2 * This file is subject to the terms and conditions of the GNU General Public
3 * License. See the file "COPYING" in the main directory of this archive
6 * Copyright (C) 1995 - 2000 by Ralf Baechle
8 #include <linux/signal.h>
9 #include <linux/sched.h>
10 #include <linux/interrupt.h>
11 #include <linux/kernel.h>
12 #include <linux/errno.h>
13 #include <linux/string.h>
14 #include <linux/types.h>
15 #include <linux/ptrace.h>
16 #include <linux/mman.h>
18 #include <linux/smp.h>
19 #include <linux/module.h>
20 #include <linux/kprobes.h>
22 #include <asm/branch.h>
23 #include <asm/mmu_context.h>
24 #include <asm/system.h>
25 #include <asm/uaccess.h>
26 #include <asm/ptrace.h>
27 #include <asm/highmem.h> /* For VMALLOC_END */
28 #include <linux/kdebug.h>
31 * This routine handles page faults. It determines the address,
32 * and the problem, and then passes it off to one of the appropriate
35 asmlinkage void __kprobes do_page_fault(struct pt_regs *regs, unsigned long write,
36 unsigned long address)
38 struct vm_area_struct * vma = NULL;
39 struct task_struct *tsk = current;
40 struct mm_struct *mm = tsk->mm;
41 const int field = sizeof(unsigned long) * 2;
46 printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", raw_smp_processor_id(),
47 current->comm, current->pid, field, address, write,
48 field, regs->cp0_epc);
53 * This is to notify the fault handler of the kprobes. The
54 * exception code is redundant as it is also carried in REGS,
55 * but we pass it anyhow.
57 if (notify_die(DIE_PAGE_FAULT, "page fault", regs, -1,
58 (regs->cp0_cause >> 2) & 0x1f, SIGSEGV) == NOTIFY_STOP)
62 info.si_code = SEGV_MAPERR;
65 * We fault-in kernel-space virtual memory on-demand. The
66 * 'reference' page table is init_mm.pgd.
68 * NOTE! We MUST NOT take any locks for this case. We may
69 * be in an interrupt or a critical region, and should
70 * only copy the information from the master page table,
74 # define VMALLOC_FAULT_TARGET no_context
76 # define VMALLOC_FAULT_TARGET vmalloc_fault
79 if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END))
80 goto VMALLOC_FAULT_TARGET;
82 if (unlikely(address >= MODULE_START && address < MODULE_END))
83 goto VMALLOC_FAULT_TARGET;
87 * If we're in an interrupt or have no user
88 * context, we must not take the fault..
90 if (in_atomic() || !mm)
91 goto bad_area_nosemaphore;
93 down_read(&mm->mmap_sem);
94 vma = find_vma(mm, address);
97 if (vma->vm_start <= address)
99 if (!(vma->vm_flags & VM_GROWSDOWN))
101 if (expand_stack(vma, address))
104 * Ok, we have a good vm_area for this memory access, so
108 info.si_code = SEGV_ACCERR;
111 if (!(vma->vm_flags & VM_WRITE))
114 if (kernel_uses_smartmips_rixi) {
115 if (address == regs->cp0_epc && !(vma->vm_flags & VM_EXEC)) {
117 pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] XI violation\n",
118 raw_smp_processor_id(),
119 current->comm, current->pid,
120 field, address, write,
121 field, regs->cp0_epc);
125 if (!(vma->vm_flags & VM_READ)) {
127 pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] RI violation\n",
128 raw_smp_processor_id(),
129 current->comm, current->pid,
130 field, address, write,
131 field, regs->cp0_epc);
136 if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
142 * If for any reason at all we couldn't handle the fault,
143 * make sure we exit gracefully rather than endlessly redo
146 fault = handle_mm_fault(mm, vma, address, write ? FAULT_FLAG_WRITE : 0);
147 if (unlikely(fault & VM_FAULT_ERROR)) {
148 if (fault & VM_FAULT_OOM)
150 else if (fault & VM_FAULT_SIGBUS)
154 if (fault & VM_FAULT_MAJOR)
159 up_read(&mm->mmap_sem);
163 * Something tried to access memory that isn't in our memory map..
164 * Fix it, but check if it's kernel or user first..
167 up_read(&mm->mmap_sem);
169 bad_area_nosemaphore:
170 /* User mode accesses just cause a SIGSEGV */
171 if (user_mode(regs)) {
172 tsk->thread.cp0_badvaddr = address;
173 tsk->thread.error_code = write;
175 printk("do_page_fault() #2: sending SIGSEGV to %s for "
176 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
178 write ? "write access to" : "read access from",
180 field, (unsigned long) regs->cp0_epc,
181 field, (unsigned long) regs->regs[31]);
183 info.si_signo = SIGSEGV;
185 /* info.si_code has been set above */
186 info.si_addr = (void __user *) address;
187 force_sig_info(SIGSEGV, &info, tsk);
192 /* Are we prepared to handle this kernel fault? */
193 if (fixup_exception(regs)) {
194 current->thread.cp0_baduaddr = address;
199 * Oops. The kernel tried to access some bad page. We'll have to
200 * terminate things with extreme prejudice.
204 printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
205 "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
206 raw_smp_processor_id(), field, address, field, regs->cp0_epc,
207 field, regs->regs[31]);
212 * We ran out of memory, call the OOM killer, and return the userspace
213 * (which will retry the fault, or kill us if we got oom-killed).
215 up_read(&mm->mmap_sem);
216 pagefault_out_of_memory();
220 up_read(&mm->mmap_sem);
222 /* Kernel mode? Handle exceptions or die */
223 if (!user_mode(regs))
227 * Send a sigbus, regardless of whether we were in kernel
231 printk("do_page_fault() #3: sending SIGBUS to %s for "
232 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
234 write ? "write access to" : "read access from",
236 field, (unsigned long) regs->cp0_epc,
237 field, (unsigned long) regs->regs[31]);
239 tsk->thread.cp0_badvaddr = address;
240 info.si_signo = SIGBUS;
242 info.si_code = BUS_ADRERR;
243 info.si_addr = (void __user *) address;
244 force_sig_info(SIGBUS, &info, tsk);
251 * Synchronize this task's top level page-table
252 * with the 'reference' page table.
254 * Do _not_ use "tsk" here. We might be inside
255 * an interrupt in the middle of a task switch..
257 int offset = __pgd_offset(address);
263 pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset;
264 pgd_k = init_mm.pgd + offset;
266 if (!pgd_present(*pgd_k))
268 set_pgd(pgd, *pgd_k);
270 pud = pud_offset(pgd, address);
271 pud_k = pud_offset(pgd_k, address);
272 if (!pud_present(*pud_k))
275 pmd = pmd_offset(pud, address);
276 pmd_k = pmd_offset(pud_k, address);
277 if (!pmd_present(*pmd_k))
279 set_pmd(pmd, *pmd_k);
281 pte_k = pte_offset_kernel(pmd_k, address);
282 if (!pte_present(*pte_k))