[deliverable/linux.git] / arch / s390 / mm / fault.c

/*
 *  arch/s390/mm/fault.c
 *
 *  S390 version
 *    Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
 *    Author(s): Hartmut Penner (hp@de.ibm.com)
 *               Ulrich Weigand (uweigand@de.ibm.com)
 *
 *  Derived from "arch/i386/mm/fault.c"
 *    Copyright (C) 1995  Linus Torvalds
 */

#include <linux/signal.h>
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/errno.h>
#include <linux/string.h>
#include <linux/types.h>
#include <linux/ptrace.h>
#include <linux/mman.h>
#include <linux/mm.h>
#include <linux/smp.h>
#include <linux/kdebug.h>
#include <linux/smp_lock.h>
#include <linux/init.h>
#include <linux/console.h>
#include <linux/module.h>
#include <linux/hardirq.h>
#include <linux/kprobes.h>
#include <linux/uaccess.h>

#include <asm/system.h>
#include <asm/pgtable.h>
#include <asm/s390_ext.h>

#ifndef CONFIG_64BIT
#define __FAIL_ADDR_MASK 0x7ffff000
#define __FIXUP_MASK 0x7fffffff
#define __SUBCODE_MASK 0x0200
#define __PF_RES_FIELD 0ULL
#else /* CONFIG_64BIT */
#define __FAIL_ADDR_MASK -4096L
#define __FIXUP_MASK ~0L
#define __SUBCODE_MASK 0x0600
#define __PF_RES_FIELD 0x8000000000000000ULL
#endif /* CONFIG_64BIT */

#ifdef CONFIG_SYSCTL
extern int sysctl_userprocess_debug;
#endif

extern void die(const char *,struct pt_regs *,long);

#ifdef CONFIG_KPROBES
static inline int notify_page_fault(struct pt_regs *regs, long err)
{
	int ret = 0;

	/* kprobe_running() needs smp_processor_id() */
	if (!user_mode(regs)) {
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 14))
			ret = 1;
		preempt_enable();
	}

	return ret;
}
#else
static inline int notify_page_fault(struct pt_regs *regs, long err)
{
	return 0;
}
#endif


/*
 * Unlock any spinlocks which will prevent us from getting the
 * message out.
 */
void bust_spinlocks(int yes)
{
	if (yes) {
		oops_in_progress = 1;
	} else {
		int loglevel_save = console_loglevel;
		console_unblank();
		oops_in_progress = 0;
		/*
		 * OK, the message is on the console.  Now we call printk()
		 * without oops_in_progress set so that printk will give klogd
		 * a poke.  Hold onto your hats...
		 */
		console_loglevel = 15;
		printk(" ");
		console_loglevel = loglevel_save;
	}
}

/*
 * Returns the address space associated with the fault.
 * Returns 0 for kernel space, 1 for user space and
 * 2 for code execution in user space with noexec=on.
 */
static inline int check_space(struct task_struct *tsk)
{
	/*
	 * The lowest two bits of S390_lowcore.trans_exc_code
	 * indicate which paging table was used.
	 */
	int desc = S390_lowcore.trans_exc_code & 3;

	if (desc == 3)	/* Home Segment Table Descriptor */
		return switch_amode == 0;
	if (desc == 2)	/* Secondary Segment Table Descriptor */
		return tsk->thread.mm_segment.ar4;
#ifdef CONFIG_S390_SWITCH_AMODE
	if (unlikely(desc == 1)) { /* STD determined via access register */
		/* %a0 always indicates primary space. */
		if (S390_lowcore.exc_access_id != 0) {
			save_access_regs(tsk->thread.acrs);
			/*
			 * An alet of 0 indicates primary space.
			 * An alet of 1 indicates secondary space.
			 * Any other alet values generate an
			 * alen-translation exception.
			 */
			if (tsk->thread.acrs[S390_lowcore.exc_access_id])
				return tsk->thread.mm_segment.ar4;
		}
	}
#endif
	/* Primary Segment Table Descriptor */
	return switch_amode << s390_noexec;
}

/*
 * Send SIGSEGV to task.  This is an external routine
 * to keep the stack usage of do_page_fault small.
 */
static void do_sigsegv(struct pt_regs *regs, unsigned long error_code,
		       int si_code, unsigned long address)
{
	struct siginfo si;

#if defined(CONFIG_SYSCTL) || defined(CONFIG_PROCESS_DEBUG)
#if defined(CONFIG_SYSCTL)
	if (sysctl_userprocess_debug)
#endif
	{
		printk("User process fault: interruption code 0x%lX\n",
		       error_code);
		printk("failing address: %lX\n", address);
		show_regs(regs);
	}
#endif
	si.si_signo = SIGSEGV;
	si.si_code = si_code;
	si.si_addr = (void __user *) address;
	force_sig_info(SIGSEGV, &si, current);
}

static void do_no_context(struct pt_regs *regs, unsigned long error_code,
			  unsigned long address)
{
	const struct exception_table_entry *fixup;

