[linux-2.6-block.git] / arch / microblaze / mm / fault.c

/*
 *  arch/microblaze/mm/fault.c
 *
 *    Copyright (C) 2007 Xilinx, Inc.  All rights reserved.
 *
 *  Derived from "arch/ppc/mm/fault.c"
 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
 *
 *  Derived from "arch/i386/mm/fault.c"
 *    Copyright (C) 1991, 1992, 1993, 1994  Linus Torvalds
 *
 *  Modified by Cort Dougan and Paul Mackerras.
 *
 * This file is subject to the terms and conditions of the GNU General
 * Public License.  See the file COPYING in the main directory of this
 * archive for more details.
 *
 */

#include <linux/extable.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/interrupt.h>

#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/mmu.h>
#include <linux/mmu_context.h>
#include <linux/uaccess.h>
#include <asm/exceptions.h>

static unsigned long pte_misses;	/* updated by do_page_fault() */
static unsigned long pte_errors;	/* updated by do_page_fault() */

/*
 * Check whether the instruction at regs->pc is a store using
 * an update addressing form which will update r1.
 */
static int store_updates_sp(struct pt_regs *regs)
{
	unsigned int inst;

	if (get_user(inst, (unsigned int __user *)regs->pc))
		return 0;
	/* check for 1 in the rD field */
	if (((inst >> 21) & 0x1f) != 1)
		return 0;
	/* check for store opcodes */
	if ((inst & 0xd0000000) == 0xd0000000)
		return 1;
	return 0;
}


/*
 * bad_page_fault is called when we have a bad access from the kernel.
 * It is called from do_page_fault above and from some of the procedures
 * in traps.c.
 */
void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
{
	const struct exception_table_entry *fixup;
/* MS: no context */
	/* Are we prepared to handle this fault?  */
	fixup = search_exception_tables(regs->pc);
	if (fixup) {
		regs->pc = fixup->fixup;
		return;
	}

	/* kernel has accessed a bad area */
	die("kernel access of bad area", regs, sig);
}

/*
 * The error_code parameter is ESR for a data fault,
 * 0 for an instruction fault.
 */
void do_page_fault(struct pt_regs *regs, unsigned long address,
		   unsigned long error_code)
{
	struct vm_area_struct *vma;
	struct mm_struct *mm = current->mm;
	int code = SEGV_MAPERR;
	int is_write = error_code & ESR_S;
	vm_fault_t fault;
	unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;

	regs->ear = address;
	regs->esr = error_code;

	/* On a kernel SLB miss we can only check for a valid exception entry */
	if (unlikely(kernel_mode(regs) && (address >= TASK_SIZE))) {
		pr_warn("kernel task_size exceed");
		_exception(SIGSEGV, regs, code, address);
	}

	/* for instr TLB miss and instr storage exception ESR_S is undefined */
	if ((error_code & 0x13) == 0x13 || (error_code & 0x11) == 0x11)
		is_write = 0;

	if (unlikely(faulthandler_disabled() || !mm)) {
		if (kernel_mode(regs))
			goto bad_area_nosemaphore;

		/* faulthandler_disabled() in user mode is really bad,
		   as is current->mm == NULL. */
		pr_emerg("Page fault in user mode with faulthandler_disabled(), mm = %p\n",
			 mm);
		pr_emerg("r15 = %lx  MSR = %lx\n",
		       regs->r15, regs->msr);
		die("Weird page fault", regs, SIGSEGV);
	}

	if (user_mode(regs))
		flags |= FAULT_FLAG_USER;

	/* When running in the kernel we expect faults to occur only to
	 * addresses in user space.  All other faults represent errors in the
	 * kernel and should generate an OOPS.  Unfortunately, in the case of an
	 * erroneous fault occurring in a code path which already holds mmap_sem
	 * we will deadlock attempting to validate the fault against the
	 * address space.  Luckily the kernel only validly references user
	 * space from well defined areas of code, which are listed in the
	 * exceptions table.
	 *
	 * As the vast majority of faults will be valid we will only perform
	 * the source reference check when there is a possibility of a deadlock.
	 * Attempt to lock the address space, if we cannot we then validate the
	 * source.  If this is invalid we can skip the address space check,
	 * thus avoiding the deadlock.
	 */
	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
		if (kernel_mode(regs) && !search_exception_tables(regs->pc))
			goto bad_area_nosemaphore;

retry:
		down_read(&mm->mmap_sem);
	}

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

	if (vma->vm_start <= address)
		goto good_area;

