Re: kern/21547 (Reading short block from ugen device can lock up USB)

[Andreas Gustafsson <gson%gson.org@localhost>]

The following reply was made to PR kern/21547; it has been noted by GNATS.

From: Andreas Gustafsson <gson%gson.org@localhost>
To: dholland%NetBSD.org@localhost
Cc: gnats-bugs%netbsd.org@localhost
Subject: Re: kern/21547 (Reading short block from ugen device can lock up USB)
Date: Tue, 3 Jan 2023 17:41:05 +0200

 dholland%NetBSD.org@localhost wrote:
 > how does this look compared to 2004? it's been a while and vast numbers
 > of fixes have gone into the usb code...
 
 Sadly, the bug is still there.  I no longer have the PC that it
 originally affected, but my main USB test machine contains multiple
 PCI cards with different USB controller chips, and one of them happens
 to be the affected VIA VT83C572.  I booted a NetBSD-current/amd64
 kernel built from CVS source date 2023.01.03.00.00.45, and it still
 behaves as reported in the original PR.
 
 To be precise, this card has four ports and two UHCI companion
 controllers each shared by two ports, and a short read on one of the
 ports causes that port and the one sharing the same UCHI companion
 controller to lock up, but the other two ports remain functional until
 a short read is done on one of them, too.
 
 The bug is of little practical significance at this point as the
 VT83C572 is thoroughly obsolete by now, but in any case, the offer I
 made in the original PR to ship a USB device suitable for reproducing
 the bug to anyone serious about looking into it still stands.
 VT83C572 USB controller PCI cards can still be found on eBay.
 -- 
 Andreas Gustafsson, gson%gson.org@localhost