fault.c

h内核

                return;
        }

/*
 * Oops. The kernel tried to access some bad page. We'll have to
 * terminate things with extreme prejudice.
 */
        if (user_address == 0)
                printk(KERN_ALERT "Unable to handle kernel pointer dereference"
        	       " at virtual kernel address %p\n", (void *)address);
        else
                printk(KERN_ALERT "Unable to handle kernel paging request"
		       " at virtual user address %p\n", (void *)address);

        die("Oops", regs, error_code);
        do_exit(SIGKILL);


/*
 * We ran out of memory, or some other thing happened to us that made
 * us unable to handle the page fault gracefully.
*/
out_of_memory:
	up_read(&mm->mmap_sem);
	if (tsk->pid == 1) {
		yield();
		goto survive;
	}
	printk("VM: killing process %s\n", tsk->comm);
	if (regs->psw.mask & PSW_MASK_PSTATE)
		do_exit(SIGKILL);
	goto no_context;

do_sigbus:
	up_read(&mm->mmap_sem);

	/*
	 * Send a sigbus, regardless of whether we were in kernel
	 * or user mode.
	 */
        tsk->thread.prot_addr = address;
        tsk->thread.trap_no = error_code;
	force_sig(SIGBUS, tsk);

	/* Kernel mode? Handle exceptions or die */
	if (!(regs->psw.mask & PSW_MASK_PSTATE))
		goto no_context;
}

void do_protection_exception(struct pt_regs *regs, unsigned long error_code)
{
	regs->psw.addr -= (error_code >> 16);
	do_exception(regs, 4, 1);
}

void do_dat_exception(struct pt_regs *regs, unsigned long error_code)
{
	do_exception(regs, error_code & 0xff, 0);
}

#ifndef CONFIG_ARCH_S390X

typedef struct _pseudo_wait_t {
       struct _pseudo_wait_t *next;
       wait_queue_head_t queue;
       unsigned long address;
       int resolved;
} pseudo_wait_t;

static pseudo_wait_t *pseudo_lock_queue = NULL;
static spinlock_t pseudo_wait_spinlock; /* spinlock to protect lock queue */

/*
 * This routine handles 'pagex' pseudo page faults.
 */
asmlinkage void
do_pseudo_page_fault(struct pt_regs *regs, unsigned long error_code)
{
        pseudo_wait_t wait_struct;
        pseudo_wait_t *ptr, *last, *next;
        unsigned long address;

        /*
         * get the failing address
         * more specific the segment and page table portion of
         * the address
         */
        address = S390_lowcore.trans_exc_code & 0xfffff000;

        if (address & 0x80000000) {
                /* high bit set -> a page has been swapped in by VM */
                address &= 0x7fffffff;
                spin_lock(&pseudo_wait_spinlock);
                last = NULL;
                ptr = pseudo_lock_queue;
                while (ptr != NULL) {
                        next = ptr->next;
                        if (address == ptr->address) {
				 /*
                                 * This is one of the processes waiting
                                 * for the page. Unchain from the queue.
                                 * There can be more than one process
                                 * waiting for the same page. VM presents
                                 * an initial and a completion interrupt for
                                 * every process that tries to access a 
                                 * page swapped out by VM. 
                                 */
                                if (last == NULL)
                                        pseudo_lock_queue = next;
                                else
                                        last->next = next;
                                /* now wake up the process */
                                ptr->resolved = 1;
                                wake_up(&ptr->queue);
                        } else
                                last = ptr;
                        ptr = next;
                }
                spin_unlock(&pseudo_wait_spinlock);
        } else {
                /* Pseudo page faults in kernel mode is a bad idea */
                if (!(regs->psw.mask & PSW_MASK_PSTATE)) {
                        /*
			 * VM presents pseudo page faults if the interrupted
			 * state was not disabled for interrupts. So we can
			 * get pseudo page fault interrupts while running
			 * in kernel mode. We simply access the page here
			 * while we are running disabled. VM will then swap
			 * in the page synchronously.
                         */
                         if (check_user_space(regs, error_code) == 0)
                                 /* dereference a virtual kernel address */
                                 __asm__ __volatile__ (
                                         "  ic 0,0(%0)"
                                         : : "a" (address) : "0");
                         else
                                 /* dereference a virtual user address */
                                 __asm__ __volatile__ (
                                         "  la   2,0(%0)\n"
                                         "  sacf 512\n"
                                         "  ic   2,0(2)\n"
					 "0:sacf 0\n"
					 ".section __ex_table,\"a\"\n"
					 "  .align 4\n"
					 "  .long  0b,0b\n"
					 ".previous"
                                         : : "a" (address) : "2" );

