Re: [PATCH] x86: fix system without memory on node0

[Yinghai Lu]

Andi Kleen wrote:
>> +#if defined(CONFIG_X86_64) && defined(CONFIG_NUMA)
>> +	/*
>> +	 * make sure boot cpu node_number is right, when boot cpu is on the
>> +	 * node that doesn't have mem installed
>> +	 */
>> +	per_cpu(node_number, boot_cpu_id) = cpu_to_node(boot_cpu_id);
>> +#endif
> 
> Seems like a quite crappy hac^wpatch. Why is it ever set to the wrong
> value? And why is that only the case on NUMA and 64bit? 

two places touched that per_cpu(node_number,), 
1. in cpu/common.c::cpu_init() and it is not for BP
#ifdef CONFIG_NUMA
        if (cpu != 0 && percpu_read(node_number) == 0 &&
            cpu_to_node(cpu) != NUMA_NO_NODE)
                percpu_write(node_number, cpu_to_node(cpu));
#endif
for BP is traps_init ==> cpu_init
for AP is start_secondary ==> cpu_init

2. cpu/intel.c or amd.c::srat_detect_node via numa_set_node() and they are called via identify_cpu
for BP:
check_bugs ==> identify_boot_cpu ==> identify_cpu() that is rather later before numa_node_id() is used for BP...
for AP:
start_secondary=>smp_callin=>smp_store_cpu_info()=>identify_secondary_cpu ==> identify_cpu()

so only try to set that for BP more early in setup_per_cpu_areas, and don't bother set that for APs there.
(and don't want to mess the 0 before the copying BP per_cpu to APs)

or you check set the per_cpu(node_number) is early enough with setup_per_cpu_areas();


setup_percpu.c in arch/x86/kernel/ is used on lot of conf.

and in arch/x86/include/asm/topology.h, we have

#ifdef CONFIG_NUMA
#include <linux/cpumask.h>
#include <asm/mpspec.h>

#ifdef CONFIG_X86_32

/* Mappings between logical cpu number and node number */
extern int cpu_to_node_map[];

/* Returns the number of the node containing CPU 'cpu' */
static inline int cpu_to_node(int cpu)
{
        return cpu_to_node_map[cpu];
}
#define early_cpu_to_node(cpu)  cpu_to_node(cpu)

#else /* CONFIG_X86_64 */

/* Mappings between logical cpu number and node number */
DECLARE_EARLY_PER_CPU(int, x86_cpu_to_node_map);

/* Returns the number of the current Node. */
DECLARE_PER_CPU(int, node_number);
#define numa_node_id()          percpu_read(node_number)
...

so we need to for NUMA and 64 bit.


YH
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/