[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/perf_event.h>
#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/compat.h>
#include <linux/smp.h>
#include <linux/kdebug.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 <linux/hugetlb.h>
#include <asm/system.h>
#include <asm/pgtable.h>
#include <asm/s390_ext.h>
#include <asm/mmu_context.h>
#include <asm/compat.h>
#include "../kernel/entry.h"

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

#ifdef CONFIG_SYSCTL
extern int sysctl_userprocess_debug;
#endif

#define VM_FAULT_BADCONTEXT	0x010000
#define VM_FAULT_BADMAP		0x020000
#define VM_FAULT_BADACCESS	0x040000

static inline int notify_page_fault(struct pt_regs *regs)
{
	int ret = 0;

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


/*
 * 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 and 1 for user space.
 */
static inline int user_space_fault(unsigned long trans_exc_code)
{
	/*
	 * The lowest two bits of the translation exception
	 * identification indicate which paging table was used.
	 */
	trans_exc_code &= 3;
	if (trans_exc_code == 2)
		/* Access via secondary space, set_fs setting decides */
		return current->thread.mm_segment.ar4;
	if (user_mode == HOME_SPACE_MODE)
		/* User space if the access has been done via home space. */
		return trans_exc_code == 3;
	/*
	 * If the user space is not the home space the kernel runs in home
	 * space. Access via secondary space has already been covered,
	 * access via primary space or access register is from user space
	 * and access via home space is from the kernel.
	 */
	return trans_exc_code != 3;
}

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

	address = trans_exc_code & __FAIL_ADDR_MASK;
	current->thread.prot_addr = address;
	current->thread.trap_no = int_code;
#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",
		       int_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 noinline void do_no_context(struct pt_regs *regs, long int_code,
				   unsigned long trans_exc_code)
{
	const struct exception_table_entry *fixup;
	unsigned long address;

	/* Are we prepared to handle this kernel fault?  */
	fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN);
	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.
	 */
	address = trans_exc_code & __FAIL_ADDR_MASK;
	if (!user_space_fault(trans_exc_code))
		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, int_code);
	do_exit(SIGKILL);
}

static noinline void do_low_address(struct pt_regs *regs, long int_code,
				    unsigned long trans_exc_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, int_code);
		do_exit(SIGKILL);
	}

	do_no_context(regs, int_code, trans_exc_code);
}

static noinline void do_sigbus(struct pt_regs *regs, long int_code,
			       unsigned long trans_exc_code)
{
	struct task_struct *tsk = current;

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

#ifdef CONFIG_S390_EXEC_PROTECT
static noinline int signal_return(struct pt_regs *regs, long int_code,
				  unsigned long trans_exc_code)
{
	u16 instruction;
	int rc;

	rc = __get_user(instruction, (u16 __user *) regs->psw.addr);

	if (!rc && instruction == 0x0a77) {
		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
		if (is_compat_task())
			sys32_sigreturn();
		else
			sys_sigreturn();
	} else if (!rc && instruction == 0x0aad) {
		clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
		if (is_compat_task())
			sys32_rt_sigreturn();
		else
			sys_rt_sigreturn();
	} else
		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
	return 0;
}
#endif /* CONFIG_S390_EXEC_PROTECT */

static noinline void do_fault_error(struct pt_regs *regs, long int_code,
				    unsigned long trans_exc_code, int fault)
{
	int si_code;

	switch (fault) {
	case VM_FAULT_BADACCESS:
#ifdef CONFIG_S390_EXEC_PROTECT
		if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
		    (trans_exc_code & 3) == 0) {
			signal_return(regs, int_code, trans_exc_code);
			break;
		}
#endif /* CONFIG_S390_EXEC_PROTECT */
	case VM_FAULT_BADMAP:
		/* Bad memory access. Check if it is kernel or user space. */
		if (regs->psw.mask & PSW_MASK_PSTATE) {
			/* User mode accesses just cause a SIGSEGV */
			si_code = (fault == VM_FAULT_BADMAP) ?
				SEGV_MAPERR : SEGV_ACCERR;
			do_sigsegv(regs, int_code, si_code, trans_exc_code);
			return;
		}
	case VM_FAULT_BADCONTEXT:
		do_no_context(regs, int_code, trans_exc_code);
		break;
	default: /* fault & VM_FAULT_ERROR */
		if (fault & VM_FAULT_OOM)
			pagefault_out_of_memory();
		else if (fault & VM_FAULT_SIGBUS) {
			do_sigbus(regs, int_code, trans_exc_code);
			/* Kernel mode? Handle exceptions or die */
			if (!(regs->psw.mask & PSW_MASK_PSTATE))
				do_no_context(regs, int_code, trans_exc_code);
		} else
			BUG();
		break;
	}
}