	/* Are we prepared to handle this kernel fault?  */
	fixup = search_exception_tables(regs->psw.addr & __FIXUP_MASK);
	if (fixup) {
		regs->psw.addr = fixup->fixup | PSW_ADDR_AMODE;
		return;
	}

	/*
	 * Oops. The kernel tried to access some bad page. We'll have to
	 * terminate things with extreme prejudice.
	 */
	if (check_space(current) == 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);
}

static void do_low_address(struct pt_regs *regs, unsigned long error_code)
{
	/* Low-address protection hit in kernel mode means
	   NULL pointer write access in kernel mode.  */
	if (regs->psw.mask & PSW_MASK_PSTATE) {
		/* Low-address protection hit in user mode 'cannot happen'. */
		die ("Low-address protection", regs, error_code);
		do_exit(SIGKILL);
	}

	do_no_context(regs, error_code, 0);
}

/*
 * We ran out of memory, or some other thing happened to us that made
 * us unable to handle the page fault gracefully.
 */
static int do_out_of_memory(struct pt_regs *regs, unsigned long error_code,
			    unsigned long address)
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	up_read(&mm->mmap_sem);
	if (is_init(tsk)) {
		yield();
		down_read(&mm->mmap_sem);
		return 1;
	}
	printk("VM: killing process %s\n", tsk->comm);
	if (regs->psw.mask & PSW_MASK_PSTATE)
		do_exit(SIGKILL);
	do_no_context(regs, error_code, address);
	return 0;
}

static void do_sigbus(struct pt_regs *regs, unsigned long error_code,
		      unsigned long address)
{
	struct task_struct *tsk = current;
	struct mm_struct *mm = tsk->mm;

	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))
		do_no_context(regs, error_code, address);
}

#ifdef CONFIG_S390_EXEC_PROTECT
extern long sys_sigreturn(struct pt_regs *regs);
extern long sys_rt_sigreturn(struct pt_regs *regs);
extern long sys32_sigreturn(struct pt_regs *regs);
extern long sys32_rt_sigreturn(struct pt_regs *regs);

static int signal_return(struct mm_struct *mm, struct pt_regs *regs,
			 unsigned long address, unsigned long error_code)
{
	u16 instruction;
	int rc;
#ifdef CONFIG_COMPAT
	int compat;
#endif

	pagefault_disable();
	rc = __get_user(instruction, (u16 __user *) regs->psw.addr);
	pagefault_enable();
	if (rc)
		return -EFAULT;

	up_read(&mm->mmap_sem);
	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
#ifdef CONFIG_COMPAT
	compat = test_tsk_thread_flag(current, TIF_31BIT);
	if (compat && instruction == 0x0a77)
		sys32_sigreturn(regs);
	else if (compat && instruction == 0x0aad)
		sys32_rt_sigreturn(regs);
	else
#endif
	if (instruction == 0x0a77)
		sys_sigreturn(regs);
	else if (instruction == 0x0aad)
		sys_rt_sigreturn(regs);
	else {
		current->thread.prot_addr = address;
		current->thread.trap_no = error_code;
		do_sigsegv(regs, error_code, SEGV_MAPERR, address);
	}
	return 0;
}
#endif /* CONFIG_S390_EXEC_PROTECT */

/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 *
 * error_code:
 *   04       Protection           ->  Write-Protection  (suprression)
 *   10       Segment translation  ->  Not present       (nullification)
 *   11       Page translation     ->  Not present       (nullification)
 *   3b       Region third trans.  ->  Not present       (nullification)
 */
static inline void
do_exception(struct pt_regs *regs, unsigned long error_code, int write)
{
	struct task_struct *tsk;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
	unsigned long address;
	int space;
	int si_code;
	int fault;

	if (notify_page_fault(regs, error_code))
		return;

	tsk = current;
	mm = tsk->mm;

	/* get the failing address and the affected space */
	address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
	space = check_space(tsk);

	/*
	 * Verify that the fault happened in user space, that
	 * we are not in an interrupt and that there is a 
	 * user context.
	 */
	if (unlikely(space == 0 || in_atomic() || !mm))
		goto no_context;

	/*
	 * When we get here, the fault happened in the current
	 * task's user address space, so we can switch on the
	 * interrupts again and then search the VMAs
	 */
	local_irq_enable();

	down_read(&mm->mmap_sem);

	si_code = SEGV_MAPERR;
	vma = find_vma(mm, address);
	if (!vma)
		goto bad_area;

