"Shannon C. Dealy" wrote: > > On Wed, 17 Nov 2004, Robert M. Solovay wrote: > > > On Tue, 16 Nov 2004, Shannon C. Dealy wrote: > > > > . Now if they could just fix the [EMAIL PROTECTED] problems with > > USB in the latest 2.6 kernels. > > > > What are these problems? I was planning to upgrade to Fedora > > Core 3 [so that I can use Firefox.] But if one can't use > > USB 2.0 that's a showstopper. > > With kernel 2.6.8, I can't get it to work with either of my > external USB 2.0 drives (two different vendors), after some > digging I was able to find some discussion of the problem, > which was basically one or more people involved in the USB > core code claiming that vendors are violating USB specs and > that their code expects compliance. Don't know if this is > actually the case (and don't currently have time to check), > but, ALOT of people are apparently having problems with this, > late 2.4 and early 2.6 kernels did work with USB 2.0 (i.e. > both of my drives), and regardless of whether or not their > assessment is correct, their handling of this needs to be > changed.
I've heard this claim oodles of times on [EMAIL PROTECTED], which I've been following since 2001. I'm also familiar with the qualification process for USB. Basically if it works under Windoze then it is qualified. And since vendors can release special drivers for their products, they can (and do) work around any flaws in their implementation. So very likely the claims are correct. For example I understand that Windoze doesn't bother to verify the CRC from the packets. Products that get this wrong are broken, but Windoze doesn't flag them. When folks go to the vendor and point out that they have this wrong in their firmware, they reply that "it works under Windoze", to it is by definition OK. "What's your problem?" > Overall, I have never been overly impressed with > the USB drive handling, they keep forgetting to put timeouts > and error recovery in their code, it should NEVER be necessary > to reboot the system to clear an error, particularly in a > module loaded at run time (worst case you should simply be > able to unload and reload the module), but unfortunately, USB > handling has problems in this area and a reboot is the only > cure (even in 2.4 kernels though it doesn't happen to me very > often). If I had any time I'd fix it myself as this kind of > problem shouldn't be that complicated, it's mostly just the > time it takes to learn someone else's code that's the killer. I wish you had the time! I have neither the time nor the skills. I agree that USB is not handled well. In their defense, it's not suprising considering what they have to work with. Altho the USB device (i.e. peripheral) spec is publically available, the USB host spec is only available to a half a dozen companies. Of course, that doesn't excuse screwing up the running OS so that it requires a reboot. > Anyway, I'm back using 2.4 kernels, 2.6 still isn't ready for > prime time, power management is also at least partially broken, > and I've seen other problems. Frankly, 2.6.4 seemed to work > better than 2.6.8, but user-mode-linux (UML) didn't work with > the older 2.6 kernels and I use UML heavily. > > FWIW. > > Shannon C. Dealy | DeaTech Research Inc. Perhaps its just as well that I wasn't able to compile the 2.6.8 kernel. :-) -- Allen Brown work: Agilent Technologies non-work: http://www.peak.org/~abrown/ [EMAIL PROTECTED] [EMAIL PROTECTED] Isn't it a bit unnerving that doctors call what they do "practice"? --- George Carlin _______________________________________________ EUGLUG mailing list [EMAIL PROTECTED] http://www.euglug.org/mailman/listinfo/euglug
