Re: [PATCH v6 0/4] mm/zsmalloc: reduce lock contention in zs_free()

[Sergey Senozhatsky <senozhatsky@chromium.org>]

On (26/07/25 11:57), Barry Song (Xiaomi) wrote:
> On Fri, Jul 24, 2026 at 12:55 PM Sergey Senozhatsky <senozhatsky@chromium.org> wrote:
> >
> > Apologies for the delay.
> >
> > On (26/06/27 21:36), Andrew Morton wrote:
> > [..]
> > > > Raspberry Pi 4B (4-core ARM64 Cortex-A72):
> > > >
> > > >   mode        Base       Patched     Speedup
> > > >   single      59.0ms     56.0ms      1.05x
> > > >   multi 2p    94.6ms     66.7ms      1.42x
> > > >   multi 4p    202.9ms    110.6ms     1.83x
> > > >
> > > > x86 (20-core Intel i7-12700, 16 concurrent processes):
> > > >
> > > >   mode        Base       Patched     Speedup
> > > >   single      11.7ms     9.8ms       1.19x
> > > >   multi 2p    24.1ms     17.2ms      1.40x
> > > >   multi 4p    63.0ms     45.3ms      1.39x
> > >
> > > Well that's a nice result.
> >
> > Indeed.
> >
> > > Sashiko AI review said ....  nothing.  I don't recall seeing that
> > > before ;)
> >
> > grep-ing the code, I wonder if we maybe want to tweak some comments.
> > Namely, zs_page_migrate() says:
> >
> >         The pool migrate_lock protects the race between zpage migration
> >         and zs_free
> >
> > which is not entirely true anymore, as zs_free() has a lockless path
> > that doesn't take pool look (when ZS_OBJ_CLASS_BITS permit).
> >
> > Likewise, __zs_compact() says the following about pool migrate_lock:
> >
> >         protect the race between zpage migration and zs_free
> >
> > None of these are critical, very minor.
> 
> Does it look good to you if we squash the below change into patch2/4:

Sure, looks good to me!