Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20201217105409.2gacwgg7rco2ft3m@box> <CAHk-=wiyPTnQ9E1dT9LJtNxeVmLaykursk_MSecUqFjSb3gwAw@mail.gmail.com>
 <20201218110400.yve45r3zsv7qgfa3@box> <CAHk-=wgO2LsoKhX7MjSECo+Xrj1-Me7tzRfNcsdEZBRwJW1cQg@mail.gmail.com>
 <20201219124103.w6isern3ywc7xbur@box> <CAHk-=wifcVaxaTn_RbZ=idfYFazTPwm8t5cB1rY6xEBjbcfO5g@mail.gmail.com>
 <20201222100047.p5zdb4ghagncq2oe@box> <alpine.LSU.2.11.2012231905300.5723@eggly.anvils>
 <20201225113157.e7hmluffh56fszfc@box> <CAHk-=wiT50aEErZgZOrbFQ=GhFuM3MnBmHoSBSScN9rmXMnOKQ@mail.gmail.com>
 <20201226204335.dikqkrkezqet6oqf@box>
In-Reply-To: <20201226204335.dikqkrkezqet6oqf@box>
From:   Linus Torvalds <torvalds@linux-foundation.org>
Date:   Sat, 26 Dec 2020 13:07:14 -0800
Message-ID: <CAHk-=wgbUdvTShO5woNpM8w9rd655XwxKwHoz_r_HWReyx1oAw@mail.gmail.com>
Subject: Re: [PATCH 1/2] mm: Allow architectures to request 'old' entries when prefaulting
To:     "Kirill A. Shutemov" <kirill@shutemov.name>
Cc:     Hugh Dickins <hughd@google.com>,
        Matthew Wilcox <willy@infradead.org>,
        "Kirill A. Shutemov" <kirill.shutemov@linux.intel.com>,
        Will Deacon <will@kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Linux-MM <linux-mm@kvack.org>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        Catalin Marinas <catalin.marinas@arm.com>,
        Jan Kara <jack@suse.cz>, Minchan Kim <minchan@kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Vinayak Menon <vinmenon@codeaurora.org>,
        Android Kernel Team <kernel-team@android.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

I was going to just apply this patch, because I like it so much, but
then I decided to take one last look, and:

On Sat, Dec 26, 2020 at 12:43 PM Kirill A. Shutemov
<kirill@shutemov.name> wrote:
>
> +static bool filemap_map_pmd(struct vm_fault *vmf, struct page *page)
> +{
> +       struct mm_struct *mm = vmf->vma->vm_mm;
> +
> +       /* Huge page is mapped? No need to proceed. */
> +       if (pmd_trans_huge(*vmf->pmd))
> +               return true;

doesn't this cause us to leak a locked page?

I get the feeling that every single "return true" case here should
always unlock the page and - with the exception of a successful
do_set_pmd() - do a "put_page()".

Which kind of argues that we should just do it in the caller (and get
an extra ref in the do_set_pmd() case, so that the caller can always
do

        if (filemap_map_pmd(..)) {
                unlock_page(page);
                put_page(page);
                rcu_read_unlock();
                return;
        }

andf then there are no odd cases inside that filemap_map_pmd() function. Hmm?

Other than that, I really find it all much more legible.

Of course, if I'm wrong about the above, that just proves that I'm
missing something and it wasn't so legible after all..

              Linus