On 2 December 2014 at 17:32, Jeremiah Mahler <jmmah...@gmail.com> wrote:
> John,
>
> On Mon, Nov 24, 2014 at 08:35:45PM -0800, John Stultz wrote:
>> In commit 6067dc5a8c2b ("time: Avoid possible NTP adjustment mult
>> overflow") a new check was added to watch for adjustments that could
>> cause a mult overflow.
>>
>> Unfortunately the check compares a signed with unsigned value and
>> ignored the case where the adjustment was negative, which causes
>> spurious warn-ons on some systems (and seems like it would result in
>> problematic time adjustments there as well, due to the early
>> return).
>>
>> Thus this patch adds a check to make sure the adjustment is positive
>> before we check for an overflow, and resovles the issue in my
>> testing.
>>
>> Cc: pang.xunlei <pang.xun...@linaro.org>
>> Cc: Fengguang Wu <fengguang...@intel.com>
>> Cc: Thomas Gleixner <t...@linutronix.de>
>> Cc: Ingo Molnar <mi...@kernel.org>
>> Reported-by: Fengguang Wu <fengguang...@intel.com>
>> Debugged-by: pang.xunlei <pang.xun...@linaro.org>
>> Signed-off-by: John Stultz <john.stu...@linaro.org>
>> ---
>> kernel/time/timekeeping.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/kernel/time/timekeeping.c b/kernel/time/timekeeping.c
>> index 29a7d67..2dc0646 100644
>> --- a/kernel/time/timekeeping.c
>> +++ b/kernel/time/timekeeping.c
>> @@ -1330,7 +1330,7 @@ static __always_inline void
>> timekeeping_apply_adjustment(struct timekeeper *tk,
>> *
>> * XXX - TODO: Doc ntp_error calculation.
>> */
>> - if (tk->tkr.mult + mult_adj < mult_adj) {
>> + if ((mult_adj > 0) && (tk->tkr.mult + mult_adj < mult_adj)) {
>> /* NTP adjustment caused clocksource mult overflow */
>> WARN_ON_ONCE(1);
>> return;
>
> This change does quiet the warning but I think it does so for the wrong
> reason.
>
> mult_adj is a signed number and tk->tkr.mult is an unsigned number.
> Adding the check that (mult_adj > 0) limits the test to only positive
> numbers. A positive number plus a positive number will never be less
> than either of the two positive numbers. The test is always false.
Hi Jeremiah,
The result of a positive number plus a positive number may overflow,
for example, u32 (0xFFFF_FFF0 + 0x20) will be 0x10,
that's what this patch is dealing with.
Thanks,
Xunlei
>
> --
> - Jeremiah Mahler
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
