On Oct 29, 2022, at 12:14 PM, Linus Torvalds <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: > On Sat, Oct 29, 2022 at 11:58 AM Linus Torvalds > <torvalds@xxxxxxxxxxxxxxxxxxxx> wrote: >> Still not tested in any way, shape, or form. I decided I wanted to >> send this one before booting into this and possibly blowing up ;^) > > Well, it boots, and I see no difference with your PoC code. > > It didn't fail for me before, it doesn't fail for me with those patches. > > Again, the "it doesn't fail for me" is probably because I'm running it > incorrectly, although for all I know there can also be hardware > differences. Please give me some time to test it. I presume you ran it with block ram device (not tmpfs) and not on a virtual machine (which can affect besides Intel/AMD implementation differences). But even if your patches work and the tests pass, I am not sure it means that everything is fine. I did not try to trigger a race with shrink_page_list(), and doing that might be harder than the race I tried to create before. I need to do some tracing to understand what I was missing in my understanding of the shrink_page_list() - assuming that I am mistaken about the buffers being potentially released. I would note that my concern about releasing the buffers is partially driven by to issues that were reported before [1]. I am actually not sure how they were resolved. [1] https://lore.kernel.org/all/20180103100430.GE4911@xxxxxxxxxxxxxx/
