Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
MIME-Version: 1.0
References: <20220423100751.1870771-1-catalin.marinas@arm.com>
 <CAHk-=wgafGgBC9JEu397YxFD8o8qiCZHQS+f5i+BSXOkOFqX3w@mail.gmail.com> <YmRIKJSr0xSRHliN@arm.com>
In-Reply-To: <YmRIKJSr0xSRHliN@arm.com>
From:   Andreas Gruenbacher <agruenba@redhat.com>
Date:   Mon, 25 Apr 2022 13:08:32 +0200
Message-ID: <CAHc6FU5Sm6jR-NM5t7qEEHs3z42BzLHgZ5FG-kjg5VjUwXJjNQ@mail.gmail.com>
Subject: Re: [PATCH v4 0/3] Avoid live-lock in btrfs fault-in+uaccess loop
To:     Catalin Marinas <catalin.marinas@arm.com>
Cc:     Linus Torvalds <torvalds@linux-foundation.org>,
        Andrew Morton <akpm@linux-foundation.org>,
        Josef Bacik <josef@toxicpanda.com>,
        Al Viro <viro@zeniv.linux.org.uk>, Chris Mason <clm@fb.com>,
        David Sterba <dsterba@suse.com>, Will Deacon <will@kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        Linux ARM <linux-arm-kernel@lists.infradead.org>,
        linux-btrfs <linux-btrfs@vger.kernel.org>,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Hi Catalin,

On Sat, Apr 23, 2022 at 8:40 PM Catalin Marinas <catalin.marinas@arm.com> wrote:
> On Sat, Apr 23, 2022 at 09:35:42AM -0700, Linus Torvalds wrote:
> > On Sat, Apr 23, 2022 at 3:07 AM Catalin Marinas <catalin.marinas@arm.com> wrote:
> > >
> > > The series introduces fault_in_subpage_writeable() together with the
> > > arm64 probing counterpart and the btrfs fix.
> >
> > Looks fine to me - and I think it can probably go through the arm64
> > tree since you'd be the only one really testing it anyway.
>
> I'll queue it via arm64 then.

sounds good to me, thank you.

> > I assume you checked that btrfs is the only one that uses
> > fault_in_writeable() in this way? Everybody else updates to the right
> > byte boundary and retries (or returns immediately)?
>
> I couldn't find any other places (by inspection or testing). The
> buffered file I/O can already make progress in current fault_in_*() +
> copy_*_user() loops.

This started working correctly with commit bc1bb416bbb9
("generic_perform_write()/iomap_write_actor(): saner logics for short
copy") by Al from last May.

> O_DIRECT either goes via GUP (and memcpy() doesn't
> fault) or, if the user buffer is not PAGE aligned, it may fall back to
> buffered I/O. That's why I simplified the series, AFAICT it's only btrfs
> search_ioctl() with this problem.

Thanks,
Andreas