#include <linux/sched.h>
#include <linux/kernel_stat.h>
#include <linux/smp_lock.h>
-#include <linux/irq.h>
#include <linux/bootmem.h>
+#include <linux/notifier.h>
+#include <linux/cpu.h>
+#include <linux/percpu.h>
#include <linux/delay.h>
#include <linux/mc146818rtc.h>
#include <smpboot_hooks.h>
/* Set if we find a B stepping CPU */
-static int __initdata smp_b_stepping;
+static int __devinitdata smp_b_stepping;
/* Number of siblings per CPU package */
int smp_num_siblings = 1;
-int phys_proc_id[NR_CPUS]; /* Package ID of each logical CPU */
-EXPORT_SYMBOL(phys_proc_id);
-int cpu_core_id[NR_CPUS]; /* Core ID of each logical CPU */
-EXPORT_SYMBOL(cpu_core_id);
+#ifdef CONFIG_X86_HT
+EXPORT_SYMBOL(smp_num_siblings);
+#endif
+
+/* Package ID of each logical CPU */
+int phys_proc_id[NR_CPUS] __read_mostly = {[0 ... NR_CPUS-1] = BAD_APICID};
+
+/* Core ID of each logical CPU */
+int cpu_core_id[NR_CPUS] __read_mostly = {[0 ... NR_CPUS-1] = BAD_APICID};
+
+/* Last level cache ID of each logical CPU */
+int cpu_llc_id[NR_CPUS] __cpuinitdata = {[0 ... NR_CPUS-1] = BAD_APICID};
+
+/* representing HT siblings of each logical CPU */
+cpumask_t cpu_sibling_map[NR_CPUS] __read_mostly;
+EXPORT_SYMBOL(cpu_sibling_map);
+
+/* representing HT and core siblings of each logical CPU */
+cpumask_t cpu_core_map[NR_CPUS] __read_mostly;
+EXPORT_SYMBOL(cpu_core_map);
/* bitmap of online cpus */
-cpumask_t cpu_online_map;
+cpumask_t cpu_online_map __read_mostly;
+EXPORT_SYMBOL(cpu_online_map);
cpumask_t cpu_callin_map;
cpumask_t cpu_callout_map;
+EXPORT_SYMBOL(cpu_callout_map);
+cpumask_t cpu_possible_map;
+EXPORT_SYMBOL(cpu_possible_map);
static cpumask_t smp_commenced_mask;
+/* TSC's upper 32 bits can't be written in eariler CPU (before prescott), there
+ * is no way to resync one AP against BP. TBD: for prescott and above, we
+ * should use IA64's algorithm
+ */
+static int __devinitdata tsc_sync_disabled;
+
/* Per CPU bogomips and other parameters */
struct cpuinfo_x86 cpu_data[NR_CPUS] __cacheline_aligned;
+EXPORT_SYMBOL(cpu_data);
-u8 x86_cpu_to_apicid[NR_CPUS] =
+u8 x86_cpu_to_apicid[NR_CPUS] __read_mostly =
{ [0 ... NR_CPUS-1] = 0xff };
EXPORT_SYMBOL(x86_cpu_to_apicid);
static void map_cpu_to_logical_apicid(void);
+/* State of each CPU. */
+DEFINE_PER_CPU(int, cpu_state) = { 0 };
+
/*
* Currently trivial. Write the real->protected mode
* bootstrap into the page concerned. The caller
* has made sure it's suitably aligned.
