1/*
  2 * Copyright (C) 2009 Wind River Systems Inc
  3 *   Implemented by fredrik.markstrom@gmail.com and ivarholmqvist@gmail.com
  4 *
  5 * based on arch/mips/mm/fault.c which is:
  6 *
  7 * Copyright (C) 1995-2000 Ralf Baechle
  8 *
  9 * This file is subject to the terms and conditions of the GNU General Public
 10 * License.  See the file "COPYING" in the main directory of this archive
 11 * for more details.
 12 */
 13
 14#include <linux/signal.h>
 15#include <linux/sched.h>
 16#include <linux/sched/debug.h>
 17#include <linux/interrupt.h>
 18#include <linux/kernel.h>
 19#include <linux/errno.h>
 20#include <linux/string.h>
 21#include <linux/types.h>
 22#include <linux/ptrace.h>
 23#include <linux/mman.h>
 24#include <linux/mm.h>
 25#include <linux/extable.h>
 26#include <linux/uaccess.h>
 27#include <linux/perf_event.h>
 28
 29#include <asm/mmu_context.h>
 30#include <asm/traps.h>
 31
 32#define EXC_SUPERV_INSN_ACCESS	9  /* Supervisor only instruction address */
 33#define EXC_SUPERV_DATA_ACCESS	11 /* Supervisor only data address */
 34#define EXC_X_PROTECTION_FAULT	13 /* TLB permission violation (x) */
 35#define EXC_R_PROTECTION_FAULT	14 /* TLB permission violation (r) */
 36#define EXC_W_PROTECTION_FAULT	15 /* TLB permission violation (w) */
 37
 38/*
 39 * This routine handles page faults.  It determines the address,
 40 * and the problem, and then passes it off to one of the appropriate
 41 * routines.
 42 */
 43asmlinkage void do_page_fault(struct pt_regs *regs, unsigned long cause,
 44				unsigned long address)
 45{
 46	struct vm_area_struct *vma = NULL;
 47	struct task_struct *tsk = current;
 48	struct mm_struct *mm = tsk->mm;
 49	int code = SEGV_MAPERR;
 50	vm_fault_t fault;
 51	unsigned int flags = FAULT_FLAG_DEFAULT;
 52
 53	cause >>= 2;
 54
 55	/* Restart the instruction */
 56	regs->ea -= 4;
 57
 58	/*
 59	 * We fault-in kernel-space virtual memory on-demand. The
 60	 * 'reference' page table is init_mm.pgd.
 61	 *
 62	 * NOTE! We MUST NOT take any locks for this case. We may
 63	 * be in an interrupt or a critical region, and should
 64	 * only copy the information from the master page table,
 65	 * nothing more.
 66	 */
 67	if (unlikely(address >= VMALLOC_START && address <= VMALLOC_END)) {
 68		if (user_mode(regs))
 69			goto bad_area_nosemaphore;
 70		else
 71			goto vmalloc_fault;
 72	}
 73
 74	if (unlikely(address >= TASK_SIZE))
 75		goto bad_area_nosemaphore;
 76
 77	/*
 78	 * If we're in an interrupt or have no user
 79	 * context, we must not take the fault..
 80	 */
 81	if (faulthandler_disabled() || !mm)
 82		goto bad_area_nosemaphore;
 83
 84	if (user_mode(regs))
 85		flags |= FAULT_FLAG_USER;
 86
 87	perf_sw_event(PERF_COUNT_SW_PAGE_FAULTS, 1, regs, address);
 88
 89	if (!mmap_read_trylock(mm)) {
 90		if (!user_mode(regs) && !search_exception_tables(regs->ea))
 91			goto bad_area_nosemaphore;
 92retry:
 93		mmap_read_lock(mm);
 94	}
 95
 96	vma = find_vma(mm, address);
 97	if (!vma)
 98		goto bad_area;
 99	if (vma->vm_start <= address)
100		goto good_area;
101	if (!(vma->vm_flags & VM_GROWSDOWN))
102		goto bad_area;
103	if (expand_stack(vma, address))
104		goto bad_area;
105/*
106 * Ok, we have a good vm_area for this memory access, so
107 * we can handle it..
108 */
109good_area:
110	code = SEGV_ACCERR;
111
112	switch (cause) {
113	case EXC_SUPERV_INSN_ACCESS:
114		goto bad_area;
115	case EXC_SUPERV_DATA_ACCESS:
116		goto bad_area;
117	case EXC_X_PROTECTION_FAULT:
118		if (!(vma->vm_flags & VM_EXEC))
119			goto bad_area;
120		break;
121	case EXC_R_PROTECTION_FAULT:
122		if (!(vma->vm_flags & VM_READ))
123			goto bad_area;
124		break;
125	case EXC_W_PROTECTION_FAULT:
126		if (!(vma->vm_flags & VM_WRITE))
127			goto bad_area;
128		flags = FAULT_FLAG_WRITE;
129		break;
130	}
131
132	/*
133	 * If for any reason at all we couldn't handle the fault,
134	 * make sure we exit gracefully rather than endlessly redo
135	 * the fault.
