Index | Thread | Search

From:
Kirill A. Korinsky <kirill@korins.ky>
Subject:
Re: sys/xhci: avoid clear to zero maxBurst for Enhanced SuperSpeed
To:
Peter Hessler <phessler@theapt.org>
Cc:
tech@openbsd.org
Date:
Fri, 06 Nov 2026 14:06:26 +0100

Download raw body.

Thread
On Fri, 06 Nov 2026 13:54:12 +0100,
Peter Hessler <phessler@theapt.org> wrote:
> 
> On 2026 Nov 06 (Fri) at 11:54:32 +0100 (+0100), Kirill A. Korinsky wrote:
> :tech@,
> :
> :I'm reading xhci and nearby to understand why ure is capped to ~1.5 Gbit and
> :so far I think I had spolied one missed case in xhci.
> :
> :Section 6.2.3.4 of eXtensible Host Controller Interface for Universal Serial
> :Bus states that for for Enhanced SuperSpeed endpoints maxBurts for all
> :endpoints should be setup into bMaxBurst. At least I read it this way. See
> :page 455 here:
> :https://www.intel.com/content/dam/www/public/us/en/documents/technical-specifications/extensible-host-controler-interface-usb-xhci.pdf
> :
> :Am I wrong?
> :
> 
> Reading the spec it says all endpoints (transfer types), so I think the
> check should simply be
> "if (esscd)"
> 
> Or if you want to specifically exclude control endpoints, then use
> "if (esscd && xfertype != UE_CONTROL)"
> 

Thanks for suggestion.

Yes, I think if esscd is that is actually written in the spec.

Index: sys/dev/usb/xhci.c
===================================================================
RCS file: /home/cvs/src/sys/dev/usb/xhci.c,v
diff -u -p -r1.136 xhci.c
--- sys/dev/usb/xhci.c	1 Mar 2025 14:43:03 -0000	1.136
+++ sys/dev/usb/xhci.c	6 Nov 2026 13:05:21 -0000
@@ -1455,9 +1455,9 @@ xhci_pipe_maxburst(struct usbd_pipe *pip
 			maxb = UE_GET_TRANS(mps);
 		break;
 	case USB_SPEED_SUPER:
-		if (esscd &&
-		    (xfertype == UE_ISOCHRONOUS || xfertype == UE_INTERRUPT))
+		if (esscd)
 			maxb = esscd->bMaxBurst;
+		break;
 	default:
 		break;
 	}


-- 
wbr, Kirill