1 #include <linux/errno.h>
2 #include <linux/kernel.h>
5 #include <linux/prctl.h>
6 #include <linux/slab.h>
7 #include <linux/sched.h>
8 #include <linux/module.h>
10 #include <linux/clockchips.h>
11 #include <linux/random.h>
12 #include <linux/user-return-notifier.h>
13 #include <linux/dmi.h>
14 #include <linux/utsname.h>
15 #include <trace/events/power.h>
16 #include <linux/hw_breakpoint.h>
17 #include <asm/system.h>
19 #include <asm/syscalls.h>
21 #include <asm/uaccess.h>
24 #include <asm/debugreg.h>
26 unsigned long idle_halt
;
27 EXPORT_SYMBOL(idle_halt
);
28 unsigned long idle_nomwait
;
29 EXPORT_SYMBOL(idle_nomwait
);
31 struct kmem_cache
*task_xstate_cachep
;
33 int arch_dup_task_struct(struct task_struct
*dst
, struct task_struct
*src
)
36 if (src
->thread
.xstate
) {
37 dst
->thread
.xstate
= kmem_cache_alloc(task_xstate_cachep
,
39 if (!dst
->thread
.xstate
)
41 WARN_ON((unsigned long)dst
->thread
.xstate
& 15);
42 memcpy(dst
->thread
.xstate
, src
->thread
.xstate
, xstate_size
);
47 void free_thread_xstate(struct task_struct
*tsk
)
49 if (tsk
->thread
.xstate
) {
50 kmem_cache_free(task_xstate_cachep
, tsk
->thread
.xstate
);
51 tsk
->thread
.xstate
= NULL
;
54 WARN(tsk
->thread
.ds_ctx
, "leaking DS context\n");
57 void free_thread_info(struct thread_info
*ti
)
59 free_thread_xstate(ti
->task
);
60 free_pages((unsigned long)ti
, get_order(THREAD_SIZE
));
63 void arch_task_cache_init(void)
66 kmem_cache_create("task_xstate", xstate_size
,
67 __alignof__(union thread_xstate
),
68 SLAB_PANIC
| SLAB_NOTRACK
, NULL
);
72 * Free current thread data structures etc..
74 void exit_thread(void)
76 struct task_struct
*me
= current
;
77 struct thread_struct
*t
= &me
->thread
;
78 unsigned long *bp
= t
->io_bitmap_ptr
;
81 struct tss_struct
*tss
= &per_cpu(init_tss
, get_cpu());
83 t
->io_bitmap_ptr
= NULL
;
84 clear_thread_flag(TIF_IO_BITMAP
);
86 * Careful, clear this in the TSS too:
88 memset(tss
->io_bitmap
, 0xff, t
->io_bitmap_max
);
95 void show_regs_common(void)
97 const char *board
, *product
;
99 board
= dmi_get_system_info(DMI_BOARD_NAME
);
102 product
= dmi_get_system_info(DMI_PRODUCT_NAME
);
106 printk(KERN_CONT
"\n");
107 printk(KERN_DEFAULT
"Pid: %d, comm: %.20s %s %s %.*s %s/%s\n",
108 current
->pid
, current
->comm
, print_tainted(),
109 init_utsname()->release
,
110 (int)strcspn(init_utsname()->version
, " "),
111 init_utsname()->version
, board
, product
);
114 void flush_thread(void)
116 struct task_struct
*tsk
= current
;
118 flush_ptrace_hw_breakpoint(tsk
);
119 memset(tsk
->thread
.tls_array
, 0, sizeof(tsk
->thread
.tls_array
));
121 * Forget coprocessor state..
123 tsk
->fpu_counter
= 0;
128 static void hard_disable_TSC(void)
130 write_cr4(read_cr4() | X86_CR4_TSD
);
133 void disable_TSC(void)
136 if (!test_and_set_thread_flag(TIF_NOTSC
))
138 * Must flip the CPU state synchronously with
139 * TIF_NOTSC in the current running context.
145 static void hard_enable_TSC(void)
147 write_cr4(read_cr4() & ~X86_CR4_TSD
);
150 static void enable_TSC(void)
153 if (test_and_clear_thread_flag(TIF_NOTSC
))
155 * Must flip the CPU state synchronously with
156 * TIF_NOTSC in the current running context.
