[net-next-2.6.git] / arch / sparc64 / mm / fault.c

/* $Id: fault.c,v 1.59 2002/02/09 19:49:31 davem Exp $
 * arch/sparc64/mm/fault.c: Page fault handlers for the 64-bit Sparc.
 *
 * Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
 * Copyright (C) 1997, 1999 Jakub Jelinek (jj@ultra.linux.cz)
 */

#include <asm/head.h>

#include <linux/string.h>
#include <linux/types.h>
#include <linux/sched.h>
#include <linux/ptrace.h>
#include <linux/mman.h>
#include <linux/signal.h>
#include <linux/mm.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/interrupt.h>
#include <linux/kprobes.h>
#include <linux/kallsyms.h>
#include <linux/kdebug.h>

#include <asm/page.h>
#include <asm/pgtable.h>
#include <asm/openprom.h>
#include <asm/oplib.h>
#include <asm/uaccess.h>
#include <asm/asi.h>
#include <asm/lsu.h>
#include <asm/sections.h>
#include <asm/mmu_context.h>

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

	/* kprobe_running() needs smp_processor_id() */
	if (!user_mode(regs)) {
		preempt_disable();
		if (kprobe_running() && kprobe_fault_handler(regs, 0))
			ret = 1;
		preempt_enable();
	}
	return ret;
}
#else
static inline int notify_page_fault(struct pt_regs *regs)
{
	return 0;
}
#endif

/*
 * To debug kernel to catch accesses to certain virtual/physical addresses.
 * Mode = 0 selects physical watchpoints, mode = 1 selects virtual watchpoints.
 * flags = VM_READ watches memread accesses, flags = VM_WRITE watches memwrite accesses.
 * Caller passes in a 64bit aligned addr, with mask set to the bytes that need to be
 * watched. This is only useful on a single cpu machine for now. After the watchpoint
 * is detected, the process causing it will be killed, thus preventing an infinite loop.
 */
void set_brkpt(unsigned long addr, unsigned char mask, int flags, int mode)
{
	unsigned long lsubits;

	__asm__ __volatile__("ldxa [%%g0] %1, %0"
			     : "=r" (lsubits)
			     : "i" (ASI_LSU_CONTROL));
	lsubits &= ~(LSU_CONTROL_PM | LSU_CONTROL_VM |
		     LSU_CONTROL_PR | LSU_CONTROL_VR |
		     LSU_CONTROL_PW | LSU_CONTROL_VW);

	__asm__ __volatile__("stxa	%0, [%1] %2\n\t"
			     "membar	#Sync"
			     : /* no outputs */
			     : "r" (addr), "r" (mode ? VIRT_WATCHPOINT : PHYS_WATCHPOINT),
			       "i" (ASI_DMMU));

	lsubits |= ((unsigned long)mask << (mode ? 25 : 33));
	if (flags & VM_READ)
		lsubits |= (mode ? LSU_CONTROL_VR : LSU_CONTROL_PR);
	if (flags & VM_WRITE)
		lsubits |= (mode ? LSU_CONTROL_VW : LSU_CONTROL_PW);
	__asm__ __volatile__("stxa %0, [%%g0] %1\n\t"
			     "membar #Sync"
			     : /* no outputs */
			     : "r" (lsubits), "i" (ASI_LSU_CONTROL)
			     : "memory");
}

static void __kprobes unhandled_fault(unsigned long address,
				      struct task_struct *tsk,
				      struct pt_regs *regs)
{
	if ((unsigned long) address < PAGE_SIZE) {
		printk(KERN_ALERT "Unable to handle kernel NULL "
		       "pointer dereference\n");
	} else {
		printk(KERN_ALERT "Unable to handle kernel paging request "
		       "at virtual address %016lx\n", (unsigned long)address);
	}
	printk(KERN_ALERT "tsk->{mm,active_mm}->context = %016lx\n",
	       (tsk->mm ?
		CTX_HWBITS(tsk->mm->context) :
		CTX_HWBITS(tsk->active_mm->context)));
	printk(KERN_ALERT "tsk->{mm,active_mm}->pgd = %016lx\n",
	       (tsk->mm ? (unsigned long) tsk->mm->pgd :
		          (unsigned long) tsk->active_mm->pgd));
	die_if_kernel("Oops", regs);
}

static void bad_kernel_pc(struct pt_regs *regs, unsigned long vaddr)
{
	unsigned long *ksp;

	printk(KERN_CRIT "OOPS: Bogus kernel PC [%016lx] in fault handler\n",
	       regs->tpc);
	printk(KERN_CRIT "OOPS: RPC [%016lx]\n", regs->u_regs[15]);
	print_symbol("RPC: <%s>\n", regs->u_regs[15]);
	printk(KERN_CRIT "OOPS: Fault was to vaddr[%lx]\n", vaddr);
	__asm__("mov %%sp, %0" : "=r" (ksp));
	show_stack(current, ksp);
	unhandled_fault(regs->tpc, current, regs);
}

