ntp: Fixup adjtimex freq validation on 32-bit systems
authorJohn Stultz <john.stultz@linaro.org>
Tue, 10 Feb 2015 07:30:36 +0000 (23:30 -0800)
committerIngo Molnar <mingo@kernel.org>
Wed, 18 Feb 2015 13:50:10 +0000 (14:50 +0100)
Additional validation of adjtimex freq values to avoid
potential multiplication overflows were added in commit
5e5aeb4367b (time: adjtimex: Validate the ADJ_FREQUENCY values)

Unfortunately the patch used LONG_MAX/MIN instead of
LLONG_MAX/MIN, which was fine on 64-bit systems, but being
much smaller on 32-bit systems caused false positives
resulting in most direct frequency adjustments to fail w/
EINVAL.

ntpd only does direct frequency adjustments at startup, so
the issue was not as easily observed there, but other time
sync applications like ptpd and chrony were more effected by
the bug.

See bugs:

  https://bugzilla.kernel.org/show_bug.cgi?id=92481
  https://bugzilla.redhat.com/show_bug.cgi?id=1188074

This patch changes the checks to use LLONG_MAX for
clarity, and additionally the checks are disabled
on 32-bit systems since LLONG_MAX/PPM_SCALE is always
larger then the 32-bit long freq value, so multiplication
overflows aren't possible there.

Reported-by: Josh Boyer <jwboyer@fedoraproject.org>
Reported-by: George Joseph <george.joseph@fairview5.com>
Tested-by: George Joseph <george.joseph@fairview5.com>
Signed-off-by: John Stultz <john.stultz@linaro.org>
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: <stable@vger.kernel.org> # v3.19+
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Sasha Levin <sasha.levin@oracle.com>
Link: http://lkml.kernel.org/r/1423553436-29747-1-git-send-email-john.stultz@linaro.org
[ Prettified the changelog and the comments a bit. ]
Signed-off-by: Ingo Molnar <mingo@kernel.org>
kernel/time/ntp.c

index 4b585e0fdd22e16288f688baa1051395836461d5..0f60b08a4f073e9246ced1dc3b5de5f50efd7cf4 100644 (file)
@@ -633,10 +633,14 @@ int ntp_validate_timex(struct timex *txc)
        if ((txc->modes & ADJ_SETOFFSET) && (!capable(CAP_SYS_TIME)))
                return -EPERM;
 
-       if (txc->modes & ADJ_FREQUENCY) {
-               if (LONG_MIN / PPM_SCALE > txc->freq)
+       /*
+        * Check for potential multiplication overflows that can
+        * only happen on 64-bit systems:
+        */
+       if ((txc->modes & ADJ_FREQUENCY) && (BITS_PER_LONG == 64)) {
+               if (LLONG_MIN / PPM_SCALE > txc->freq)
                        return -EINVAL;
-               if (LONG_MAX / PPM_SCALE < txc->freq)
+               if (LLONG_MAX / PPM_SCALE < txc->freq)
                        return -EINVAL;
        }