Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / arch / m32r / mm / fault.c
diff --git a/kernel/arch/m32r/mm/fault.c b/kernel/arch/m32r/mm/fault.c
new file mode 100644 (file)
index 0000000..8f9875b
--- /dev/null
@@ -0,0 +1,549 @@
+/*
+ *  linux/arch/m32r/mm/fault.c
+ *
+ *  Copyright (c) 2001, 2002  Hitoshi Yamamoto, and H. Kondo
+ *  Copyright (c) 2004  Naoto Sugai, NIIBE Yutaka
+ *
+ *  Some code taken from i386 version.
+ *    Copyright (C) 1995  Linus Torvalds
+ */
+
+#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/smp.h>
+#include <linux/interrupt.h>
+#include <linux/init.h>
+#include <linux/tty.h>
+#include <linux/vt_kern.h>             /* For unblank_screen() */
+#include <linux/highmem.h>
+#include <linux/module.h>
+#include <linux/uaccess.h>
+
+#include <asm/m32r.h>
+#include <asm/hardirq.h>
+#include <asm/mmu_context.h>
+#include <asm/tlbflush.h>
+
+extern void die(const char *, struct pt_regs *, long);
+
+#ifndef CONFIG_SMP
+asmlinkage unsigned int tlb_entry_i_dat;
+asmlinkage unsigned int tlb_entry_d_dat;
+#define tlb_entry_i tlb_entry_i_dat
+#define tlb_entry_d tlb_entry_d_dat
+#else
+unsigned int tlb_entry_i_dat[NR_CPUS];
+unsigned int tlb_entry_d_dat[NR_CPUS];
+#define tlb_entry_i tlb_entry_i_dat[smp_processor_id()]
+#define tlb_entry_d tlb_entry_d_dat[smp_processor_id()]
+#endif
+
+extern void init_tlb(void);
+
+/*======================================================================*
+ * do_page_fault()
+ *======================================================================*
+ * This routine handles page faults.  It determines the address,
+ * and the problem, and then passes it off to one of the appropriate
+ * routines.
+ *
+ * ARGUMENT:
+ *  regs       : M32R SP reg.
+ *  error_code : See below
+ *  address    : M32R MMU MDEVA reg. (Operand ACE)
+ *             : M32R BPC reg. (Instruction ACE)
+ *
+ * error_code :
+ *  bit 0 == 0 means no page found, 1 means protection fault
+ *  bit 1 == 0 means read, 1 means write
+ *  bit 2 == 0 means kernel, 1 means user-mode
+ *  bit 3 == 0 means data, 1 means instruction
+ *======================================================================*/
+#define ACE_PROTECTION         1
+#define ACE_WRITE              2
+#define ACE_USERMODE           4
+#define ACE_INSTRUCTION                8
+
+asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long error_code,
+  unsigned long address)
+{
+       struct task_struct *tsk;
+       struct mm_struct *mm;
+       struct vm_area_struct * vma;
+       unsigned long page, addr;
+       unsigned long flags = 0;
+       int fault;
+       siginfo_t info;
+
+       /*
+        * If BPSW IE bit enable --> set PSW IE bit
+        */
+       if (regs->psw & M32R_PSW_BIE)
+               local_irq_enable();
+
+       tsk = current;
+
+       info.si_code = SEGV_MAPERR;
+
+       /*
+        * We fault-in kernel-space virtual memory on-demand. The
+        * 'reference' page table is init_mm.pgd.
+        *
+        * NOTE! We MUST NOT take any locks for this case. We may
+        * be in an interrupt or a critical region, and should
+        * only copy the information from the master page table,
+        * nothing more.
+        *
+        * This verifies that the fault happens in kernel space
+        * (error_code & ACE_USERMODE) == 0, and that the fault was not a
+        * protection error (error_code & ACE_PROTECTION) == 0.
+        */
+       if (address >= TASK_SIZE && !(error_code & ACE_USERMODE))
+               goto vmalloc_fault;
+
+       mm = tsk->mm;
+
+       /*
+        * If we're in an interrupt or have no user context or have pagefaults
+        * disabled then we must not take the fault.
