[deliverable/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 <asm/mmu_context.h>
#include <asm/system.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;

	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))) {
		printk(KERN_WARNING "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. */
		printk(KERN_EMERG "Page fault in user mode with "
		       "in_atomic(), mm = %p\n", mm);
		printk(KERN_EMERG "r15 = %lx  MSR = %lx\n",
		       regs->r15, regs->msr);
		die("Weird page fault", regs, SIGSEGV);
	}

	/* 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;

		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;
	/* 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, is_write ? FAULT_FLAG_WRITE : 0);
	if (unlikely(fault & VM_FAULT_ERROR)) {
		if (fault & VM_FAULT_OOM)
			goto out_of_memory;
		else if (fault & VM_FAULT_SIGBUS)
			goto do_sigbus;
		BUG();
	}
	if (unlikely(fault & VM_FAULT_MAJOR))
		current->maj_flt++;
	else
		current->min_flt++;
	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>
	35	#include <asm/mmu_context.h>
	36	#include <asm/system.h>
	37	#include <linux/uaccess.h>
	38	#include <asm/exceptions.h>
	39
5de96121 MS	40	static unsigned long pte_misses; /* updated by do_page_fault() */
	41	static unsigned long pte_errors; /* updated by do_page_fault() */
	42
	43	/*
	44	* Check whether the instruction at regs->pc is a store using
	45	* an update addressing form which will update r1.
	46	*/
	47	static int store_updates_sp(struct pt_regs *regs)
	48	{
	49	unsigned int inst;
	50
5213a9c3	51	if (get_user(inst, (unsigned int __user *)regs->pc))
5de96121 MS	52	return 0;
	53	/* check for 1 in the rD field */
	54	if (((inst >> 21) & 0x1f) != 1)
	55	return 0;
	56	/* check for store opcodes */
	57	if ((inst & 0xd0000000) == 0xd0000000)
	58	return 1;
	59	return 0;
	60	}
	61
	62
	63	/*
	64	* bad_page_fault is called when we have a bad access from the kernel.
	65	* It is called from do_page_fault above and from some of the procedures
	66	* in traps.c.
	67	*/
3863dbce	68	void bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
5de96121 MS	69	{
	70	const struct exception_table_entry *fixup;
	71	/* MS: no context */
	72	/* Are we prepared to handle this fault? */
	73	fixup = search_exception_tables(regs->pc);
	74	if (fixup) {
	75	regs->pc = fixup->fixup;
	76	return;
	77	}
	78
	79	/* kernel has accessed a bad area */
5de96121 MS	80	die("kernel access of bad area", regs, sig);
	81	}
	82
	83	/*
	84	* The error_code parameter is ESR for a data fault,
	85	* 0 for an instruction fault.
	86	*/
	87	void do_page_fault(struct pt_regs *regs, unsigned long address,
	88	unsigned long error_code)
	89	{
	90	struct vm_area_struct *vma;
	91	struct mm_struct *mm = current->mm;
	92	siginfo_t info;
	93	int code = SEGV_MAPERR;
	94	int is_write = error_code & ESR_S;
	95	int fault;
	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))) {
5de96121 MS	102	printk(KERN_WARNING "kernel task_size exceed");
	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,
	115	as is current->mm == NULL. */
	116	printk(KERN_EMERG "Page fault in user mode with "
	117	"in_atomic(), mm = %p\n", mm);
	118	printk(KERN_EMERG "r15 = %lx MSR = %lx\n",
	119	regs->r15, regs->msr);
	120	die("Weird page fault", regs, SIGSEGV);
	121	}
	122
	123	/* When running in the kernel we expect faults to occur only to
	124	* addresses in user space. All other faults represent errors in the
	125	* kernel and should generate an OOPS. Unfortunately, in the case of an
	126	* erroneous fault occurring in a code path which already holds mmap_sem
	127	* we will deadlock attempting to validate the fault against the
	128	* address space. Luckily the kernel only validly references user
	129	* space from well defined areas of code, which are listed in the
	130	* exceptions table.
	131	*
	132	* As the vast majority of faults will be valid we will only perform
	133	* the source reference check when there is a possibility of a deadlock.
	134	* Attempt to lock the address space, if we cannot we then validate the
	135	* source. If this is invalid we can skip the address space check,
	136	* thus avoiding the deadlock.
	137	*/
78ebfa88	138	if (unlikely(!down_read_trylock(&mm->mmap_sem))) {