*/
-static unsigned long __init setup_trampoline(void)
+static unsigned long __devinit setup_trampoline(void)
{
memcpy(trampoline_base, trampoline_data, trampoline_end - trampoline_data);
return virt_to_phys(trampoline_base);
* a given CPU
*/
-static void __init smp_store_cpu_info(int id)
+static void __devinit smp_store_cpu_info(int id)
{
struct cpuinfo_x86 *c = cpu_data + id;
goto valid_k7;
/* If we get here, it's not a certified SMP capable AMD system. */
- tainted |= TAINT_UNSAFE_SMP;
+ add_taint(TAINT_UNSAFE_SMP);
}
valid_k7:
unsigned long long t0;
unsigned long long sum, avg;
long long delta;
- unsigned long one_usec;
+ unsigned int one_usec;
int buggy = 0;
printk(KERN_INFO "checking TSC synchronization across %u CPUs: ", num_booting_cpus());
if (tsc_values[i] < avg)
realdelta = -realdelta;
- printk(KERN_INFO "CPU#%d had %ld usecs TSC skew, fixed it up.\n", i, realdelta);
+ if (realdelta > 0)
+ printk(KERN_INFO "CPU#%d had %ld usecs TSC "
+ "skew, fixed it up.\n", i, realdelta);
}
sum += delta;
static atomic_t init_deasserted;
-static void __init smp_callin(void)
+static void __devinit smp_callin(void)
{
int cpuid, phys_id;
unsigned long timeout;
/*
* Synchronize the TSC with the BP
*/
- if (cpu_has_tsc && cpu_khz)
+ if (cpu_has_tsc && cpu_khz && !tsc_sync_disabled)
synchronize_tsc_ap();
}
static int cpucount;
+/* maps the cpu to the sched domain representing multi-core */
+cpumask_t cpu_coregroup_map(int cpu)
+{
+ struct cpuinfo_x86 *c = cpu_data + cpu;
+ /*
+ * For perf, we return last level cache shared map.
+ * TBD: when power saving sched policy is added, we will return
+ * cpu_core_map when power saving policy is enabled
+ */
+ return c->llc_shared_map;
+}
+
+/* representing cpus for which sibling maps can be computed */
+static cpumask_t cpu_sibling_setup_map;
+
+static inline void
+set_cpu_sibling_map(int cpu)
+{
+ int i;
+ struct cpuinfo_x86 *c = cpu_data;
+
+ cpu_set(cpu, cpu_sibling_setup_map);
+
+ if (smp_num_siblings > 1) {
+ for_each_cpu_mask(i, cpu_sibling_setup_map) {
+ if (phys_proc_id[cpu] == phys_proc_id[i] &&
+ cpu_core_id[cpu] == cpu_core_id[i]) {
+ cpu_set(i, cpu_sibling_map[cpu]);
+ cpu_set(cpu, cpu_sibling_map[i]);
+ cpu_set(i, cpu_core_map[cpu]);
+ cpu_set(cpu, cpu_core_map[i]);
+ cpu_set(i, c[cpu].llc_shared_map);
+ cpu_set(cpu, c[i].llc_shared_map);
+ }
+ }
+ } else {
+ cpu_set(cpu, cpu_sibling_map[cpu]);
+ }
+
+ cpu_set(cpu, c[cpu].llc_shared_map);
+
+ if (current_cpu_data.x86_max_cores == 1) {
+ cpu_core_map[cpu] = cpu_sibling_map[cpu];
+ c[cpu].booted_cores = 1;
+ return;
+ }
+
+ for_each_cpu_mask(i, cpu_sibling_setup_map) {
+ if (cpu_llc_id[cpu] != BAD_APICID &&
+ cpu_llc_id[cpu] == cpu_llc_id[i]) {
+ cpu_set(i, c[cpu].llc_shared_map);
+ cpu_set(cpu, c[i].llc_shared_map);
+ }
+ if (phys_proc_id[cpu] == phys_proc_id[i]) {
+ cpu_set(i, cpu_core_map[cpu]);
+ cpu_set(cpu, cpu_core_map[i]);
+ /*
+ * Does this new cpu bringup a new core?
+ */
+ if (cpus_weight(cpu_sibling_map[cpu]) == 1) {
+ /*
+ * for each core in package, increment
+ * the booted_cores for this new cpu
+ */
+ if (first_cpu(cpu_sibling_map[i]) == i)
+ c[cpu].booted_cores++;
+ /*
+ * increment the core count for all
+ * the other cpus in this package
+ */
+ if (i != cpu)
+ c[i].booted_cores++;
+ } else if (i != cpu && !c[cpu].booted_cores)
+ c[cpu].booted_cores = c[i].booted_cores;
+ }
+ }
+}
+
/*
* Activate a secondary processor.
