[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/smp_lock.h>
#include <linux/init.h>
#include <linux/console.h>
#include <linux/module.h>
#include <linux/hardirq.h>
#include <linux/kprobes.h>

#include <asm/system.h>
#include <asm/uaccess.h>
#include <asm/pgtable.h>
#include <asm/kdebug.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 ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
int register_page_fault_notifier(struct notifier_block *nb)
{
	return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
}

int unregister_page_fault_notifier(struct notifier_block *nb)
{
	return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
}

static inline int notify_page_fault(enum die_val val, const char *str,
			struct pt_regs *regs, long err, int trap, int sig)
{
	struct die_args args = {
		.regs = regs,
		.str = str,
		.err = err,
		.trapnr = trap,
		.signr = sig
	};
	return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
}
#else
static inline int notify_page_fault(enum die_val val, const char *str,
			struct pt_regs *regs, long err, int trap, int sig)
{
	return NOTIFY_DONE;
}
#endif

extern spinlock_t timerlist_lock;

/*
 * Unlock any spinlocks which will prevent us from getting the
 * message out (timerlist_lock is acquired through the
 * console unblank code)
 */
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;
	}
}

/*
 * Check which address space is addressed by the access
 * register in S390_lowcore.exc_access_id.
 * Returns 1 for user space and 0 for kernel space.
 */
static int __check_access_register(struct pt_regs *regs, int error_code)
{
	int areg = S390_lowcore.exc_access_id;

	if (areg == 0)
		/* Access via access register 0 -> kernel address */
		return 0;
	save_access_regs(current->thread.acrs);
	if (regs && areg < NUM_ACRS && current->thread.acrs[areg] <= 1)
		/*
		 * access register contains 0 -> kernel address,
		 * access register contains 1 -> user space address
		 */
		return current->thread.acrs[areg];

	/* Something unhealthy was done with the access registers... */
	die("page fault via unknown access register", regs, error_code);
	do_exit(SIGKILL);
	return 0;
}

/*
 * Check which address space the address belongs to.
 * May return 1 or 2 for user space and 0 for kernel space.
 * Returns 2 for user space in primary addressing mode with
 * CONFIG_S390_EXEC_PROTECT on and kernel parameter noexec=on.
 */
static inline int check_user_space(struct pt_regs *regs, int error_code)
{
	/*
	 * The lowest two bits of S390_lowcore.trans_exc_code indicate
	 * which paging table was used:
	 *   0: Primary Segment Table Descriptor
	 *   1: STD determined via access register
	 *   2: Secondary Segment Table Descriptor
	 *   3: Home Segment Table Descriptor
	 */
	int descriptor = S390_lowcore.trans_exc_code & 3;
	if (unlikely(descriptor == 1))
		return __check_access_register(regs, error_code);
	if (descriptor == 2)
		return current->thread.mm_segment.ar4;
	return ((descriptor != 0) ^ (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);
}

#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 inline void do_sigreturn(struct mm_struct *mm, struct pt_regs *regs,
				int rt)
{
	up_read(&mm->mmap_sem);
	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
#ifdef CONFIG_COMPAT
	if (test_tsk_thread_flag(current, TIF_31BIT)) {
		if (rt)
			sys32_rt_sigreturn(regs);
		else
			sys32_sigreturn(regs);
		return;
	}
#endif /* CONFIG_COMPAT */
	if (rt)
		sys_rt_sigreturn(regs);
	else
		sys_sigreturn(regs);
	return;
}

static int signal_return(struct mm_struct *mm, struct pt_regs *regs,
			 unsigned long address, unsigned long error_code)
{
	pgd_t *pgd;
	pmd_t *pmd;
	pte_t *pte;
	u16 *instruction;
	unsigned long pfn, uaddr = regs->psw.addr;

	spin_lock(&mm->page_table_lock);
	pgd = pgd_offset(mm, uaddr);
	if (pgd_none(*pgd) || unlikely(pgd_bad(*pgd)))
		goto out_fault;
	pmd = pmd_offset(pgd, uaddr);
	if (pmd_none(*pmd) || unlikely(pmd_bad(*pmd)))
		goto out_fault;
	pte = pte_offset_map(pmd_offset(pgd_offset(mm, uaddr), uaddr), uaddr);
	if (!pte || !pte_present(*pte))
		goto out_fault;
	pfn = pte_pfn(*pte);
	if (!pfn_valid(pfn))
		goto out_fault;
	spin_unlock(&mm->page_table_lock);

