Commit 167c0fd7 authored by Linus Torvalds's avatar Linus Torvalds

Merge bk://are.twiddle.net:8080/linux-alpha-2.5

into home.transmeta.com:/home/torvalds/v2.5/linux
parents 6c645061 c8075847
...@@ -62,6 +62,7 @@ static struct { ...@@ -62,6 +62,7 @@ static struct {
enum ipi_message_type { enum ipi_message_type {
IPI_RESCHEDULE, IPI_RESCHEDULE,
IPI_MIGRATION,
IPI_CALL_FUNC, IPI_CALL_FUNC,
IPI_CPU_STOP, IPI_CPU_STOP,
}; };
...@@ -69,7 +70,7 @@ enum ipi_message_type { ...@@ -69,7 +70,7 @@ enum ipi_message_type {
spinlock_t kernel_flag __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED; spinlock_t kernel_flag __cacheline_aligned_in_smp = SPIN_LOCK_UNLOCKED;
/* Set to a secondary's cpuid when it comes online. */ /* Set to a secondary's cpuid when it comes online. */
static unsigned long smp_secondary_alive; static int smp_secondary_alive __initdata = 0;
/* Which cpus ids came online. */ /* Which cpus ids came online. */
unsigned long cpu_present_mask; unsigned long cpu_present_mask;
...@@ -82,6 +83,7 @@ int smp_num_probed; /* Internal processor count */ ...@@ -82,6 +83,7 @@ int smp_num_probed; /* Internal processor count */
int smp_num_cpus = 1; /* Number that came online. */ int smp_num_cpus = 1; /* Number that came online. */
int smp_threads_ready; /* True once the per process idle is forked. */ int smp_threads_ready; /* True once the per process idle is forked. */
cycles_t cacheflush_time; cycles_t cacheflush_time;
unsigned long cache_decay_ticks;
int __cpu_number_map[NR_CPUS]; int __cpu_number_map[NR_CPUS];
int __cpu_logical_map[NR_CPUS]; int __cpu_logical_map[NR_CPUS];
...@@ -156,13 +158,6 @@ smp_callin(void) ...@@ -156,13 +158,6 @@ smp_callin(void)
{ {
int cpuid = hard_smp_processor_id(); int cpuid = hard_smp_processor_id();
if (current != init_tasks[cpu_number_map(cpuid)]) {
printk("BUG: smp_calling: cpu %d current %p init_tasks[cpu_number_map(cpuid)] %p\n",
cpuid, current, init_tasks[cpu_number_map(cpuid)]);
}
DBGS(("CALLIN %d state 0x%lx\n", cpuid, current->state));
/* Turn on machine checks. */ /* Turn on machine checks. */
wrmces(7); wrmces(7);
...@@ -175,22 +170,21 @@ smp_callin(void) ...@@ -175,22 +170,21 @@ smp_callin(void)
/* Get our local ticker going. */ /* Get our local ticker going. */
smp_setup_percpu_timer(cpuid); smp_setup_percpu_timer(cpuid);
/* All kernel threads share the same mm context. */
atomic_inc(&init_mm.mm_count);
current->active_mm = &init_mm;
/* Must have completely accurate bogos. */ /* Must have completely accurate bogos. */
__sti(); __sti();
/* /* Wait boot CPU to stop with irq enabled before running
* Wait boot CPU to stop with irq enabled before calibrate_delay. */
* running calibrate_delay().
*/
wait_boot_cpu_to_stop(cpuid); wait_boot_cpu_to_stop(cpuid);
mb(); mb();
calibrate_delay(); calibrate_delay();
smp_store_cpu_info(cpuid); smp_store_cpu_info(cpuid);
/* /* Allow master to continue only after we written loops_per_jiffy. */
* Allow master to continue only after we written
* the loops_per_jiffy.
