Hi Dietmar, Lukasz,
Thanks for digging this bug and fixing it.
On Wed, Nov 20, 2019 at 10:42:12AM +0000, Dietmar Eggemann wrote:
> Commit ca74b316df96 ("arm: Use common cpu_topology structure and
> functions.") changed cpu_coregroup_mask() from the ARM32 specific
> implementation in arch/arm/include/asm/topology.h to the one shared
> with ARM64 and RISCV in drivers/base/arch_topology.c.
>
> Currently on Arm32 (TC2 w/ CONFIG_SCHED_MC) the task scheduler setup
> code (w/ CONFIG_SCHED_DEBUG) shows this during CPU hotplug:
>
> ERROR: groups don't span domain->span
>
> It happens to CPUs of the cluster of the CPU which gets hot-plugged
> out on scheduler domain MC.
>
> Turns out that the shared cpu_coregroup_mask() requires that the
> hot-plugged CPU is removed from the core_sibling mask via
> remove_cpu_topology(). Otherwise the 'is core_sibling subset of
> cpumask_of_node()' doesn't work. In this case the task scheduler has to
> deal with cpumask_of_node instead of core_sibling which is wrong on
> scheduler domain MC.
>
> e.g. CPU3 hot-plugged out on TC2 [cluster0: 0,3-4 cluster1: 1-2]:
>
> cpu_coregroup_mask(): CPU3 cpumask_of_node=0-2,4 core_sibling=0,3-4
> ^
> should be:
>
> cpu_coregroup_mask(): CPU3 cpumask_of_node=0-2,4 core_sibling=0,4
>
> Add remove_cpu_topology() to __cpu_disable() to remove the CPU from the
> topology masks in case of a CPU hotplug out operation.
>
> At the same time tweak store_cpu_topology() slightly so it will call
> update_siblings_masks() in case of CPU hotplug in operation via
> secondary_start_kernel()->smp_store_cpu_info().
>
> This aligns the Arm32 implementation with the Arm64 one.
>
> Fixes: ca74b316df96 ("arm: Use common cpu_topology structure and functions")
Apart from the minor nit below,
Reviewed-by: Sudeep Holla <[email protected]>
> Signed-off-by: Dietmar Eggemann <[email protected]>
> ---
> arch/arm/kernel/smp.c | 2 ++
> arch/arm/kernel/topology.c | 15 +++++++--------
> 2 files changed, 9 insertions(+), 8 deletions(-)
>
> diff --git a/arch/arm/kernel/smp.c b/arch/arm/kernel/smp.c
> index 4b0bab2607e4..139c0d98fa29 100644
> --- a/arch/arm/kernel/smp.c
> +++ b/arch/arm/kernel/smp.c
> @@ -240,6 +240,8 @@ int __cpu_disable(void)
> if (ret)
> return ret;
>
> + remove_cpu_topology(cpu);
> +
> /*
> * Take this CPU offline. Once we clear this, we can't return,
> * and we must not schedule until we're ready to give up the cpu.
> diff --git a/arch/arm/kernel/topology.c b/arch/arm/kernel/topology.c
> index 5b9faba03afb..b37b0a340991 100644
> --- a/arch/arm/kernel/topology.c
> +++ b/arch/arm/kernel/topology.c
> @@ -196,9 +196,8 @@ void store_cpu_topology(unsigned int cpuid)
> struct cpu_topology *cpuid_topo = &cpu_topology[cpuid];
> unsigned int mpidr;
>
> - /* If the cpu topology has been already set, just return */
> - if (cpuid_topo->core_id != -1)
> - return;
> + if (cpuid_topo->package_id != -1)
> + goto topology_populated;
>
> mpidr = read_cpuid_mpidr();
>
> @@ -231,14 +230,14 @@ void store_cpu_topology(unsigned int cpuid)
> cpuid_topo->package_id = -1;
> }
>
> - update_siblings_masks(cpuid);
> -
> update_cpu_capacity(cpuid);
>
> pr_info("CPU%u: thread %d, cpu %d, socket %d, mpidr %x\n",
> - cpuid, cpu_topology[cpuid].thread_id,
> - cpu_topology[cpuid].core_id,
> - cpu_topology[cpuid].package_id, mpidr);
> + cpuid, cpuid_topo->thread_id, cpuid_topo->core_id,
> + cpuid_topo->package_id, mpidr);
> +
[nit] The above is a clearly cosmetic cleanup and shouldn't be part of
this fix as they will be backported automatically. So I prefer to drop
this or make it separate patch if required.
--
Regards,
Sudeep
On 20.11.19 11:09, Sudeep Holla wrote:
> Hi Dietmar, Lukasz,
>
> Thanks for digging this bug and fixing it.
>
> On Wed, Nov 20, 2019 at 10:42:12AM +0000, Dietmar Eggemann wrote:
[...]
>> @@ -231,14 +230,14 @@ void store_cpu_topology(unsigned int cpuid)
>> cpuid_topo->package_id = -1;
>> }
>>
>> - update_siblings_masks(cpuid);
>> -
>> update_cpu_capacity(cpuid);
>>
>> pr_info("CPU%u: thread %d, cpu %d, socket %d, mpidr %x\n",
>> - cpuid, cpu_topology[cpuid].thread_id,
>> - cpu_topology[cpuid].core_id,
>> - cpu_topology[cpuid].package_id, mpidr);
>> + cpuid, cpuid_topo->thread_id, cpuid_topo->core_id,
>> + cpuid_topo->package_id, mpidr);
>> +
>
> [nit] The above is a clearly cosmetic cleanup and shouldn't be part of
> this fix as they will be backported automatically. So I prefer to drop
> this or make it separate patch if required.
I can get rid of this cleanup in v2. Thanks for the review!