/*
 * This routine handles page faults.  It determines the address,
 * and the problem, and then passes it off to one of the appropriate
 * routines.
 *
 * interruption code (int_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 int do_exception(struct pt_regs *regs, int access,
			       unsigned long trans_exc_code)
{
	struct task_struct *tsk;
	struct mm_struct *mm;
	struct vm_area_struct *vma;
	unsigned long address;
	int fault;

	if (notify_page_fault(regs))
		return 0;

	tsk = current;
	mm = tsk->mm;

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

	address = trans_exc_code & __FAIL_ADDR_MASK;
	/*
	 * 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();
	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
	down_read(&mm->mmap_sem);

	fault = VM_FAULT_BADMAP;
	vma = find_vma(mm, address);
	if (!vma)
		goto out_up;

	if (unlikely(vma->vm_start > address)) {
		if (!(vma->vm_flags & VM_GROWSDOWN))
			goto out_up;
		if (expand_stack(vma, address))
			goto out_up;
	}

	/*
	 * Ok, we have a good vm_area for this memory access, so
	 * we can handle it..
	 */
	fault = VM_FAULT_BADACCESS;
	if (unlikely(!(vma->vm_flags & access)))
		goto out_up;

	if (is_vm_hugetlb_page(vma))
		address &= HPAGE_MASK;
	/*
	 * 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,
				(access == VM_WRITE) ? FAULT_FLAG_WRITE : 0);
	if (unlikely(fault & VM_FAULT_ERROR))
		goto out_up;

	if (fault & VM_FAULT_MAJOR) {
		tsk->maj_flt++;
		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
				     regs, address);
	} else {
		tsk->min_flt++;
		perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
				     regs, address);
	}
	/*
	 * 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);
	fault = 0;
out_up:
	up_read(&mm->mmap_sem);
out:
	return fault;
}

void __kprobes do_protection_exception(struct pt_regs *regs, long int_code)
{
	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
	int fault;

	/* Protection exception is supressing, decrement psw address. */
	regs->psw.addr -= (int_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(!(trans_exc_code & 4))) {
		do_low_address(regs, int_code, trans_exc_code);
		return;
	}
	fault = do_exception(regs, VM_WRITE, trans_exc_code);
	if (unlikely(fault))
		do_fault_error(regs, 4, trans_exc_code, fault);
}

void __kprobes do_dat_exception(struct pt_regs *regs, long int_code)
{
	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
	int access, fault;

	access = VM_READ | VM_EXEC | VM_WRITE;
#ifdef CONFIG_S390_EXEC_PROTECT
	if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
	    (trans_exc_code & 3) == 0)
		access = VM_EXEC;
#endif
	fault = do_exception(regs, access, trans_exc_code);
	if (unlikely(fault))
		do_fault_error(regs, int_code & 255, trans_exc_code, fault);
}

#ifdef CONFIG_64BIT
void __kprobes do_asce_exception(struct pt_regs *regs, long int_code)
{
	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;

	if (unlikely(!user_space_fault(trans_exc_code) || in_atomic() || !mm))
		goto no_context;

	local_irq_enable();

	down_read(&mm->mmap_sem);
	vma = find_vma(mm, trans_exc_code & __FAIL_ADDR_MASK);
	up_read(&mm->mmap_sem);

	if (vma) {
		update_mm(mm, current);
		return;
	}

	/* User mode accesses just cause a SIGSEGV */
	if (regs->psw.mask & PSW_MASK_PSTATE) {
		do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
		return;
	}

no_context:
	do_no_context(regs, int_code, trans_exc_code);
}
#endif

int __handle_fault(unsigned long uaddr, unsigned long int_code, int write_user)
{
	struct pt_regs regs;
	int access, fault;