162 int get_tsc_mode(unsigned long adr
)
166 if (test_thread_flag(TIF_NOTSC
))
167 val
= PR_TSC_SIGSEGV
;
171 return put_user(val
, (unsigned int __user
*)adr
);
174 int set_tsc_mode(unsigned int val
)
176 if (val
== PR_TSC_SIGSEGV
)
178 else if (val
== PR_TSC_ENABLE
)
186 void __switch_to_xtra(struct task_struct
*prev_p
, struct task_struct
*next_p
,
187 struct tss_struct
*tss
)
189 struct thread_struct
*prev
, *next
;
191 prev
= &prev_p
->thread
;
192 next
= &next_p
->thread
;
194 if (test_tsk_thread_flag(next_p
, TIF_DS_AREA_MSR
) ||
195 test_tsk_thread_flag(prev_p
, TIF_DS_AREA_MSR
))
196 ds_switch_to(prev_p
, next_p
);
197 else if (next
->debugctlmsr
!= prev
->debugctlmsr
)
198 update_debugctlmsr(next
->debugctlmsr
);
200 if (test_tsk_thread_flag(prev_p
, TIF_NOTSC
) ^
201 test_tsk_thread_flag(next_p
, TIF_NOTSC
)) {
202 /* prev and next are different */
203 if (test_tsk_thread_flag(next_p
, TIF_NOTSC
))
209 if (test_tsk_thread_flag(next_p
, TIF_IO_BITMAP
)) {
211 * Copy the relevant range of the IO bitmap.
212 * Normally this is 128 bytes or less:
214 memcpy(tss
->io_bitmap
, next
->io_bitmap_ptr
,
215 max(prev
->io_bitmap_max
, next
->io_bitmap_max
));
216 } else if (test_tsk_thread_flag(prev_p
, TIF_IO_BITMAP
)) {
218 * Clear any possible leftover bits:
220 memset(tss
->io_bitmap
, 0xff, prev
->io_bitmap_max
);
222 propagate_user_return_notify(prev_p
, next_p
);
225 int sys_fork(struct pt_regs
*regs
)
227 return do_fork(SIGCHLD
, regs
->sp
, regs
, 0, NULL
, NULL
);
231 * This is trivial, and on the face of it looks like it
232 * could equally well be done in user mode.
234 * Not so, for quite unobvious reasons - register pressure.
235 * In user mode vfork() cannot have a stack frame, and if
236 * done by calling the "clone()" system call directly, you
237 * do not have enough call-clobbered registers to hold all
238 * the information you need.
240 int sys_vfork(struct pt_regs
*regs
)
242 return do_fork(CLONE_VFORK
| CLONE_VM
| SIGCHLD
, regs
->sp
, regs
, 0,
247 sys_clone(unsigned long clone_flags
, unsigned long newsp
,
248 void __user
*parent_tid
, void __user
*child_tid
, struct pt_regs
*regs
)
252 return do_fork(clone_flags
, newsp
, regs
, 0, parent_tid
, child_tid
);
256 * This gets run with %si containing the
257 * function to call, and %di containing
260 extern void kernel_thread_helper(void);
263 * Create a kernel thread
265 int kernel_thread(int (*fn
)(void *), void *arg
, unsigned long flags
)
269 memset(®s
, 0, sizeof(regs
));
271 regs
.si
= (unsigned long) fn
;
272 regs
.di
= (unsigned long) arg
;
277 regs
.fs
= __KERNEL_PERCPU
;
278 regs
.gs
= __KERNEL_STACK_CANARY
;
280 regs
.ss
= __KERNEL_DS
;
284 regs
.ip
= (unsigned long) kernel_thread_helper
;
285 regs
.cs
= __KERNEL_CS
| get_kernel_rpl();
286 regs
.flags
= X86_EFLAGS_IF
| 0x2;
288 /* Ok, create the new process.. */
289 return do_fork(flags
| CLONE_VM
| CLONE_UNTRACED
, 0, ®s
, 0, NULL
, NULL
);
291 EXPORT_SYMBOL(kernel_thread
);
294 * sys_execve() executes a new program.
296 long sys_execve(char __user
*name
, char __user
* __user
*argv
,
297 char __user
* __user
*envp
, struct pt_regs
*regs
)
302 filename
= getname(name
);
303 error
= PTR_ERR(filename
);
304 if (IS_ERR(filename
))
306 error
= do_execve(filename
, argv
, envp
, regs
);
310 /* Make sure we don't return using sysenter.. */
311 set_thread_flag(TIF_IRET
);
320 * Idle related variables and functions
322 unsigned long boot_option_idle_override
= 0;
323 EXPORT_SYMBOL(boot_option_idle_override
);
326 * Powermanagement idle function, if any..