	instruction = (u16 *) ((pfn << PAGE_SHIFT) + (uaddr & (PAGE_SIZE-1)));
	if (*instruction == 0x0a77)
		do_sigreturn(mm, regs, 0);
	else if (*instruction == 0x0aad)
		do_sigreturn(mm, regs, 1);
	else {
		printk("- XXX - do_exception: task = %s, primary, NO EXEC "
		       "-> SIGSEGV\n", current->comm);
		up_read(&mm->mmap_sem);
		current->thread.prot_addr = address;
		current->thread.trap_no = error_code;
		do_sigsegv(regs, error_code, SEGV_MAPERR, address);
	}
	return 0;
out_fault:
	spin_unlock(&mm->page_table_lock);
	return -EFAULT;
}
#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 __kprobes
do_exception(struct pt_regs *regs, unsigned long error_code, int is_protection)
{
        struct task_struct *tsk;
        struct mm_struct *mm;
        struct vm_area_struct * vma;
        unsigned long address;
	int user_address;
	const struct exception_table_entry *fixup;
	int si_code = SEGV_MAPERR;

        tsk = current;
        mm = tsk->mm;
	
	if (notify_page_fault(DIE_PAGE_FAULT, "page fault", regs, error_code, 14,
					SIGSEGV) == NOTIFY_STOP)
		return;

	/* 
         * 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 (is_protection && !(S390_lowcore.trans_exc_code & 4)) {

		/* Low-address protection hit in kernel mode means 
		   NULL pointer write access in kernel mode.  */
 		if (!(regs->psw.mask & PSW_MASK_PSTATE)) {
			address = 0;
			user_address = 0;
			goto no_context;
		}

		/* Low-address protection hit in user mode 'cannot happen'.  */
		die ("Low-address protection", regs, error_code);
        	do_exit(SIGKILL);
	}

        /* 
         * get the failing address 
         * more specific the segment and page table portion of 
         * the address 
         */
        address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
	user_address = check_user_space(regs, error_code);

	/*
	 * Verify that the fault happened in user space, that
	 * we are not in an interrupt and that there is a 
	 * user context.
	 */
        if (user_address == 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);

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

#ifdef CONFIG_S390_EXEC_PROTECT
	if (unlikely((user_address == 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 (!is_protection) {
		/* 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.
	 */
	switch (handle_mm_fault(mm, vma, address, is_protection)) {
	case VM_FAULT_MINOR:
		tsk->min_flt++;
		break;
	case VM_FAULT_MAJOR:
		tsk->maj_flt++;
		break;
	case VM_FAULT_SIGBUS:
		goto do_sigbus;
	case VM_FAULT_OOM:
		goto out_of_memory;
	default:
		BUG();
	}

        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(current, 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:
        /* 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 (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 (is_init(tsk)) {
		yield();
		down_read(&mm->mmap_sem);
		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);
}

