I have a Transmonde Vivante laptop with a PCMCIA modem running Debian 2.1. I recompiled kernel 2.0.36 to use apm, and recompiled the kernel modules as well. For the most part, my modem is fine. If I type 'cardctl suspend', all is well. However, if I then type
% cardctl resume; cardctl resume then my machine crashes. (I think cardctl forks to the background, so this is like running two copies at about the same time.) The crashes come in different flavors, but almost always happen. Occasionally I get what is probably the correct behavior, namely an error message "ioctl(): Device or resource busy". But most of the time I get a crash. Usually thousands of lines of numbers scroll by (if I'm logged into a virtual terminal when running cardctl), including things like "Call Trace: ..." and "Code: ..." and "Aiee, killing interrupt handler". Then the machine freezes, and won't respond to anything but the power off switch. Once I managed to get syslog to capture a bit of info. In my /var/log/syslog is May 8 19:36:03 scratchy kernel: out during reset May 8 19:36:03 scratchy kernel: cs: socket 0 timed out during reset (This is followed by 14 null characters, and then a fragment of the directory listing of /dev. This probably indicates some filesystem corruption from the crash.) In case you are wondering how I discovered this, it comes up because of the way the apmd package is set-up. In /etc/apm/resume.d/pcmcia is a script which runs 'cardctl resume' if a file /etc/pcmcia/apm.opts exists and contains the line "APM=suspend". It occured to me that I should suspend and resume my cards when suspending and resuming my machine, so I created a file apm.opts containing that line. This is when the crashes started. I think what is happening is that apmd is running the equivalent of 'cardctl resume' already. Thus I think that those scripts should be removed from the distribution, as they cause serious problems when used. Please correct me if I am misunderstanding something. The more fundamental problem is with cardctl, which should never cause the machine to crash. Could people test this with their machine? I recommend unmounting unused filesystems and typing sync before doing so. Anyways, the workaround is simply to not run two copies of 'cardctl resume' at the same time, but I thought I would report this to the list in the hopes that the root of the problem is fixed. Comments or requests for further information, to the list or directly to me, are welcome. Should I submit this to the bug-tracking system? I've never done this -- is it self-explanatory? Dan -- Dan Christensen [EMAIL PROTECTED]
