I have been working on porting CIPE over to 2.3.x now for a while. With
the help of Russell (the Netfilter guy), Alan, and Dave I got it working
fairly well for 2.3.x<=33.
I believe I have it most of the way now. Problem is it seems to stall
here and there. I believe this is partially because rt->peer is null in
one place where I need to make sure I have a new, correct ip_id (used to
be iph->id=ip_id_count++; I now have it iph->id=oldiph_id; as per
ipip.c). The problem is, ipip.c seems to assume that each
port/socket/connection's ip_id will ALWAYS be higher than any others
last. So it seems CIPE gets slow because you get stuff like this:
16:14:04.895312 > site1.port > site2.port: udp 280 (DF)
16:14:05.480679 < site2.port > site1.port: udp 272
16:14:05.570671 < site2.port > site1.port: udp 264
16:15:23.069578 > site1.port > site2.port: udp 168
16:15:23.074521 > site1.port > site2.port: udp 216 (DF)
16:15:23.599351 < site2.port > site1.port: udp 216
16:15:23.679350 < site2.port > site1.port: udp 288
How can I go about using the peer->ip_id_count or secure_ip_id (couldn't
find the macro or function using grep and rgrep)? Can any kernel
hackers help out? I am not touching the crypto code, I am afraid though
that my patch does have links to the crypto calls, but no crypto itself.
Patch is available on request. Olaf Titz, I see you here from time to
time, you ignore me elsewhere. Could you please tell me what you think
of this patch and if you have any alternative fixes? This code should
work on older boxes as well; I have tried hard to keep the ifdefs to
your standard.
Trever Adams
-- For the finest in family and value oriented products: http://www.daysofyore.com- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu Please read the FAQ at http://www.tux.org/lkml/
This archive was generated by hypermail 2b29 : Tue Mar 07 2000 - 21:00:13 EST