	regs.psw.mask = psw_kernel_bits;
	if (!irqs_disabled())
		regs.psw.mask |= PSW_MASK_IO | PSW_MASK_EXT;
	regs.psw.addr = (unsigned long) __builtin_return_address(0);
	regs.psw.addr |= PSW_ADDR_AMODE;
	uaddr &= PAGE_MASK;
	access = write_user ? VM_WRITE : VM_READ;
	fault = do_exception(&regs, access, uaddr | 2);
	if (unlikely(fault)) {
		if (fault & VM_FAULT_OOM) {
			pagefault_out_of_memory();
			fault = 0;
		} else if (fault & VM_FAULT_SIGBUS)
			do_sigbus(&regs, int_code, uaddr);
	}
	return fault ? -EFAULT : 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, aligned(8))) 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 int_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
cdd6c482	13	#include <linux/perf_event.h>
1da177e4 LT	14	#include <linux/signal.h>
	15	#include <linux/sched.h>
	16	#include <linux/kernel.h>
	17	#include <linux/errno.h>
	18	#include <linux/string.h>
	19	#include <linux/types.h>
	20	#include <linux/ptrace.h>
	21	#include <linux/mman.h>
	22	#include <linux/mm.h>
7757591a	23	#include <linux/compat.h>
1da177e4	24	#include <linux/smp.h>
1eeb66a1	25	#include <linux/kdebug.h>
1da177e4 LT	26	#include <linux/init.h>
	27	#include <linux/console.h>
	28	#include <linux/module.h>
	29	#include <linux/hardirq.h>
4ba069b8	30	#include <linux/kprobes.h>
be5ec363	31	#include <linux/uaccess.h>
53492b1d	32	#include <linux/hugetlb.h>
1da177e4	33	#include <asm/system.h>
1da177e4	34	#include <asm/pgtable.h>
29b08d2b	35	#include <asm/s390_ext.h>
6252d702	36	#include <asm/mmu_context.h>
50d7280d	37	#include <asm/compat.h>
a806170e	38	#include "../kernel/entry.h"
1da177e4	39
347a8dc3	40	#ifndef CONFIG_64BIT
1da177e4	41	#define __FAIL_ADDR_MASK 0x7ffff000
1da177e4 LT	42	#define __SUBCODE_MASK 0x0200
1da177e4 LT	43	#define __PF_RES_FIELD 0ULL
347a8dc3	44	#else /* CONFIG_64BIT */
1da177e4	45	#define __FAIL_ADDR_MASK -4096L
1da177e4 LT	46	#define __SUBCODE_MASK 0x0600
1da177e4 LT	47	#define __PF_RES_FIELD 0x8000000000000000ULL
347a8dc3	48	#endif /* CONFIG_64BIT */
1da177e4 LT	49
	50	#ifdef CONFIG_SYSCTL
	51	extern int sysctl_userprocess_debug;
	52	#endif
	53
50d7280d MS	54	#define VM_FAULT_BADCONTEXT 0x010000
	55	#define VM_FAULT_BADMAP 0x020000
	56	#define VM_FAULT_BADACCESS 0x040000
	57
7ecb344a	58	static inline int notify_page_fault(struct pt_regs *regs)
10c1031f	59	{
33464e3b CH	60	int ret = 0;
33464e3b CH	61
7ecb344a	62	#ifdef CONFIG_KPROBES
33464e3b CH	63	/* kprobe_running() needs smp_processor_id() */
	64	if (!user_mode(regs)) {
	65	preempt_disable();
	66	if (kprobe_running() && kprobe_fault_handler(regs, 14))
	67	ret = 1;
	68	preempt_enable();
	69	}
7ecb344a	70	#endif
33464e3b	71	return ret;
4ba069b8	72	}
4ba069b8	73
1da177e4 LT	74
	75	/*
	76	* Unlock any spinlocks which will prevent us from getting the
cefc8be8	77	* message out.
1da177e4 LT	78	*/
	79	void bust_spinlocks(int yes)
	80	{
	81	if (yes) {
	82	oops_in_progress = 1;
	83	} else {
	84	int loglevel_save = console_loglevel;
	85	console_unblank();
	86	oops_in_progress = 0;
	87	/*
	88	* OK, the message is on the console. Now we call printk()
	89	* without oops_in_progress set so that printk will give klogd
	90	* a poke. Hold onto your hats...
	91	*/
	92	console_loglevel = 15;
	93	printk(" ");
	94	console_loglevel = loglevel_save;
	95	}
	96	}
	97
	98	/*
482b05dd	99	* Returns the address space associated with the fault.
61365e13	100	* Returns 0 for kernel space and 1 for user space.