                        return;
                }
		/* initialize and add element to pseudo_lock_queue */
                init_waitqueue_head (&wait_struct.queue);
                wait_struct.address = address;
                wait_struct.resolved = 0;
                spin_lock(&pseudo_wait_spinlock);
                wait_struct.next = pseudo_lock_queue;
                pseudo_lock_queue = &wait_struct;
                spin_unlock(&pseudo_wait_spinlock);
		/*
		 * The instruction that caused the program check will
		 * be repeated. Don't signal single step via SIGTRAP.
		 */
		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
                /* go to sleep */
                wait_event(wait_struct.queue, wait_struct.resolved);
        }
}
#endif /* CONFIG_ARCH_S390X */

#ifdef CONFIG_PFAULT 
/*
 * 'pfault' pseudo page faults routines.
 */
static int pfault_disable = 0;

static int __init nopfault(char *str)
{
	pfault_disable = 1;
	return 1;
}

__setup("nopfault", nopfault);

typedef struct {
	__u16 refdiagc;
	__u16 reffcode;
	__u16 refdwlen;
	__u16 refversn;
	__u64 refgaddr;
	__u64 refselmk;
	__u64 refcmpmk;
	__u64 reserved;
} __attribute__ ((packed)) pfault_refbk_t;

int pfault_init(void)
{
	pfault_refbk_t refbk =
		{ 0x258, 0, 5, 2, __LC_CURRENT, 1ULL << 48, 1ULL << 48,
		  __PF_RES_FIELD };
        int rc;

	if (pfault_disable)
		return -1;
        __asm__ __volatile__(
                "    diag  %1,%0,0x258\n"
		"0:  j     2f\n"
		"1:  la    %0,8\n"
		"2:\n"
		".section __ex_table,\"a\"\n"
		"   .align 4\n"
#ifndef CONFIG_ARCH_S390X
		"   .long  0b,1b\n"
#else /* CONFIG_ARCH_S390X */
		"   .quad  0b,1b\n"
#endif /* CONFIG_ARCH_S390X */
		".previous"
                : "=d" (rc) : "a" (&refbk) : "cc" );
        __ctl_set_bit(0, 9);
        return rc;
}

void pfault_fini(void)
{
	pfault_refbk_t refbk =
	{ 0x258, 1, 5, 2, 0ULL, 0ULL, 0ULL, 0ULL };

	if (pfault_disable)
		return;
	__ctl_clear_bit(0,9);
        __asm__ __volatile__(
                "    diag  %0,0,0x258\n"
		"0:\n"
		".section __ex_table,\"a\"\n"
		"   .align 4\n"
#ifndef CONFIG_ARCH_S390X
		"   .long  0b,0b\n"
#else /* CONFIG_ARCH_S390X */
		"   .quad  0b,0b\n"
#endif /* CONFIG_ARCH_S390X */
		".previous"
		: : "a" (&refbk) : "cc" );
}

asmlinkage void
pfault_interrupt(struct pt_regs *regs, __u16 error_code)
{
	struct task_struct *tsk;
	__u16 subcode;

	/*
	 * Get the external interruption subcode & pfault
	 * initial/completion signal bit. VM stores this 
	 * in the 'cpu address' field associated with the
         * external interrupt. 
	 */
	subcode = S390_lowcore.cpu_addr;
	if ((subcode & 0xff00) != __SUBCODE_MASK)
		return;

	/*
	 * Get the token (= address of the task structure of the affected task).
	 */
	tsk = *(struct task_struct **) __LC_PFAULT_INTPARM;

	if (subcode & 0x0080) {
		/* signal bit is set -> a page has been swapped in by VM */
		if (xchg(&tsk->thread.pfault_wait, -1) != 0) {
			/* Initial interrupt was faster than the completion
			 * interrupt. pfault_wait is valid. Set pfault_wait
			 * back to zero and wake up the process. This can
			 * safely be done because the task is still sleeping
			 * and can't procude new pfaults. */
			tsk->thread.pfault_wait = 0;
			wake_up_process(tsk);
		}
	} else {
		/* signal bit not set -> a real page is missing. */
		set_task_state(tsk, TASK_UNINTERRUPTIBLE);
		if (xchg(&tsk->thread.pfault_wait, 1) != 0) {
			/* Completion interrupt was faster than the initial
			 * interrupt (swapped in a -1 for pfault_wait). Set
			 * pfault_wait back to zero and exit. This can be
			 * done safely because tsk is running in kernel 
			 * mode and can't produce new pfaults. */
			tsk->thread.pfault_wait = 0;
			set_task_state(tsk, TASK_RUNNING);
		} else
			set_tsk_need_resched(tsk);
	}
}
#endif