2.6.26 Breaks USB Fax Modem

2.6.26 Breaks USB Fax Modem

[Frank Peters]

Hello,

After updating the kernel from 2.6.25 to 2.6.26, I now experience
severe problems when trying to send a fax using an external USB
modem.  Using either HylaFAX or efax facsimile software, the fax
transmission will just suddenly hang at some point before completion.
Neither the fax log or the kernel log will show any error or other
anomaly.  Rebooting the system with kernel 2.6.25 will allow me
to send faxes without problem.

Nothing on my machine has changed between 2.6.25 and 2.6.26 and
the same configuration was used to compile both versions. 

The cdc-acm module is responsible for communicating with the USB
modem and I suspect that it is at fault.

Has this problem been reported by anyone else?  How can I further
debug this problem?  As I mentioned, no log shows any error message
of any kind.  The logs from the fax software may indicate an
"unspecified error" but will usually just show that the data
stream has been abruptly halted.

Please cc to me directly since I am not subscribed to the list.

Frank Peters

Re: 2.6.26 Breaks USB Fax Modem

[Andrew Morton]

On Tue, 19 Aug 2008 21:10:39 -0400 Frank Peters <frank.peters@comcast.net> wrote:

> Hello,
> 
> After updating the kernel from 2.6.25 to 2.6.26, I now experience
> severe problems when trying to send a fax using an external USB
> modem.  Using either HylaFAX or efax facsimile software, the fax
> transmission will just suddenly hang at some point before completion.
> Neither the fax log or the kernel log will show any error or other
> anomaly.  Rebooting the system with kernel 2.6.25 will allow me
> to send faxes without problem.
> 
> Nothing on my machine has changed between 2.6.25 and 2.6.26 and
> the same configuration was used to compile both versions. 
> 
> The cdc-acm module is responsible for communicating with the USB
> modem and I suspect that it is at fault.
> 
> Has this problem been reported by anyone else?  How can I further
> debug this problem?  As I mentioned, no log shows any error message
> of any kind.  The logs from the fax software may indicate an
> "unspecified error" but will usually just show that the data
> stream has been abruptly halted.
> 
> Please cc to me directly since I am not subscribed to the list.
> 

(cc linux-usb)

Are any processes left hanging in `D' state afterwards?

If so, please run

	echo w > /proc/sysrq-trigger
	dmesg -s 1000000 > foo

then send foo.

Thanks.

Re: 2.6.26 Breaks USB Fax Modem

[Frank Peters]

On Wed, 20 Aug 2008 07:39:51 +0200
Oliver Neukum <oliver@neukum.org> wrote:

> 
> This is completely new. Please compile 2.6.26 with CONFIG_USB_DEBUG
> and reexamine the log. It should be more verbose.
> 

In menuconfig, I enabled the USB_DEBUG option and then compiled
kernel 2.6.26.2.  The kernel log during a fax transmission is,
to my eyes, still not very revealing.

Before the fax is transmitted, the cdc-acm module is first loaded
with the commands:

modprobe cdc-acm
mount -t usbfs none /proc/bus/usb

Then the fax is sent with HylaFAX software.  (The problem is not
unique to  HylaFAX, however; it will occur with any fax software.)
HylaFAX will initialize the USB modem and then send the fax.  Here
is the kernel log:

Aug 20 10:53:22 (none) kernel: cdc_acm 4-2:2.0: usb_probe_interface
Aug 20 10:53:22 (none) kernel: cdc_acm 4-2:2.0: usb_probe_interface - got id
Aug 20 10:53:22 (none) kernel: cdc_acm 4-2:2.0: ttyACM0: USB ACM device
Aug 20 10:53:22 (none) kernel: usbcore: registered new interface driver cdc_acm
Aug 20 10:53:22 (none) kernel: cdc_acm: v0.26:USB Abstract Control Model driver for USB modems and ISDN adapters
Aug 20 10:53:42 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 10:54:10 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:01:30 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:01:33 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:01:33 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:03:34 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:03:34 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:03:38 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:03:38 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:09:31 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us

Notice the long gap from 11:03:38 to 11:09:31.  During this time the modem
is hung.  Nothing is happening.  The HylaFAX transmission log shows only an
abrupt break in the data stream:

Aug 20 11:04:18.04: [  717]: <-- data [2]
Aug 20 11:04:18.11: [  717]: <-- data [261]
Aug 20 11:04:18.19: [  717]: <-- data [2]
Aug 20 11:04:18.26: [  717]: <-- data [260]
Aug 20 11:04:18.33: [  717]: <-- data [2]
Aug 20 11:04:18.33: [  717]: <-- data [262]
Aug 20 11:05:18.33: [  717]: MODEM TIMEOUT: writing to modem
Aug 20 11:05:18.33: [  717]: <-- data [2]
Aug 20 11:06:18.33: [  717]: MODEM TIMEOUT: writing to modem

The break occurs at 11:04:18.33.  After a minute a MODEM TIMEOUT
is reported.

To get to this point, however, requires a lot of prior communication
with the modem to establish the parameters and protocol. Somehow,
the modem becomes unresponsive only later in the transmission.

HylaFAX will eventually recover but the modem remains unresponsive.
A reboot is necessary to restore the modem functionality.

The kernel log after this point shows only the same:

Aug 20 11:09:34 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:09:38 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:09:38 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:10:11 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:10:11 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:10:42 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:12:56 (none) kernel: uhci_hcd 0000:00:1a.1: reserve dev 2 ep84-INT, period 128, phase 0, 36 us
Aug 20 11:12:59 (none) kernel: uhci_hcd 0000:00:1a.1: release dev 2 ep84-INT, period 128, phase 0, 36 us

If I immediately reboot the machine with kernel 2.6.25, the same fax
can be sent to the same fax machine with no problem.

>
> Are any processes left hanging in `D' state afterwards?
>

The command "ps ax" shows no process in the "D" state at any time.

I did not report this to the HylaFAX list because it does not
appear to be a fault of HylaFAX or other fax software, but I think
I should indicate this thread to the HylaFAX list.  They may
be able to provide some additional insight about the point of
failure.

Frank Peters

Re: 2.6.26 Breaks USB Fax Modem

[Alan Stern]

On Wed, 20 Aug 2008, Frank Peters wrote:

> > This is completely new. Please compile 2.6.26 with CONFIG_USB_DEBUG
> > and reexamine the log. It should be more verbose.

> If I immediately reboot the machine with kernel 2.6.25, the same fax
> can be sent to the same fax machine with no problem.

If all else fails, you could try bisecting the kernels between
2.6.25 and 2.6.26 to locate the problem.

Alan Stern