#ifdef CONFIG_S390_EXEC_PROTECT
	if (unlikely((space == 2) && !(vma->vm_flags & VM_EXEC)))
		if (!signal_return(mm, regs, address, error_code))
			/*
			 * signal_return() has done an up_read(&mm->mmap_sem)
			 * if it returns 0.
			 */
			return;
#endif

	if (vma->vm_start <= address)
		goto good_area;
	if (!(vma->vm_flags & VM_GROWSDOWN))
		goto bad_area;
	if (expand_stack(vma, address))
		goto bad_area;
/*
 * Ok, we have a good vm_area for this memory access, so
 * we can handle it..
 */
good_area:
	si_code = SEGV_ACCERR;
	if (!write) {
		/* page not present, check vm flags */
		if (!(vma->vm_flags & (VM_READ | VM_EXEC | VM_WRITE)))
			goto bad_area;
	} else {
		if (!(vma->vm_flags & VM_WRITE))
			goto bad_area;
	}

survive:
	/*
	 * If for any reason at all we couldn't handle the fault,
	 * make sure we exit gracefully rather than endlessly redo
	 * the fault.
	 */
	fault = handle_mm_fault(mm, vma, address, write);
	if (unlikely(fault & VM_FAULT_ERROR)) {
		if (fault & VM_FAULT_OOM) {
			if (do_out_of_memory(regs, error_code, address))
				goto survive;
			return;
		} else if (fault & VM_FAULT_SIGBUS) {
			do_sigbus(regs, error_code, address);
			return;
		}
		BUG();
	}
	if (fault & VM_FAULT_MAJOR)
		tsk->maj_flt++;
	else
		tsk->min_flt++;

        up_read(&mm->mmap_sem);
	/*
	 * The instruction that caused the program check will
	 * be repeated. Don't signal single step via SIGTRAP.
	 */
	clear_tsk_thread_flag(tsk, TIF_SINGLE_STEP);
        return;

/*
 * Something tried to access memory that isn't in our memory map..
 * Fix it, but check if it's kernel or user first..
 */
bad_area:
	up_read(&mm->mmap_sem);

	/* User mode accesses just cause a SIGSEGV */
	if (regs->psw.mask & PSW_MASK_PSTATE) {
		tsk->thread.prot_addr = address;
		tsk->thread.trap_no = error_code;
		do_sigsegv(regs, error_code, si_code, address);
		return;
	}

no_context:
	do_no_context(regs, error_code, address);
}

void __kprobes do_protection_exception(struct pt_regs *regs,
				       unsigned long error_code)
{
	/* Protection exception is supressing, decrement psw address. */
	regs->psw.addr -= (error_code >> 16);
	/*
	 * Check for low-address protection.  This needs to be treated
	 * as a special case because the translation exception code
	 * field is not guaranteed to contain valid data in this case.
	 */
	if (unlikely(!(S390_lowcore.trans_exc_code & 4))) {
		do_low_address(regs, error_code);
		return;
	}
	do_exception(regs, 4, 1);
}

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

#ifdef CONFIG_PFAULT 
/*
 * 'pfault' pseudo page faults routines.
 */
static ext_int_info_t ext_int_pfault;
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 (!MACHINE_IS_VM || pfault_disable)
		return -1;
	asm volatile(
		"	diag	%1,%0,0x258\n"
		"0:	j	2f\n"
		"1:	la	%0,8\n"
		"2:\n"
		EX_TABLE(0b,1b)
		: "=d" (rc) : "a" (&refbk), "m" (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 (!MACHINE_IS_VM || pfault_disable)
		return;
	__ctl_clear_bit(0,9);
	asm volatile(
		"	diag	%0,0,0x258\n"
		"0:\n"
		EX_TABLE(0b,0b)
		: : "a" (&refbk), "m" (refbk) : "cc");
}

static void pfault_interrupt(__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 produce new pfaults. */
			tsk->thread.pfault_wait = 0;
			wake_up_process(tsk);
			put_task_struct(tsk);
		}
	} else {
		/* signal bit not set -> a real page is missing. */
		get_task_struct(tsk);
		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);
			put_task_struct(tsk);
		} else
			set_tsk_need_resched(tsk);
	}
}

void __init pfault_irq_init(void)
{
	if (!MACHINE_IS_VM)
		return;

	/*
	 * Try to get pfault pseudo page faults going.
	 */
	if (register_early_external_interrupt(0x2603, pfault_interrupt,
					      &ext_int_pfault) != 0)
		panic("Couldn't request external interrupt 0x2603");

	if (pfault_init() == 0)
		return;