*/
-static void __init start_secondary(void *unused)
+static void __devinit start_secondary(void *unused)
{
/*
* Dont put anything before smp_callin(), SMP
* things done here to the most necessary things.
*/
cpu_init();
+ preempt_disable();
smp_callin();
while (!cpu_isset(smp_processor_id(), smp_commenced_mask))
rep_nop();
* the local TLBs too.
*/
local_flush_tlb();
+
+ /* This must be done before setting cpu_online_map */
+ set_cpu_sibling_map(raw_smp_processor_id());
+ wmb();
+
+ /*
+ * We need to hold call_lock, so there is no inconsistency
+ * between the time smp_call_function() determines number of
+ * IPI receipients, and the time when the determination is made
+ * for which cpus receive the IPI. Holding this
+ * lock helps us to not include this cpu in a currently in progress
+ * smp_call_function().
+ */
+ lock_ipi_call_lock();
cpu_set(smp_processor_id(), cpu_online_map);
+ unlock_ipi_call_lock();
+ per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
/* We can take interrupts now: we're officially "up". */
local_irq_enable();
* from the task structure
* This function must not return.
*/
-void __init initialize_secondary(void)
+void __devinit initialize_secondary(void)
{
/*
* We don't actually need to load the full TSS,
#ifdef CONFIG_NUMA
/* which logical CPUs are on which nodes */
-cpumask_t node_2_cpu_mask[MAX_NUMNODES] =
+cpumask_t node_2_cpu_mask[MAX_NUMNODES] __read_mostly =
{ [0 ... MAX_NUMNODES-1] = CPU_MASK_NONE };
/* which node each logical CPU is on */
-int cpu_2_node[NR_CPUS] = { [0 ... NR_CPUS-1] = 0 };
+int cpu_2_node[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = 0 };
EXPORT_SYMBOL(cpu_2_node);
/* set up a mapping between cpu and node. */
#endif /* CONFIG_NUMA */
-u8 cpu_2_logical_apicid[NR_CPUS] = { [0 ... NR_CPUS-1] = BAD_APICID };
+u8 cpu_2_logical_apicid[NR_CPUS] __read_mostly = { [0 ... NR_CPUS-1] = BAD_APICID };
static void map_cpu_to_logical_apicid(void)
{
printk("Inquiring remote APIC #%d...\n", apicid);
- for (i = 0; i < sizeof(regs) / sizeof(*regs); i++) {
+ for (i = 0; i < ARRAY_SIZE(regs); i++) {
printk("... APIC #%d %s: ", apicid, names[i]);
/*
* INIT, INIT, STARTUP sequence will reset the chip hard for us, and this
* won't ... remember to clear down the APIC, etc later.
*/
-static int __init
+static int __devinit
wakeup_secondary_cpu(int logical_apicid, unsigned long start_eip)
{
unsigned long send_status = 0, accept_status = 0;
#endif /* WAKE_SECONDARY_VIA_NMI */
#ifdef WAKE_SECONDARY_VIA_INIT
-static int __init
+static int __devinit
wakeup_secondary_cpu(int phys_apicid, unsigned long start_eip)
{
unsigned long send_status = 0, accept_status = 0;
#endif /* WAKE_SECONDARY_VIA_INIT */
extern cpumask_t cpu_initialized;
+static inline int alloc_cpu_id(void)
+{
+ cpumask_t tmp_map;
+ int cpu;
+ cpus_complement(tmp_map, cpu_present_map);
+ cpu = first_cpu(tmp_map);
+ if (cpu >= NR_CPUS)
+ return -ENODEV;
+ return cpu;
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+static struct task_struct * __devinitdata cpu_idle_tasks[NR_CPUS];
+static inline struct task_struct * alloc_idle_task(int cpu)
+{
+ struct task_struct *idle;
-static int __init do_boot_cpu(int apicid)
+ if ((idle = cpu_idle_tasks[cpu]) != NULL) {
+ /* initialize thread_struct. we really want to avoid destroy
+ * idle tread
+ */
+ idle->thread.esp = (unsigned long)task_pt_regs(idle);
+ init_idle(idle, cpu);
+ return idle;
+ }
+ idle = fork_idle(cpu);
+
+ if (!IS_ERR(idle))
+ cpu_idle_tasks[cpu] = idle;
+ return idle;
+}
+#else
+#define alloc_idle_task(cpu) fork_idle(cpu)
+#endif
+
+static int __devinit do_boot_cpu(int apicid, int cpu)
/*
* NOTE - on most systems this is a PHYSICAL apic ID, but on multiquad
* (ie clustered apic addressing mode), this is a LOGICAL apic ID.
