Updated CPU Hotplug patches for IA64 [Patch 1/7]
Subsystems (3)
Hi Andrew
Here are the updated patches for CPU hotplug support for IA64. I reworked some of the patches, and merged/split them based on feedback from David Mosberger-Tang, and couple others from the list.
Summary of changes.
- Removed kdb patch that got in by mistake :-(
- split migrate_irq.patch to separate the bug fix v.s. feature required for cpu hotplug support. (Thanks to Kanashige-san (fujitsu) for giving me that extra work :-)).
- Removed one per-cpu global that was also initialized to 0, which was not necessary.
- Removed some hunks that were white space in hotcpu_ia64.patch
- Sorted the newly added headers as per David Moshberger's requirements.
- Merged the rest of migrate_irq.patch with hotcpu_ia64.patch as it was integral part of the cpu hotplug patches.
- arch/ia64/dig/topology.c is added only for non-numa, numa adds it in arch/ia64/mm/numa.c (Probably better to move to a common file in future)
Mostly the same, except some minor restructuring and cleanup...
Oh.. this time, i have learned to keep the lines together...:-) and not introducing line breaks... (Hopefully)
Cheers,
Ashok Raj
- Linux Core Software Group
Name: core_kernel_init.patch
Author: Ashok Raj (Intel Corporation)
D: This patch changes __init to __devinit to init_idle so that when a new cpu
D: arrives, it can call these functions at a later time.
---
linux-2.6.5-lhcs-root/init/main.c | 2 +- linux-2.6.5-lhcs-root/kernel/sched.c | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
kernel/sched.c renamed from kernel/sched.c~core_kernel_init +1 -1
@@ -2657,7 +2657,7 @@ void show_state(void)read_unlock (&tasklist_lock); } -void __init init_idle (task_t *idle,int cpu) +void __devinit init_idle (task_t *idle,int cpu) {runqueue_t *idle_rq =cpu_rq (cpu), *rq =cpu_rq (task_cpu (idle));unsigned long flags;
init/main.c renamed from init/main.c~core_kernel_init +1 -1
@@ -181,7 +181,7 @@ EXPORT_SYMBOL(loops_per_jiffy);better than 1% */ #define LPS_PREC 8 -void __init calibrate_delay (void ) +void __devinit calibrate_delay (void ) {unsigned long ticks, loopbit;int lps_precision =LPS_PREC ;
_
Updated CPU hotplug patches for IA64 [Patch 2/7]
Name: ia64_init_removal.patch
Author: Ashok Raj (Intel Corporation)
Status: Experimental
D: Contains changes from __init to __devinit to support cpu hotplug
D: Changes only arch/ia64 portions of the kernel tree.
---
linux-2.6.5-lhcs-root/arch/ia64/kernel/setup.c | 2 +- linux-2.6.5-lhcs-root/arch/ia64/kernel/smpboot.c | 12 ++++++------ linux-2.6.5-lhcs-root/arch/ia64/kernel/time.c | 2 +- linux-2.6.5-lhcs-root/arch/ia64/mm/init.c | 4 ++-- linux-2.6.5-lhcs-root/arch/ia64/mm/tlb.c | 2 +- linux-2.6.5-lhcs-root/include/asm-ia64/smp.h | 3 ++- 6 files changed, 13 insertions(+), 12 deletions(-)
arch/ia64/kernel/setup.c renamed from arch/ia64/kernel/setup.c~init_removal_ia64 +1 -1
@@ -575,7 +575,7 @@ get_max_cacheline_size (void)void cpu_init (void ) { -extern void __init ia64_mmu_init (void *); +extern void __devinit ia64_mmu_init (void *);unsigned long num_phys_stacked;pal_vm_info_2_u_t vmi;unsigned int max_ctx;
arch/ia64/kernel/smpboot.c renamed from arch/ia64/kernel/smpboot.c~init_removal_ia64 +6 -6
@@ -69,7 +69,7 @@ static volatile unsigned long go[SLAVE +#define DEBUG_ITC_SYNC 0 -extern void __init calibrate_delay (void ); +extern void __devinit calibrate_delay (void );extern void start_ap (void );extern unsigned long ia64_iobase;
@@ -262,12 +262,12 @@ ia64_sync_itc (unsigned int master)/* * Ideally sets up per-cpu profiling hooks. Doesn't do much now... */ -static inline void __init +static inline void __devinit smp_setup_percpu_timer (void ) { } -static void __init +static void __devinit smp_callin (void ) {int cpuid, phys_id;
