linux-kernel - Re: [PATCH v2 1/2] nvme: Fix compat NVME_IOCTL_SUBMIT

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <CADyTPEyDbV9WiOc9SOFvzPX9ccd7mhGG_Fj8QNmxGEY9UgkiRQ@mail.gmail.com>
Date:   Sat, 28 Mar 2020 09:56:28 -0400
From:   Nick Bowler <nbowler@...conx.ca>
To:     Christoph Hellwig <hch@...radead.org>
Cc:     linux-nvme@...ts.infradead.org, linux-kernel@...r.kernel.org,
        Sagi Grimberg <sagi@...mberg.me>,
        Keith Busch <kbusch@...nel.org>
Subject: Re: [PATCH v2 1/2] nvme: Fix compat NVME_IOCTL_SUBMIT_IO numbering

On 28/03/2020, Christoph Hellwig <hch@...radead.org> wrote:
> On Sat, Mar 28, 2020 at 01:09:08AM -0400, Nick Bowler wrote:
>> When __u64 has 64-bit alignment, the nvme_user_io structure has trailing
>> padding.  This causes problems in the compat case with 32-bit userspace
>> that has less strict alignment because the size of the structure differs.
>>
>> Since the NVME_IOCTL_SUBMIT_IO macro encodes the structure size itself,
>> the result is that this ioctl does not work at all in such a scenario:
>>
>>   # nvme read /dev/nvme0n1 -z 512
>>   submit-io: Inappropriate ioctl for device
>>
>> But by the same token, this makes it easy to handle both cases and
>> since the structures differ only in unused trailing padding bytes
>> we can simply not read those bytes.
>>
>> Signed-off-by: Nick Bowler <nbowler@...conx.ca>
>
> I think we already have a similar patch titled
> "nvme: Add compat_ioctl handler for NVME_IOCTL_SUBMIT_IO" in
> linux-next, with the difference of actually implementing the
> .compat_ioctl entry point.

OK, I found that one and it looks to solve the same problem.

I'm not sure about copying the nonexistent trailing padding from 32-bit
userspace but that may not be a problem in practice.

So feel free to drop this patch.

Thanks,
  Nick