diff options
Diffstat (limited to 'arch/s390/kernel/smp.c')
| -rw-r--r-- | arch/s390/kernel/smp.c | 1322 |
1 files changed, 791 insertions, 531 deletions
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c index 35edbef1d22..243c7e51260 100644 --- a/arch/s390/kernel/smp.c +++ b/arch/s390/kernel/smp.c @@ -1,25 +1,24 @@ /* - * arch/s390/kernel/smp.c + * SMP related functions * - * Copyright IBM Corp. 1999,2007 - * Author(s): Denis Joseph Barrow (djbarrow@de.ibm.com,barrow_dj@yahoo.com), - * Martin Schwidefsky (schwidefsky@de.ibm.com) - * Heiko Carstens (heiko.carstens@de.ibm.com) + * Copyright IBM Corp. 1999, 2012 + * Author(s): Denis Joseph Barrow, + * Martin Schwidefsky <schwidefsky@de.ibm.com>, + * Heiko Carstens <heiko.carstens@de.ibm.com>, * * based on other smp stuff by * (c) 1995 Alan Cox, CymruNET Ltd <alan@cymru.net> * (c) 1998 Ingo Molnar * - * We work with logical cpu numbering everywhere we can. The only - * functions using the real cpu address (got from STAP) are the sigp - * functions. For all other functions we use the identity mapping. - * That means that cpu_number_map[i] == i for every cpu. cpu_number_map is - * used e.g. to find the idle task belonging to a logical cpu. Every array - * in the kernel is sorted by the logical cpu number and not by the physical - * one which is causing all the confusion with __cpu_logical_map and - * cpu_number_map in other architectures. + * The code outside of smp.c uses logical cpu numbers, only smp.c does + * the translation of logical to physical cpu ids. All new code that + * operates on physical cpu numbers needs to go into smp.c. */ +#define KMSG_COMPONENT "cpu" +#define pr_fmt(fmt) KMSG_COMPONENT ": " fmt + +#include <linux/workqueue.h> #include <linux/module.h> #include <linux/init.h> #include <linux/mm.h> @@ -27,310 +26,434 @@ #include <linux/spinlock.h> #include <linux/kernel_stat.h> #include <linux/delay.h> -#include <linux/cache.h> #include <linux/interrupt.h> +#include <linux/irqflags.h> #include <linux/cpu.h> -#include <linux/timex.h> -#include <linux/bootmem.h> +#include <linux/slab.h> +#include <linux/crash_dump.h> +#include <asm/asm-offsets.h> +#include <asm/switch_to.h> +#include <asm/facility.h> #include <asm/ipl.h> #include <asm/setup.h> -#include <asm/sigp.h> -#include <asm/pgalloc.h> #include <asm/irq.h> -#include <asm/s390_ext.h> -#include <asm/cpcmd.h> #include <asm/tlbflush.h> -#include <asm/timer.h> +#include <asm/vtimer.h> #include <asm/lowcore.h> +#include <asm/sclp.h> +#include <asm/vdso.h> +#include <asm/debug.h> +#include <asm/os_info.h> +#include <asm/sigp.h> +#include "entry.h" -/* - * An array with a pointer the lowcore of every CPU. - */ -struct _lowcore *lowcore_ptr[NR_CPUS]; -EXPORT_SYMBOL(lowcore_ptr); +enum { + ec_schedule = 0, + ec_call_function_single, + ec_stop_cpu, +}; -cpumask_t cpu_online_map = CPU_MASK_NONE; -EXPORT_SYMBOL(cpu_online_map); +enum { + CPU_STATE_STANDBY, + CPU_STATE_CONFIGURED, +}; -cpumask_t cpu_possible_map = CPU_MASK_NONE; -EXPORT_SYMBOL(cpu_possible_map); +struct pcpu { + struct cpu *cpu; + struct _lowcore *lowcore; /* lowcore page(s) for the cpu */ + unsigned long async_stack; /* async stack for the cpu */ + unsigned long panic_stack; /* panic stack for the cpu */ + unsigned long ec_mask; /* bit mask for ec_xxx functions */ + int state; /* physical cpu state */ + int polarization; /* physical polarization */ + u16 address; /* physical cpu address */ +}; -static struct task_struct *current_set[NR_CPUS]; +static u8 boot_cpu_type; +static u16 boot_cpu_address; +static struct pcpu pcpu_devices[NR_CPUS]; -static void smp_ext_bitcall(int, ec_bit_sig); +/* + * The smp_cpu_state_mutex must be held when changing the state or polarization + * member of a pcpu data structure within the pcpu_devices arreay. + */ +DEFINE_MUTEX(smp_cpu_state_mutex); /* - * Structure and data for __smp_call_function_map(). This is designed to - * minimise static memory requirements. It also looks cleaner. + * Signal processor helper functions. */ -static DEFINE_SPINLOCK(call_lock); - -struct call_data_struct { - void (*func) (void *info); - void *info; - cpumask_t started; - cpumask_t finished; - int wait; -}; +static inline int __pcpu_sigp_relax(u16 addr, u8 order, u32 parm, u32 *status) +{ + int cc; + + while (1) { + cc = __pcpu_sigp(addr, order, parm, NULL); + if (cc != SIGP_CC_BUSY) + return cc; + cpu_relax(); + } +} + +static int pcpu_sigp_retry(struct pcpu *pcpu, u8 order, u32 parm) +{ + int cc, retry; -static struct call_data_struct *call_data; + for (retry = 0; ; retry++) { + cc = __pcpu_sigp(pcpu->address, order, parm, NULL); + if (cc != SIGP_CC_BUSY) + break; + if (retry >= 3) + udelay(10); + } + return cc; +} + +static inline int pcpu_stopped(struct pcpu *pcpu) +{ + u32 uninitialized_var(status); + + if (__pcpu_sigp(pcpu->address, SIGP_SENSE, + 0, &status) != SIGP_CC_STATUS_STORED) + return 0; + return !!