1da177e4	101	*/
61365e13	102	static inline int user_space_fault(unsigned long trans_exc_code)
1da177e4 LT	103	{
1da177e4 LT	104	/*
61365e13 MS	105	* The lowest two bits of the translation exception
61365e13 MS	106	* identification indicate which paging table was used.
1da177e4	107	*/
61365e13 MS	108	trans_exc_code &= 3;
	109	if (trans_exc_code == 2)
	110	/* Access via secondary space, set_fs setting decides */
	111	return current->thread.mm_segment.ar4;
b11b5334	112	if (user_mode == HOME_SPACE_MODE)
61365e13 MS	113	/* User space if the access has been done via home space. */
	114	return trans_exc_code == 3;
	115	/*
	116	* If the user space is not the home space the kernel runs in home
	117	* space. Access via secondary space has already been covered,
	118	* access via primary space or access register is from user space
	119	* and access via home space is from the kernel.
	120	*/
	121	return trans_exc_code != 3;
1da177e4 LT	122	}
	123
	124	/*
	125	* Send SIGSEGV to task. This is an external routine
	126	* to keep the stack usage of do_page_fault small.
	127	*/
50d7280d MS	128	static noinline void do_sigsegv(struct pt_regs *regs, long int_code,
50d7280d MS	129	int si_code, unsigned long trans_exc_code)
1da177e4 LT	130	{
1da177e4 LT	131	struct siginfo si;
50d7280d	132	unsigned long address;
1da177e4	133
50d7280d MS	134	address = trans_exc_code & __FAIL_ADDR_MASK;
	135	current->thread.prot_addr = address;
	136	current->thread.trap_no = int_code;
1da177e4 LT	137	#if defined(CONFIG_SYSCTL) \|\| defined(CONFIG_PROCESS_DEBUG)
	138	#if defined(CONFIG_SYSCTL)
	139	if (sysctl_userprocess_debug)
	140	#endif
	141	{
	142	printk("User process fault: interruption code 0x%lX\n",
50d7280d	143	int_code);
1da177e4 LT	144	printk("failing address: %lX\n", address);
	145	show_regs(regs);
	146	}
	147	#endif
	148	si.si_signo = SIGSEGV;
	149	si.si_code = si_code;
d2c993d8	150	si.si_addr = (void __user *) address;
1da177e4 LT	151	force_sig_info(SIGSEGV, &si, current);
	152	}
	153
50d7280d MS	154	static noinline void do_no_context(struct pt_regs *regs, long int_code,
50d7280d MS	155	unsigned long trans_exc_code)
10c1031f MS	156	{
10c1031f MS	157	const struct exception_table_entry *fixup;
61365e13	158	unsigned long address;
10c1031f MS	159
10c1031f MS	160	/* Are we prepared to handle this kernel fault? */
50d7280d	161	fixup = search_exception_tables(regs->psw.addr & PSW_ADDR_INSN);
10c1031f MS	162	if (fixup) {
	163	regs->psw.addr = fixup->fixup \| PSW_ADDR_AMODE;
	164	return;
	165	}
	166
	167	/*
	168	* Oops. The kernel tried to access some bad page. We'll have to
	169	* terminate things with extreme prejudice.
	170	*/
61365e13	171	address = trans_exc_code & __FAIL_ADDR_MASK;
b11b5334	172	if (!user_space_fault(trans_exc_code))
10c1031f MS	173	printk(KERN_ALERT "Unable to handle kernel pointer dereference"
	174	" at virtual kernel address %p\n", (void *)address);
	175	else
	176	printk(KERN_ALERT "Unable to handle kernel paging request"
	177	" at virtual user address %p\n", (void *)address);
	178
50d7280d	179	die("Oops", regs, int_code);
10c1031f MS	180	do_exit(SIGKILL);
	181	}
	182
50d7280d MS	183	static noinline void do_low_address(struct pt_regs *regs, long int_code,
50d7280d MS	184	unsigned long trans_exc_code)
10c1031f MS	185	{
	186	/* Low-address protection hit in kernel mode means
	187	NULL pointer write access in kernel mode. */
	188	if (regs->psw.mask & PSW_MASK_PSTATE) {
