[net-next-2.6.git] / arch / ia64 / mm / fault.c

/*
 * MMU fault handling support.
 *
 * Copyright (C) 1998-2002 Hewlett-Packard Co
 *	David Mosberger-Tang <davidm@hpl.hp.com>
 */
#include <linux/sched.h>
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/smp_lock.h>
#include <linux/interrupt.h>
#include <linux/kprobes.h>

#include <asm/pgtable.h>
#include <asm/processor.h>
#include <asm/system.h>
#include <asm/uaccess.h>
#include <asm/kdebug.h>

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

#ifdef CONFIG_KPROBES
ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);

/* Hook to register for page fault notifications */
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

/*
 * Return TRUE if ADDRESS points at a page in the kernel's mapped segment
 * (inside region 5, on ia64) and that page is present.
 */
static int
mapped_kernel_page_is_present (unsigned long address)
{
	pgd_t *pgd;
	pud_t *pud;
	pmd_t *pmd;
	pte_t *ptep, pte;

	pgd = pgd_offset_k(address);
	if (pgd_none(*pgd) || pgd_bad(*pgd))
		return 0;

	pud = pud_offset(pgd, address);
	if (pud_none(*pud) || pud_bad(*pud))
		return 0;

	pmd = pmd_offset(pud, address);
	if (pmd_none(*pmd) || pmd_bad(*pmd))
		return 0;

	ptep = pte_offset_kernel(pmd, address);
	if (!ptep)
		return 0;

	pte = *ptep;
	return pte_present(pte);
}

void __kprobes
ia64_do_page_fault (unsigned long address, unsigned long isr, struct pt_regs *regs)
{
	int signal = SIGSEGV, code = SEGV_MAPERR;
	struct vm_area_struct *vma, *prev_vma;
	struct mm_struct *mm = current->mm;
	struct siginfo si;
	unsigned long mask;

	/* mmap_sem is performance critical.... */
	prefetchw(&mm->mmap_sem);

	/*
	 * If we're in an interrupt or have no user context, we must not take the fault..
	 */
	if (in_atomic() || !mm)
		goto no_context;

#ifdef CONFIG_VIRTUAL_MEM_MAP
	/*
	 * If fault is in region 5 and we are in the kernel, we may already
	 * have the mmap_sem (pfn_valid macro is called during mmap). There
	 * is no vma for region 5 addr's anyway, so skip getting the semaphore
	 * and go directly to the exception handling code.
	 */

	if ((REGION_NUMBER(address) == 5) && !user_mode(regs))
		goto bad_area_no_up;
#endif

	/*
	 * This is to handle the kprobes on user space access instructions
	 */
	if (notify_page_fault(DIE_PAGE_FAULT, "page fault", regs, code, TRAP_BRKPT,
					SIGSEGV) == NOTIFY_STOP)
		return;

	down_read(&mm->mmap_sem);

	vma = find_vma_prev(mm, address, &prev_vma);
	if (!vma)
		goto bad_area;

	/* find_vma_prev() returns vma such that address < vma->vm_end or NULL */
	if (address < vma->vm_start)
		goto check_expansion;

  good_area:
	code = SEGV_ACCERR;

	/* OK, we've got a good vm_area for this memory area.  Check the access permissions: */

#	define VM_READ_BIT	0
#	define VM_WRITE_BIT	1
#	define VM_EXEC_BIT	2

#	if (((1 << VM_READ_BIT) != VM_READ || (1 << VM_WRITE_BIT) != VM_WRITE) \
	    || (1 << VM_EXEC_BIT) != VM_EXEC)
#		error File is out of sync with <linux/mm.h>.  Please update.
#	endif

	if (((isr >> IA64_ISR_R_BIT) & 1UL) && (!(vma->vm_flags & (VM_READ | VM_WRITE))))
		goto bad_area;