+        */
+       if (faulthandler_disabled() || !mm)
+               goto bad_area_nosemaphore;
+
+       if (error_code & ACE_USERMODE)
+               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 (!down_read_trylock(&mm->mmap_sem)) {
+               if ((error_code & ACE_USERMODE) == 0 &&
+                   !search_exception_tables(regs->psw))
+                       goto bad_area_nosemaphore;
+               down_read(&mm->mmap_sem);
+       }
+
+       vma = find_vma(mm, address);
+       if (!vma)
+               goto bad_area;
+       if (vma->vm_start <= address)
+               goto good_area;
+       if (!(vma->vm_flags & VM_GROWSDOWN))
+               goto bad_area;
+
+       if (error_code & ACE_USERMODE) {
+               /*
+                * accessing the stack below "spu" is always a bug.
+                * The "+ 4" is there due to the push instruction
+                * doing pre-decrement on the stack and that
+                * doesn't show up until later..
+                */
+               if (address + 4 < regs->spu)
+                       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:
+       info.si_code = SEGV_ACCERR;
+       switch (error_code & (ACE_WRITE|ACE_PROTECTION)) {
+               default:        /* 3: write, present */
+                       /* fall through */
+               case ACE_WRITE: /* write, not present */
+                       if (!(vma->vm_flags & VM_WRITE))
+                               goto bad_area;
+                       flags |= FAULT_FLAG_WRITE;
+                       break;
+               case ACE_PROTECTION:    /* read, present */
+               case 0:         /* read, not present */
+                       if (!(vma->vm_flags & (VM_READ | VM_EXEC)))
+                               goto bad_area;
+       }
+
+       /*
+        * For instruction access exception, check if the area is executable
+        */
+       if ((error_code & ACE_INSTRUCTION) && !(vma->vm_flags & 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.
+        */
+       addr = (address & PAGE_MASK);
+       set_thread_fault_code(error_code);
+       fault = handle_mm_fault(mm, vma, addr, flags);
+       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 (fault & VM_FAULT_MAJOR)
+               tsk->maj_flt++;
+       else
+               tsk->min_flt++;
+       set_thread_fault_code(0);
+       up_read(&mm->mmap_sem);
+       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:
+       up_read(&mm->mmap_sem);
+
+bad_area_nosemaphore:
+       /* User mode accesses just cause a SIGSEGV */
+       if (error_code & ACE_USERMODE) {
+               tsk->thread.address = address;
+               tsk->thread.error_code = error_code | (address >= TASK_SIZE);
+               tsk->thread.trap_no = 14;
+               info.si_signo = SIGSEGV;
+               info.si_errno = 0;
+               /* info.si_code has been set above */
+               info.si_addr = (void __user *)address;
+               force_sig_info(SIGSEGV, &info, tsk);
+               return;
+       }
+
+no_context:
+       /* Are we prepared to handle this kernel fault?  */
+       if (fixup_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");
+       else
+               printk(KERN_ALERT "Unable to handle kernel paging request");
+       printk(" at virtual address %08lx\n",address);
+       printk(KERN_ALERT " printing bpc:\n");
+       printk("%08lx\n", regs->bpc);
+       page = *(unsigned long *)MPTB;
+       page = ((unsigned long *) page)[address >> PGDIR_SHIFT];
+       printk(KERN_ALERT "*pde = %08lx\n", page);
+       if (page & _PAGE_PRESENT) {
+               page &= PAGE_MASK;
+               address &= 0x003ff000;
+               page = ((unsigned long *) __va(page))[address >> PAGE_SHIFT];
+               printk(KERN_ALERT "*pte = %08lx\n", page);
+       }
+       die("Oops", regs, error_code);
+       bust_spinlocks(0);
+       do_exit(SIGKILL);
+
+/*
+ * 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 (!(error_code & ACE_USERMODE))
+               goto no_context;
+       pagefault_out_of_memory();
+       return;
+
+do_sigbus:
+       up_read(&mm->mmap_sem);
+
+       /* Kernel mode? Handle exception or die */
+       if (!(error_code & ACE_USERMODE))
+               goto no_context;
+
+       tsk->thread.address = address;
+       tsk->thread.error_code = error_code;
+       tsk->thread.trap_no = 14;
+       info.si_signo = SIGBUS;
+       info.si_errno = 0;
+       info.si_code = BUS_ADRERR;
+       info.si_addr = (void __user *)address;
+       force_sig_info(SIGBUS, &info, tsk);
+       return;
+
+vmalloc_fault:
+       {
+               /*
+                * Synchronize this task's top level page-table
+                * with the 'reference' page table.
+                *
+                * Do _not_ use "tsk" here. We might be inside
+                * an interrupt in the middle of a task switch..