#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>
	23	#include <linux/smp_lock.h>
	24	#include <linux/init.h>
	25	#include <linux/console.h>
	26	#include <linux/module.h>
	27	#include <linux/hardirq.h>
4ba069b8	28	#include <linux/kprobes.h>
1da177e4 LT	29
	30	#include <asm/system.h>
	31	#include <asm/uaccess.h>
	32	#include <asm/pgtable.h>
4ba069b8	33	#include <asm/kdebug.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
2b67fc46	55	static ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
4ba069b8 MG	56	int register_page_fault_notifier(struct notifier_block *nb)
	57	{
	58	return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
	59	}
	60
	61	int unregister_page_fault_notifier(struct notifier_block *nb)
	62	{
	63	return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
	64	}
	65
	66	static inline int notify_page_fault(enum die_val val, const char *str,
	67	struct pt_regs *regs, long err, int trap, int sig)
	68	{
	69	struct die_args args = {
	70	.regs = regs,
	71	.str = str,
	72	.err = err,
	73	.trapnr = trap,
	74	.signr = sig
	75	};
	76	return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
	77	}
	78	#else
	79	static inline int notify_page_fault(enum die_val val, const char *str,
	80	struct pt_regs *regs, long err, int trap, int sig)
	81	{
	82	return NOTIFY_DONE;
	83	}
	84	#endif
	85
1da177e4 LT	86	extern spinlock_t timerlist_lock;
	87
	88	/*
	89	* Unlock any spinlocks which will prevent us from getting the
	90	* message out (timerlist_lock is acquired through the
	91	* console unblank code)
	92	*/
	93	void bust_spinlocks(int yes)
	94	{
	95	if (yes) {
	96	oops_in_progress = 1;
	97	} else {
	98	int loglevel_save = console_loglevel;
	99	console_unblank();
	100	oops_in_progress = 0;
	101	/*
	102	* OK, the message is on the console. Now we call printk()
	103	* without oops_in_progress set so that printk will give klogd
	104	* a poke. Hold onto your hats...
	105	*/
	106	console_loglevel = 15;
	107	printk(" ");
	108	console_loglevel = loglevel_save;
	109	}
	110	}
	111
	112	/*
	113	* Check which address space is addressed by the access
	114	* register in S390_lowcore.exc_access_id.
	115	* Returns 1 for user space and 0 for kernel space.
	116	*/
	117	static int __check_access_register(struct pt_regs *regs, int error_code)
	118	{
	119	int areg = S390_lowcore.exc_access_id;
	120
	121	if (areg == 0)
	122	/* Access via access register 0 -> kernel address */
	123	return 0;
	124	save_access_regs(current->thread.acrs);
	125	if (regs && areg < NUM_ACRS && current->thread.acrs[areg] <= 1)
	126	/*
	127	* access register contains 0 -> kernel address,
	128	* access register contains 1 -> user space address
	129	*/
	130	return current->thread.acrs[areg];
	131
	132	/* Something unhealthy was done with the access registers... */
	133	die("page fault via unknown access register", regs, error_code);
	134	do_exit(SIGKILL);
	135	return 0;
	136	}
	137
	138	/*
	139	* Check which address space the address belongs to.
c1821c2e GS	140	* May return 1 or 2 for user space and 0 for kernel space.
	141	* Returns 2 for user space in primary addressing mode with
	142	* CONFIG_S390_EXEC_PROTECT on and kernel parameter noexec=on.
1da177e4 LT	143	*/
	144	static inline int check_user_space(struct pt_regs *regs, int error_code)
	145	{
	146	/*
	147	* The lowest two bits of S390_lowcore.trans_exc_code indicate
	148	* which paging table was used:
	149	* 0: Primary Segment Table Descriptor