@@ -333,7 +333,7 @@ smp_callin (void)/* * Activate a secondary processor. head.S calls this. */ -int __init +int __devinit start_secondary (void *unused) {extern int cpu_idle (void );
@@ -346,7 +346,7 @@ start_secondary (void *unused)return cpu_idle (); } -static struct task_struct *__init +static struct task_struct *__devinit fork_by_hand (void ) {/*
@@ -356,7 +356,7 @@ fork_by_hand (void)return copy_process (CLONE_VM |CLONE_IDLETASK ,0 ,0 ,0 ,NULL ,NULL ); } -static int __init +static int __devinit do_boot_cpu (int sapicid,int cpu) {struct task_struct *idle;
arch/ia64/kernel/time.c renamed from arch/ia64/kernel/time.c~init_removal_ia64 +1 -1
@@ -326,7 +326,7 @@ ia64_cpu_local_tick (void)ia64_set_itm (local_cpu_data->itm_next); } -void __init +void __devinit ia64_init_itm (void ) {unsigned long platform_base_freq, itc_freq;
arch/ia64/mm/init.c renamed from arch/ia64/mm/init.c~init_removal_ia64 +2 -2
@@ -274,11 +274,11 @@ setup_gate (void)ia64_patch_gate (); } -void __init +void __devinit ia64_mmu_init (void *my_cpu_data) {unsigned long psr, pta, impl_va_bits; -extern void __init tlb_init (void ); +extern void __devinit tlb_init (void );int cpu;#ifdef CONFIG_DISABLE_VHPT
arch/ia64/mm/tlb.c renamed from arch/ia64/mm/tlb.c~init_removal_ia64 +1 -1
@@ -166,7 +166,7 @@ flush_tlb_range (struct vm_area_struct * }EXPORT_SYMBOL (flush_tlb_range); -void __init +void __devinit ia64_tlb_init (void ) {ia64_ptce_info_t ptce_info;
include/asm-ia64/smp.h renamed from include/asm-ia64/smp.h~init_removal_ia64 +2 -1
@@ -36,7 +36,7 @@ extern struct smp_boot_data {int cpu_phys_id[NR_CPUS ]; } smp_boot_data__initdata ; -extern char no_int_routing__initdata ; +extern char no_int_routing__devinitdata ;extern cpumask_t phys_cpu_present_map;extern cpumask_t cpu_online_map;
@@ -113,6 +113,7 @@ hard_smp_processor_id (void)/* Upping and downing of CPUs */ extern int __cpu_disable (void );extern void __cpu_die (unsigned int cpu); +extern void cpu_die (void ) __attribute__ ((noreturn));extern int __cpu_up (unsigned int cpu);extern void __init smp_build_cpu_map (void );
_
Updated CPU hotplug patches for IA64 [Patch 3/7]
In reply to Ashok RajName: sysfs_ia64.patch
Author: Ashok Raj (Intel Corporation)
D: Creation of sysfs via topology_init() creates sysfs entries. The creation of
D: the online control file is created separately when the cpu_up is invoked
D: in arch independent code.
---
---
linux-2.6.5-lhcs-root/arch/ia64/dig/Makefile | 5 ++ linux-2.6.5-lhcs-root/arch/ia64/dig/topology.c | 43 +++++++++++++++++++++++++ linux-2.6.5-lhcs-root/include/asm-ia64/cpu.h | 17 +++++++++ 3 files changed, 65 insertions(+)
arch/ia64/dig/Makefile renamed from arch/ia64/dig/Makefile~sysfs_ia64 +5 -0
@@ -6,4 +6,9 @@ # obj-y := setup.o + +ifndef CONFIG_NUMA +obj-$(CONFIG_IA64_DIG) += topology.o +endif + obj-$(CONFIG_IA64_GENERIC) += machvec.o
arch/ia64/dig/topology.c new file +43 -0
43 lines not on this page · raw message
include/asm-ia64/cpu.h new file +17 -0
@@ -0,0 +1,17 @@ +#ifndef _ASM_IA64_CPU_H_ +#define _ASM_IA64_CPU_H_ + +#include <linux/device.h> +#include <linux/cpu.h> +#include <linux/topology.h> +#include <linux/percpu.h> + +struct ia64_cpu { +struct cpu cpu; +}; + +DECLARE_PER_CPU(struct ia64_cpu, cpu_devices); + +DECLARE_PER_CPU (int, cpu_state); + +#endif /* _ASM_IA64_CPU_H_ */
_
Updated CPU hotplug patches for IA64 [Patch 4/7]
In reply to Ashok RajName: ia64_irq_affinity_fix.patch
Author: Ashok Raj (Intel Corporation)
Status: Tested with Hotplug CPU Code
D: irq affinity setting via /proc was forcing iosapic rte programming
D: by force. The correct way to do this is to perform this when a interrupt
D: is pending.