	/* Tough luck, no pfault. */
	pfault_disable = 1;
	unregister_early_external_interrupt(0x2603, pfault_interrupt,
					    &ext_int_pfault);
}
#endif
Commit	Line	Data
1da177e4 LT	1	/*
	2	* arch/s390/mm/fault.c
	3	*
	4	* S390 version
	5	* Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
	6	* Author(s): Hartmut Penner (hp@de.ibm.com)
	7	* Ulrich Weigand (uweigand@de.ibm.com)
	8	*
	9	* Derived from "arch/i386/mm/fault.c"
	10	* Copyright (C) 1995 Linus Torvalds
	11	*/
	12
1da177e4 LT	13	#include <linux/signal.h>
	14	#include <linux/sched.h>
	15	#include <linux/kernel.h>
	16	#include <linux/errno.h>
	17	#include <linux/string.h>
	18	#include <linux/types.h>
	19	#include <linux/ptrace.h>
	20	#include <linux/mman.h>
	21	#include <linux/mm.h>
	22	#include <linux/smp.h>
1eeb66a1	23	#include <linux/kdebug.h>
1da177e4 LT	24	#include <linux/smp_lock.h>
	25	#include <linux/init.h>
	26	#include <linux/console.h>
	27	#include <linux/module.h>
	28	#include <linux/hardirq.h>
4ba069b8	29	#include <linux/kprobes.h>
be5ec363	30	#include <linux/uaccess.h>
1da177e4 LT	31
1da177e4 LT	32	#include <asm/system.h>
1da177e4	33	#include <asm/pgtable.h>
29b08d2b	34	#include <asm/s390_ext.h>
1da177e4	35
347a8dc3	36	#ifndef CONFIG_64BIT
1da177e4 LT	37	#define __FAIL_ADDR_MASK 0x7ffff000
	38	#define __FIXUP_MASK 0x7fffffff
	39	#define __SUBCODE_MASK 0x0200
	40	#define __PF_RES_FIELD 0ULL
347a8dc3	41	#else /* CONFIG_64BIT */
1da177e4 LT	42	#define __FAIL_ADDR_MASK -4096L
	43	#define __FIXUP_MASK ~0L
	44	#define __SUBCODE_MASK 0x0600
	45	#define __PF_RES_FIELD 0x8000000000000000ULL
347a8dc3	46	#endif /* CONFIG_64BIT */
1da177e4 LT	47
	48	#ifdef CONFIG_SYSCTL
	49	extern int sysctl_userprocess_debug;
	50	#endif
	51
	52	extern void die(const char ,struct pt_regs ,long);
	53
4ba069b8	54	#ifdef CONFIG_KPROBES
10c1031f MS	55	static inline int notify_page_fault(struct pt_regs *regs, long err)
10c1031f MS	56	{
33464e3b CH	57	int ret = 0;
	58
	59	/* kprobe_running() needs smp_processor_id() */
	60	if (!user_mode(regs)) {
	61	preempt_disable();
	62	if (kprobe_running() && kprobe_fault_handler(regs, 14))
	63	ret = 1;
	64	preempt_enable();
	65	}
	66
	67	return ret;
4ba069b8 MG	68	}
4ba069b8 MG	69	#else
10c1031f	70	static inline int notify_page_fault(struct pt_regs *regs, long err)
4ba069b8	71	{
33464e3b	72	return 0;
4ba069b8 MG	73	}
	74	#endif
	75
1da177e4 LT	76
	77	/*
	78	* Unlock any spinlocks which will prevent us from getting the
cefc8be8	79	* message out.
1da177e4 LT	80	*/
	81	void bust_spinlocks(int yes)
	82	{
	83	if (yes) {
	84	oops_in_progress = 1;
	85	} else {
	86	int loglevel_save = console_loglevel;
	87	console_unblank();
	88	oops_in_progress = 0;
	89	/*
	90	* OK, the message is on the console. Now we call printk()
	91	* without oops_in_progress set so that printk will give klogd
	92	* a poke. Hold onto your hats...
	93	*/
	94	console_loglevel = 15;
	95	printk(" ");
	96	console_loglevel = loglevel_save;
	97	}
	98	}
	99
	100	/*
482b05dd GS	101	* Returns the address space associated with the fault.
	102	* Returns 0 for kernel space, 1 for user space and
	103	* 2 for code execution in user space with noexec=on.
1da177e4	104	*/
482b05dd	105	static inline int check_space(struct task_struct *tsk)
1da177e4 LT	106	{
1da177e4 LT	107	/*
482b05dd GS	108	* The lowest two bits of S390_lowcore.trans_exc_code
482b05dd GS	109	* indicate which paging table was used.
1da177e4	110	*/
482b05dd GS	111	int desc = S390_lowcore.trans_exc_code & 3;
	112
	113	if (desc == 3) /* Home Segment Table Descriptor */
	114	return switch_amode == 0;
	115	if (desc == 2) /* Secondary Segment Table Descriptor */
	116	return tsk->thread.mm_segment.ar4;
