[linux-2.6-block.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/interrupt.h>
#include <linux/kprobes.h>
#include <linux/kdebug.h>

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

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

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

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

	return ret;
}
#else
static inline int notify_page_fault(struct pt_regs *regs, int trap)
{
	return 0;
}
#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;
	int fault;

	/* 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(regs, TRAP_BRKPT))
		return;

	down_read(&mm->mmap_sem);

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

        /*
         * find_vma_prev() returns vma such that address < vma->vm_end or NULL
         *
         * May find no vma, but could be that the last vm area is the
         * register backing store that needs to expand upwards, in
         * this case vma will be null, but prev_vma will ne non-null
         */
        if (( !vma && prev_vma ) || (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.
	 */
	fault = handle_mm_fault(mm, vma, address, (mask & VM_WRITE) != 0);
	if (unlikely(fault & VM_FAULT_ERROR)) {
		/*
		 * We ran out of memory, or some other thing happened
		 * to us that made us unable to handle the page fault
		 * gracefully.
		 */
		if (fault & VM_FAULT_OOM) {
			goto out_of_memory;
		} else if (fault & VM_FAULT_SIGBUS) {
			signal = SIGBUS;
			goto bad_area;
		}
		BUG();
	}
	if (fault & VM_FAULT_MAJOR)
		current->maj_flt++;
	else
		current->min_flt++;
	up_read(&mm->mmap_sem);
	return;

  check_expansion:
	if (!(prev_vma && (prev_vma->vm_flags & VM_GROWSUP) && (address == prev_vma->vm_end))) {
		if (!vma)
			goto bad_area;
		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);
	if (die("Oops", regs, isr))
		regs = NULL;
	bust_spinlocks(0);
	if (regs)
		do_exit(SIGKILL);
	return;

