Re: [RFC PATCH v8 2/2] mm: introduce MAP_SHARED_VALIDATE, a mechanism to safely define new mmap flags

[Dan Williams]

On Mon, Sep 11, 2017 at 2:47 AM, Jan Kara <jack@xxxxxxx> wrote:
> On Fri 08-09-17 12:35:13, Dan Williams wrote:
>> The mmap(2) syscall suffers from the ABI anti-pattern of not validating
>> unknown flags. However, proposals like MAP_SYNC and MAP_DIRECT need a
>> mechanism to define new behavior that is known to fail on older kernels
>> without the support. Define a new MAP_SHARED_VALIDATE flag pattern that
>> is guaranteed to fail on all legacy mmap implementations.
>>
>> With this in place new flags can be defined as:
>>
>>     #define MAP_new (MAP_SHARED_VALIDATE | val)
>
> Is this changelog stale? Given MAP_SHARED_VALIDATE will be new mapping
> type, I'd expect we define new flags just as any other mapping flags...
> I see no reason why MAP_SHARED_VALIDATE should be or'ed to that.

True, it will just by a new MAP_TYPE plus new flags. I will fix this up comment.

[..]
>> diff --git a/arch/alpha/include/uapi/asm/mman.h b/arch/alpha/include/uapi/asm/mman.h
>> index 3b26cc62dadb..c32276c4196a 100644
>> --- a/arch/alpha/include/uapi/asm/mman.h
>> +++ b/arch/alpha/include/uapi/asm/mman.h
>> @@ -14,6 +14,7 @@
>>  #define MAP_TYPE     0x0f            /* Mask for type of mapping (OSF/1 is _wrong_) */
>>  #define MAP_FIXED    0x100           /* Interpret addr exactly */
>>  #define MAP_ANONYMOUS        0x10            /* don't use a file */
>> +#define MAP_SHARED_VALIDATE (MAP_SHARED|MAP_PRIVATE) /* validate extension flags */
>
> And I'd explicitely define MAP_SHARED_VALIDATE as the first unused value
> among mapping types (which is in fact enum embedded inside mapping flags).
> I.e. 0x03 on alpha, x86, and probably all other archs - it has nothing to
> do with MAP_SHARED|MAP_PRIVATE - it is just another type of the mapping
> which happens to have most of the MAP_SHARED semantics...

Ok, I'll make it 0x3 everywhere.