---
linux-2.6.5-lhcs-root/arch/ia64/kernel/iosapic.c | 19 +++++++++++++++++++ linux-2.6.5-lhcs-root/arch/ia64/kernel/irq.c | 12 +++++++++--- 2 files changed, 28 insertions(+), 3 deletions(-)
arch/ia64/kernel/irq.c renamed from arch/ia64/kernel/irq.c~ia64_irq_affinity_fix +9 -3
@@ -46,7 +46,7 @@#include <asm/delay.h> #include <asm/irq.h> - +extern cpumask_t __cacheline_aligned pending_irq_cpumask[NR_IRQS];/* * Linux has a controller-independent x86 interrupt architecture.
@@ -938,7 +938,9 @@ void set_irq_affinity_info (unsigned intstatic int irq_affinity_read_proc (char *page,char **start,off_t off,int count,int *eof,void *data) { -int len =cpumask_scnprintf (page, count, irq_affinity[(long )data]); +int len =sprintf (page,"%s" , irq_redir[(long )data] ?"r " :"" ); + + len +=cpumask_scnprintf (page+len, count, irq_affinity[(long )data]);if (count - len <2 )return -EINVAL ; len +=sprintf (page + len,"\n" );
@@ -956,6 +958,7 @@ static int irq_affinity_write_proc (struint rlen;int prelen;irq_desc_t *desc =irq_descp (irq); +unsigned long flags;if (!desc->handler->set_affinity)return -EIO ;
@@ -994,7 +997,10 @@ static int irq_affinity_write_proc (struif (cpus_empty (tmp))return -EINVAL ; - desc->handler->set_affinity (irq, new_value); +spin_lock_irqsave (&desc->lock, flags); + pending_irq_cpumask[irq] = new_value; +spin_unlock_irqrestore (&desc->lock, flags); +return full_count; }
arch/ia64/kernel/iosapic.c renamed from arch/ia64/kernel/iosapic.c~ia64_irq_affinity_fix +19 -0
@@ -98,6 +98,7 @@#endif static spinlock_t iosapic_lock =SPIN_LOCK_UNLOCKED ; +cpumask_t __cacheline_aligned pending_irq_cpumask[NR_IRQS];/* These tables map IA-64 vectors to the IOSAPIC pin that generates this vector. */
@@ -327,6 +328,21 @@ iosapic_set_affinity (unsigned int irq,#endif } +static inline void move_irq (int irq) +{ +/* note - we hold desc->lock */ +cpumask_t tmp; +irq_desc_t *desc =irq_descp (irq); + +if (!cpus_empty (pending_irq_cpumask[irq])) { +cpus_and (tmp, pending_irq_cpumask[irq], cpu_online_map); +if (unlikely (!cpus_empty (tmp))) { + desc->handler->set_affinity (irq, pending_irq_cpumask[irq]); + } +cpus_clear (pending_irq_cpumask[irq]); + } +} +/* * Handlers for level-triggered interrupts. */
@@ -343,6 +359,7 @@ iosapic_end_level_irq (unsigned int irq) {ia64_vector vec =irq_to_vector (irq); +move_irq (irq);writel (vec, iosapic_intr_info[vec].addr +IOSAPIC_EOI ); }
@@ -382,6 +399,8 @@ static void iosapic_ack_edge_irq (unsigned int irq) {irq_desc_t *idesc =irq_descp (irq); + +move_irq (irq);/* * Once we have recorded IRQ_PENDING already, we can mask the * interrupt for real. This prevents IRQ storms from unhandled
_
Updated CPU hotplug patches for IA64 [Patch 5/7]
In reply to Ashok RajName: hotcpu_ia64.patch
Author: Ashok Raj (Intel Corporation)
D: Supports basic ability to enable hotplug functions for IA64.
D: Code is just evolving, and there are several loose ends to tie up. D:
D: What this code drop does
D: - Support logical online and offline
D: - Handles interrupt migration without loss of interrupts.
D: - Handles stress fine > 24+ hrs with make -j/ftp/rcp workloads
D: - Handles irq migration from a dying cpu without loss of interrupts.
D: What needs to be done
D: - Boot CPU removal support, with platform level authentication
D: - Putting cpu being removed in BOOT_RENDEZ mode.
