Thread (17 messages) 17 messages, 2 authors, 2021-01-28

Re: USB2 / USB3 compatibility problems: xhci_hcd 0000:00:06.0: WARN Wrong bounce buffer write length: 0 != 512

From: Andreas Hartmann <hidden>
Date: 2021-01-27 12:26:24

Hello Mathias!

On 26.01.21 at 18:29 Mathias Nyman wrote:
quoted hunk ↗ jump to hunk
quoted
quoted
I'm not sure if it's important for you to know: The driver doesn't use struct scatterlist or num_mapped_sgs at all (if it's meant to be used by the sender at all).

But it sets URB_NO_TRANSFER_DMA_MAP (for data transfer among others).

Mlme packets are sent w/o bulk and w/o setting URB_NO_TRANSFER_DMA_MAP. All other packets are sent with URB_NO_TRANSFER_DMA_MAP turned on.
Ok, thanks, I see what's going on here.

Short recap of xhci alignment requirements.
1. Data pointed to by a transfer request block (TRB) may not span 64k boundary
2. If a transfer contains several TRBs, and spans over two TRB ringbuffer
   segments, then the sum of the TRB data in the first segment must be a 
   multiple of max packets in size.
 
Code assumes that if transfer is split into several blocks,(TRBs) and a block
in the middle of a transfer is smaller than max packet size, then it must be sg list.

But this is not necessarily the case if data was already DMA mapped beforehand.
Data might start just before a 64k boundary, causing first TRB to be less than
packet size.

I'll start implementing a fix for this.
Got a first iteration ready,
any change you could try it out?

Thanks
-Mathias

8<---
diff --git a/drivers/usb/host/xhci-ring.c b/drivers/usb/host/xhci-ring.c
index 5677b81c0915..8df30618aaf1 100644
--- a/drivers/usb/host/xhci-ring.c
+++ b/drivers/usb/host/xhci-ring.c
@@ -699,11 +699,16 @@ static void xhci_unmap_td_bounce_buffer(struct xhci_hcd *xhci,
 	dma_unmap_single(dev, seg->bounce_dma, ring->bounce_buf_len,
 			 DMA_FROM_DEVICE);
 	/* for in tranfers we need to copy the data from bounce to sg */
-	len = sg_pcopy_from_buffer(urb->sg, urb->num_sgs, seg->bounce_buf,
-			     seg->bounce_len, seg->bounce_offs);
-	if (len != seg->bounce_len)
-		xhci_warn(xhci, "WARN Wrong bounce buffer read length: %zu != %d\n",
-				len, seg->bounce_len);
+	if (urb->num_sgs) {
+		len = sg_pcopy_from_buffer(urb->sg, urb->num_sgs, seg->bounce_buf,
+					   seg->bounce_len, seg->bounce_offs);
+		if (len != seg->bounce_len)
+			xhci_warn(xhci, "WARN Wrong bounce buffer read length: %zu != %d\n",
+				  len, seg->bounce_len);
+	} else {
+		memcpy(urb->transfer_buffer + seg->bounce_offs, seg->bounce_buf,
+		       seg->bounce_len);
+	}
 	seg->bounce_len = 0;
 	seg->bounce_offs = 0;
 }
@@ -3275,12 +3280,16 @@ static int xhci_align_td(struct xhci_hcd *xhci, struct urb *urb, u32 enqd_len,
 
 	/* create a max max_pkt sized bounce buffer pointed to by last trb */
 	if (usb_urb_dir_out(urb)) {
-		len = sg_pcopy_to_buffer(urb->sg, urb->num_sgs,
-				   seg->bounce_buf, new_buff_len, enqd_len);
-		if (len != new_buff_len)
-			xhci_warn(xhci,
-				"WARN Wrong bounce buffer write length: %zu != %d\n",
-				len, new_buff_len);
+		if (urb->num_sgs) {
+			len = sg_pcopy_to_buffer(urb->sg, urb->num_sgs,
+						 seg->bounce_buf, new_buff_len, enqd_len);
+			if (len != new_buff_len)
+				xhci_warn(xhci, "WARN Wrong bounce buffer write length: %zu != %d\n",
+					  len, new_buff_len);
+		} else {
+			memcpy(seg->bounce_buf, urb->transfer_buffer + enqd_len, new_buff_len);
+		}
+
If I'm understanding it correctly, you're always creating a bounce
buffer though it is not necessary (at least in my case - my test patch
proofed, that no changes at all are necessary). Why aren't you checking
for URB_NO_TRANSFER_DMA_MAP at the very beginning? Or is it your purpose
to first basically test your new code path? That would be ok.

I tested with the notebook (in both directions) - it seems to work - I
didn't get any problems though I used 24 kB bulk packets. Throughput was
unaltered high.

I'm doing the same test tomorrow with the other USB 3.1 controller!
 		seg->bounce_dma = dma_map_single(dev, seg->bounce_buf,
 						 max_pkt, DMA_TO_DEVICE);
 	} else {
Thanks
Andreas
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help