	150	* 1: STD determined via access register
	151	* 2: Secondary Segment Table Descriptor
	152	* 3: Home Segment Table Descriptor
	153	*/
	154	int descriptor = S390_lowcore.trans_exc_code & 3;
	155	if (unlikely(descriptor == 1))
	156	return __check_access_register(regs, error_code);
	157	if (descriptor == 2)
	158	return current->thread.mm_segment.ar4;
c1821c2e	159	return ((descriptor != 0) ^ (switch_amode)) << s390_noexec;
1da177e4 LT	160	}
	161
	162	/*
	163	* Send SIGSEGV to task. This is an external routine
	164	* to keep the stack usage of do_page_fault small.
	165	*/
	166	static void do_sigsegv(struct pt_regs *regs, unsigned long error_code,
	167	int si_code, unsigned long address)
	168	{
	169	struct siginfo si;
	170
	171	#if defined(CONFIG_SYSCTL) \|\| defined(CONFIG_PROCESS_DEBUG)
	172	#if defined(CONFIG_SYSCTL)
	173	if (sysctl_userprocess_debug)
	174	#endif
	175	{
	176	printk("User process fault: interruption code 0x%lX\n",
	177	error_code);
	178	printk("failing address: %lX\n", address);
	179	show_regs(regs);
	180	}
	181	#endif
	182	si.si_signo = SIGSEGV;
	183	si.si_code = si_code;
d2c993d8	184	si.si_addr = (void __user *) address;
1da177e4 LT	185	force_sig_info(SIGSEGV, &si, current);
	186	}
	187
c1821c2e GS	188	#ifdef CONFIG_S390_EXEC_PROTECT
	189	extern long sys_sigreturn(struct pt_regs *regs);
	190	extern long sys_rt_sigreturn(struct pt_regs *regs);
	191	extern long sys32_sigreturn(struct pt_regs *regs);
	192	extern long sys32_rt_sigreturn(struct pt_regs *regs);
	193
	194	static inline void do_sigreturn(struct mm_struct mm, struct pt_regs regs,
	195	int rt)
	196	{
	197	up_read(&mm->mmap_sem);
	198	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
	199	#ifdef CONFIG_COMPAT
	200	if (test_tsk_thread_flag(current, TIF_31BIT)) {
	201	if (rt)
	202	sys32_rt_sigreturn(regs);
	203	else
	204	sys32_sigreturn(regs);
	205	return;
	206	}
	207	#endif /* CONFIG_COMPAT */
	208	if (rt)
	209	sys_rt_sigreturn(regs);
	210	else
	211	sys_sigreturn(regs);
	212	return;
	213	}
	214
	215	static int signal_return(struct mm_struct mm, struct pt_regs regs,
	216	unsigned long address, unsigned long error_code)
	217	{
	218	pgd_t *pgd;
	219	pmd_t *pmd;
	220	pte_t *pte;
	221	u16 *instruction;
	222	unsigned long pfn, uaddr = regs->psw.addr;
	223
	224	spin_lock(&mm->page_table_lock);
	225	pgd = pgd_offset(mm, uaddr);
	226	if (pgd_none(pgd) \|\| unlikely(pgd_bad(pgd)))
	227	goto out_fault;
	228	pmd = pmd_offset(pgd, uaddr);
	229	if (pmd_none(pmd) \|\| unlikely(pmd_bad(pmd)))
	230	goto out_fault;
	231	pte = pte_offset_map(pmd_offset(pgd_offset(mm, uaddr), uaddr), uaddr);
	232	if (!pte \|\| !pte_present(*pte))
	233	goto out_fault;
	234	pfn = pte_pfn(*pte);
	235	if (!pfn_valid(pfn))
	236	goto out_fault;
	237	spin_unlock(&mm->page_table_lock);
	238
	239	instruction = (u16 *) ((pfn << PAGE_SHIFT) + (uaddr & (PAGE_SIZE-1)));
	240	if (*instruction == 0x0a77)
	241	do_sigreturn(mm, regs, 0);
	242	else if (*instruction == 0x0aad)
	243	do_sigreturn(mm, regs, 1);
	244	else {
	245	printk("- XXX - do_exception: task = %s, primary, NO EXEC "
	246	"-> SIGSEGV\n", current->comm);
	247	up_read(&mm->mmap_sem);
	248	current->thread.prot_addr = address;
	249	current->thread.trap_no = error_code;
	250	do_sigsegv(regs, error_code, SEGV_MAPERR, address);
	251	}
252	return 0;
253	out_fault:
254	spin_unlock(&mm->page_table_lock);
255	return -EFAULT;
256	}
257	#endif /* CONFIG_S390_EXEC_PROTECT */
258
1da177e4 LT	259	/*