*/
wmb(); wmb();
smp_secondary_alive = 1; smp_secondary_alive = 1;
...@@ -198,15 +192,9 @@ smp_callin(void) ...@@ -198,15 +192,9 @@ smp_callin(void)
while (!smp_threads_ready) while (!smp_threads_ready)
barrier(); barrier();
DBGS(("smp_callin: commencing CPU %d current %p\n", DBGS(("smp_callin: commencing CPU %d current %p active_mm %p\n",
cpuid, current)); cpuid, current, current->active_mm));
/* Setup the scheduler for this processor. */
init_idle();
/* ??? This should be in init_idle. */
atomic_inc(&init_mm.mm_count);
current->active_mm = &init_mm;
/* Do nothing. */ /* Do nothing. */
cpu_idle(); cpu_idle();
} }
...@@ -222,8 +210,9 @@ static void __init ...@@ -222,8 +210,9 @@ static void __init
smp_tune_scheduling (int cpuid) smp_tune_scheduling (int cpuid)
{ {
struct percpu_struct *cpu; struct percpu_struct *cpu;
unsigned long on_chip_cache; unsigned long on_chip_cache; /* kB */
unsigned long freq; unsigned long freq; /* Hz */
unsigned long bandwidth = 350; /* MB/s */
cpu = (struct percpu_struct*)((char*)hwrpb + hwrpb->processor_offset cpu = (struct percpu_struct*)((char*)hwrpb + hwrpb->processor_offset
+ cpuid * hwrpb->processor_size); + cpuid * hwrpb->processor_size);
...@@ -244,43 +233,54 @@ smp_tune_scheduling (int cpuid) ...@@ -244,43 +233,54 @@ smp_tune_scheduling (int cpuid)
case EV6_CPU: case EV6_CPU:
case EV67_CPU: case EV67_CPU:
on_chip_cache = 64 + 64;
break;
default: default:
on_chip_cache = 8 + 8; on_chip_cache = 64 + 64;
break; break;
} }
freq = hwrpb->cycle_freq ? : est_cycle_freq; freq = hwrpb->cycle_freq ? : est_cycle_freq;
#if 0 cacheflush_time = (freq / 1000000) * (on_chip_cache << 10) / bandwidth;
/* Magic estimation stolen from x86 port. */ cache_decay_ticks = cacheflush_time / (freq / 1000) * HZ / 1000;
cacheflush_time = freq / 1024L * on_chip_cache / 5000L;
printk("Using heuristic of %d cycles.\n", printk("per-CPU timeslice cutoff: %ld.%02ld usecs.\n",
cacheflush_time); cacheflush_time/(freq/1000000),
#else (cacheflush_time*100/(freq/1000000)) % 100);
/* Magic value to force potential preemption of other CPUs. */ printk("task migration cache decay timeout: %ld msecs.\n",
cacheflush_time = INT_MAX; (cache_decay_ticks + 1) * 1000 / HZ);
}
printk("Using heuristic of %d cycles.\n", /* Wait until hwrpb->txrdy is clear for cpu. Return -1 on timeout. */
cacheflush_time); static int __init
#endif wait_for_txrdy (unsigned long cpumask)
{
unsigned long timeout;
if (!(hwrpb->txrdy & cpumask))
return 0;
timeout = jiffies + 10*HZ;
while (time_before(jiffies, timeout)) {
if (!(hwrpb->txrdy & cpumask))
return 0;
udelay(10);
barrier();
}
return -1;
} }
/* /*
* Send a message to a secondary's console. "START" is one such * Send a message to a secondary's console. "START" is one such
* interesting message. ;-) * interesting message. ;-)
*/ */
static void static void __init
send_secondary_console_msg(char *str, int cpuid) send_secondary_console_msg(char *str, int cpuid)
{ {
struct percpu_struct *cpu; struct percpu_struct *cpu;
register char *cp1, *cp2; register char *cp1, *cp2;
unsigned long cpumask; unsigned long cpumask;
size_t len; size_t len;
long timeout;
cpu = (struct percpu_struct *) cpu = (struct percpu_struct *)
((char*)hwrpb ((char*)hwrpb
...@@ -288,9 +288,8 @@ send_secondary_console_msg(char *str, int cpuid) ...@@ -288,9 +288,8 @@ send_secondary_console_msg(char *str, int cpuid)
+ cpuid * hwrpb->processor_size); + cpuid * hwrpb->processor_size);
cpumask = (1UL << cpuid); cpumask = (1UL << cpuid);
if (hwrpb->txrdy & cpumask) if (wait_for_txrdy(cpumask))
goto delay1; goto timeout;
ready1:
cp2 = str; cp2 = str;
len = strlen(cp2); len = strlen(cp2);
...@@ -302,34 +301,12 @@ send_secondary_console_msg(char *str, int cpuid) ...@@ -302,34 +301,12 @@ send_secondary_console_msg(char *str, int cpuid)
wmb(); wmb();
set_bit(cpuid, &hwrpb->rxrdy); set_bit(cpuid, &hwrpb->rxrdy);
if (hwrpb->txrdy & cpumask) if (wait_for_txrdy(cpumask))
goto delay2;
ready2:
return;
delay1:
/* Wait 10 seconds. Note that jiffies aren't ticking yet. */
for (timeout = 1000000; timeout > 0; --timeout) {
if (!(hwrpb->txrdy & cpumask))
goto ready1;
udelay(10);
barrier();
}
goto timeout;
