[people/ms/linux.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/module.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;
	siginfo_t info;
	int code = SEGV_MAPERR;
	int is_write = error_code & ESR_S;
	int 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(in_atomic() || !mm)) {
		if (kernel_mode(regs))
			goto bad_area_nosemaphore;

		/* in_atomic() in user mode is really bad,
		   as is current->mm == NULL. */
		pr_emerg("Page fault in user mode with in_atomic(), 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(mm, 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);
/*		info.si_signo = SIGSEGV;
		info.si_errno = 0;
		info.si_code = code;
		info.si_addr = (void *) address;
		force_sig_info(SIGSEGV, &info, current);*/
		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)) {
		info.si_signo = SIGBUS;
		info.si_errno = 0;
		info.si_code = BUS_ADRERR;
		info.si_addr = (void __user *)address;
		force_sig_info(SIGBUS, &info, current);
		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
	20	#include <linux/module.h>
	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;
	91	siginfo_t info;
	92	int code = SEGV_MAPERR;
	93	int is_write = error_code & ESR_S;
	94	int fault;
759496ba	95	unsigned int flags = FAULT_FLAG_ALLOW_RETRY \| FAULT_FLAG_KILLABLE;
5de96121 MS	96
	97	regs->ear = address;
	98	regs->esr = error_code;
	99
	100	/* On a kernel SLB miss we can only check for a valid exception entry */
78ebfa88	101	if (unlikely(kernel_mode(regs) && (address >= TASK_SIZE))) {
6bd55f0b	102	pr_warn("kernel task_size exceed");
5de96121 MS	103	_exception(SIGSEGV, regs, code, address);
	104	}
	105
	106	/* for instr TLB miss and instr storage exception ESR_S is undefined */
	107	if ((error_code & 0x13) == 0x13 \|\| (error_code & 0x11) == 0x11)
	108	is_write = 0;
	109
78ebfa88	110	if (unlikely(in_atomic() \|\| !mm)) {
f10eca6e MS	111	if (kernel_mode(regs))
	112	goto bad_area_nosemaphore;
	113
5de96121 MS	114	/* in_atomic() in user mode is really bad,
5de96121 MS	115	as is current->mm == NULL. */
6bd55f0b MS	116	pr_emerg("Page fault in user mode with in_atomic(), mm = %p\n",
	117	mm);
	118	pr_emerg("r15 = %lx MSR = %lx\n",
5de96121 MS	119	regs->r15, regs->msr);
	120	die("Weird page fault", regs, SIGSEGV);
	121	}
	122
759496ba JW	123	if (user_mode(regs))
	124	flags \|= FAULT_FLAG_USER;
	125
5de96121 MS	126	/* When running in the kernel we expect faults to occur only to
	127	* addresses in user space. All other faults represent errors in the
	128	* kernel and should generate an OOPS. Unfortunately, in the case of an
	129	* erroneous fault occurring in a code path which already holds mmap_sem
	130	* we will deadlock attempting to validate the fault against the
	131	* address space. Luckily the kernel only validly references user
	132	* space from well defined areas of code, which are listed in the
	133	* exceptions table.
	134	*
	135	* As the vast majority of faults will be valid we will only perform
	136	* the source reference check when there is a possibility of a deadlock.
	137	* Attempt to lock the address space, if we cannot we then validate the
	138	* source. If this is invalid we can skip the address space check,
	139	* thus avoiding the deadlock.
	140	*/
78ebfa88	141	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
5de96121 MS	142	if (kernel_mode(regs) && !search_exception_tables(regs->pc))
	143	goto bad_area_nosemaphore;
	144
f397c077	145	retry:
5de96121 MS	146	down_read(&mm->mmap_sem);
	147	}
	148
	149	vma = find_vma(mm, address);
78ebfa88	150	if (unlikely(!vma))
5de96121 MS	151	goto bad_area;
	152
	153	if (vma->vm_start <= address)
	154	goto good_area;
	155
78ebfa88	156	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
5de96121 MS	157	goto bad_area;
5de96121 MS	158
78ebfa88	159	if (unlikely(!is_write))
5de96121 MS	160	goto bad_area;
	161
	162	/*
	163	* N.B. The ABI allows programs to access up to
	164	* a few hundred bytes below the stack pointer (TBD).
	165	* The kernel signal delivery code writes up to about 1.5kB
	166	* below the stack pointer (r1) before decrementing it.
	167	* The exec code can write slightly over 640kB to the stack
	168	* before setting the user r1. Thus we allow the stack to
	169	* expand to 1MB without further checks.
	170	*/
78ebfa88	171	if (unlikely(address + 0x100000 < vma->vm_end)) {
5de96121 MS	172
	173	/* get user regs even if this fault is in kernel mode */
	174	struct pt_regs *uregs = current->thread.regs;
	175	if (uregs == NULL)
	176	goto bad_area;
	177
	178	/*
	179	* A user-mode access to an address a long way below
	180	* the stack pointer is only valid if the instruction
	181	* is one which would update the stack pointer to the
	182	* address accessed if the instruction completed,
	183	* i.e. either stwu rs,n(r1) or stwux rs,r1,rb
	184	* (or the byte, halfword, float or double forms).
	185	*
	186	* If we don't check this then any write to the area
	187	* between the last mapped region and the stack will
	188	* expand the stack rather than segfaulting.
	189	*/
	190	if (address + 2048 < uregs->r1
	191	&& (kernel_mode(regs) \|\| !store_updates_sp(regs)))
	192	goto bad_area;
	193	}
	194	if (expand_stack(vma, address))
	195	goto bad_area;
	196
	197	good_area:
	198	code = SEGV_ACCERR;
	199
	200	/* a write */
78ebfa88 MS	201	if (unlikely(is_write)) {
78ebfa88 MS	202	if (unlikely(!(vma->vm_flags & VM_WRITE)))
5de96121	203	goto bad_area;
759496ba	204	flags \|= FAULT_FLAG_WRITE;
5de96121 MS	205	/* a read */
	206	} else {
	207	/* protection fault */
78ebfa88	208	if (unlikely(error_code & 0x08000000))
5de96121	209	goto bad_area;
78ebfa88	210	if (unlikely(!(vma->vm_flags & (VM_READ \| VM_EXEC))))
5de96121 MS	211	goto bad_area;
	212	}
	213
	214	/*
	215	* If for any reason at all we couldn't handle the fault,
	216	* make sure we exit gracefully rather than endlessly redo
	217	* the fault.
	218	*/
f397c077 KC	219	fault = handle_mm_fault(mm, vma, address, flags);
	220
	221	if ((fault & VM_FAULT_RETRY) && fatal_signal_pending(current))
	222	return;
	223
5de96121 MS	224	if (unlikely(fault & VM_FAULT_ERROR)) {
	225	if (fault & VM_FAULT_OOM)
	226	goto out_of_memory;
1868445f LT	227	else if (fault & VM_FAULT_SIGSEGV)
1868445f LT	228	goto bad_area;
5de96121 MS	229	else if (fault & VM_FAULT_SIGBUS)
	230	goto do_sigbus;
	231	BUG();
	232	}
f397c077 KC	233
	234	if (flags & FAULT_FLAG_ALLOW_RETRY) {
	235	if (unlikely(fault & VM_FAULT_MAJOR))
	236	current->maj_flt++;
	237	else
	238	current->min_flt++;
	239	if (fault & VM_FAULT_RETRY) {
	240	flags &= ~FAULT_FLAG_ALLOW_RETRY;
45cac65b	241	flags \|= FAULT_FLAG_TRIED;
f397c077 KC	242
	243	/*
	244	* No need to up_read(&mm->mmap_sem) as we would
	245	* have already released it in __lock_page_or_retry
	246	* in mm/filemap.c.
	247	*/
	248
	249	goto retry;
	250	}
	251	}
	252
5de96121	253	up_read(&mm->mmap_sem);
f397c077	254
5de96121 MS	255	/*
	256	* keep track of tlb+htab misses that are good addrs but
	257	* just need pte's created via handle_mm_fault()
	258	* -- Cort
	259	*/
	260	pte_misses++;
	261	return;
	262
	263	bad_area:
	264	up_read(&mm->mmap_sem);
	265
	266	bad_area_nosemaphore:
	267	pte_errors++;
	268
	269	/* User mode accesses cause a SIGSEGV */
	270	if (user_mode(regs)) {
	271	_exception(SIGSEGV, regs, code, address);
	272	/* info.si_signo = SIGSEGV;
	273	info.si_errno = 0;
	274	info.si_code = code;
	275	info.si_addr = (void *) address;
	276	force_sig_info(SIGSEGV, &info, current);*/
	277	return;
	278	}
	279
	280	bad_page_fault(regs, address, SIGSEGV);
	281	return;
	282
	283	/*
	284	* We ran out of memory, or some other thing happened to us that made
	285	* us unable to handle the page fault gracefully.
	286	*/
	287	out_of_memory:
5de96121	288	up_read(&mm->mmap_sem);
21e1c936 NP	289	if (!user_mode(regs))
	290	bad_page_fault(regs, address, SIGKILL);
	291	else
	292	pagefault_out_of_memory();
5de96121 MS	293	return;
	294
	295	do_sigbus:
	296	up_read(&mm->mmap_sem);
	297	if (user_mode(regs)) {
	298	info.si_signo = SIGBUS;
	299	info.si_errno = 0;
	300	info.si_code = BUS_ADRERR;
	301	info.si_addr = (void __user *)address;
	302	force_sig_info(SIGBUS, &info, current);
	303	return;
	304	}
	305	bad_page_fault(regs, address, SIGBUS);
	306	}