On Fri, Jul 24, 2026 at 03:07:16PM -0700, Jakub Kicinski wrote: > On Fri, 24 Jul 2026 14:41:47 -0700 Stanislav Fomichev wrote: > > > > Talked to Stan and he mentioned the original idea was not necessarily > > > > just for link down, but any test that could be considered disruptive to > > > > other tests, with the idea to eventually introduce some scheduling > > > > (postpone disruptive until the end). > > > > > > Perhaps have such conversations on the list? IDK what Stan has in mind. > > > > I don't have much to add so I didn't post on the list. In my mind, > > disruptive > > started as "anything that can lead to a soft reset". But since the > > sockets tend to survive these well, I do agree that your definition > > of "link goes down" is a better fit. > > My recollection was basically that the test is not safe to run over > SSH (without tmux). Hence the link down concern. If we say soft reset > that'd cover 95% of the tests so it's not very useful. > > The existing doc and README already document the link down AFAICT.
Indeed, so I'll just test and confirm if devmem's existing ksft_disruptive tests do or can cause link down, and submit a patch removing for those that don't.
