Re: [PATCH] mmap.2: document new MAP_FIXED_NOREPLACE flag

[Jann Horn <jannh@google.com>]

On Wed, Apr 11, 2018 at 6:36 PM, Michal Hocko <mhocko@kernel.org> wrote:
> On Wed 11-04-18 17:37:46, Jann Horn wrote:
>> On Wed, Apr 11, 2018 at 2:04 PM,  <mhocko@kernel.org> wrote:
>> > From: Michal Hocko <mhocko@suse.com>
>> >
>> > 4.17+ kernels offer a new MAP_FIXED_NOREPLACE flag which allows the caller to
>> > atomicaly probe for a given address range.
>> >
>> > [wording heavily updated by John Hubbard <jhubbard@nvidia.com>]
>> > Signed-off-by: Michal Hocko <mhocko@suse.com>
>> > ---
>> > Hi,
>> > Andrew's sent the MAP_FIXED_NOREPLACE to Linus for the upcoming merge
>> > window. So here we go with the man page update.
>> >
>> >  man2/mmap.2 | 27 +++++++++++++++++++++++++++
>> >  1 file changed, 27 insertions(+)
>> >
>> > diff --git a/man2/mmap.2 b/man2/mmap.2
>> > index ea64eb8f0dcc..f702f3e4eba2 100644
>> > --- a/man2/mmap.2
>> > +++ b/man2/mmap.2
>> > @@ -261,6 +261,27 @@ Examples include
>> >  and the PAM libraries
>> >  .UR http://www.linux-pam.org
>> >  .UE .
>> > +Newer kernels
>> > +(Linux 4.17 and later) have a
>> > +.B MAP_FIXED_NOREPLACE
>> > +option that avoids the corruption problem; if available, MAP_FIXED_NOREPLACE
>> > +should be preferred over MAP_FIXED.
>>
>> This still looks wrong to me. There are legitimate uses for MAP_FIXED,
>> and for most users of MAP_FIXED that I'm aware of, MAP_FIXED_NOREPLACE
>> wouldn't work while MAP_FIXED works perfectly well.
>>
>> MAP_FIXED is for when you have already reserved the targeted memory
>> area using another VMA; MAP_FIXED_NOREPLACE is for when you haven't.
>> Please don't make it sound as if MAP_FIXED is always wrong.
>
> Well, this was suggested by John. I think, nobody is objecting that
> MAP_FIXED has legitimate usecases. The above text just follows up on
> the previous section which emphasises the potential memory corruption
> problems and it suggests that a new flag is safe with that regards.
>
> If you have specific wording that would be better I am open for changes.

I guess I'd probably also want to change the previous text; so I
should probably send a followup patch once this one has landed.

>> > +.TP
>> > +.BR MAP_FIXED_NOREPLACE " (since Linux 4.17)"
>> > +Similar to MAP_FIXED with respect to the
>> > +.I
>> > +addr
>> > +enforcement, but different in that MAP_FIXED_NOREPLACE never clobbers a pre-existing
>> > +mapped range. If the requested range would collide with an existing
>> > +mapping, then this call fails with
>> > +.B EEXIST.
>> > +This flag can therefore be used as a way to atomically (with respect to other
>> > +threads) attempt to map an address range: one thread will succeed; all others
>> > +will report failure. Please note that older kernels which do not recognize this
>> > +flag will typically (upon detecting a collision with a pre-existing mapping)
>> > +fall back to a "non-MAP_FIXED" type of behavior: they will return an address that
>> > +is different than the requested one. Therefore, backward-compatible software
>> > +should check the returned address against the requested address.
>> >  .TP
>> >  .B MAP_GROWSDOWN
>> >  This flag is used for stacks.
>> > @@ -487,6 +508,12 @@ is not a valid file descriptor (and
>> >  .B MAP_ANONYMOUS
>> >  was not set).
>> >  .TP
>> > +.B EEXIST
>> > +range covered by
>> > +.IR addr ,
>> > +.IR length
>> > +is clashing with an existing mapping.
>>
>> Maybe add something like ", and MAP_FIXED_NOREPLACE was specified"? I
>> think most manpages explicitly document which error conditions can be
>> triggered by which flags.
>
> sure, no objection from me.
>
> --
> Michal Hocko
> SUSE Labs