On Sat, Dec 10, 2005 at 07:39:04PM +0100, Thiemo Seufer wrote: > Andrew Suffield wrote: > > This doesn't look like #223110 to me - that bug is specific to signal > > handlers. > > After a closer look I agree. > > > I just tried the test on casals against 5.8.7-6 though, and > > I can't get it to break (although it is still breaking on the buildd). > > > > What do you know about it? I need to duplicate the pesky thing > > somehow... > > Hm, it is easily reproducible on SWARM, as well as on a Octane with 2 > CPUs. Which probably means it is a SMP-induced race condition in fork.
My bet is that it's this kernel bug: I don't know if you've been following, but it was recently discoverd that on smp, if multiple processes read from /dev/urandom at the same time, they can get the same data. Theodore Tytso posted a patch to fix this for 2.6, and someone else told me this problem has existed all the way back to 1.3. Fixed in 2.4.29 and 2.6.5. What kernel version have you got there? If it's 2.4.27 (sarge) then I think we have our problem found. -- .''`. ** Debian GNU/Linux ** | Andrew Suffield : :' : http://www.debian.org/ | `. `' | `- -><- |
signature.asc
Description: Digital signature
