Re: WARNINGs in usb-serial.c

[Alan Cox <alan@linux.intel.com>]

On Fri, 4 Sep 2009 10:17:30 -0400 (EDT)
Alan Stern <stern@rowland.harvard.edu> wrote:

> On Fri, 4 Sep 2009, Miklos Szeredi wrote:
> 
> > On Thu, 3 Sep 2009, Greg KH wrote:
> > > On Thu, Sep 03, 2009 at 04:13:34PM +0200, Miklos Szeredi wrote:
> > > > I get the following warnings in 2.6.31-rc.  A lot of them:
> > > > 
> > > > WARNING: at drivers/usb/serial/usb-serial.c:414
> > > > serial_write_room+0x59/0x6e [usbserial]() WARNING: at
> > > > drivers/usb/serial/usb-serial.c:401 serial_write+0x77/0x9d
> > > > [usbserial]()
> > > > 
> > > > This is a huawei-3565 modem.
> > > > 
> > > > The messages don't seem to hurt, but there's lot of trouble
> > > > with this gadget.  Sometimes it just doesn't work after a while
> > > > and a reboot is needed.  Sometimes it hangs the kernel after
> > > > suspend, etc.  Not sure if this is related to the warnings...
> > > 
> > > {sigh}
> > > 
> > > The tty layer changes are being a pain here :(
> > > 
> > > Alan Stern posted a set of patches to the linux-usb list to
> > > hopefully address stuff like this, is there any way you could
> > > test them out to see if they help or not?
> > 
> > Nope, it doesn't seem to help.  Still got the same warnings with the
> > patched kernel (latest git + ghk tree + usbserial patches).
> 
> What were you doing when these warnings appeared?

The usual way to get it was

	close
		users = 0
		ldisc processing
		echo
		write room
			hey we are closed
				Spew

I suspect the actual check in write_room is now simply bogus with it
all being refcounted (the ldisc will hold a tty kref at that point)