Date:	Tue, 15 Apr 2008 13:37:33 -0700
From:	Mike Travis <travis@....com>
To:	Ingo Molnar <mingo@...e.hu>
CC:	Thomas Gleixner <tglx@...utronix.de>,
	"H. Peter Anvin" <hpa@...or.com>,
	Andrew Morton <akpm@...ux-foundation.org>,
	linux-kernel@...r.kernel.org
Subject: Re: [PATCH 2/2] x86: cleanup early per cpu variables/accesses

Hi Ingo,

Would you please hold off on this patch.  I just discovered I got caught
in an ifdef/ifndef conflict in topology.h shortly after I updated to
the latest x86.git code base.  I'll resubmit shortly.

Thanks,
Mike


Mike Travis wrote:
>   * Introduce a new PER_CPU macro called "EARLY_PER_CPU".  This is
>     used by some per_cpu variables that are initialized and accessed
>     before there are per_cpu areas allocated.
> 
>     ["Early" in respect to per_cpu variables is "earlier than the per_cpu
>     areas have been setup".]
> 
>     This patchset adds these new macros:
> 
> 	DEFINE_EARLY_PER_CPU(_type, _name, _initvalue)
> 	EXPORT_EARLY_PER_CPU_SYMBOL(_name)
> 	DECLARE_EARLY_PER_CPU(_type, _name)
> 
> 	early_per_cpu_ptr(_name)
> 	early_per_cpu_map(_name, _idx)
> 	early_per_cpu(_name, _cpu)
> 
>     The DEFINE macro defines the per_cpu variable as well as the early
>     map and pointer.  It also initializes the per_cpu variable and map
>     elements to "_initvalue".  The early_* macros provide access to
>     the initial map (usually setup during system init) and the early
>     pointer.  This pointer is initialized to point to the early map
>     but is then NULL'ed when the actual per_cpu areas are setup.  After
>     that the per_cpu variable is the correct access to the variable.
> 
>     The early_per_cpu() macro is not very efficient but does show how to
>     access the variable if you have a function that can be called both
>     "early" and "late".  It tests the early ptr to be NULL, and if not
>     then it's still valid.  Otherwise, the per_cpu variable is used
>     instead:
> 
> 	#define early_per_cpu(_name, _cpu) 			\
> 		(early_per_cpu_ptr(_name) ?			\
> 			early_per_cpu_ptr(_name)[_cpu] :	\
> 			per_cpu(_name, _cpu))
> 
> 
>     A better method is to actually check the pointer manually.  In the
>     case below, numa_set_node can be called both "early" and "late":
> 
> 	void __cpuinit numa_set_node(int cpu, int node)
> 	{
> 	    int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map);
> 
> 	    if (cpu_to_node_map)
> 		    cpu_to_node_map[cpu] = node;
> 	    else 
> 		    per_cpu(x86_cpu_to_node_map, cpu) = node;
> 	}
> 
>   * Additionally, some scattered x86 code variations are consolidated
>     aiding the merger of i386/x86_64.  The affected variables are:
> 
> 	x86_bios_cpu_apicid
> 	x86_cpu_to_apicid
> 	x86_cpu_to_node_map
> 	node_to_cpumask_map
> 	cpu_to_node_map(removed)
> 
>   * Change some calls from early_cpu_to_node to (non-early) cpu_to_node.
>     This reduces the amount of code generated.  An example is:
> 
>      /* Returns the number of the current Node. */
>     -#define numa_node_id()	(early_cpu_to_node(raw_smp_processor_id()))
>     +#define numa_node_id()	(__get_cpu_var(x86_cpu_to_node_map))
> 
>     In this case,  x86_cpu_to_node_map will always be correct.  If it's
>     running on the boot cpu (which is always granted cpu index of 0 and
>     it's node will also be zero) then it is initialized correctly.
>     If, for some reason the system needs to re-boot say a guest OS, with
>     a cpu/node other than zero, then the initial cpu and node indices
>     should be adjusted before continuing startup.
> 
>   * The coverage of CONFIG_DEBUG_PER_CPU_MAPS has been increased while
>     the non-debug case has been optimized a bit.
> 
> Based on:
> 	git://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux-2.6.git
>     +   x86/latest          .../x86/linux-2.6-x86.git
>     +   sched-devel/latest  .../mingo/linux-2.6-sched-devel.git
> 
> # for checkpatch "incidents":
> Cc:      Andy Whitcroft <apw@...dowen.org>
> Cc:      Randy Dunlap <rdunlap@...otime.net>
> Cc:      Joel Schopp <jschopp@...tin.ibm.com>
> 
> Signed-off-by: Mike Travis <travis@....com>
> ---
> checkpatch.pl results: 3 "errors"/1 "warning"...
