suspend can take minutes with yenta_socket loaded

This is the output of dmesg including a couple of suspend-resume cycles with yenta_socket blacklisted and not loaded.

If you're a user who just wants a fast suspend, here's how you can work around it. It does disable your PCMCIA socket, however.
1. Create a file /etc/modprobe.d/pcmcia-suspend-workaround (the exact name is not important) with contents: blacklist yenta_socket
2. Reboot
This prevents loading of the PCMCIA kernel drivers, so it also prevents the suspend delay bug.

This may be related: http://lists.infradead.org/pipermail/linux-pcmcia/2003-October/000323.html
someone running kernel 2.6.0 where yenta_socket prevents the machine (IBM ThinkPad T20) from suspending; no fix though

dmesg from suspend-resume cycle with yenta_socket loaded, first time

dmesg from suspend-resume cycle with yenta_socket loaded, second time

dmesg from suspend-resume cycle with yenta_socket loaded, third time
this time the suspend delay was some tens of seconds, while the second time it took more than a minute

it seems that the suspend-resume logs are not too different; btw http://meld.sourceforge.net/ can diff three files at once

Oops, I made a mistake in dmesg-susp-y01: it contains the last part of a suspend-resume cycle without yenta_socket. This file has the first part deleted.

With Gutsy's kernel 2.6.22-8-generic this problem doesn't appear to occur anymore. Also, the pcmcia-related kernel message "unable to remove socket power" has vanished.

Doesn't appear anymore in kernel 2.6.22.