	260	* This routine handles page faults. It determines the address,
	261	* and the problem, and then passes it off to one of the appropriate
	262	* routines.
	263	*
	264	* error_code:
	265	* 04 Protection -> Write-Protection (suprression)
	266	* 10 Segment translation -> Not present (nullification)
	267	* 11 Page translation -> Not present (nullification)
	268	* 3b Region third trans. -> Not present (nullification)
	269	*/
4ba069b8	270	static inline void __kprobes
1da177e4 LT	271	do_exception(struct pt_regs *regs, unsigned long error_code, int is_protection)
	272	{
	273	struct task_struct *tsk;
	274	struct mm_struct *mm;
	275	struct vm_area_struct * vma;
	276	unsigned long address;
	277	int user_address;
	278	const struct exception_table_entry *fixup;
	279	int si_code = SEGV_MAPERR;
	280
	281	tsk = current;
	282	mm = tsk->mm;
	283
4ba069b8 MG	284	if (notify_page_fault(DIE_PAGE_FAULT, "page fault", regs, error_code, 14,
	285	SIGSEGV) == NOTIFY_STOP)
	286	return;
	287
1da177e4 LT	288	/*
	289	* Check for low-address protection. This needs to be treated
	290	* as a special case because the translation exception code
	291	* field is not guaranteed to contain valid data in this case.
	292	*/
	293	if (is_protection && !(S390_lowcore.trans_exc_code & 4)) {
	294
	295	/* Low-address protection hit in kernel mode means
	296	NULL pointer write access in kernel mode. */
	297	if (!(regs->psw.mask & PSW_MASK_PSTATE)) {
	298	address = 0;
	299	user_address = 0;
	300	goto no_context;
	301	}
	302
	303	/* Low-address protection hit in user mode 'cannot happen'. */
	304	die ("Low-address protection", regs, error_code);
	305	do_exit(SIGKILL);
	306	}
	307
	308	/*
	309	* get the failing address
	310	* more specific the segment and page table portion of
	311	* the address
	312	*/
	313	address = S390_lowcore.trans_exc_code & __FAIL_ADDR_MASK;
	314	user_address = check_user_space(regs, error_code);
	315
	316	/*
	317	* Verify that the fault happened in user space, that
	318	* we are not in an interrupt and that there is a
	319	* user context.
	320	*/
595bf2aa	321	if (user_address == 0 \|\| in_atomic() \|\| !mm)
1da177e4 LT	322	goto no_context;
	323
	324	/*
	325	* When we get here, the fault happened in the current
	326	* task's user address space, so we can switch on the
	327	* interrupts again and then search the VMAs
	328	*/
	329	local_irq_enable();
	330
	331	down_read(&mm->mmap_sem);
	332
	333	vma = find_vma(mm, address);
	334	if (!vma)
	335	goto bad_area;
c1821c2e GS	336
	337	#ifdef CONFIG_S390_EXEC_PROTECT
	338	if (unlikely((user_address == 2) && !(vma->vm_flags & VM_EXEC)))
	339	if (!signal_return(mm, regs, address, error_code))
	340	/*
	341	* signal_return() has done an up_read(&mm->mmap_sem)
	342	* if it returns 0.
	343	*/
	344	return;
	345	#endif
	346
1da177e4 LT	347	if (vma->vm_start <= address)
	348	goto good_area;
	349	if (!(vma->vm_flags & VM_GROWSDOWN))
	350	goto bad_area;
	351	if (expand_stack(vma, address))
	352	goto bad_area;
	353	/*
	354	* Ok, we have a good vm_area for this memory access, so
	355	* we can handle it..
	356	*/
	357	good_area:
	358	si_code = SEGV_ACCERR;
	359	if (!is_protection) {
	360	/* page not present, check vm flags */
	361	if (!(vma->vm_flags & (VM_READ \| VM_EXEC \| VM_WRITE)))
	362	goto bad_area;
	363	} else {
	364	if (!(vma->vm_flags & VM_WRITE))
	365	goto bad_area;
	366	}
	367
	368	survive:
	369	/*
	370	* If for any reason at all we couldn't handle the fault,