	117	#ifdef CONFIG_S390_SWITCH_AMODE
	118	if (unlikely(desc == 1)) { /* STD determined via access register */
	119	/* %a0 always indicates primary space. */
	120	if (S390_lowcore.exc_access_id != 0) {
	121	save_access_regs(tsk->thread.acrs);
	122	/*
	123	* An alet of 0 indicates primary space.
	124	* An alet of 1 indicates secondary space.
	125	* Any other alet values generate an
	126	* alen-translation exception.
	127	*/
	128	if (tsk->thread.acrs[S390_lowcore.exc_access_id])
	129	return tsk->thread.mm_segment.ar4;
	130	}
	131	}
	132	#endif
	133	/* Primary Segment Table Descriptor */
	134	return switch_amode << s390_noexec;
1da177e4 LT	135	}
	136
	137	/*
	138	* Send SIGSEGV to task. This is an external routine
	139	* to keep the stack usage of do_page_fault small.
	140	*/
	141	static void do_sigsegv(struct pt_regs *regs, unsigned long error_code,
	142	int si_code, unsigned long address)
	143	{
	144	struct siginfo si;
	145
	146	#if defined(CONFIG_SYSCTL) \|\| defined(CONFIG_PROCESS_DEBUG)
	147	#if defined(CONFIG_SYSCTL)
	148	if (sysctl_userprocess_debug)
	149	#endif
	150	{
	151	printk("User process fault: interruption code 0x%lX\n",
	152	error_code);
	153	printk("failing address: %lX\n", address);
	154	show_regs(regs);
	155	}
	156	#endif
	157	si.si_signo = SIGSEGV;
	158	si.si_code = si_code;
d2c993d8	159	si.si_addr = (void __user *) address;
1da177e4 LT	160	force_sig_info(SIGSEGV, &si, current);
	161	}
	162
10c1031f MS	163	static void do_no_context(struct pt_regs *regs, unsigned long error_code,
	164	unsigned long address)
	165	{
	166	const struct exception_table_entry *fixup;
	167
	168	/* Are we prepared to handle this kernel fault? */
	169	fixup = search_exception_tables(regs->psw.addr & __FIXUP_MASK);
	170	if (fixup) {
	171	regs->psw.addr = fixup->fixup \| PSW_ADDR_AMODE;
	172	return;
	173	}
	174
	175	/*
	176	* Oops. The kernel tried to access some bad page. We'll have to
	177	* terminate things with extreme prejudice.
	178	*/
	179	if (check_space(current) == 0)
	180	printk(KERN_ALERT "Unable to handle kernel pointer dereference"
	181	" at virtual kernel address %p\n", (void *)address);
	182	else
	183	printk(KERN_ALERT "Unable to handle kernel paging request"
	184	" at virtual user address %p\n", (void *)address);
	185
	186	die("Oops", regs, error_code);
	187	do_exit(SIGKILL);
	188	}
	189
	190	static void do_low_address(struct pt_regs *regs, unsigned long error_code)
	191	{
	192	/* Low-address protection hit in kernel mode means
	193	NULL pointer write access in kernel mode. */
	194	if (regs->psw.mask & PSW_MASK_PSTATE) {
	195	/* Low-address protection hit in user mode 'cannot happen'. */
	196	die ("Low-address protection", regs, error_code);
	197	do_exit(SIGKILL);
	198	}
	199
	200	do_no_context(regs, error_code, 0);
	201	}
	202
	203	/*
	204	* We ran out of memory, or some other thing happened to us that made
	205	* us unable to handle the page fault gracefully.
	206	*/
	207	static int do_out_of_memory(struct pt_regs *regs, unsigned long error_code,
	208	unsigned long address)
	209	{
	210	struct task_struct *tsk = current;
	211	struct mm_struct *mm = tsk->mm;
	212
	213	up_read(&mm->mmap_sem);
	214	if (is_init(tsk)) {
	215	yield();
	216	down_read(&mm->mmap_sem);
	217	return 1;
	218	}
	219	printk("VM: killing process %s\n", tsk->comm);
	220	if (regs->psw.mask & PSW_MASK_PSTATE)
	221	do_exit(SIGKILL);
	222	do_no_context(regs, error_code, address);
	223	return 0;
	224	}
	225
	226	static void do_sigbus(struct pt_regs *regs, unsigned long error_code,
227	unsigned long address)
228	{
229	struct task_struct *tsk = current;
230	struct mm_struct *mm = tsk->mm;