---
linux-2.6.5-lhcs-root/arch/ia64/Kconfig | 9 + linux-2.6.5-lhcs-root/arch/ia64/kernel/iosapic.c | 7 - linux-2.6.5-lhcs-root/arch/ia64/kernel/irq.c | 103 +++++++++++++++++ linux-2.6.5-lhcs-root/arch/ia64/kernel/irq_ia64.c | 60 +++++++++- linux-2.6.5-lhcs-root/arch/ia64/kernel/process.c | 43 +++++++ linux-2.6.5-lhcs-root/arch/ia64/kernel/sal.c | 13 ++ linux-2.6.5-lhcs-root/arch/ia64/kernel/smp.c | 26 ++++ linux-2.6.5-lhcs-root/arch/ia64/kernel/smpboot.c | 130 ++++++++++++++++++++-- linux-2.6.5-lhcs-root/arch/ia64/kernel/time.c | 5 linux-2.6.5-lhcs-root/include/asm-ia64/smp.h | 2 10 files changed, 381 insertions(+), 17 deletions(-)
arch/ia64/Kconfig renamed from arch/ia64/Kconfig~hotcpu_ia64 +9 -0
@@ -367,6 +367,15 @@ config PCI_DOMAINS source "drivers/pci/Kconfig" +config HOTPLUG_CPU + bool "Support for hot-pluggable CPUs (EXPERIMENTAL)" + depends on SMP && HOTPLUG && EXPERIMENTAL + default n + ---help--- + Say Y here to experiment with turning CPUs off and on. CPUs + can be controlled through /sys/devices/system/cpu/cpu# + Say N. + source "drivers/pci/hotplug/Kconfig" source "drivers/pcmcia/Kconfig"
arch/ia64/kernel/irq.c renamed from arch/ia64/kernel/irq.c~hotcpu_ia64 +103 -0
135 lines not on this page · raw message
arch/ia64/kernel/process.c renamed from arch/ia64/kernel/process.c~hotcpu_ia64 +42 -1
86 lines not on this page · raw message
arch/ia64/kernel/smpboot.c renamed from arch/ia64/kernel/smpboot.c~hotcpu_ia64 +122 -8
203 lines not on this page · raw message
arch/ia64/kernel/smp.c renamed from arch/ia64/kernel/smp.c~hotcpu_ia64 +25 -1
@@ -71,10 +71,23 @@ static volatile struct call_data_struct/* This needs to be cacheline aligned because it is written to by *other* CPUs. */ static DEFINE_PER_CPU(u64, ipi_operation) ____cacheline_aligned; +extern void cpu_halt (void ); + +void +lock_ipi_calllock (void ) +{ +spin_lock_irq (&call_lock); +} + +void +unlock_ipi_calllock (void ) +{ +spin_unlock_irq (&call_lock); +} +static void stop_this_cpu (void ) { -extern void cpu_halt (void );/* * Remove this CPU: */
@@ -84,6 +97,17 @@ stop_this_cpu (void)cpu_halt (); } +void +cpu_die (void ) +{ +max_xtp (); +local_irq_disable (); +cpu_halt (); +/* Should never be here */ +BUG (); +for (;;); +} +irqreturn_t handle_IPI (int irq,void *dev_id,struct pt_regs *regs) {
arch/ia64/kernel/time.c renamed from arch/ia64/kernel/time.c~hotcpu_ia64 +5 -0
@@ -10,6 +10,7 @@*/ #include <linux/config.h> +#include <linux/cpu.h> #include <linux/init.h> #include <linux/kernel.h> #include <linux/module.h>
@@ -244,6 +245,10 @@ timer_interrupt (int irq, void *dev_id, {unsigned long new_itm; +if (unlikely (cpu_is_offline (smp_processor_id ()))) { +return IRQ_HANDLED ; + } +platform_timer_interrupt (irq, dev_id, regs); new_itm = local_cpu_data->itm_next;
include/asm-ia64/smp.h renamed from include/asm-ia64/smp.h~hotcpu_ia64 +2 -0
@@ -123,6 +123,8 @@ extern void smp_do_timer (struct pt_regsextern int smp_call_function_single (int cpuid,void (*func) (void *info),void *info,int retry,int wait);extern void smp_send_reschedule (int cpu); +extern void lock_ipi_calllock (void ); +extern void unlock_ipi_calllock (void );#endif /* CONFIG_SMP */ #endif /* _ASM_IA64_SMP_H */
arch/ia64/kernel/iosapic.c renamed from arch/ia64/kernel/iosapic.c~hotcpu_ia64 +5 -2
