traps.c

linux 内核源代码

/*
 * Architecture-specific trap handling.
 *
 * Copyright (C) 1998-2003 Hewlett-Packard Co
 *	David Mosberger-Tang <[e-mail]>
 *
 * 05/12/00 grao <[e-mail]> : added isr in siginfo for SIGFPE
 */

#include <linux/kernel.h>
#include <linux/init.h>
#include <linux/sched.h>
#include <linux/tty.h>
#include <linux/vt_kern.h>		/* For unblank_screen() */
#include <linux/module.h>       /* for EXPORT_SYMBOL */
#include <linux/hardirq.h>
#include <linux/kprobes.h>
#include <linux/delay.h>		/* for ssleep() */
#include <linux/kdebug.h>

#include <asm/fpswa.h>
#include <asm/ia32.h>
#include <asm/intrinsics.h>
#include <asm/processor.h>
#include <asm/uaccess.h>

fpswa_interface_t *fpswa_interface;
EXPORT_SYMBOL(fpswa_interface);

void __init
trap_init (void)
{
	if (ia64_boot_param->fpswa)
		/* FPSWA fixup: make the interface pointer a kernel virtual address: */
		fpswa_interface = __va(ia64_boot_param->fpswa);
}

void
die (const char *str, struct pt_regs *regs, long err)
{
	static struct {
		spinlock_t lock;
		u32 lock_owner;
		int lock_owner_depth;
	} die = {
		.lock =	__SPIN_LOCK_UNLOCKED(die.lock),
		.lock_owner = -1,
		.lock_owner_depth = 0
	};
	static int die_counter;
	int cpu = get_cpu();

	if (die.lock_owner != cpu) {
		console_verbose();
		spin_lock_irq(&die.lock);
		die.lock_owner = cpu;
		die.lock_owner_depth = 0;
		bust_spinlocks(1);
	}
	put_cpu();

	if (++die.lock_owner_depth < 3) {
		printk("%s[%d]: %s %ld [%d]\n",
		current->comm, task_pid_nr(current), str, err, ++die_counter);
		(void) notify_die(DIE_OOPS, (char *)str, regs, err, 255, SIGSEGV);
		show_regs(regs);
  	} else
		printk(KERN_ERR "Recursive die() failure, output suppressed\n");

	bust_spinlocks(0);
	die.lock_owner = -1;
	add_taint(TAINT_DIE);
	spin_unlock_irq(&die.lock);

	if (panic_on_oops)
		panic("Fatal exception");

  	do_exit(SIGSEGV);
}

void
die_if_kernel (char *str, struct pt_regs *regs, long err)
{
	if (!user_mode(regs))
		die(str, regs, err);
}

void
__kprobes ia64_bad_break (unsigned long break_num, struct pt_regs *regs)
{
	siginfo_t siginfo;
	int sig, code;

	/* SIGILL, SIGFPE, SIGSEGV, and SIGBUS want these field initialized: */
	siginfo.si_addr = (void __user *) (regs->cr_iip + ia64_psr(regs)->ri);
	siginfo.si_imm = break_num;
	siginfo.si_flags = 0;		/* clear __ISR_VALID */
	siginfo.si_isr = 0;

	switch (break_num) {
	      case 0: /* unknown error (used by GCC for __builtin_abort()) */
		if (notify_die(DIE_BREAK, "break 0", regs, break_num, TRAP_BRKPT, SIGTRAP)
			       	== NOTIFY_STOP)
			return;
		die_if_kernel("bugcheck!", regs, break_num);
		sig = SIGILL; code = ILL_ILLOPC;
		break;

	      case 1: /* integer divide by zero */
		sig = SIGFPE; code = FPE_INTDIV;
		break;

	      case 2: /* integer overflow */
		sig = SIGFPE; code = FPE_INTOVF;
		break;

	      case 3: /* range check/bounds check */
		sig = SIGFPE; code = FPE_FLTSUB;
		break;

	      case 4: /* null pointer dereference */
		sig = SIGSEGV; code = SEGV_MAPERR;
		break;

	      case 5: /* misaligned data */
		sig = SIGSEGV; code = BUS_ADRALN;
		break;

	      case 6: /* decimal overflow */
		sig = SIGFPE; code = __FPE_DECOVF;
		break;

	      case 7: /* decimal divide by zero */
		sig = SIGFPE; code = __FPE_DECDIV;
		break;

	      case 8: /* packed decimal error */
		sig = SIGFPE; code = __FPE_DECERR;
		break;

	      case 9: /* invalid ASCII digit */
		sig = SIGFPE; code = __FPE_INVASC;
		break;

	      case 10: /* invalid decimal digit */
		sig = SIGFPE; code = __FPE_INVDEC;
		break;

	      case 11: /* paragraph stack overflow */
		sig = SIGSEGV; code = __SEGV_PSTKOVF;
		break;

	      case 0x3f000 ... 0x3ffff:	/* bundle-update in progress */
		sig = SIGILL; code = __ILL_BNDMOD;
		break;

	      default:
		if (break_num < 0x40000 || break_num > 0x100000)
			die_if_kernel("Bad break", regs, break_num);

		if (break_num < 0x80000) {
			sig = SIGILL; code = __ILL_BREAK;
		} else {
			if (notify_die(DIE_BREAK, "bad break", regs, break_num, TRAP_BRKPT, SIGTRAP)
					== NOTIFY_STOP)
				return;
			sig = SIGTRAP; code = TRAP_BRKPT;
		}
	}
	siginfo.si_signo = sig;
	siginfo.si_errno = 0;
	siginfo.si_code = code;
	force_sig_info(sig, &siginfo, current);
}

