On Mon, Jul 16, 2018 at 10:44:46AM +0200, Alexander Sverdlin wrote: > On 16/07/18 10:21, Lars-Peter Clausen wrote: > > You'd run readwriter() before checking the timeout and reset the timeout if > > it is able to make progress. Only if there was no progress you'd check the > > timeout. > I think your point is valid, but I don't feel only this improvement will be enough. > What do you think about both increasing the timeout to 60-120sec and resetting > timeout if a progress has been made? > Then I can prepare v2. That's sounding like an extremely high timeout, long enough that people are likely to think that the system has locked up (and it'd probably be triggering the scheduler warnings too). It feels like either whatever is consuming the CPU has a problem that needs fixing or we need some system wide indication that there's something intentionally doing this so other tasks should lift any timeout checks that they have.
Attachment:
signature.asc
Description: PGP signature