delay2:
/* Wait 10 seconds. */
for (timeout = 1000000; timeout > 0; --timeout) {
if (!(hwrpb->txrdy & cpumask))
goto ready2;
udelay(10);
barrier();
}
goto timeout; goto timeout;
return;
timeout: timeout:
printk("Processor %x not ready\n", cpuid); printk("Processor %x not ready\n", cpuid);
return;
} }
/* /*
...@@ -392,7 +369,7 @@ static int __init ...@@ -392,7 +369,7 @@ static int __init
secondary_cpu_start(int cpuid, struct task_struct *idle) secondary_cpu_start(int cpuid, struct task_struct *idle)
{ {
struct percpu_struct *cpu; struct percpu_struct *cpu;
struct pcb_struct *hwpcb; struct pcb_struct *hwpcb, *ipcb;
long timeout; long timeout;
cpu = (struct percpu_struct *) cpu = (struct percpu_struct *)
...@@ -400,18 +377,19 @@ secondary_cpu_start(int cpuid, struct task_struct *idle) ...@@ -400,18 +377,19 @@ secondary_cpu_start(int cpuid, struct task_struct *idle)
+ hwrpb->processor_offset + hwrpb->processor_offset
+ cpuid * hwrpb->processor_size); + cpuid * hwrpb->processor_size);
hwpcb = (struct pcb_struct *) cpu->hwpcb; hwpcb = (struct pcb_struct *) cpu->hwpcb;
ipcb = &idle->thread_info->pcb;
/* Initialize the CPU's HWPCB to something just good enough for /* Initialize the CPU's HWPCB to something just good enough for
us to get started. Immediately after starting, we'll swpctx us to get started. Immediately after starting, we'll swpctx
to the target idle task's ptb. Reuse the stack in the mean to the target idle task's pcb. Reuse the stack in the mean
time. Precalculate the target PCBB. */ time. Precalculate the target PCBB. */
hwpcb->ksp = (unsigned long) idle + sizeof(union task_union) - 16; hwpcb->ksp = (unsigned long)ipcb + sizeof(union thread_union) - 16;
hwpcb->usp = 0; hwpcb->usp = 0;
hwpcb->ptbr = idle->thread.ptbr; hwpcb->ptbr = ipcb->ptbr;
hwpcb->pcc = 0; hwpcb->pcc = 0;
hwpcb->asn = 0; hwpcb->asn = 0;
hwpcb->unique = virt_to_phys(&idle->thread); hwpcb->unique = virt_to_phys(ipcb);
hwpcb->flags = idle->thread.pal_flags; hwpcb->flags = ipcb->flags;
hwpcb->res1 = hwpcb->res2 = 0; hwpcb->res1 = hwpcb->res2 = 0;
#if 0 #if 0
...@@ -419,7 +397,7 @@ secondary_cpu_start(int cpuid, struct task_struct *idle) ...@@ -419,7 +397,7 @@ secondary_cpu_start(int cpuid, struct task_struct *idle)
hwpcb->ksp, hwpcb->ptbr, hwrpb->vptb, hwpcb->unique)); hwpcb->ksp, hwpcb->ptbr, hwrpb->vptb, hwpcb->unique));
#endif #endif
DBGS(("Starting secondary cpu %d: state 0x%lx pal_flags 0x%lx\n", DBGS(("Starting secondary cpu %d: state 0x%lx pal_flags 0x%lx\n",
cpuid, idle->state, idle->thread.pal_flags)); cpuid, idle->state, ipcb->flags));
/* Setup HWRPB fields that SRM uses to activate secondary CPU */ /* Setup HWRPB fields that SRM uses to activate secondary CPU */
hwrpb->CPU_restart = __smp_callin; hwrpb->CPU_restart = __smp_callin;
...@@ -439,9 +417,9 @@ secondary_cpu_start(int cpuid, struct task_struct *idle) ...@@ -439,9 +417,9 @@ secondary_cpu_start(int cpuid, struct task_struct *idle)
send_secondary_console_msg("START\r\n", cpuid); send_secondary_console_msg("START\r\n", cpuid);
/* Wait 10 seconds for an ACK from the console. Note that jiffies /* Wait 10 seconds for an ACK from the console. */
aren't ticking yet. */ timeout = jiffies + 10*HZ;
for (timeout = 1000000; timeout > 0; timeout--) { while (time_before(jiffies, timeout)) {
if (cpu->flags & 1) if (cpu->flags & 1)
goto started; goto started;
udelay(10); udelay(10);
...@@ -450,18 +428,17 @@ secondary_cpu_start(int cpuid, struct task_struct *idle) ...@@ -450,18 +428,17 @@ secondary_cpu_start(int cpuid, struct task_struct *idle)
printk(KERN_ERR "SMP: Processor %d failed to start.\n", cpuid); printk(KERN_ERR "SMP: Processor %d failed to start.\n", cpuid);
return -1; return -1;
started: started:
DBGS(("secondary_cpu_start: SUCCESS for CPU %d!!!\n", cpuid)); DBGS(("secondary_cpu_start: SUCCESS for CPU %d!!!\n", cpuid));
return 0; return 0;
} }
static int __init fork_by_hand(void) static int __init
fork_by_hand(void)
{ {
/* Don't care about the contents of regs since we'll never
reschedule the forked task. */
struct pt_regs regs; struct pt_regs regs;
/*
* don't care about the regs settings since
* we'll never reschedule the forked task.