	189	/* Low-address protection hit in user mode 'cannot happen'. */
50d7280d	190	die ("Low-address protection", regs, int_code);
10c1031f MS	191	do_exit(SIGKILL);
	192	}
	193
50d7280d	194	do_no_context(regs, int_code, trans_exc_code);
10c1031f MS	195	}
10c1031f MS	196
50d7280d MS	197	static noinline void do_sigbus(struct pt_regs *regs, long int_code,
50d7280d MS	198	unsigned long trans_exc_code)
10c1031f MS	199	{
10c1031f MS	200	struct task_struct *tsk = current;
10c1031f	201
10c1031f MS	202	/*
	203	* Send a sigbus, regardless of whether we were in kernel
	204	* or user mode.
	205	*/
61365e13	206	tsk->thread.prot_addr = trans_exc_code & __FAIL_ADDR_MASK;
50d7280d	207	tsk->thread.trap_no = int_code;
10c1031f	208	force_sig(SIGBUS, tsk);
10c1031f MS	209	}
10c1031f MS	210
c1821c2e	211	#ifdef CONFIG_S390_EXEC_PROTECT
50d7280d MS	212	static noinline int signal_return(struct pt_regs *regs, long int_code,
50d7280d MS	213	unsigned long trans_exc_code)
c1821c2e	214	{
be5ec363	215	u16 instruction;
490f03d6	216	int rc;
be5ec363	217
be5ec363	218	rc = __get_user(instruction, (u16 __user *) regs->psw.addr);
be5ec363	219
50d7280d MS	220	if (!rc && instruction == 0x0a77) {
	221	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
	222	if (is_compat_task())
	223	sys32_sigreturn();
	224	else
	225	sys_sigreturn();
	226	} else if (!rc && instruction == 0x0aad) {
	227	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
	228	if (is_compat_task())
	229	sys32_rt_sigreturn();
	230	else
	231	sys_rt_sigreturn();
	232	} else
	233	do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
c1821c2e	234	return 0;
c1821c2e GS	235	}
	236	#endif /* CONFIG_S390_EXEC_PROTECT */
	237
50d7280d MS	238	static noinline void do_fault_error(struct pt_regs *regs, long int_code,
	239	unsigned long trans_exc_code, int fault)
	240	{
	241	int si_code;
	242
	243	switch (fault) {
	244	case VM_FAULT_BADACCESS:
	245	#ifdef CONFIG_S390_EXEC_PROTECT
	246	if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
	247	(trans_exc_code & 3) == 0) {
	248	signal_return(regs, int_code, trans_exc_code);
	249	break;
	250	}
	251	#endif /* CONFIG_S390_EXEC_PROTECT */
	252	case VM_FAULT_BADMAP:
	253	/* Bad memory access. Check if it is kernel or user space. */
	254	if (regs->psw.mask & PSW_MASK_PSTATE) {
	255	/* User mode accesses just cause a SIGSEGV */
	256	si_code = (fault == VM_FAULT_BADMAP) ?
	257	SEGV_MAPERR : SEGV_ACCERR;
	258	do_sigsegv(regs, int_code, si_code, trans_exc_code);
	259	return;
	260	}
	261	case VM_FAULT_BADCONTEXT:
	262	do_no_context(regs, int_code, trans_exc_code);
	263	break;
	264	default: /* fault & VM_FAULT_ERROR */
	265	if (fault & VM_FAULT_OOM)
	266	pagefault_out_of_memory();
	267	else if (fault & VM_FAULT_SIGBUS) {
	268	do_sigbus(regs, int_code, trans_exc_code);
	269	/* Kernel mode? Handle exceptions or die */
	270	if (!(regs->psw.mask & PSW_MASK_PSTATE))
	271	do_no_context(regs, int_code, trans_exc_code);
	272	} else
	273	BUG();
	274	break;
	275	}
	276	}
	277
1da177e4 LT	278	/*
	279	* This routine handles page faults. It determines the address,
	280	* and the problem, and then passes it off to one of the appropriate
	281	* routines.
	282	*
50d7280d	283	* interruption code (int_code):
1da177e4 LT	284	* 04 Protection -> Write-Protection (suprression)
	285	* 10 Segment translation -> Not present (nullification)
	286	* 11 Page translation -> Not present (nullification)
