Commit f9c497c4aea28be826b8450a5339952f2ffc705e

Authored by Nick Piggin
Committed by Linus Torvalds
1 parent b7e5d1f041

frv: invoke oom-killer from page fault

As explained in commit 1c0fe6e3bd ("mm: invoke oom-killer from page
fault") , we want to call the architecture independent oom killer when
getting an unexplained OOM from handle_mm_fault, rather than simply
killing current.

Signed-off-by: Nick Piggin <npiggin@suse.de>
Acked-by: David Howells <dhowells@redhat.com>
Acked-by: David Rientjes <rientjes@google.com>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

Showing 1 changed file with 4 additions and 4 deletions Inline Diff

1 /* 1 /*
2 * linux/arch/frv/mm/fault.c 2 * linux/arch/frv/mm/fault.c
3 * 3 *
4 * Copyright (C) 2003 Red Hat, Inc. All Rights Reserved. 4 * Copyright (C) 2003 Red Hat, Inc. All Rights Reserved.
5 * - Written by David Howells (dhowells@redhat.com) 5 * - Written by David Howells (dhowells@redhat.com)
6 * - Derived from arch/m68knommu/mm/fault.c 6 * - Derived from arch/m68knommu/mm/fault.c
7 * - Copyright (C) 1998 D. Jeff Dionne <jeff@lineo.ca>, 7 * - Copyright (C) 1998 D. Jeff Dionne <jeff@lineo.ca>,
8 * - Copyright (C) 2000 Lineo, Inc. (www.lineo.com) 8 * - Copyright (C) 2000 Lineo, Inc. (www.lineo.com)
9 * 9 *
10 * Based on: 10 * Based on:
11 * 11 *
12 * linux/arch/m68k/mm/fault.c 12 * linux/arch/m68k/mm/fault.c
13 * 13 *
14 * Copyright (C) 1995 Hamish Macdonald 14 * Copyright (C) 1995 Hamish Macdonald
15 */ 15 */
16 16
17 #include <linux/mman.h> 17 #include <linux/mman.h>
18 #include <linux/mm.h> 18 #include <linux/mm.h>
19 #include <linux/kernel.h> 19 #include <linux/kernel.h>
20 #include <linux/ptrace.h> 20 #include <linux/ptrace.h>
21 #include <linux/hardirq.h> 21 #include <linux/hardirq.h>
22 22
23 #include <asm/system.h> 23 #include <asm/system.h>
24 #include <asm/pgtable.h> 24 #include <asm/pgtable.h>
25 #include <asm/uaccess.h> 25 #include <asm/uaccess.h>
26 #include <asm/gdb-stub.h> 26 #include <asm/gdb-stub.h>
27 27
28 /*****************************************************************************/ 28 /*****************************************************************************/
29 /* 29 /*
30 * This routine handles page faults. It determines the problem, and 30 * This routine handles page faults. It determines the problem, and
31 * then passes it off to one of the appropriate routines. 31 * then passes it off to one of the appropriate routines.
32 */ 32 */
33 asmlinkage void do_page_fault(int datammu, unsigned long esr0, unsigned long ear0) 33 asmlinkage void do_page_fault(int datammu, unsigned long esr0, unsigned long ear0)
34 { 34 {
35 struct vm_area_struct *vma; 35 struct vm_area_struct *vma;
36 struct mm_struct *mm; 36 struct mm_struct *mm;
37 unsigned long _pme, lrai, lrad, fixup; 37 unsigned long _pme, lrai, lrad, fixup;
38 siginfo_t info; 38 siginfo_t info;
39 pgd_t *pge; 39 pgd_t *pge;
40 pud_t *pue; 40 pud_t *pue;
41 pte_t *pte; 41 pte_t *pte;
42 int write; 42 int write;
43 int fault; 43 int fault;
44 44
45 #if 0 45 #if 0
46 const char *atxc[16] = { 46 const char *atxc[16] = {
47 [0x0] = "mmu-miss", [0x8] = "multi-dat", [0x9] = "multi-sat", 47 [0x0] = "mmu-miss", [0x8] = "multi-dat", [0x9] = "multi-sat",
48 [0xa] = "tlb-miss", [0xc] = "privilege", [0xd] = "write-prot", 48 [0xa] = "tlb-miss", [0xc] = "privilege", [0xd] = "write-prot",
49 }; 49 };
50 50
51 printk("do_page_fault(%d,%lx [%s],%lx)\n", 51 printk("do_page_fault(%d,%lx [%s],%lx)\n",
52 datammu, esr0, atxc[esr0 >> 20 & 0xf], ear0); 52 datammu, esr0, atxc[esr0 >> 20 & 0xf], ear0);
53 #endif 53 #endif
54 54
55 mm = current->mm; 55 mm = current->mm;
56 56
57 /* 57 /*
58 * We fault-in kernel-space virtual memory on-demand. The 58 * We fault-in kernel-space virtual memory on-demand. The
59 * 'reference' page table is init_mm.pgd. 59 * 'reference' page table is init_mm.pgd.
60 * 60 *
