The Answer Gang

By Jim Dennis, Karl-Heinz Herrmann, Breen, Chris, and... (meet the Gang) ... the Editors of Linux Gazette... and You!

---

Kernel 2.6 and PCMCIA troubles

From Antoun Kanawati

Answered By: Thomas Adam, John Karns.

Dear Answer Gang,

The problem is: "/etc/init.d/pcmcia start" reports 0 sockets being watched when use any 2.6 kernel other Mandrake 10.0's 2.6.3-7.

[Thomas] My first suggestion at reading that is Mandrake, like RH (and SuSE to an extent) have patched that kernel to high heaven (read that as 'mangled') such that it is no longer a kernel... grr, stock kernela are evil.

[John] It's been a while since I last dug around for info on the pcmcia subsystem, but last I looked at the pcmcia pkg docs from Hinds - it has been a while, maybe 18 months or more, there were some things that the stand-alone pcmcia pkg did better than the integrated kernel version. His recommendation was that in circumstances where there were problems, one should try compiling the kernel without the pcmcia options, and compile the pkg from source, and run it from a system init script. It might be worth taking a look to see if that situation still holds.

I've had this problem with Fedora 2, Suse 9.1, and Manrdake 10. The only kernel that get my PCMCIA right is 2.6.3-7 from the Mandrake 10 distro.

This happens on two of my laptops, a fujitsu lifebook 765DX (Pentium 166MMX), and an NEC Versa LX (PII-233).

The other possibly related irritant is that "/etc/init.d/pcmcia stop" is not working right; more precisely, when I get the PCMICA card started, the card's lights go on and eth0 is brought up; when I stop the card, eth0 is brought down, but the card remains ON. So, if I reboot, even with kernel 2.6.3-7 mdk 10.0, the next "pcmcia start" fails to notice the two slots. If I power down and then boot, all works fine.

[Thomas] If the pcmcia initscript is the same as it was in Mandrake 7.0 (which was my only fleeting contact with it) then that should actually be calling cardmgr . Cardmgr normally works just fine.

[John] Yes, agreed. Instead of stopping the pcmcia service, try the command

"cardctl eject"

as root. Before I discovered the card mgrs eject command (back in the days of the 2.2 kernels), I used to stop the pcmcia service, and it would sometimes hang, and otherwise misbehave. I found the eject command to be a better option.

To summarize: "/etc/init.d/pcmcia start" find 2 sockets and works with only one very specific kernel. The "stop" doesn't stop all the way.

I looked around the net for a while, but didn't find an answer.

I am guessing that this is one of those "older machine" things that require a slightly exceptional configuration clause somewhere.

[Thomas] Actually, I would be more inclined to say that you should look at compiling your own kernel. Since it works with a specific kernel you can almost certainly rule out hardware issues. So the trick is to look at the specific working kernel's config file for clues. The config files for kernels should be in /boot as:

/boot/config-$(uname -r)

The $(uname -r) interpolates to the current running kernel's version, but the principle is the same -- the file name is /boot/config-<version> . If you can do the following for that kernel version:

grep -i pcmcia /boot/config-<ver>

And send it to us, that would help. Along with that, you should repeat that same command on a file for a kernel that does not work, for comparison.

If you think this is long winded, you'd be right. I would definitely look at compiling your own kernel. It's not that hard, and there's plenty of references to it, here's two:

• http://www.hantslug.org.uk/cgi-bin/wiki.pl?LinuxHints/KernelBuilding
• http://www.kernelnewbies.org

Pointers will be greatly appreciated. Thanks in advance.

[Thomas] Hope that helps.

---

This page edited and maintained by the Editors of Linux Gazette
HTML script maintained by Heather Stern of Starshine Technical Services, http://www.starshine.org/

---