Messages in this thread | | | Date | Mon, 15 Aug 2022 03:53:49 -0400 | From | "Michael S. Tsirkin" <> | Subject | Re: upstream kernel crashes |
| |
On Mon, Aug 15, 2022 at 12:46:36AM -0700, Andres Freund wrote: > Hi, > > On Mon, Aug 15, 2022, at 00:29, Michael S. Tsirkin wrote: > > On Mon, Aug 15, 2022 at 12:11:43AM -0700, Andres Freund wrote: > >> Hi, > >> > >> On 2022-08-14 20:18:44 -0700, Linus Torvalds wrote: > >> > On Sun, Aug 14, 2022 at 6:36 PM Andres Freund <andres@anarazel.de> wrote: > >> > > > >> > > Some of the symptoms could be related to the issue in this thread, hence > >> > > listing them here > >> > > >> > Smells like slab corruption to me, and the problems may end up being > >> > then largely random just depending on who ends up using the allocation > >> > that gets trampled on. > >> > > >> > I wouldn't be surprised if it's all the same thing - including your > >> > network issue. > >> > >> Yea. As I just wrote in > >> https://postgr.es/m/20220815070203.plwjx7b3cyugpdt7%40awork3.anarazel.de I > >> bisected it down to one commit (762faee5a267). With that commit I only see the > >> networking issue across a few reboots, but with ebcce4926365 some boots oops > >> badly and other times it' "just" network not working. > >> > >> > >> [oopses] > > >> If somebody knowledgeable staring at 762faee5a267 doesn't surface somebody I > >> can create a kernel with some more debugging stuff enabled, if somebody tells > >> me what'd work best here. > >> > >> > >> Greetings, > >> > >> Andres Freund > > > > Thanks a lot for the work! > > Just a small clarification: > > > > So IIUC you see several issues, right? > > Yes, although they might be related, as theorized by Linus upthread. > > > With 762faee5a2678559d3dc09d95f8f2c54cd0466a7 you see networking issues. > > Yes. > > > > With ebcce492636506443e4361db6587e6acd1a624f9 you see crashes. > > Changed between rebooting. Sometimes the network issue, sometimes the crashes in the email you're replying to. >
OK I think we can safely drop the others on the Cc, and I will add a bunch of people and relevant lists instead. I want to fix this quickly so testing can proceed.
> > MST
| |