  out_of_memory:
	up_read(&mm->mmap_sem);
	if (is_global_init(current)) {
		yield();
		down_read(&mm->mmap_sem);
		goto survive;
	}
	printk(KERN_CRIT "VM: killing process %s\n", current->comm);
	if (user_mode(regs))
		do_group_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>
1da177e4	10	#include <linux/interrupt.h>
1f7ad57b	11	#include <linux/kprobes.h>
1eeb66a1	12	#include <linux/kdebug.h>
1da177e4 LT	13
	14	#include <asm/pgtable.h>
	15	#include <asm/processor.h>
	16	#include <asm/system.h>
	17	#include <asm/uaccess.h>
	18
620de2f5	19	extern int die(char , struct pt_regs , long);
1da177e4	20
ae9a5b85	21	#ifdef CONFIG_KPROBES
576fe0bd	22	static inline int notify_page_fault(struct pt_regs *regs, int trap)
ae9a5b85	23	{
576fe0bd	24	int ret = 0;
ae9a5b85	25
576fe0bd CH	26	if (!user_mode(regs)) {
	27	/* kprobe_running() needs smp_processor_id() */
	28	preempt_disable();
45e18c22	29	if (kprobe_running() && kprobe_fault_handler(regs, trap))
576fe0bd CH	30	ret = 1;
	31	preempt_enable();
	32	}
ae9a5b85	33
576fe0bd	34	return ret;
ae9a5b85 AK	35	}
ae9a5b85 AK	36	#else
576fe0bd	37	static inline int notify_page_fault(struct pt_regs *regs, int trap)
ae9a5b85	38	{
576fe0bd	39	return 0;
ae9a5b85 AK	40	}
	41	#endif
	42
1da177e4 LT	43	/*
	44	* Return TRUE if ADDRESS points at a page in the kernel's mapped segment
	45	* (inside region 5, on ia64) and that page is present.
	46	*/
	47	static int
	48	mapped_kernel_page_is_present (unsigned long address)
	49	{
	50	pgd_t *pgd;
	51	pud_t *pud;
	52	pmd_t *pmd;
	53	pte_t *ptep, pte;
	54
	55	pgd = pgd_offset_k(address);
	56	if (pgd_none(pgd) \|\| pgd_bad(pgd))
	57	return 0;
	58
	59	pud = pud_offset(pgd, address);
	60	if (pud_none(pud) \|\| pud_bad(pud))
	61	return 0;
	62
	63	pmd = pmd_offset(pud, address);
	64	if (pmd_none(pmd) \|\| pmd_bad(pmd))
	65	return 0;
	66
	67	ptep = pte_offset_kernel(pmd, address);
	68	if (!ptep)
	69	return 0;
	70
	71	pte = *ptep;
	72	return pte_present(pte);
	73	}
	74
1f7ad57b	75	void __kprobes
1da177e4 LT	76	ia64_do_page_fault (unsigned long address, unsigned long isr, struct pt_regs *regs)
	77	{
	78	int signal = SIGSEGV, code = SEGV_MAPERR;
	79	struct vm_area_struct vma, prev_vma;
	80	struct mm_struct *mm = current->mm;
	81	struct siginfo si;
	82	unsigned long mask;
83c54070	83	int fault;
1da177e4	84
0ffe9849 CL	85	/* mmap_sem is performance critical.... */
	86	prefetchw(&mm->mmap_sem);
	87
1da177e4 LT	88	/*
	89	* If we're in an interrupt or have no user context, we must not take the fault..
	90	*/
	91	if (in_atomic() \|\| !mm)
	92	goto no_context;
	93
	94	#ifdef CONFIG_VIRTUAL_MEM_MAP
	95	/*
	96	* If fault is in region 5 and we are in the kernel, we may already
	97	* have the mmap_sem (pfn_valid macro is called during mmap). There
	98	* is no vma for region 5 addr's anyway, so skip getting the semaphore
	99	* and go directly to the exception handling code.
	100	*/
	101
	102	if ((REGION_NUMBER(address) == 5) && !user_mode(regs))
	103	goto bad_area_no_up;
	104	#endif
	105
7213b252 AK	106	/*
	107	* This is to handle the kprobes on user space access instructions
	108	*/
576fe0bd	109	if (notify_page_fault(regs, TRAP_BRKPT))
7213b252 AK	110	return;
7213b252 AK	111
1da177e4 LT	112	down_read(&mm->mmap_sem);
	113
	114	vma = find_vma_prev(mm, address, &prev_vma);
e8c59c0c	115	if (!vma && !prev_vma )
1da177e4 LT	116	goto bad_area;
1da177e4 LT	117
e8c59c0c AB	118	/*
	119	* find_vma_prev() returns vma such that address < vma->vm_end or NULL
	120	*
	121	* May find no vma, but could be that the last vm area is the
	122	* register backing store that needs to expand upwards, in
	123	* this case vma will be null, but prev_vma will ne non-null
	124	*/
	125	if (( !vma && prev_vma ) \|\| (address < vma->vm_start) )
1da177e4 LT	126	goto check_expansion;
	127
	128	good_area:
	129	code = SEGV_ACCERR;
	130
	131	/* OK, we've got a good vm_area for this memory area. Check the access permissions: */
	132
	133	# define VM_READ_BIT 0
	134	# define VM_WRITE_BIT 1
	135	# define VM_EXEC_BIT 2
	136
	137	# if (((1 << VM_READ_BIT) != VM_READ \|\| (1 << VM_WRITE_BIT) != VM_WRITE) \
	138	\|\| (1 << VM_EXEC_BIT) != VM_EXEC)
	139	# error File is out of sync with <linux/mm.h>. Please update.
	140	# endif
	141
df67b3da JB	142	if (((isr >> IA64_ISR_R_BIT) & 1UL) && (!(vma->vm_flags & (VM_READ \| VM_WRITE))))
	143	goto bad_area;
	144
1da177e4	145	mask = ( (((isr >> IA64_ISR_X_BIT) & 1UL) << VM_EXEC_BIT)
df67b3da	146	\| (((isr >> IA64_ISR_W_BIT) & 1UL) << VM_WRITE_BIT));
1da177e4 LT	147
	148	if ((vma->vm_flags & mask) != mask)
	149	goto bad_area;