	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
		goto bad_area;

	if (unlikely(!is_write))
		goto bad_area;

	/*
	 * N.B. The ABI allows programs to access up to
	 * a few hundred bytes below the stack pointer (TBD).
	 * The kernel signal delivery code writes up to about 1.5kB
	 * below the stack pointer (r1) before decrementing it.
	 * The exec code can write slightly over 640kB to the stack
	 * before setting the user r1.  Thus we allow the stack to
	 * expand to 1MB without further checks.
	 */
	if (unlikely(address + 0x100000 < vma->vm_end)) {

		/* get user regs even if this fault is in kernel mode */
		struct pt_regs *uregs = current->thread.regs;
		if (uregs == NULL)
			goto bad_area;

		/*
		 * A user-mode access to an address a long way below
		 * the stack pointer is only valid if the instruction
		 * is one which would update the stack pointer to the
		 * address accessed if the instruction completed,
		 * i.e. either stwu rs,n(r1) or stwux rs,r1,rb
		 * (or the byte, halfword, float or double forms).
		 *
		 * If we don't check this then any write to the area
		 * between the last mapped region and the stack will
		 * expand the stack rather than segfaulting.
		 */
		if (address + 2048 < uregs->r1
			&& (kernel_mode(regs) || !store_updates_sp(regs)))
				goto bad_area;
	}
	if (expand_stack(vma, address))
		goto bad_area;

good_area:
	code = SEGV_ACCERR;

	/* a write */
	if (unlikely(is_write)) {
		if (unlikely(!(vma->vm_flags & VM_WRITE)))
			goto bad_area;
		flags |= FAULT_FLAG_WRITE;
	/* a read */
	} else {
		/* protection fault */
		if (unlikely(error_code & 0x08000000))
			goto bad_area;
		if (unlikely(!(vma->vm_flags & (VM_READ | VM_EXEC))))
			goto bad_area;
	}

	/*
	 * 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(vma, address, flags);

	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
		return;

	if (unlikely(fault & VM_FAULT_ERROR)) {
		if (fault & VM_FAULT_OOM)
			goto out_of_memory;
		else if (fault & VM_FAULT_SIGSEGV)
			goto bad_area;
		else if (fault & VM_FAULT_SIGBUS)
			goto do_sigbus;
		BUG();
	}

	if (flags & FAULT_FLAG_ALLOW_RETRY) {
		if (unlikely(fault & VM_FAULT_MAJOR))
			current->maj_flt++;
		else
			current->min_flt++;
		if (fault & VM_FAULT_RETRY) {
			flags &= ~FAULT_FLAG_ALLOW_RETRY;
			flags |= FAULT_FLAG_TRIED;

			/*
			 * No need to up_read(&mm->mmap_sem) as we would
			 * have already released it in __lock_page_or_retry
			 * in mm/filemap.c.
			 */

			goto retry;
		}
	}

	up_read(&mm->mmap_sem);

	/*
	 * keep track of tlb+htab misses that are good addrs but
	 * just need pte's created via handle_mm_fault()
	 * -- Cort
	 */
	pte_misses++;
	return;

bad_area:
	up_read(&mm->mmap_sem);

bad_area_nosemaphore:
	pte_errors++;

	/* User mode accesses cause a SIGSEGV */
	if (user_mode(regs)) {
		_exception(SIGSEGV, regs, code, address);
		return;
	}

	bad_page_fault(regs, address, SIGSEGV);
	return;