(status & (SIGP_STATUS_CHECK_STOP|SIGP_STATUS_STOPPED)); +} + +static inline int pcpu_running(struct pcpu *pcpu) +{ + if (__pcpu_sigp(pcpu->address, SIGP_SENSE_RUNNING, + 0, NULL) != SIGP_CC_STATUS_STORED) + return 1; + /* Status stored condition code is equivalent to cpu not running. */ + return 0; +} /* - * 'Call function' interrupt callback + * Find struct pcpu by cpu address. */ -static void do_call_function(void) +static struct pcpu *pcpu_find_address(const struct cpumask *mask, int address) { - void (*func) (void *info) = call_data->func; - void *info = call_data->info; - int wait = call_data->wait; + int cpu; - cpu_set(smp_processor_id(), call_data->started); - (*func)(info); - if (wait) - cpu_set(smp_processor_id(), call_data->finished);; + for_each_cpu(cpu, mask) + if (pcpu_devices[cpu].address == address) + return pcpu_devices + cpu; + return NULL; } -static void __smp_call_function_map(void (*func) (void *info), void *info, - int nonatomic, int wait, cpumask_t map) +static void pcpu_ec_call(struct pcpu *pcpu, int ec_bit) { - struct call_data_struct data; - int cpu, local = 0; + int order; - /* - * Can deadlock when interrupts are disabled or if in wrong context. - */ - WARN_ON(irqs_disabled() || in_irq()); + if (test_and_set_bit(ec_bit, &pcpu->ec_mask)) + return; + order = pcpu_running(pcpu) ? SIGP_EXTERNAL_CALL : SIGP_EMERGENCY_SIGNAL; + pcpu_sigp_retry(pcpu, order, 0); +} - /* - * Check for local function call. We have to have the same call order - * as in on_each_cpu() because of machine_restart_smp(). - */ - if (cpu_isset(smp_processor_id(), map)) { - local = 1; - cpu_clear(smp_processor_id(), map); +static int pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu) +{ + struct _lowcore *lc; + + if (pcpu != &pcpu_devices[0]) { + pcpu->lowcore = (struct _lowcore *) + __get_free_pages(GFP_KERNEL | GFP_DMA, LC_ORDER); + pcpu->async_stack = __get_free_pages(GFP_KERNEL, ASYNC_ORDER); + pcpu->panic_stack = __get_free_page(GFP_KERNEL); + if (!pcpu->lowcore || !pcpu->panic_stack || !pcpu->async_stack) + goto out; } - - cpus_and(map, map, cpu_online_map); - if (cpus_empty(map)) + lc = pcpu->lowcore; + memcpy(lc, &S390_lowcore, 512); + memset((char *) lc + 512, 0, sizeof(*lc) - 512); + lc->async_stack = pcpu->async_stack + ASYNC_SIZE + - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); + lc->panic_stack = pcpu->panic_stack + PAGE_SIZE + - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); + lc->cpu_nr = cpu; + lc->spinlock_lockval = arch_spin_lockval(cpu); +#ifndef CONFIG_64BIT + if (MACHINE_HAS_IEEE) { + lc->extended_save_area_addr = get_zeroed_page(GFP_KERNEL); + if (!lc->extended_save_area_addr) + goto out; + } +#else + if (vdso_alloc_per_cpu(lc)) goto out; +#endif + lowcore_ptr[cpu] = lc; + pcpu_sigp_retry(pcpu, SIGP_SET_PREFIX, (u32)(unsigned long) lc); + return 0; +out: + if (pcpu != &pcpu_devices[0]) { + free_page(pcpu->panic_stack); + free_pages(pcpu->async_stack, ASYNC_ORDER); + free_pages((unsigned long) pcpu->lowcore, LC_ORDER); + } + return -ENOMEM; +} - data.func = func; - data.info = info; - data.started = CPU_MASK_NONE; - data.wait = wait; - if (wait) - data.finished = CPU_MASK_NONE; - - spin_lock(&call_lock); - call_data = &data; +#ifdef CONFIG_HOTPLUG_CPU - for_each_cpu_mask(cpu, map) - smp_ext_bitcall(cpu, ec_call_function); +static void pcpu_free_lowcore(struct pcpu *pcpu) +{ + pcpu_sigp_retry(pcpu, SIGP_SET_PREFIX, 0); + lowcore_ptr[pcpu - pcpu_devices] = NULL; +#ifndef CONFIG_64BIT + if (MACHINE_HAS_IEEE) { + struct _lowcore *lc = pcpu->lowcore; - /* Wait for response */ - while (!cpus_equal(map, data.started)) - cpu_relax(); - if (wait) - while (!cpus_equal(map, data.finished)) - cpu_relax(); - spin_unlock(&call_lock); -out: - if (local) { - local_irq_disable(); - func(info); - local_irq_enable(); + free_page((unsigned long) lc->extended_save_area_addr); + lc->extended_save_area_addr = 0; } +#else + vdso_free_per_cpu(pcpu->lowcore); +#endif + if (pcpu != &pcpu_devices[0]) { + free_page(pcpu->panic_stack); + free_pages(pcpu->async_stack, ASYNC_ORDER); + free_pages((unsigned long) pcpu->lowcore, LC_ORDER); + } +} + +#endif /* CONFIG_HOTPLUG_CPU */ + +static void pcpu_prepare_secondary(struct pcpu *pcpu, int cpu) +{ + struct _lowcore *lc = pcpu->lowcore; + + if (MACHINE_HAS_TLB_LC) + cpumask_set_cpu(cpu, &init_mm.context.cpu_attach_mask); + cpumask_set_cpu(cpu, mm_cpumask(&init_mm)); + atomic_inc(&init_mm.context.attach_count); + lc->cpu_nr = cpu; + lc->spinlock_lockval = arch_spin_lockval(cpu); + lc->percpu_offset = __per_cpu_offset[cpu]; + lc->kernel_asce = S390_lowcore.kernel_asce; + lc->machine_flags = S390_lowcore.machine_flags; + lc->ftrace_func = S390_lowcore.ftrace_func; + lc->user_timer = lc->system_timer = lc->steal_timer = 0; + __ctl_store(lc->cregs_save_area, 0, 15); + save_access_regs((unsigned int *) lc->access_regs_save_area); + memcpy(lc->stfle_fac_list, S390_lowcore.stfle_fac_list, + MAX_FACILITY_BIT/8); +} + +static void pcpu_attach_task(struct pcpu *pcpu, struct task_struct *tsk) +{ + struct _lowcore *lc = pcpu->lowcore; + struct thread_info *ti = task_thread_info(tsk); + + lc->kernel_stack = (unsigned long) task_stack_page(tsk) + + THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs); + lc->thread_info = (unsigned long) task_thread_info(tsk); + lc->current_task = (unsigned long) tsk; + lc->user_timer = ti->user_timer; + lc->system_timer = ti->system_timer; + lc->steal_timer = 0; +} + +static void pcpu_start_fn(struct pcpu *pcpu, void (*func)(void *), void *data) +{ + struct _lowcore *lc = pcpu->lowcore; + + lc->restart_stack = lc->kernel_stack; + lc->restart_fn = (unsigned long) func; + lc->restart_data = (unsigned long) data; + lc->restart_source = -1UL; + pcpu_sigp_retry(pcpu, SIGP_RESTART, 0); } /* - * smp_call_function: - * @func: the function to run; this must be fast and non-blocking - * @info: an arbitrary pointer to pass to the function - * @nonatomic: unused - * @wait: if true, wait (atomically) until function has completed on other CPUs - * - * Run a function on all other CPUs. - * - * You must not call this function with disabled interrupts, from a - * hardware interrupt