	150
	151	survive:
	152	/*
	153	* If for any reason at all we couldn't handle the fault, make
	154	* sure we exit gracefully rather than endlessly redo the
	155	* fault.
	156	*/
83c54070 NP	157	fault = handle_mm_fault(mm, vma, address, (mask & VM_WRITE) != 0);
83c54070 NP	158	if (unlikely(fault & VM_FAULT_ERROR)) {
1da177e4 LT	159	/*
	160	* We ran out of memory, or some other thing happened
	161	* to us that made us unable to handle the page fault
	162	* gracefully.
	163	*/
83c54070 NP	164	if (fault & VM_FAULT_OOM) {
	165	goto out_of_memory;
	166	} else if (fault & VM_FAULT_SIGBUS) {
	167	signal = SIGBUS;
	168	goto bad_area;
	169	}
1da177e4 LT	170	BUG();
1da177e4 LT	171	}
83c54070 NP	172	if (fault & VM_FAULT_MAJOR)
	173	current->maj_flt++;
	174	else
	175	current->min_flt++;
1da177e4 LT	176	up_read(&mm->mmap_sem);
	177	return;
	178
	179	check_expansion:
	180	if (!(prev_vma && (prev_vma->vm_flags & VM_GROWSUP) && (address == prev_vma->vm_end))) {
e8c59c0c AB	181	if (!vma)
e8c59c0c AB	182	goto bad_area;
1da177e4 LT	183	if (!(vma->vm_flags & VM_GROWSDOWN))
	184	goto bad_area;
	185	if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
	186	\|\| REGION_OFFSET(address) >= RGN_MAP_LIMIT)
	187	goto bad_area;
	188	if (expand_stack(vma, address))
	189	goto bad_area;
	190	} else {
	191	vma = prev_vma;
	192	if (REGION_NUMBER(address) != REGION_NUMBER(vma->vm_start)
	193	\|\| REGION_OFFSET(address) >= RGN_MAP_LIMIT)
	194	goto bad_area;
46dea3d0 HD	195	/*
	196	* Since the register backing store is accessed sequentially,
	197	* we disallow growing it by more than a page at a time.
	198	*/
	199	if (address > vma->vm_end + PAGE_SIZE - sizeof(long))
	200	goto bad_area;
	201	if (expand_upwards(vma, address))
1da177e4 LT	202	goto bad_area;
	203	}
	204	goto good_area;
	205
	206	bad_area:
	207	up_read(&mm->mmap_sem);
	208	#ifdef CONFIG_VIRTUAL_MEM_MAP
	209	bad_area_no_up:
	210	#endif
	211	if ((isr & IA64_ISR_SP)
	212	\|\| ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
	213	{
	214	/*
	215	* This fault was due to a speculative load or lfetch.fault, set the "ed"
	216	* bit in the psr to ensure forward progress. (Target register will get a
	217	* NaT for ld.s, lfetch will be canceled.)
	218	*/
	219	ia64_psr(regs)->ed = 1;
	220	return;
	221	}
	222	if (user_mode(regs)) {
	223	si.si_signo = signal;
	224	si.si_errno = 0;
	225	si.si_code = code;
	226	si.si_addr = (void __user *) address;
	227	si.si_isr = isr;
	228	si.si_flags = __ISR_VALID;
	229	force_sig_info(signal, &si, current);
	230	return;
	231	}
	232
	233	no_context:
f0a8d3c9 TL	234	if ((isr & IA64_ISR_SP)
	235	\|\| ((isr & IA64_ISR_NA) && (isr & IA64_ISR_CODE_MASK) == IA64_ISR_CODE_LFETCH))
	236	{
1da177e4	237	/*
f0a8d3c9 TL	238	* This fault was due to a speculative load or lfetch.fault, set the "ed"
	239	* bit in the psr to ensure forward progress. (Target register will get a
	240	* NaT for ld.s, lfetch will be canceled.)
1da177e4 LT	241	*/
	242	ia64_psr(regs)->ed = 1;
	243	return;
	244	}
	245
1da177e4 LT	246	/*
	247	* Since we have no vma's for region 5, we might get here even if the address is
	248	* valid, due to the VHPT walker inserting a non present translation that becomes
	249	* stale. If that happens, the non present fault handler already purged the stale
	250	* translation, which fixed the problem. So, we check to see if the translation is
	251	* valid, and return if it is.
	252	*/
	253	if (REGION_NUMBER(address) == 5 && mapped_kernel_page_is_present(address))
	254	return;
	255
63028aa7 KU	256	if (ia64_done_with_exception(regs))
	257	return;
	258
1da177e4 LT	259	/*
	260	* Oops. The kernel tried to access some bad page. We'll have to terminate things
	261	* with extreme prejudice.
	262	*/
	263	bust_spinlocks(1);
	264
	265	if (address < PAGE_SIZE)
	266	printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference (address %016lx)\n", address);
	267	else
	268	printk(KERN_ALERT "Unable to handle kernel paging request at "
	269	"virtual address %016lx\n", address);
620de2f5 JB	270	if (die("Oops", regs, isr))
620de2f5 JB	271	regs = NULL;
1da177e4	272	bust_spinlocks(0);
620de2f5 JB	273	if (regs)
620de2f5 JB	274	do_exit(SIGKILL);
1da177e4 LT	275	return;
	276
	277	out_of_memory:
	278	up_read(&mm->mmap_sem);
b460cbc5	279	if (is_global_init(current)) {
1da177e4 LT	280	yield();
	281	down_read(&mm->mmap_sem);
	282	goto survive;
	283	}
	284	printk(KERN_CRIT "VM: killing process %s\n", current->comm);
	285	if (user_mode(regs))
dcca2bde	286	do_group_exit(SIGKILL);
1da177e4 LT	287	goto no_context;
1da177e4 LT	288	}