	287	* 3b Region third trans. -> Not present (nullification)
	288	*/
1ab947de	289	static inline int do_exception(struct pt_regs *regs, int access,
50d7280d	290	unsigned long trans_exc_code)
1da177e4	291	{
10c1031f MS	292	struct task_struct *tsk;
	293	struct mm_struct *mm;
	294	struct vm_area_struct *vma;
	295	unsigned long address;
83c54070	296	int fault;
1da177e4	297
7ecb344a	298	if (notify_page_fault(regs))
50d7280d	299	return 0;
4ba069b8	300
10c1031f MS	301	tsk = current;
10c1031f MS	302	mm = tsk->mm;
1da177e4	303
1da177e4 LT	304	/*
	305	* Verify that the fault happened in user space, that
	306	* we are not in an interrupt and that there is a
	307	* user context.
	308	*/
50d7280d	309	fault = VM_FAULT_BADCONTEXT;
61365e13	310	if (unlikely(!user_space_fault(trans_exc_code) \|\| in_atomic() \|\| !mm))
50d7280d	311	goto out;
1da177e4	312
61365e13	313	address = trans_exc_code & __FAIL_ADDR_MASK;
1da177e4 LT	314	/*
	315	* When we get here, the fault happened in the current
	316	* task's user address space, so we can switch on the
	317	* interrupts again and then search the VMAs
	318	*/
	319	local_irq_enable();
cdd6c482	320	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
10c1031f	321	down_read(&mm->mmap_sem);
1da177e4	322
50d7280d	323	fault = VM_FAULT_BADMAP;
482b05dd GS	324	vma = find_vma(mm, address);
482b05dd GS	325	if (!vma)
50d7280d	326	goto out_up;
c1821c2e	327
50d7280d MS	328	if (unlikely(vma->vm_start > address)) {
	329	if (!(vma->vm_flags & VM_GROWSDOWN))
	330	goto out_up;
	331	if (expand_stack(vma, address))
	332	goto out_up;
	333	}
	334
	335	/*
	336	* Ok, we have a good vm_area for this memory access, so
	337	* we can handle it..
	338	*/
	339	fault = VM_FAULT_BADACCESS;
1ab947de	340	if (unlikely(!(vma->vm_flags & access)))
50d7280d	341	goto out_up;
1da177e4	342
53492b1d GS	343	if (is_vm_hugetlb_page(vma))
53492b1d GS	344	address &= HPAGE_MASK;
1da177e4 LT	345	/*
	346	* If for any reason at all we couldn't handle the fault,
	347	* make sure we exit gracefully rather than endlessly redo
	348	* the fault.
	349	*/
1ab947de MS	350	fault = handle_mm_fault(mm, vma, address,
1ab947de MS	351	(access == VM_WRITE) ? FAULT_FLAG_WRITE : 0);
50d7280d MS	352	if (unlikely(fault & VM_FAULT_ERROR))
	353	goto out_up;
	354
bde69af2	355	if (fault & VM_FAULT_MAJOR) {
83c54070	356	tsk->maj_flt++;
cdd6c482	357	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ, 1, 0,
bde69af2 HC	358	regs, address);
bde69af2 HC	359	} else {
83c54070	360	tsk->min_flt++;
cdd6c482	361	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN, 1, 0,
bde69af2 HC	362	regs, address);
bde69af2 HC	363	}
1da177e4 LT	364	/*
	365	* The instruction that caused the program check will
	366	* be repeated. Don't signal single step via SIGTRAP.
	367	*/
482b05dd	368	clear_tsk_thread_flag(tsk, TIF_SINGLE_STEP);
50d7280d MS	369	fault = 0;
50d7280d MS	370	out_up:
10c1031f	371	up_read(&mm->mmap_sem);
50d7280d MS	372	out:
50d7280d MS	373	return fault;
1da177e4 LT	374	}
1da177e4 LT	375
50d7280d	376	void __kprobes do_protection_exception(struct pt_regs *regs, long int_code)
1da177e4	377	{
61365e13	378	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
50d7280d	379	int fault;
61365e13	380
10c1031f	381	/* Protection exception is supressing, decrement psw address. */
50d7280d	382	regs->psw.addr -= (int_code >> 16);
10c1031f MS	383	/*
	384	* Check for low-address protection. This needs to be treated