	mask = (  (((isr >> IA64_ISR_X_BIT) & 1UL) << VM_EXEC_BIT)
		| (((isr >> IA64_ISR_W_BIT) & 1UL) << VM_WRITE_BIT));

	if ((vma->vm_flags & mask) != mask)
		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, (mask & VM_WRITE) != 0)) {
	      case VM_FAULT_MINOR:
		++current->min_flt;
		break;
	      case VM_FAULT_MAJOR:
		++current->maj_flt;
		break;
	      case VM_FAULT_SIGBUS:
		/*
		 * We ran out of memory, or some other thing happened
		 * to us that made us unable to handle the page fault
		 * gracefully.
		 */
		signal = SIGBUS;
		goto bad_area;
	      case VM_FAULT_OOM:
		goto out_of_memory;
	      default:
		BUG();
	}
	up_read(&mm->mmap_sem);
	return;

  check_expansion:
	if (!(prev_vma && (prev_vma->vm_flags & VM_GROWSUP) && (address == prev_vma->vm_end))) {
		if (!(vma->vm_flags & VM_GROWSDOWN))
			goto bad_area;
		if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
		    || REGION_OFFSET(address) >= RGN_MAP_LIMIT)
			goto bad_area;
		if (expand_stack(vma, address))
			goto bad_area;
	} else {
		vma = prev_vma;
		if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
		    || REGION_OFFSET(address) >= RGN_MAP_LIMIT)
			goto bad_area;
		/*
		 * Since the register backing store is accessed sequentially,
		 * we disallow growing it by more than a page at a time.
		 */
		if (address > vma->vm_end + PAGE_SIZE - sizeof(long))
			goto bad_area;
		if (expand_upwards(vma, address))
			goto bad_area;
	}
	goto good_area;

  bad_area:
	up_read(&mm->mmap_sem);
#ifdef CONFIG_VIRTUAL_MEM_MAP
  bad_area_no_up:
#endif
	if ((isr & IA64_ISR_SP)
	    || ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
	{
		/*
		 * This fault was due to a speculative load or lfetch.fault, set the "ed"
		 * bit in the psr to ensure forward progress.  (Target register will get a
		 * NaT for ld.s, lfetch will be canceled.)
		 */
		ia64_psr(regs)->ed = 1;
		return;
	}
	if (user_mode(regs)) {
		si.si_signo = signal;
		si.si_errno = 0;
		si.si_code = code;
		si.si_addr = (void __user *) address;
		si.si_isr = isr;
		si.si_flags = __ISR_VALID;
		force_sig_info(signal, &si, current);
		return;
	}

  no_context:
	if ((isr & IA64_ISR_SP)
	    || ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
	{
		/*
		 * This fault was due to a speculative load or lfetch.fault, set the "ed"
		 * bit in the psr to ensure forward progress.  (Target register will get a
		 * NaT for ld.s, lfetch will be canceled.)
		 */
		ia64_psr(regs)->ed = 1;
		return;
	}

	/*
	 * Since we have no vma's for region 5, we might get here even if the address is
	 * valid, due to the VHPT walker inserting a non present translation that becomes
	 * stale. If that happens, the non present fault handler already purged the stale
	 * translation, which fixed the problem. So, we check to see if the translation is
	 * valid, and return if it is.
	 */
	if (REGION_NUMBER(address) == 5 && mapped_kernel_page_is_present(address))
		return;

	if (ia64_done_with_exception(regs))
		return;