> 
> I'm not sure how to reformat this to get rid of these errors/warnings:
> 
> #define	DEFINE_EARLY_PER_CPU(_type, _name, _initvalue)			\
> 	DEFINE_PER_CPU(_type, _name) = _initvalue;			\
> 	__typeof__(_type) _name##_early_map[NR_CPUS] __initdata =	\
> 				{ [0 ... NR_CPUS-1] = _initvalue };	\
> 	__typeof__(_type) *_name##_early_ptr = _name##_early_map
> 
> #define DECLARE_EARLY_PER_CPU(_type, _name)			\
> 	DECLARE_PER_CPU(_type, _name);				\
> 	extern __typeof__(_type) *_name##_early_ptr;		\
> 	extern __typeof__(_type)  _name##_early_map[]
> 
> ERROR: Macros with multiple statements should be enclosed in a do - while loop
> #601: FILE: include/asm-x86/percpu.h:156:
> +	DEFINE_PER_CPU(_type, _name) = _initvalue;			\
> 
> WARNING: usage of NR_CPUS is often wrong - consider using cpu_possible(), num_possible_cpus(), for_each_possible_cpu(), etc
> #602: FILE: include/asm-x86/percpu.h:157:
> +	__typeof__(_type) _name##_early_map[NR_CPUS] __initdata =	\
> 
> WARNING: usage of NR_CPUS is often wrong - consider using cpu_possible(), num_possible_cpus(), for_each_possible_cpu(), etc
> #603: FILE: include/asm-x86/percpu.h:158:
> +				{ [0 ... NR_CPUS-1] = _initvalue };	\
> 
> ERROR: Macros with multiple statements should be enclosed in a do - while loop
> #610: FILE: include/asm-x86/percpu.h:165:
> +	DECLARE_PER_CPU(_type, _name);				\
> 
> total: 3 errors, 2 warnings, 659 lines checked
> ---
>  arch/x86/kernel/apic_32.c  |    9 +--
>  arch/x86/kernel/apic_64.c  |   11 +---
>  arch/x86/kernel/setup.c    |   79 ++++++++++++++++++++++++++++----
>  arch/x86/kernel/setup_32.c |   24 ----------
>  arch/x86/kernel/setup_64.c |    9 ---
>  arch/x86/kernel/smpboot.c  |   41 +++--------------
>  arch/x86/mm/numa_64.c      |   43 ++++-------------
>  include/asm-x86/numa.h     |   28 +++++++++++
>  include/asm-x86/numa_64.h  |   15 ------
>  include/asm-x86/percpu.h   |   46 +++++++++++++++++++
>  include/asm-x86/smp.h      |   15 +-----
>  include/asm-x86/topology.h |  108 ++++++++++++++++++++-------------------------
>  12 files changed, 221 insertions(+), 207 deletions(-)
> 
> --- linux-2.6.x86.sched.orig/arch/x86/kernel/apic_32.c
> +++ linux-2.6.x86.sched/arch/x86/kernel/apic_32.c
> @@ -52,9 +52,6 @@
>  
>  unsigned long mp_lapic_addr;
>  
> -DEFINE_PER_CPU(u16, x86_bios_cpu_apicid) = BAD_APICID;
> -EXPORT_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
> -
>  /*
>   * Knob to control our willingness to enable the local APIC.
>   *
> @@ -1533,9 +1530,9 @@ void __cpuinit generic_processor_info(in
>  	}
>  #ifdef CONFIG_SMP
>  	/* are we being called early in kernel startup? */
> -	if (x86_cpu_to_apicid_early_ptr) {
> -		u16 *cpu_to_apicid = x86_cpu_to_apicid_early_ptr;
> -		u16 *bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
> +	if (early_per_cpu_ptr(x86_cpu_to_apicid)) {
> +		u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid);
> +		u16 *bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid);
>  
>  		cpu_to_apicid[cpu] = apicid;
>  		bios_cpu_apicid[cpu] = apicid;
> --- linux-2.6.x86.sched.orig/arch/x86/kernel/apic_64.c
> +++ linux-2.6.x86.sched/arch/x86/kernel/apic_64.c
> @@ -87,9 +87,6 @@ static unsigned long apic_phys;
>  
>  unsigned long mp_lapic_addr;
>  
> -DEFINE_PER_CPU(u16, x86_bios_cpu_apicid) = BAD_APICID;
> -EXPORT_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
> -
>  unsigned int __cpuinitdata maxcpus = NR_CPUS;
>  /*
>   * Get the LAPIC version
> @@ -1090,9 +1087,9 @@ void __cpuinit generic_processor_info(in
>  		cpu = 0;
>  	}
>  	/* are we being called early in kernel startup? */
> -	if (x86_cpu_to_apicid_early_ptr) {
> -		u16 *cpu_to_apicid = x86_cpu_to_apicid_early_ptr;