*/
return do_fork(CLONE_VM|CLONE_PID, 0, &regs, 0); return do_fork(CLONE_VM|CLONE_PID, 0, &regs, 0);
} }
...@@ -474,67 +451,57 @@ smp_boot_one_cpu(int cpuid, int cpunum) ...@@ -474,67 +451,57 @@ smp_boot_one_cpu(int cpuid, int cpunum)
struct task_struct *idle; struct task_struct *idle;
long timeout; long timeout;
/* Cook up an idler for this guy. Note that the address we give /* Cook up an idler for this guy. Note that the address we
to kernel_thread is irrelevant -- it's going to start where give to kernel_thread is irrelevant -- it's going to start
HWRPB.CPU_restart says to start. But this gets all the other where HWRPB.CPU_restart says to start. But this gets all
task-y sort of data structures set up like we wish. */ the other task-y sort of data structures set up like we
/* wish. We can't use kernel_thread since we must avoid
* We can't use kernel_thread since we must avoid to rescheduling the child. */
* reschedule the child.
*/
if (fork_by_hand() < 0) if (fork_by_hand() < 0)
panic("failed fork for CPU %d", cpuid); panic("failed fork for CPU %d", cpuid);
idle = init_task.prev_task; idle = init_task.prev_task;
if (!idle) if (!idle)
panic("No idle process for CPU %d", cpuid); panic("No idle process for CPU %d", cpuid);
if (idle == &init_task)
panic("idle process is init_task for CPU %d", cpuid);
idle->processor = cpuid; init_idle(idle, cpuid);
idle->cpus_runnable = 1 << cpuid; /* we schedule the first task manually */ unhash_process(idle);
__cpu_logical_map[cpunum] = cpuid; __cpu_logical_map[cpunum] = cpuid;
__cpu_number_map[cpuid] = cpunum; __cpu_number_map[cpuid] = cpunum;
del_from_runqueue(idle);
unhash_process(idle);
init_tasks[cpunum] = idle;
DBGS(("smp_boot_one_cpu: CPU %d state 0x%lx flags 0x%lx\n", DBGS(("smp_boot_one_cpu: CPU %d state 0x%lx flags 0x%lx\n",
cpuid, idle->state, idle->flags)); cpuid, idle->state, idle->flags));
/* The secondary will change this once it is happy. Note that /* Signal the secondary to wait a moment. */
secondary_cpu_start contains the necessary memory barrier. */
smp_secondary_alive = -1; smp_secondary_alive = -1;
/* Whirrr, whirrr, whirrrrrrrrr... */ /* Whirrr, whirrr, whirrrrrrrrr... */
if (secondary_cpu_start(cpuid, idle)) if (secondary_cpu_start(cpuid, idle))
return -1; return -1;
/* Notify the secondary CPU it can run calibrate_delay. */
mb(); mb();
/* Notify the secondary CPU it can run calibrate_delay() */
smp_secondary_alive = 0; smp_secondary_alive = 0;
/* We've been acked by the console; wait one second for the task /* We've been acked by the console; wait one second for
to start up for real. Note that jiffies aren't ticking yet. */ the task to start up for real. */
for (timeout = 0; timeout < 1000000; timeout++) { timeout = jiffies + 1*HZ;
while (time_before(jiffies, timeout)) {
if (smp_secondary_alive == 1) if (smp_secondary_alive == 1)
goto alive; goto alive;
udelay(10); udelay(10);
barrier(); barrier();
} }
/* we must invalidate our stuff as we failed to boot the CPU */ /* We must invalidate our stuff as we failed to boot the CPU. */
__cpu_logical_map[cpunum] = -1; __cpu_logical_map[cpunum] = -1;
__cpu_number_map[cpuid] = -1; __cpu_number_map[cpuid] = -1;
/* the idle task is local to us so free it as we don't use it */
free_task_struct(idle);
printk(KERN_ERR "SMP: Processor %d is stuck.\n", cpuid); printk(KERN_ERR "SMP: Processor %d is stuck.\n", cpuid);
return -1; return -1;
alive: alive:
/* Another "Red Snapper". */ /* Another "Red Snapper". */
return 0; return 0;
} }
...@@ -605,20 +572,15 @@ smp_boot_cpus(void) ...@@ -605,20 +572,15 @@ smp_boot_cpus(void)