@@ -32,6 +32,8 @@* 03/02/19 B. Helgaas Make pcat_compat system-wide, not per-IOSAPIC. * Remove iosapic_address & gsi_base from external interfaces. * Rationalize __init/__devinit attributes. +* 04/12/04 Ashok Raj <[email protected]> Intel Corporation 2004 +* Updated to work with irq migration necessary for CPU Hotplug */ /* * Here is what the interrupt logic between a PCI device and the kernel looks like:
@@ -189,8 +191,10 @@ set_rte (unsigned int vector, unsigned i pol = iosapic_intr_info[vector].polarity; trigger = iosapic_intr_info[vector].trigger; dmode = iosapic_intr_info[vector].dmode; + vector &= (~IA64_IRQ_REDIRECTED ); redir = (dmode =IOSAPIC_LOWEST_PRIORITY ) ?1 :0 ; +#ifdef CONFIG_SMP {unsigned int irq;
@@ -308,9 +312,8 @@ iosapic_set_affinity (unsigned int irq,spin_lock_irqsave (&iosapic_lock, flags); { -/* get current delivery mode by reading the low32 */ -writel (IOSAPIC_RTE_LOW (rte_index), addr +IOSAPIC_REG_SELECT ); low32 = iosapic_intr_info[vec].low32 & ~(7 <<IOSAPIC_DELIVERY_SHIFT ); +if (redir)/* change delivery mode to lowest priority */ low32 |= (IOSAPIC_LOWEST_PRIORITY <<IOSAPIC_DELIVERY_SHIFT );
arch/ia64/kernel/irq_ia64.c renamed from arch/ia64/kernel/irq_ia64.c~hotcpu_ia64 +55 -5
84 lines not on this page · raw message
arch/ia64/kernel/sal.c renamed from arch/ia64/kernel/sal.c~hotcpu_ia64 +13 -0
@@ -122,10 +122,23 @@ sal_desc_entry_point (void *p)static void __init set_smp_redirect (int flag) { +#ifndef CONFIG_HOTPLUG_CPU if (no_int_routing) smp_int_redirect &= ~flag;else smp_int_redirect |= flag; +#else +/* +* For CPU Hotplug we dont want to do any chipset supported +* interrupt redirection. The reason is this would require that +* All interrupts be stopped and hard bind the irq to a cpu. +* Later when the interrupt is fired we need to set the redir hint +* on again in the vector. This is combersome for something that the +* user mode irq balancer will solve anyways. +*/ + no_int_routing=1 ; + smp_int_redirect &= ~flag; +#endif }#else #define set_smp_redirect (flag) do { } while (0)
_
Updated CPU hotplug patches for IA64 [Patch 6/7]
In reply to Ashok RajName: ia64_palinfo.patch
Author: Ashok Raj (Intel Corporation)
Status: Tested with Hotplug CPU Code
D: Changes proc entries for cpu hotplug to be created via the cpu
D: hotplug notifier callbacks. Also fixed a bug in the removal code
D: that did not remove proc entries as expected.
---
linux-2.6.5-lhcs-root/arch/ia64/kernel/palinfo.c | 134 +++++++++++++++++------ 1 files changed, 104 insertions(+), 30 deletions(-)
arch/ia64/kernel/palinfo.c renamed from arch/ia64/kernel/palinfo.c~ia64_palinfo +103 -30
@@ -8,11 +8,14 @@* * Copyright (C) 2000-2001, 2003 Hewlett-Packard Co * Stephane Eranian <[email protected]> +* Copyright (C) 2004 Intel Corporation +* Ashok Raj <[email protected]> * * 05/26/2000 S.Eranian initial release * 08/21/2000 S.Eranian updated to July 2000 PAL specs * 02/05/2001 S.Eranian fixed module support * 10/23/2001 S.Eranian updated pal_perf_mon_info bug fixes +* 03/24/2004 Ashok Raj updated to work with CPU Hotplug */ #include <linux/config.h> #include <linux/types.h>
@@ -22,6 +25,9 @@#include <linux/mm.h> #include <linux/module.h> #include <linux/efi.h> +#include <linux/notifier.h> +#include <linux/cpu.h> +#include <linux/cpumask.h> #include <asm/pal.h> #include <asm/sal.h>
