On Wed, Jul 15, 2020 at 10:58:03AM +0200, David Guillen Fandos wrote: > Hello linux-usb, > > I think I might have found a kernel bug related to the USB subsystem > (cdc_acm perhaps). > > Context: I was playing around with a device I'm creating, essentially a > USB quad modem device that exposes four modems to the host system. This > device is still a prototype so there's a few bugs here and there, most > likely in the USB descriptors and control requests. > > What happens: After plugging the device the system starts spitting > warnings and BUGs and it locks up. Most of the time some CPUs get into > some spinloop and never comes back (you can see it being detected by > the watchdog after a few seconds). Generally after that the USB devices > stop working completely and at some point the machine freezes > completely. In a couple of ocasions I managed to see a bug in dmesg > saying "unable to handle page fault for address XXX" and "Supervisor > read access in kernel mode" "error code (0x0000) not present page". I > could not get a trace for that one since the kernel died completely and > my log files were truncated/lost. > > Since it is happening to my two machines (both Intel but rather > different controllers, Sunrise Point-LP USB 3.0 vs 8 Series/C220) and > with different kernel versions I suspect this might be a bug in the > kernel. > > I have 4 logs that I collected, they are sort of long-ish, not sure how > to best send them to the list. Send the crashes with the callback list, that should be quite small, right? We don't need the full log. The first crash is the most important, the others can be from the first one and are not reliable. thanks, greg k-h