> -		u16 *bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
> +	if (early_per_cpu_ptr(x86_cpu_to_apicid)) {
> +		u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid);
> +		u16 *bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid);
>  
>  		cpu_to_apicid[cpu] = apicid;
>  		bios_cpu_apicid[cpu] = apicid;
> @@ -1268,7 +1265,7 @@ __cpuinit int apic_is_clustered_box(void
>  	if ((boot_cpu_data.x86_vendor == X86_VENDOR_AMD) && !is_vsmp_box())
>  		return 0;
>  
> -	bios_cpu_apicid = x86_bios_cpu_apicid_early_ptr;
> +	bios_cpu_apicid = early_per_cpu_ptr(x86_bios_cpu_apicid);
>  	bitmap_zero(clustermap, NUM_APIC_CLUSTERS);
>  
>  	for (i = 0; i < NR_CPUS; i++) {
> --- linux-2.6.x86.sched.orig/arch/x86/kernel/setup.c
> +++ linux-2.6.x86.sched/arch/x86/kernel/setup.c
> @@ -18,13 +18,21 @@ unsigned disabled_cpus __cpuinitdata;
>  unsigned int boot_cpu_physical_apicid = -1U;
>  EXPORT_SYMBOL(boot_cpu_physical_apicid);
>  
> +/* Bitmask of physically existing CPUs */
>  physid_mask_t phys_cpu_present_map;
>  
> -DEFINE_PER_CPU(u16, x86_cpu_to_apicid) = BAD_APICID;
> -EXPORT_PER_CPU_SYMBOL(x86_cpu_to_apicid);
> +/* map cpu index to physical APIC ID */
> +DEFINE_EARLY_PER_CPU(u16, x86_cpu_to_apicid, BAD_APICID);
> +DEFINE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid, BAD_APICID);
>  
> -/* Bitmask of physically existing CPUs */
> -physid_mask_t phys_cpu_present_map;
> +EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_apicid);
> +EXPORT_EARLY_PER_CPU_SYMBOL(x86_bios_cpu_apicid);
> +
> +/* map cpu index to node index */
> +#ifdef CONFIG_NUMA
> +DEFINE_EARLY_PER_CPU(int, x86_cpu_to_node_map, NUMA_NO_NODE);
> +EXPORT_EARLY_PER_CPU_SYMBOL(x86_cpu_to_node_map);
> +#endif
>  
>  #if defined(CONFIG_HAVE_SETUP_PER_CPU_AREA) && defined(CONFIG_SMP)
>  /*
> @@ -37,20 +45,21 @@ static void __init setup_per_cpu_maps(vo
>  	int cpu;
>  
>  	for_each_possible_cpu(cpu) {
> -		per_cpu(x86_cpu_to_apicid, cpu) = x86_cpu_to_apicid_init[cpu];
> +		per_cpu(x86_cpu_to_apicid, cpu) =
> +				early_per_cpu_map(x86_cpu_to_apicid, cpu);
>  		per_cpu(x86_bios_cpu_apicid, cpu) =
> -						x86_bios_cpu_apicid_init[cpu];
> +				early_per_cpu_map(x86_bios_cpu_apicid, cpu);
>  #ifdef CONFIG_NUMA
>  		per_cpu(x86_cpu_to_node_map, cpu) =
> -						x86_cpu_to_node_map_init[cpu];
> +				early_per_cpu_map(x86_cpu_to_node_map, cpu);
>  #endif
>  	}
>  
>  	/* indicate the early static arrays will soon be gone */
> -	x86_cpu_to_apicid_early_ptr = NULL;
> -	x86_bios_cpu_apicid_early_ptr = NULL;
> +	early_per_cpu_ptr(x86_cpu_to_apicid) = NULL;
> +	early_per_cpu_ptr(x86_bios_cpu_apicid) = NULL;
>  #ifdef CONFIG_NUMA
> -	x86_cpu_to_node_map_early_ptr = NULL;
> +	early_per_cpu_ptr(x86_cpu_to_node_map) = NULL;
>  #endif
>  }
>  
> @@ -137,3 +146,53 @@ void __init setup_per_cpu_areas(void)
>  }
>  
>  #endif
> +
> +#if defined(CONFIG_NUMA)
> +void __cpuinit numa_set_node(int cpu, int node)
> +{
> +	int *cpu_to_node_map = early_per_cpu_ptr(x86_cpu_to_node_map);
> +
> +	if (cpu_to_node_map)
> +		cpu_to_node_map[cpu] = node;
> +
> +	else if (per_cpu_offset(cpu))
> +		per_cpu(x86_cpu_to_node_map, cpu) = node;
> +
> +	else
> +		Dprintk(KERN_INFO "Setting node for non-present cpu %d\n", cpu);
> +}
> +
> +void __cpuinit numa_add_cpu(int cpu)
> +{
> +	cpu_set(cpu, node_to_cpumask_map[early_cpu_to_node(cpu)]);
> +}
> +
> +#endif
> +
> +#ifdef CONFIG_DEBUG_PER_CPU_MAPS
> +int cpu_to_node(int cpu)
> +{
> +	if (early_per_cpu_ptr(x86_cpu_to_node_map)) {
> +		printk(KERN_WARNING
> +			"cpu_to_node(%d): usage too early!\n", cpu);
> +		dump_stack();
> +		return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
