* Dave Jones : Report invalid combinations of Athlon CPUs.
* Rusty Russell : Hacked into shape for new "hotplug" boot process. */
+
+/* SMP boot always wants to use real time delay to allow sufficient time for
+ * the APs to come online */
+#define USE_REAL_TIME_DELAY
+
#include <linux/module.h>
#include <linux/init.h>
#include <linux/kernel.h>
#include <asm/desc.h>
#include <asm/arch_hooks.h>
#include <asm/nmi.h>
+#include <asm/pda.h>
+#include <asm/genapic.h>
#include <mach_apic.h>
#include <mach_wakecpu.h>
#include <smpboot_hooks.h>
+#include <asm/vmi.h>
/* Set if we find a B stepping CPU */
static int __devinitdata smp_b_stepping;
/* Number of siblings per CPU package */
int smp_num_siblings = 1;
-#ifdef CONFIG_X86_HT
EXPORT_SYMBOL(smp_num_siblings);
-#endif
/* Last level cache ID of each logical CPU */
int cpu_llc_id[NR_CPUS] __cpuinitdata = {[0 ... NR_CPUS-1] = BAD_APICID};
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);
* a given CPU
*/
-static void __devinit smp_store_cpu_info(int id)
+static void __cpuinit smp_store_cpu_info(int id)
{
struct cpuinfo_x86 *c = cpu_data + id;
;
}
-/*
- * TSC synchronization.
- *
- * We first check whether all CPUs have their TSC's synchronized,
- * then we print a warning if not, and always resync.
- */
-
-static struct {
- atomic_t start_flag;
- atomic_t count_start;
- atomic_t count_stop;
- unsigned long long values[NR_CPUS];
-} tsc __initdata = {
- .start_flag = ATOMIC_INIT(0),
- .count_start = ATOMIC_INIT(0),
- .count_stop = ATOMIC_INIT(0),
-};
-
-#define NR_LOOPS 5
-
-static void __init synchronize_tsc_bp(void)
-{
- int i;
- unsigned long long t0;
- unsigned long long sum, avg;
- long long delta;
- unsigned int one_usec;
- int buggy = 0;
-
- printk(KERN_INFO "checking TSC synchronization across %u CPUs: ", num_booting_cpus());
-
- /* convert from kcyc/sec to cyc/usec */
- one_usec = cpu_khz / 1000;
-
- atomic_set(&tsc.start_flag, 1);
- wmb();
-
- /*
- * We loop a few times to get a primed instruction cache,
- * then the last pass is more or less synchronized and
- * the BP and APs set their cycle counters to zero all at
- * once. This reduces the chance of having random offsets
- * between the processors, and guarantees that the maximum
- * delay between the cycle counters is never bigger than
- * the latency of information-passing (cachelines) between
- * two CPUs.
- */
- for (i = 0; i < NR_LOOPS; i++) {
- /*
- * all APs synchronize but they loop on '== num_cpus'
- */
- while (atomic_read(&tsc.count_start) != num_booting_cpus()-1)
- cpu_relax();
- atomic_set(&tsc.count_stop, 0);
- wmb();
- /*
- * this lets the APs save their current TSC:
- */
- atomic_inc(&tsc.count_start);
-
- rdtscll(tsc.values[smp_processor_id()]);
- /*
- * We clear the TSC in the last loop:
- */
- if (i == NR_LOOPS-1)
- write_tsc(0, 0);
-
- /*
- * Wait for all APs to leave the synchronization point:
- */
- while (atomic_read(&tsc.count_stop) != num_booting_cpus()-1)
- cpu_relax();
- atomic_set(&tsc.count_start, 0);
- wmb();
- atomic_inc(&tsc.count_stop);
- }
-
- sum = 0;
- for (i = 0; i < NR_CPUS; i++) {
- if (cpu_isset(i, cpu_callout_map)) {
- t0 = tsc.values[i];
- sum += t0;
- }
- }
- avg = sum;
- do_div(avg, num_booting_cpus());
-
- for (i = 0; i < NR_CPUS; i++) {
- if (!cpu_isset(i, cpu_callout_map))
- continue;
- delta = tsc.values[i] - avg;
- if (delta < 0)
- delta = -delta;
- /*
- * We report bigger than 2 microseconds clock differences.