handler or from a bottom half. + * Call function via PSW restart on pcpu and stop the current cpu. */ -int smp_call_function(void (*func) (void *info), void *info, int nonatomic, - int wait) +static void pcpu_delegate(struct pcpu *pcpu, void (*func)(void *), + void *data, unsigned long stack) { - cpumask_t map; + struct _lowcore *lc = lowcore_ptr[pcpu - pcpu_devices]; + unsigned long source_cpu = stap(); + + __load_psw_mask(PSW_KERNEL_BITS); + if (pcpu->address == source_cpu) + func(data); /* should not return */ + /* Stop target cpu (if func returns this stops the current cpu). */ + pcpu_sigp_retry(pcpu, SIGP_STOP, 0); + /* Restart func on the target cpu and stop the current cpu. */ + mem_assign_absolute(lc->restart_stack, stack); + mem_assign_absolute(lc->restart_fn, (unsigned long) func); + mem_assign_absolute(lc->restart_data, (unsigned long) data); + mem_assign_absolute(lc->restart_source, source_cpu); + asm volatile( + "0: sigp 0,%0,%2 # sigp restart to target cpu\n" + " brc 2,0b # busy, try again\n" + "1: sigp 0,%1,%3 # sigp stop to current cpu\n" + " brc 2,1b # busy, try again\n" + : : "d" (pcpu->address), "d" (source_cpu), + "K" (SIGP_RESTART), "K" (SIGP_STOP) + : "0", "1", "cc"); + for (;;) ; +} - preempt_disable(); - map = cpu_online_map; - cpu_clear(smp_processor_id(), map); - __smp_call_function_map(func, info, nonatomic, wait, map); - preempt_enable(); - return 0; +/* + * Call function on an online CPU. + */ +void smp_call_online_cpu(void (*func)(void *), void *data) +{ + struct pcpu *pcpu; + + /* Use the current cpu if it is online. */ + pcpu = pcpu_find_address(cpu_online_mask, stap()); + if (!pcpu) + /* Use the first online cpu. */ + pcpu = pcpu_devices + cpumask_first(cpu_online_mask); + pcpu_delegate(pcpu, func, data, (unsigned long) restart_stack); } -EXPORT_SYMBOL(smp_call_function); /* - * smp_call_function_single: - * @cpu: the CPU where func should run - * @func: the function to run; this must be fast and non-blocking - * @info: an arbitrary pointer to pass to the function - * @nonatomic: unused - * @wait: if true, wait (atomically) until function has completed on other CPUs - * - * Run a function on one processor. - * - * You must not call this function with disabled interrupts, from a - * hardware interrupt handler or from a bottom half. + * Call function on the ipl CPU. */ -int smp_call_function_single(int cpu, void (*func) (void *info), void *info, - int nonatomic, int wait) +void smp_call_ipl_cpu(void (*func)(void *), void *data) { - preempt_disable(); - __smp_call_function_map(func, info, nonatomic, wait, - cpumask_of_cpu(cpu)); - preempt_enable(); - return 0; + pcpu_delegate(&pcpu_devices[0], func, data, + pcpu_devices->panic_stack + PAGE_SIZE); } -EXPORT_SYMBOL(smp_call_function_single); -static void do_send_stop(void) +int smp_find_processor_id(u16 address) { - int cpu, rc; + int cpu; - /* stop all processors */ - for_each_online_cpu(cpu) { - if (cpu == smp_processor_id()) - continue; - do { - rc = signal_processor(cpu, sigp_stop); - } while (rc == sigp_busy); - } + for_each_present_cpu(cpu) + if (pcpu_devices[cpu].address == address) + return cpu; + return -1; } -static void do_store_status(void) +int smp_vcpu_scheduled(int cpu) { - int cpu, rc; + return pcpu_running(pcpu_devices + cpu); +} - /* store status of all processors in their lowcores (real 0) */ - for_each_online_cpu(cpu) { - if (cpu == smp_processor_id()) - continue; - do { - rc = signal_processor_p( - (__u32)(unsigned long) lowcore_ptr[cpu], cpu, - sigp_store_status_at_address); - } while (rc == sigp_busy); - } +void smp_yield(void) +{ + if (MACHINE_HAS_DIAG44) + asm volatile("diag 0,0,0x44"); } -static void do_wait_for_stop(void) +void smp_yield_cpu(int cpu) +{ + if (MACHINE_HAS_DIAG9C) + asm volatile("diag %0,0,0x9c" + : : "d" (pcpu_devices[cpu].address)); + else if (MACHINE_HAS_DIAG44) + asm volatile("diag 0,0,0x44"); +} + +/* + * Send cpus emergency shutdown signal. This gives the cpus the + * opportunity to complete outstanding interrupts. + */ +static void smp_emergency_stop(cpumask_t *cpumask) { + u64 end; int cpu; - /* Wait for all other cpus to enter stopped state */ - for_each_online_cpu(cpu) { - if (cpu == smp_processor_id()) - continue; - while (!smp_cpu_not_running(cpu)) + end = get_tod_clock() + (1000000UL << 12); + for_each_cpu(cpu, cpumask) { + struct pcpu *pcpu = pcpu_devices + cpu; + set_bit(ec_stop_cpu, &pcpu->ec_mask); + while (__pcpu_sigp(pcpu->address, SIGP_EMERGENCY_SIGNAL, + 0, NULL) == SIGP_CC_BUSY && + get_tod_clock() < end) cpu_relax(); } + while (get_tod_clock() < end) { + for_each_cpu(cpu, cpumask) + if (pcpu_stopped(pcpu_devices + cpu)) + cpumask_clear_cpu(cpu, cpumask); + if (cpumask_empty(cpumask)) + break; + cpu_relax(); + } } /* - * this function sends a 'stop' sigp to all other CPUs in the system. - * it goes straight through. + * Stop all cpus but the current one. */ void smp_send_stop(void) { - /* Disable all interrupts/machine checks */ - __load_psw_mask(psw_kernel_bits & ~PSW_MASK_MCHECK); + cpumask_t cpumask; + int cpu; - /* write magic number to zero page (absolute 0) */ - lowcore_ptr[smp_processor_id()]->panic_magic = __PANIC_MAGIC; + /* Disable all interrupts/machine checks */ + __load_psw_mask(PSW_KERNEL_BITS | PSW_MASK_DAT); + trace_hardirqs_off(); - /* stop other processors. */ - do_send_stop(); + debug_set_critical(); + cpumask_copy(&cpumask, cpu_online_mask); + cpumask_clear_cpu(smp_processor_id(), &cpumask); - /* wait until other processors are stopped */ - do_wait_for_stop(); + if (oops_in_progress) + smp_emergency_stop(&cpumask); - /* store status of other processors. */ - do_store_status(); + /* stop all processors */ + for_each_cpu(cpu, &cpumask) { + struct pcpu *pcpu = pcpu_devices + cpu; + pcpu_sigp_retry(pcpu, SIGP_STOP, 0); + while (!pcpu_stopped(pcpu)) + cpu_relax(); + } } /* - * Reboot, halt and power_off routines for SMP. + * This is the main routine where commands issued by other + * cpus are handled. */ -void machine_restart_smp(char *__unused) +static void smp_handle_ext_call(void) { - smp_send_stop(); - do_reipl(); -} + unsigned long bits; -void machine_halt_smp(void) -{ - smp_send_stop(); - if (MACHINE_IS_VM && strlen(vmhalt_cmd) > 0) - __cpcmd(vmhalt_cmd, NULL, 0, NULL); - signal_processor(smp_processor_id(), sigp_stop_and_store_status); - for (;;); + /* handle bit signal external calls */ + bits = xchg(&pcpu_devices[smp_processor_id()].ec_mask, 0); + if (test_bit(ec_stop_cpu, &bits)) + smp_stop_cpu(); + if (test_bit(ec_schedule, &bits)) + scheduler_ipi(); + if (test_bit(ec_call_function_single, &bits)) + generic_smp_call_function_single_interrupt(); } -void machine_power_off_smp(void) +static void do_ext_call_interrupt(struct ext_code ext_code, + unsigned int param32, unsigned long param64) { - smp_send_stop(); - if (MACHINE_IS_VM && strlen(vmpoff_cmd) > 0) - __cpcmd(vmpoff_cmd, NULL, 0, NULL); - signal_processor(smp_processor_id(), sigp_stop_and_store_status); - for (;;); + inc_irq_stat(ext_code.code == 0x1202 ? IRQEXT_EXC : IRQEXT_EMS); + smp_handle_ext_call(); } -/* - * This is the main routine where commands issued by other - * cpus are handled. - */ - -static void do_ext_call_interrupt(__u16 code) +void arch_send_call_function_ipi_mask(const struct cpumask *mask) { - unsigned long bits; - - /* - * handle bit signal external calls - * - * For the ec_schedule signal we have to do nothing. All the work - * is done automatically when we return from the interrupt. - */ - bits = xchg(&S390_lowcore.ext_call_fast, 0); + int cpu; - if (test_bit(ec_call_function, &bits)) - do_call_function(); + for_each_cpu(cpu, mask) + pcpu_ec_call(pcpu_devices + cpu, ec_call_function_single); } -/* - * Send an external call sigp to another cpu and return without waiting - * for its completion. - */ -static void smp_ext_bitcall(int cpu, ec_bit_sig sig) +void arch_send_call_function_single_ipi(int cpu) { - /* - * Set signaling bit in lowcore of target cpu and kick it - */ - set_bit(sig, (unsigned long *) &lowcore_ptr[cpu]->ext_call_fast); - while (signal_processor(cpu, sigp_emergency_signal) == sigp_busy) - udelay(10); + pcpu_ec_call(pcpu_devices + cpu, ec_call_function_single); } #ifndef CONFIG_64BIT /* * this function sends a 'purge tlb' signal to another CPU. */ -void smp_ptlb_callback(void *info) +static void smp_ptlb_callback(void *info) { - local_flush_tlb(); + __tlb_flush_local(); } void smp_ptlb_all(void) { - on_each_cpu(smp_ptlb_callback, NULL, 0, 1); + on_each_cpu(smp_ptlb_callback, NULL, 1); } EXPORT_SYMBOL(smp_ptlb_all); #endif /* ! CONFIG_64BIT */ @@ -342,15 +465,16 @@ EXPORT_SYMBOL(smp_ptlb_all); */ void smp_send_reschedule(int cpu) { - smp_ext_bitcall(cpu, ec_schedule); + pcpu_ec_call(pcpu_devices + cpu, ec_schedule); } /* * parameter area for the set/clear control bit callbacks */ struct ec_creg_mask_parms { - unsigned long orvals[16]; - unsigned long andvals[16]; + unsigned long orval; + unsigned long andval; + int cr; }; /* @@ -360,11 +484,9 @@ static void smp_ctl_bit_callback(void *info) { struct ec_creg_mask_parms *pp = info; unsigned long cregs[16]; - int i; __ctl_store(cregs, 0, 15); - for (i = 0; i <= 15; i++) - cregs[i] = (cregs[i] & pp->andvals[i]) | pp->orvals[i]; + cregs[pp->cr] = (cregs[pp->cr] & pp->andval) | pp->orval; __ctl_load(cregs, 0, 15); } @@ -373,12 +495,9 @@ static void smp_ctl_bit_callback(void *info) */ void smp_ctl_set_bit(int cr, int bit) { - struct ec_creg_mask_parms parms; + struct ec_creg_mask_parms parms = { 1UL << bit, -1UL, cr }; - memset(&parms.orvals, 0, sizeof(parms.orvals)); - memset(&parms.andvals, 0xff, sizeof(parms.andvals)); - parms.orvals[cr] = 1 << bit; - on_each_cpu(smp_ctl_bit_callback, &parms, 0, 1); + on_each_cpu(smp_ctl_bit_callback, &parms, 1); } EXPORT_SYMBOL(smp_ctl_set_bit); @@ -387,348 +506,301 @@ EXPORT_SYMBOL(smp_ctl_set_bit); */ void smp_ctl_clear_bit(int cr, int bit) { - struct ec_creg_mask_parms parms; + struct ec_creg_mask_parms parms = { 0, ~(1UL << bit), cr }; - memset(&parms.orvals, 0, sizeof(parms.orvals)); - memset(&parms.andvals, 0xff, sizeof(parms.andvals)); - parms.andvals[cr] = ~(1L << bit); - on_each_cpu(smp_ctl_bit_callback, &parms, 0, 1); + on_each_cpu(smp_ctl_bit_callback, &parms, 1); } EXPORT_SYMBOL(smp_ctl_clear_bit); -#if defined(CONFIG_ZFCPDUMP) || defined(CONFIG_ZFCPDUMP_MODULE) - -/* - * zfcpdump_prefix_array holds prefix registers for the following scenario: - * 64 bit zfcpdump kernel and 31 bit kernel which is to be dumped. We have to - * save its prefix registers, since they get lost, when switching from 31 bit - * to 64 bit. - */ -unsigned int zfcpdump_prefix_array[NR_CPUS + 1] \ - __attribute__((__section__(".data"))); +#ifdef CONFIG_CRASH_DUMP -static void __init smp_get_save_area(unsigned int cpu, unsigned int phy_cpu) +static void __init smp_get_save_area(int cpu, u16 address) { - if (ipl_info.type != IPL_TYPE_FCP_DUMP) + void *lc = pcpu_devices[0].lowcore; + struct save_area *save_area; + + if (is_kdump_kernel()) + return; + if (!OLDMEM_BASE && (address == boot_cpu_address || + ipl_info.type != IPL_TYPE_FCP_DUMP)) return; - if (cpu >= NR_CPUS) { - printk(KERN_WARNING "Registers for cpu %i not saved since dump " - "kernel was compiled with NR_CPUS=%i\n", cpu, NR_CPUS); + save_area = dump_save_area_create(cpu); + if (!save_area) + panic("could not allocate memory for save area\n"); + if (address == boot_cpu_address) { + /* Copy the registers of the boot cpu. */ + copy_oldmem_page(1, (void *) save_area, sizeof(*save_area), + SAVE_AREA_BASE - PAGE_SIZE, 0); return; } - zfcpdump_save_areas[cpu] = alloc_bootmem(sizeof(union save_area)); - __cpu_logical_map[1] = (__u16) phy_cpu; - while (signal_processor(1, sigp_stop_and_store_status) == sigp_busy) - cpu_relax(); - memcpy(zfcpdump_save_areas[cpu], - (void *)(unsigned long) store_prefix() + SAVE_AREA_BASE, - SAVE_AREA_SIZE); -#ifdef CONFIG_64BIT - /* copy original prefix register */ - zfcpdump_save_areas[cpu]->s390x.pref_reg = zfcpdump_prefix_array[cpu]; -#endif + /* Get the registers of a non-boot cpu. */ + __pcpu_sigp_relax(address, SIGP_STOP_AND_STORE_STATUS, 0, NULL); + memcpy_real(save_area, lc + SAVE_AREA_BASE, sizeof(*save_area)); } -union save_area *zfcpdump_save_areas[NR_CPUS + 1]; -EXPORT_SYMBOL_GPL(zfcpdump_save_areas); +int smp_store_status(int cpu) +{ + struct pcpu *pcpu; -#else + pcpu = pcpu_devices + cpu; + if (__pcpu_sigp_relax(pcpu->address, SIGP_STOP_AND_STORE_STATUS, + 0, NULL) != SIGP_CC_ORDER_CODE_ACCEPTED) + return -EIO; + return 0; +} -static inline void smp_get_save_area(unsigned int cpu, unsigned int phy_cpu) { } +#else /* CONFIG_CRASH_DUMP */ -#endif /* CONFIG_ZFCPDUMP || CONFIG_ZFCPDUMP_MODULE */ +static inline void smp_get_save_area(int cpu, u16 address) { } -/* - * Lets check how many CPUs we have. - */ -static unsigned int __init smp_count_cpus(void) -{ - unsigned int cpu, num_cpus; - __u16 boot_cpu_addr; - - /* - * cpu 0 is the boot cpu. See smp_prepare_boot_cpu. - */ - boot_cpu_addr = S390_lowcore.cpu_data.cpu_addr; - current_thread_info()->cpu = 0; - num_cpus = 1; - for (cpu = 0; cpu <= 65535; cpu++) { - if ((__u16) cpu == boot_cpu_addr) +#endif /* CONFIG_CRASH_DUMP */ + +void smp_cpu_set_polarization(int cpu, int val) +{ + pcpu_devices[cpu].polarization = val; +} + +int smp_cpu_get_polarization(int cpu) +{ + return pcpu_devices[cpu].polarization; +} + +static struct sclp_cpu_info *smp_get_cpu_info(void) +{ + static int use_sigp_detection; + struct sclp_cpu_info *info; + int address; + + info = kzalloc(sizeof(*info), GFP_KERNEL); + if (info && (use_sigp_detection || sclp_get_cpu_info(info))) { + use_sigp_detection = 1; + for (address = 0; address <= MAX_CPU_ADDRESS; address++) { + if (__pcpu_sigp_relax(address, SIGP_SENSE, 0, NULL) == + SIGP_CC_NOT_OPERATIONAL) + continue; + info->cpu[info->configured].address = address; + info->configured++; + } + info->combined = info->configured; + } + return info; +} + +static int smp_add_present_cpu(int cpu); + +static int __smp_rescan_cpus(struct sclp_cpu_info *info, int sysfs_add) +{ + struct pcpu *pcpu; + cpumask_t avail; + int cpu, nr, i; + + nr = 0; + cpumask_xor(&avail, cpu_possible_mask, cpu_present_mask); + cpu = cpumask_first(&avail); + for (i = 0; (i < info->combined) && (cpu < nr_cpu_ids); i++) { + if (info->has_cpu_type && info->cpu[i].type != boot_cpu_type) + continue; + if (pcpu_find_address(cpu_present_mask, info->cpu[i].address)) continue; - __cpu_logical_map[1] = (__u16) cpu; - if (signal_processor(1, sigp_sense) == sigp_not_operational) + pcpu = pcpu_devices + cpu; + pcpu->address = info->cpu[i].address; + pcpu->state = (i >= info->configured) ? + CPU_STATE_STANDBY : CPU_STATE_CONFIGURED; + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + set_cpu_present(cpu, true); + if (sysfs_add && smp_add_present_cpu(cpu) != 0) + set_cpu_present(cpu, false); + else + nr++; + cpu = cpumask_next(cpu, &avail); + } + return nr; +} + +static void __init smp_detect_cpus(void) +{ + unsigned int cpu, c_cpus, s_cpus; + struct sclp_cpu_info *info; + + info = smp_get_cpu_info(); + if (!info) + panic("smp_detect_cpus failed to allocate memory\n"); + if (info->has_cpu_type) { + for (cpu = 0; cpu < info->combined; cpu++) { + if (info->cpu[cpu].address != boot_cpu_address) + continue; + /* The boot cpu dictates the cpu type. */ + boot_cpu_type = info->cpu[cpu].type; + break; + } + } + c_cpus = s_cpus = 0; + for (cpu = 0; cpu < info->combined; cpu++) { + if (info->has_cpu_type && info->cpu[cpu].type != boot_cpu_type) continue; - smp_get_save_area(num_cpus, cpu); - num_cpus++; + if (cpu < info->configured) { + smp_get_save_area(c_cpus, info->cpu[cpu].address); + c_cpus++; + } else + s_cpus++; } - printk("Detected %d CPU's\n", (int) num_cpus); - printk("Boot cpu address %2X\n", boot_cpu_addr); - return num_cpus; + pr_info("%d configured CPUs, %d standby CPUs\n", c_cpus, s_cpus); + get_online_cpus(); + __smp_rescan_cpus(info, 0); + put_online_cpus(); + kfree(info); } /* * Activate a secondary processor. */ -int __cpuinit start_secondary(void *cpuvoid) +static void smp_start_secondary(void *cpuvoid) { - /* Setup the cpu */ + S390_lowcore.last_update_clock = get_tod_clock(); + S390_lowcore.restart_stack = (unsigned long) restart_stack; + S390_lowcore.restart_fn = (unsigned long) do_restart; + S390_lowcore.restart_data = 0; + S390_lowcore.restart_source = -1UL; + restore_access_regs(S390_lowcore.access_regs_save_area); + __ctl_load(S390_lowcore.cregs_save_area, 0, 15); + __load_psw_mask(PSW_KERNEL_BITS | PSW_MASK_DAT); cpu_init(); preempt_disable(); - /* Enable TOD clock interrupts on the secondary