Distribution: ArchLinux Hardware Environment: Thinkpad A31p Software Environment: i686 since upgrade to 2.6.9-mm1 PCMCIA is broken ("0 sockets found") dmesg says this: Device 'i823650' does not have a release() function, it is broken and must be fixed.
i found that 2.6.10-rc1-mm2 is out ... compiling it right now ... hope it solves this
rc3-mm1 ... still the same trouble! this bug must be solved for 2.6.10, as PCMCIA is very essential for some users (like me)
the config used for 2.6.10-rc3-mm1 can be found here: http://daperi.freelinuxhost.com/archlinux/lns/kernel26mm/config
please try a "modprobe yenta_socket" before starting the pcmcia init script. It seems to me you're using the wrong PCMCIA socket driver. Possibly the PCIC setting in the pcmcia init script configuration file needs updating.
Addendum: I'll take a look at the release function, but that doesn't seem to be critical.
on comment #4: i'll try yenta_socket on next reboot ...
yenta_socket works fine ... on pcmcia-start the i82365 complains on a non-existent device and the nonexistent release() but this is now does not concern me much, as pcmcia works again (note: under 2.6.9, it worked without having to load yenta_socket --- so most probably you are right about the script conf that needs to be upgraded ... archlinux is using pcmcia-cs 3.2.7 atm )
*** Bug 4302 has been marked as a duplicate of this bug. ***
This warning is a reminder to developers that i82365 is doing something which may be a problem in some circumstances with the driver model. However, there is no clean way at present to resolve this (and merely quietening down the warning by providing an empty release function is not acceptable.) Brodo: essentially, the i82365 platform device must not be statically allocated. There's a function (platform_device_register_simple) which dynamically allocates the device structure, but unfortunately you can't attach anything other than resources to it.
an alternative to loading yenta_socket by hand is editing /etc/conf.d/pcmcia and changing the "PCIC=i82365" line to be "PCIC=yenta_socket" (kernel >2.6.9)
Additionally, when i82365.ko fails to load, 2 port area at i365_base may remain allocted (it is allocted in isa_probe, but not deallocated when isa_probe does not find anything). The result is kernel OOPSes later when accessing /proc/ioports.
Created attachment 6525 [details] patch for 2.6.14-git-current
*** Bug 5657 has been marked as a duplicate of this bug. ***
Dominik, could you submit your patch for 2.6.14.4?
No, not this one -- in fact, it was another bug which was triggered, and I sent that one (instead of this mostly harmless bug) to the stable team.