Alan Dekok <aland@deployingradius.com> said
Yes. For simplicity:
$ script valgrind.log $ valgrind radiusd -xxx ... $ <ctrl-c> $ exit
Okie Dokie. I was going to use --log-file witrh valgrind, but 'script' works. Done. So far the only errors I'm seeing are these: ==29820== Thread 2: ==29820== Invalid write of size 1 ==29820== at 0x4819294: strNcpy (misc.c:187) ==29820== by 0x4CC43F3: sqlippool_postauth (rlm_sqlippool.c:527) ==29820== by 0xD469: modcall (modcall.c:236) ==29820== by 0xDA73: call_one (modcall.c:269) ==29820== by 0xD701: modcall (modcall.c:324) ==29820== by 0xC8F7: indexed_modcall (modules.c:469) ==29820== by 0x5865: rad_postauth (auth.c:430) ==29820== by 0x6101: rad_authenticate (auth.c:1029) ==29820== by 0xF6D4: rad_respond (radiusd.c:1669) ==29820== by 0x13240: request_handler_thread (threads.c:570) ==29820== by 0x3B4370: start_thread (in /lib/tls/libpthread-2.3.4.so) ==29820== by 0x208FFD: clone (in /lib/tls/libc-2.3.4.so) ==29820== Address 0x4A87AA2 is 0 bytes after a block of size 10 alloc'd ==29820== at 0x4804405: malloc (vg_replace_malloc.c:149) ==29820== by 0x1AA02F: strdup (in /lib/tls/libc-2.3.4.so) ==29820== by 0x4CC42AF: sqlippool_instantiate (rlm_sqlippool.c:486) ==29820== by 0xC5DE: find_module_instance (modules.c:358) ==29820== by 0xDD9C: do_compile_modsingle (modcall.c:1005) ==29820== by 0xCD6E: setup_modules (modules.c:580) ==29820== by 0x103E1: main (radiusd.c:965) There are typically three or four similar events for each authentication, a mix of invalid writes and reads. They are always out of sqlippool_postauth.
Alan DeKok.
-- hugh
Hugh Messenger wrote:
Alan Dekok <aland@deployingradius.com> said ... So far the only errors I'm seeing are these:
==29820== Thread 2: ==29820== Invalid write of size 1 ==29820== at 0x4819294: strNcpy (misc.c:187) ==29820== by 0x4CC43F3: sqlippool_postauth (rlm_sqlippool.c:527)
That's... fairly broken. Barring severe code changes to rlm_sqlippool, I would suggest not using it in 1.1.6. Sorry. Try 2.0.0-pre, at least the rlm_sqlippool module is fixed there. Alan DeKok.
On Tue 19 Jun 2007, Alan DeKok wrote:
Hugh Messenger wrote:
Alan Dekok <aland@deployingradius.com> said
...
So far the only errors I'm seeing are these:
==29820== Thread 2: ==29820== Invalid write of size 1 ==29820== at 0x4819294: strNcpy (misc.c:187) ==29820== by 0x4CC43F3: sqlippool_postauth (rlm_sqlippool.c:527)
That's... fairly broken.
Barring severe code changes to rlm_sqlippool, I would suggest not using it in 1.1.6. Sorry.
Try 2.0.0-pre, at least the rlm_sqlippool module is fixed there.
Yes. There have been numerous changes to rlm_sqlippool in 2.0. Enough so that I think we will probably mark it as a stable module... On the other hand rlm_sqlippool in 1.1.x should work but could be horribly broken. Thats why it's marked as experimental. It was a conscious decision on my part after consultation with Alan not to backport the rlm_sqlippool code changes from cvs HEAD to the 1.1.x branch (Because of a lack of bandwidth on my side). If you can break rlm_sqlippool in cvs head/2.0preX in the same way it is breaking in 1.1.x then we will have a stab at trying to fix it, but otherwise.. Sorry.. Thats what experimental modules are for.. You are of course welcome to submit a patch to fix the problem or backport patches from cvs HEAD.. It's not a huge amount of work, but enough that I didnt want to do it :-) Cheers -- Peter Nixon http://www.peternixon.net/ PGP Key: http://www.peternixon.net/public.asc
participants (3)
-
Alan DeKok -
Hugh Messenger -
Peter Nixon