Re: [PATCH] x86-64: Restore LD_PREFER_MAP_32BIT_EXEC support [BZ #28656]

[Carlos O'Donell <carlos@redhat.com>]

On 2/10/23 11:50, H.J. Lu wrote:
> On Thu, Feb 9, 2023 at 6:35 AM Florian Weimer <fweimer@redhat.com> wrote:
>>
>> * H. J. Lu via Libc-alpha:
>>
>>> Crossing 2GB boundaries with indirect calls and jumps can use more
>>> branch prediction resources on Intel Golden Cove CPU (see the
>>> "Misprediction for Branches >2GB" section in Intel 64 and IA-32
>>> Architectures Optimization Reference Manual.)  There is visible
>>> performance improvement on workloads with many PLT calls when executable
>>> and shared libraries are mmapped below 2GB.  Add the Prefer_MAP_32BIT_EXEC
>>> bit so that mmap will try to map executable or denywrite pages in shared
>>> libraries with MAP_32BIT first.
>>>
>>> NB: Prefer_MAP_32BIT_EXEC reduces bits available for address space
>>> layout randomization (ASLR), which is always disabled for SUID programs
>>> and can only be enabled by setting environment variable,
>>> LD_PREFER_MAP_32BIT_EXEC.  LD_PREFER_MAP_32BIT_EXEC works only between
>>> shared libraries or between shared libraries and executables with
>>> addresses below 2GB.  PIEs are usually mapped above 4GB by the kernel.
>>
>> I still think we should fix this in the kernel, using MAP_DENYWRITE as a
>> hint for placement.  This way, it's easier to turn it on unconditionally
>> for the whole system because the lower 4 GiB will not be polluted by
>> code mappings.
>>
> 
> Kernel MM change may take a long time and this mitigation may only be needed
> for a few specific applications.   On the other hand, it is simpler to
> use tunable
> instead.

I tend to agree with you here that a Kernel MM change is going to take too long and may
be possibly fragile.

I would like to see this use a tunable with an alias.

Note that the Linux man pages continue to document LD_PREFER_MAP_32BIT_EXEC :-)

-- 
Cheers,
Carlos.