	385	* as a special case because the translation exception code
	386	* field is not guaranteed to contain valid data in this case.
	387	*/
61365e13	388	if (unlikely(!(trans_exc_code & 4))) {
50d7280d	389	do_low_address(regs, int_code, trans_exc_code);
10c1031f MS	390	return;
10c1031f MS	391	}
1ab947de	392	fault = do_exception(regs, VM_WRITE, trans_exc_code);
50d7280d MS	393	if (unlikely(fault))
50d7280d MS	394	do_fault_error(regs, 4, trans_exc_code, fault);
1da177e4 LT	395	}
1da177e4 LT	396
50d7280d	397	void __kprobes do_dat_exception(struct pt_regs *regs, long int_code)
1da177e4	398	{
50d7280d	399	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
1ab947de	400	int access, fault;
50d7280d	401
1ab947de MS	402	access = VM_READ \| VM_EXEC \| VM_WRITE;
	403	#ifdef CONFIG_S390_EXEC_PROTECT
	404	if ((regs->psw.mask & PSW_MASK_ASC) == PSW_ASC_SECONDARY &&
	405	(trans_exc_code & 3) == 0)
	406	access = VM_EXEC;
	407	#endif
	408	fault = do_exception(regs, access, trans_exc_code);
50d7280d MS	409	if (unlikely(fault))
50d7280d MS	410	do_fault_error(regs, int_code & 255, trans_exc_code, fault);
1da177e4 LT	411	}
1da177e4 LT	412
6252d702	413	#ifdef CONFIG_64BIT
50d7280d	414	void __kprobes do_asce_exception(struct pt_regs *regs, long int_code)
6252d702	415	{
61365e13	416	unsigned long trans_exc_code = S390_lowcore.trans_exc_code;
50d7280d	417	struct mm_struct *mm = current->mm;
6252d702	418	struct vm_area_struct *vma;
6252d702	419
61365e13	420	if (unlikely(!user_space_fault(trans_exc_code) \|\| in_atomic() \|\| !mm))
6252d702 MS	421	goto no_context;
	422
	423	local_irq_enable();
	424
	425	down_read(&mm->mmap_sem);
50d7280d	426	vma = find_vma(mm, trans_exc_code & __FAIL_ADDR_MASK);
6252d702 MS	427	up_read(&mm->mmap_sem);
	428
	429	if (vma) {
	430	update_mm(mm, current);
	431	return;
	432	}
	433
	434	/* User mode accesses just cause a SIGSEGV */
	435	if (regs->psw.mask & PSW_MASK_PSTATE) {
50d7280d	436	do_sigsegv(regs, int_code, SEGV_MAPERR, trans_exc_code);
6252d702 MS	437	return;
	438	}
	439
	440	no_context:
50d7280d	441	do_no_context(regs, int_code, trans_exc_code);
6252d702 MS	442	}
	443	#endif
	444
6c1e3e79 GS	445	int __handle_fault(unsigned long uaddr, unsigned long int_code, int write_user)
	446	{
	447	struct pt_regs regs;
	448	int access, fault;
	449
	450	regs.psw.mask = psw_kernel_bits;
	451	if (!irqs_disabled())
	452	regs.psw.mask \|= PSW_MASK_IO \| PSW_MASK_EXT;
	453	regs.psw.addr = (unsigned long) __builtin_return_address(0);
	454	regs.psw.addr \|= PSW_ADDR_AMODE;
	455	uaddr &= PAGE_MASK;
	456	access = write_user ? VM_WRITE : VM_READ;
	457	fault = do_exception(&regs, access, uaddr \| 2);
	458	if (unlikely(fault)) {
	459	if (fault & VM_FAULT_OOM) {
	460	pagefault_out_of_memory();
	461	fault = 0;
	462	} else if (fault & VM_FAULT_SIGBUS)
	463	do_sigbus(&regs, int_code, uaddr);
	464	}
	465	return fault ? -EFAULT : 0;
	466	}
	467
1da177e4 LT	468	#ifdef CONFIG_PFAULT
	469	/*
	470	* 'pfault' pseudo page faults routines.
	471	*/
29b08d2b	472	static ext_int_info_t ext_int_pfault;
1da177e4 LT	473	static int pfault_disable = 0;
	474
	475	static int __init nopfault(char *str)
	476	{
	477	pfault_disable = 1;
	478	return 1;
	479	}
	480
	481	__setup("nopfault", nopfault);
	482
	483	typedef struct {
	484	__u16 refdiagc;
	485	__u16 reffcode;
	486	__u16 refdwlen;
	487	__u16 refversn;
	488	__u64 refgaddr;
	489	__u64 refselmk;
	490	__u64 refcmpmk;
	491	__u64 reserved;
