On Mon, 17 Oct 2022 18:34:50 +0200 Oleksandr Natalenko <oleksandr@xxxxxxxxxxxxxx> wrote: > > > > I'll give it a try. The only problem is that for me the issue is not reproducible at will, it can take 1 day, or it can take 2 weeks before the panic is hit. > > > > > > > > > [1] https://lore.kernel.org/linux-mm/YxDyZLfBdFHK1Y1P@bfoster/ > > > > > [2] https://lore.kernel.org/linux-mm/20220906190602.1626037-1-bfoster@xxxxxxxxxx/ > > > > > > So far, I haven't reproduced this issue with your patch. I haven't run the machine sufficiently long, just under a week, so this is rather to let you know that I haven't abandoned testing. > > > > > > > Thanks for the update. Is this still going well, or reached a point > > where you typically see the problem? I can still reproduce the original > > problem so I may have to ping the patch again.. > > So far, no issue observed with your patch. Thanks. It's actually unclear (to me) why Matthew's b653db77350c73 ("mm: Clear page->private when splitting or migrating a page") was considered necessary. What problem did it solve? https://lore.kernel.org/linux-mm/20220906190602.1626037-1-bfoster@xxxxxxxxxx/ is a partial undoing of that change, but should we simply revert b653db77350c73?