/*
 * We now make sure that mmap_sem is held in all paths that call 
 * this. Additionally, to prevent kswapd from ripping ptes from
 * under us, raise interrupts around the time that we look at the
 * pte, kswapd will have to wait to get his smp ipi response from
 * us. vmtruncate likewise. This saves us having to get pte lock.
 */
static unsigned int get_user_insn(unsigned long tpc)
{
	pgd_t *pgdp = pgd_offset(current->mm, tpc);
	pud_t *pudp;
	pmd_t *pmdp;
	pte_t *ptep, pte;
	unsigned long pa;
	u32 insn = 0;
	unsigned long pstate;

	if (pgd_none(*pgdp))
		goto outret;
	pudp = pud_offset(pgdp, tpc);
	if (pud_none(*pudp))
		goto outret;
	pmdp = pmd_offset(pudp, tpc);
	if (pmd_none(*pmdp))
		goto outret;

	/* This disables preemption for us as well. */
	__asm__ __volatile__("rdpr %%pstate, %0" : "=r" (pstate));
	__asm__ __volatile__("wrpr %0, %1, %%pstate"
				: : "r" (pstate), "i" (PSTATE_IE));
	ptep = pte_offset_map(pmdp, tpc);
	pte = *ptep;
	if (!pte_present(pte))
		goto out;

	pa  = (pte_pfn(pte) << PAGE_SHIFT);
	pa += (tpc & ~PAGE_MASK);

	/* Use phys bypass so we don't pollute dtlb/dcache. */
	__asm__ __volatile__("lduwa [%1] %2, %0"
			     : "=r" (insn)
			     : "r" (pa), "i" (ASI_PHYS_USE_EC));

out:
	pte_unmap(ptep);
	__asm__ __volatile__("wrpr %0, 0x0, %%pstate" : : "r" (pstate));
outret:
	return insn;
}

extern unsigned long compute_effective_address(struct pt_regs *, unsigned int, unsigned int);

static void do_fault_siginfo(int code, int sig, struct pt_regs *regs,
			     unsigned int insn, int fault_code)
{
	siginfo_t info;

	info.si_code = code;
	info.si_signo = sig;
	info.si_errno = 0;
	if (fault_code & FAULT_CODE_ITLB)
		info.si_addr = (void __user *) regs->tpc;
	else
		info.si_addr = (void __user *)
			compute_effective_address(regs, insn, 0);
	info.si_trapno = 0;
	force_sig_info(sig, &info, current);
}

extern int handle_ldf_stq(u32, struct pt_regs *);
extern int handle_ld_nf(u32, struct pt_regs *);

static unsigned int get_fault_insn(struct pt_regs *regs, unsigned int insn)
{
	if (!insn) {
		if (!regs->tpc || (regs->tpc & 0x3))
			return 0;
		if (regs->tstate & TSTATE_PRIV) {
			insn = *(unsigned int *) regs->tpc;
		} else {
			insn = get_user_insn(regs->tpc);
		}
	}
	return insn;
}

static void do_kernel_fault(struct pt_regs *regs, int si_code, int fault_code,
			    unsigned int insn, unsigned long address)
{
	unsigned char asi = ASI_P;
 
	if ((!insn) && (regs->tstate & TSTATE_PRIV))
		goto cannot_handle;

	/* If user insn could be read (thus insn is zero), that
	 * is fine.  We will just gun down the process with a signal
	 * in that case.
	 */

	if (!(fault_code & (FAULT_CODE_WRITE|FAULT_CODE_ITLB)) &&
	    (insn & 0xc0800000) == 0xc0800000) {
		if (insn & 0x2000)
			asi = (regs->tstate >> 24);
		else
			asi = (insn >> 5);
		if ((asi & 0xf2) == 0x82) {
			if (insn & 0x1000000) {
				handle_ldf_stq(insn, regs);
			} else {
				/* This was a non-faulting load. Just clear the
				 * destination register(s) and continue with the next
				 * instruction. -jj
				 */
				handle_ld_nf(insn, regs);
			}
			return;
		}
	}
		
	/* Is this in ex_table? */
	if (regs->tstate & TSTATE_PRIV) {
		const struct exception_table_entry *entry;

		if (asi == ASI_P && (insn & 0xc0800000) == 0xc0800000) {
			if (insn & 0x2000)
				asi = (regs->tstate >> 24);
			else
				asi = (insn >> 5);
		}
	
		/* Look in asi.h: All _S asis have LS bit set */
		if ((asi & 0x1) &&
		    (entry = search_exception_tables(regs->tpc))) {
			regs->tpc = entry->fixup;
			regs->tnpc = regs->tpc + 4;
			return;
		}
	} else {
		/* The si_code was set to make clear whether
		 * this was a SEGV_MAPERR or SEGV_ACCERR fault.
		 */
		do_fault_siginfo(si_code, SIGSEGV, regs, insn, fault_code);
		return;
	}

cannot_handle:
	unhandled_fault (address, current, regs);
}

asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
{
	struct mm_struct *mm = current->mm;
	struct vm_area_struct *vma;
	unsigned int insn = 0;
	int si_code, fault_code, fault;
	unsigned long address, mm_rss;

	fault_code = get_thread_fault_code();

	if (notify_page_fault(regs))
		return;

	si_code = SEGV_MAPERR;
	address = current_thread_info()->fault_address;

	if ((fault_code & FAULT_CODE_ITLB) &&
	    (fault_code & FAULT_CODE_DTLB))
		BUG();

	if (regs->tstate & TSTATE_PRIV) {
		unsigned long tpc = regs->tpc;

		/* Sanity check the PC. */
		if ((tpc >= KERNBASE && tpc < (unsigned long) _etext) ||
		    (tpc >= MODULES_VADDR && tpc < MODULES_END)) {
			/* Valid, no problems... */
		} else {
			bad_kernel_pc(regs, address);
			return;
		}
	}

	/*
	 * If we're in an interrupt or have no user
	 * context, we must not take the fault..
	 */
	if (in_atomic() || !mm)
		goto intr_or_no_mm;

	if (test_thread_flag(TIF_32BIT)) {
		if (!(regs->tstate & TSTATE_PRIV))
			regs->tpc &= 0xffffffff;
		address &= 0xffffffff;
	}

	if (!down_read_trylock(&mm->mmap_sem)) {
		if ((regs->tstate & TSTATE_PRIV) &&
		    !search_exception_tables(regs->tpc)) {
			insn = get_fault_insn(regs, insn);
			goto handle_kernel_fault;
		}
		down_read(&mm->mmap_sem);
	}

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

	/* Pure DTLB misses do not tell us whether the fault causing
	 * load/store/atomic was a write or not, it only says that there
	 * was no match.  So in such a case we (carefully) read the
	 * instruction to try and figure this out.  It's an optimization
	 * so it's ok if we can't do this.
	 *
	 * Special hack, window spill/fill knows the exact fault type.
	 */
	if (((fault_code &
	      (FAULT_CODE_DTLB | FAULT_CODE_WRITE | FAULT_CODE_WINFIXUP)) == FAULT_CODE_DTLB) &&
	    (vma->vm_flags & VM_WRITE) != 0) {
		insn = get_fault_insn(regs, 0);
		if (!insn)
			goto continue_fault;
		/* All loads, stores and atomics have bits 30 and 31 both set
		 * in the instruction.  Bit 21 is set in all stores, but we
		 * have to avoid prefetches which also have bit 21 set.
		 */
		if ((insn & 0xc0200000) == 0xc0200000 &&
		    (insn & 0x01780000) != 0x01680000) {
			/* Don't bother updating thread struct value,
			 * because update_mmu_cache only cares which tlb
			 * the access came from.
			 */
			fault_code |= FAULT_CODE_WRITE;
		}
	}
continue_fault:

	if (vma->vm_start <= address)
		goto good_area;
	if (!(vma->vm_flags & VM_GROWSDOWN))
		goto bad_area;
	if (!(fault_code & FAULT_CODE_WRITE)) {
		/* Non-faulting loads shouldn't expand stack. */
		insn = get_fault_insn(regs, insn);
		if ((insn & 0xc0800000) == 0xc0800000) {
			unsigned char asi;

			if (insn & 0x2000)
				asi = (regs->tstate >> 24);
			else
				asi = (insn >> 5);
			if ((asi & 0xf2) == 0x82)
				goto bad_area;
		}
	}
	if (expand_stack(vma, address))
		goto bad_area;
	/*
	 * Ok, we have a good vm_area for this memory access, so
	 * we can handle it..
	 */
good_area:
	si_code = SEGV_ACCERR;

	/* If we took a ITLB miss on a non-executable page, catch
	 * that here.
	 */
	if ((fault_code & FAULT_CODE_ITLB) && !(vma->vm_flags & VM_EXEC)) {
		BUG_ON(address != regs->tpc);
		BUG_ON(regs->tstate & TSTATE_PRIV);
		goto bad_area;
	}

	if (fault_code & FAULT_CODE_WRITE) {
		if (!(vma->vm_flags & VM_WRITE))
			goto bad_area;

		/* Spitfire has an icache which does not snoop
		 * processor stores.  Later processors do...
		 */
		if (tlb_type == spitfire &&
		    (vma->vm_flags & VM_EXEC) != 0 &&
		    vma->vm_file != NULL)
			set_thread_fault_code(fault_code |
					      FAULT_CODE_BLKCOMMIT);
	} else {
		/* Allow reads even for write-only mappings */
		if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
			goto bad_area;
	}

	fault = handle_mm_fault(mm, vma, address, (fault_code & FAULT_CODE_WRITE));
	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 (fault & VM_FAULT_MAJOR)
		current->maj_flt++;
	else
		current->min_flt++;

	up_read(&mm->mmap_sem);

	mm_rss = get_mm_rss(mm);
#ifdef CONFIG_HUGETLB_PAGE
	mm_rss -= (mm->context.huge_pte_count * (HPAGE_SIZE / PAGE_SIZE));
#endif
	if (unlikely(mm_rss >
		     mm->context.tsb_block[MM_TSB_BASE].tsb_rss_limit))
		tsb_grow(mm, MM_TSB_BASE, mm_rss);
#ifdef CONFIG_HUGETLB_PAGE
	mm_rss = mm->context.huge_pte_count;
	if (unlikely(mm_rss >
		     mm->context.tsb_block[MM_TSB_HUGE].tsb_rss_limit))
		tsb_grow(mm, MM_TSB_HUGE, mm_rss);
#endif
	return;

	/*
	 * Something tried to access memory that isn't in our memory map..
	 * Fix it, but check if it's kernel or user first..
	 */
bad_area:
	insn = get_fault_insn(regs, insn);
	up_read(&mm->mmap_sem);

handle_kernel_fault:
	do_kernel_fault(regs, si_code, fault_code, insn, address);
	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:
	insn = get_fault_insn(regs, insn);
	up_read(&mm->mmap_sem);
	printk("VM: killing process %s\n", current->comm);
	if (!(regs->tstate & TSTATE_PRIV))
		do_exit(SIGKILL);
	goto handle_kernel_fault;

intr_or_no_mm:
	insn = get_fault_insn(regs, 0);
	goto handle_kernel_fault;

do_sigbus:
	insn = get_fault_insn(regs, insn);
	up_read(&mm->mmap_sem);

	/*
	 * Send a sigbus, regardless of whether we were in kernel
	 * or user mode.
	 */
	do_fault_siginfo(BUS_ADRERR, SIGBUS, regs, insn, fault_code);

	/* Kernel mode? Handle exceptions or die */
	if (regs->tstate & TSTATE_PRIV)
		goto handle_kernel_fault;
}
Commit	Line	Data
1da177e4 LT	1	/* $Id: fault.c,v 1.59 2002/02/09 19:49:31 davem Exp $
	2	* arch/sparc64/mm/fault.c: Page fault handlers for the 64-bit Sparc.
	3	*
	4	* Copyright (C) 1996 David S. Miller (davem@caip.rutgers.edu)
	5	* Copyright (C) 1997, 1999 Jakub Jelinek (jj@ultra.linux.cz)
	6	*/
	7
	8	#include <asm/head.h>
	9
	10	#include <linux/string.h>
	11	#include <linux/types.h>
	12	#include <linux/sched.h>
	13	#include <linux/ptrace.h>
	14	#include <linux/mman.h>
	15	#include <linux/signal.h>
	16	#include <linux/mm.h>
	17	#include <linux/module.h>
1da177e4 LT	18	#include <linux/init.h>
1da177e4 LT	19	#include <linux/interrupt.h>
05e14cb3	20	#include <linux/kprobes.h>
eb398d10	21	#include <linux/kallsyms.h>
1eeb66a1	22	#include <linux/kdebug.h>
1da177e4 LT	23
	24	#include <asm/page.h>
	25	#include <asm/pgtable.h>
	26	#include <asm/openprom.h>
	27	#include <asm/oplib.h>
	28	#include <asm/uaccess.h>
	29	#include <asm/asi.h>
	30	#include <asm/lsu.h>
	31	#include <asm/sections.h>
7a1ac526	32	#include <asm/mmu_context.h>
1da177e4	33
d98f8f05	34	#ifdef CONFIG_KPROBES
127cda1e	35	static inline int notify_page_fault(struct pt_regs *regs)
d98f8f05	36	{
127cda1e DM	37	int ret = 0;
	38
	39	/* kprobe_running() needs smp_processor_id() */
	40	if (!user_mode(regs)) {
	41	preempt_disable();
	42	if (kprobe_running() && kprobe_fault_handler(regs, 0))
	43	ret = 1;
	44	preempt_enable();
	45	}
	46	return ret;
d98f8f05 AK	47	}
d98f8f05 AK	48	#else
127cda1e	49	static inline int notify_page_fault(struct pt_regs *regs)
d98f8f05	50	{
127cda1e	51	return 0;
d98f8f05 AK	52	}
	53	#endif
	54
1da177e4 LT	55	/*
	56	* To debug kernel to catch accesses to certain virtual/physical addresses.
	57	* Mode = 0 selects physical watchpoints, mode = 1 selects virtual watchpoints.
	58	* flags = VM_READ watches memread accesses, flags = VM_WRITE watches memwrite accesses.
	59	* Caller passes in a 64bit aligned addr, with mask set to the bytes that need to be
	60	* watched. This is only useful on a single cpu machine for now. After the watchpoint
	61	* is detected, the process causing it will be killed, thus preventing an infinite loop.
	62	*/
	63	void set_brkpt(unsigned long addr, unsigned char mask, int flags, int mode)
	64	{
	65	unsigned long lsubits;
	66
	67	__asm__ __volatile__("ldxa [%%g0] %1, %0"
	68	: "=r" (lsubits)
	69	: "i" (ASI_LSU_CONTROL));
	70	lsubits &= ~(LSU_CONTROL_PM \| LSU_CONTROL_VM \|
	71	LSU_CONTROL_PR \| LSU_CONTROL_VR \|
	72	LSU_CONTROL_PW \| LSU_CONTROL_VW);
	73
	74	__asm__ __volatile__("stxa %0, [%1] %2\n\t"
	75	"membar #Sync"
	76	: /* no outputs */
	77	: "r" (addr), "r" (mode ? VIRT_WATCHPOINT : PHYS_WATCHPOINT),
	78	"i" (ASI_DMMU));
	79
	80	lsubits \|= ((unsigned long)mask << (mode ? 25 : 33));
	81	if (flags & VM_READ)
	82	lsubits \|= (mode ? LSU_CONTROL_VR : LSU_CONTROL_PR);
	83	if (flags & VM_WRITE)
	84	lsubits \|= (mode ? LSU_CONTROL_VW : LSU_CONTROL_PW);
	85	__asm__ __volatile__("stxa %0, [%%g0] %1\n\t"
	86	"membar #Sync"
	87	: /* no outputs */
	88	: "r" (lsubits), "i" (ASI_LSU_CONTROL)
	89	: "memory");
	90	}
	91
05e14cb3 PP	92	static void __kprobes unhandled_fault(unsigned long address,
	93	struct task_struct *tsk,
	94	struct pt_regs *regs)
1da177e4 LT	95	{
	96	if ((unsigned long) address < PAGE_SIZE) {
	97	printk(KERN_ALERT "Unable to handle kernel NULL "
	98	"pointer dereference\n");
	99	} else {
	100	printk(KERN_ALERT "Unable to handle kernel paging request "
	101	"at virtual address %016lx\n", (unsigned long)address);
	102	}
	103	printk(KERN_ALERT "tsk->{mm,active_mm}->context = %016lx\n",
	104	(tsk->mm ?
	105	CTX_HWBITS(tsk->mm->context) :
	106	CTX_HWBITS(tsk->active_mm->context)));
	107	printk(KERN_ALERT "tsk->{mm,active_mm}->pgd = %016lx\n",
	108	(tsk->mm ? (unsigned long) tsk->mm->pgd :
	109	(unsigned long) tsk->active_mm->pgd));
1da177e4 LT	110	die_if_kernel("Oops", regs);
	111	}
	112
bf941d6c	113	static void bad_kernel_pc(struct pt_regs *regs, unsigned long vaddr)
1da177e4 LT	114	{
	115	unsigned long *ksp;
	116
	117	printk(KERN_CRIT "OOPS: Bogus kernel PC [%016lx] in fault handler\n",
	118	regs->tpc);
eb398d10 DM	119	printk(KERN_CRIT "OOPS: RPC [%016lx]\n", regs->u_regs[15]);
eb398d10 DM	120	print_symbol("RPC: <%s>\n", regs->u_regs[15]);
bf941d6c	121	printk(KERN_CRIT "OOPS: Fault was to vaddr[%lx]\n", vaddr);
1da177e4 LT	122	__asm__("mov %%sp, %0" : "=r" (ksp));
	123	show_stack(current, ksp);
	124	unhandled_fault(regs->tpc, current, regs);
	125	}
	126
	127	/*
	128	* We now make sure that mmap_sem is held in all paths that call
	129	* this. Additionally, to prevent kswapd from ripping ptes from
	130	* under us, raise interrupts around the time that we look at the
	131	* pte, kswapd will have to wait to get his smp ipi response from
da160546	132	* us. vmtruncate likewise. This saves us having to get pte lock.
1da177e4 LT	133	*/
	134	static unsigned int get_user_insn(unsigned long tpc)
	135	{
	136	pgd_t *pgdp = pgd_offset(current->mm, tpc);
	137	pud_t *pudp;
	138	pmd_t *pmdp;
	139	pte_t *ptep, pte;
	140	unsigned long pa;
	141	u32 insn = 0;
	142	unsigned long pstate;
	143
	144	if (pgd_none(*pgdp))
	145	goto outret;
	146	pudp = pud_offset(pgdp, tpc);
	147	if (pud_none(*pudp))
	148	goto outret;
	149	pmdp = pmd_offset(pudp, tpc);
	150	if (pmd_none(*pmdp))
	151	goto outret;
	152
	153	/* This disables preemption for us as well. */
	154	__asm__ __volatile__("rdpr %%pstate, %0" : "=r" (pstate));
	155	__asm__ __volatile__("wrpr %0, %1, %%pstate"
	156	: : "r" (pstate), "i" (PSTATE_IE));
	157	ptep = pte_offset_map(pmdp, tpc);
	158	pte = *ptep;
	159	if (!pte_present(pte))
	160	goto out;
	161
c4bce90e	162	pa = (pte_pfn(pte) << PAGE_SHIFT);
1da177e4 LT	163	pa += (tpc & ~PAGE_MASK);
	164
	165	/* Use phys bypass so we don't pollute dtlb/dcache. */
	166	__asm__ __volatile__("lduwa [%1] %2, %0"
	167	: "=r" (insn)
	168	: "r" (pa), "i" (ASI_PHYS_USE_EC));
	169
	170	out:
	171	pte_unmap(ptep);
	172	__asm__ __volatile__("wrpr %0, 0x0, %%pstate" : : "r" (pstate));
	173	outret:
	174	return insn;
	175	}
	176
	177	extern unsigned long compute_effective_address(struct pt_regs *, unsigned int, unsigned int);
	178
	179	static void do_fault_siginfo(int code, int sig, struct pt_regs *regs,
	180	unsigned int insn, int fault_code)
	181	{
	182	siginfo_t info;
	183
	184	info.si_code = code;
	185	info.si_signo = sig;
	186	info.si_errno = 0;
	187	if (fault_code & FAULT_CODE_ITLB)
	188	info.si_addr = (void __user *) regs->tpc;
	189	else
	190	info.si_addr = (void __user *)
	191	compute_effective_address(regs, insn, 0);
	192	info.si_trapno = 0;
	193	force_sig_info(sig, &info, current);
	194	}
	195
	196	extern int handle_ldf_stq(u32, struct pt_regs *);
	197	extern int handle_ld_nf(u32, struct pt_regs *);
	198
	199	static unsigned int get_fault_insn(struct pt_regs *regs, unsigned int insn)
	200	{
	201	if (!insn) {
	202	if (!regs->tpc \|\| (regs->tpc & 0x3))
	203	return 0;
	204	if (regs->tstate & TSTATE_PRIV) {
	205	insn = (unsigned int ) regs->tpc;
	206	} else {
	207	insn = get_user_insn(regs->tpc);
	208	}
	209	}
	210	return insn;
	211	}
	212
	213	static void do_kernel_fault(struct pt_regs *regs, int si_code, int fault_code,
	214	unsigned int insn, unsigned long address)
	215	{
1da177e4 LT	216	unsigned char asi = ASI_P;
	217
	218	if ((!insn) && (regs->tstate & TSTATE_PRIV))
	219	goto cannot_handle;
	220
	221	/* If user insn could be read (thus insn is zero), that
	222	* is fine. We will just gun down the process with a signal
	223	* in that case.
	224	*/
	225
	226	if (!(fault_code & (FAULT_CODE_WRITE\|FAULT_CODE_ITLB)) &&
	227	(insn & 0xc0800000) == 0xc0800000) {
	228	if (insn & 0x2000)
	229	asi = (regs->tstate >> 24);
	230	else
	231	asi = (insn >> 5);
	232	if ((asi & 0xf2) == 0x82) {
	233	if (insn & 0x1000000) {
	234	handle_ldf_stq(insn, regs);
	235	} else {
	236	/* This was a non-faulting load. Just clear the
	237	* destination register(s) and continue with the next
	238	* instruction. -jj
	239	*/
	240	handle_ld_nf(insn, regs);
	241	}
	242	return;
	243	}
	244	}
	245
1da177e4 LT	246	/* Is this in ex_table? */
1da177e4 LT	247	if (regs->tstate & TSTATE_PRIV) {
8cf14af0	248	const struct exception_table_entry *entry;
1da177e4 LT	249
	250	if (asi == ASI_P && (insn & 0xc0800000) == 0xc0800000) {
	251	if (insn & 0x2000)
	252	asi = (regs->tstate >> 24);
	253	else
	254	asi = (insn >> 5);
	255	}
	256
	257	/* Look in asi.h: All _S asis have LS bit set */
	258	if ((asi & 0x1) &&
8cf14af0 DM	259	(entry = search_exception_tables(regs->tpc))) {
8cf14af0 DM	260	regs->tpc = entry->fixup;
1da177e4	261	regs->tnpc = regs->tpc + 4;
1da177e4 LT	262	return;
	263	}
	264	} else {
	265	/* The si_code was set to make clear whether
	266	* this was a SEGV_MAPERR or SEGV_ACCERR fault.
	267	*/
	268	do_fault_siginfo(si_code, SIGSEGV, regs, insn, fault_code);
	269	return;
	270	}
	271
	272	cannot_handle:
	273	unhandled_fault (address, current, regs);
	274	}
	275
05e14cb3	276	asmlinkage void __kprobes do_sparc64_fault(struct pt_regs *regs)
1da177e4 LT	277	{
	278	struct mm_struct *mm = current->mm;
	279	struct vm_area_struct *vma;
	280	unsigned int insn = 0;
83c54070	281	int si_code, fault_code, fault;
7a1ac526	282	unsigned long address, mm_rss;
1da177e4 LT	283
	284	fault_code = get_thread_fault_code();
	285
127cda1e	286	if (notify_page_fault(regs))
1da177e4 LT	287	return;
	288
	289	si_code = SEGV_MAPERR;
	290	address = current_thread_info()->fault_address;
	291
	292	if ((fault_code & FAULT_CODE_ITLB) &&
	293	(fault_code & FAULT_CODE_DTLB))
	294	BUG();
	295
	296	if (regs->tstate & TSTATE_PRIV) {
	297	unsigned long tpc = regs->tpc;
	298
	299	/* Sanity check the PC. */
	300	if ((tpc >= KERNBASE && tpc < (unsigned long) _etext) \|\|
	301	(tpc >= MODULES_VADDR && tpc < MODULES_END)) {
	302	/* Valid, no problems... */
	303	} else {
bf941d6c	304	bad_kernel_pc(regs, address);
1da177e4 LT	305	return;
	306	}
	307	}
	308
	309	/*
	310	* If we're in an interrupt or have no user
	311	* context, we must not take the fault..
	312	*/
	313	if (in_atomic() \|\| !mm)
	314	goto intr_or_no_mm;
	315
	316	if (test_thread_flag(TIF_32BIT)) {
	317	if (!(regs->tstate & TSTATE_PRIV))
	318	regs->tpc &= 0xffffffff;
	319	address &= 0xffffffff;
	320	}
	321
	322	if (!down_read_trylock(&mm->mmap_sem)) {
	323	if ((regs->tstate & TSTATE_PRIV) &&
	324	!search_exception_tables(regs->tpc)) {
	325	insn = get_fault_insn(regs, insn);
	326	goto handle_kernel_fault;
	327	}
	328	down_read(&mm->mmap_sem);
	329	}
	330
	331	vma = find_vma(mm, address);
	332	if (!vma)
	333	goto bad_area;
	334
	335	/* Pure DTLB misses do not tell us whether the fault causing
	336	* load/store/atomic was a write or not, it only says that there
	337	* was no match. So in such a case we (carefully) read the
	338	* instruction to try and figure this out. It's an optimization
	339	* so it's ok if we can't do this.
	340	*
	341	* Special hack, window spill/fill knows the exact fault type.
	342	*/
	343	if (((fault_code &
	344	(FAULT_CODE_DTLB \| FAULT_CODE_WRITE \| FAULT_CODE_WINFIXUP)) == FAULT_CODE_DTLB) &&
	345	(vma->vm_flags & VM_WRITE) != 0) {
	346	insn = get_fault_insn(regs, 0);
	347	if (!insn)
	348	goto continue_fault;
73c50a27 DM	349	/* All loads, stores and atomics have bits 30 and 31 both set
	350	* in the instruction. Bit 21 is set in all stores, but we
	351	* have to avoid prefetches which also have bit 21 set.
	352	*/
1da177e4	353	if ((insn & 0xc0200000) == 0xc0200000 &&
73c50a27	354	(insn & 0x01780000) != 0x01680000) {
1da177e4 LT	355	/* Don't bother updating thread struct value,
	356	* because update_mmu_cache only cares which tlb
