patches to make bind9 with TKEY/GSS updates easier to configure

[Mark Andrews]

In message <19703.10388.533135.227163 at samba.org>, tridge at samba.org writes:
> Hi Mark,
> 
>  > This is not what I would expect it the patch applied cleanly.  dstkey is
>  > now dstkeyp (third arguement).
> 
> ahh, well spotted. I was running the new binary under gdb, but it was
> picking up the libs from the install prefix. I hadn't done a install,
> or set LD_LIBRARY_PATH.
> 
> Using the right libs means it doesn't abort at startup, but I still
> get the INSIST() assert that indicates a memory leak when I do a
> TSIG-GSS udpate:
> 
>   02-Dec-2010 15:55:28.336 mem.c:1074: INSIST((((ctx->debuglist[i]).head == (
> (void *)0)) ? isc_boolean_true : isc_boolean_false)) failed, back trace
> 
> The full backtrace is here:
> 
>  http://samba.org/tridge/bind9-patches/backtrace2.txt
> 
> Sorry for the false alarm!
> 
> Cheers, Tridge
> 
> PS: Are you on #bind on IRC?

No.

"-T clientest" is designed to catch long term memory usage which
shouldn't be associated with a clients memory context (short term).

The offending memory is allocated here.

	ptr 0x7fcab83defd0 size 62 file name.c line 2153
	ptr 0x7fcab83db0b0 size 80 file dst_api.c line 1210
	ptr 0x7fcab83df010 size 176 file dst_api.c line 1203

tkey.c:process_gsstkey() uses msg->mctx rather than ring->mctx to create
the key.  Once this is corrected msg is no longer needed.

Mark
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org