> +	}
> +	return per_cpu(x86_cpu_to_node_map, cpu);
> +}
> +EXPORT_SYMBOL(cpu_to_node);
> +
> +int early_cpu_to_node(int cpu)
> +{
> +	if (early_per_cpu_ptr(x86_cpu_to_node_map))
> +		return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
> +
> +	if (!per_cpu_offset(cpu)) {
> +		printk(KERN_WARNING
> +			"early_cpu_to_node(%d): no per_cpu area!\n", cpu);
> +			dump_stack();
> +		return NUMA_NO_NODE;
> +	}
> +	return per_cpu(x86_cpu_to_node_map, cpu);
> +}
> +#endif
> --- linux-2.6.x86.sched.orig/arch/x86/kernel/setup_32.c
> +++ linux-2.6.x86.sched/arch/x86/kernel/setup_32.c
> @@ -724,18 +724,6 @@ char * __init __attribute__((weak)) memo
>  	return machine_specific_memory_setup();
>  }
>  
> -#ifdef CONFIG_NUMA
> -/*
> - * In the golden day, when everything among i386 and x86_64 will be
> - * integrated, this will not live here
> - */
> -void *x86_cpu_to_node_map_early_ptr;
> -int x86_cpu_to_node_map_init[NR_CPUS] = {
> -	[0 ... NR_CPUS-1] = NUMA_NO_NODE
> -};
> -DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
> -#endif
> -
>  /*
>   * Determine if we were loaded by an EFI loader.  If so, then we have also been
>   * passed the efi memmap, systab, etc., so we should use these data structures
> @@ -869,18 +857,6 @@ void __init setup_arch(char **cmdline_p)
>  
>  	io_delay_init();
>  
> -#ifdef CONFIG_X86_SMP
> -	/*
> -	 * setup to use the early static init tables during kernel startup
> -	 * X86_SMP will exclude sub-arches that don't deal well with it.
> -	 */
> -	x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
> -	x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
> -#ifdef CONFIG_NUMA
> -	x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
> -#endif
> -#endif
> -
>  #ifdef CONFIG_X86_GENERICARCH
>  	generic_apic_probe();
>  #endif
> --- linux-2.6.x86.sched.orig/arch/x86/kernel/setup_64.c
> +++ linux-2.6.x86.sched/arch/x86/kernel/setup_64.c
> @@ -385,15 +385,6 @@ void __init setup_arch(char **cmdline_p)
>  
>  	io_delay_init();
>  
> -#ifdef CONFIG_SMP
> -	/* setup to use the early static init tables during kernel startup */
> -	x86_cpu_to_apicid_early_ptr = (void *)x86_cpu_to_apicid_init;
> -	x86_bios_cpu_apicid_early_ptr = (void *)x86_bios_cpu_apicid_init;
> -#ifdef CONFIG_NUMA
> -	x86_cpu_to_node_map_early_ptr = (void *)x86_cpu_to_node_map_init;
> -#endif
> -#endif
> -
>  #ifdef CONFIG_ACPI
>  	/*
>  	 * Initialize the ACPI boot-time table parser (gets the RSDP and SDT).
> --- linux-2.6.x86.sched.orig/arch/x86/kernel/smpboot.c
> +++ linux-2.6.x86.sched/arch/x86/kernel/smpboot.c
> @@ -67,22 +67,6 @@
>  #include <mach_wakecpu.h>
>  #include <smpboot_hooks.h>
>  
> -/*
> - * FIXME: For x86_64, those are defined in other files. But moving them here,
> - * would make the setup areas dependent on smp, which is a loss. When we
> - * integrate apic between arches, we can probably do a better job, but
> - * right now, they'll stay here -- glommer
> - */
> -
> -/* which logical CPU number maps to which CPU (physical APIC ID) */
> -u16 x86_cpu_to_apicid_init[NR_CPUS] __initdata =
> -			{ [0 ... NR_CPUS-1] = BAD_APICID };
> -void *x86_cpu_to_apicid_early_ptr;
> -
> -u16 x86_bios_cpu_apicid_init[NR_CPUS] __initdata
> -				= { [0 ... NR_CPUS-1] = BAD_APICID };
> -void *x86_bios_cpu_apicid_early_ptr;
> -
>  #ifdef CONFIG_X86_32
>  u8 apicid_2_node[MAX_APICID];
>  #endif
> @@ -147,32 +131,25 @@ static cpumask_t cpu_sibling_setup_map;
>  int __cpuinitdata smp_b_stepping;
>  
>  #if defined(CONFIG_NUMA) && defined(CONFIG_X86_32)
> -
>  /* which logical CPUs are on which nodes */
>  cpumask_t node_to_cpumask_map[MAX_NUMNODES] __read_mostly =
>  				{ [0 ... MAX_NUMNODES-1] = CPU_MASK_NONE };