/*
 * 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 (!user_mode(regs))
		bad_page_fault(regs, address, SIGKILL);
	else
		pagefault_out_of_memory();
	return;

do_sigbus:
	up_read(&mm->mmap_sem);
	if (user_mode(regs)) {
		force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
		return;
	}
	bad_page_fault(regs, address, SIGBUS);
}
Commit	Line	Data
5de96121 MS	1	/*
	2	* arch/microblaze/mm/fault.c
	3	*
	4	* Copyright (C) 2007 Xilinx, Inc. All rights reserved.
	5	*
	6	* Derived from "arch/ppc/mm/fault.c"
	7	* Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
	8	*
	9	* Derived from "arch/i386/mm/fault.c"
	10	* Copyright (C) 1991, 1992, 1993, 1994 Linus Torvalds
	11	*
	12	* Modified by Cort Dougan and Paul Mackerras.
	13	*
	14	* This file is subject to the terms and conditions of the GNU General
	15	* Public License. See the file COPYING in the main directory of this
	16	* archive for more details.
	17	*
	18	*/
	19
a75f3108	20	#include <linux/extable.h>
5de96121 MS	21	#include <linux/signal.h>
	22	#include <linux/sched.h>
	23	#include <linux/kernel.h>
	24	#include <linux/errno.h>
	25	#include <linux/string.h>
	26	#include <linux/types.h>
	27	#include <linux/ptrace.h>
	28	#include <linux/mman.h>
	29	#include <linux/mm.h>
	30	#include <linux/interrupt.h>
	31
	32	#include <asm/page.h>
	33	#include <asm/pgtable.h>
	34	#include <asm/mmu.h>
6bd55f0b	35	#include <linux/mmu_context.h>
5de96121 MS	36	#include <linux/uaccess.h>
	37	#include <asm/exceptions.h>
	38
5de96121 MS	39	static unsigned long pte_misses; /* updated by do_page_fault() */
	40	static unsigned long pte_errors; /* updated by do_page_fault() */
	41
	42	/*
	43	* Check whether the instruction at regs->pc is a store using
	44	* an update addressing form which will update r1.
	45	*/
	46	static int store_updates_sp(struct pt_regs *regs)
	47	{
	48	unsigned int inst;
	49
5213a9c3	50	if (get_user(inst, (unsigned int __user *)regs->pc))
5de96121 MS	51	return 0;
	52	/* check for 1 in the rD field */
	53	if (((inst >> 21) & 0x1f) != 1)
	54	return 0;
	55	/* check for store opcodes */
	56	if ((inst & 0xd0000000) == 0xd0000000)
	57	return 1;
	58	return 0;
	59	}
	60
	61
	62	/*
	63	* bad_page_fault is called when we have a bad access from the kernel.
	64	* It is called from do_page_fault above and from some of the procedures
	65	* in traps.c.
	66	*/
3863dbce	67	void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
5de96121 MS	68	{
	69	const struct exception_table_entry *fixup;
	70	/* MS: no context */
	71	/* Are we prepared to handle this fault? */
	72	fixup = search_exception_tables(regs->pc);
	73	if (fixup) {
	74	regs->pc = fixup->fixup;
	75	return;
	76	}
	77
	78	/* kernel has accessed a bad area */
5de96121 MS	79	die("kernel access of bad area", regs, sig);
	80	}
	81
	82	/*
	83	* The error_code parameter is ESR for a data fault,
	84	* 0 for an instruction fault.
	85	*/
	86	void do_page_fault(struct pt_regs *regs, unsigned long address,
	87	unsigned long error_code)
	88	{
	89	struct vm_area_struct *vma;
	90	struct mm_struct *mm = current->mm;
5de96121 MS	91	int code = SEGV_MAPERR;
5de96121 MS	92	int is_write = error_code & ESR_S;
50a7ca3c	93	vm_fault_t fault;
759496ba	94	unsigned int flags = FAULT_FLAG_ALLOW_RETRY \| FAULT_FLAG_KILLABLE;
5de96121 MS	95
	96	regs->ear = address;
	97	regs->esr = error_code;
