Skip to contentSkip to conversation

ipv6 activity causing system hang in kernel 2.4.4

DISCUSSION Tim Haynes 3 participants 3 messages May 5, 2001 – May 8, 2001

lore
Share & tools
lore

Tim Haynes

author
#0 lore raw

[I can't see evidence of this being reported before; apols if I'm
wrong. Please also Cc: me as I only read l-k intermittently but would like to help out more.]

Hi,

I've been making very tentative forays into IPv6. However, in my simple experiments thus far I appear to have located a bug:

1/ configure 2 machines with site-local IP#s - I'm using

        ifconfig eth0 inet6 add fec0:1234:5:6::n 

2/ flood-ping from one to the other
3/ after about 15s, watch one box hang, needing magic-sysreq or hard reset

This is only with kernel 2.4.4; 2.4.2, 2.4.3 and NetBSD boxes are not affected. It is independent of platform; I've reproduced it at will on a lowly p75, an athlon, a p3-800 and on a powerbook/PPC.

All kernels are compiled to have ipv6 modular, netfilter modular... everything with which I'm playing, modular.

Compiler versions:

| zsh, 12:06AM % gcc -v
| Reading specs from /usr/lib/gcc-lib/powerpc-linux/2.95.4/specs
| gcc version 2.95.4 20010319 (Debian prerelease)
| zsh, 12:06AM % gcc -v
| Reading specs from /usr/lib/gcc-lib/i386-linux/2.95.4/specs
| gcc version 2.95.4 20010319 (Debian prerelease)

(I'm tracking Debian/Unstable here.)

I have tcpdump logs (<http://spodzone.org.uk/~tim/ipv6/> - they're 570K apiece); the `victim' machine receives nothing but ping-requests and sends nothing but ping-replies until the file is truncated; the surviving box sends nothing but requests and receives nothing but replies until it becomes requests-only. (IOW there is no evidence of ARP, fragmentation traffic, only the pings.)

The Changelog lists an `IPv6 packet re-assembly fix' in -pre2; my suspicions lie in this area or with my compiler.

If there's anything else I can provide by way of diagnostics, please let me know.

Cheers,

~Tim

Signature · 1 line
<http://spodzone.org.uk/>
#1 lore raw

Try this patch, posted the other day. I bet if you inspected, you'd find OOPSes in your logs:

ipv6/ndisc.c renamed from vanilla/linux/net/ipv6/ndisc.c +1 -1
@@ -394,7 +382,7 @@
 	int send_llinfo;
 
 	len = sizeof(struct icmp6hdr) + sizeof(struct in6_addr);
-	send_llinfo = dev->addr_len && ipv6_addr_type(saddr) != IPV6_ADDR_ANY;
+	send_llinfo = dev->addr_len && saddr && ipv6_addr_type(saddr) != IPV6_ADDR_ANY;
 	if (send_llinfo)
 		len += NDISC_OPT_SPACE(dev->addr_len);
 
#2 lore raw
In reply to Tim Haynes
Quoted 3 lines · In article <[email protected]> you wrote:

This is only with kernel 2.4.4; 2.4.2, 2.4.3 and NetBSD boxes are not affected. It is independent of platform; I've reproduced it at will on a lowly p75, an athlon, a p3-800 and on a powerbook/PPC.

I have just reproduced that on 2.4.5pre-1. It was only one ping (ping6) (from the other side of ipv6 over ipv4 tunnel.

Quoted 2 lines

All kernels are compiled to have ipv6 modular, netfilter modular... everything with which I'm playing, modular.

My configuration is without any ipv6/netfilter modules - all build in kernel.

portraits:~# gcc -v
Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs
gcc version 2.96 20000731 (Red Hat Linux 7.1 2.96-81)

glibc-2.2.2-10 (i686) - RedHat 7.1, 1GB RAM, 2x Pentium III

Signature · 2 lines
*[ Łukasz Trąbiński ]*
SysAdmin @wsisiz.edu.pl