[00:21:01] --- deason has become available [01:03:09] --- deason has left [01:03:10] --- deason has become available [01:14:12] --- deason has left [01:59:25] --- Simon Wilkinson has become available [02:10:17] --- Simon Wilkinson has left [02:46:00] --- Simon Wilkinson has become available [02:51:48] --- Simon Wilkinson has left [13:00:58] --- Simon Wilkinson has become available [13:10:15] --- Simon Wilkinson has left [13:12:04] --- Simon Wilkinson has become available [13:54:04] I think I'm confusing myself about how our syscall interface works. I added a new AFSOP_SEED_ENTROPY that's supposed to take 5 additional arguments (each just a 32-bit word of random data, though I don't think this will be the final form). So, we should be invoking the OS syscall() with (our syscall number), AFSCALL_CALL, AFSOP_SEED_ENTROPY, [additional parameters]. afsd_syscall_populate seems to be able to handle five additional parameters, and so does afs_syscall_call. I'm testing on FreeBSD, and the interface for afs3_syscall from the system headers is: int afs3_syscall(long syscall, \ 678 long parm1, long parm2, long parm3, \ 679 long parm4, long parm5, long parm6); which should still give me five "additional parameters", right? [13:54:48] In practice, on my freebsd box, when I print what I received in afs_syscall_call, the last "additional parameter" is always the same. [14:00:33] ... except we actually use the syscall number for nnpfs_syscall (aka pioctl aka arla's), for ~historical compatibility. I wonder if that has anything to do with it. [14:07:32] vague recollection that there's a limit on number of syscall parameters for some reason in the fbsd syscall interface [14:08:07] I had a vague such recollection as well, but did not find documentation in some quick looking. [14:09:01] In any case, changing which syscall number was used did not have an effect. [14:14:06] bleh, trying to remember context and just realized that it was a discussion so long ago that it may well not be relevant (linux has completely replaced its low level syscall interface at least twice since then, and I've not been tracking whether fbsd has done so or not) [14:15:39] I don't think it's been completely revamped. We still support running FreeBSD 1.x binaries, after all. [14:16:07] sure but that's just the personalities subsystem [14:16:44] (or could be at least :) [14:17:29] linux was still using int 0x80 back then, I think fbsd was using an lcall [15:07:53] --- Simon Wilkinson has left [15:16:13] --- Simon Wilkinson has become available [17:02:34] --- ballbery has left [23:31:43] --- Simon Wilkinson has left