@@ -768,13 +774,12 @@ static palinfo_entry_t palinfo_entries[]* does not do recursion of deletion * * Notes: -* - first +1 accounts for the cpuN entry -* - second +1 account for toplevel palinfo -* +* - +1 accounts for the cpuN directory entry in /proc/pal */ -#define NR_PALINFO_PROC_ENTRIES (NR_CPUS*(NR_PALINFO_ENTRIES+1)+1) +#define NR_PALINFO_PROC_ENTRIES (NR_CPUS*(NR_PALINFO_ENTRIES+1))static struct proc_dir_entry *palinfo_proc_entries[NR_PALINFO_PROC_ENTRIES ]; +static struct proc_dir_entry *palinfo_dir;/* * This data structure is used to pass which cpu,function is being requested
@@ -888,47 +893,107 @@ palinfo_read_entry(char *page, char **streturn len; } -static int __init -palinfo_init (void ) +static void +create_palinfo_proc_entries (unsigned int cpu) {# define CPUSTR "cpu%d"pal_func_cpu_u_t f; -struct proc_dir_entry **pdir = palinfo_proc_entries; -struct proc_dir_entry *palinfo_dir, *cpu_dir; -int i, j; +struct proc_dir_entry **pdir; +struct proc_dir_entry *cpu_dir; +int j;char cpustr[sizeof (CPUSTR )]; -printk (KERN_INFO "PAL Information Facility v%s\n" ,PALINFO_VERSION ); - - palinfo_dir =proc_mkdir ("pal" ,NULL );/* * we keep track of created entries in a depth-first order for * cleanup purposes. Each entry is stored into palinfo_proc_entries */ -for (i=0 ; i <NR_CPUS ; i++) { +sprintf (cpustr,CPUSTR , cpu); -if (!cpu_online (i))continue ; + cpu_dir =proc_mkdir (cpustr, palinfo_dir); -sprintf (cpustr,CPUSTR , i); + f.req_cpu = cpu; - cpu_dir =proc_mkdir (cpustr, palinfo_dir); +/* +* Compute the location to store per cpu entries +* We dont store the top level entry in this list, but +* remove it finally after removing all cpu entries. +*/ + pdir = &palinfo_proc_entries[cpu*(NR_PALINFO_ENTRIES +1 )]; + *pdir++ = cpu_dir; +for (j=0 ; j <NR_PALINFO_ENTRIES ; j++) { + f.func_id = j; + *pdir =create_proc_read_entry ( + palinfo_entries[j].name,0 , cpu_dir, + palinfo_read_entry, (void *)f.value); +if (*pdir) + (*pdir)->owner =THIS_MODULE ; + pdir++; + } +} - f.req_cpu = i; +static void +remove_palinfo_proc_entries (unsigned int hcpu) +{ +int j; +struct proc_dir_entry *cpu_dir, **pdir; -for (j=0 ; j <NR_PALINFO_ENTRIES ; j++) { - f.func_id = j; - *pdir =create_proc_read_entry ( - palinfo_entries[j].name,0 , cpu_dir, - palinfo_read_entry, (void *)f.value); -if (*pdir) - (*pdir)->owner =THIS_MODULE ; - pdir++; + pdir = &palinfo_proc_entries[hcpu*(NR_PALINFO_ENTRIES +1 )]; + cpu_dir = *pdir; + *pdir++=NULL ; +for (j=0 ; j < (NR_PALINFO_ENTRIES ); j++) { +if ((*pdir)) { +remove_proc_entry ((*pdir)->name, cpu_dir); + *pdir ++=NULL ; } - *pdir++ = cpu_dir; } - *pdir = palinfo_dir; + +if (cpu_dir) { +remove_proc_entry (cpu_dir->name, palinfo_dir); + } +} + +static int __devinit palinfo_cpu_callback (struct notifier_block *nfb, +unsigned long action, +void *hcpu) +{ +unsigned int hotcpu = (unsigned long )hcpu; + +switch (action) { +case CPU_ONLINE : +create_palinfo_proc_entries (hotcpu); +break ; +#ifdef CONFIG_HOTPLUG_CPU +case CPU_DEAD : +remove_palinfo_proc_entries (hotcpu); +break ; +#endif + } +return NOTIFY_OK ; +} + +static struct notifier_block palinfo_cpu_notifier +{ + .notifier_call = palinfo_cpu_callback, + .priority = 0, +}; + +static int __init +palinfo_init (void ) +{ +int i =0 ; + +printk (KERN_INFO "PAL Information Facility v%s\n" ,PALINFO_VERSION ); + palinfo_dir =proc_mkdir ("pal" ,NULL ); + +/* Create palinfo dirs in /proc for all online cpus */ +for_each_online_cpu (i) { +create_palinfo_proc_entries (i); + } + +/* Register for future delivery via notify registration */ +register_cpu_notifier (&palinfo_cpu_notifier);return 0 ; }