c41fbc69	492	} __attribute__ ((packed, aligned(8))) pfault_refbk_t;
1da177e4 LT	493
	494	int pfault_init(void)
	495	{
	496	pfault_refbk_t refbk =
	497	{ 0x258, 0, 5, 2, __LC_CURRENT, 1ULL << 48, 1ULL << 48,
	498	__PF_RES_FIELD };
	499	int rc;
	500
29b08d2b	501	if (!MACHINE_IS_VM \|\| pfault_disable)
1da177e4	502	return -1;
94c12cc7 MS	503	asm volatile(
	504	" diag %1,%0,0x258\n"
	505	"0: j 2f\n"
	506	"1: la %0,8\n"
1da177e4	507	"2:\n"
94c12cc7 MS	508	EX_TABLE(0b,1b)
94c12cc7 MS	509	: "=d" (rc) : "a" (&refbk), "m" (refbk) : "cc");
1da177e4 LT	510	__ctl_set_bit(0, 9);
	511	return rc;
	512	}
	513
	514	void pfault_fini(void)
	515	{
	516	pfault_refbk_t refbk =
	517	{ 0x258, 1, 5, 2, 0ULL, 0ULL, 0ULL, 0ULL };
	518
29b08d2b	519	if (!MACHINE_IS_VM \|\| pfault_disable)
1da177e4 LT	520	return;
1da177e4 LT	521	__ctl_clear_bit(0,9);
94c12cc7 MS	522	asm volatile(
94c12cc7 MS	523	" diag %0,0,0x258\n"
1da177e4	524	"0:\n"
94c12cc7 MS	525	EX_TABLE(0b,0b)
94c12cc7 MS	526	: : "a" (&refbk), "m" (refbk) : "cc");
1da177e4 LT	527	}
1da177e4 LT	528
50d7280d	529	static void pfault_interrupt(__u16 int_code)
1da177e4 LT	530	{
	531	struct task_struct *tsk;
	532	__u16 subcode;
	533
	534	/*
	535	* Get the external interruption subcode & pfault
	536	* initial/completion signal bit. VM stores this
	537	* in the 'cpu address' field associated with the
	538	* external interrupt.
	539	*/
	540	subcode = S390_lowcore.cpu_addr;
	541	if ((subcode & 0xff00) != __SUBCODE_MASK)
	542	return;
	543
	544	/*
	545	* Get the token (= address of the task structure of the affected task).
	546	*/
	547	tsk = (struct task_struct *) __LC_PFAULT_INTPARM;
	548
	549	if (subcode & 0x0080) {
	550	/* signal bit is set -> a page has been swapped in by VM */
	551	if (xchg(&tsk->thread.pfault_wait, -1) != 0) {
	552	/* Initial interrupt was faster than the completion
	553	* interrupt. pfault_wait is valid. Set pfault_wait
	554	* back to zero and wake up the process. This can
	555	* safely be done because the task is still sleeping
b6d09449	556	* and can't produce new pfaults. */
1da177e4 LT	557	tsk->thread.pfault_wait = 0;
1da177e4 LT	558	wake_up_process(tsk);
b6d09449	559	put_task_struct(tsk);
1da177e4 LT	560	}
	561	} else {
	562	/* signal bit not set -> a real page is missing. */
b6d09449	563	get_task_struct(tsk);
1da177e4 LT	564	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	565	if (xchg(&tsk->thread.pfault_wait, 1) != 0) {
	566	/* Completion interrupt was faster than the initial
	567	* interrupt (swapped in a -1 for pfault_wait). Set
	568	* pfault_wait back to zero and exit. This can be
	569	* done safely because tsk is running in kernel
	570	* mode and can't produce new pfaults. */
	571	tsk->thread.pfault_wait = 0;
	572	set_task_state(tsk, TASK_RUNNING);
b6d09449	573	put_task_struct(tsk);
1da177e4 LT	574	} else
	575	set_tsk_need_resched(tsk);
	576	}
	577	}
1da177e4	578
29b08d2b HC	579	void __init pfault_irq_init(void)
	580	{
	581	if (!MACHINE_IS_VM)
	582	return;
	583
	584	/*
	585	* Try to get pfault pseudo page faults going.
	586	*/
	587	if (register_early_external_interrupt(0x2603, pfault_interrupt,
	588	&ext_int_pfault) != 0)
	589	panic("Couldn't request external interrupt 0x2603");
	590
	591	if (pfault_init() == 0)
	592	return;
	593
	594	/* Tough luck, no pfault. */
	595	pfault_disable = 1;
	596	unregister_early_external_interrupt(0x2603, pfault_interrupt,
	597	&ext_int_pfault);
	598	}
	599	#endif