kernel-fxtec-pro1x/arch/avr32/kernel/process.c
Thomas Gleixner b8f8c3cf0a nohz: prevent tick stop outside of the idle loop
Jack Ren and Eric Miao tracked down the following long standing
problem in the NOHZ code:

	scheduler switch to idle task
	enable interrupts

Window starts here

	----> interrupt happens (does not set NEED_RESCHED)
	      	irq_exit() stops the tick

	----> interrupt happens (does set NEED_RESCHED)

	return from schedule()
	
	cpu_idle(): preempt_disable();

Window ends here

The interrupts can happen at any point inside the race window. The
first interrupt stops the tick, the second one causes the scheduler to
rerun and switch away from idle again and we end up with the tick
disabled.

The fact that it needs two interrupts where the first one does not set
NEED_RESCHED and the second one does made the bug obscure and extremly
hard to reproduce and analyse. Kudos to Jack and Eric.

Solution: Limit the NOHZ functionality to the idle loop to make sure
that we can not run into such a situation ever again.

cpu_idle()
{
	preempt_disable();

	while(1) {
		 tick_nohz_stop_sched_tick(1); <- tell NOHZ code that we
		 			          are in the idle loop

		 while (!need_resched())
		       halt();

		 tick_nohz_restart_sched_tick(); <- disables NOHZ mode
		 preempt_enable_no_resched();
		 schedule();
		 preempt_disable();
	}
}

In hindsight we should have done this forever, but ... 

/me grabs a large brown paperbag.

Debugged-by: Jack Ren <jack.ren@marvell.com>, 
Debugged-by: eric miao <eric.y.miao@gmail.com>
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
2008-07-18 18:10:28 +02:00

