[05:23:48] --- SecureEndpoints has left: Replaced by new connection
[06:23:55] --- matt has become available
[06:24:02] --- matt has left
[10:58:56] --- Mike Garrison has become available
[11:21:48] <Mike Garrison> rather than banging my head against the wall, does anyone have any suggestions on how to further debug "Warning: failed to find address of system call table"
[11:22:05] <Simon Wilkinson> Which kernel?
[11:22:27] <Mike Garrison> home built 2.6.27.3
[11:22:59] <shadow@gmail.com/owlF80E49DA> the simplest way is "build your own kernel and export it"
[11:23:06] <Simon Wilkinson> Okay. So your first option is to turn on keyring support, ignore the error and get on with life.
[11:23:24] <shadow@gmail.com/owlF80E49DA> EXPORT_SYMBOL(sys_call_table);
[11:23:38] <Simon Wilkinson> You need to disable the memory protection features, too.
[11:23:48] <Rrrrred> oh really?
[11:24:22] <Rrrrred> y does one need to disable mem protection?
[11:24:38] <Mike Garrison> hmm
[11:24:46] <Simon Wilkinson> Because, with the debug memory protection on, you can't _write_ to the syscall table.
[11:25:00] <Simon Wilkinson> So AFS can't patch it, and there's not really much value in having found it!
[11:26:14] <Rrrrred> this doesn't sound sustainable
[11:26:36] <Simon Wilkinson> It's not.
[11:26:52] <Simon Wilkinson> Which is why we haven't relied on being able to patch the syscall table for ages.
[11:27:05] <Mike Garrison> so keyring support is really the best scenario
[11:27:14] <Simon Wilkinson> picotls are implement using an ioctl on a file in /proc. PAGs are implemented with keyrings.
[11:27:15] <Simon Wilkinson> Yes.
[11:30:10] <Simon Wilkinson> Personally, I think we should just kill support for patching the syscall table in 1.5.x
[11:31:35] --- Russ has become available
[11:32:00] <Rrrrred> does that break compatibility with some older version of kernels? (which we might not care about anymore)
[11:32:27] <Simon Wilkinson> Almost certainly. But there has to be a point that we draw the line.
[11:32:42] <Mike Garrison> hmm, is the only thing that needs to be enabled in the kernel CONFIG_KEYS ?
[11:34:34] <Simon Wilkinson> I think so. But we use RHEL kernels which have it enabled out of the box, so I've never had to think about it.
[11:35:41] <Mike Garrison> looks like RHEL also has CONFIG_KEYS_DEBUG_PROC_KEYS enabled
[11:36:43] <Simon Wilkinson> Hmmm. We don't need that for OpenAFS.
[11:36:47] <Simon Wilkinson> In fact, hmmm.
[11:38:01] <Simon Wilkinson> You might want to enable it for debugging purposes. I don't think it discloses any information that it shouldn't.
[11:38:11] <Mike Garrison> yeah
[13:35:11] --- sxw has become available
[13:36:25] --- sxw has left
[13:47:31] --- Russ has left: Disconnected
[14:17:21] --- Mike Garrison has left
[14:20:04] --- Russ has become available
[22:30:22] --- reuteras has become available