328 void (*pm_idle
)(void);
329 EXPORT_SYMBOL(pm_idle
);
333 * This halt magic was a workaround for ancient floppy DMA
334 * wreckage. It should be safe to remove.
336 static int hlt_counter
;
337 void disable_hlt(void)
341 EXPORT_SYMBOL(disable_hlt
);
343 void enable_hlt(void)
347 EXPORT_SYMBOL(enable_hlt
);
349 static inline int hlt_use_halt(void)
351 return (!hlt_counter
&& boot_cpu_data
.hlt_works_ok
);
354 static inline int hlt_use_halt(void)
361 * We use this if we don't have any better
364 void default_idle(void)
366 if (hlt_use_halt()) {
367 trace_power_start(POWER_CSTATE
, 1);
368 current_thread_info()->status
&= ~TS_POLLING
;
370 * TS_POLLING-cleared state must be visible before we
376 safe_halt(); /* enables interrupts racelessly */
379 current_thread_info()->status
|= TS_POLLING
;
382 /* loop is done by the caller */
386 #ifdef CONFIG_APM_MODULE
387 EXPORT_SYMBOL(default_idle
);
390 void stop_this_cpu(void *dummy
)
396 set_cpu_online(smp_processor_id(), false);
397 disable_local_APIC();
400 if (hlt_works(smp_processor_id()))
405 static void do_nothing(void *unused
)
410 * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
411 * pm_idle and update to new pm_idle value. Required while changing pm_idle
412 * handler on SMP systems.
414 * Caller must have changed pm_idle to the new value before the call. Old
415 * pm_idle value will not be used by any CPU after the return of this function.
417 void cpu_idle_wait(void)
420 /* kick all the CPUs so that they exit out of pm_idle */
421 smp_call_function(do_nothing
, NULL
, 1);
423 EXPORT_SYMBOL_GPL(cpu_idle_wait
);
426 * This uses new MONITOR/MWAIT instructions on P4 processors with PNI,
427 * which can obviate IPI to trigger checking of need_resched.
428 * We execute MONITOR against need_resched and enter optimized wait state
429 * through MWAIT. Whenever someone changes need_resched, we would be woken
430 * up from MWAIT (without an IPI).
432 * New with Core Duo processors, MWAIT can take some hints based on CPU
435 void mwait_idle_with_hints(unsigned long ax
, unsigned long cx
)
437 trace_power_start(POWER_CSTATE
, (ax
>>4)+1);
438 if (!need_resched()) {
439 if (cpu_has(¤t_cpu_data
, X86_FEATURE_CLFLUSH_MONITOR
))
440 clflush((void *)¤t_thread_info()->flags
);
442 __monitor((void *)¤t_thread_info()->flags
, 0, 0);
449 /* Default MONITOR/MWAIT with no hints, used for default C1 state */
450 static void mwait_idle(void)
452 if (!need_resched()) {
453 trace_power_start(POWER_CSTATE
, 1);
454 if (cpu_has(¤t_cpu_data
, X86_FEATURE_CLFLUSH_MONITOR
))
455 clflush((void *)¤t_thread_info()->flags
);
457 __monitor((void *)¤t_thread_info()->flags
, 0, 0);
468 * On SMP it's slightly faster (but much more power-consuming!)
469 * to poll the ->work.need_resched flag instead of waiting for the
470 * cross-CPU IPI to arrive. Use this option with caution.
472 static void poll_idle(void)
474 trace_power_start(POWER_CSTATE
, 0);
476 while (!need_resched())
482 * mwait selection logic:
484 * It depends on the CPU. For AMD CPUs that support MWAIT this is
485 * wrong. Family 0x10 and 0x11 CPUs will enter C1 on HLT. Powersavings
486 * then depend on a clock divisor and current Pstate of the core. If
487 * all cores of a processor are in halt state (C1) the processor can
488 * enter the C1E (C1 enhanced) state. If mwait is used this will never
491 * idle=mwait overrides this decision and forces the usage of mwait.
