static int __init pnv_init_idle_states(void)
{
supported_cpuidle_states = 0;
if (cpuidle_disable != IDLE_NO_OVERRIDE)
goto out;
pnv_probe_idle_states();
if (!(supported_cpuidle_states & OPAL_PM_SLEEP_ENABLED_ER1)) {
patch_instruction(
(unsigned int *)pnv_fastsleep_workaround_at_entry,
PPC_INST_NOP);
patch_instruction(
(unsigned int *)pnv_fastsleep_workaround_at_exit,
PPC_INST_NOP);
} else {
/*
* OPAL_PM_SLEEP_ENABLED_ER1 is set. It indicates that
* workaround is needed to use fastsleep. Provide sysfs
* control to choose how this workaround has to be applied.
*/
device_create_file(cpu_subsys.dev_root,
&dev_attr_fastsleep_workaround_applyonce);
}
pnv_alloc_idle_core_states();
/*
* For each CPU, record its PACA address in each of it's
* sibling thread's PACA at the slot corresponding to this
* CPU's index in the core.
*/
if (cpu_has_feature(CPU_FTR_POWER9_DD1)) {
int cpu;
pr_info("powernv: idle: Saving PACA pointers of all CPUs in their thread sibling PACA\n");
for_each_possible_cpu(cpu) {
int base_cpu = cpu_first_thread_sibling(cpu);
int idx = cpu_thread_in_core(cpu);
int i;
for (i = 0; i < threads_per_core; i++) {
int j = base_cpu + i;
paca[j].thread_sibling_pacas[idx] = &paca[cpu];
}
}
}
if (supported_cpuidle_states & OPAL_PM_NAP_ENABLED)
ppc_md.power_save = power7_idle;
out:
return 0;
}
static unsigned int steal_context_smp(unsigned int id)
{
struct mm_struct *mm;
unsigned int cpu, max, i;
max = last_context - first_context;
/* Attempt to free next_context first and then loop until we manage */
while (max--) {
/* Pick up the victim mm */
mm = context_mm[id];
/* We have a candidate victim, check if it's active, on SMP
* we cannot steal active contexts
*/
if (mm->context.active) {
id++;
if (id > last_context)
id = first_context;
continue;
}
pr_hardcont(" | steal %d from 0x%p", id, mm);
/* Mark this mm has having no context anymore */
mm->context.id = MMU_NO_CONTEXT;
/* Mark it stale on all CPUs that used this mm. For threaded
* implementations, we set it on all threads on each core
* represented in the mask. A future implementation will use
* a core map instead but this will do for now.
*/
for_each_cpu(cpu, mm_cpumask(mm)) {
for (i = cpu_first_thread_sibling(cpu);
i <= cpu_last_thread_sibling(cpu); i++) {
if (stale_map[i])
__set_bit(id, stale_map[i]);
}
cpu = i - 1;
}
return id;
}
/* This will happen if you have more CPUs than available contexts,
* all we can do here is wait a bit and try again
*/
raw_spin_unlock(&context_lock);
cpu_relax();
raw_spin_lock(&context_lock);
/* This will cause the caller to try again */
return MMU_NO_CONTEXT;
}
static void verify_cpu_node_mapping(int cpu, int node)
{
int base, sibling, i;
/* Verify that all the threads in the core belong to the same node */
base = cpu_first_thread_sibling(cpu);
for (i = 0; i < threads_per_core; i++) {
sibling = base + i;
if (sibling == cpu || cpu_is_offline(sibling))
continue;
if (cpu_to_node(sibling) != node) {
WARN(1, "CPU thread siblings %d and %d don't belong"
" to the same node!\n", cpu, sibling);
break;
}
}
}
int __cpu_disable(void)
{
struct device_node *l2_cache;
int cpu = smp_processor_id();
int base, i;
int err;
if (!smp_ops->cpu_disable)
return -ENOSYS;
err = smp_ops->cpu_disable();
if (err)
return err;
/* Update sibling maps */
base = cpu_first_thread_sibling(cpu);
for (i = 0; i < threads_per_core; i++) {
cpumask_clear_cpu(cpu, cpu_sibling_mask(base + i));
cpumask_clear_cpu(base + i, cpu_sibling_mask(cpu));
cpumask_clear_cpu(cpu, cpu_core_mask(base + i));
cpumask_clear_cpu(base + i, cpu_core_mask(cpu));
}
l2_cache = cpu_to_l2cache(cpu);
for_each_present_cpu(i) {
struct device_node *np = cpu_to_l2cache(i);
if (!np)
continue;
if (np == l2_cache) {