__cpu_number_map[boot_cpuid] = 0; __cpu_number_map[boot_cpuid] = 0;
__cpu_logical_map[0] = boot_cpuid; __cpu_logical_map[0] = boot_cpuid;
current->processor = boot_cpuid; current_thread_info()->cpu = boot_cpuid;
smp_store_cpu_info(boot_cpuid); smp_store_cpu_info(boot_cpuid);
smp_tune_scheduling(boot_cpuid); smp_tune_scheduling(boot_cpuid);
smp_setup_percpu_timer(boot_cpuid); smp_setup_percpu_timer(boot_cpuid);
init_idle();
/* ??? This should be in init_idle. */
atomic_inc(&init_mm.mm_count);
current->active_mm = &init_mm;
/* Nothing to do on a UP box, or when told not to. */ /* Nothing to do on a UP box, or when told not to. */
if (smp_num_probed == 1 || max_cpus == 0) { if (smp_num_probed == 1 || max_cpus == 0) {
cpu_present_mask = 1UL << boot_cpuid;
printk(KERN_INFO "SMP mode deactivated.\n"); printk(KERN_INFO "SMP mode deactivated.\n");
return; return;
} }
...@@ -707,26 +669,35 @@ setup_profiling_timer(unsigned int multiplier) ...@@ -707,26 +669,35 @@ setup_profiling_timer(unsigned int multiplier)
static void static void
send_ipi_message(unsigned long to_whom, enum ipi_message_type operation) send_ipi_message(unsigned long to_whom, enum ipi_message_type operation)
{ {
long i, j; unsigned long i, set, n;
/* Reduce the number of memory barriers by doing two loops,
one to set the bits, one to invoke the interrupts. */
mb(); /* Order out-of-band data and bit setting. */
for (i = 0, j = 1; i < NR_CPUS; ++i, j <<= 1) { set = to_whom & -to_whom;
if (to_whom & j) if (to_whom == set) {
set_bit(operation, &ipi_data[i].bits); n = __ffs(set);
mb();
set_bit(operation, &ipi_data[n].bits);
mb();
wripir(n);
} else {
mb();
for (i = to_whom; i ; i &= ~set) {
set = i & -i;
n = __ffs(set);
set_bit(operation, &ipi_data[n].bits);
} }
mb(); /* Order bit setting and interrupt. */ mb();
for (i = to_whom; i ; i &= ~set) {
for (i = 0, j = 1; i < NR_CPUS; ++i, j <<= 1) { set = i & -i;
if (to_whom & j) n = __ffs(set);
wripir(i); wripir(n);
}
} }
} }
/* Data for IPI_MIGRATION. */
static task_t *migration_task;
/* Structure and data for smp_call_function. This is designed to /* Structure and data for smp_call_function. This is designed to
minimize static memory requirements. Plus it looks cleaner. */ minimize static memory requirements. Plus it looks cleaner. */
...@@ -743,13 +714,13 @@ static struct smp_call_struct *smp_call_function_data; ...@@ -743,13 +714,13 @@ static struct smp_call_struct *smp_call_function_data;
/* Atomicly drop data into a shared pointer. The pointer is free if /* Atomicly drop data into a shared pointer. The pointer is free if
it is initially locked. If retry, spin until free. */ it is initially locked. If retry, spin until free. */
static inline int static int
pointer_lock (void *lock, void *data, int retry) pointer_lock (void *lock, void *data, int retry)
{ {
void *old, *tmp; void *old, *tmp;
mb(); mb();
again: again:
/* Compare and swap with zero. */ /* Compare and swap with zero. */
asm volatile ( asm volatile (
"1: ldq_l %0,%1\n" "1: ldq_l %0,%1\n"
...@@ -792,13 +763,25 @@ handle_ipi(struct pt_regs *regs) ...@@ -792,13 +763,25 @@ handle_ipi(struct pt_regs *regs)
which = ops & -ops; which = ops & -ops;
ops &= ~which; ops &= ~which;
which = ffz(~which); which = __ffs(which);
if (which == IPI_RESCHEDULE) { switch (which) {
case IPI_RESCHEDULE:
/* Reschedule callback. Everything to be done /* Reschedule callback. Everything to be done
is done by the interrupt return path. */ is done by the interrupt return path. */
break;
case IPI_MIGRATION:
{
task_t *t = migration_task;
mb();
migration_task = 0;
sched_task_migrated(t);
break;
} }
else if (which == IPI_CALL_FUNC) {
case IPI_CALL_FUNC:
{
struct smp_call_struct *data; struct smp_call_struct *data;