446 lines
11 KiB
C

/*
* Copyright (C) 2004-2006 Atmel Corporation
*
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License version 2 as
* published by the Free Software Foundation.
*/
#include <linux/sched.h>
#include <linux/module.h>
#include <linux/kallsyms.h>
#include <linux/fs.h>
#include <linux/ptrace.h>
#include <linux/reboot.h>
#include <linux/tick.h>
#include <linux/uaccess.h>
#include <linux/unistd.h>
#include <asm/sysreg.h>
#include <asm/ocd.h>
#include <asm/arch/pm.h>
void (*pm_power_off)(void) = NULL;
EXPORT_SYMBOL(pm_power_off);
/*
* This file handles the architecture-dependent parts of process handling..
*/
void cpu_idle(void)
{
/* endless idle loop with no priority at all */
while (1) {
tick_nohz_stop_sched_tick(1);
while (!need_resched())
cpu_idle_sleep();
tick_nohz_restart_sched_tick();
preempt_enable_no_resched();
schedule();
preempt_disable();
}
}
void machine_halt(void)
{
/*
* Enter Stop mode. The 32 kHz oscillator will keep running so
* the RTC will keep the time properly and the system will
* boot quickly.
*/
asm volatile("sleep 3\n\t"
"sub pc, -2");
}
void machine_power_off(void)
{
if (pm_power_off)
pm_power_off();
}
void machine_restart(char *cmd)
{
ocd_write(DC, (1 << OCD_DC_DBE_BIT));
ocd_write(DC, (1 << OCD_DC_RES_BIT));
while (1) ;
}
/*
* PC is actually discarded when returning from a system call -- the
* return address must be stored in LR. This function will make sure
* LR points to do_exit before starting the thread.
*
* Also, when returning from fork(), r12 is 0, so we must copy the
* argument as well.
*
* r0 : The argument to the main thread function
* r1 : The address of do_exit
* r2 : The address of the main thread function
*/
asmlinkage extern void kernel_thread_helper(void);
__asm__(" .type kernel_thread_helper, @function\n"
"kernel_thread_helper:\n"
" mov r12, r0\n"
" mov lr, r2\n"
" mov pc, r1\n"
" .size kernel_thread_helper, . - kernel_thread_helper");
int kernel_thread(int (*fn)(void *), void *arg, unsigned long flags)
{
struct pt_regs regs;
memset(&regs, 0, sizeof(regs));
regs.r0 = (unsigned long)arg;
regs.r1 = (unsigned long)fn;
regs.r2 = (unsigned long)do_exit;
regs.lr = (unsigned long)kernel_thread_helper;
regs.pc = (unsigned long)kernel_thread_helper;
regs.sr = MODE_SUPERVISOR;
return do_fork(flags | CLONE_VM | CLONE_UNTRACED,
0, &regs, 0, NULL, NULL);
}
EXPORT_SYMBOL(kernel_thread);
/*
* Free current thread data structures etc
*/
void exit_thread(void)
{
ocd_disable(current);
}
void flush_thread(void)
{
/* nothing to do */
}
void release_thread(struct task_struct *dead_task)
{
/* do nothing */
}
static void dump_mem(const char *str, const char *log_lvl,
unsigned long bottom, unsigned long top)
{
unsigned long p;
int i;
printk("%s%s(0x%08lx to 0x%08lx)\n", log_lvl, str, bottom, top);
for (p = bottom & ~31; p < top; ) {
printk("%s%04lx: ", log_lvl, p & 0xffff);
for (i = 0; i < 8; i++, p += 4) {
unsigned int val;
if (p < bottom || p >= top)
printk(" ");
else {
if (__get_user(val, (unsigned int __user *)p)) {
printk("\n");
goto out;
}
printk("%08x ", val);
}
}
printk("\n");
}
out:
return;
}
static inline int valid_stack_ptr(struct thread_info *tinfo, unsigned long p)
{
return (p > (unsigned long)tinfo)
&& (p < (unsigned long)tinfo + THREAD_SIZE - 3);
}
#ifdef CONFIG_FRAME_POINTER
static void show_trace_log_lvl(struct task_struct *tsk, unsigned long *sp,
struct pt_regs *regs, const char *log_lvl)
{
unsigned long lr, fp;
struct thread_info *tinfo;
if (regs)
fp = regs->r7;
else if (tsk == current)
asm("mov %0, r7" : "=r"(fp));
else
fp = tsk->thread.cpu_context.r7;
/*
* Walk the stack as long as the frame pointer (a) is within
* the kernel stack of the task, and (b) it doesn't move
* downwards.
*/
tinfo = task_thread_info(tsk);
printk("%sCall trace:\n", log_lvl);
while (valid_stack_ptr(tinfo, fp)) {
unsigned long new_fp;
lr = *(unsigned long *)fp;
#ifdef CONFIG_KALLSYMS
printk("%s [<%08lx>] ", log_lvl, lr);
#else
printk(" [<%08lx>] ", lr);
#endif
print_symbol("%s\n", lr);
new_fp = *(unsigned long *)(fp + 4);
if (new_fp <= fp)
break;
fp = new_fp;
}
printk("\n");
}
#else
static void show_trace_log_lvl(struct task_struct *tsk, unsigned long *sp,
struct pt_regs *regs, const char *log_lvl)
{
unsigned long addr;
printk("%sCall trace:\n", log_lvl);
while (!kstack_end(sp)) {
addr = *sp++;
if (kernel_text_address(addr)) {
#ifdef CONFIG_KALLSYMS
printk("%s [<%08lx>] ", log_lvl, addr);
#else
printk(" [<%08lx>] ", addr);
#endif
print_symbol("%s\n", addr);
}
}
printk("\n");
}
#endif
void show_stack_log_lvl(struct task_struct *tsk, unsigned long sp,
struct pt_regs *regs, const char *log_lvl)
{
struct thread_info *tinfo;
if (sp == 0) {
if (tsk)
sp = tsk->thread.cpu_context.ksp;
else
sp = (unsigned long)&tinfo;
}
if (!tsk)
tsk = current;
tinfo = task_thread_info(tsk);
if (valid_stack_ptr(tinfo, sp)) {
dump_mem("Stack: ", log_lvl, sp,
THREAD_SIZE + (unsigned long)tinfo);
show_trace_log_lvl(tsk, (unsigned long *)sp, regs, log_lvl);
}
}
void show_stack(struct task_struct *tsk, unsigned long *stack)
{
show_stack_log_lvl(tsk, (unsigned long)stack, NULL, "");
}
void dump_stack(void)
{
unsigned long stack;
