> Try applying the patch below, and enable CONFIG_USB_DEBUG. Let's see > what the dmesg log shows after the driver stops working -- there should > be a line printed for every OHCI interrupt. > > Alan Stern > > > > Index: usb-3.4/drivers/usb/host/ohci-hcd.c > =================================================================== > --- usb-3.4.orig/drivers/usb/host/ohci-hcd.c > +++ usb-3.4/drivers/usb/host/ohci-hcd.c > @@ -755,6 +755,8 @@ static irqreturn_t ohci_irq (struct usb_ > * work on all systems (edge triggering for OHCI can be a factor). > */ > ints = ohci_readl(ohci, ®s->intrstatus); > + ohci_dbg(ohci, "ohci_irq: ints %04x enable %04x\n", ints, > + ohci_readl(ohci, ®s->intrenable)); > > /* Check for an all 1's result which is a typical consequence > * of dead, unclocked, or unplugged (CardBus...) devices > thanks Alan, sorry for late response but with your patch I'm unable to reproduce the bug. On bare linux-3.3 I can reproduce the bug immediately and always. Enabling CONFIG_USB_DEBUG on this very same kernel makes no difference also. After applying your patch the bug is not present anymore - it disappears. I'm not able to reproduce it or observe. It looks this single read from ohci controller stabilized the controller I/O during interrupt service. The question is how much ohci is stable now. This is what I'm going to test now. thanks again Alan. Thanks to your patch I do not have to use noapic parameter anymore. Raplacing fasteoi apic with level apic is also not needed. BTW I sent today an email to nvidia nforce linux team. They reproduced this bug 2 years ago. I hope they have a patch for kernel to share with us or at least any findings which will help us fixing this. -- To unsubscribe from this list: send the line "unsubscribe linux-usb" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