>  EXPORT_SYMBOL(node_to_cpumask_map);
> -/* which node each logical CPU is on */
> -int cpu_to_node_map[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
> -EXPORT_SYMBOL(cpu_to_node_map);
>  
>  /* set up a mapping between cpu and node. */
>  static void map_cpu_to_node(int cpu, int node)
>  {
>  	printk(KERN_INFO "Mapping cpu %d to node %d\n", cpu, node);
> -	cpu_set(cpu, node_to_cpumask_map[node]);
> -	cpu_to_node_map[cpu] = node;
> +	numa_set_node(cpu, node);
> +	numa_add_cpu(cpu);
>  }
>  
>  /* undo a mapping between cpu and node. */
>  static void unmap_cpu_to_node(int cpu)
>  {
> -	int node;
> -
>  	printk(KERN_INFO "Unmapping cpu %d from all nodes\n", cpu);
> -	for (node = 0; node < MAX_NUMNODES; node++)
> -		cpu_clear(cpu, node_to_cpumask_map[node]);
> -	cpu_to_node_map[cpu] = 0;
> +	numa_remove_cpu(cpu);
> +	numa_clear_node(cpu);
>  }
>  #else /* !(CONFIG_NUMA && CONFIG_X86_32) */
>  #define map_cpu_to_node(cpu, node)	({})
> @@ -193,13 +170,13 @@ void map_cpu_to_logical_apicid(void)
>  		node = first_online_node;
>  
>  	cpu_2_logical_apicid[cpu] = apicid;
> -	map_cpu_to_node(cpu, node);
> +	numa_set_node(cpu, node);
>  }
>  
>  void unmap_cpu_to_logical_apicid(int cpu)
>  {
>  	cpu_2_logical_apicid[cpu] = BAD_APICID;
> -	unmap_cpu_to_node(cpu);
> +	numa_clear_node(cpu);
>  }
>  #else
>  #define unmap_cpu_to_logical_apicid(cpu) do {} while (0)
> @@ -973,9 +950,7 @@ do_rest:
>  	if (boot_error) {
>  		/* Try to put things back the way they were before ... */
>  		unmap_cpu_to_logical_apicid(cpu);
> -#ifdef CONFIG_X86_64
> -		clear_node_cpumask(cpu); /* was set by numa_add_cpu */
> -#endif
> +		numa_remove_cpu(cpu); /* was set by numa_add_cpu */
>  		cpu_clear(cpu, cpu_callout_map); /* was set by do_boot_cpu() */
>  		cpu_clear(cpu, cpu_initialized); /* was set by cpu_init() */
>  		cpu_clear(cpu, cpu_possible_map);
> @@ -1354,7 +1329,7 @@ static void __ref remove_cpu_from_maps(i
>  	cpu_clear(cpu, cpu_callin_map);
>  	/* was set by cpu_init() */
>  	clear_bit(cpu, (unsigned long *)&cpu_initialized);
> -	clear_node_cpumask(cpu);
> +	numa_remove_cpu(cpu);
>  #endif
>  }
>  
> --- linux-2.6.x86.sched.orig/arch/x86/mm/numa_64.c
> +++ linux-2.6.x86.sched/arch/x86/mm/numa_64.c
> @@ -31,16 +31,6 @@ bootmem_data_t plat_node_bdata[MAX_NUMNO
>  
>  struct memnode memnode;
>  
> -#ifdef CONFIG_SMP
> -int x86_cpu_to_node_map_init[NR_CPUS] = {
> -	[0 ... NR_CPUS-1] = NUMA_NO_NODE
> -};
> -void *x86_cpu_to_node_map_early_ptr;
> -EXPORT_SYMBOL(x86_cpu_to_node_map_early_ptr);
> -#endif
> -DEFINE_PER_CPU(int, x86_cpu_to_node_map) = NUMA_NO_NODE;
> -EXPORT_PER_CPU_SYMBOL(x86_cpu_to_node_map);
> -
>  s16 apicid_to_node[MAX_LOCAL_APIC] __cpuinitdata = {
>  	[0 ... MAX_LOCAL_APIC-1] = NUMA_NO_NODE
>  };
> @@ -578,24 +568,6 @@ void __init numa_initmem_init(unsigned l
>  	setup_node_bootmem(0, start_pfn << PAGE_SHIFT, end_pfn << PAGE_SHIFT);
>  }
>  
> -__cpuinit void numa_add_cpu(int cpu)
> -{
> -	set_bit(cpu,
> -		(unsigned long *)&node_to_cpumask_map[early_cpu_to_node(cpu)]);
> -}
> -
> -void __cpuinit numa_set_node(int cpu, int node)
> -{
> -	int *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
> -
> -	if(cpu_to_node_map)
> -		cpu_to_node_map[cpu] = node;
> -	else if(per_cpu_offset(cpu))
> -		per_cpu(x86_cpu_to_node_map, cpu) = node;
> -	else
> -		Dprintk(KERN_INFO "Setting node for non-present cpu %d\n", cpu);
> -}
> -
>  unsigned long __init numa_free_all_bootmem(void)
>  {
>  	unsigned long pages = 0;
> @@ -642,6 +614,7 @@ static __init int numa_setup(char *opt)
>  }
>  early_param("numa", numa_setup);
>  
> +#ifdef CONFIG_NUMA
>  /*
>   * Setup early cpu_to_node.