/*
 * disabled_fph_fault() is called when a user-level process attempts to access f32..f127
 * and it doesn't own the fp-high register partition.  When this happens, we save the
 * current fph partition in the task_struct of the fpu-owner (if necessary) and then load
 * the fp-high partition of the current task (if necessary).  Note that the kernel has
 * access to fph by the time we get here, as the IVT's "Disabled FP-Register" handler takes
 * care of clearing psr.dfh.
 */
static inline void
disabled_fph_fault (struct pt_regs *regs)
{
	struct ia64_psr *psr = ia64_psr(regs);

	/* first, grant user-level access to fph partition: */
	psr->dfh = 0;

	/*
	 * Make sure that no other task gets in on this processor
	 * while we're claiming the FPU
	 */
	preempt_disable();
#ifndef CONFIG_SMP
	{
		struct task_struct *fpu_owner
			= (struct task_struct *)ia64_get_kr(IA64_KR_FPU_OWNER);

		if (ia64_is_local_fpu_owner(current)) {
			preempt_enable_no_resched();
			return;
		}

		if (fpu_owner)
			ia64_flush_fph(fpu_owner);
	}
#endif /* !CONFIG_SMP */
	ia64_set_local_fpu_owner(current);
	if ((current->thread.flags & IA64_THREAD_FPH_VALID) != 0) {
		__ia64_load_fpu(current->thread.fph);
		psr->mfh = 0;
	} else {
		__ia64_init_fpu();
		/*
		 * Set mfh because the state in thread.fph does not match the state in
		 * the fph partition.
		 */
		psr->mfh = 1;
	}
	preempt_enable_no_resched();
}

static inline int
fp_emulate (int fp_fault, void *bundle, long *ipsr, long *fpsr, long *isr, long *pr, long *ifs,
	    struct pt_regs *regs)
{
	fp_state_t fp_state;
	fpswa_ret_t ret;

	if (!fpswa_interface)
		return -1;

	memset(&fp_state, 0, sizeof(fp_state_t));

	/*
	 * compute fp_state.  only FP registers f6 - f11 are used by the
	 * kernel, so set those bits in the mask and set the low volatile
	 * pointer to point to these registers.
	 */
	fp_state.bitmask_low64 = 0xfc0;  /* bit6..bit11 */

	fp_state.fp_state_low_volatile = (fp_state_low_volatile_t *) &regs->f6;
	/*
	 * unsigned long (*EFI_FPSWA) (
	 *      unsigned long    trap_type,
	 *	void             *Bundle,
	 *	unsigned long    *pipsr,
	 *	unsigned long    *pfsr,
	 *	unsigned long    *pisr,
	 *	unsigned long    *ppreds,
	 *	unsigned long    *pifs,
	 *	void             *fp_state);
	 */
	ret = (*fpswa_interface->fpswa)((unsigned long) fp_fault, bundle,
					(unsigned long *) ipsr, (unsigned long *) fpsr,
					(unsigned long *) isr, (unsigned long *) pr,
					(unsigned long *) ifs, &fp_state);

	return ret.status;
}

struct fpu_swa_msg {
	unsigned long count;
	unsigned long time;
};
static DEFINE_PER_CPU(struct fpu_swa_msg, cpulast);
DECLARE_PER_CPU(struct fpu_swa_msg, cpulast);
static struct fpu_swa_msg last __cacheline_aligned;


/*
 * Handle floating-point assist faults and traps.
 */
static int
handle_fpu_swa (int fp_fault, struct pt_regs *regs, unsigned long isr)
{
	long exception, bundle[2];
	unsigned long fault_ip;
	struct siginfo siginfo;

	fault_ip = regs->cr_iip;
	if (!fp_fault && (ia64_psr(regs)->ri == 0))
		fault_ip -= 16;
	if (copy_from_user(bundle, (void __user *) fault_ip, sizeof(bundle)))
		return -1;

	if (!(current->thread.flags & IA64_THREAD_FPEMU_NOPRINT))  {
		unsigned long count, current_jiffies = jiffies;
		struct fpu_swa_msg *cp = &__get_cpu_var(cpulast);

		if (unlikely(current_jiffies > cp->time))
			cp->count = 0;
		if (unlikely(cp->count < 5)) {
			cp->count++;
			cp->time = current_jiffies + 5 * HZ;

			/* minimize races by grabbing a copy of count BEFORE checking last.time. */
			count = last.count;
			barrier();

			/*
			 * Lower 4 bits are used as a count. Upper bits are a sequence
			 * number that is updated when count is reset. The cmpxchg will
			 * fail is seqno has changed. This minimizes mutiple cpus
			 * resetting the count.
			 */
			if (current_jiffies > last.time)
				(void) cmpxchg_acq(&last.count, count, 16 + (count & ~15));

			/* used fetchadd to atomically update the count */
			if ((last.count & 15) < 5 && (ia64_fetchadd(1, &last.count, acq) & 15) < 5) {
				last.time = current_jiffies + 5 * HZ;
				printk(KERN_WARNING
		       			"%s(%d): floating-point assist fault at ip %016lx, isr %016lx\n",
		       			current->comm, task_pid_nr(current), regs->cr_iip + ia64_psr(regs)->ri, isr);
			}
		}
	}

	exception = fp_emulate(fp_fault, bundle, &regs->cr_ipsr, &regs->ar_fpsr, &isr, &regs->pr,
			       &regs->cr_ifs, regs);
	if (fp_fault) {