	98
	99	/* On a kernel SLB miss we can only check for a valid exception entry */
78ebfa88	100	if (unlikely(kernel_mode(regs) && (address >= TASK_SIZE))) {
6bd55f0b	101	pr_warn("kernel task_size exceed");
5de96121 MS	102	_exception(SIGSEGV, regs, code, address);
	103	}
	104
	105	/* for instr TLB miss and instr storage exception ESR_S is undefined */
	106	if ((error_code & 0x13) == 0x13 \|\| (error_code & 0x11) == 0x11)
	107	is_write = 0;
	108
70ffdb93	109	if (unlikely(faulthandler_disabled() \|\| !mm)) {
f10eca6e MS	110	if (kernel_mode(regs))
	111	goto bad_area_nosemaphore;
	112
70ffdb93	113	/* faulthandler_disabled() in user mode is really bad,
5de96121	114	as is current->mm == NULL. */
70ffdb93 DH	115	pr_emerg("Page fault in user mode with faulthandler_disabled(), mm = %p\n",
70ffdb93 DH	116	mm);
6bd55f0b	117	pr_emerg("r15 = %lx MSR = %lx\n",
5de96121 MS	118	regs->r15, regs->msr);
	119	die("Weird page fault", regs, SIGSEGV);
	120	}
	121
759496ba JW	122	if (user_mode(regs))
	123	flags \|= FAULT_FLAG_USER;
	124
5de96121 MS	125	/* When running in the kernel we expect faults to occur only to
	126	* addresses in user space. All other faults represent errors in the
	127	* kernel and should generate an OOPS. Unfortunately, in the case of an
	128	* erroneous fault occurring in a code path which already holds mmap_sem
	129	* we will deadlock attempting to validate the fault against the
	130	* address space. Luckily the kernel only validly references user
	131	* space from well defined areas of code, which are listed in the
	132	* exceptions table.
	133	*
	134	* As the vast majority of faults will be valid we will only perform
	135	* the source reference check when there is a possibility of a deadlock.
	136	* Attempt to lock the address space, if we cannot we then validate the
	137	* source. If this is invalid we can skip the address space check,
	138	* thus avoiding the deadlock.
	139	*/
78ebfa88	140	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
5de96121 MS	141	if (kernel_mode(regs) && !search_exception_tables(regs->pc))
	142	goto bad_area_nosemaphore;
	143
f397c077	144	retry:
5de96121 MS	145	down_read(&mm->mmap_sem);
	146	}
	147
	148	vma = find_vma(mm, address);
78ebfa88	149	if (unlikely(!vma))
5de96121 MS	150	goto bad_area;
	151
	152	if (vma->vm_start <= address)
	153	goto good_area;
	154
78ebfa88	155	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
5de96121 MS	156	goto bad_area;
5de96121 MS	157
78ebfa88	158	if (unlikely(!is_write))
5de96121 MS	159	goto bad_area;
	160
	161	/*
	162	* N.B. The ABI allows programs to access up to
	163	* a few hundred bytes below the stack pointer (TBD).
	164	* The kernel signal delivery code writes up to about 1.5kB
	165	* below the stack pointer (r1) before decrementing it.
	166	* The exec code can write slightly over 640kB to the stack
	167	* before setting the user r1. Thus we allow the stack to
	168	* expand to 1MB without further checks.
	169	*/
78ebfa88	170	if (unlikely(address + 0x100000 < vma->vm_end)) {
5de96121 MS	171
	172	/* get user regs even if this fault is in kernel mode */
	173	struct pt_regs *uregs = current->thread.regs;
	174	if (uregs == NULL)
	175	goto bad_area;
	176
	177	/*
	178	* A user-mode access to an address a long way below
	179	* the stack pointer is only valid if the instruction
