Messages like this, devoted to programming issues, generally belong on the linux-usb-devel mailing list. Relatively few developers follow linux-usb-users.
Alan Stern P.S.: Have you checked to see whether this problem has already been resolved in 2.6.9-rc1? ---------- Forwarded message ---------- Date: Fri, 3 Sep 2004 21:44:16 -0500 (EST) From: [EMAIL PROTECTED] To: [EMAIL PROTECTED] Subject: [Linux-usb-users] OHCI bug introduced in 2.6.7-rc1 With these ingredients: HP OfficeJet 4110 printer/scanner combo hpoj 0.91 (hpoj.sourceforge.net) SiS OHCI controller libsane 1.0.7-4 (debian package) libusb-0.1-4 1:0.1.5-4 (debian package) Since 2.6.7-rc1, the scanner stopped working. Attempting to use it results in this stuff being logged: usb 1-1: usbfs: USBDEVFS_BULK failed ep 0x85 len 4096 ret -110 hub 1-0:1.0: port 1 disabled by hub (EMI?), re-enabling... usb 1-1: usbfs: USBDEVFS_BULK failed ep 0x85 len 4096 ret -110 usb 1-1: USB disconnect, address 2 The first line repeats several times, then the last few lines come, it all happens very quickly. There is an odd workaround: if I unload and reload ohci_hcd, it works. ohci_hcd must be loaded twice before it will work. I traced the bug to the periodic_reinit() function in drivers/usb/host/ohci.h, and sprinkled it with printk's, which revealed that on the first module load, ohci->fminterval is 0xbff82edf, and on the second load it is 0x27782edf. >From there, I followed the trail to hc_reset() in drivers/usb/host/ohci-hcd.c, where ohci->fminterval gets initialized. Sprinkling that one with printk's, I found that the difference comes from the following line: temp = ohci_readl (&ohci->regs->fminterval); On the first module load, temp is 0xbff82edf. On the second module load, it is 0x00002edf. Replacing the first line of periodic_reinit() by its counterpart from 2.6.6 fixes the bug: /*GOOD:*/ writel ((((6 * (FI - 210)) / 7) << 16) | FI, &ohci->regs->fminterval); /*BAD:*/ /*writel (ohci->fminterval, &ohci->regs->fminterval);*/ This was a rough one... I discovered the unload-reload workaround by applying the 2.6.7-rc1 piece-by-piece, reloading the modules after recompiling them, without knowing that I needed to reboot in order to trigger the bug again. ------------------------------------------------------- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=5047&alloc_id=10808&op=click _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-users ------------------------------------------------------- This SF.Net email is sponsored by BEA Weblogic Workshop FREE Java Enterprise J2EE developer tools! Get your free copy of BEA WebLogic Workshop 8.1 today. http://ads.osdn.com/?ad_id=5047&alloc_id=10808&op=click _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: https://lists.sourceforge.net/lists/listinfo/linux-usb-devel
