On Tue, 11 November 2003 18:27:12 +0100, Thomas Habets wrote: > > On Tuesday 11 November 2003 17:27, Jörn Engel wrote: > > Looks familiar. Can you recreate and send the output from > > /proc/slabinfo? > > Oh, I didn't notice that file. :-) > Recreating is just booting and waiting for a week, but the box is still up. > > This is the line that stands out (complete file attached, 137 lines). > > tcp6_sock 111663 111664 960 4 1 : tunables 54 27 0 : > slabdata 27916 27916 0 > > I seem to remember a changelog mentioning a leak being fixed in ipv6 code, > but it looks like there may be another one? The only ipv6 service running is > sshd, and the mrtg-sshs that go off every 5 minutes are NOT over ipv6. > netstat -na shows nothing interesting. Only the ssh I connect with uses a bit > of ipv6 (ffff:1.2.3.4). So, one listening socket, and one established. Holger Kiehl found a similar leak somewhere in the ipv4 code, possibly route.c. We couldn't nail the problem down yet and it could still be inside a seperate device driver (GPL, outside of kernel tree). Holger's problem is independent of the one found by the stanford checker some weeks ago, so there seem to be or have been at least three memory leaks in the network code. Could be a single developer that is simply not careful enough wrt. resource leaks. David, can you think of an easy way to find all of those leaks? Easier than waiting for bug reports and hunting them one by one? Jörn -- Fantasy is more important than knowlegde. Knowlegde is limited, while fantasy embraces the whole world. -- Albert Einstein - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html