5de96121 MS	139	if (kernel_mode(regs) && !search_exception_tables(regs->pc))
	140	goto bad_area_nosemaphore;
	141
	142	down_read(&mm->mmap_sem);
	143	}
	144
	145	vma = find_vma(mm, address);
78ebfa88	146	if (unlikely(!vma))
5de96121 MS	147	goto bad_area;
	148
	149	if (vma->vm_start <= address)
	150	goto good_area;
	151
78ebfa88	152	if (unlikely(!(vma->vm_flags & VM_GROWSDOWN)))
5de96121 MS	153	goto bad_area;
5de96121 MS	154
78ebfa88	155	if (unlikely(!is_write))
5de96121 MS	156	goto bad_area;
	157
	158	/*
	159	* N.B. The ABI allows programs to access up to
	160	* a few hundred bytes below the stack pointer (TBD).
	161	* The kernel signal delivery code writes up to about 1.5kB
	162	* below the stack pointer (r1) before decrementing it.
	163	* The exec code can write slightly over 640kB to the stack
	164	* before setting the user r1. Thus we allow the stack to
	165	* expand to 1MB without further checks.
	166	*/
78ebfa88	167	if (unlikely(address + 0x100000 < vma->vm_end)) {
5de96121 MS	168
	169	/* get user regs even if this fault is in kernel mode */
	170	struct pt_regs *uregs = current->thread.regs;
	171	if (uregs == NULL)
	172	goto bad_area;
	173
	174	/*
	175	* A user-mode access to an address a long way below
	176	* the stack pointer is only valid if the instruction
	177	* is one which would update the stack pointer to the
	178	* address accessed if the instruction completed,
	179	* i.e. either stwu rs,n(r1) or stwux rs,r1,rb
	180	* (or the byte, halfword, float or double forms).
	181	*
	182	* If we don't check this then any write to the area
	183	* between the last mapped region and the stack will
	184	* expand the stack rather than segfaulting.
	185	*/
	186	if (address + 2048 < uregs->r1
	187	&& (kernel_mode(regs) \|\| !store_updates_sp(regs)))
	188	goto bad_area;
	189	}
	190	if (expand_stack(vma, address))
	191	goto bad_area;
	192
	193	good_area:
	194	code = SEGV_ACCERR;
	195
	196	/* a write */
78ebfa88 MS	197	if (unlikely(is_write)) {
78ebfa88 MS	198	if (unlikely(!(vma->vm_flags & VM_WRITE)))
5de96121 MS	199	goto bad_area;
	200	/* a read */
	201	} else {
	202	/* protection fault */
78ebfa88	203	if (unlikely(error_code & 0x08000000))
5de96121	204	goto bad_area;
78ebfa88	205	if (unlikely(!(vma->vm_flags & (VM_READ \| VM_EXEC))))
5de96121 MS	206	goto bad_area;
	207	}
	208
	209	/*
	210	* If for any reason at all we couldn't handle the fault,
	211	* make sure we exit gracefully rather than endlessly redo
	212	* the fault.
	213	*/
d06063cc	214	fault = handle_mm_fault(mm, vma, address, is_write ? FAULT_FLAG_WRITE : 0);
5de96121 MS	215	if (unlikely(fault & VM_FAULT_ERROR)) {
	216	if (fault & VM_FAULT_OOM)
	217	goto out_of_memory;
	218	else if (fault & VM_FAULT_SIGBUS)
	219	goto do_sigbus;
	220	BUG();
	221	}
78ebfa88	222	if (unlikely(fault & VM_FAULT_MAJOR))
5de96121 MS	223	current->maj_flt++;
	224	else
	225	current->min_flt++;
	226	up_read(&mm->mmap_sem);
	227	/*
	228	* keep track of tlb+htab misses that are good addrs but
	229	* just need pte's created via handle_mm_fault()
	230	* -- Cort
	231	*/
	232	pte_misses++;
	233	return;
	234
	235	bad_area:
	236	up_read(&mm->mmap_sem);
	237
	238	bad_area_nosemaphore:
	239	pte_errors++;
	240
	241	/* User mode accesses cause a SIGSEGV */
	242	if (user_mode(regs)) {
	243	_exception(SIGSEGV, regs, code, address);
	244	/* info.si_signo = SIGSEGV;
	245	info.si_errno = 0;
	246	info.si_code = code;
	247	info.si_addr = (void *) address;
	248	force_sig_info(SIGSEGV, &info, current);*/
	249	return;
	250	}
	251
	252	bad_page_fault(regs, address, SIGSEGV);
	253	return;
	254
	255	/*
	256	* We ran out of memory, or some other thing happened to us that made
	257	* us unable to handle the page fault gracefully.
	258	*/
	259	out_of_memory:
5de96121	260	up_read(&mm->mmap_sem);
21e1c936 NP	261	if (!user_mode(regs))
	262	bad_page_fault(regs, address, SIGKILL);
	263	else
	264	pagefault_out_of_memory();
5de96121 MS	265	return;
	266
	267	do_sigbus:
	268	up_read(&mm->mmap_sem);
	269	if (user_mode(regs)) {
	270	info.si_signo = SIGBUS;
	271	info.si_errno = 0;
	272	info.si_code = BUS_ADRERR;
	273	info.si_addr = (void __user *)address;
	274	force_sig_info(SIGBUS, &info, current);
	275	return;
	276	}
	277	bad_page_fault(regs, address, SIGBUS);
	278	}