Linux ASLR Weakness Addressed

Credit: Hector Marco
Risk: Low
Local: No
Remote: Yes
CWE: CWE-254

CVSS Base Score: 4.6/10
Impact Subscore: 6.4/10
Exploitability Subscore: 3.9/10
Exploit range: Local
Attack complexity: Low
Authentication: No required
Confidentiality impact: Partial
Integrity impact: Partial
Availability impact: Partial

CVE-2016-3672 - Unlimiting the stack not longer disables ASLR Authors: Hector Marco & Ismael Ripoll CVE: CVE-2016-3672 Dates: April 2016 Description We have fixed an old and very known weakness in the Linux ASLR implementation. Any user able to running 32-bit applications in a x86 machine can disable the ASLR by setting the RLIMIT_STACK resource to unlimited. Following are the steps to test whether your system is vulnerable or not: 1) Create a dummy program which shows its memory map: #include <stdio.h> int main(int argc, const char *argv[]) { char cmd[256]; sprintf(cmd, "cat /proc/%d/maps", getpid()); system(cmd); return 0; } 2) Compile it: $ gcc show_maps.c -o show_maps # In a i386 machine $ gcc show_maps.c -o show_maps -m32 # In a 64-bit machine 3) Run the application to check that ASLR is working $ for i in `seq 1 10`; do ./show_maps | grep "r-xp.*libc"; done f75c4000-f7769000 r-xp 00000000 08:01 784726 /lib32/ f75db000-f7780000 r-xp 00000000 08:01 784726 /lib32/ f7557000-f76fc000 r-xp 00000000 08:01 784726 /lib32/ f7595000-f773a000 r-xp 00000000 08:01 784726 /lib32/ f7574000-f7719000 r-xp 00000000 08:01 784726 /lib32/ f75af000-f7754000 r-xp 00000000 08:01 784726 /lib32/ f7530000-f76d5000 r-xp 00000000 08:01 784726 /lib32/ f7529000-f76ce000 r-xp 00000000 08:01 784726 /lib32/ f75c2000-f7767000 r-xp 00000000 08:01 784726 /lib32/ f75fe000-f77a3000 r-xp 00000000 08:01 784726 /lib32/ The library is mapped at random positions, so, the ASLR is working properly. Now, we run the same test but setting the stack to unlimited: $ ulimit -a | grep stack stack size (kbytes, -s) 8192 $ ulimit -s unlimited stack size (kbytes, -s) unlimited $ for i in `seq 1 10`; do ./show_maps | grep "r-xp.*libc"; done 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ 5559a000-5573f000 r-xp 00000000 08:01 784726 /lib32/ The library is mapped at the same position in all executions: the ASLR has been disabled. This is a very old trick to disable ASLR, but unfortunately it was still present in current Linux systems. Vulnerable packages The weakness, IFAIK is present from the first version of current Linux GIT repository. The first version on this resposiroty is Linux-2.6.12-rc2 dated on April 2005. Impact An attacker capable of running 32-bit system applications in a x86 machine is able to disable the ASLR of any application, including sensitive applications such as setuid and setgid. Note that it is not a exploitable vulnerability by itself but a trick to disable the ASLR. This weakness can be use by an attacker when trying to exploit some other bug. Since the i386 is still very used, the number of systems and affected users could be extremely huge. The wekaness The issue arises because the ASLR Linux implementation does not randomize always the mmap base address when the stack size is set to unlimited. Concretely, on i386 and on X86_64 when emulating X86_32 in legacy mode, only the stack and the executable are randomized but not other mmapped files (libraries, vDSO, etc.). And depending in the Linux version, the executable is neither randomized. The function to calculate the libraries position when the stack is set to unlimited is mmap_legacy_base(): static unsigned long mmap_legacy_base(void) { if (mmap_is_ia32()) return TASK_UNMAPPED_BASE; else return TASK_UNMAPPED_BASE + mmap_rnd(); } The function doesn't add any random offset when the system is running in a native 32-bit system (i386) or a 32-bit emulated system (x86_32). Exploit To exploit this weakness, the attacker just need to set to unlimited the stack and then execute a 32-bit application. Obviously the idea is to execute (attack) privileged applications such as setuid/setgid. FIX We have created a patch to fix this issue: diff --git a/arch/x86/mm/mmap.c b/arch/x86/mm/mmap.c index 96bd1e2..389939f 100644 --- a/arch/x86/mm/mmap.c +++ b/arch/x86/mm/mmap.c @@ -94,18 +94,6 @@ static unsigned long mmap_base(unsigned long rnd) } /* - * Bottom-up (legacy) layout on X86_32 did not support randomization, X86_64 - * does, but not when emulating X86_32 - */ -static unsigned long mmap_legacy_base(unsigned long rnd) -{ - if (mmap_is_ia32()) - return TASK_UNMAPPED_BASE; - else - return TASK_UNMAPPED_BASE + rnd; -} - -/* * This function, called very early during the creation of a new * process VM image, sets up which VM layout function to use: */ @@ -116,7 +104,7 @@ void arch_pick_mmap_layout(struct mm_struct *mm) if (current->flags & PF_RANDOMIZE) random_factor = arch_mmap_rnd(); - mm->mmap_legacy_base = mmap_legacy_base(random_factor); + mm->mmap_legacy_base = TASK_UNMAPPED_BASE + random_factor; if (mmap_is_legacy()) { mm->mmap_base = mm->mmap_legacy_base; The patch enables randomization for the libraries, vDSO and mmap requests on i386 and in X86_32 in legacy mode. We already sent the patch to Linux mantainers and the issue will be not problem in incomming Linux versions: Enable full randomization on i386 and X86_32 Discussion Although this vulnerability is not exploitable by itself, the truth is that the ASLR protection mechanism is useless on local attacks for i386 and x86_32 systems when the attackers are able to attack applications that they can lauch. Hector Marco -

Vote for this issue:


Thanks for you vote!


Thanks for you comment!
Your message is in quarantine 48 hours.

Comment it here.

(*) - required fields.  
{{ x.nick }} | Date: {{ x.ux * 1000 | date:'yyyy-MM-dd' }} {{ x.ux * 1000 | date:'HH:mm' }} CET+1
{{ x.comment }}

Copyright 2019,


Back to Top