> After my initial post about 5 days ago about random oopses in both 2.4.2 > and 2.4.3, several replies were sent, one by David Brownell about > reverting to kernel 2.4.2, with Alan Cox's last pre-patch (ac28), and > apply pci-pool and ohci patches from 03/23, and enable kernel debugging > and slab poisoning, one from Pete Zaitcev pointing to a patch dealing with > device initialization, and another from Mark McClelland about ov511 v1.35 > being available. > > I took David's approach first, and it did two things, not the least of > which *seemed* to be eliminating the oopses, as it ran for a solid 24 > hours before I started the experiment. But, it also seemed to result in > total garbage from the camera, with mostly green, pink, and blue > "compression squares" the entire JPEG image (from vidcat 0.62). I then > quick upgraded to ov511 v1.35 with the same results. If you got total garbage, that suggests the slab poisoning exposed a bug somewhere in how the I/O buffers are accessed. I'd guess ov511 before an ISO-specific issue with OHCI, but that's just a guess for now. The reason I suggested using that slab poisoning in addition to the current ov511 was specifically in case such problems turned up. Your original oops should still be lurking, just hiding because of the way slab poisoning changed device interaction timing, unless the MediaGX does PCI DMA differently from other x86 systems. > However, with Kernel 2.4.2-ac28 and enabling kernel debugging and slab > poisoning (CONFIG_DEBUG_KERNEL=y, CONFIG_DEBUG_SLAB=y), I load usbcore > first, and then get an immediate oops when loading the usb-ohci module. That's because you needed the other patches (pci_pool and ohci using it) to run correctly with slab debugging. > Also, taking that same compiled kernel (2.4.2-ac28 with kernel debugging > and slab poisoning enabled) it doesn't work correctly either (but doesn't > oops), with this in the message log: > > Apr 7 17:04:31 devel-uhci kernel: usb.c: registered new driver usbdevfs > Apr 7 17:04:31 devel-uhci kernel: usb.c: registered new driver hub > Apr 7 17:04:31 devel-uhci kernel: usb-uhci.c: $Revision: 1.251 $ time 15:56:36 Apr 7 2001 You'd need the patch to let usb-uhci use slab poisoning too ... it's in list archives for March 24 I think. - Dave _______________________________________________ [EMAIL PROTECTED] To unsubscribe, use the last form field at: http://lists.sourceforge.net/lists/listinfo/linux-usb-devel
