Messages in this thread | | | Date | Wed, 17 Nov 2021 16:55:45 -0500 | Subject | Re: [PATCH v2 0/4] clocksource: Avoid incorrect hpet fallback | From | Waiman Long <> |
| |
On 11/17/21 16:25, Paul E. McKenney wrote: > On Wed, Nov 17, 2021 at 01:51:51PM -0500, Waiman Long wrote: >> On 11/17/21 11:54, Paul E. McKenney wrote: >>> On Tue, Nov 16, 2021 at 06:44:22PM -0500, Waiman Long wrote: >>> A few questions: >>> >>> 1. Once you have all the patches in place, is the increase in >>> WATCHDOG_MAX_SKEW from 50us to 100us necessary? >> I think so. Using Feng's reproducer, I was able to cause a hpet-hpet delay >> of more than 90us on a 1-socket system. With a default 50us >> WATCHDOG_MAX_SKEW, the chance of a warning showing up will be much higher. >> Trying to minimize the chance that a warning may appear is my primary reason >> to increase WATCHDOG_MAX_SKEW. > Should we downgrade the "had to retry read" complain to pr_info(), > and make the only real warning be the case where a large number of > consecutive read attempts fail? I believe that Heiner Kallweit was > looking for something like this. Sure. I will downgrade it to pr_info(). > >>> 2. The reason for having cs->uncertainty_margin set to >>> 2*WATCHDOG_MAX_SKEW was to allow for worst-case skew from both >>> the previous and the current reading. Are you sure that >>> dropping back to WATCHDOG_MAX_SKEW avoids false positives? >> I can remove the hunk of changing cs->uncertainty_margin. It is critical for >> this patch. > Assuming "not critical", good!
Yes, it is "not critical". Somehow I missed the "not" :-)
Cheers, Longman
| |