@@ -939,10 +1004,19 @@ palinfo_exit(void)int i =0 ;/* remove all nodes: depth first pass. Could optimize this */ -for (i=0 ; i<NR_PALINFO_PROC_ENTRIES ; i++) { -if (palinfo_proc_entries[i]) -remove_proc_entry (palinfo_proc_entries[i]->name,NULL ); +for_each_online_cpu (i) { +remove_palinfo_proc_entries (i); } + +/* +* Remove the top level entry finally +*/ +remove_proc_entry (palinfo_dir->name,NULL ); + +/* +* Unregister from cpu notifier callbacks +*/ +unregister_cpu_notifier (&palinfo_cpu_notifier); }module_init (palinfo_init);
_
Updated CPU hotplug patches for IA64 [Patch 7/7]
In reply to Ashok RajName: cpu_present_map.patch
Author: Ashok Raj (Intel Corporation)
D: This patch adds cpu_present_map, cpu_present() and for_each_cpu_present()
D: to distinguish between possible cpu's in a system and cpu's physically
D: present in a system. Before cpu hotplug was introduced cpu_possible()
D: represented cpu's physically present in the system. With hotplug capable
D: Kernel, there is a requirement to distinguish a cpu as possible verses a
D: CPU physically present in the system. This is required so thta when
D: smp_init() attempts to start all cpu's it should now only attempt
D: to start cpu's present in the system. When a hotplug cpu is
D: physically inserted cpu_present_map will have bits updated
D: dynamically.
---
arch/ia64/kernel/smp.c | 0 linux-2.6.5-lhcs-root/arch/ia64/kernel/smpboot.c | 21 +++++++++++++++++++-- linux-2.6.5-lhcs-root/fs/proc/proc_misc.c | 4 ++-- linux-2.6.5-lhcs-root/include/asm-ia64/cpumask.h | 1 + linux-2.6.5-lhcs-root/include/asm-ia64/smp.h | 5 +++++ linux-2.6.5-lhcs-root/include/linux/cpumask.h | 16 ++++++++++++++++ linux-2.6.5-lhcs-root/init/main.c | 2 +- linux-2.6.5-lhcs-root/kernel/cpu.c | 2 +- 8 files changed, 45 insertions(+), 6 deletions(-)
include/asm-ia64/smp.h renamed from include/asm-ia64/smp.h~cpu_present +5 -0
@@ -48,7 +48,12 @@ extern volatile int ia64_cpu_to_sapicid[extern unsigned long ap_wakeup_vector; +#ifdef CONFIG_HOTPLUG_CPU +extern cpumask_t cpu_possible_map; +#define cpu_present_map phys_cpu_present_map +#else #define cpu_possible_map phys_cpu_present_map +#endif /* * Function to map hard smp processor id to logical id. Slow, so don't use this in
arch/ia64/kernel/smpboot.c renamed from arch/ia64/kernel/smpboot.c~cpu_present +19 -2
@@ -87,9 +87,15 @@ DEFINE_PER_CPU(int, cpu_state);/* Bitmask of currently online CPUs */ cpumask_t cpu_online_map;EXPORT_SYMBOL (cpu_online_map); +cpumask_t phys_cpu_present_map;EXPORT_SYMBOL (phys_cpu_present_map); +#ifdef CONFIG_HOTPLUG_CPU +cpumask_t cpu_possible_map; +EXPORT_SYMBOL (cpu_possible_map); +#endif +/* which logical CPU number maps to which CPU (physical APIC ID) */ volatile int ia64_cpu_to_sapicid[NR_CPUS ];EXPORT_SYMBOL (ia64_cpu_to_sapicid);
@@ -108,6 +114,7 @@ static int __init nointroute (char *str) { no_int_routing =1 ; +printk ("no_int_routing on\n" );return 1 ; }
