Re: [PATCH 0/4] move per-vma lock into vm_area_struct

From: Davidlohr Bueso
Date: Mon Nov 11 2024 - 17:46:45 EST


On Mon, 11 Nov 2024, Suren Baghdasaryan wrote:

To minimize memory overhead, vm_lock implementation is changed from
using rw_semaphore (40 bytes) to an atomic (8 bytes) and several
vm_area_struct members are moved into the last cacheline, resulting
in a less fragmented structure:

I am not a fan of building a custom lock, replacing a standard one.
How much do we really care about this? rwsems are quite optimized
and are known to heavily affect mm performance altogether.

...

Performance measurements using pft test on x86 do not show considerable
difference, on Pixel 6 running Android it results in 3-5% improvement in
faults per second.

pft is a very micro benchmark, these results do not justify this change, imo.

Thanks,
Davidlohr