	/*
	 * Oops. The kernel tried to access some bad page. We'll have to terminate things
	 * with extreme prejudice.
	 */
	bust_spinlocks(1);

	if (address < PAGE_SIZE)
		printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference (address %016lx)\n", address);
	else
		printk(KERN_ALERT "Unable to handle kernel paging request at "
		       "virtual address %016lx\n", address);
	die("Oops", regs, isr);
	bust_spinlocks(0);
	do_exit(SIGKILL);
	return;

  out_of_memory:
	up_read(&mm->mmap_sem);
	if (current->pid == 1) {
		yield();
		down_read(&mm->mmap_sem);
		goto survive;
	}
	printk(KERN_CRIT "VM: killing process %s\n", current->comm);
	if (user_mode(regs))
		do_exit(SIGKILL);
	goto no_context;
}
Commit	Line	Data
1da177e4 LT	1	/*
	2	* MMU fault handling support.
	3	*
	4	* Copyright (C) 1998-2002 Hewlett-Packard Co
	5	* David Mosberger-Tang <davidm@hpl.hp.com>
	6	*/
	7	#include <linux/sched.h>
	8	#include <linux/kernel.h>
	9	#include <linux/mm.h>
	10	#include <linux/smp_lock.h>
	11	#include <linux/interrupt.h>
1f7ad57b	12	#include <linux/kprobes.h>
1da177e4 LT	13
	14	#include <asm/pgtable.h>
	15	#include <asm/processor.h>
	16	#include <asm/system.h>
	17	#include <asm/uaccess.h>
7213b252	18	#include <asm/kdebug.h>
1da177e4 LT	19
	20	extern void die (char , struct pt_regs , long);
	21
ae9a5b85 AK	22	#ifdef CONFIG_KPROBES
	23	ATOMIC_NOTIFIER_HEAD(notify_page_fault_chain);
	24
	25	/* Hook to register for page fault notifications */
	26	int register_page_fault_notifier(struct notifier_block *nb)
	27	{
	28	return atomic_notifier_chain_register(&notify_page_fault_chain, nb);
	29	}
	30
	31	int unregister_page_fault_notifier(struct notifier_block *nb)
	32	{
	33	return atomic_notifier_chain_unregister(&notify_page_fault_chain, nb);
	34	}
	35
	36	static inline int notify_page_fault(enum die_val val, const char *str,
	37	struct pt_regs *regs, long err, int trap, int sig)
	38	{
	39	struct die_args args = {
	40	.regs = regs,
	41	.str = str,
	42	.err = err,
	43	.trapnr = trap,
	44	.signr = sig
	45	};
	46	return atomic_notifier_call_chain(&notify_page_fault_chain, val, &args);
	47	}
	48	#else
	49	static inline int notify_page_fault(enum die_val val, const char *str,
	50	struct pt_regs *regs, long err, int trap, int sig)
	51	{
	52	return NOTIFY_DONE;
	53	}
	54	#endif
	55
1da177e4 LT	56	/*
	57	* Return TRUE if ADDRESS points at a page in the kernel's mapped segment
	58	* (inside region 5, on ia64) and that page is present.
	59	*/
	60	static int
	61	mapped_kernel_page_is_present (unsigned long address)
	62	{
	63	pgd_t *pgd;
	64	pud_t *pud;
	65	pmd_t *pmd;
	66	pte_t *ptep, pte;
	67
	68	pgd = pgd_offset_k(address);
	69	if (pgd_none(pgd) \|\| pgd_bad(pgd))
	70	return 0;
	71
	72	pud = pud_offset(pgd, address);
	73	if (pud_none(pud) \|\| pud_bad(pud))
	74	return 0;
	75
	76	pmd = pmd_offset(pud, address);
	77	if (pmd_none(pmd) \|\| pmd_bad(pmd))
	78	return 0;
	79
	80	ptep = pte_offset_kernel(pmd, address);
	81	if (!ptep)
	82	return 0;
	83
	84	pte = *ptep;
	85	return pte_present(pte);
	86	}
	87
1f7ad57b	88	void __kprobes
1da177e4 LT	89	ia64_do_page_fault (unsigned long address, unsigned long isr, struct pt_regs *regs)
	90	{