cpu. */ init_cpu_timer(); -#ifdef CONFIG_VIRT_TIMER - /* Enable cpu timer interrupts on the secondary cpu. */ init_cpu_vtimer(); -#endif - /* Enable pfault pseudo page faults on this cpu. */ pfault_init(); - - /* Mark this cpu as online */ - cpu_set(smp_processor_id(), cpu_online_map); - /* Switch on interrupts */ + notify_cpu_starting(smp_processor_id()); + set_cpu_online(smp_processor_id(), true); + inc_irq_stat(CPU_RST); local_irq_enable(); - /* Print info about this processor */ - print_cpu_info(&S390_lowcore.cpu_data); - /* cpu_idle will call schedule for us */ - cpu_idle(); - return 0; -} - -static void __init smp_create_idle(unsigned int cpu) -{ - struct task_struct *p; - - /* - * don't care about the psw and regs settings since we'll never - * reschedule the forked task. - */ - p = fork_idle(cpu); - if (IS_ERR(p)) - panic("failed fork for CPU %u: %li", cpu, PTR_ERR(p)); - current_set[cpu] = p; -} - -static int cpu_stopped(int cpu) -{ - __u32 status; - - /* Check for stopped state */ - if (signal_processor_ps(&status, 0, cpu, sigp_sense) == - sigp_status_stored) { - if (status & 0x40) - return 1; - } - return 0; + cpu_startup_entry(CPUHP_ONLINE); } /* Upping and downing of CPUs */ - -int __cpu_up(unsigned int cpu) +int __cpu_up(unsigned int cpu, struct task_struct *tidle) { - struct task_struct *idle; - struct _lowcore *cpu_lowcore; - struct stack_frame *sf; - sigp_ccode ccode; - int curr_cpu; - - for (curr_cpu = 0; curr_cpu <= 65535; curr_cpu++) { - __cpu_logical_map[cpu] = (__u16) curr_cpu; - if (cpu_stopped(cpu)) - break; - } - - if (!cpu_stopped(cpu)) - return -ENODEV; + struct pcpu *pcpu; + int rc; - ccode = signal_processor_p((__u32)(unsigned long)(lowcore_ptr[cpu]), - cpu, sigp_set_prefix); - if (ccode) { - printk("sigp_set_prefix failed for cpu %d " - "with condition code %d\n", - (int) cpu, (int) ccode); + pcpu = pcpu_devices + cpu; + if (pcpu->state != CPU_STATE_CONFIGURED) + return -EIO; + if (pcpu_sigp_retry(pcpu, SIGP_INITIAL_CPU_RESET, 0) != + SIGP_CC_ORDER_CODE_ACCEPTED) return -EIO; - } - - idle = current_set[cpu]; - cpu_lowcore = lowcore_ptr[cpu]; - cpu_lowcore->kernel_stack = (unsigned long) - task_stack_page(idle) + THREAD_SIZE; - sf = (struct stack_frame *) (cpu_lowcore->kernel_stack - - sizeof(struct pt_regs) - - sizeof(struct stack_frame)); - memset(sf, 0, sizeof(struct stack_frame)); - sf->gprs[9] = (unsigned long) sf; - cpu_lowcore->save_area[15] = (unsigned long) sf; - __ctl_store(cpu_lowcore->cregs_save_area[0], 0, 15); - asm volatile( - " stam 0,15,0(%0)" - : : "a" (&cpu_lowcore->access_regs_save_area) : "memory"); - cpu_lowcore->percpu_offset = __per_cpu_offset[cpu]; - cpu_lowcore->current_task = (unsigned long) idle; - cpu_lowcore->cpu_data.cpu_nr = cpu; - eieio(); - - while (signal_processor(cpu, sigp_restart) == sigp_busy) - udelay(10); + rc = pcpu_alloc_lowcore(pcpu, cpu); + if (rc) + return rc; + pcpu_prepare_secondary(pcpu, cpu); + pcpu_attach_task(pcpu, tidle); + pcpu_start_fn(pcpu, smp_start_secondary, NULL); while (!cpu_online(cpu)) cpu_relax(); return 0; } -static unsigned int __initdata additional_cpus; -static unsigned int __initdata possible_cpus; - -void __init smp_setup_cpu_possible_map(void) -{ - unsigned int phy_cpus, pos_cpus, cpu; - - phy_cpus = smp_count_cpus(); - pos_cpus = min(phy_cpus + additional_cpus, (unsigned int) NR_CPUS); - - if (possible_cpus) - pos_cpus = min(possible_cpus, (unsigned int) NR_CPUS); - - for (cpu = 0; cpu < pos_cpus; cpu++) - cpu_set(cpu, cpu_possible_map); - - phy_cpus = min(phy_cpus, pos_cpus); - - for (cpu = 0; cpu < phy_cpus; cpu++) - cpu_set(cpu, cpu_present_map); -} - -#ifdef CONFIG_HOTPLUG_CPU +static unsigned int setup_possible_cpus __initdata; -static int __init setup_additional_cpus(char *s) +static int __init _setup_possible_cpus(char *s) { - additional_cpus = simple_strtoul(s, NULL, 0); + get_option(&s, &setup_possible_cpus); return 0; } -early_param("additional_cpus", setup_additional_cpus); +early_param("possible_cpus", _setup_possible_cpus); -static int __init setup_possible_cpus(char *s) -{ - possible_cpus = simple_strtoul(s, NULL, 0); - return 0; -} -early_param("possible_cpus", setup_possible_cpus); +#ifdef CONFIG_HOTPLUG_CPU int __cpu_disable(void) { - struct ec_creg_mask_parms cr_parms; - int cpu = smp_processor_id(); - - cpu_clear(cpu, cpu_online_map); + unsigned long cregs[16]; - /* Disable pfault pseudo page faults on this cpu. */ + /* Handle possible pending IPIs */ + smp_handle_ext_call(); + set_cpu_online(smp_processor_id(), false); + /* Disable pseudo page faults on this cpu. */ pfault_fini(); - - memset(&cr_parms.orvals, 0, sizeof(cr_parms.orvals)); - memset(&cr_parms.andvals, 0xff, sizeof(cr_parms.andvals)); - - /* disable all external interrupts */ - cr_parms.orvals[0] = 0; - cr_parms.andvals[0] = ~(1 << 15 | 1 << 14 | 1 << 13 | 1 << 12 | - 1 << 11 | 1 << 10 | 1 << 6 | 1 << 4); - /* disable all I/O interrupts */ - cr_parms.orvals[6] = 0; - cr_parms.andvals[6] = ~(1 << 31 | 1 << 30 | 1 << 29 | 1 << 28 | - 1 << 27 | 1 << 26 | 1 << 25 | 1 << 24); - /* disable most machine checks */ - cr_parms.orvals[14] = 0; - cr_parms.andvals[14] = ~(1 << 28 | 1 << 27 | 1 << 26 | - 1 << 25 | 1 << 24); - - smp_ctl_bit_callback(&cr_parms); - + /* Disable interrupt sources via control register. */ + __ctl_store(cregs, 0, 15); + cregs[0] &= ~0x0000ee70UL; /* disable all external interrupts */ + cregs[6] &= ~0xff000000UL; /* disable all I/O interrupts */ + cregs[14] &= ~0x1f000000UL; /* disable most machine checks */ + __ctl_load(cregs, 