>   *
> @@ -653,14 +626,19 @@ early_param("numa", numa_setup);
>   * is already initialized in a round robin manner at numa_init_array,
>   * prior to this call, and this initialization is good enough
>   * for the fake NUMA cases.
> + *
> + * Called before the per_cpu areas are setup.
>   */
>  void __init init_cpu_to_node(void)
>  {
> -	int i;
> +	int cpu;
> +	u16 *cpu_to_apicid = early_per_cpu_ptr(x86_cpu_to_apicid);
>  
> -	for (i = 0; i < NR_CPUS; i++) {
> +	BUG_ON(cpu_to_apicid == NULL);
> +
> +	for_each_possible_cpu(cpu) {
>  		int node;
> -		u16 apicid = x86_cpu_to_apicid_init[i];
> +		u16 apicid = cpu_to_apicid[cpu];
>  
>  		if (apicid == BAD_APICID)
>  			continue;
> @@ -669,8 +647,9 @@ void __init init_cpu_to_node(void)
>  			continue;
>  		if (!node_online(node))
>  			continue;
> -		numa_set_node(i, node);
> +		numa_set_node(cpu, node);
>  	}
>  }
> +#endif
>  
>  
> --- linux-2.6.x86.sched.orig/include/asm-x86/numa.h
> +++ linux-2.6.x86.sched/include/asm-x86/numa.h
> @@ -1,5 +1,33 @@
> +#ifndef _ASM_X86_NUMA_H
> +#define _ASM_X86_NUMA_H 1
> +
>  #ifdef CONFIG_X86_32
>  # include "numa_32.h"
>  #else
>  # include "numa_64.h"
>  #endif
> +
> +#ifndef CONFIG_NUMA
> +static void inline init_cpu_to_node(void)		{ }
> +static void inline numa_set_node(int cpu, int node)	{ }
> +static void inline numa_clear_node(int cpu)		{ }
> +static void inline numa_add_cpu(int cpu, int node)	{ }
> +static void inline numa_remove_cpu(int cpu)		{ }
> +
> +#else
> +extern void __init init_cpu_to_node(void);
> +extern void __cpuinit numa_set_node(int cpu, int node);
> +extern void numa_add_cpu(int cpu);
> +
> +static inline void numa_clear_node(int cpu)
> +{
> +	numa_set_node(cpu, NUMA_NO_NODE);
> +}
> +
> +static inline void numa_remove_cpu(int cpu)
> +{
> +	cpu_clear(cpu, node_to_cpumask_map[cpu_to_node(cpu)]);
> +}
> +#endif
> +
> +#endif /* _ASM_X86_NUMA_H */
> --- linux-2.6.x86.sched.orig/include/asm-x86/numa_64.h
> +++ linux-2.6.x86.sched/include/asm-x86/numa_64.h
> @@ -14,11 +14,9 @@ extern int compute_hash_shift(struct boo
>  
>  #define ZONE_ALIGN (1UL << (MAX_ORDER+PAGE_SHIFT))
>  
> -extern void numa_add_cpu(int cpu);
>  extern void numa_init_array(void);
>  extern int numa_off;
>  
> -extern void numa_set_node(int cpu, int node);
>  extern void srat_reserve_add_area(int nodeid);
>  extern int hotadd_percent;
>  
> @@ -29,17 +27,4 @@ extern unsigned long numa_free_all_bootm
>  extern void setup_node_bootmem(int nodeid, unsigned long start,
>  			       unsigned long end);
>  
> -#ifdef CONFIG_NUMA
> -extern void __init init_cpu_to_node(void);
> -
> -static inline void clear_node_cpumask(int cpu)
> -{
> -	clear_bit(cpu, (unsigned long *)&node_to_cpumask_map[cpu_to_node(cpu)]);
> -}
> -
> -#else
> -#define init_cpu_to_node() do {} while (0)
> -#define clear_node_cpumask(cpu) do {} while (0)
> -#endif
> -
>  #endif
> --- linux-2.6.x86.sched.orig/include/asm-x86/percpu.h
> +++ linux-2.6.x86.sched/include/asm-x86/percpu.h
> @@ -143,4 +143,50 @@ do {							\
>  #define x86_or_percpu(var, val) percpu_to_op("or", per_cpu__##var, val)
>  #endif /* !__ASSEMBLY__ */
>  #endif /* !CONFIG_X86_64 */
> +
> +#ifdef CONFIG_SMP
> +
> +/*
> + * Define the "EARLY_PER_CPU" macros.  These are used for some per_cpu
> + * variables that are initialized and accessed before there are per_cpu
> + * areas allocated.