+                */
+               int offset = pgd_index(address);
+               pgd_t *pgd, *pgd_k;
+               pmd_t *pmd, *pmd_k;
+               pte_t *pte_k;
+
+               pgd = (pgd_t *)*(unsigned long *)MPTB;
+               pgd = offset + (pgd_t *)pgd;
+               pgd_k = init_mm.pgd + offset;
+
+               if (!pgd_present(*pgd_k))
+                       goto no_context;
+
+               /*
+                * set_pgd(pgd, *pgd_k); here would be useless on PAE
+                * and redundant with the set_pmd() on non-PAE.
+                */
+
+               pmd = pmd_offset(pgd, address);
+               pmd_k = pmd_offset(pgd_k, address);
+               if (!pmd_present(*pmd_k))
+                       goto no_context;
+               set_pmd(pmd, *pmd_k);
+
+               pte_k = pte_offset_kernel(pmd_k, address);
+               if (!pte_present(*pte_k))
+                       goto no_context;
+
+               addr = (address & PAGE_MASK);
+               set_thread_fault_code(error_code);
+               update_mmu_cache(NULL, addr, pte_k);
+               set_thread_fault_code(0);
+               return;
+       }
+}
+
+/*======================================================================*
+ * update_mmu_cache()
+ *======================================================================*/
+#define TLB_MASK       (NR_TLB_ENTRIES - 1)
+#define ITLB_END       (unsigned long *)(ITLB_BASE + (NR_TLB_ENTRIES * 8))
+#define DTLB_END       (unsigned long *)(DTLB_BASE + (NR_TLB_ENTRIES * 8))
+void update_mmu_cache(struct vm_area_struct *vma, unsigned long vaddr,
+       pte_t *ptep)
+{
+       volatile unsigned long *entry1, *entry2;
+       unsigned long pte_data, flags;
+       unsigned int *entry_dat;
+       int inst = get_thread_fault_code() & ACE_INSTRUCTION;
+       int i;
+
+       /* Ptrace may call this routine. */
+       if (vma && current->active_mm != vma->vm_mm)
+               return;
+
+       local_irq_save(flags);
+
+       vaddr = (vaddr & PAGE_MASK) | get_asid();
+
+       pte_data = pte_val(*ptep);
+
+#ifdef CONFIG_CHIP_OPSP
+       entry1 = (unsigned long *)ITLB_BASE;
+       for (i = 0; i < NR_TLB_ENTRIES; i++) {
+               if (*entry1++ == vaddr) {
+                       set_tlb_data(entry1, pte_data);
+                       break;
+               }
+               entry1++;
+       }
+       entry2 = (unsigned long *)DTLB_BASE;
+       for (i = 0; i < NR_TLB_ENTRIES; i++) {
+               if (*entry2++ == vaddr) {
+                       set_tlb_data(entry2, pte_data);
+                       break;
+               }
+               entry2++;
+       }
+#else
+       /*
+        * Update TLB entries
+        *  entry1: ITLB entry address
+        *  entry2: DTLB entry address
+        */
+       __asm__ __volatile__ (
+               "seth   %0, #high(%4)   \n\t"
+               "st     %2, @(%5, %0)   \n\t"
+               "ldi    %1, #1          \n\t"
+               "st     %1, @(%6, %0)   \n\t"
+               "add3   r4, %0, %7      \n\t"
+               ".fillinsn              \n"
+               "1:                     \n\t"
+               "ld     %1, @(%6, %0)   \n\t"
+               "bnez   %1, 1b          \n\t"
+               "ld     %0, @r4+        \n\t"
+               "ld     %1, @r4         \n\t"
+               "st     %3, @+%0        \n\t"
+               "st     %3, @+%1        \n\t"
+               : "=&r" (entry1), "=&r" (entry2)
+               : "r" (vaddr), "r" (pte_data), "i" (MMU_REG_BASE),
+               "i" (MSVA_offset), "i" (MTOP_offset), "i" (MIDXI_offset)
+               : "r4", "memory"
+       );
+#endif
+
+       if ((!inst && entry2 >= DTLB_END) || (inst && entry1 >= ITLB_END))
+               goto notfound;
+
+found:
+       local_irq_restore(flags);
+
+       return;
+
+       /* Valid entry not found */
+notfound:
+       /*
+        * Update ITLB or DTLB entry
+        *  entry1: TLB entry address
+        *  entry2: TLB base address
+        */
+       if (!inst) {
+               entry2 = (unsigned long *)DTLB_BASE;
+               entry_dat = &tlb_entry_d;
+       } else {
+               entry2 = (unsigned long *)ITLB_BASE;
+               entry_dat = &tlb_entry_i;