	371	* make sure we exit gracefully rather than endlessly redo
	372	* the fault.
	373	*/
	374	switch (handle_mm_fault(mm, vma, address, is_protection)) {
	375	case VM_FAULT_MINOR:
	376	tsk->min_flt++;
	377	break;
	378	case VM_FAULT_MAJOR:
	379	tsk->maj_flt++;
	380	break;
	381	case VM_FAULT_SIGBUS:
	382	goto do_sigbus;
	383	case VM_FAULT_OOM:
	384	goto out_of_memory;
	385	default:
	386	BUG();
	387	}
	388
	389	up_read(&mm->mmap_sem);
	390	/*
	391	* The instruction that caused the program check will
	392	* be repeated. Don't signal single step via SIGTRAP.
	393	*/
	394	clear_tsk_thread_flag(current, TIF_SINGLE_STEP);
	395	return;
	396
	397	/*
	398	* Something tried to access memory that isn't in our memory map..
	399	* Fix it, but check if it's kernel or user first..
	400	*/
	401	bad_area:
	402	up_read(&mm->mmap_sem);
	403
	404	/* User mode accesses just cause a SIGSEGV */
	405	if (regs->psw.mask & PSW_MASK_PSTATE) {
	406	tsk->thread.prot_addr = address;
	407	tsk->thread.trap_no = error_code;
	408	do_sigsegv(regs, error_code, si_code, address);
	409	return;
	410	}
411
412	no_context:
413	/* Are we prepared to handle this kernel fault? */
414	fixup = search_exception_tables(regs->psw.addr & __FIXUP_MASK);
415	if (fixup) {
416	regs->psw.addr = fixup->fixup \| PSW_ADDR_AMODE;
417	return;
418	}
419
420	/*
421	* Oops. The kernel tried to access some bad page. We'll have to
422	* terminate things with extreme prejudice.
423	*/
424	if (user_address == 0)
425	printk(KERN_ALERT "Unable to handle kernel pointer dereference"
426	" at virtual kernel address %p\n", (void *)address);
427	else
428	printk(KERN_ALERT "Unable to handle kernel paging request"
429	" at virtual user address %p\n", (void *)address);
430
431	die("Oops", regs, error_code);
432	do_exit(SIGKILL);
433
434
435	/*
436	* We ran out of memory, or some other thing happened to us that made
437	* us unable to handle the page fault gracefully.
438	*/
439	out_of_memory:
440	up_read(&mm->mmap_sem);
f400e198	441	if (is_init(tsk)) {
1da177e4	442	yield();
bac9c66c	443	down_read(&mm->mmap_sem);
1da177e4 LT	444	goto survive;
	445	}
	446	printk("VM: killing process %s\n", tsk->comm);
	447	if (regs->psw.mask & PSW_MASK_PSTATE)
	448	do_exit(SIGKILL);
	449	goto no_context;
	450
	451	do_sigbus:
	452	up_read(&mm->mmap_sem);
	453
	454	/*
	455	* Send a sigbus, regardless of whether we were in kernel
	456	* or user mode.
	457	*/
	458	tsk->thread.prot_addr = address;
	459	tsk->thread.trap_no = error_code;
	460	force_sig(SIGBUS, tsk);
	461
	462	/* Kernel mode? Handle exceptions or die */
	463	if (!(regs->psw.mask & PSW_MASK_PSTATE))
	464	goto no_context;
	465	}
	466
	467	void do_protection_exception(struct pt_regs *regs, unsigned long error_code)
	468	{
	469	regs->psw.addr -= (error_code >> 16);
	470	do_exception(regs, 4, 1);
	471	}
	472
	473	void do_dat_exception(struct pt_regs *regs, unsigned long error_code)
	474	{
	475	do_exception(regs, error_code & 0xff, 0);
	476	}
	477
1da177e4 LT	478	#ifdef CONFIG_PFAULT
	479	/*
	480	* 'pfault' pseudo page faults routines.
	481	*/
29b08d2b	482	static ext_int_info_t ext_int_pfault;
1da177e4 LT	483	static int pfault_disable = 0;
	484
	485	static int __init nopfault(char *str)
	486	{
	487	pfault_disable = 1;
	488	return 1;
	489	}
	490
	491	__setup("nopfault", nopfault);
	492
	493	typedef struct {
	494	__u16 refdiagc;
	495	__u16 reffcode;
	496	__u16 refdwlen;
	497	__u16 refversn;
	498	__u64 refgaddr;