0, 15); return 0; } void __cpu_die(unsigned int cpu) { + struct pcpu *pcpu; + /* Wait until target cpu is down */ - while (!smp_cpu_not_running(cpu)) + pcpu = pcpu_devices + cpu; + while (!pcpu_stopped(pcpu)) cpu_relax(); - printk("Processor %d spun down\n", cpu); + pcpu_free_lowcore(pcpu); + atomic_dec(&init_mm.context.attach_count); + cpumask_clear_cpu(cpu, mm_cpumask(&init_mm)); + if (MACHINE_HAS_TLB_LC) + cpumask_clear_cpu(cpu, &init_mm.context.cpu_attach_mask); } -void cpu_die(void) +void __noreturn cpu_die(void) { idle_task_exit(); - signal_processor(smp_processor_id(), sigp_stop); - BUG(); - for (;;); + pcpu_sigp_retry(pcpu_devices + smp_processor_id(), SIGP_STOP, 0); + for (;;) ; } #endif /* CONFIG_HOTPLUG_CPU */ -/* - * Cycle through the processors and setup structures. - */ +void __init smp_fill_possible_mask(void) +{ + unsigned int possible, sclp, cpu; + + sclp = sclp_get_max_cpu() ?: nr_cpu_ids; + possible = setup_possible_cpus ?: nr_cpu_ids; + possible = min(possible, sclp); + for (cpu = 0; cpu < possible && cpu < nr_cpu_ids; cpu++) + set_cpu_possible(cpu, true); +} void __init smp_prepare_cpus(unsigned int max_cpus) { - unsigned long stack; - unsigned int cpu; - int i; - /* request the 0x1201 emergency signal external interrupt */ - if (register_external_interrupt(0x1201, do_ext_call_interrupt) != 0) + if (register_external_irq(EXT_IRQ_EMERGENCY_SIG, do_ext_call_interrupt)) panic("Couldn't request external interrupt 0x1201"); - memset(lowcore_ptr, 0, sizeof(lowcore_ptr)); - /* - * Initialize prefix pages and stacks for all possible cpus - */ - print_cpu_info(&S390_lowcore.cpu_data); - - for_each_possible_cpu(i) { - lowcore_ptr[i] = (struct _lowcore *) - __get_free_pages(GFP_KERNEL | GFP_DMA, - sizeof(void*) == 8 ? 1 : 0); - stack = __get_free_pages(GFP_KERNEL, ASYNC_ORDER); - if (!lowcore_ptr[i] || !stack) - panic("smp_boot_cpus failed to allocate memory\n"); - - *(lowcore_ptr[i]) = S390_lowcore; - lowcore_ptr[i]->async_stack = stack + ASYNC_SIZE; - stack = __get_free_pages(GFP_KERNEL, 0); - if (!stack) - panic("smp_boot_cpus failed to allocate memory\n"); - lowcore_ptr[i]->panic_stack = stack + PAGE_SIZE; -#ifndef CONFIG_64BIT - if (MACHINE_HAS_IEEE) { - lowcore_ptr[i]->extended_save_area_addr = - (__u32) __get_free_pages(GFP_KERNEL, 0); - if (!lowcore_ptr[i]->extended_save_area_addr) - panic("smp_boot_cpus failed to " - "allocate memory\n"); - } -#endif - } -#ifndef CONFIG_64BIT - if (MACHINE_HAS_IEEE) - ctl_set_bit(14, 29); /* enable extended save area */ -#endif - set_prefix((u32)(unsigned long) lowcore_ptr[smp_processor_id()]); - - for_each_possible_cpu(cpu) - if (cpu != smp_processor_id()) - smp_create_idle(cpu); + /* request the 0x1202 external call external interrupt */ + if (register_external_irq(EXT_IRQ_EXTERNAL_CALL, do_ext_call_interrupt)) + panic("Couldn't request external interrupt 0x1202"); + smp_detect_cpus(); } void __init smp_prepare_boot_cpu(void) { - BUG_ON(smp_processor_id() != 0); - - cpu_set(0, cpu_online_map); + struct pcpu *pcpu = pcpu_devices; + + boot_cpu_address = stap(); + pcpu->state = CPU_STATE_CONFIGURED; + pcpu->address = boot_cpu_address; + pcpu->lowcore = (struct _lowcore *)(unsigned long) store_prefix(); + pcpu->async_stack = S390_lowcore.async_stack - ASYNC_SIZE + + STACK_FRAME_OVERHEAD + sizeof(struct pt_regs); + pcpu->panic_stack = S390_lowcore.panic_stack - PAGE_SIZE + + STACK_FRAME_OVERHEAD + sizeof(struct pt_regs); S390_lowcore.percpu_offset = __per_cpu_offset[0]; - current_set[0] = current; + smp_cpu_set_polarization(0, POLARIZATION_UNKNOWN); + set_cpu_present(0, true); + set_cpu_online(0, true); } void __init smp_cpus_done(unsigned int max_cpus) { - cpu_present_map = cpu_possible_map; +} + +void __init smp_setup_processor_id(void) +{ + S390_lowcore.cpu_nr = 0; + S390_lowcore.spinlock_lockval = arch_spin_lockval(0); } /* @@ -742,62 +814,250 @@ int setup_profiling_timer(unsigned int multiplier) return 0; } -static DEFINE_PER_CPU(struct cpu, cpu_devices); +#ifdef CONFIG_HOTPLUG_CPU +static ssize_t cpu_configure_show(struct device *dev, + struct device_attribute *attr, char *buf) +{ + ssize_t count; + + mutex_lock(&smp_cpu_state_mutex); + count = sprintf(buf, "%d\n", pcpu_devices[dev->id].state); + mutex_unlock(&smp_cpu_state_mutex); + return count; +} -static ssize_t show_capability(struct sys_device *dev, char *buf) +static ssize_t cpu_configure_store(struct device *dev, + struct device_attribute *attr, + const char *buf, size_t count) { - unsigned int capability; - int rc; + struct pcpu *pcpu; + int cpu, val, rc; + char delim; + + if (sscanf(buf, "%d %c", &val, &delim) != 1) + return -EINVAL; + if (val != 0 && val != 1) + return -EINVAL; + get_online_cpus(); + mutex_lock(&smp_cpu_state_mutex); + rc = -EBUSY; + /* disallow configuration changes of online cpus and cpu 0 */ + cpu = dev->id; + if (cpu_online(cpu) || cpu == 0) + goto out; + pcpu = pcpu_devices + cpu; + rc = 0; + switch (val) { + case 0: + if (pcpu->state != CPU_STATE_CONFIGURED) + break; + rc = sclp_cpu_deconfigure(pcpu->address); + if (rc) + break; + pcpu->state = CPU_STATE_STANDBY; + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + topology_expect_change(); + break; + case 1: + if (pcpu->state != CPU_STATE_STANDBY) + break; + rc = sclp_cpu_configure(pcpu->address); + if (rc) + break; + pcpu->state = CPU_STATE_CONFIGURED; + smp_cpu_set_polarization(cpu, POLARIZATION_UNKNOWN); + topology_expect_change(); + break; + default: + break; + } +out: + mutex_unlock(&smp_cpu_state_mutex); + put_online_cpus(); + return