Received-SPF: pass (google.com: best guess record for domain of linux-ext4-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200317113153.7945-1-linus.walleij@linaro.org>
 <CAFEAcA9mXE+gPnvM6HZ-w0+BhbpeuH=osFH-9NUzCLv=w-c7HQ@mail.gmail.com>
 <CACRpkdZtLNUwiZEMiJEoB0ojOBckyGcZeyFkR6MC69qv-ry9EA@mail.gmail.com>
 <CAFEAcA-gdwi=KSW6LqVdEJWSo9VEL5abYQs9LoHd4mKE_-h=Aw@mail.gmail.com>
 <CACRpkdYuZgZUznVxt1AHCSJa_GAXy8N0SduE5OrjDnE1s_L7Zg@mail.gmail.com> <20200324023431.GD53396@mit.edu>
In-Reply-To: <20200324023431.GD53396@mit.edu>
From:   Peter Maydell <peter.maydell@linaro.org>
Date:   Tue, 24 Mar 2020 09:29:58 +0000
Message-ID: <CAFEAcA_6RY1XFVNJCo5=tTkv2GQpXZRqh_Zz4dYadq-8MJZgTQ@mail.gmail.com>
Subject: Re: [PATCH] ext4: Give 32bit personalities 32bit hashes
To:     "Theodore Y. Ts'o" <tytso@mit.edu>
Cc:     Linus Walleij <linus.walleij@linaro.org>,
        "Suzuki K. Poulose" <suzuki.poulose@arm.com>,
        Andreas Dilger <adilger.kernel@dilger.ca>,
        Ext4 Developers List <linux-ext4@vger.kernel.org>,
        linux-fsdevel <linux-fsdevel@vger.kernel.org>,
        Linux API <linux-api@vger.kernel.org>,
        QEMU Developers <qemu-devel@nongnu.org>,
        Florian Weimer <fw@deneb.enyo.de>,
        Andy Lutomirski <luto@kernel.org>,
        stable <stable@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-ext4-owner@vger.kernel.org
Precedence: bulk

On Tue, 24 Mar 2020 at 02:34, Theodore Y. Ts'o <tytso@mit.edu> wrote:
> Another possibility, which would be messier for qemu, would be use a
> flag set via fcntl.  That would require qemu from noticing when the
> guest is calling open, openat, or openat2, and then inserting a fcntl
> system call to set the 32-bit readdir mode.  That's cleaner from the
> kernel interface complexity perspective, but it's messier for qemu.

On the contrary, that would be a much better interface for QEMU.
We always know when we're doing an open-syscall on behalf
of the guest, and it would be trivial to make the fcntl() call then.
That would ensure that we don't accidentally get the
'32-bit semantics' on file descriptors QEMU opens for its own
purposes, and wouldn't leave us open to the risk in future that
setting the PER_LINUX32 flag for all of QEMU causes
unexpected extra behaviour in future kernels that would be correct
for the guest binary but wrong/broken for QEMU's own internals.

thanks
-- PMM