	91	int signal = SIGSEGV, code = SEGV_MAPERR;
	92	struct vm_area_struct vma, prev_vma;
	93	struct mm_struct *mm = current->mm;
	94	struct siginfo si;
	95	unsigned long mask;
	96
0ffe9849 CL	97	/* mmap_sem is performance critical.... */
	98	prefetchw(&mm->mmap_sem);
	99
1da177e4 LT	100	/*
	101	* If we're in an interrupt or have no user context, we must not take the fault..
	102	*/
	103	if (in_atomic() \|\| !mm)
	104	goto no_context;
	105
	106	#ifdef CONFIG_VIRTUAL_MEM_MAP
	107	/*
	108	* If fault is in region 5 and we are in the kernel, we may already
	109	* have the mmap_sem (pfn_valid macro is called during mmap). There
	110	* is no vma for region 5 addr's anyway, so skip getting the semaphore
	111	* and go directly to the exception handling code.
	112	*/
	113
	114	if ((REGION_NUMBER(address) == 5) && !user_mode(regs))
	115	goto bad_area_no_up;
	116	#endif
	117
7213b252 AK	118	/*
	119	* This is to handle the kprobes on user space access instructions
	120	*/
ae9a5b85	121	if (notify_page_fault(DIE_PAGE_FAULT, "page fault", regs, code, TRAP_BRKPT,
7213b252 AK	122	SIGSEGV) == NOTIFY_STOP)
	123	return;
	124
1da177e4 LT	125	down_read(&mm->mmap_sem);
	126
	127	vma = find_vma_prev(mm, address, &prev_vma);
	128	if (!vma)
	129	goto bad_area;
	130
	131	/* find_vma_prev() returns vma such that address < vma->vm_end or NULL */
	132	if (address < vma->vm_start)
	133	goto check_expansion;
	134
	135	good_area:
	136	code = SEGV_ACCERR;
	137
	138	/* OK, we've got a good vm_area for this memory area. Check the access permissions: */
	139
	140	# define VM_READ_BIT 0
	141	# define VM_WRITE_BIT 1
	142	# define VM_EXEC_BIT 2
	143
	144	# if (((1 << VM_READ_BIT) != VM_READ \|\| (1 << VM_WRITE_BIT) != VM_WRITE) \
	145	\|\| (1 << VM_EXEC_BIT) != VM_EXEC)
	146	# error File is out of sync with <linux/mm.h>. Please update.
	147	# endif
	148
df67b3da JB	149	if (((isr >> IA64_ISR_R_BIT) & 1UL) && (!(vma->vm_flags & (VM_READ \| VM_WRITE))))
	150	goto bad_area;
	151
1da177e4	152	mask = ( (((isr >> IA64_ISR_X_BIT) & 1UL) << VM_EXEC_BIT)
df67b3da	153	\| (((isr >> IA64_ISR_W_BIT) & 1UL) << VM_WRITE_BIT));
1da177e4 LT	154
	155	if ((vma->vm_flags & mask) != mask)
	156	goto bad_area;
	157
	158	survive:
	159	/*
	160	* If for any reason at all we couldn't handle the fault, make
	161	* sure we exit gracefully rather than endlessly redo the
	162	* fault.
	163	*/
	164	switch (handle_mm_fault(mm, vma, address, (mask & VM_WRITE) != 0)) {
	165	case VM_FAULT_MINOR:
	166	++current->min_flt;
	167	break;
	168	case VM_FAULT_MAJOR:
	169	++current->maj_flt;
	170	break;
	171	case VM_FAULT_SIGBUS:
	172	/*
	173	* We ran out of memory, or some other thing happened
	174	* to us that made us unable to handle the page fault
	175	* gracefully.
	176	*/
	177	signal = SIGBUS;
	178	goto bad_area;
	179	case VM_FAULT_OOM:
	180	goto out_of_memory;
	181	default:
	182	BUG();
	183	}
	184	up_read(&mm->mmap_sem);
	185	return;
	186
	187	check_expansion:
	188	if (!(prev_vma && (prev_vma->vm_flags & VM_GROWSUP) && (address == prev_vma->vm_end))) {
	189	if (!(vma->vm_flags & VM_GROWSDOWN))
	190	goto bad_area;
	191	if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
	192	\|\| REGION_OFFSET(address) >= RGN_MAP_LIMIT)
