Re: WARNING in pvr2_i2c_core_done

[Greg Kroah-Hartman]

On Thu, Sep 26, 2019 at 05:44:31PM -0400, Alan Stern wrote:
> On Wed, 25 Sep 2019, Andrey Konovalov wrote:
> 
> > On Wed, Sep 25, 2019 at 4:10 PM Alan Stern <stern@xxxxxxxxxxxxxxxxxxx> wrote:
> > >
> > > On Wed, 25 Sep 2019, syzbot wrote:
> > >
> > > > Hello,
> > > >
> > > > syzbot found the following crash on:
> > > >
> > > > HEAD commit:    d9e63adc usb-fuzzer: main usb gadget fuzzer driver
> > > > git tree:       https://github.com/google/kasan.git usb-fuzzer
> > > > console output: https://syzkaller.appspot.com/x/log.txt?x=16b5fcd5600000
> > > > kernel config:  https://syzkaller.appspot.com/x/.config?x=f4fa60e981ee8e6a
> > > > dashboard link: https://syzkaller.appspot.com/bug?extid=e74a998ca8f1df9cc332
> > > > compiler:       gcc (GCC) 9.0.0 20181231 (experimental)
> > > > syz repro:      https://syzkaller.appspot.com/x/repro.syz?x=16ec07b1600000
> > > > C reproducer:   https://syzkaller.appspot.com/x/repro.c?x=13ff0871600000
> > > >
> > > > IMPORTANT: if you fix the bug, please add the following tag to the commit:
> > > > Reported-by: syzbot+e74a998ca8f1df9cc332@xxxxxxxxxxxxxxxxxxxxxxxxx
> > > >
> > > > pvrusb2: Device being rendered inoperable
> > > > cx25840 0-0044: Unable to detect h/w, assuming cx23887
> > > > cx25840 0-0044: cx23887 A/V decoder found @ 0x88 (pvrusb2_a)
> > > > pvrusb2: Attached sub-driver cx25840
> > > > pvrusb2: ***WARNING*** pvrusb2 device hardware appears to be jammed and I
> > > > can't clear it.
> > > > pvrusb2: You might need to power cycle the pvrusb2 device in order to
> > > > recover.
> > > > ------------[ cut here ]------------
> > > > sysfs group 'power' not found for kobject 'i2c-0'
> > > > WARNING: CPU: 0 PID: 102 at fs/sysfs/group.c:278 sysfs_remove_group
> > > > fs/sysfs/group.c:278 [inline]
> > > > WARNING: CPU: 0 PID: 102 at fs/sysfs/group.c:278
> > > > sysfs_remove_group+0x155/0x1b0 fs/sysfs/group.c:269
> > >
> > > I have seen a lot of error messages like this one (i.e., "group 'power'
> > > not found for kobject"), in runs that involved fuzzing a completely
> > > different USB driver.  Initial testing failed to find a cause.
> > >
> > > This leads me to wonder whether the problem might lie somewhere else
> > > entirely.  A bug in some core kernel code?  Memory corruption?
> > 
> > AFAICS so far this has only been triggered from the usbvision driver
> > [1] and from the pvrusb2 driver (this report).
> > 
> > I wanted to loop in sysfs maintainers, but it seems that Greg and
> > Rafael are already cc'ed on this.
> > 
> > [1] https://syzkaller.appspot.com/bug?extid=7fa38a608b1075dfd634
> 
> It turns out the reason for this error is simple: The driver 
> unregisters its subdevices in the release handler instead of in the 
> disconnect handler.  There probably is documentation about this 
> somewhere, but I don't know exactly where -- maybe Greg remembers.

Nope, I don't remember.  It should happen in the disconnect handler, odd
of it to be in release, but maybe that's the "easiest" way for v4l to
handle this?

thanks,

greg k-h