void (*func)(void *info); void (*func)(void *info);
void *info; void *info;
...@@ -821,13 +804,16 @@ handle_ipi(struct pt_regs *regs) ...@@ -821,13 +804,16 @@ handle_ipi(struct pt_regs *regs)
/* Notify the sending CPU that the task is done. */ /* Notify the sending CPU that the task is done. */
mb(); mb();
if (wait) atomic_dec (&data->unfinished_count); if (wait) atomic_dec (&data->unfinished_count);
break;
} }
else if (which == IPI_CPU_STOP) {
case IPI_CPU_STOP:
halt(); halt();
}
else { default:
printk(KERN_CRIT "Unknown IPI on CPU %d: %lu\n", printk(KERN_CRIT "Unknown IPI on CPU %d: %lu\n",
this_cpu, which); this_cpu, which);
break;
} }
} while (ops); } while (ops);
...@@ -851,10 +837,23 @@ smp_send_reschedule(int cpu) ...@@ -851,10 +837,23 @@ smp_send_reschedule(int cpu)
send_ipi_message(1UL << cpu, IPI_RESCHEDULE); send_ipi_message(1UL << cpu, IPI_RESCHEDULE);
} }
void
smp_migrate_task(int cpu, task_t *t)
{
#if DEBUG_IPI_MSG
if (cpu == hard_smp_processor_id())
printk(KERN_WARNING
"smp_migrate_task: Sending IPI to self.\n");
#endif
/* Acquire the migration_task mutex. */
pointer_lock(&migration_task, t, 1);
send_ipi_message(1UL << cpu, IPI_MIGRATION);
}
void void
smp_send_stop(void) smp_send_stop(void)
{ {
unsigned long to_whom = cpu_present_mask ^ (1UL << smp_processor_id()); unsigned long to_whom = cpu_present_mask & ~(1UL << smp_processor_id());
#if DEBUG_IPI_MSG #if DEBUG_IPI_MSG
if (hard_smp_processor_id() != boot_cpu_id) if (hard_smp_processor_id() != boot_cpu_id)
printk(KERN_WARNING "smp_send_stop: Not on boot cpu.\n"); printk(KERN_WARNING "smp_send_stop: Not on boot cpu.\n");
...@@ -881,16 +880,13 @@ smp_call_function_on_cpu (void (*func) (void *info), void *info, int retry, ...@@ -881,16 +880,13 @@ smp_call_function_on_cpu (void (*func) (void *info), void *info, int retry,
struct smp_call_struct data; struct smp_call_struct data;
long timeout; long timeout;
int num_cpus_to_call; int num_cpus_to_call;
long i,j;
data.func = func; data.func = func;
data.info = info; data.info = info;
data.wait = wait; data.wait = wait;
to_whom &= ~(1L << smp_processor_id()); to_whom &= ~(1L << smp_processor_id());
for (i = 0, j = 1, num_cpus_to_call = 0; i < NR_CPUS; ++i, j <<= 1) num_cpus_to_call = hweight64(to_whom);
if (to_whom & j)
num_cpus_to_call++;
atomic_set(&data.unstarted_count, num_cpus_to_call); atomic_set(&data.unstarted_count, num_cpus_to_call);
atomic_set(&data.unfinished_count, num_cpus_to_call); atomic_set(&data.unfinished_count, num_cpus_to_call);
...@@ -1094,7 +1090,7 @@ flush_icache_page(struct vm_area_struct *vma, struct page *page) ...@@ -1094,7 +1090,7 @@ flush_icache_page(struct vm_area_struct *vma, struct page *page)
#ifdef CONFIG_DEBUG_SPINLOCK #ifdef CONFIG_DEBUG_SPINLOCK
void void
spin_unlock(spinlock_t * lock) _raw_spin_unlock(spinlock_t * lock)
{ {
mb(); mb();
lock->lock = 0; lock->lock = 0;
...@@ -1185,7 +1181,7 @@ debug_spin_trylock(spinlock_t * lock, const char *base_file, int line_no) ...@@ -1185,7 +1181,7 @@ debug_spin_trylock(spinlock_t * lock, const char *base_file, int line_no)
#endif /* CONFIG_DEBUG_SPINLOCK */ #endif /* CONFIG_DEBUG_SPINLOCK */
#ifdef CONFIG_DEBUG_RWLOCK #ifdef CONFIG_DEBUG_RWLOCK
void write_lock(rwlock_t * lock) void _raw_write_lock(rwlock_t * lock)
{ {
long regx, regy; long regx, regy;
int stuck_lock, stuck_reader; int stuck_lock, stuck_reader;
...@@ -1230,7 +1226,7 @@ void write_lock(rwlock_t * lock) ...@@ -1230,7 +1226,7 @@ void write_lock(rwlock_t * lock)
} }
} }
void read_lock(rwlock_t * lock) void _raw_read_lock(rwlock_t * lock)
{ {
long regx; long regx;
int stuck_lock; int stuck_lock;
......
