[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <541d89b3aa2092382405672b70b58b5b3d263f1e.camel@svanheule.net>
Date: Tue, 07 Jun 2022 14:06:38 +0200
From: Sander Vanheule <sander@...nheule.net>
To: Yury Norov <yury.norov@...il.com>
Cc: Peter Zijlstra <peterz@...radead.org>,
Andrew Morton <akpm@...ux-foundation.org>,
Valentin Schneider <vschneid@...hat.com>,
Thomas Gleixner <tglx@...utronix.de>,
Greg Kroah-Hartman <gregkh@...uxfoundation.org>,
Marco Elver <elver@...gle.com>, linux-kernel@...r.kernel.org,
Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
Subject: Re: [PATCH v3 1/4] cpumask: Fix invalid uniprocessor mask assumption
On Mon, 2022-06-06 at 12:14 -0700, Yury Norov wrote:
> On Sun, Jun 05, 2022 at 08:22:38AM +0200, Sander Vanheule wrote:
> > On uniprocessor builds, any CPU mask is assumed to contain exactly one
> > CPU (cpu0). This assumption ignores the existence of empty masks,
> > resulting in incorrect behaviour.
> > cpumask_first_zero(), cpumask_next_zero(), and for_each_cpu_not() don't
> > provide behaviour matching the assumption that a UP mask is always "1",
> > and instead provide behaviour matching the empty mask.
> >
> > Drop the incorrectly optimised code and use the generic implementations
> > in all cases.
> >
> > Signed-off-by: Sander Vanheule <sander@...nheule.net>
> > ---
[...]
> > -static inline unsigned int cpumask_local_spread(unsigned int i, int node)
> > -{
> > - return 0;
> > -}
> > -
> > -static inline int cpumask_any_and_distribute(const struct cpumask *src1p,
> > - const struct cpumask *src2p) {
> > - return cpumask_first_and(src1p, src2p);
> > -}
> > -
> > -static inline int cpumask_any_distribute(const struct cpumask *srcp)
> > -{
> > - return cpumask_first(srcp);
> > -}
>
> It looks like cpumask_local_spread, cpumask_any_and_distribute and
> cpumask_any_distribute were correct and better optimized in UP case.
> cpumask_local_spread - for sure. I think it's worth keeping them
> optimized.
Yes, these were correct and we can keep them. I will have to add an #ifded CONFIG_SMP (or #if
NR_CPUS > 1) guard in lib/cpumask.c around these functions, so they don't collide with the inlined
UP versions.
Best,
Sander
Powered by blists - more mailing lists