	193	goto bad_area;
	194	if (expand_stack(vma, address))
	195	goto bad_area;
	196	} else {
	197	vma = prev_vma;
	198	if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
	199	\|\| REGION_OFFSET(address) >= RGN_MAP_LIMIT)
	200	goto bad_area;
46dea3d0 HD	201	/*
	202	* Since the register backing store is accessed sequentially,
	203	* we disallow growing it by more than a page at a time.
	204	*/
	205	if (address > vma->vm_end + PAGE_SIZE - sizeof(long))
	206	goto bad_area;
	207	if (expand_upwards(vma, address))
1da177e4 LT	208	goto bad_area;
	209	}
	210	goto good_area;
	211
	212	bad_area:
	213	up_read(&mm->mmap_sem);
	214	#ifdef CONFIG_VIRTUAL_MEM_MAP
	215	bad_area_no_up:
	216	#endif
	217	if ((isr & IA64_ISR_SP)
	218	\|\| ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
	219	{
	220	/*
	221	* This fault was due to a speculative load or lfetch.fault, set the "ed"
	222	* bit in the psr to ensure forward progress. (Target register will get a
	223	* NaT for ld.s, lfetch will be canceled.)
	224	*/
	225	ia64_psr(regs)->ed = 1;
	226	return;
	227	}
	228	if (user_mode(regs)) {
	229	si.si_signo = signal;
	230	si.si_errno = 0;
	231	si.si_code = code;
	232	si.si_addr = (void __user *) address;
	233	si.si_isr = isr;
	234	si.si_flags = __ISR_VALID;
	235	force_sig_info(signal, &si, current);
	236	return;
	237	}
	238
	239	no_context:
f0a8d3c9 TL	240	if ((isr & IA64_ISR_SP)
	241	\|\| ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
	242	{
1da177e4	243	/*
f0a8d3c9 TL	244	* This fault was due to a speculative load or lfetch.fault, set the "ed"
	245	* bit in the psr to ensure forward progress. (Target register will get a
	246	* NaT for ld.s, lfetch will be canceled.)
1da177e4 LT	247	*/
	248	ia64_psr(regs)->ed = 1;
	249	return;
	250	}
	251
1da177e4 LT	252	/*
	253	* Since we have no vma's for region 5, we might get here even if the address is
	254	* valid, due to the VHPT walker inserting a non present translation that becomes
	255	* stale. If that happens, the non present fault handler already purged the stale
	256	* translation, which fixed the problem. So, we check to see if the translation is
	257	* valid, and return if it is.
	258	*/
	259	if (REGION_NUMBER(address) == 5 && mapped_kernel_page_is_present(address))
	260	return;
	261
63028aa7 KU	262	if (ia64_done_with_exception(regs))
	263	return;
	264
1da177e4 LT	265	/*
	266	* Oops. The kernel tried to access some bad page. We'll have to terminate things
	267	* with extreme prejudice.
	268	*/
	269	bust_spinlocks(1);
	270
	271	if (address < PAGE_SIZE)
	272	printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference (address %016lx)\n", address);
	273	else
	274	printk(KERN_ALERT "Unable to handle kernel paging request at "
	275	"virtual address %016lx\n", address);
	276	die("Oops", regs, isr);
	277	bust_spinlocks(0);
	278	do_exit(SIGKILL);
	279	return;
	280
	281	out_of_memory:
	282	up_read(&mm->mmap_sem);
	283	if (current->pid == 1) {
	284	yield();
	285	down_read(&mm->mmap_sem);
	286	goto survive;
	287	}
	288	printk(KERN_CRIT "VM: killing process %s\n", current->comm);
	289	if (user_mode(regs))
	290	do_exit(SIGKILL);
	291	goto no_context;
	292	}