cpumask_clear_cpu(cpu, cpu_core_mask(i));
cpumask_clear_cpu(i, cpu_core_mask(cpu));
}
of_node_put(np);
}
of_node_put(l2_cache);
return 0;
}
static unsigned int steal_context_smp(unsigned int id)
{
struct mm_struct *mm;
unsigned int cpu, max, i;
max = last_context - first_context;
while (max--) {
mm = context_mm[id];
if (mm->context.active) {
id++;
if (id > last_context)
id = first_context;
continue;
}
pr_hardcont(" | steal %d from 0x%p", id, mm);
mm->context.id = MMU_NO_CONTEXT;
for_each_cpu(cpu, mm_cpumask(mm)) {
for (i = cpu_first_thread_sibling(cpu);
i <= cpu_last_thread_sibling(cpu); i++)
__set_bit(id, stale_map[i]);
cpu = i - 1;
}
return id;
}
raw_spin_unlock(&context_lock);
cpu_relax();
raw_spin_lock(&context_lock);
return MMU_NO_CONTEXT;
}
extern void radix_kvm_prefetch_workaround(struct mm_struct *mm)
{
unsigned int pid = mm->context.id;
if (unlikely(pid == MMU_NO_CONTEXT))
return;
/*
* If this context hasn't run on that CPU before and KVM is
* around, there's a slim chance that the guest on another
* CPU just brought in obsolete translation into the TLB of
* this CPU due to a bad prefetch using the guest PID on
* the way into the hypervisor.
*
* We work around this here. If KVM is possible, we check if
* any sibling thread is in KVM. If it is, the window may exist
* and thus we flush that PID from the core.
*
* A potential future improvement would be to mark which PIDs
* have never been used on the system and avoid it if the PID
* is new and the process has no other cpumask bit set.
*/
if (cpu_has_feature(CPU_FTR_HVMODE) && radix_enabled()) {
int cpu = smp_processor_id();
int sib = cpu_first_thread_sibling(cpu);
bool flush = false;
for (; sib <= cpu_last_thread_sibling(cpu) && !flush; sib++) {
if (sib == cpu)
continue;
if (paca[sib].kvm_hstate.kvm_vcpu)
flush = true;
}
if (flush)
_tlbiel_pid(pid, RIC_FLUSH_ALL);
}
}
void switch_mmu_context(struct mm_struct *prev, struct mm_struct *next,
struct task_struct *tsk)
{
unsigned int i, id, cpu = smp_processor_id();
unsigned long *map;
/* No lockless fast path .. yet */
raw_spin_lock(&context_lock);
pr_hard("[%d] activating context for mm @%p, active=%d, id=%d",
cpu, next, next->context.active, next->context.id);
#ifdef CONFIG_SMP
/* Mark us active and the previous one not anymore */
next->context.active++;
if (prev) {
pr_hardcont(" (old=0x%p a=%d)", prev, prev->context.active);
WARN_ON(prev->context.active < 1);
prev->context.active--;
}
again:
#endif /* CONFIG_SMP */
/* If we already have a valid assigned context, skip all that */
id = next->context.id;
if (likely(id != MMU_NO_CONTEXT)) {
#ifdef DEBUG_MAP_CONSISTENCY
if (context_mm[id] != next)
pr_err("MMU: mm 0x%p has id %d but context_mm[%d] says 0x%p\n",
next, id, id, context_mm[id]);
#endif
goto ctxt_ok;
}
/* We really don't have a context, let's try to acquire one */
id = next_context;
if (id > last_context)
id = first_context;
map = context_map;
/* No more free contexts, let's try to steal one */
if (nr_free_contexts == 0) {
#ifdef CONFIG_SMP
if (num_online_cpus() > 1) {
id = steal_context_smp(id);
if (id == MMU_NO_CONTEXT)
goto again;
goto stolen;
}
#endif /* CONFIG_SMP */
if (no_selective_tlbil)
id = steal_all_contexts();
else
id = steal_context_up(id);
goto stolen;
}
nr_free_contexts--;
/* We know there's at least one free context, try to find it */
while (__test_and_set_bit(id, map)) {
id = find_next_zero_bit(map, last_context+1, id);
if (id > last_context)
id = first_context;
}
stolen:
next_context = id + 1;
context_mm[id] = next;
next->context.id = id;
pr_hardcont(" | new id=%d,nrf=%d", id, nr_free_contexts);
context_check_map();
ctxt_ok:
/* If that context got marked stale on this CPU, then flush the