- */
- if (delta > 2*one_usec) {
- long long realdelta;
-
- if (!buggy) {
- buggy = 1;
- printk("\n");
- }
- realdelta = delta;
- do_div(realdelta, one_usec);
- if (tsc.values[i] < avg)
- realdelta = -realdelta;
-
- if (realdelta)
- printk(KERN_INFO "CPU#%d had %Ld usecs TSC "
- "skew, fixed it up.\n", i, realdelta);
- }
- }
- if (!buggy)
- printk("passed.\n");
-}
-
-static void __init synchronize_tsc_ap(void)
-{
- int i;
-
- /*
- * Not every cpu is online at the time
- * this gets called, so we first wait for the BP to
- * finish SMP initialization:
- */
- while (!atomic_read(&tsc.start_flag))
- cpu_relax();
-
- for (i = 0; i < NR_LOOPS; i++) {
- atomic_inc(&tsc.count_start);
- while (atomic_read(&tsc.count_start) != num_booting_cpus())
- cpu_relax();
-
- rdtscll(tsc.values[smp_processor_id()]);
- if (i == NR_LOOPS-1)
- write_tsc(0, 0);
-
- atomic_inc(&tsc.count_stop);
- while (atomic_read(&tsc.count_stop) != num_booting_cpus())
- cpu_relax();
- }
-}
-#undef NR_LOOPS
-
extern void calibrate_delay(void);
static atomic_t init_deasserted;
-static void __devinit smp_callin(void)
+static void __cpuinit smp_callin(void)
{
int cpuid, phys_id;
unsigned long timeout;
/*
* Save our processor parameters
*/
- smp_store_cpu_info(cpuid);
-
- disable_APIC_timer();
+ smp_store_cpu_info(cpuid);
/*
* Allow the master to continue.
*/
cpu_set(cpuid, cpu_callin_map);
-
- /*
- * Synchronize the TSC with the BP
- */
- if (cpu_has_tsc && cpu_khz && !tsc_sync_disabled)
- synchronize_tsc_ap();
}
static int cpucount;
/*
* Activate a secondary processor.
*/
-static void __devinit start_secondary(void *unused)
+static void __cpuinit start_secondary(void *unused)
{
/*
- * Dont put anything before smp_callin(), SMP
+ * Don't put *anything* before secondary_cpu_init(), SMP
* booting is too fragile that we want to limit the
* things done here to the most necessary things.
*/
- cpu_init();
+#ifdef CONFIG_VMI
+ vmi_bringup();
+#endif
+ secondary_cpu_init();
preempt_disable();
smp_callin();
while (!cpu_isset(smp_processor_id(), smp_commenced_mask))
rep_nop();
- setup_secondary_APIC_clock();
+ /*
+ * Check TSC synchronization with the BP:
+ */
+ check_tsc_sync_target();
+
+ setup_secondary_clock();
if (nmi_watchdog == NMI_IO_APIC) {
disable_8259A_irq(0);
enable_NMI_through_LVT0(NULL);
enable_8259A_irq(0);
}
- enable_APIC_timer();
/*
* low-memory mappings have been cleared, flush them from
* the local TLBs too.
*/
void __devinit initialize_secondary(void)
{
+ /*
+ * switch to the per CPU GDT we already set up
+ * in do_boot_cpu()
+ */
+ cpu_set_gdt(current_thread_info()->cpu);
+
/*
* We don't actually need to load the full TSS,
* basically just the stack pointer and the eip.
"movl %0,%%esp\n\t"
"jmp *%1"
:
- :"r" (current->thread.esp),"r" (current->thread.eip));
+ :"m" (current->thread.esp),"m" (current->thread.eip));
}
+/* Static state in head.S used to set up a CPU */
extern struct {
void * esp;
unsigned short ss;
} stack_start;
+extern struct i386_pda *start_pda;
#ifdef CONFIG_NUMA
/*
* Due to the Pentium erratum 3AP.
*/
- maxlvt = get_maxlvt();
+ maxlvt = lapic_get_maxlvt();
if (maxlvt > 3) {
apic_read_around(APIC_SPIV);
apic_write(APIC_ESR, 0);
else
num_starts = 0;
+ /*
+ * Paravirt / VMI wants a startup IPI hook here to set up the
+ * target processor state.
+ */
+ startup_ipi_hook(phys_apicid, (unsigned long) start_secondary,
+ (unsigned long) stack_start.esp);
+
/*
* Run STARTUP IPI loop.