493 static int __cpuinitdata force_mwait
;
495 #define MWAIT_INFO 0x05
496 #define MWAIT_ECX_EXTENDED_INFO 0x01
497 #define MWAIT_EDX_C1 0xf0
499 static int __cpuinit
mwait_usable(const struct cpuinfo_x86
*c
)
501 u32 eax
, ebx
, ecx
, edx
;
506 if (c
->cpuid_level
< MWAIT_INFO
)
509 cpuid(MWAIT_INFO
, &eax
, &ebx
, &ecx
, &edx
);
510 /* Check, whether EDX has extended info about MWAIT */
511 if (!(ecx
& MWAIT_ECX_EXTENDED_INFO
))
515 * edx enumeratios MONITOR/MWAIT extensions. Check, whether
518 return (edx
& MWAIT_EDX_C1
);
522 * Check for AMD CPUs, which have potentially C1E support
524 static int __cpuinit
check_c1e_idle(const struct cpuinfo_x86
*c
)
526 if (c
->x86_vendor
!= X86_VENDOR_AMD
)
532 /* Family 0x0f models < rev F do not have C1E */
533 if (c
->x86
== 0x0f && c
->x86_model
< 0x40)
539 static cpumask_var_t c1e_mask
;
540 static int c1e_detected
;
542 void c1e_remove_cpu(int cpu
)
544 if (c1e_mask
!= NULL
)
545 cpumask_clear_cpu(cpu
, c1e_mask
);
549 * C1E aware idle routine. We check for C1E active in the interrupt
550 * pending message MSR. If we detect C1E, then we handle it the same
551 * way as C3 power states (local apic timer and TSC stop)
553 static void c1e_idle(void)
561 rdmsr(MSR_K8_INT_PENDING_MSG
, lo
, hi
);
562 if (lo
& K8_INTP_C1E_ACTIVE_MASK
) {
564 if (!boot_cpu_has(X86_FEATURE_NONSTOP_TSC
))
565 mark_tsc_unstable("TSC halt in AMD C1E");
566 printk(KERN_INFO
"System has AMD C1E enabled\n");
567 set_cpu_cap(&boot_cpu_data
, X86_FEATURE_AMDC1E
);
572 int cpu
= smp_processor_id();
574 if (!cpumask_test_cpu(cpu
, c1e_mask
)) {
575 cpumask_set_cpu(cpu
, c1e_mask
);
577 * Force broadcast so ACPI can not interfere.
579 clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_FORCE
,
581 printk(KERN_INFO
"Switch to broadcast mode on CPU%d\n",
584 clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_ENTER
, &cpu
);
589 * The switch back from broadcast mode needs to be
590 * called with interrupts disabled.
593 clockevents_notify(CLOCK_EVT_NOTIFY_BROADCAST_EXIT
, &cpu
);
599 void __cpuinit
select_idle_routine(const struct cpuinfo_x86
*c
)
602 if (pm_idle
== poll_idle
&& smp_num_siblings
> 1) {
603 printk(KERN_WARNING
"WARNING: polling idle and HT enabled,"
604 " performance may degrade.\n");
610 if (cpu_has(c
, X86_FEATURE_MWAIT
) && mwait_usable(c
)) {
612 * One CPU supports mwait => All CPUs supports mwait
614 printk(KERN_INFO
"using mwait in idle threads.\n");
615 pm_idle
= mwait_idle
;
616 } else if (check_c1e_idle(c
)) {
617 printk(KERN_INFO
"using C1E aware idle routine\n");
620 pm_idle
= default_idle
;
623 void __init
init_c1e_mask(void)
625 /* If we're using c1e_idle, we need to allocate c1e_mask. */
626 if (pm_idle
== c1e_idle
)
627 zalloc_cpumask_var(&c1e_mask
, GFP_KERNEL
);
630 static int __init
idle_setup(char *str
)
635 if (!strcmp(str
, "poll")) {
636 printk("using polling idle threads.\n");
638 } else if (!strcmp(str
, "mwait"))
640 else if (!strcmp(str
, "halt")) {
642 * When the boot option of idle=halt is added, halt is
643 * forced to be used for CPU idle. In such case CPU C2/C3
644 * won't be used again.
645 * To continue to load the CPU idle driver, don't touch
646 * the boot_option_idle_override.
648 pm_idle
= default_idle
;
651 } else if (!strcmp(str
, "nomwait")) {
653 * If the boot option of "idle=nomwait" is added,
654 * it means that mwait will be disabled for CPU C2/C3
655 * states. In such case it won't touch the variable
656 * of boot_option_idle_override.
663 boot_option_idle_override
= 1;
666 early_param("idle", idle_setup
);
668 unsigned long arch_align_stack(unsigned long sp
)
670 if (!(current
->personality
& ADDR_NO_RANDOMIZE
) && randomize_va_space
)
671 sp
-= get_random_int() % 8192;
675 unsigned long arch_randomize_brk(struct mm_struct
*mm
)
677 unsigned long range_end
= mm
->brk
+ 0x02000000;
678 return randomize_range(mm
->brk
, range_end
, 0) ? : mm
->brk
;