[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: leaks



--On Thursday, September 02, 2004 12:06 PM -0700 Quanah Gibson-Mount <quanah@stanford.edu> wrote:

Unfortunately, this patch has not solved the issue.  However, on servers
that haven't suddenly consumed all available memory, they are running
with about 200MB more free memory than before. ;)

Not using SSL frees up another good chunk of memory (there are some interesting leaks in OpenSSL). However, I continue to see slapd's consumation of available resources. On Monday, it affected 3 servers. So far today, it has affected one server. I don't see anything glaring in the profiling data I've gathered. More troubling is that slapd has been locking up a bit (ITS #3321), and that has rather annoying side affects (ITS #3322).

The patch I applied to 2.2.15 is:

<http://www.openldap.org/devel/cvsweb.cgi/servers/slapd/saslauthz.c.diff?r1=1.88.2.18&r2=1.88.2.19&hideattic=1&sortbydate=0&f=h>

In looking at it, I have a question on one bit:

op.o_req_dn.bv_val is freed by ch_free in the first & third group of changes, but is freed by sl_free in the second and fourth group of changes. Is that correct?

From the patch (1st & 3rd group): 

if( !BER_BVISNULL( &op.o_req_dn ) ) ch_free( op.o_req_dn.bv_val );

From the patch (2nd & 4th group):

if( !BER_BVISNULL( &op.o_req_dn ) ) sl_free( op.o_req_dn.bv_val, opx->o_tmpmemctx );


--Quanah

--
Quanah Gibson-Mount
Principal Software Developer
ITSS/Shared Services
Stanford University
GnuPG Public Key: http://www.stanford.edu/~quanah/pgp.html