> + */
> +
> +#define	DEFINE_EARLY_PER_CPU(_type, _name, _initvalue)			\
> +	DEFINE_PER_CPU(_type, _name) = _initvalue;			\
> +	__typeof__(_type) _name##_early_map[NR_CPUS] __initdata =	\
> +				{ [0 ... NR_CPUS-1] = _initvalue };	\
> +	__typeof__(_type) *_name##_early_ptr = _name##_early_map
> +
> +#define EXPORT_EARLY_PER_CPU_SYMBOL(_name)			\
> +	EXPORT_PER_CPU_SYMBOL(_name)
> +
> +#define DECLARE_EARLY_PER_CPU(_type, _name)			\
> +	DECLARE_PER_CPU(_type, _name);				\
> +	extern __typeof__(_type) *_name##_early_ptr;		\
> +	extern __typeof__(_type)  _name##_early_map[]
> +
> +#define	early_per_cpu_ptr(_name) (_name##_early_ptr)
> +#define	early_per_cpu_map(_name, _idx) (_name##_early_map[_idx])
> +#define	early_per_cpu(_name, _cpu) 				\
> +	(early_per_cpu_ptr(_name) ?				\
> +		early_per_cpu_ptr(_name)[_cpu] :		\
> +		per_cpu(_name, _cpu))
> +
> +#else	/* !CONFIG_SMP */
> +#define	DEFINE_EARLY_PER_CPU(_type, _name, _initvalue)		\
> +	DEFINE_PER_CPU(_type, _name) = _initvalue
> +
> +#define EXPORT_EARLY_PER_CPU_SYMBOL(_name)			\
> +	EXPORT_PER_CPU_SYMBOL(_name)
> +
> +#define DECLARE_EARLY_PER_CPU(_type, _name)			\
> +	DECLARE_PER_CPU(_type, _name)
> +
> +#define	early_per_cpu(_name, _cpu) per_cpu(_name, _cpu)
> +#define	early_per_cpu_ptr(_name) NULL
> +/* no early_per_cpu_map() */
> +
> +#endif	/* !CONFIG_SMP */
> +
>  #endif /* _ASM_X86_PERCPU_H_ */
> --- linux-2.6.x86.sched.orig/include/asm-x86/smp.h
> +++ linux-2.6.x86.sched/include/asm-x86/smp.h
> @@ -29,21 +29,12 @@ extern int smp_num_siblings;
>  extern unsigned int num_processors;
>  extern cpumask_t cpu_initialized;
>  
> -#ifdef CONFIG_SMP
> -extern u16 x86_cpu_to_apicid_init[];
> -extern u16 x86_bios_cpu_apicid_init[];
> -extern void *x86_cpu_to_apicid_early_ptr;
> -extern void *x86_bios_cpu_apicid_early_ptr;
> -#else
> -#define x86_cpu_to_apicid_early_ptr NULL
> -#define x86_bios_cpu_apicid_early_ptr NULL
> -#endif
> -
>  DECLARE_PER_CPU(cpumask_t, cpu_sibling_map);
>  DECLARE_PER_CPU(cpumask_t, cpu_core_map);
>  DECLARE_PER_CPU(u16, cpu_llc_id);
> -DECLARE_PER_CPU(u16, x86_cpu_to_apicid);
> -DECLARE_PER_CPU(u16, x86_bios_cpu_apicid);
> +
> +DECLARE_EARLY_PER_CPU(u16, x86_cpu_to_apicid);
> +DECLARE_EARLY_PER_CPU(u16, x86_bios_cpu_apicid);
>  
>  /* Static state in head.S used to set up a CPU */
>  extern struct {
> --- linux-2.6.x86.sched.orig/include/asm-x86/topology.h
> +++ linux-2.6.x86.sched/include/asm-x86/topology.h
> @@ -31,82 +31,64 @@ struct pci_bus;
>  #include <asm/mpspec.h>
>  
>  /* Mappings between logical cpu number and node number */
> -#ifdef CONFIG_X86_32
> -extern int cpu_to_node_map[];
> -#else
> -/* Returns the number of the current Node. */
> -#define numa_node_id()		(early_cpu_to_node(raw_smp_processor_id()))
> -#endif
> +DECLARE_EARLY_PER_CPU(int, x86_cpu_to_node_map);
>  
> -DECLARE_PER_CPU(int, x86_cpu_to_node_map);
> +#define NUMA_NO_NODE	(-1)
>  
> -#ifdef CONFIG_SMP
> -extern int x86_cpu_to_node_map_init[];
> -extern void *x86_cpu_to_node_map_early_ptr;
> -#else
> -#define x86_cpu_to_node_map_early_ptr NULL
> -#endif
> +#ifndef CONFIG_NUMA
> +#define numa_node_id()		0
> +#define	cpu_to_node(cpu)	0
> +#define	early_cpu_to_node(cpu)	0
> +#define node_to_cpumask_ptr(v, node)		\
> +		cpumask_t *v = &cpu_online_map
> +#define node_to_cpumask_ptr_next(v, node)	\
> +			   v = &cpu_online_map
