Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20191008091508.2682-1-thomas_os@shipmail.org> <20191008091508.2682-4-thomas_os@shipmail.org>
 <20191009152737.p42w7w456zklxz72@box>
In-Reply-To: <20191009152737.p42w7w456zklxz72@box>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Wed, 9 Oct 2019 09:21:08 -0700
Message-ID: <CAHk-=wh4waroKr-Xtcv+5pTxBcHxGEj-g73eQvXVawML_C0EXw@mail.gmail.com>
Subject: Re: [PATCH v4 3/9] mm: pagewalk: Don't split transhuge pmds when a
 pmd_entry is present
To:     "Kirill A. Shutemov" <kirill@shutemov.name>
Cc:     =?UTF-8?Q?Thomas_Hellstr=C3=B6m_=28VMware=29?= 
        <thomas_os@shipmail.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux-MM <linux-mm@kvack.org>,
        Thomas Hellstrom <thellstrom@vmware.com>,
        Matthew Wilcox <willy@infradead.org>,
        Will Deacon <will.deacon@arm.com>,
        Peter Zijlstra <peterz@infradead.org>,
        Rik van Riel <riel@surriel.com>,
        Minchan Kim <minchan@kernel.org>,
        Michal Hocko <mhocko@suse.com>,
        Huang Ying <ying.huang@intel.com>,
        =?UTF-8?B?SsOpcsO0bWUgR2xpc3Nl?= <jglisse@redhat.com>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Wed, Oct 9, 2019 at 8:27 AM Kirill A. Shutemov <kirill@shutemov.name> wrote:
>
> Do we have any current user that expect split_huge_pmd() in this scenario.

No. There are no current users of the pmd callback and the pte
callback at all, that I could find.

But it looks like the new drm use does want a "I can't handle the
hugepage, please split it and I'll fo the ptes instead".

> That's hacky.
>
> Maybe just use an error code for this? -EAGAIN?

I actually like the PAGE_WALK_FALLBACK thing as more documentation
than "it's an error, but not one you return", although I do detest the
particular value chosen, which is just a nasty bitpattern.

Maybe it could use an error value, just one that makes no sense, and
is hidden by the PAGE_WALK_FALLBACK define, ie something like

  #define PAGE_WALK_FALLBACK (-ECHILD)

or something like that.

And I suspect the conditional would be cleaner if it was written something like

        if (!err)
                continue;
        if (err != PAGE_WALK_FALLBACK)
                break;
        err = 0;
        if (pmd_trans_unstable(pmd))
                goto again;
        .. do the split ..

and skip the WARN_ON() and the odd "non-zero but smaller than MAX test"

            Linus