Add the rt linux 4.1.3-rt3 as base
[kvmfornfv.git] / kernel / arch / xtensa / mm / fault.c
diff --git a/kernel/arch/xtensa/mm/fault.c b/kernel/arch/xtensa/mm/fault.c
new file mode 100644 (file)
index 0000000..83a44a3
--- /dev/null
@@ -0,0 +1,260 @@
+// TODO VM_EXEC flag work-around, cache aliasing
+/*
+ * arch/xtensa/mm/fault.c
+ *
+ * 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.
+ *
+ * Copyright (C) 2001 - 2010 Tensilica Inc.
+ *
+ * Chris Zankel <chris@zankel.net>
+ * Joe Taylor  <joe@tensilica.com, joetylr@yahoo.com>
+ */
+
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/hardirq.h>
+#include <linux/uaccess.h>
+#include <asm/mmu_context.h>
+#include <asm/cacheflush.h>
+#include <asm/hardirq.h>
+#include <asm/pgalloc.h>
+
+DEFINE_PER_CPU(unsigned long, asid_cache) = ASID_USER_FIRST;
+void bad_page_fault(struct pt_regs*, unsigned long, int);
+
+#undef DEBUG_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.
+ *
+ * Note: does not handle Miss and MultiHit.
+ */
+
+void do_page_fault(struct pt_regs *regs)
+{
+       struct vm_area_struct * vma;
+       struct mm_struct *mm = current->mm;
+       unsigned int exccause = regs->exccause;
+       unsigned int address = regs->excvaddr;
+       siginfo_t info;
+
+       int is_write, is_exec;
+       int fault;
+       unsigned int flags = FAULT_FLAG_ALLOW_RETRY | FAULT_FLAG_KILLABLE;
+
+       info.si_code = SEGV_MAPERR;
+
+       /* We fault-in kernel-space virtual memory on-demand. The
+        * 'reference' page table is init_mm.pgd.
+        */
+       if (address >= TASK_SIZE && !user_mode(regs))
+               goto vmalloc_fault;
+
+       /* If we're in an interrupt or have no user
+        * context, we must not take the fault..
+        */
+       if (faulthandler_disabled() || !mm) {
+               bad_page_fault(regs, address, SIGSEGV);
+               return;
+       }
+
+       is_write = (exccause == EXCCAUSE_STORE_CACHE_ATTRIBUTE) ? 1 : 0;
+       is_exec =  (exccause == EXCCAUSE_ITLB_PRIVILEGE ||
+                   exccause == EXCCAUSE_ITLB_MISS ||
+                   exccause == EXCCAUSE_FETCH_CACHE_ATTRIBUTE) ? 1 : 0;
+
+#ifdef DEBUG_PAGE_FAULT
+       printk("[%s:%d:%08x:%d:%08x:%s%s]\n", current->comm, current->pid,
+              address, exccause, regs->pc, is_write? "w":"", is_exec? "x":"");
+#endif
+
+       if (user_mode(regs))
+               flags |= FAULT_FLAG_USER;
+retry:
+       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 (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;
+
+       if (is_write) {
+               if (!(vma->vm_flags & VM_WRITE))
+                       goto bad_area;
+               flags |= FAULT_FLAG_WRITE;
+       } else if (is_exec) {
+               if (!(vma->vm_flags & VM_EXEC))
+                       goto bad_area;
+       } else  /* Allow read even from write-only pages. */
+               if (!(vma->vm_flags & (VM_READ | VM_WRITE)))
+                       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 (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);
+       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);
+       if (user_mode(regs)) {
+               current->thread.bad_vaddr = address;
+               current->thread.error_code = is_write;
+               info.si_signo = SIGSEGV;
+               info.si_errno = 0;
+               /* info.si_code has been set above */
+               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);
+
+       /* Send a sigbus, regardless of whether we were in kernel
+        * or user mode.
+        */
+       current->thread.bad_vaddr = address;
+       info.si_code = SIGBUS;
+       info.si_errno = 0;
+       info.si_code = BUS_ADRERR;
+       info.si_addr = (void *) address;
+       force_sig_info(SIGBUS, &info, current);
+
+       /* Kernel mode? Handle exceptions or die */
+       if (!user_mode(regs))
+               bad_page_fault(regs, address, SIGBUS);
+       return;
+
+vmalloc_fault:
+       {
+               /* Synchronize this task's top level page-table
+                * with the 'reference' page table.
+                */
+               struct mm_struct *act_mm = current->active_mm;
+               int index = pgd_index(address);
+               pgd_t *pgd, *pgd_k;
+               pmd_t *pmd, *pmd_k;
+               pte_t *pte_k;
+
+               if (act_mm == NULL)
+                       goto bad_page_fault;
+
+               pgd = act_mm->pgd + index;
+               pgd_k = init_mm.pgd + index;
+
+               if (!pgd_present(*pgd_k))
+                       goto bad_page_fault;
+
+               pgd_val(*pgd) = pgd_val(*pgd_k);
+
+               pmd = pmd_offset(pgd, address);
+               pmd_k = pmd_offset(pgd_k, address);
+               if (!pmd_present(*pmd) || !pmd_present(*pmd_k))
+                       goto bad_page_fault;
+
+               pmd_val(*pmd) = pmd_val(*pmd_k);
+               pte_k = pte_offset_kernel(pmd_k, address);
+
+               if (!pte_present(*pte_k))
+                       goto bad_page_fault;
+               return;
+       }
+bad_page_fault:
+       bad_page_fault(regs, address, SIGKILL);
+       return;
+}
+
+
+void
+bad_page_fault(struct pt_regs *regs, unsigned long address, int sig)
+{
+       extern void die(const char*, struct pt_regs*, long);
+       const struct exception_table_entry *entry;
+
+       /* Are we prepared to handle this kernel fault?  */
+       if ((entry = search_exception_tables(regs->pc)) != NULL) {
+#ifdef DEBUG_PAGE_FAULT
+               printk(KERN_DEBUG "%s: Exception at pc=%#010lx (%lx)\n",
+                               current->comm, regs->pc, entry->fixup);
+#endif
+               current->thread.bad_uaddr = address;
+               regs->pc = entry->fixup;
+               return;
+       }
+
+       /* Oops. The kernel tried to access some bad page. We'll have to
+        * terminate things with extreme prejudice.
+        */
+       printk(KERN_ALERT "Unable to handle kernel paging request at virtual "
+              "address %08lx\n pc = %08lx, ra = %08lx\n",
+              address, regs->pc, regs->areg[0]);
+       die("Oops", regs, sig);
+       do_exit(sig);
+}