[RFC PATCH 1/2] powerpc/numa: fix cpu_to_node() usage during boot

[David Rientjes]

On Thu, 2 Jul 2015, Nishanth Aravamudan wrote:

> Much like on x86, now that powerpc is using USE_PERCPU_NUMA_NODE_ID, we
> have an ordering issue during boot with early calls to cpu_to_node().
> The value returned by those calls now depend on the per-cpu area being
> setup, but that is not guaranteed to be the case during boot. Instead,
> we need to add an early_cpu_to_node() which doesn't use the per-CPU area
> and call that from certain spots that are known to invoke cpu_to_node()
> before the per-CPU areas are not configured.
> 
> On an example 2-node NUMA system with the following topology:
> 
> available: 2 nodes (0-1)
> node 0 cpus: 0 1 2 3
> node 0 size: 2029 MB
> node 0 free: 1753 MB
> node 1 cpus: 4 5 6 7
> node 1 size: 2045 MB
> node 1 free: 1945 MB
> node distances:
> node   0   1 
>   0:  10  40 
>   1:  40  10 
> 
> we currently emit at boot:
> 
> [    0.000000] pcpu-alloc: [0] 0 1 2 3 [0] 4 5 6 7 
> 
> After this commit, we correctly emit:
> 
> [    0.000000] pcpu-alloc: [0] 0 1 2 3 [1] 4 5 6 7 
> 
> Signed-off-by: Nishanth Aravamudan <nacc at linux.vnet.ibm.com>
> 
> diff --git a/arch/powerpc/include/asm/topology.h b/arch/powerpc/include/asm/topology.h
> index 5f1048e..f2c4c89 100644
> --- a/arch/powerpc/include/asm/topology.h
> +++ b/arch/powerpc/include/asm/topology.h
> @@ -39,6 +39,8 @@ static inline int pcibus_to_node(struct pci_bus *bus)
>  extern int __node_distance(int, int);
>  #define node_distance(a, b) __node_distance(a, b)
>  
> +extern int early_cpu_to_node(int);
> +
>  extern void __init dump_numa_cpu_topology(void);
>  
>  extern int sysfs_add_device_to_node(struct device *dev, int nid);
> diff --git a/arch/powerpc/kernel/setup_64.c b/arch/powerpc/kernel/setup_64.c
> index c69671c..23a2cf3 100644
> --- a/arch/powerpc/kernel/setup_64.c
> +++ b/arch/powerpc/kernel/setup_64.c
> @@ -715,8 +715,8 @@ void __init setup_arch(char **cmdline_p)
>  
>  static void * __init pcpu_fc_alloc(unsigned int cpu, size_t size, size_t align)
>  {
> -	return __alloc_bootmem_node(NODE_DATA(cpu_to_node(cpu)), size, align,
> -				    __pa(MAX_DMA_ADDRESS));
> +	return __alloc_bootmem_node(NODE_DATA(early_cpu_to_node(cpu)), size,
> +				    align, __pa(MAX_DMA_ADDRESS));
>  }
>  
>  static void __init pcpu_fc_free(void *ptr, size_t size)
> @@ -726,7 +726,7 @@ static void __init pcpu_fc_free(void *ptr, size_t size)
>  
>  static int pcpu_cpu_distance(unsigned int from, unsigned int to)
>  {
> -	if (cpu_to_node(from) == cpu_to_node(to))
> +	if (early_cpu_to_node(from) == early_cpu_to_node(to))
>  		return LOCAL_DISTANCE;
>  	else
>  		return REMOTE_DISTANCE;
> diff --git a/arch/powerpc/mm/numa.c b/arch/powerpc/mm/numa.c
> index 5e80621..9ffabf4 100644
> --- a/arch/powerpc/mm/numa.c
> +++ b/arch/powerpc/mm/numa.c
> @@ -157,6 +157,11 @@ static void map_cpu_to_node(int cpu, int node)
>  		cpumask_set_cpu(cpu, node_to_cpumask_map[node]);
>  }
>  
> +int early_cpu_to_node(int cpu)
> +{
> +	return numa_cpu_lookup_table[cpu];
> +}
> +
>  #if defined(CONFIG_HOTPLUG_CPU) || defined(CONFIG_PPC_SPLPAR)
>  static void unmap_cpu_from_node(unsigned long cpu)
>  {
> 
> 

early_cpu_to_node() looks like it's begging to be __init since we 
shouldn't have a need to reference to numa_cpu_lookup_table after boot and 
that appears like it can be done if pcpu_cpu_distance() is made __init in 
this patch and smp_prepare_boot_cpu() is made __init in the next patch.  
So I think this is fine, but those functions and things like 
reset_numa_cpu_lookup_table() should be in init.text.

After the percpu areas on initialized and cpu_to_node() is correct, it 
would be really nice to be able to make numa_cpu_lookup_table[] be 
__initdata since it shouldn't be necessary anymore.  That probably has cpu 
callbacks that need to be modified to no longer look at 
numa_cpu_lookup_table[] or pass the value in, but it would make it much 
cleaner.  Then nobody will have to worry about figuring out whether 
early_cpu_to_node() or cpu_to_node() is the right one to call.