@@ -472,13 +479,23 @@ smp_build_cpu_map (void)int sapicid, cpu, i;int boot_cpu_id =hard_smp_processor_id (); -for (cpu =0 ; cpu <NR_CPUS ; cpu++) +/* +* Since ACPI does not give us whats MAX cpu's +* possible in the entire platform, we will just have to use +* NR_CPUS set for possible map. cpu_present_map a.k.a +* phys_cpu_present_map can grow based on a new cpu +* becomming available to the system. +*/ +cpus_clear (cpu_possible_map); +for (cpu=0 ; cpu <NR_CPUS ; cpu++) { ia64_cpu_to_sapicid[cpu] =-1 ; +cpu_set (cpu, cpu_possible_map); + } ia64_cpu_to_sapicid[0 ] = boot_cpu_id; +cpus_clear (phys_cpu_present_map);cpu_set (0 , phys_cpu_present_map); -for (cpu =1 , i =0 ; i < smp_boot_data.cpu_count; i++) { sapicid = smp_boot_data.cpu_phys_id[i];if (sapicid = boot_cpu_id)
include/linux/cpumask.h renamed from include/linux/cpumask.h~cpu_present +16 -0
@@ -16,6 +16,16 @@ extern cpumask_t cpu_possible_map;#define cpu_online (cpu) cpu_isset(cpu, cpu_online_map)#define cpu_possible (cpu) cpu_isset(cpu, cpu_possible_map) +#ifdef ARCH_HAS_CPU_PRESENT_MAP +extern cpumask_t cpu_present_map; +#define cpu_present (cpu) cpu_isset(cpu, cpu_present_map) +#define num_present_cpus () cpus_weight(cpu_present_map) +#else +#define cpu_present_map cpu_possible_map +#define cpu_present (x) cpu_possible(x) +#define num_present_cpus () num_possible_cpus() +#endif +#define for_each_cpu_mask (cpu, mask) \ for (cpu = first_cpu_const(mk_cpumask_const(mask)); \ cpu < NR_CPUS; \
@@ -23,13 +33,19 @@ extern cpumask_t cpu_possible_map;#define for_each_cpu (cpu) for_each_cpu_mask(cpu, cpu_possible_map)#define for_each_online_cpu (cpu) for_each_cpu_mask(cpu, cpu_online_map) +#define for_each_present_cpu (cpu) for_each_cpu_mask(cpu, cpu_present_map)#else #define cpu_online_map cpumask_of_cpu(0)#define cpu_possible_map cpumask_of_cpu(0) +#define cpu_present_map cpumask_of_cpu(0) +#define num_online_cpus () 1#define num_possible_cpus () 1 +#define num_present_cpus 1 +#define cpu_online (cpu) ({ BUG_ON((cpu) != 0); 1; })#define cpu_possible (cpu) ({ BUG_ON((cpu) != 0); 1; }) +#define cpu_present (cpu) ({ BUG_ON((cpu) != 0); 1; })#define for_each_cpu (cpu) for (cpu = 0; cpu < 1; cpu++)#define for_each_online_cpu (cpu) for (cpu = 0; cpu < 1; cpu++)
init/main.c renamed from init/main.c~cpu_present +1 -1
@@ -360,7 +360,7 @@ static void __init smp_init(void)for (i =0 ; i <NR_CPUS ; i++) {if (num_online_cpus () >= max_cpus)break ; -if (cpu_possible (i) && !cpu_online (i)) { +if (cpu_possible (i) && !cpu_online (i) &&cpu_present (i)) {cpu_up (i); j++; }
kernel/cpu.c renamed from kernel/cpu.c~cpu_present +1 -1
@@ -169,7 +169,7 @@ int __devinit cpu_up(unsigned int cpu)if ((ret =down_interruptible (&cpucontrol)) !=0 )return ret; -if (cpu_online (cpu)) { +if (cpu_online (cpu) || !cpu_present (cpu)) { ret = -EINVAL ; goto out; }
diff -puN arch/ia64/kernel/smp.c~cpu_present arch/ia64/kernel/smp.c
fs/proc/proc_misc.c renamed from fs/proc/proc_misc.c~cpu_present +2 -2
@@ -368,7 +368,7 @@ int show_stat(struct seq_file *p, void *if (wall_to_monotonic.tv_nsec) --jif; -for_each_cpu (i) { +for_each_online_cpu (i) {int j; user +=kstat_cpu (i).cpustat.user;
@@ -390,7 +390,7 @@ int show_stat(struct seq_file *p, void * (unsigned long long )jiffies_64_to_clock_t (iowait), (unsigned long long )jiffies_64_to_clock_t (irq), (unsigned long long)jiffies_64_to_clock_t(softirq)); -for_each_cpu (i) { +for_each_online_cpu (i) {/* Copy values here to work around gcc-2.95.3, gcc-2.96 */ user =kstat_cpu (i).cpustat.user;
include/asm-ia64/cpumask.h renamed from include/asm-ia64/cpumask.h~cpu_present +1 -0
@@ -1,6 +1,7 @@#ifndef _ASM_IA64_CPUMASK_H#define _ASM_IA64_CPUMASK_H +#define ARCH_HAS_CPU_PRESENT_MAP #include <asm-generic/cpumask.h> #endif /* _ASM_IA64_CPUMASK_H */
_