{
struct task_struct *idle;
unsigned long boot_error;
- int timeout, cpu;
+ int timeout;
unsigned long start_eip;
unsigned short nmi_high = 0, nmi_low = 0;
- cpu = ++cpucount;
+ ++cpucount;
+ alternatives_smp_switch(1);
+
/*
* We can't use kernel_thread since we must avoid to
* reschedule the child.
*/
- idle = fork_idle(cpu);
+ idle = alloc_idle_task(cpu);
if (IS_ERR(idle))
panic("failed fork for CPU %d", cpu);
idle->thread.eip = (unsigned long) start_secondary;
inquire_remote_apic(apicid);
}
}
- x86_cpu_to_apicid[cpu] = apicid;
+
if (boot_error) {
/* Try to put things back the way they were before ... */
unmap_cpu_to_logical_apicid(cpu);
cpu_clear(cpu, cpu_callout_map); /* was set here (do_boot_cpu()) */
cpu_clear(cpu, cpu_initialized); /* was set by cpu_init() */
cpucount--;
+ } else {
+ x86_cpu_to_apicid[cpu] = apicid;
+ cpu_set(cpu, cpu_present_map);
}
/* mark "stuck" area as not stuck */
return boot_error;
}
+#ifdef CONFIG_HOTPLUG_CPU
+void cpu_exit_clear(void)
+{
+ int cpu = raw_smp_processor_id();
+
+ idle_task_exit();
+
+ cpucount --;
+ cpu_uninit();
+ irq_ctx_exit(cpu);
+
+ cpu_clear(cpu, cpu_callout_map);
+ cpu_clear(cpu, cpu_callin_map);
+
+ cpu_clear(cpu, smp_commenced_mask);
+ unmap_cpu_to_logical_apicid(cpu);
+}
+
+struct warm_boot_cpu_info {
+ struct completion *complete;
+ int apicid;
+ int cpu;
+};
+
+static void __cpuinit do_warm_boot_cpu(void *p)
+{
+ struct warm_boot_cpu_info *info = p;
+ do_boot_cpu(info->apicid, info->cpu);
+ complete(info->complete);
+}
+
+static int __cpuinit __smp_prepare_cpu(int cpu)
+{
+ DECLARE_COMPLETION(done);
+ struct warm_boot_cpu_info info;
+ struct work_struct task;
+ int apicid, ret;
+
+ apicid = x86_cpu_to_apicid[cpu];
+ if (apicid == BAD_APICID) {
+ ret = -ENODEV;
+ goto exit;
+ }
+
+ info.complete = &done;
+ info.apicid = apicid;
+ info.cpu = cpu;
+ INIT_WORK(&task, do_warm_boot_cpu, &info);
+
+ tsc_sync_disabled = 1;
+
+ /* init low mem mapping */
+ clone_pgd_range(swapper_pg_dir, swapper_pg_dir + USER_PGD_PTRS,
+ KERNEL_PGD_PTRS);
+ flush_tlb_all();
+ schedule_work(&task);
+ wait_for_completion(&done);
+
+ tsc_sync_disabled = 0;
+ zap_low_mappings();
+ ret = 0;
+exit:
+ return ret;
+}
+#endif
+
static void smp_tune_scheduling (void)
{
unsigned long cachesize; /* kB */
cachesize = 16; /* Pentiums, 2x8kB cache */
bandwidth = 100;
}
+ max_cache_size = cachesize * 1024;
}
}
static int boot_cpu_logical_apicid;
/* Where the IO area was mapped on multiquad, always 0 otherwise */
void *xquad_portio;
-
-cpumask_t cpu_sibling_map[NR_CPUS] __cacheline_aligned;
-cpumask_t cpu_core_map[NR_CPUS] __cacheline_aligned;
-EXPORT_SYMBOL(cpu_core_map);
+#ifdef CONFIG_X86_NUMAQ
+EXPORT_SYMBOL(xquad_portio);
+#endif
static void __init smp_boot_cpus(unsigned int max_cpus)
{
current_thread_info()->cpu = 0;
smp_tune_scheduling();
- cpus_clear(cpu_sibling_map[0]);
- cpu_set(0, cpu_sibling_map[0]);
- cpus_clear(cpu_core_map[0]);
- cpu_set(0, cpu_core_map[0]);
+ set_cpu_sibling_map(0);
/*
* If we couldn't find an SMP configuration at boot time,