rc ? rc : count; +} +static DEVICE_ATTR(configure, 0644, cpu_configure_show, cpu_configure_store); +#endif /* CONFIG_HOTPLUG_CPU */ - rc = get_cpu_capability(&capability); - if (rc) - return rc; - return sprintf(buf, "%u\n", capability); +static ssize_t show_cpu_address(struct device *dev, + struct device_attribute *attr, char *buf) +{ + return sprintf(buf, "%d\n", pcpu_devices[dev->id].address); +} +static DEVICE_ATTR(address, 0444, show_cpu_address, NULL); + +static struct attribute *cpu_common_attrs[] = { +#ifdef CONFIG_HOTPLUG_CPU + &dev_attr_configure.attr, +#endif + &dev_attr_address.attr, + NULL, +}; + +static struct attribute_group cpu_common_attr_group = { + .attrs = cpu_common_attrs, +}; + +static ssize_t show_idle_count(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct s390_idle_data *idle = &per_cpu(s390_idle, dev->id); + unsigned long long idle_count; + unsigned int sequence; + + do { + sequence = ACCESS_ONCE(idle->sequence); + idle_count = ACCESS_ONCE(idle->idle_count); + if (ACCESS_ONCE(idle->clock_idle_enter)) + idle_count++; + } while ((sequence & 1) || (ACCESS_ONCE(idle->sequence) != sequence)); + return sprintf(buf, "%llu\n", idle_count); } -static SYSDEV_ATTR(capability, 0444, show_capability, NULL); +static DEVICE_ATTR(idle_count, 0444, show_idle_count, NULL); + +static ssize_t show_idle_time(struct device *dev, + struct device_attribute *attr, char *buf) +{ + struct s390_idle_data *idle = &per_cpu(s390_idle, dev->id); + unsigned long long now, idle_time, idle_enter, idle_exit; + unsigned int sequence; + + do { + now = get_tod_clock(); + sequence = ACCESS_ONCE(idle->sequence); + idle_time = ACCESS_ONCE(idle->idle_time); + idle_enter = ACCESS_ONCE(idle->clock_idle_enter); + idle_exit = ACCESS_ONCE(idle->clock_idle_exit); + } while ((sequence & 1) || (ACCESS_ONCE(idle->sequence) != sequence)); + idle_time += idle_enter ? ((idle_exit ? : now) - idle_enter) : 0; + return sprintf(buf, "%llu\n", idle_time >> 12); +} +static DEVICE_ATTR(idle_time_us, 0444, show_idle_time, NULL); + +static struct attribute *cpu_online_attrs[] = { + &dev_attr_idle_count.attr, + &dev_attr_idle_time_us.attr, + NULL, +}; + +static struct attribute_group cpu_online_attr_group = { + .attrs = cpu_online_attrs, +}; -static int __cpuinit smp_cpu_notify(struct notifier_block *self, - unsigned long action, void *hcpu) +static int smp_cpu_notify(struct notifier_block *self, unsigned long action, + void *hcpu) { unsigned int cpu = (unsigned int)(long)hcpu; - struct cpu *c = &per_cpu(cpu_devices, cpu); - struct sys_device *s = &c->sysdev; + struct cpu *c = pcpu_devices[cpu].cpu; + struct device *s = &c->dev; + int err = 0; - switch (action) { + switch (action & ~CPU_TASKS_FROZEN) { case CPU_ONLINE: - case CPU_ONLINE_FROZEN: - if (sysdev_create_file(s, &attr_capability)) - return NOTIFY_BAD; + err = sysfs_create_group(&s->kobj, &cpu_online_attr_group); break; case CPU_DEAD: - case CPU_DEAD_FROZEN: - sysdev_remove_file(s, &attr_capability); + sysfs_remove_group(&s->kobj, &cpu_online_attr_group); break; } - return NOTIFY_OK; + return notifier_from_errno(err); } -static struct notifier_block __cpuinitdata smp_cpu_nb = { - .notifier_call = smp_cpu_notify, -}; +static int smp_add_present_cpu(int cpu) +{ + struct device *s; + struct cpu *c; + int rc; + + c = kzalloc(sizeof(*c), GFP_KERNEL); + if (!c) + return -ENOMEM; + pcpu_devices[cpu].cpu = c; + s = &c->dev; + c->hotpluggable = 1; + rc = register_cpu(c, cpu); + if (rc) + goto out; + rc = sysfs_create_group(&s->kobj, &cpu_common_attr_group); + if (rc) + goto out_cpu; + if (cpu_online(cpu)) { + rc = sysfs_create_group(&s->kobj, &cpu_online_attr_group); + if (rc) + goto out_online; + } + rc = topology_cpu_init(c); + if (rc) + goto out_topology; + return 0; + +out_topology: + if (cpu_online(cpu)) + sysfs_remove_group(&s->kobj, &cpu_online_attr_group); +out_online: + sysfs_remove_group(&s->kobj, &cpu_common_attr_group); +out_cpu: +#ifdef CONFIG_HOTPLUG_CPU + unregister_cpu(c); +#endif +out: + return rc; +} + +#ifdef CONFIG_HOTPLUG_CPU -static int __init topology_init(void) +int __ref smp_rescan_cpus(void) { - int cpu; + struct sclp_cpu_info *info; + int nr; + + info = smp_get_cpu_info(); + if (!info) + return -ENOMEM; + get_online_cpus(); + mutex_lock(&smp_cpu_state_mutex); + nr = __smp_rescan_cpus(info, 1); + mutex_unlock(&smp_cpu_state_mutex); + put_online_cpus(); + kfree(info); + if (nr) + topology_schedule_update(); + return 0; +} - register_cpu_notifier(&smp_cpu_nb); +static ssize_t __ref rescan_store(struct device *dev, + struct device_attribute *attr, + const char *buf, + size_t count) +{ + int rc; - for_each_possible_cpu(cpu) { - struct cpu *c = &per_cpu(cpu_devices, cpu); - struct sys_device *s = &c->sysdev; + rc = smp_rescan_cpus(); + return rc ? rc : count; +} +static DEVICE_ATTR(rescan, 0200, NULL, rescan_store); +#endif /* CONFIG_HOTPLUG_CPU */ - c->hotpluggable = 1; - register_cpu(c, cpu); - if (!cpu_online(cpu)) - continue; - s = &c->sysdev; - sysdev_create_file(s, &attr_capability); +static int __init s390_smp_init(void) +{ + int cpu, rc = 0; + +#ifdef CONFIG_HOTPLUG_CPU + rc = device_create_file(cpu_subsys.dev_root, &dev_attr_rescan); + if (rc) + return rc; +#endif + cpu_notifier_register_begin(); + for_each_present_cpu(cpu) { + rc = smp_add_present_cpu(cpu); + if (rc) + goto out; } - return 0; + + __hotcpu_notifier(smp_cpu_notify, 0); + +out: + cpu_notifier_register_done(); + return rc; } -subsys_initcall(topology_init); +subsys_initcall(s390_smp_init); |