show_trace_log_lvl(current, &stack, NULL, "");
}
EXPORT_SYMBOL(dump_stack);
static const char *cpu_modes[] = {
"Application", "Supervisor", "Interrupt level 0", "Interrupt level 1",
"Interrupt level 2", "Interrupt level 3", "Exception", "NMI"
};
void show_regs_log_lvl(struct pt_regs *regs, const char *log_lvl)
{
unsigned long sp = regs->sp;
unsigned long lr = regs->lr;
unsigned long mode = (regs->sr & MODE_MASK) >> MODE_SHIFT;
if (!user_mode(regs)) {
sp = (unsigned long)regs + FRAME_SIZE_FULL;
printk("%s", log_lvl);
print_symbol("PC is at %s\n", instruction_pointer(regs));
printk("%s", log_lvl);
print_symbol("LR is at %s\n", lr);
}
printk("%spc : [<%08lx>] lr : [<%08lx>] %s\n"
"%ssp : %08lx r12: %08lx r11: %08lx\n",
log_lvl, instruction_pointer(regs), lr, print_tainted(),
log_lvl, sp, regs->r12, regs->r11);
printk("%sr10: %08lx r9 : %08lx r8 : %08lx\n",
log_lvl, regs->r10, regs->r9, regs->r8);
printk("%sr7 : %08lx r6 : %08lx r5 : %08lx r4 : %08lx\n",
log_lvl, regs->r7, regs->r6, regs->r5, regs->r4);
printk("%sr3 : %08lx r2 : %08lx r1 : %08lx r0 : %08lx\n",
log_lvl, regs->r3, regs->r2, regs->r1, regs->r0);
printk("%sFlags: %c%c%c%c%c\n", log_lvl,
regs->sr & SR_Q ? 'Q' : 'q',
regs->sr & SR_V ? 'V' : 'v',
regs->sr & SR_N ? 'N' : 'n',
regs->sr & SR_Z ? 'Z' : 'z',
regs->sr & SR_C ? 'C' : 'c');
printk("%sMode bits: %c%c%c%c%c%c%c%c%c%c\n", log_lvl,
regs->sr & SR_H ? 'H' : 'h',
regs->sr & SR_J ? 'J' : 'j',
regs->sr & SR_DM ? 'M' : 'm',
regs->sr & SR_D ? 'D' : 'd',
regs->sr & SR_EM ? 'E' : 'e',
regs->sr & SR_I3M ? '3' : '.',
regs->sr & SR_I2M ? '2' : '.',
regs->sr & SR_I1M ? '1' : '.',
regs->sr & SR_I0M ? '0' : '.',
regs->sr & SR_GM ? 'G' : 'g');
printk("%sCPU Mode: %s\n", log_lvl, cpu_modes[mode]);
printk("%sProcess: %s [%d] (task: %p thread: %p)\n",
log_lvl, current->comm, current->pid, current,
task_thread_info(current));
}
void show_regs(struct pt_regs *regs)
{
unsigned long sp = regs->sp;
if (!user_mode(regs))
sp = (unsigned long)regs + FRAME_SIZE_FULL;
show_regs_log_lvl(regs, "");
show_trace_log_lvl(current, (unsigned long *)sp, regs, "");
}
EXPORT_SYMBOL(show_regs);
/* Fill in the fpu structure for a core dump. This is easy -- we don't have any */
int dump_fpu(struct pt_regs *regs, elf_fpregset_t *fpu)
{
/* Not valid */
return 0;
}
asmlinkage void ret_from_fork(void);
int copy_thread(int nr, unsigned long clone_flags, unsigned long usp,
unsigned long unused,
struct task_struct *p, struct pt_regs *regs)
{
struct pt_regs *childregs;
childregs = ((struct pt_regs *)(THREAD_SIZE + (unsigned long)task_stack_page(p))) - 1;
*childregs = *regs;
if (user_mode(regs))
childregs->sp = usp;
else
childregs->sp = (unsigned long)task_stack_page(p) + THREAD_SIZE;
childregs->r12 = 0; /* Set return value for child */
p->thread.cpu_context.sr = MODE_SUPERVISOR | SR_GM;
p->thread.cpu_context.ksp = (unsigned long)childregs;
p->thread.cpu_context.pc = (unsigned long)ret_from_fork;
clear_tsk_thread_flag(p, TIF_DEBUG);
if ((clone_flags & CLONE_PTRACE) && test_thread_flag(TIF_DEBUG))
ocd_enable(p);
return 0;
}
/* r12-r8 are dummy parameters to force the compiler to use the stack */
asmlinkage int sys_fork(struct pt_regs *regs)
{
return do_fork(SIGCHLD, regs->sp, regs, 0, NULL, NULL);
}
asmlinkage int sys_clone(unsigned long clone_flags, unsigned long newsp,
unsigned long parent_tidptr,
unsigned long child_tidptr, struct pt_regs *regs)
{
if (!newsp)
newsp = regs->sp;
return do_fork(clone_flags, newsp, regs, 0,
(int __user *)parent_tidptr,
(int __user *)child_tidptr);
}
asmlinkage int sys_vfork(struct pt_regs *regs)
{
return do_fork(CLONE_VFORK | CLONE_VM | SIGCHLD, regs->sp, regs,
0, NULL, NULL);
}
asmlinkage int sys_execve(char __user *ufilename, char __user *__user *uargv,
char __user *__user *uenvp, struct pt_regs *regs)
{
int error;
char *filename;
filename = getname(ufilename);
error = PTR_ERR(filename);
if (IS_ERR(filename))
goto out;
error = do_execve(filename, uargv, uenvp, regs);
if (error == 0)
current->ptrace &= ~PT_DTRACE;
putname(filename);
out:
return error;
}
/*
* This function is supposed to answer the question "who called
* schedule()?"
*/
unsigned long get_wchan(struct task_struct *p)
{
unsigned long pc;
unsigned long stack_page;
if (!p || p == current || p->state == TASK_RUNNING)
return 0;
stack_page = (unsigned long)task_stack_page(p);
BUG_ON(!stack_page);
/*
* The stored value of PC is either the address right after
* the call to __switch_to() or ret_from_fork.
*/
pc = thread_saved_pc(p);
if (in_sched_functions(pc)) {
#ifdef CONFIG_FRAME_POINTER
unsigned long fp = p->thread.cpu_context.r7;
BUG_ON(fp < stack_page || fp > (THREAD_SIZE + stack_page));
pc = *(unsigned long *)fp;
#else
/*
* We depend on the frame size of schedule here, which
* is actually quite ugly. It might be possible to
* determine the frame size automatically at build
* time by doing this:
* - compile sched.c
* - disassemble the resulting sched.o
* - look for 'sub sp,??' shortly after '<schedule>:'
*/
unsigned long sp = p->thread.cpu_context.ksp + 16;
BUG_ON(sp < stack_page || sp > (THREAD_SIZE + stack_page));
pc = *(unsigned long *)sp;
#endif
}
return pc;
}