[LEAPSECS] happy anniversary pips
imp at bsdimp.com
Tue Feb 11 21:55:28 EST 2014
On Feb 11, 2014, at 5:34 PM, Greg Hennessy wrote:
>> People have been working for the past 15 years to make leap seconds
>> better, yet in the last leap second all Linux kernels crashed due to
>> a subtle bug that is only triggered when there was a leap second.
> Um, that is false. All linux kernels did not crash, in fact NONE
> of mine did.
"all" here was an overstatement, but the impact of the leap second should never be "your kernel crashes" even if your personal kernels didn't. It is a consequence that's much larger than the extra second...
> Did *some* multiprocessing kernels go into a spin lock
> when they issued a printk? Yes. I admit to wanting to know
> why the cry isn't "Symmetric multiprocessing is hard"
> rather than "leap seconds are hard".
MP is hard, sure, but that's not the root cause of this issue. Most of the changes that get into the base linux base are well tested because they are easy to test. Bugs that are this easy to trigger are quite rare. Part of what made this bug last as long as it did in the baseline was (a) the rarity of leap seconds and (b) the difficulty in testing them.
More information about the LEAPSECS