On Wed, Jul 02, 2003 at 02:29:21PM -0400, Jim Sandoz wrote: > eh? > > (ronin)$ pwd > /ronin/home0/jds/tools/ipf/ip_fil3.4.31 > (ronin)$ grep NAT_TABLE_SZ * > ip_nat.c:u_int ipf_nattable_sz = NAT_TABLE_SZ; > ip_nat.h:#ifndef NAT_TABLE_SZ > ip_nat.h:# define NAT_TABLE_SZ 127 > ip_nat.h:#undef NAT_TABLE_SZ > ip_nat.h:#define NAT_TABLE_SZ 16383
Looks may be deceiving. Take a look at ip_nat.c, and you'll see that NAT_TABLE_SZ controls the number of hash buckets, not a maximum number of table entries allocated. Upon flush/unload, you'll see a bulk KFREE of the hash table, but the actual entries are maintained in the nat_instances list, whose items are iteratively freed afterwards. ip_state.c has the whole fr_statemax thing going on; there's no analog for ip_nat.c (... yet?). -- ryan beasley <[EMAIL PROTECTED]> GPG ID: 0x16EFBD48
pgp00000.pgp
Description: PGP signature