	499	__u64 refselmk;
	500	__u64 refcmpmk;
	501	__u64 reserved;
	502	} __attribute__ ((packed)) pfault_refbk_t;
	503
	504	int pfault_init(void)
	505	{
	506	pfault_refbk_t refbk =
	507	{ 0x258, 0, 5, 2, __LC_CURRENT, 1ULL << 48, 1ULL << 48,
	508	__PF_RES_FIELD };
	509	int rc;
	510
29b08d2b	511	if (!MACHINE_IS_VM \|\| pfault_disable)
1da177e4	512	return -1;
94c12cc7 MS	513	asm volatile(
	514	" diag %1,%0,0x258\n"
	515	"0: j 2f\n"
	516	"1: la %0,8\n"
1da177e4	517	"2:\n"
94c12cc7 MS	518	EX_TABLE(0b,1b)
94c12cc7 MS	519	: "=d" (rc) : "a" (&refbk), "m" (refbk) : "cc");
1da177e4 LT	520	__ctl_set_bit(0, 9);
	521	return rc;
	522	}
	523
	524	void pfault_fini(void)
	525	{
	526	pfault_refbk_t refbk =
	527	{ 0x258, 1, 5, 2, 0ULL, 0ULL, 0ULL, 0ULL };
	528
29b08d2b	529	if (!MACHINE_IS_VM \|\| pfault_disable)
1da177e4 LT	530	return;
1da177e4 LT	531	__ctl_clear_bit(0,9);
94c12cc7 MS	532	asm volatile(
94c12cc7 MS	533	" diag %0,0,0x258\n"
1da177e4	534	"0:\n"
94c12cc7 MS	535	EX_TABLE(0b,0b)
94c12cc7 MS	536	: : "a" (&refbk), "m" (refbk) : "cc");
1da177e4 LT	537	}
1da177e4 LT	538
2b67fc46	539	static void pfault_interrupt(__u16 error_code)
1da177e4 LT	540	{
	541	struct task_struct *tsk;
	542	__u16 subcode;
	543
	544	/*
	545	* Get the external interruption subcode & pfault
	546	* initial/completion signal bit. VM stores this
	547	* in the 'cpu address' field associated with the
	548	* external interrupt.
	549	*/
	550	subcode = S390_lowcore.cpu_addr;
	551	if ((subcode & 0xff00) != __SUBCODE_MASK)
	552	return;
	553
	554	/*
	555	* Get the token (= address of the task structure of the affected task).
	556	*/
	557	tsk = (struct task_struct *) __LC_PFAULT_INTPARM;
	558
	559	if (subcode & 0x0080) {
	560	/* signal bit is set -> a page has been swapped in by VM */
	561	if (xchg(&tsk->thread.pfault_wait, -1) != 0) {
	562	/* Initial interrupt was faster than the completion
	563	* interrupt. pfault_wait is valid. Set pfault_wait
	564	* back to zero and wake up the process. This can
	565	* safely be done because the task is still sleeping
b6d09449	566	* and can't produce new pfaults. */
1da177e4 LT	567	tsk->thread.pfault_wait = 0;
1da177e4 LT	568	wake_up_process(tsk);
b6d09449	569	put_task_struct(tsk);
1da177e4 LT	570	}
	571	} else {
	572	/* signal bit not set -> a real page is missing. */
b6d09449	573	get_task_struct(tsk);
1da177e4 LT	574	set_task_state(tsk, TASK_UNINTERRUPTIBLE);
	575	if (xchg(&tsk->thread.pfault_wait, 1) != 0) {
	576	/* Completion interrupt was faster than the initial
	577	* interrupt (swapped in a -1 for pfault_wait). Set
	578	* pfault_wait back to zero and exit. This can be
	579	* done safely because tsk is running in kernel
	580	* mode and can't produce new pfaults. */
	581	tsk->thread.pfault_wait = 0;
	582	set_task_state(tsk, TASK_RUNNING);
b6d09449	583	put_task_struct(tsk);
1da177e4 LT	584	} else
	585	set_tsk_need_resched(tsk);
	586	}
	587	}
1da177e4	588
29b08d2b HC	589	void __init pfault_irq_init(void)
	590	{
	591	if (!MACHINE_IS_VM)
	592	return;
	593
	594	/*
	595	* Try to get pfault pseudo page faults going.
	596	*/
	597	if (register_early_external_interrupt(0x2603, pfault_interrupt,
	598	&ext_int_pfault) != 0)
	599	panic("Couldn't request external interrupt 0x2603");
	600
	601	if (pfault_init() == 0)
	602	return;
	603
	604	/* Tough luck, no pfault. */
	605	pfault_disable = 1;
	606	unregister_early_external_interrupt(0x2603, pfault_interrupt,
	607	&ext_int_pfault);
	608	}
	609	#endif