* local TLB for it and unmark it before we use it
*/
if (test_bit(id, stale_map[cpu])) {
pr_hardcont(" | stale flush %d [%d..%d]",
id, cpu_first_thread_sibling(cpu),
cpu_last_thread_sibling(cpu));
local_flush_tlb_mm(next);
/* XXX This clear should ultimately be part of local_flush_tlb_mm */
for (i = cpu_first_thread_sibling(cpu);
i <= cpu_last_thread_sibling(cpu); i++) {
if (stale_map[i])
__clear_bit(id, stale_map[i]);
}
}
/* Flick the MMU and release lock */
pr_hardcont(" -> %d\n", id);
set_context(id, next->pgd);
raw_spin_unlock(&context_lock);
}
/* Activate a secondary processor. */
__cpuinit void start_secondary(void *unused)
{
unsigned int cpu = smp_processor_id();
struct device_node *l2_cache;
int i, base;
atomic_inc(&init_mm.mm_count);
current->active_mm = &init_mm;
smp_store_cpu_info(cpu);
set_dec(tb_ticks_per_jiffy);
preempt_disable();
cpu_callin_map[cpu] = 1;
if (smp_ops->setup_cpu)
smp_ops->setup_cpu(cpu);
if (smp_ops->take_timebase)
smp_ops->take_timebase();
secondary_cpu_time_init();
#ifdef CONFIG_PPC64
if (system_state == SYSTEM_RUNNING)
vdso_data->processorCount++;
vdso_getcpu_init();
#endif
notify_cpu_starting(cpu);
set_cpu_online(cpu, true);
/* Update sibling maps */
base = cpu_first_thread_sibling(cpu);
for (i = 0; i < threads_per_core; i++) {
if (cpu_is_offline(base + i))
continue;
cpumask_set_cpu(cpu, cpu_sibling_mask(base + i));
cpumask_set_cpu(base + i, cpu_sibling_mask(cpu));
/* cpu_core_map should be a superset of
* cpu_sibling_map even if we don't have cache
* information, so update the former here, too.
*/
cpumask_set_cpu(cpu, cpu_core_mask(base + i));
cpumask_set_cpu(base + i, cpu_core_mask(cpu));
}
l2_cache = cpu_to_l2cache(cpu);
for_each_online_cpu(i) {
struct device_node *np = cpu_to_l2cache(i);
if (!np)
continue;
if (np == l2_cache) {
cpumask_set_cpu(cpu, cpu_core_mask(i));
cpumask_set_cpu(i, cpu_core_mask(cpu));
}
of_node_put(np);
}
of_node_put(l2_cache);
local_irq_enable();
cpu_startup_entry(CPUHP_ONLINE);
BUG();
}
void switch_mmu_context(struct mm_struct *prev, struct mm_struct *next)
{
unsigned int i, id, cpu = smp_processor_id();
unsigned long *map;
raw_spin_lock(&context_lock);
pr_hard("[%d] activating context for mm @%p, active=%d, id=%d",
cpu, next, next->context.active, next->context.id);
#ifdef CONFIG_SMP
next->context.active++;
if (prev) {
pr_hardcont(" (old=0x%p a=%d)", prev, prev->context.active);
WARN_ON(prev->context.active < 1);
prev->context.active--;
}
again:
#endif
id = next->context.id;
if (likely(id != MMU_NO_CONTEXT)) {
#ifdef DEBUG_MAP_CONSISTENCY
if (context_mm[id] != next)
pr_err("MMU: mm 0x%p has id %d but context_mm[%d] says 0x%p\n",
next, id, id, context_mm[id]);
#endif
goto ctxt_ok;
}
id = next_context;
if (id > last_context)
id = first_context;
map = context_map;
if (nr_free_contexts == 0) {
#ifdef CONFIG_SMP
if (num_online_cpus() > 1) {
id = steal_context_smp(id);
if (id == MMU_NO_CONTEXT)
goto again;
goto stolen;
}
#endif
id = steal_context_up(id);
goto stolen;
}
nr_free_contexts--;
while (__test_and_set_bit(id, map)) {
id = find_next_zero_bit(map, last_context+1, id);
if (id > last_context)
id = first_context;
}
stolen:
next_context = id + 1;
context_mm[id] = next;
next->context.id = id;
pr_hardcont(" | new id=%d,nrf=%d", id, nr_free_contexts);
context_check_map();
ctxt_ok:
if (test_bit(id, stale_map[cpu])) {
pr_hardcont(" | stale flush %d [%d..%d]",
id, cpu_first_thread_sibling(cpu),
cpu_last_thread_sibling(cpu));
local_flush_tlb_mm(next);
for (i = cpu_first_thread_sibling(cpu);
i <= cpu_last_thread_sibling(cpu); i++) {
__clear_bit(id, stale_map[i]);
}
}
pr_hardcont(" -> %d\n", id);
set_context(id, next->pgd);
raw_spin_unlock(&context_lock);
}