysdev,
+					   urb->transfer_buffer,
+					   urb->transfer_buffer_length,
+					   dir);
+
+	if (dma_mapping_error(hcd->self.sysdev,
+			      urb->transfer_dma)) {
+		ret = -EAGAIN;
+		kfree(temp);
+	} else {
+		urb->transfer_flags |= URB_DMA_MAP_SINGLE;

Not sure if using URB_DMA_MAP_SINGLE to flag that this urb is boucebuffered is appropriate.

Quoted 5 lines

Quoted 4 lines

If Greg or Alan don't object then it's fine for me as well.

@Greg/Alan do you suggest any change for the flag here?

This requires care, because the core will already have set other flags for this URB. I don't remember the details and I don't have time to check them now. But it wouldn't be at all surprising if the core doesn't expect an URB both to use SG and to use DMA_MAP_SINGLE.

Quoted 8 lines

Quoted 2 lines

When this quirk is enabled and it is required to consolidate SG list to single buffer no other flags are set for the URB.

I don't believe that. What about URB_DMA_MAP_SG or URB_DMA_MAP_PAGE? The USB core certainly would have set one of them already. Not only would you need to clear that flag, you want also need to undo the mapping.

Quoted 5 lines

URB_DMA_MAP_SG or URB_DMA_MAP_PAGE flags are set for an URB in core within function usb_hcd_map_urb_for_dma(), with my patch the xhci_map_urb_for_dma() does not call usb_hcd_map_urb_for_dma() when SG list consolidation is required but it calls the newly added function xhci_map_temp_buffer() within that function only dma_map_single() is called to map consolidated SG list.

Ah, in your patch xhci-hcd's map_urb_for_dma() routine bypasses usb_hcd_map_urb_for_dma(). I hadn't noticed that.

Then yes, your use of the URB flags is okay.

Alan Stern