Re: [PATCH v4 2/5] cpumask: Fix invalid uniprocessor mask assumption

From: Sander Vanheule
Date: Sun Jul 03 2022 - 10:17:43 EST


On Sat, 2022-07-02 at 14:42 -0700, Yury Norov wrote:
> On Sat, Jul 02, 2022 at 06:08:25PM +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@xxxxxxxxxxxxx>
>
> Suggested-by: Yury Norov <yury.norov@xxxxxxxxx>
>

To be honest, I was somewhat taken aback by this Suggested-by. I certainly appreciate the feedback
you've provided, and I think the patch has gotten better for it, so I do want to acknowledge that.
During reviews of other patches however, I haven't had a reviewer request this for suggestions that
couldn't be spun off into a separate patch. If you are OK with this patch, a Reviewed-by would
definitely be warranted and seems more appropriate IMHO.


Best,
Sander

> > ---
> >
> > Notes:
> >     Changes since v3:
> >     - Add back UP-optimised cpumask_local_spread, cpumask_any_distribute,
> >       cpumask_any_and_distribute
> >    
> >     Changes since v1:
> >     - Drop UP implementations instead of trying to fix them
> >
> >  include/linux/cpumask.h | 99 ++++++++---------------------------------
> >  lib/Makefile            |  3 +-
> >  lib/cpumask.c           |  2 +
> >  3 files changed, 22 insertions(+), 82 deletions(-)
> >
> > diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
> > index fe29ac7cc469..7fbef41b3093 100644
> > --- a/include/linux/cpumask.h
> > +++ b/include/linux/cpumask.h
> > @@ -116,85 +116,6 @@ static __always_inline unsigned int cpumask_check(unsigned int cpu)
> >         return cpu;
> >  }
> >  
> > -#if NR_CPUS == 1
> > -/* Uniprocessor.  Assume all masks are "1". */
> > -static inline unsigned int cpumask_first(const struct cpumask *srcp)
> > -{
> > -       return 0;
> > -}
> > -
> > -static inline unsigned int cpumask_first_zero(const struct cpumask *srcp)
> > -{
> > -       return 0;
> > -}
> > -
> > -static inline unsigned int cpumask_first_and(const struct cpumask *srcp1,
> > -                                            const struct cpumask *srcp2)
> > -{
> > -       return 0;
> > -}
> > -
> > -static inline unsigned int cpumask_last(const struct cpumask *srcp)
> > -{
> > -       return 0;
> > -}
> > -
> > -/* Valid inputs for n are -1 and 0. */
> > -static inline unsigned int cpumask_next(int n, const struct cpumask *srcp)
> > -{
> > -       return n+1;
> > -}
> > -
> > -static inline unsigned int cpumask_next_zero(int n, const struct cpumask *srcp)
> > -{
> > -       return n+1;
> > -}
> > -
> > -static inline unsigned int cpumask_next_and(int n,
> > -                                           const struct cpumask *srcp,
> > -                                           const struct cpumask *andp)
> > -{
> > -       return n+1;
> > -}
> > -
> > -static inline unsigned int cpumask_next_wrap(int n, const struct cpumask *mask,
> > -                                            int start, bool wrap)
> > -{
> > -       /* cpu0 unless stop condition, wrap and at cpu0, then nr_cpumask_bits */
> > -       return (wrap && n == 0);
> > -}
> > -
> > -/* cpu must be a valid cpu, ie 0, so there's no other choice. */
> > -static inline unsigned int cpumask_any_but(const struct cpumask *mask,
> > -                                          unsigned int cpu)
> > -{
> > -       return 1;
> > -}
> > -
> > -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);
> > -}
> > -
> > -#define for_each_cpu(cpu, mask)                        \
> > -       for ((cpu) = 0; (cpu) < 1; (cpu)++, (void)mask)
> > -#define for_each_cpu_not(cpu, mask)            \
> > -       for ((cpu) = 0; (cpu) < 1; (cpu)++, (void)mask)
> > -#define for_each_cpu_wrap(cpu, mask, start)    \
> > -       for ((cpu) = 0; (cpu) < 1; (cpu)++, (void)mask, (void)(start))
> > -#define for_each_cpu_and(cpu, mask1, mask2)    \
> > -       for ((cpu) = 0; (cpu) < 1; (cpu)++, (void)mask1, (void)mask2)
> > -#else
> >  /**
> >   * cpumask_first - get the first cpu in a cpumask
> >   * @srcp: the cpumask pointer
> > @@ -260,10 +181,29 @@ static inline unsigned int cpumask_next_zero(int n, const struct cpumask
> > *srcp)
> >  
> >  int __pure cpumask_next_and(int n, const struct cpumask *, const struct cpumask *);
> >  int __pure cpumask_any_but(const struct cpumask *mask, unsigned int cpu);
> > +
> > +#if NR_CPUS == 1
> > +/* Uniprocessor: there is only one valid CPU */
> > +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);
> > +}
> > +#else
> >  unsigned int cpumask_local_spread(unsigned int i, int node);
> >  int cpumask_any_and_distribute(const struct cpumask *src1p,
> >                                const struct cpumask *src2p);
> >  int cpumask_any_distribute(const struct cpumask *srcp);
> > +#endif /* NR_CPUS */
> >  
> >  /**
> >   * for_each_cpu - iterate over every cpu in a mask
> > @@ -324,7 +264,6 @@ extern int cpumask_next_wrap(int n, const struct cpumask *mask, int start,
> > bool
> >         for ((cpu) = -1;                                                \
> >                 (cpu) = cpumask_next_and((cpu), (mask1), (mask2)),      \
> >                 (cpu) < nr_cpu_ids;)
> > -#endif /* SMP */
> >  
> >  #define CPU_BITS_NONE                                          \
> >  {                                                              \
> > diff --git a/lib/Makefile b/lib/Makefile
> > index f99bf61f8bbc..bcc7e8ea0cde 100644
> > --- a/lib/Makefile
> > +++ b/lib/Makefile
> > @@ -34,10 +34,9 @@ lib-y := ctype.o string.o vsprintf.o cmdline.o \
> >          is_single_threaded.o plist.o decompress.o kobject_uevent.o \
> >          earlycpio.o seq_buf.o siphash.o dec_and_lock.o \
> >          nmi_backtrace.o nodemask.o win_minmax.o memcat_p.o \
> > -        buildid.o
> > +        buildid.o cpumask.o
> >  
> >  lib-$(CONFIG_PRINTK) += dump_stack.o
> > -lib-$(CONFIG_SMP) += cpumask.o
> >  
> >  lib-y  += kobject.o klist.o
> >  obj-y  += lockref.o
> > diff --git a/lib/cpumask.c b/lib/cpumask.c
> > index a971a82d2f43..b9728513a4d4 100644
> > --- a/lib/cpumask.c
> > +++ b/lib/cpumask.c
> > @@ -192,6 +192,7 @@ void __init free_bootmem_cpumask_var(cpumask_var_t mask)
> >  }
> >  #endif
> >  
> > +#if NR_CPUS > 1
> >  /**
> >   * cpumask_local_spread - select the i'th cpu with local numa cpu's first
> >   * @i: index number
> > @@ -279,3 +280,4 @@ int cpumask_any_distribute(const struct cpumask *srcp)
> >         return next;
> >  }
> >  EXPORT_SYMBOL(cpumask_any_distribute);
> > +#endif /* NR_CPUS */
> > --
> > 2.36.1