...@@ -3,6 +3,7 @@ ...@@ -3,6 +3,7 @@
#include <linux/config.h> #include <linux/config.h>
#include <linux/kernel.h> #include <linux/kernel.h>
#include <asm/compiler.h>
/* /*
* Copyright 1994, Linus Torvalds. * Copyright 1994, Linus Torvalds.
...@@ -60,14 +61,14 @@ clear_bit(unsigned long nr, volatile void * addr) ...@@ -60,14 +61,14 @@ clear_bit(unsigned long nr, volatile void * addr)
__asm__ __volatile__( __asm__ __volatile__(
"1: ldl_l %0,%3\n" "1: ldl_l %0,%3\n"
" and %0,%2,%0\n" " bic %0,%2,%0\n"
" stl_c %0,%1\n" " stl_c %0,%1\n"
" beq %0,2f\n" " beq %0,2f\n"
".subsection 2\n" ".subsection 2\n"
"2: br 1b\n" "2: br 1b\n"
".previous" ".previous"
:"=&r" (temp), "=m" (*m) :"=&r" (temp), "=m" (*m)
:"Ir" (~(1UL << (nr & 31))), "m" (*m)); :"Ir" (1UL << (nr & 31)), "m" (*m));
} }
/* /*
...@@ -246,12 +247,15 @@ test_bit(int nr, volatile void * addr) ...@@ -246,12 +247,15 @@ test_bit(int nr, volatile void * addr)
*/ */
static inline unsigned long ffz_b(unsigned long x) static inline unsigned long ffz_b(unsigned long x)
{ {
unsigned long sum = 0; unsigned long sum, x1, x2, x4;
x = ~x & -~x; /* set first 0 bit, clear others */ x = ~x & -~x; /* set first 0 bit, clear others */
if (x & 0xF0) sum += 4; x1 = x & 0xAA;
if (x & 0xCC) sum += 2; x2 = x & 0xCC;
if (x & 0xAA) sum += 1; x4 = x & 0xF0;
sum = x2 ? 2 : 0;
sum += (x4 != 0) * 4;
sum += (x1 != 0);
return sum; return sum;
} }
...@@ -268,7 +272,7 @@ static inline unsigned long ffz(unsigned long word) ...@@ -268,7 +272,7 @@ static inline unsigned long ffz(unsigned long word)
__asm__("cmpbge %1,%2,%0" : "=r"(bits) : "r"(word), "r"(~0UL)); __asm__("cmpbge %1,%2,%0" : "=r"(bits) : "r"(word), "r"(~0UL));
qofs = ffz_b(bits); qofs = ffz_b(bits);
__asm__("extbl %1,%2,%0" : "=r"(bits) : "r"(word), "r"(qofs)); bits = __kernel_extbl(word, qofs);
bofs = ffz_b(bits); bofs = ffz_b(bits);
return qofs*8 + bofs; return qofs*8 + bofs;
...@@ -290,7 +294,7 @@ static inline unsigned long __ffs(unsigned long word) ...@@ -290,7 +294,7 @@ static inline unsigned long __ffs(unsigned long word)
__asm__("cmpbge $31,%1,%0" : "=r"(bits) : "r"(word)); __asm__("cmpbge $31,%1,%0" : "=r"(bits) : "r"(word));
qofs = ffz_b(bits); qofs = ffz_b(bits);
__asm__("extbl %1,%2,%0" : "=r"(bits) : "r"(word), "r"(qofs)); bits = __kernel_extbl(word, qofs);
bofs = ffz_b(~bits); bofs = ffz_b(~bits);
return qofs*8 + bofs; return qofs*8 + bofs;
...@@ -349,6 +353,14 @@ static inline unsigned long hweight64(unsigned long w) ...@@ -349,6 +353,14 @@ static inline unsigned long hweight64(unsigned long w)
#define hweight16(x) hweight64((x) & 0xfffful) #define hweight16(x) hweight64((x) & 0xfffful)
#define hweight8(x) hweight64((x) & 0xfful) #define hweight8(x) hweight64((x) & 0xfful)
#else #else
static inline unsigned long hweight64(unsigned long w)
{
unsigned long result;
for (result = 0; w ; w >>= 1)
result += (w & 1);
return result;
}
#define hweight32(x) generic_hweight32(x) #define hweight32(x) generic_hweight32(x)
#define hweight16(x) generic_hweight16(x) #define hweight16(x) generic_hweight16(x)
#define hweight8(x) generic_hweight8(x) #define hweight8(x) generic_hweight8(x)
......
