Re: kernel locks due to USB I/O

[Alan Stern <stern@rowland.harvard.edu>]

On Mon, Nov 16, 2020 at 11:53:38AM -0500, Alberto Sentieri wrote:
> The objective of this email is to report the current status of my findings.
> 
> I loaded netconsole on both machines I was having problems with. I tried 3
> times on the machine with kernel 5.0.0-37 and twice with on the machine with
> kernel 5.3.0-62. Each attempt consisted of running the program which lock
> the kernel until it locked (about 3 minutes after stating the program). The
> referred program had the "semphore code" commented out. Nothing was sent to
> netconsole on all the 5 attempts I made when the kernel locked.
> 
> Just to be clear about my use of netconsole, before loading the netconsole
> kernel module, I ran "dmesg -n 8". When netconsole module was loaded I could
> clearly see about 9 message lines on the computer receiving the netconsole
> messages telling me that netconsole was loaded (and how it was configured),
> so no doubts about the correct netconsole setup. The "netconsole server" was
> a machine on the same local network.
> 
> My next attempt will be to compile kernel 5.9, as you suggest, and try it.

While this may not generate any useful information, one way to create a 
bunch of log output while running your test is to set the usbfs_snoop 
module parameter for usbcore to true.

You can also enable dynamic debugging for usbcore, although in a stable 
environment like yours it probably won't produce much output.

Alan Stern