+       }
+       entry1 = entry2 + (((*entry_dat - 1) & TLB_MASK) << 1);
+
+       for (i = 0 ; i < NR_TLB_ENTRIES ; i++) {
+               if (!(entry1[1] & 2))   /* Valid bit check */
+                       break;
+
+               if (entry1 != entry2)
+                       entry1 -= 2;
+               else
+                       entry1 += TLB_MASK << 1;
+       }
+
+       if (i >= NR_TLB_ENTRIES) {      /* Empty entry not found */
+               entry1 = entry2 + (*entry_dat << 1);
+               *entry_dat = (*entry_dat + 1) & TLB_MASK;
+       }
+       *entry1++ = vaddr;      /* Set TLB tag */
+       set_tlb_data(entry1, pte_data);
+
+       goto found;
+}
+
+/*======================================================================*
+ * flush_tlb_page() : flushes one page
+ *======================================================================*/
+void local_flush_tlb_page(struct vm_area_struct *vma, unsigned long page)
+{
+       if (vma->vm_mm && mm_context(vma->vm_mm) != NO_CONTEXT) {
+               unsigned long flags;
+
+               local_irq_save(flags);
+               page &= PAGE_MASK;
+               page |= (mm_context(vma->vm_mm) & MMU_CONTEXT_ASID_MASK);
+               __flush_tlb_page(page);
+               local_irq_restore(flags);
+       }
+}
+
+/*======================================================================*
+ * flush_tlb_range() : flushes a range of pages
+ *======================================================================*/
+void local_flush_tlb_range(struct vm_area_struct *vma, unsigned long start,
+       unsigned long end)
+{
+       struct mm_struct *mm;
+
+       mm = vma->vm_mm;
+       if (mm_context(mm) != NO_CONTEXT) {
+               unsigned long flags;
+               int size;
+
+               local_irq_save(flags);
+               size = (end - start + (PAGE_SIZE - 1)) >> PAGE_SHIFT;
+               if (size > (NR_TLB_ENTRIES / 4)) { /* Too many TLB to flush */
+                       mm_context(mm) = NO_CONTEXT;
+                       if (mm == current->mm)
+                               activate_context(mm);
+               } else {
+                       unsigned long asid;
+
+                       asid = mm_context(mm) & MMU_CONTEXT_ASID_MASK;
+                       start &= PAGE_MASK;
+                       end += (PAGE_SIZE - 1);
+                       end &= PAGE_MASK;
+
+                       start |= asid;
+                       end   |= asid;
+                       while (start < end) {
+                               __flush_tlb_page(start);
+                               start += PAGE_SIZE;
+                       }
+               }
+               local_irq_restore(flags);
+       }
+}
+
+/*======================================================================*
+ * flush_tlb_mm() : flushes the specified mm context TLB's
+ *======================================================================*/
+void local_flush_tlb_mm(struct mm_struct *mm)
+{
+       /* Invalidate all TLB of this process. */
+       /* Instead of invalidating each TLB, we get new MMU context. */
+       if (mm_context(mm) != NO_CONTEXT) {
+               unsigned long flags;
+
+               local_irq_save(flags);
+               mm_context(mm) = NO_CONTEXT;
+               if (mm == current->mm)
+                       activate_context(mm);
+               local_irq_restore(flags);
+       }
+}
+
+/*======================================================================*
+ * flush_tlb_all() : flushes all processes TLBs
+ *======================================================================*/
+void local_flush_tlb_all(void)
+{
+       unsigned long flags;
+
+       local_irq_save(flags);
+       __flush_tlb_all();
+       local_irq_restore(flags);
+}
+
+/*======================================================================*
+ * init_mmu()
+ *======================================================================*/
+void __init init_mmu(void)
+{
+       tlb_entry_i = 0;
+       tlb_entry_d = 0;
+       mmu_context_cache = MMU_CONTEXT_FIRST_VERSION;
+       set_asid(mmu_context_cache & MMU_CONTEXT_ASID_MASK);
+       *(volatile unsigned long *)MPTB = (unsigned long)swapper_pg_dir;
+}