Re: [linus:master] [mm] efa7df3e3b: kernel_BUG_at_include/linux/page_ref.h

[Yang Shi <shy828301@gmail.com>]

On Tue, Jun 11, 2024 at 11:01 PM Oliver Sang <oliver.sang@intel.com> wrote:
>
> hi, Yang Shi,
>
> On Wed, Jun 05, 2024 at 08:44:37PM -0700, Yang Shi wrote:
> > Oliver Sang <oliver.sang@intel.com>于2024年6月5日 周三19:16写道：
> >
> > > hi, Yang Shi,
> > >
> > > On Tue, Jun 04, 2024 at 04:53:56PM -0700, Yang Shi wrote:
> > > > On Mon, Jun 3, 2024 at 9:54 AM Yang Shi <shy828301@gmail.com> wrote:
> > > > >
> > > > > On Mon, Jun 3, 2024 at 7:02 AM Oliver Sang <oliver.sang@intel.com>
> > > wrote:
> > > > > >
> > > > > > hi, Yang Shi,
> > > > > >
> > > > > > On Fri, May 31, 2024 at 01:57:06PM -0700, Yang Shi wrote:
> > > > > > > Hi Oliver,
> > > > > > >
> > > > > > > I just came up with a quick patch (just build test) per the
> > > discussion
> > > > > > > and attached, can you please to give it a try? Once it is
> > > verified, I
> > > > > > > will refine the patch and submit for review.
> > > > > >
> > > > > > what's the base of this patch? I tried to apply it upon efa7df3e3b or
> > > > > > v6.10-rc2. both failed.
> > > > >
> > > > > Its base is mm-unstable. The head commit is 8e06d6b9274d ("mm: add
> > > > > swappiness= arg to memory.reclaim"). Sorry for the confusion, I should
> > > > > have mentioned this.
> > > >
> > > > I just figured out a bug in the patch. Anyway, we are going to take a
> > > > different approach to fix the issue per the discussion. I already sent
> > > > the series to the mailing list. Please refer to
> > > >
> > > https://lore.kernel.org/linux-mm/20240604234858.948986-1-yang@os.amperecomputing.com/
> > >
> > > got it. seems you will submit v2? should we wait v2 to do the tests?
> >
> >
> > The real fix is patch #1, that doesn’t need v2. So you just need to test
> > that.
>
> we've finished tests and confirmed patch #1 fixed the issue.
> we also tested upon patch #2, still clean.

Thanks for testing. Sorry for the late reply, just came back from
vacation. It seems like Andrew didn't take the fix yet. I will resend
the patch with your tested-by tag. And I will drop the patch #2 since
it is just a clean up and I didn't receive any review comments. In
addition, the undergoing hugepd clean up may make this clean up
easier, so I will put the clean up on the back burner for now.

>
> our bot applied your patch upon 306dde9ce5c951 as below
>
> 5d45cc9b1beb57 mm: gup: do not call try_grab_folio() in slow path
> fd3fc964468925 mm: page_ref: remove folio_try_get_rcu()
> 306dde9ce5c951 foo
>
> on 306dde9ce5c951, we still observed the issue we reported. clean on both patch
> #1 and #2
>
> 306dde9ce5c9516d fd3fc96446892528af48d6271a3 5d45cc9b1beb57386992c005669
> ---------------- --------------------------- ---------------------------
>        fail:runs  %reproduction    fail:runs  %reproduction    fail:runs
>            |             |             |             |             |
>          47:50         -94%            :50         -94%            :50    dmesg.Kernel_panic-not_syncing:Fatal_exception
>          47:50         -94%            :50         -94%            :50    dmesg.Oops:invalid_opcode:#[##]KASAN
>          47:50         -94%            :50         -94%            :50    dmesg.RIP:try_get_folio
>          47:50         -94%            :50         -94%            :50    dmesg.kernel_BUG_at_include/linux/page_ref.h
>
>
> >
> > For patch #2, I haven’t received any comment yet and I’m going to travel so
> > I’m not going to submit v2 soon .
> >
> > And I heard if hugepd is going to be gone soon, so I may wait for that then
> > rebase on top of it. Anyway it is just a clean up.
> >
> >
> >
> > >
> > > sorry that due to resource constraint, we cannot respond test request very
> > > quickly now.
> > >
> > > >
> > > > >
> > > > > >
> > > > > > >
> > > > > > > Thanks,
> > > > > > > Yang
> > > > > > >
> > > > > > > >
> > > > > > > > >
> > > > > > > > >
> > > > > > > > > --
> > > > > > > > > Cheers,
> > > > > > > > >
> > > > > > > > > David / dhildenb
> > > > > > > > >
> > > > > >
> > > > > >
> > > >
> > >