136	 */
137	fault = handle_mm_fault(vma, address, flags, regs);
138
139	if (fault_signal_pending(fault, regs))
140		return;
141
142	if (unlikely(fault & VM_FAULT_ERROR)) {
143		if (fault & VM_FAULT_OOM)
144			goto out_of_memory;
145		else if (fault & VM_FAULT_SIGSEGV)
146			goto bad_area;
147		else if (fault & VM_FAULT_SIGBUS)
148			goto do_sigbus;
149		BUG();
150	}
151
152	if (flags & FAULT_FLAG_ALLOW_RETRY) {
153		if (fault & VM_FAULT_RETRY) {
154			flags |= FAULT_FLAG_TRIED;
155
156			/*
157			 * No need to mmap_read_unlock(mm) as we would
158			 * have already released it in __lock_page_or_retry
159			 * in mm/filemap.c.
160			 */
161
162			goto retry;
163		}
164	}
165
166	mmap_read_unlock(mm);
167	return;
168
169/*
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..
172 */
173bad_area:
174	mmap_read_unlock(mm);
175
176bad_area_nosemaphore:
177	/* User mode accesses just cause a SIGSEGV */
178	if (user_mode(regs)) {
179		if (unhandled_signal(current, SIGSEGV) && printk_ratelimit()) {
180			pr_info("%s: unhandled page fault (%d) at 0x%08lx, "
181				"cause %ld\n", current->comm, SIGSEGV, address, cause);
182			show_regs(regs);
183		}
184		_exception(SIGSEGV, regs, code, address);
185		return;
186	}
187
188no_context:
189	/* Are we prepared to handle this kernel fault? */
190	if (fixup_exception(regs))
191		return;
192
193	/*
194	 * Oops. The kernel tried to access some bad page. We'll have to
195	 * terminate things with extreme prejudice.
196	 */
197	bust_spinlocks(1);
198
199	pr_alert("Unable to handle kernel %s at virtual address %08lx",
200		address < PAGE_SIZE ? "NULL pointer dereference" :
201		"paging request", address);
202	pr_alert("ea = %08lx, ra = %08lx, cause = %ld\n", regs->ea, regs->ra,
203		cause);
204	panic("Oops");
205	return;
206
207/*
208 * We ran out of memory, or some other thing happened to us that made
209 * us unable to handle the page fault gracefully.
210 */
211out_of_memory:
212	mmap_read_unlock(mm);
213	if (!user_mode(regs))
214		goto no_context;
215	pagefault_out_of_memory();
216	return;
217
218do_sigbus:
219	mmap_read_unlock(mm);
220
221	/* Kernel mode? Handle exceptions or die */
222	if (!user_mode(regs))
223		goto no_context;
224
225	_exception(SIGBUS, regs, BUS_ADRERR, address);
226	return;
227
228vmalloc_fault:
229	{
230		/*
231		 * Synchronize this task's top level page-table
232		 * with the 'reference' page table.
233		 *
234		 * Do _not_ use "tsk" here. We might be inside
235		 * an interrupt in the middle of a task switch..
236		 */
237		int offset = pgd_index(address);
238		pgd_t *pgd, *pgd_k;
239		p4d_t *p4d, *p4d_k;
240		pud_t *pud, *pud_k;
241		pmd_t *pmd, *pmd_k;
242		pte_t *pte_k;
243
244		pgd = pgd_current + offset;
245		pgd_k = init_mm.pgd + offset;
246
247		if (!pgd_present(*pgd_k))
248			goto no_context;
249		set_pgd(pgd, *pgd_k);
250
251		p4d = p4d_offset(pgd, address);
252		p4d_k = p4d_offset(pgd_k, address);
253		if (!p4d_present(*p4d_k))
254			goto no_context;
255		pud = pud_offset(p4d, address);
256		pud_k = pud_offset(p4d_k, address);
257		if (!pud_present(*pud_k))
258			goto no_context;
259		pmd = pmd_offset(pud, address);
260		pmd_k = pmd_offset(pud_k, address);
261		if (!pmd_present(*pmd_k))
262			goto no_context;
263		set_pmd(pmd, *pmd_k);
264
265		pte_k = pte_offset_kernel(pmd_k, address);
266		if (!pte_present(*pte_k))
267			goto no_context;
268
269		flush_tlb_kernel_page(address);
270		return;
271	}
272}