Re: Trouble getting udev rule to fire, works in udevtest

[Alan Jenkins <alan-jenkins@tuffmail.co.uk>]

On 7/18/09, Ben Beuchler <insyte@gmail.com> wrote:
> I'm trying to write a rule that detects the unplugging of a particular
> USB device and unloads the kernel module when that happens.  The
> reason is that the module appears to be buggy in that if the device is
> reconnected without first unloading the module, the module hangs and
> the device is not activated.  At this point even forcibly unloading
> the module makes no difference.
>
> The device is a MultiTech GSM modem.   The udevinfo used to build the
> rule is pasted below.  Here is the rule I've created (and installed in
> /etc/udev/rules.d/85-mts_gsm_3410_remove.rules):
>
> SUBSYSTEMS="usb", ACTION="remove", ATTRS{idVendor}="06e0",
> ATTRS{idProduct}="f111", RUN+="modprobe -r ti_usb_3410_5052"
>
> For testing purposes, I changed the modprobe to "echo hello from
> ti_usb_remove >> /tmp/udev.works"
>
> udevtest seems to indicate that the rule works:
>
> # udevtest --action=remove
> /devices/pci0000:00/0000:00:1d.3/usb4/4-1/4-1:2.0/ttyUSB0 | grep -v
> ^parse_file
> [:snipped udevtest explanation:]
> import_uevent_var: import into environment: 'DRIVER=ti_usb_3410_5052_1'
> udevtest: looking at device
> '/devices/pci0000:00/0000:00:1d.3/usb4/4-1/4-1:2.0/ttyUSB0' from
> subsystem 'usb-serial'
> udevtest: run: '/bin/sh -c 'echo 2 >
>
/sys/devices/pci0000:00/0000:00:1d.3/usb4/4-1/4-1:2.0/ttyUSB0/device/bConfigurationValue''
> udevtest: run: 'echo hello from ti_usb_remove >> /tmp/udev.works'
> udevtest: run: 'socket:/org/kernel/udev/monitor'
>
> However if I actually unplug the device, /tmp/udev.works doesn't show
> up.   On the off chance that my kernel doesn't support inotify, I also
> tried running "udevcontrol reload_rules".  Still failed.  The output
> of udevmonitor, captured while unplugging the device, is pasted below.
>  For giggles, I also tried running a "logger hello" command instead of
> the echo.  Does not appear to run.
>
> Any idea what I'm doing wrong?

ATTR won't work on a real remove, because the sysfs files are removed
before udev processes the event.

Apparently you can change the match to ACTION="add" and set
ENV{REMOVE_CMD}="modprobe -r ti_usb_3410_5052" instead.  This feature
was recently mentioned here
<http://article.gmane.org/gmane.linux.hotplug.devel/14221>.

I don't know when REMOVE_CMD was added to the rules.  If your rules are
too old, you should be able implement it yourself.  It is a really
simple idea, you just need another rule to run the command on remove.

ACTION="remove", ENV{REMOVE_CMD}="?*", RUN+="$env{REMOVE_CMD}"

Oh, and this is a pretty major bug in the module, please make sure it
has been reported to the appropriate people :-).

Alan