231
232	up_read(&mm->mmap_sem);
233	/*
234	* Send a sigbus, regardless of whether we were in kernel
235	* or user mode.
236	*/
237	tsk->thread.prot_addr = address;
238	tsk->thread.trap_no = error_code;
239	force_sig(SIGBUS, tsk);
240
241	/* Kernel mode? Handle exceptions or die */
242	if (!(regs->psw.mask & PSW_MASK_PSTATE))
243	do_no_context(regs, error_code, address);
244	}
245
c1821c2e GS	246	#ifdef CONFIG_S390_EXEC_PROTECT
	247	extern long sys_sigreturn(struct pt_regs *regs);
	248	extern long sys_rt_sigreturn(struct pt_regs *regs);
	249	extern long sys32_sigreturn(struct pt_regs *regs);
	250	extern long sys32_rt_sigreturn(struct pt_regs *regs);
	251
be5ec363 MS	252	static int signal_return(struct mm_struct mm, struct pt_regs regs,
be5ec363 MS	253	unsigned long address, unsigned long error_code)
c1821c2e	254	{
be5ec363	255	u16 instruction;
490f03d6 HC	256	int rc;
	257	#ifdef CONFIG_COMPAT
	258	int compat;
	259	#endif
be5ec363 MS	260
	261	pagefault_disable();
	262	rc = __get_user(instruction, (u16 __user *) regs->psw.addr);
	263	pagefault_enable();
	264	if (rc)
	265	return -EFAULT;
	266
c1821c2e GS	267	up_read(&mm->mmap_sem);
	268	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
	269	#ifdef CONFIG_COMPAT
be5ec363 MS	270	compat = test_tsk_thread_flag(current, TIF_31BIT);
	271	if (compat && instruction == 0x0a77)
	272	sys32_sigreturn(regs);
	273	else if (compat && instruction == 0x0aad)
	274	sys32_rt_sigreturn(regs);
c1821c2e	275	else
be5ec363 MS	276	#endif
be5ec363 MS	277	if (instruction == 0x0a77)
c1821c2e	278	sys_sigreturn(regs);
be5ec363 MS	279	else if (instruction == 0x0aad)
be5ec363 MS	280	sys_rt_sigreturn(regs);
c1821c2e	281	else {
c1821c2e GS	282	current->thread.prot_addr = address;
	283	current->thread.trap_no = error_code;
	284	do_sigsegv(regs, error_code, SEGV_MAPERR, address);
	285	}
	286	return 0;
c1821c2e GS	287	}
	288	#endif /* CONFIG_S390_EXEC_PROTECT */
	289
1da177e4 LT	290	/*
	291	* This routine handles page faults. It determines the address,
	292	* and the problem, and then passes it off to one of the appropriate
	293	* routines.
	294	*
	295	* error_code:
	296	* 04 Protection -> Write-Protection (suprression)
	297	* 10 Segment translation -> Not present (nullification)
	298	* 11 Page translation -> Not present (nullification)
	299	* 3b Region third trans. -> Not present (nullification)
	300	*/
482b05dd	301	static inline void
10c1031f	302	do_exception(struct pt_regs *regs, unsigned long error_code, int write)
1da177e4	303	{
10c1031f MS	304	struct task_struct *tsk;
	305	struct mm_struct *mm;
	306	struct vm_area_struct *vma;
	307	unsigned long address;
482b05dd	308	int space;
10c1031f	309	int si_code;
83c54070	310	int fault;
1da177e4	311
33464e3b	312	if (notify_page_fault(regs, error_code))
4ba069b8 MG	313	return;
4ba069b8 MG	314
10c1031f MS	315	tsk = current;
10c1031f MS	316	mm = tsk->mm;
1da177e4	317
10c1031f MS	318	/* get the failing address and the affected space */
10c1031f MS	319	address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
482b05dd	320	space = check_space(tsk);
1da177e4 LT	321
	322	/*
	323	* Verify that the fault happened in user space, that
	324	* we are not in an interrupt and that there is a
	325	* user context.
	326	*/
482b05dd GS	327	if (unlikely(space == 0 \|\| in_atomic() \|\| !mm))
482b05dd GS	328	goto no_context;
1da177e4 LT	329
	330	/*
	331	* When we get here, the fault happened in the current
	332	* task's user address space, so we can switch on the
	333	* interrupts again and then search the VMAs
	334	*/
	335	local_irq_enable();
	336
10c1031f	337	down_read(&mm->mmap_sem);
1da177e4	338
482b05dd GS	339	si_code = SEGV_MAPERR;
