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>
21 #include <linux/perf_event.h>
23 #include <asm/branch.h>
24 #include <asm/mmu_context.h>
25 #include <asm/system.h>
26 #include <asm/uaccess.h>
27 #include <asm/ptrace.h>
28 #include <asm/highmem.h> /* For VMALLOC_END */
29 #include <linux/kdebug.h>
32 * This routine handles page faults. It determines the address,
33 * and the problem, and then passes it off to one of the appropriate
36 asmlinkage void __kprobes do_page_fault(struct pt_regs *regs, unsigned long write,
37 unsigned long address)
39 struct vm_area_struct * vma = NULL;
40 struct task_struct *tsk = current;
41 struct mm_struct *mm = tsk->mm;
42 const int field = sizeof(unsigned long) * 2;
47 printk("Cpu%d[%s:%d:%0*lx:%ld:%0*lx]\n", raw_smp_processor_id(),
48 current->comm, current->pid, field, address, write,
49 field, regs->cp0_epc);
54 * This is to notify the fault handler of the kprobes. The
55 * exception code is redundant as it is also carried in REGS,
56 * but we pass it anyhow.
58 if (notify_die(DIE_PAGE_FAULT, "page fault", regs, -1,
59 (regs->cp0_cause >> 2) & 0x1f, SIGSEGV) == NOTIFY_STOP)
63 info.si_code = SEGV_MAPERR;
66 * We fault-in kernel-space virtual memory on-demand. The
67 * 'reference' page table is init_mm.pgd.
69 * NOTE! We MUST NOT take any locks for this case. We may
70 * be in an interrupt or a critical region, and should
71 * only copy the information from the master page table,
75 # define VMALLOC_FAULT_TARGET no_context
77 # define VMALLOC_FAULT_TARGET vmalloc_fault
80 if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END))
81 goto VMALLOC_FAULT_TARGET;
83 if (unlikely(address >= MODULE_START && address < MODULE_END))
84 goto VMALLOC_FAULT_TARGET;
88 * If we're in an interrupt or have no user
89 * context, we must not take the fault..
91 if (in_atomic() || !mm)
92 goto bad_area_nosemaphore;
94 down_read(&mm->mmap_sem);
95 vma = find_vma(mm, address);
98 if (vma->vm_start <= address)
100 if (!(vma->vm_flags & VM_GROWSDOWN))
102 if (expand_stack(vma, address))
105 * Ok, we have a good vm_area for this memory access, so
109 info.si_code = SEGV_ACCERR;
112 if (!(vma->vm_flags & VM_WRITE))
115 if (kernel_uses_smartmips_rixi) {
116 if (address == regs->cp0_epc && !(vma->vm_flags & VM_EXEC)) {
118 pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] XI violation\n",
119 raw_smp_processor_id(),
120 current->comm, current->pid,
121 field, address, write,
122 field, regs->cp0_epc);
126 if (!(vma->vm_flags & VM_READ)) {
128 pr_notice("Cpu%d[%s:%d:%0*lx:%ld:%0*lx] RI violation\n",
129 raw_smp_processor_id(),
130 current->comm, current->pid,
131 field, address, write,
132 field, regs->cp0_epc);
137 if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
143 * If for any reason at all we couldn't handle the fault,
144 * make sure we exit gracefully rather than endlessly redo
147 fault = handle_mm_fault(mm, vma, address, write ? FAULT_FLAG_WRITE : 0);
148 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, 0, regs, address);
149 if (unlikely(fault & VM_FAULT_ERROR)) {
150 if (fault & VM_FAULT_OOM)
152 else if (fault & VM_FAULT_SIGBUS)
156 if (fault & VM_FAULT_MAJOR) {
157 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MAJ,
158 1, 0, regs, address);
161 perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS_MIN,
162 1, 0, regs, address);
166 up_read(&mm->mmap_sem);
170 * Something tried to access memory that isn't in our memory map..
171 * Fix it, but check if it's kernel or user first..
174 up_read(&mm->mmap_sem);
176 bad_area_nosemaphore:
177 /* User mode accesses just cause a SIGSEGV */
178 if (user_mode(regs)) {
179 tsk->thread.cp0_badvaddr = address;
180 tsk->thread.error_code = write;
182 printk("do_page_fault() #2: sending SIGSEGV to %s for "
183 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
185 write ? "write access to" : "read access from",
187 field, (unsigned long) regs->cp0_epc,
188 field, (unsigned long) regs->regs[31]);
190 info.si_signo = SIGSEGV;
192 /* info.si_code has been set above */
193 info.si_addr = (void __user *) address;
194 force_sig_info(SIGSEGV, &info, tsk);
199 /* Are we prepared to handle this kernel fault? */
200 if (fixup_exception(regs)) {
201 current->thread.cp0_baduaddr = address;
206 * Oops. The kernel tried to access some bad page. We'll have to
207 * terminate things with extreme prejudice.
211 printk(KERN_ALERT "CPU %d Unable to handle kernel paging request at "
212 "virtual address %0*lx, epc == %0*lx, ra == %0*lx\n",
213 raw_smp_processor_id(), field, address, field, regs->cp0_epc,
214 field, regs->regs[31]);
219 * We ran out of memory, call the OOM killer, and return the userspace
220 * (which will retry the fault, or kill us if we got oom-killed).
222 up_read(&mm->mmap_sem);
223 pagefault_out_of_memory();
227 up_read(&mm->mmap_sem);
229 /* Kernel mode? Handle exceptions or die */
230 if (!user_mode(regs))
234 * Send a sigbus, regardless of whether we were in kernel
238 printk("do_page_fault() #3: sending SIGBUS to %s for "
239 "invalid %s\n%0*lx (epc == %0*lx, ra == %0*lx)\n",
241 write ? "write access to" : "read access from",
243 field, (unsigned long) regs->cp0_epc,
244 field, (unsigned long) regs->regs[31]);
246 tsk->thread.cp0_badvaddr = address;
247 info.si_signo = SIGBUS;
249 info.si_code = BUS_ADRERR;
250 info.si_addr = (void __user *) address;
251 force_sig_info(SIGBUS, &info, tsk);
258 * Synchronize this task's top level page-table
259 * with the 'reference' page table.
261 * Do _not_ use "tsk" here. We might be inside
262 * an interrupt in the middle of a task switch..
264 int offset = __pgd_offset(address);
270 pgd = (pgd_t *) pgd_current[raw_smp_processor_id()] + offset;
271 pgd_k = init_mm.pgd + offset;
273 if (!pgd_present(*pgd_k))
275 set_pgd(pgd, *pgd_k);
277 pud = pud_offset(pgd, address);
278 pud_k = pud_offset(pgd_k, address);
279 if (!pud_present(*pud_k))
282 pmd = pmd_offset(pud, address);
283 pmd_k = pmd_offset(pud_k, address);
284 if (!pmd_present(*pmd_k))
286 set_pmd(pmd, *pmd_k);
288 pte_k = pte_offset_kernel(pmd_k, address);
289 if (!pte_present(*pte_k))