Volker Armin Hemmann <[email protected]> posted [email protected], excerpted below, on Tue, 16 Dec 2008 03:02:10 +0100:
> On Dienstag 16 Dezember 2008, Volker Armin Hemmann wrote: >> On Dienstag 16 Dezember 2008, Duncan wrote: >> > David Relson <[email protected]> posted >> > 20081215074756.33119...@fit-pc, excerpted below, on Mon, 15 Dec 2008 >> > >> > 07:47:56 -0500: >> > > FWIW, I've had the board for a year or so. I learned when I got it >> > > that the USB support for the SB600 is bad (based on discussions >> > > with the kernel USB mailing list). I worked around that problem >> > > with a PCI USB card. >> > >> > 2.6.28 I believe has the fix for that. If not, .29 will. There's >> > definitely a patch ready, I know that, as I've been following a bug >> > based on someone here having that issue. >> >> the fix is in 2.6.27.8, .9, 2.6.28-rc-something > > em, the fix for sb700. Don't know about sb600. Yes, the fix for the sb700 USB was the breakthru on the sb600 as well. Apparently someone who could code kernel got an sb700, so that's why it got the fix first. Then someone on the sb600 noted the sb700 fix, and someone from AMD/ATI got involved, and next thing I saw, there was a patch for the sb600 as well. From the discussion, I believe it's basically taking the sb700 patch and generalizing it a bit, then making it apply to all the sb600s and conditionally to the early sb700s (the later ones apparently have the bug fixed in hardware). The sb700 patch is therefore a bit ahead of the more generalized patch that applies to both, and is in 2.6.27 stable now as well as 2.6.28-rc- whatever. I'm not sure if the more generalized patch has/will make it into .28 or not, but for sure it'll be in .29 (and probably in .28 stable series), if it doesn't make it into .28. That's from memory. Since I'm posting a followup, let me go try to find the bug to link, and refresh my memory... ... It looks like my memory was correct. As of the last comment on the bug, the sb600 patch hadn't been applied to mainline yet, tho it had been accepted by Greg KH. There's a report that with the patch applied, a direct no-hub transfer now pauses, resets the USB root hub, and then resumes. That's slightly worse on a direct connect, but much better with a hub since it actually works with a hub now where it wouldn't before. So it'll probably go in, but they might be testing and tweaking a bit more to see if they can fix that last hitch before going mainline with it. Meanwhile, there are several people confirming it does indeed fix the problem with a hub on the sb600. Here's the bug link. You may wish to CC yourself, to get updates. As I said, you'll see I already am CCed. http://bugzilla.kernel.org/show_bug.cgi?id=11599 -- Duncan - List replies preferred. No HTML msgs. "Every nonfree program has a lord, a master -- and if you use the program, he is your master." Richard Stallman