...@@ -38,12 +38,12 @@ typedef struct { ...@@ -38,12 +38,12 @@ typedef struct {
#define spin_unlock_wait(x) ({ do { barrier(); } while ((x)->lock); }) #define spin_unlock_wait(x) ({ do { barrier(); } while ((x)->lock); })
#if CONFIG_DEBUG_SPINLOCK #if CONFIG_DEBUG_SPINLOCK
extern void spin_unlock(spinlock_t * lock); extern void _raw_spin_unlock(spinlock_t * lock);
extern void debug_spin_lock(spinlock_t * lock, const char *, int); extern void debug_spin_lock(spinlock_t * lock, const char *, int);
extern int debug_spin_trylock(spinlock_t * lock, const char *, int); extern int debug_spin_trylock(spinlock_t * lock, const char *, int);
#define spin_lock(LOCK) debug_spin_lock(LOCK, __BASE_FILE__, __LINE__) #define _raw_spin_lock(LOCK) debug_spin_lock(LOCK, __BASE_FILE__, __LINE__)
#define spin_trylock(LOCK) debug_spin_trylock(LOCK, __BASE_FILE__, __LINE__) #define _raw_spin_trylock(LOCK) debug_spin_trylock(LOCK, __BASE_FILE__, __LINE__)
#define spin_lock_own(LOCK, LOCATION) \ #define spin_lock_own(LOCK, LOCATION) \
do { \ do { \
...@@ -54,13 +54,13 @@ do { \ ...@@ -54,13 +54,13 @@ do { \
(LOCK)->lock ? "taken" : "freed", (LOCK)->on_cpu); \ (LOCK)->lock ? "taken" : "freed", (LOCK)->on_cpu); \
} while (0) } while (0)
#else #else
static inline void spin_unlock(spinlock_t * lock) static inline void _raw_spin_unlock(spinlock_t * lock)
{ {
mb(); mb();
lock->lock = 0; lock->lock = 0;
} }
static inline void spin_lock(spinlock_t * lock) static inline void _raw_spin_lock(spinlock_t * lock)
{ {
long tmp; long tmp;
...@@ -83,7 +83,11 @@ static inline void spin_lock(spinlock_t * lock) ...@@ -83,7 +83,11 @@ static inline void spin_lock(spinlock_t * lock)
: "m"(lock->lock) : "memory"); : "m"(lock->lock) : "memory");
} }
#define spin_trylock(lock) (!test_and_set_bit(0,(lock))) static inline int _raw_spin_trylock(spinlock_t *lock)
{
return !test_and_set_bit(0, &lock->lock);
}
#define spin_lock_own(LOCK, LOCATION) ((void)0) #define spin_lock_own(LOCK, LOCATION) ((void)0)
#endif /* CONFIG_DEBUG_SPINLOCK */ #endif /* CONFIG_DEBUG_SPINLOCK */
...@@ -98,10 +102,10 @@ typedef struct { ...@@ -98,10 +102,10 @@ typedef struct {
#define rwlock_init(x) do { *(x) = RW_LOCK_UNLOCKED; } while(0) #define rwlock_init(x) do { *(x) = RW_LOCK_UNLOCKED; } while(0)
#if CONFIG_DEBUG_RWLOCK #if CONFIG_DEBUG_RWLOCK
extern void write_lock(rwlock_t * lock); extern void _raw_write_lock(rwlock_t * lock);
extern void read_lock(rwlock_t * lock); extern void _raw_read_lock(rwlock_t * lock);
#else #else
static inline void write_lock(rwlock_t * lock) static inline void _raw_write_lock(rwlock_t * lock)
{ {
long regx; long regx;
...@@ -121,7 +125,7 @@ static inline void write_lock(rwlock_t * lock) ...@@ -121,7 +125,7 @@ static inline void write_lock(rwlock_t * lock)
: "0" (*(volatile int *)lock) : "memory"); : "0" (*(volatile int *)lock) : "memory");
} }
static inline void read_lock(rwlock_t * lock) static inline void _raw_read_lock(rwlock_t * lock)
{ {
long regx; long regx;
...@@ -142,13 +146,13 @@ static inline void read_lock(rwlock_t * lock) ...@@ -142,13 +146,13 @@ static inline void read_lock(rwlock_t * lock)
} }
#endif /* CONFIG_DEBUG_RWLOCK */ #endif /* CONFIG_DEBUG_RWLOCK */
static inline void write_unlock(rwlock_t * lock) static inline void _raw_write_unlock(rwlock_t * lock)
{ {
mb(); mb();
*(volatile int *)lock = 0; *(volatile int *)lock = 0;
} }
static inline void read_unlock(rwlock_t * lock) static inline void _raw_read_unlock(rwlock_t * lock)
{ {
long regx; long regx;
__asm__ __volatile__( __asm__ __volatile__(
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment