Re: [PATCH] [iov_iter] use memmove() when copying to/from user page

[Eric Dumazet]

On Thu, May 25, 2017 at 10:04 AM, Pavel Machek <pavel@xxxxxx> wrote:
> Hi!
>
> On Tue 2017-05-16 23:48:54, Al Viro wrote:
>> On Tue, May 16, 2017 at 03:15:16PM -0700, Dmitry Vyukov wrote:
>> > > Because it's not going to be *one* call of memcpy() or memmove().  It's
>> > > one per page.
>> >
>> >
>> > I missed that.
>> >
>> > I assumed that in the case of sendfile from memfd to memfd data will
>> > be copied directly. But it goes through a pipe with multiple buffers.
>> > Does not look easily fixable.
>>
>> Which leaves us only with "will nasal demons really fly there?".
>
> It seems so:
>
> Date: Tue, 16 May 2017 14:27:34 +0200
> From: Alexander Potapenko <glider@xxxxxxxxxx>
> Subject: [PATCH] [iov_iter] use memmove() when copying to/from user
> page
>
> BUG: memcpy-param-overlap in generic_perform_write+0x551/0xa20
> __msan_memcpy(ffff88013c6e3001, ffff88013c6e3000, 105)
> CPU: 0 PID: 1040 Comm: probe Not tainted 4.11.0-rc5+ #2562
> Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS Bochs
> 01/01/2011
>
> At the very least, we do not want userland to trigger kernel
> BUG()s... so this needs fixes beyond documentation.

To be fair, this BUG() only happens because Alexander added one in memcpy() ,
testing for the cases where memmove() should have been used.

Kind of a debugging trap if you prefer.

This is not something that a pristine kernel would do.