if (max_cpus <= cpucount+1)
continue;
- if (do_boot_cpu(apicid))
+ if (((cpu = alloc_cpu_id()) <= 0) || do_boot_cpu(apicid, cpu))
printk("CPU #%d not responding - cannot use it.\n",
apicid);
else
printk(KERN_INFO
"Total of %d processors activated (%lu.%02lu BogoMIPS).\n",
cpucount+1,
- HZ*(bogosum >> 3)/62500,
- (HZ*(bogosum >> 3)/625) % 100);
+ bogosum/(500000/HZ),
+ (bogosum/(5000/HZ))%100);
Dprintk("Before bogocount - setting activated=1.\n");
cpus_clear(cpu_core_map[cpu]);
}
- for (cpu = 0; cpu < NR_CPUS; cpu++) {
- struct cpuinfo_x86 *c = cpu_data + cpu;
- int siblings = 0;
- int i;
- if (!cpu_isset(cpu, cpu_callout_map))
- continue;
-
- if (smp_num_siblings > 1) {
- for (i = 0; i < NR_CPUS; i++) {
- if (!cpu_isset(i, cpu_callout_map))
- continue;
- if (cpu_core_id[cpu] == cpu_core_id[i]) {
- siblings++;
- cpu_set(i, cpu_sibling_map[cpu]);
- }
- }
- } else {
- siblings++;
- cpu_set(cpu, cpu_sibling_map[cpu]);
- }
-
- if (siblings != smp_num_siblings) {
- printk(KERN_WARNING "WARNING: %d siblings found for CPU%d, should be %d\n", siblings, cpu, smp_num_siblings);
- smp_num_siblings = siblings;
- }
-
- if (c->x86_num_cores > 1) {
- for (i = 0; i < NR_CPUS; i++) {
- if (!cpu_isset(i, cpu_callout_map))
- continue;
- if (phys_proc_id[cpu] == phys_proc_id[i]) {
- cpu_set(i, cpu_core_map[cpu]);
- }
- }
- } else {
- cpu_core_map[cpu] = cpu_sibling_map[cpu];
- }
- }
+ cpu_set(0, cpu_sibling_map[0]);
+ cpu_set(0, cpu_core_map[0]);
smpboot_setup_io_apic();
who understands all this stuff should rewrite it properly. --RR 15/Jul/02 */
void __init smp_prepare_cpus(unsigned int max_cpus)
{
+ smp_commenced_mask = cpumask_of_cpu(0);
+ cpu_callin_map = cpumask_of_cpu(0);
+ mb();
smp_boot_cpus(max_cpus);
}
{
cpu_set(smp_processor_id(), cpu_online_map);
cpu_set(smp_processor_id(), cpu_callout_map);
+ cpu_set(smp_processor_id(), cpu_present_map);
+ cpu_set(smp_processor_id(), cpu_possible_map);
+ per_cpu(cpu_state, smp_processor_id()) = CPU_ONLINE;
}
-int __devinit __cpu_up(unsigned int cpu)
+#ifdef CONFIG_HOTPLUG_CPU
+static void
+remove_siblinginfo(int cpu)
{
- /* This only works at boot for x86. See "rewrite" above. */
- if (cpu_isset(cpu, smp_commenced_mask)) {
- local_irq_enable();
- return -ENOSYS;
+ int sibling;
+ struct cpuinfo_x86 *c = cpu_data;
+
+ for_each_cpu_mask(sibling, cpu_core_map[cpu]) {
+ cpu_clear(cpu, cpu_core_map[sibling]);
+ /*
+ * last thread sibling in this cpu core going down
+ */
+ if (cpus_weight(cpu_sibling_map[cpu]) == 1)
+ c[sibling].booted_cores--;
}
+
+ for_each_cpu_mask(sibling, cpu_sibling_map[cpu])
+ cpu_clear(cpu, cpu_sibling_map[sibling]);
+ cpus_clear(cpu_sibling_map[cpu]);
+ cpus_clear(cpu_core_map[cpu]);
+ phys_proc_id[cpu] = BAD_APICID;
+ cpu_core_id[cpu] = BAD_APICID;
+ cpu_clear(cpu, cpu_sibling_setup_map);
+}
+
+int __cpu_disable(void)
+{
+ cpumask_t map = cpu_online_map;
+ int cpu = smp_processor_id();
+
+ /*
+ * Perhaps use cpufreq to drop frequency, but that could go
+ * into generic code.