	180	* is one which would update the stack pointer to the
	181	* address accessed if the instruction completed,
	182	* i.e. either stwu rs,n(r1) or stwux rs,r1,rb
	183	* (or the byte, halfword, float or double forms).
	184	*
	185	* If we don't check this then any write to the area
	186	* between the last mapped region and the stack will
	187	* expand the stack rather than segfaulting.
	188	*/
	189	if (address + 2048 < uregs->r1
	190	&& (kernel_mode(regs) \|\| !store_updates_sp(regs)))
	191	goto bad_area;
	192	}
	193	if (expand_stack(vma, address))
	194	goto bad_area;
	195
	196	good_area:
	197	code = SEGV_ACCERR;
	198
	199	/* a write */
78ebfa88 MS	200	if (unlikely(is_write)) {
78ebfa88 MS	201	if (unlikely(!(vma->vm_flags & VM_WRITE)))
5de96121	202	goto bad_area;
759496ba	203	flags \|= FAULT_FLAG_WRITE;
5de96121 MS	204	/* a read */
	205	} else {
	206	/* protection fault */
78ebfa88	207	if (unlikely(error_code & 0x08000000))
5de96121	208	goto bad_area;
78ebfa88	209	if (unlikely(!(vma->vm_flags & (VM_READ \| VM_EXEC))))
5de96121 MS	210	goto bad_area;
	211	}
	212
	213	/*
	214	* If for any reason at all we couldn't handle the fault,
	215	* make sure we exit gracefully rather than endlessly redo
	216	* the fault.
	217	*/
dcddffd4	218	fault = handle_mm_fault(vma, address, flags);
f397c077 KC	219
	220	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
	221	return;
	222
5de96121 MS	223	if (unlikely(fault & VM_FAULT_ERROR)) {
	224	if (fault & VM_FAULT_OOM)
	225	goto out_of_memory;
33692f27 LT	226	else if (fault & VM_FAULT_SIGSEGV)
33692f27 LT	227	goto bad_area;
5de96121 MS	228	else if (fault & VM_FAULT_SIGBUS)
	229	goto do_sigbus;
	230	BUG();
	231	}
f397c077 KC	232
	233	if (flags & FAULT_FLAG_ALLOW_RETRY) {
	234	if (unlikely(fault & VM_FAULT_MAJOR))
	235	current->maj_flt++;
	236	else
	237	current->min_flt++;
	238	if (fault & VM_FAULT_RETRY) {
	239	flags &= ~FAULT_FLAG_ALLOW_RETRY;
45cac65b	240	flags \|= FAULT_FLAG_TRIED;
f397c077 KC	241
	242	/*
	243	* No need to up_read(&mm->mmap_sem) as we would
	244	* have already released it in __lock_page_or_retry
	245	* in mm/filemap.c.
	246	*/
	247
	248	goto retry;
	249	}
	250	}
	251
5de96121	252	up_read(&mm->mmap_sem);
f397c077	253
5de96121 MS	254	/*
	255	* keep track of tlb+htab misses that are good addrs but
	256	* just need pte's created via handle_mm_fault()
	257	* -- Cort
	258	*/
	259	pte_misses++;
	260	return;
	261
	262	bad_area:
	263	up_read(&mm->mmap_sem);
	264
	265	bad_area_nosemaphore:
	266	pte_errors++;
	267
	268	/* User mode accesses cause a SIGSEGV */
	269	if (user_mode(regs)) {
	270	_exception(SIGSEGV, regs, code, address);
5de96121 MS	271	return;
	272	}
	273
	274	bad_page_fault(regs, address, SIGSEGV);
	275	return;
	276
	277	/*
	278	* We ran out of memory, or some other thing happened to us that made
	279	* us unable to handle the page fault gracefully.
	280	*/
	281	out_of_memory:
5de96121	282	up_read(&mm->mmap_sem);
21e1c936 NP	283	if (!user_mode(regs))
	284	bad_page_fault(regs, address, SIGKILL);
	285	else
	286	pagefault_out_of_memory();
5de96121 MS	287	return;
	288
	289	do_sigbus:
	290	up_read(&mm->mmap_sem);
	291	if (user_mode(regs)) {
2e1661d2	292	force_sig_fault(SIGBUS, BUS_ADRERR, (void __user *)address);
5de96121 MS	293	return;
	294	}
	295	bad_page_fault(regs, address, SIGBUS);
	296	}