On Fri, Dec 11, 2020 at 12:20, Daniel Wagner wrote:
On Fri, Dec 11, 2020 at 11:11:15AM +0100, Daniel Wagner wrote:
> Anyway, I think the best way forward is to either revert the commit or
> change it to use the original clock id.
I've reverted the commit. There is no point in keeping the code around
until we figure out how to address the problem this commit to tried to
The original code used CLOCK_MONOTONIC for timeouts, which pauses during
suspend. Therefore the lease was kept long after wakeup and after the
lease was expired. To not wake up the system with CLOCK_BOOTTIME_ALARM,
we could use CLOCK_BOOTTIME instead, which at least releases the address
right after wakeup.