	340	vma = find_vma(mm, address);
	341	if (!vma)
	342	goto bad_area;
c1821c2e GS	343
c1821c2e GS	344	#ifdef CONFIG_S390_EXEC_PROTECT
482b05dd	345	if (unlikely((space == 2) && !(vma->vm_flags & VM_EXEC)))
c1821c2e GS	346	if (!signal_return(mm, regs, address, error_code))
	347	/*
	348	* signal_return() has done an up_read(&mm->mmap_sem)
	349	* if it returns 0.
	350	*/
	351	return;
	352	#endif
	353
10c1031f MS	354	if (vma->vm_start <= address)
	355	goto good_area;
	356	if (!(vma->vm_flags & VM_GROWSDOWN))
	357	goto bad_area;
	358	if (expand_stack(vma, address))
	359	goto bad_area;
1da177e4 LT	360	/*
	361	* Ok, we have a good vm_area for this memory access, so
	362	* we can handle it..
	363	*/
	364	good_area:
	365	si_code = SEGV_ACCERR;
10c1031f	366	if (!write) {
1da177e4 LT	367	/* page not present, check vm flags */
	368	if (!(vma->vm_flags & (VM_READ \| VM_EXEC \| VM_WRITE)))
	369	goto bad_area;
	370	} else {
	371	if (!(vma->vm_flags & VM_WRITE))
	372	goto bad_area;
	373	}
	374
	375	survive:
	376	/*
	377	* If for any reason at all we couldn't handle the fault,
	378	* make sure we exit gracefully rather than endlessly redo
	379	* the fault.
	380	*/
83c54070 NP	381	fault = handle_mm_fault(mm, vma, address, write);
	382	if (unlikely(fault & VM_FAULT_ERROR)) {
	383	if (fault & VM_FAULT_OOM) {
	384	if (do_out_of_memory(regs, error_code, address))
	385	goto survive;
	386	return;
	387	} else if (fault & VM_FAULT_SIGBUS) {
	388	do_sigbus(regs, error_code, address);
	389	return;
	390	}
1da177e4 LT	391	BUG();
1da177e4 LT	392	}
83c54070 NP	393	if (fault & VM_FAULT_MAJOR)
	394	tsk->maj_flt++;
	395	else
	396	tsk->min_flt++;
1da177e4 LT	397
	398	up_read(&mm->mmap_sem);
	399	/*
	400	* The instruction that caused the program check will
	401	* be repeated. Don't signal single step via SIGTRAP.
	402	*/
482b05dd	403	clear_tsk_thread_flag(tsk, TIF_SINGLE_STEP);
1da177e4 LT	404	return;
	405
	406	/*
	407	* Something tried to access memory that isn't in our memory map..
	408	* Fix it, but check if it's kernel or user first..
	409	*/
	410	bad_area:
10c1031f	411	up_read(&mm->mmap_sem);
1da177e4	412
10c1031f MS	413	/* User mode accesses just cause a SIGSEGV */
	414	if (regs->psw.mask & PSW_MASK_PSTATE) {
	415	tsk->thread.prot_addr = address;
	416	tsk->thread.trap_no = error_code;
1da177e4	417	do_sigsegv(regs, error_code, si_code, address);
10c1031f	418	return;
1da177e4 LT	419	}
	420
	421	no_context:
10c1031f	422	do_no_context(regs, error_code, address);
1da177e4 LT	423	}
1da177e4 LT	424
482b05dd GS	425	void __kprobes do_protection_exception(struct pt_regs *regs,
482b05dd GS	426	unsigned long error_code)
1da177e4	427	{
10c1031f	428	/* Protection exception is supressing, decrement psw address. */
1da177e4	429	regs->psw.addr -= (error_code >> 16);
10c1031f MS	430	/*
	431	* Check for low-address protection. This needs to be treated
	432	* as a special case because the translation exception code
	433	* field is not guaranteed to contain valid data in this case.
	434	*/
	435	if (unlikely(!(S390_lowcore.trans_exc_code & 4))) {
	436	do_low_address(regs, error_code);
	437	return;
	438	}
1da177e4 LT	439	do_exception(regs, 4, 1);
	440	}
	441
482b05dd	442	void __kprobes do_dat_exception(struct pt_regs *regs, unsigned long error_code)
1da177e4 LT	443	{
	444	do_exception(regs, error_code & 0xff, 0);
	445	}
	446
1da177e4 LT	447	#ifdef CONFIG_PFAULT
	448	/*
	449	* 'pfault' pseudo page faults routines.
	450	*/
29b08d2b	451	static ext_int_info_t ext_int_pfault;
1da177e4 LT	452	static int pfault_disable = 0;
	453
	454	static int __init nopfault(char *str)
	455	{
	456	pfault_disable = 1;