+ *
+ * We won't take down the boot processor on i386 due to some
+ * interrupts only being able to be serviced by the BSP.
+ * Especially so if we're not using an IOAPIC -zwane
+ */
+ if (cpu == 0)
+ return -EBUSY;
+
+ clear_local_APIC();
+ /* Allow any queued timer interrupts to get serviced */
+ local_irq_enable();
+ mdelay(1);
+ local_irq_disable();
+
+ remove_siblinginfo(cpu);
+
+ cpu_clear(cpu, map);
+ fixup_irqs(map);
+ /* It's now safe to remove this processor from the online map */
+ cpu_clear(cpu, cpu_online_map);
+ return 0;
+}
+
+void __cpu_die(unsigned int cpu)
+{
+ /* We don't do anything here: idle task is faking death itself. */
+ unsigned int i;
+
+ for (i = 0; i < 10; i++) {
+ /* They ack this in play_dead by setting CPU_DEAD */
+ if (per_cpu(cpu_state, cpu) == CPU_DEAD) {
+ printk ("CPU %d is now offline\n", cpu);
+ if (1 == num_online_cpus())
+ alternatives_smp_switch(0);
+ return;
+ }
+ msleep(100);
+ }
+ printk(KERN_ERR "CPU %u didn't die...\n", cpu);
+}
+#else /* ... !CONFIG_HOTPLUG_CPU */
+int __cpu_disable(void)
+{
+ return -ENOSYS;
+}
+
+void __cpu_die(unsigned int cpu)
+{
+ /* We said "no" in __cpu_disable */
+ BUG();
+}
+#endif /* CONFIG_HOTPLUG_CPU */
+
+int __devinit __cpu_up(unsigned int cpu)
+{
+#ifdef CONFIG_HOTPLUG_CPU
+ int ret=0;
+
+ /*
+ * We do warm boot only on cpus that had booted earlier
+ * Otherwise cold boot is all handled from smp_boot_cpus().
+ * cpu_callin_map is set during AP kickstart process. Its reset
+ * when a cpu is taken offline from cpu_exit_clear().
+ */
+ if (!cpu_isset(cpu, cpu_callin_map))
+ ret = __smp_prepare_cpu(cpu);
+
+ if (ret)
+ return -EIO;
+#endif
/* In case one didn't come up */
if (!cpu_isset(cpu, cpu_callin_map)) {
+ printk(KERN_DEBUG "skipping cpu%d, didn't come online\n", cpu);
local_irq_enable();
return -EIO;
}
local_irq_enable();
+ per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
/* Unleash the CPU! */
cpu_set(cpu, smp_commenced_mask);
while (!cpu_isset(cpu, cpu_online_map))
setup_ioapic_dest();
#endif
zap_low_mappings();
+#ifndef CONFIG_HOTPLUG_CPU
/*
* Disable executability of the SMP trampoline:
*/
set_kernel_exec((unsigned long)trampoline_base, trampoline_exec);
+#endif
}
void __init smp_intr_init(void)
git://git.onelab.eu / linux-2.6.git / blobdiff