*/
Dprintk("#startup loops: %d.\n", num_starts);
- maxlvt = get_maxlvt();
+ maxlvt = lapic_get_maxlvt();
for (j = 1; j <= num_starts; j++) {
Dprintk("Sending STARTUP #%d.\n",j);
#define alloc_idle_task(cpu) fork_idle(cpu)
#endif
-static int __devinit do_boot_cpu(int apicid, int cpu)
+static int __cpuinit 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.
unsigned long start_eip;
unsigned short nmi_high = 0, nmi_low = 0;
- ++cpucount;
- alternatives_smp_switch(1);
-
/*
* We can't use kernel_thread since we must avoid to
* reschedule the child.
idle = alloc_idle_task(cpu);
if (IS_ERR(idle))
panic("failed fork for CPU %d", cpu);
+
+ /* Pre-allocate and initialize the CPU's GDT and PDA so it
+ doesn't have to do any memory allocation during the
+ delicate CPU-bringup phase. */
+ if (!init_gdt(cpu, idle)) {
+ printk(KERN_INFO "Couldn't allocate GDT/PDA for CPU %d\n", cpu);
+ return -1; /* ? */
+ }
+
idle->thread.eip = (unsigned long) start_secondary;
/* start_eip had better be page-aligned! */
start_eip = setup_trampoline();
+ ++cpucount;
+ alternatives_smp_switch(1);
+
/* So we see what's up */
printk("Booting processor %d/%d eip %lx\n", cpu, apicid, start_eip);
/* Stack for startup_32 can be just as for start_secondary onwards */
info.cpu = cpu;
INIT_WORK(&info.task, do_warm_boot_cpu);
- tsc_sync_disabled = 1;
-
/* init low mem mapping */
clone_pgd_range(swapper_pg_dir, swapper_pg_dir + USER_PGD_PTRS,
- KERNEL_PGD_PTRS);
+ min_t(unsigned long, KERNEL_PGD_PTRS, USER_PGD_PTRS));
flush_tlb_all();
schedule_work(&info.task);
wait_for_completion(&done);
- tsc_sync_disabled = 0;
zap_low_mappings();
ret = 0;
exit:
}
#endif
-static void smp_tune_scheduling (void)
+static void smp_tune_scheduling(void)
{
unsigned long cachesize; /* kB */
- unsigned long bandwidth = 350; /* MB/s */
- /*
- * Rough estimation for SMP scheduling, this is the number of
- * cycles it takes for a fully memory-limited process to flush
- * the SMP-local cache.
- *
- * (For a P5 this pretty much means we will choose another idle
- * CPU almost always at wakeup time (this is due to the small
- * L1 cache), on PIIs it's around 50-100 usecs, depending on
- * the cache size)
- */
- if (!cpu_khz) {
- /*
- * this basically disables processor-affinity
- * scheduling on SMP without a TSC.
- */
- return;
- } else {
+ if (cpu_khz) {
cachesize = boot_cpu_data.x86_cache_size;
- if (cachesize == -1) {
- cachesize = 16; /* Pentiums, 2x8kB cache */
- bandwidth = 100;
- }
- max_cache_size = cachesize * 1024;
+
+ if (cachesize > 0)
+ max_cache_size = cachesize * 1024;
}
}
smpboot_setup_io_apic();
- setup_boot_APIC_clock();
-
- /*
- * Synchronize the TSC with the AP
- */
- if (cpu_has_tsc && cpucount && cpu_khz)
- synchronize_tsc_bp();
+ setup_boot_clock();
}
/* These are wrappers to interface to the new boot process. Someone
}
#endif /* CONFIG_HOTPLUG_CPU */
-int __devinit __cpu_up(unsigned int cpu)
+int __cpuinit __cpu_up(unsigned int cpu)
{
#ifdef CONFIG_HOTPLUG_CPU
int ret=0;
}
local_irq_enable();
+
per_cpu(cpu_state, cpu) = CPU_UP_PREPARE;
/* Unleash the CPU! */
cpu_set(cpu, smp_commenced_mask);
+
+ /*
+ * Check TSC synchronization with the AP:
+ */
+ check_tsc_sync_source(cpu);
+
while (!cpu_isset(cpu, cpu_online_map))
cpu_relax();
+
+#ifdef CONFIG_X86_GENERICARCH
+ if (num_online_cpus() > 8 && genapic == &apic_default)
+ panic("Default flat APIC routing can't be used with > 8 cpus\n");
+#endif
+
return 0;
}