	457	return 1;
	458	}
	459
	460	__setup("nopfault", nopfault);
	461
	462	typedef struct {
	463	__u16 refdiagc;
	464	__u16 reffcode;
	465	__u16 refdwlen;
	466	__u16 refversn;
	467	__u64 refgaddr;
	468	__u64 refselmk;
	469	__u64 refcmpmk;
	470	__u64 reserved;
	471	} __attribute__ ((packed)) pfault_refbk_t;
	472
	473	int pfault_init(void)
	474	{
	475	pfault_refbk_t refbk =
	476	{ 0x258, 0, 5, 2, __LC_CURRENT, 1ULL << 48, 1ULL << 48,
	477	__PF_RES_FIELD };
	478	int rc;
	479
29b08d2b	480	if (!MACHINE_IS_VM \|\| pfault_disable)
1da177e4	481	return -1;
94c12cc7 MS	482	asm volatile(
	483	" diag %1,%0,0x258\n"
	484	"0: j 2f\n"
	485	"1: la %0,8\n"
1da177e4	486	"2:\n"
94c12cc7 MS	487	EX_TABLE(0b,1b)
94c12cc7 MS	488	: "=d" (rc) : "a" (&refbk), "m" (refbk) : "cc");
1da177e4 LT	489	__ctl_set_bit(0, 9);
	490	return rc;
	491	}
	492
	493	void pfault_fini(void)
	494	{
	495	pfault_refbk_t refbk =
	496	{ 0x258, 1, 5, 2, 0ULL, 0ULL, 0ULL, 0ULL };
	497
29b08d2b	498	if (!MACHINE_IS_VM \|\| pfault_disable)
1da177e4 LT	499	return;
1da177e4 LT	500	__ctl_clear_bit(0,9);
94c12cc7 MS	501	asm volatile(
94c12cc7 MS	502	" diag %0,0,0x258\n"
1da177e4	503	"0:\n"
94c12cc7 MS	504	EX_TABLE(0b,0b)
94c12cc7 MS	505	: : "a" (&refbk), "m" (refbk) : "cc");
1da177e4 LT	506	}
1da177e4 LT	507
2b67fc46	508	static void pfault_interrupt(__u16 error_code)
1da177e4 LT	509	{
	510	struct task_struct *tsk;
	511	__u16 subcode;
	512
	513	/*
	514	* Get the external interruption subcode & pfault
	515	* initial/completion signal bit. VM stores this
	516	* in the 'cpu address' field associated with the
	517	* external interrupt.
	518	*/
	519	subcode = S390_lowcore.cpu_addr;
	520	if ((subcode & 0xff00) != __SUBCODE_MASK)
	521	return;
	522
	523	/*
	524	* Get the token (= address of the task structure of the affected task).
	525	*/
	526	tsk = (struct task_struct *) __LC_PFAULT_INTPARM;
	527
	528	if (subcode & 0x0080) {
	529	/* signal bit is set -> a page has been swapped in by VM */
	530	if (xchg(&tsk->thread.pfault_wait, -1) != 0) {
	531	/* Initial interrupt was faster than the completion
	532	* interrupt. pfault_wait is valid. Set pfault_wait
	533	* back to zero and wake up the process. This can
	534	* safely be done because the task is still sleeping
b6d09449	535	* and can't produce new pfaults. */
1da177e4 LT	536	tsk->thread.pfault_wait = 0;
1da177e4 LT	537	wake_up_process(tsk);
b6d09449	538	put_task_struct(tsk);
1da177e4 LT	539	}
	540	} else {
	541	/* signal bit not set -> a real page is missing. */
b6d09449	542	get_task_struct(tsk);
1da177e4 LT	543	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	544	if (xchg(&tsk->thread.pfault_wait, 1) != 0) {
	545	/* Completion interrupt was faster than the initial
	546	* interrupt (swapped in a -1 for pfault_wait). Set
	547	* pfault_wait back to zero and exit. This can be
	548	* done safely because tsk is running in kernel
	549	* mode and can't produce new pfaults. */
	550	tsk->thread.pfault_wait = 0;
	551	set_task_state(tsk, TASK_RUNNING);
b6d09449	552	put_task_struct(tsk);
1da177e4 LT	553	} else
	554	set_tsk_need_resched(tsk);
	555	}
	556	}
1da177e4	557
29b08d2b HC	558	void __init pfault_irq_init(void)
	559	{
	560	if (!MACHINE_IS_VM)
	561	return;
	562
	563	/*
	564	* Try to get pfault pseudo page faults going.
	565	*/
	566	if (register_early_external_interrupt(0x2603, pfault_interrupt,
	567	&ext_int_pfault) != 0)
	568	panic("Couldn't request external interrupt 0x2603");
	569
	570	if (pfault_init() == 0)
	571	return;
	572
	573	/* Tough luck, no pfault. */
	574	pfault_disable = 1;
	575	unregister_early_external_interrupt(0x2603, pfault_interrupt,
	576	&ext_int_pfault);
	577	}
	578	#endif