	357	* the access came from.
	358	*/
	359	fault_code \|= FAULT_CODE_WRITE;
	360	}
	361	}
	362	continue_fault:
	363
	364	if (vma->vm_start <= address)
	365	goto good_area;
	366	if (!(vma->vm_flags & VM_GROWSDOWN))
	367	goto bad_area;
	368	if (!(fault_code & FAULT_CODE_WRITE)) {
	369	/* Non-faulting loads shouldn't expand stack. */
	370	insn = get_fault_insn(regs, insn);
	371	if ((insn & 0xc0800000) == 0xc0800000) {
	372	unsigned char asi;
	373
	374	if (insn & 0x2000)
	375	asi = (regs->tstate >> 24);
	376	else
	377	asi = (insn >> 5);
	378	if ((asi & 0xf2) == 0x82)
	379	goto bad_area;
	380	}
	381	}
	382	if (expand_stack(vma, address))
	383	goto bad_area;
	384	/*
	385	* Ok, we have a good vm_area for this memory access, so
	386	* we can handle it..
	387	*/
	388	good_area:
	389	si_code = SEGV_ACCERR;
	390
	391	/* If we took a ITLB miss on a non-executable page, catch
	392	* that here.
	393	*/
	394	if ((fault_code & FAULT_CODE_ITLB) && !(vma->vm_flags & VM_EXEC)) {
	395	BUG_ON(address != regs->tpc);
	396	BUG_ON(regs->tstate & TSTATE_PRIV);
	397	goto bad_area;
	398	}
	399
	400	if (fault_code & FAULT_CODE_WRITE) {
	401	if (!(vma->vm_flags & VM_WRITE))
	402	goto bad_area;
	403
	404	/* Spitfire has an icache which does not snoop
	405	* processor stores. Later processors do...
	406	*/
	407	if (tlb_type == spitfire &&
	408	(vma->vm_flags & VM_EXEC) != 0 &&
	409	vma->vm_file != NULL)
	410	set_thread_fault_code(fault_code \|
	411	FAULT_CODE_BLKCOMMIT);
	412	} else {
	413	/* Allow reads even for write-only mappings */
	414	if (!(vma->vm_flags & (VM_READ \| VM_EXEC)))
	415	goto bad_area;
	416	}
	417
83c54070 NP	418	fault = handle_mm_fault(mm, vma, address, (fault_code & FAULT_CODE_WRITE));
	419	if (unlikely(fault & VM_FAULT_ERROR)) {
	420	if (fault & VM_FAULT_OOM)
	421	goto out_of_memory;
	422	else if (fault & VM_FAULT_SIGBUS)
	423	goto do_sigbus;
1da177e4 LT	424	BUG();
1da177e4 LT	425	}
83c54070 NP	426	if (fault & VM_FAULT_MAJOR)
	427	current->maj_flt++;
	428	else
	429	current->min_flt++;
1da177e4 LT	430
1da177e4 LT	431	up_read(&mm->mmap_sem);
7a1ac526 DM	432
7a1ac526 DM	433	mm_rss = get_mm_rss(mm);
dcc1e8dd DM	434	#ifdef CONFIG_HUGETLB_PAGE
	435	mm_rss -= (mm->context.huge_pte_count * (HPAGE_SIZE / PAGE_SIZE));
	436	#endif
7bebd83d	437	if (unlikely(mm_rss >
dcc1e8dd DM	438	mm->context.tsb_block[MM_TSB_BASE].tsb_rss_limit))
	439	tsb_grow(mm, MM_TSB_BASE, mm_rss);
	440	#ifdef CONFIG_HUGETLB_PAGE
	441	mm_rss = mm->context.huge_pte_count;
7bebd83d	442	if (unlikely(mm_rss >
dcc1e8dd DM	443	mm->context.tsb_block[MM_TSB_HUGE].tsb_rss_limit))
	444	tsb_grow(mm, MM_TSB_HUGE, mm_rss);
	445	#endif
efdc1e20	446	return;
1da177e4 LT	447
	448	/*
	449	* Something tried to access memory that isn't in our memory map..
	450	* Fix it, but check if it's kernel or user first..
	451	*/
	452	bad_area:
	453	insn = get_fault_insn(regs, insn);
	454	up_read(&mm->mmap_sem);
	455
	456	handle_kernel_fault:
	457	do_kernel_fault(regs, si_code, fault_code, insn, address);
efdc1e20	458	return;
1da177e4 LT	459
	460	/*
	461	* We ran out of memory, or some other thing happened to us that made
	462	* us unable to handle the page fault gracefully.
	463	*/
	464	out_of_memory:
	465	insn = get_fault_insn(regs, insn);
	466	up_read(&mm->mmap_sem);
	467	printk("VM: killing process %s\n", current->comm);
	468	if (!(regs->tstate & TSTATE_PRIV))
	469	do_exit(SIGKILL);
	470	goto handle_kernel_fault;
	471
	472	intr_or_no_mm:
	473	insn = get_fault_insn(regs, 0);
	474	goto handle_kernel_fault;
	475
	476	do_sigbus:
	477	insn = get_fault_insn(regs, insn);
	478	up_read(&mm->mmap_sem);
	479
	480	/*
	481	* Send a sigbus, regardless of whether we were in kernel
	482	* or user mode.
	483	*/
	484	do_fault_siginfo(BUS_ADRERR, SIGBUS, regs, insn, fault_code);
	485
	486	/* Kernel mode? Handle exceptions or die */
	487	if (regs->tstate & TSTATE_PRIV)
	488	goto handle_kernel_fault;
1da177e4	489	}