[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);

/*
 * 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;

	/*
	 * 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_die(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

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