> +static inline cpumask_t node_to_cpumask(int node)
> +{
> +	return cpu_online_map;
> +}
> +static inline int node_to_first_cpu(int node)
> +{
> +	return first_cpu(cpu_online_map);
> +}
> +
> +#else /* CONFIG_NUMA */
>  
>  extern cpumask_t node_to_cpumask_map[];
>  
> -#define NUMA_NO_NODE	(-1)
> +/* Returns the number of the current Node. */
> +#define numa_node_id()		(__get_cpu_var(x86_cpu_to_node_map))
> +
> +#ifdef CONFIG_DEBUG_PER_CPU_MAPS
> +extern int cpu_to_node(int cpu);
> +extern int early_cpu_to_node(int cpu);
> +#else
>  
>  /* Returns the number of the node containing CPU 'cpu' */
> -#ifdef CONFIG_X86_32
> -#define early_cpu_to_node(cpu)	cpu_to_node(cpu)
>  static inline int cpu_to_node(int cpu)
>  {
> -	return cpu_to_node_map[cpu];
> +	return per_cpu(x86_cpu_to_node_map, cpu);
>  }
>  
> -#else /* CONFIG_X86_64 */
> -
> -#ifdef CONFIG_SMP
> +/* Same function but used if called before per_cpu areas are setup */
>  static inline int early_cpu_to_node(int cpu)
>  {
> -	int *cpu_to_node_map = x86_cpu_to_node_map_early_ptr;
> -
> -	if (cpu_to_node_map)
> -		return cpu_to_node_map[cpu];
> -	else if (per_cpu_offset(cpu))
> -		return per_cpu(x86_cpu_to_node_map, cpu);
> -	else
> -		return NUMA_NO_NODE;
> -}
> -#else
> -#define	early_cpu_to_node(cpu)	cpu_to_node(cpu)
> -#endif
> +	if (early_per_cpu_ptr(x86_cpu_to_node_map))
> +		return early_per_cpu_ptr(x86_cpu_to_node_map)[cpu];
>  
> -static inline int cpu_to_node(int cpu)
> -{
> -#ifdef CONFIG_DEBUG_PER_CPU_MAPS
> -	if (x86_cpu_to_node_map_early_ptr) {
> -		printk("KERN_NOTICE cpu_to_node(%d): usage too early!\n",
> -		       (int)cpu);
> -		dump_stack();
> -		return ((int *)x86_cpu_to_node_map_early_ptr)[cpu];
> -	}
> -#endif
>  	return per_cpu(x86_cpu_to_node_map, cpu);
>  }
> -
> -#ifdef	CONFIG_NUMA
> +#endif /* CONFIG_DEBUG_PER_CPU_MAPS */
>  
>  /* Returns a pointer to the cpumask of CPUs on Node 'node'. */
>  #define node_to_cpumask_ptr(v, node)		\
> -		cpumask_t *v = &(node_to_cpumask_map[node])
> -
> +		cpumask_t *v = _node_to_cpumask_ptr(node)
>  #define node_to_cpumask_ptr_next(v, node)	\
> -			   v = &(node_to_cpumask_map[node])
> -#endif
> -
> -#endif /* CONFIG_X86_64 */
> -
> -/*
> - * Returns the number of the node containing Node 'node'. This
> - * architecture is flat, so it is a pretty simple function!
> - */
> -#define parent_node(node) (node)
> +			   v = _node_to_cpumask_ptr(node)
> +static inline cpumask_t *_node_to_cpumask_ptr(int node)
> +{
> +	return &node_to_cpumask_map[node];
> +}
>  
>  /* Returns a bitmask of CPUs on Node 'node'. */
>  static inline cpumask_t node_to_cpumask(int node)
> @@ -117,11 +99,19 @@ static inline cpumask_t node_to_cpumask(
>  /* Returns the number of the first CPU on Node 'node'. */
>  static inline int node_to_first_cpu(int node)
>  {
> -	cpumask_t mask = node_to_cpumask(node);
> -
> -	return first_cpu(mask);
> +	node_to_cpumask_ptr(mask, node);
> +	return first_cpu(*mask);
>  }
>  
> +#endif /* CONFIG_NUMA */
> +
> +/*
> + * Returns the number of the node containing Node 'node'. This
> + * architecture is flat, so it is a pretty simple function!
> + */
> +#define parent_node(node) (node)
> +
> +
>  #define pcibus_to_node(bus) __pcibus_to_node(bus)
>  #define pcibus_to_cpumask(bus) __pcibus_to_cpumask(bus)
>  
> 

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

Powered by blists - more mailing lists