Re: [PATCH] Fix isocpus's param handling when CPUMASK_OFFSTACK=n.

From: Ingo Molnar
Date: Mon Oct 23 2017 - 07:51:07 EST



* Rakib Mullick <rakib.mullick@xxxxxxxxx> wrote:

> > *On Fri, Oct 20, 2017 at 2:49 PM, Ingo Molnar <mingo@xxxxxxxxxx> wrote:
> >
> >> Rakib Mullick <rakib.mullick@xxxxxxxxx> wrote:
> >> include/linux/cpumask.h | 16 ++++++++++++++++
> >> kernel/sched/topology.c | 8 +++++---
> >> 2 files changed, 21 insertions(+), 3 deletions(-)
> >
> > What kernel is this against? It does not apply to the latest kernels.
>
> It was against 4.14-rc4, prepared before -rc5 release. Please, consider
> the below one, against -rc5.
>
> cpulist_parse() uses nr_cpumask_bits as limit to parse the
> passed buffer from kernel commandline. What nr_cpumask_bits
> represents varies depends upon CONFIG_CPUMASK_OFFSTACK option.
> If CONFIG_CPUMASK_OFFSTACK=n, then nr_cpumask_bits is same as
> NR_CPUS, which might not represent the # of cpus really exist
> (default 64). So, there's a chance of gap between nr_cpu_ids
> and NR_CPUS, which ultimately lead towards invalid cpulist_parse()
> operation. For example, if isolcpus=9 is passed on a 8 cpu
> system (CONFIG_CPUMASK_OFFSTACK=n) it doesn't show the error
> that it suppose to.
>
> This patch fixes this issue by effectively find out the last
> cpu of the passed isolcpus list and checking it with nr_cpu_ids.
> Also, fixes the error message where the nr_cpu_ids should be
> nr_cpu_ids-1, since the cpu numbering starts from 0.
>
> Signed-off-by: Rakib Mullick <rakib.mullick@xxxxxxxxx>
> ---
> include/linux/cpumask.h | 16 ++++++++++++++++
> kernel/sched/topology.c | 8 +++++---
> 2 files changed, 21 insertions(+), 3 deletions(-)
>
> diff --git a/include/linux/cpumask.h b/include/linux/cpumask.h
> index cd415b7..5631725 100644
> --- a/include/linux/cpumask.h
> +++ b/include/linux/cpumask.h
> @@ -130,6 +130,11 @@ static inline unsigned int cpumask_first(const struct cpumask *srcp)
> 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)
> {
> @@ -178,6 +183,17 @@ static inline unsigned int cpumask_first(const struct cpumask *srcp)
> return find_first_bit(cpumask_bits(srcp), nr_cpumask_bits);
> }
>
> +/**
> + * cpumask_last - get the last cpu in a cpumask

Please capitalize 'CPU' properly in documentation.

> + * @srcp: - the cpumask pointer
> + *
> + * Returns >= nr_cpumask_bits if no cpus set.
> + */
> +static inline unsigned int cpumask_last(const struct cpumask *srcp)
> +{
> + return find_last_bit(cpumask_bits(srcp), nr_cpumask_bits);
> +}
> +
> unsigned int cpumask_next(int n, const struct cpumask *srcp);
>
> /**
> diff --git a/kernel/sched/topology.c b/kernel/sched/topology.c
> index f1cf4f3..b9265c8 100644
> --- a/kernel/sched/topology.c
> +++ b/kernel/sched/topology.c
> @@ -466,12 +466,14 @@ cpu_attach_domain(struct sched_domain *sd, struct root_domain *rd, int cpu)
> /* Setup the mask of CPUs configured for isolated domains */
> static int __init isolated_cpu_setup(char *str)
> {
> - int ret;
> + int ret, lastcpu;
>
> alloc_bootmem_cpumask_var(&cpu_isolated_map);
> ret = cpulist_parse(str, cpu_isolated_map);
> - if (ret) {
> - pr_err("sched: Error, all isolcpus= values must be between 0 and %u\n", nr_cpu_ids);
> + lastcpu = cpumask_last(cpu_isolated_map);
> + if (ret || lastcpu >= nr_cpu_ids) {

Any reason why 'lastcpu' has to be introduced - why not just use cpumask_last()
directly in the condition? It looks obvious enough of a pattern.

> + pr_err("sched: Error, all isolcpus= values must be between 0 and %u\n",
> + nr_cpu_ids-1);

Please don't break the line mindlessly just due to checkpatch complaining - it
makes the code less readable.

Thanks,

Ingo