I tried to update my system to gentoo-sources 2.4.20-r5 and pcmcia-cs 3.2.4. kernel compilation and emerge of pcmcia-cs worked. module load of pcmcia-core, i82365 and ds worked as well. but as soon as a pcmcia nic is in the slot and /etc/init.d/pcmcia has been started the appropriate module fails loading with a message like this: "bind of #module# on socket #socket-no failed: Operation not permitted." I tried a 10mbit, a 100mbit and a wlan nic. none of their modules could be loaded. pcmcia-cs-3.2.1-r* did also fail. some weeks ago I used gentoo-sources-2.4.20-r(1-3). afaik they work ok. I think the problem is in the last gentoo-sources 2.4.20 revision (r5). Sorry, because of the missing network connection I can't really dump here the original syslog messages. Reproducible: Always Steps to Reproduce: 1.emerge gentoo-sources-2.4.20-r5; install and compile kernel+modules (without pcmcia support and with wireless networking, but without wireless drivers) 2.emerge pcmcia-cs-3.2.4 (or 3.2.1 - looks like it doesn't matter) 3.reboot system 4. load pcmcia_core, i82365 and ds 5. (re)start /etc/init.d/pcmcia 6. if not already done, insert a pcmcia nic into any slot. 7. check syslog Actual Results: pcmcia nic modules couldn't be loaded with error message: "bind of #pcmcia-nic-module on socket #socket-no. failed: Operation not permitted." Expected Results: Should be clear.
Tried vanilla-sources 2.4.20. No problems with module insertion of pcmcia-cs drivers here.
Also experiencing same problem. Rebuilding kernel after saving .config and doing a 'make mrproper' had no effect. Disabling grsecurity also had no effect
GS-sources 2.4.21_rc2 kernel works also. so i can at least use a patched kernel instead of the plain vanilla sources :)
I had this same problem, and saw <a href="http://forums.gentoo.org/viewtopic.php?t=54461&highlight=pcmciacs">this forum thread</a> which says that this only happens with Systrace turned on. This is also true for me. So, for a workaround, turn off systrace. It's in General Setup. <p>The thread also mentions that the kernel drivers do work, it's just the pcmcia-cs drivers that don't work. I don't know whether this is true or not. <p>The setting is associated with <a href="http://www.citi.umich.edu/u/provos/systrace/index.html">Systrace</a>, an interactive policy generator for system calls. It allows you to deny access to various system calls according to what program is calling them. I think what might be happening is that Systrace is denying cardmgr permission to bind the card to the socket, and because I didn't have any Systrace userland installed, I didn't heard anything about it. <p>I think a solution to this problem would be to add docs to the Systrace option. Then people would actually *know* that they need to install the userspace for it not to break things horribly, instead of assuming that it's associated with strace functionality.
This really isn't a bug, just the result of not knowing what the systrace option is, exactly. It should be closed.
closing with comment #5
