On Thu, 11 Mar 2004, Matthias Andree wrote:
> Alan,
>
> I have massive difficulties with a USB 1.1 scanner (Epson Perfection 1650) and
> a VIA KT600 based PeeCee mainboard (VT8237 south bridge) in Linux 2.6.
> Linux 2.4 is fine, FreeBSD 5-CURRENT is fine, so hardware trouble seems
> unlikely /to me/.
>
> I have updated the Linux 2.6 kernel from BitKeeper regularly, and I have
> switched from a KT133 board to a KT600 board and upgraded the CPU four weeks
> ago - the problem started around the time I upgraded the machine.
>
> If you think this could have something to do with the UHCI driver, then read
> on, if not, just tell me so, ignore the rest of this mail and apologize my
> intrusion. I've seen you're listed as the UHCI maintainer and as previous posts
> to linux-kernel and linux-usb-devel have been vain endeavor, I hope I
> can make some progress with your help.
>
> If anything of this looks unreasonable, don't hesitate to flame me, I
> can bear harsh criticism if it looks halfway justified and isn't personal.
>
> --- rest only applies if the problem might be related to a driver you maintain ---
>
> I had no troubles with my old board (Gigabyte 7ZXR 1.0, KT133, 82C686A USB
> controller, AMD Duron Spitfire 700 MHz CPU) in Early February.
>
> In Mid February, I found the board suffered from the "electrolytic capacitor"
> problem and I chose to replace the board and got an Asus A7V600-X, KT600, 8237
> USB controller, AMD Athlon XP+ 2500 (Barton core, 1833 MHz processor). At
> around that time for a reason I cannot dissect among these changes (and I don't
> currently have the chance to revert), the problems started. Things have
> improved slowly over the past weeks' worth of "bk pull" and rebuilds, in late
> February, the scanner would often drive its sled into the case, that no longer
> happens now past 2.6.4-rc3.
>
> The visible problem is that Linux 2.6 sees hangs with unkillable ("D"eep sleep
> state in ps(1)) user-space applications such as xsane or iscan. I am using
> libusb for consistency on any of Linux 2.4, 2.6, FreeBSD.
>
> The KT133 used to run the same software only with a kernel of late January or
> early February without scan problems in Linux 2.6. I am sure it worked for me
> on Jan 31st.
>
> With Linux 2.6, I can scan with www.hamrick.com's vuescan, I can sometimes scan
> with sane-frontends (xscanimage, scanimage), I have some trouble with xsane
> although that seems to have improved around 2.6.4-rc3, but I get an immediate
> and reproducible lockup with EPKOWA's iscan 1.7.0 (which is user space only).
>
> All I need to do is switch the scan source in iscan from "Flatbed" to "TPU for
> Pos.Film" and the USB subsystem disconnects the scanner and hangs. After that,
> hotplug processes start piling up as a consequence of the deadlock around the
> unkillable process and the hung USB subsystem.
>
> I can drown you in lengthy dmesg or syslog output, strace, ltrace, whatever -
> what do you need? (BTW the usb_hotplug messages I do not understand, the USB
> configuration - hardware wise - is static and the scanner powered on before
> Linux boots).
>
> I am attaching the log files gzipped as they appear a bit largish (7 kB
> uncompressed).
>
> Any help you could offer is much appreciated, I am willing to help debug in
> turn but need directions. I am a skilled C programmer and know how to operate
> gdb, but I have no relevant USB expertise and have only little experience with
> kernel hacking. The only stuff I did was a devinet.c fix for Linux 2.2 and 2.4
> way back.
>
> Don't hesitate to offer patches or ask for changes in a certain part of the
> code to enable debugging options or add printk or something.
>
> For any help you can offer, thank you in advance.
>
> Yours sincerely,
>From looking at your logs, I get the feeling that the initial problem
doesn't lies in the UHCI driver. The log shows the scanner failing to
accept some commands; in fact it looks like the scanner never received the
commands at all, or the computer never received its reply. That could be
the result of bad hardware or a bad cable, or maybe a bug in the scanner.
It is puzzling, though, because the commands needed to enumerate the
scanner when the USB drivers first detected it must have worked correctly.
Maybe the scanner isn't replying because it doesn't recognize the request?
The third error in the log shows "Babble", this time the result of a bad
interaction between the scanner and its userspace driver. In this case
the results are pretty bad. A lot of VIA UHCI controllers (not all but
many, and yours appear to be one of them) shut themselves off when they
receive babble, so nothing that happens after that can possibly work. At
this point we don't have a solution for this, short of reloading the UHCI
module, which obviously won't help you.
All in all, I think the userspace program may be the biggest problem.
Did you say that the same program works okay on the same hardware when
running under Linux 2.4? What does the dmesg log for 2.4 say? I would
expect it to contain the same sequence of errors.
Alan Stern
-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click
_______________________________________________
[EMAIL PROTECTED]
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