61 * NOTE! We MUST NOT take any locks for this case. We may 61 * NOTE! We MUST NOT take any locks for this case. We may
62 * be in an interrupt or a critical region, and should 62 * be in an interrupt or a critical region, and should
63 * only copy the information from the master page table, 63 * only copy the information from the master page table,
64 * nothing more. 64 * nothing more.
65 * 65 *
66 * This verifies that the fault happens in kernel space 66 * This verifies that the fault happens in kernel space
67 * and that the fault was a page not present (invalid) error 67 * and that the fault was a page not present (invalid) error
68 */ 68 */
69 if (!user_mode(__frame) && (esr0 & ESR0_ATXC) == ESR0_ATXC_AMRTLB_MISS) { 69 if (!user_mode(__frame) && (esr0 & ESR0_ATXC) == ESR0_ATXC_AMRTLB_MISS) {
70 if (ear0 >= VMALLOC_START && ear0 < VMALLOC_END) 70 if (ear0 >= VMALLOC_START && ear0 < VMALLOC_END)
71 goto kernel_pte_fault; 71 goto kernel_pte_fault;
72 if (ear0 >= PKMAP_BASE && ear0 < PKMAP_END) 72 if (ear0 >= PKMAP_BASE && ear0 < PKMAP_END)
73 goto kernel_pte_fault; 73 goto kernel_pte_fault;
74 } 74 }
75 75
76 info.si_code = SEGV_MAPERR; 76 info.si_code = SEGV_MAPERR;
77 77
78 /* 78 /*
79 * If we're in an interrupt or have no user 79 * If we're in an interrupt or have no user
80 * context, we must not take the fault.. 80 * context, we must not take the fault..
81 */ 81 */
82 if (in_atomic() || !mm) 82 if (in_atomic() || !mm)
83 goto no_context; 83 goto no_context;
84 84
85 down_read(&mm->mmap_sem); 85 down_read(&mm->mmap_sem);
86 86
87 vma = find_vma(mm, ear0); 87 vma = find_vma(mm, ear0);
88 if (!vma) 88 if (!vma)
89 goto bad_area; 89 goto bad_area;
90 if (vma->vm_start <= ear0) 90 if (vma->vm_start <= ear0)
91 goto good_area; 91 goto good_area;
92 if (!(vma->vm_flags & VM_GROWSDOWN)) 92 if (!(vma->vm_flags & VM_GROWSDOWN))
93 goto bad_area; 93 goto bad_area;
94 94
95 if (user_mode(__frame)) { 95 if (user_mode(__frame)) {
96 /* 96 /*
97 * accessing the stack below %esp is always a bug. 97 * accessing the stack below %esp is always a bug.
98 * The "+ 32" is there due to some instructions (like 98 * The "+ 32" is there due to some instructions (like
99 * pusha) doing post-decrement on the stack and that 99 * pusha) doing post-decrement on the stack and that
100 * doesn't show up until later.. 100 * doesn't show up until later..
101 */ 101 */
102 if ((ear0 & PAGE_MASK) + 2 * PAGE_SIZE < __frame->sp) { 102 if ((ear0 & PAGE_MASK) + 2 * PAGE_SIZE < __frame->sp) {
103 #if 0 103 #if 0
104 printk("[%d] ### Access below stack @%lx (sp=%lx)\n", 104 printk("[%d] ### Access below stack @%lx (sp=%lx)\n",
105 current->pid, ear0, __frame->sp); 105 current->pid, ear0, __frame->sp);
106 show_registers(__frame); 106 show_registers(__frame);
107 printk("[%d] ### Code: [%08lx] %02x %02x %02x %02x %02x %02x %02x %02x\n", 107 printk("[%d] ### Code: [%08lx] %02x %02x %02x %02x %02x %02x %02x %02x\n",
108 current->pid, 108 current->pid,
109 __frame->pc, 109 __frame->pc,
110 ((u8*)__frame->pc)[0], 110 ((u8*)__frame->pc)[0],
111 ((u8*)__frame->pc)[1], 111 ((u8*)__frame->pc)[1],
112 ((u8*)__frame->pc)[2], 112 ((u8*)__frame->pc)[2],
113 ((u8*)__frame->pc)[3], 113 ((u8*)__frame->pc)[3],
114 ((u8*)__frame->pc)[4], 114 ((u8*)__frame->pc)[4],
115 ((u8*)__frame->pc)[5], 115 ((u8*)__frame->pc)[5],
116 ((u8*)__frame->pc)[6], 116 ((u8*)__frame->pc)[6],
117 ((u8*)__frame->pc)[7] 117 ((u8*)__frame->pc)[7]
118 ); 118 );
119 #endif 119 #endif
120 goto bad_area; 120 goto bad_area;
121 } 121 }
122 } 122 }
123 123
124 if (expand_stack(vma, ear0)) 124 if (expand_stack(vma, ear0))
125 goto bad_area; 125 goto bad_area;
126 126
127 /* 127 /*
128 * Ok, we have a good vm_area for this memory access, so 128 * Ok, we have a good vm_area for this memory access, so
129 * we can handle it.. 129 * we can handle it..
130 */ 130 */
131 good_area: 131 good_area:
132 info.si_code = SEGV_ACCERR; 132 info.si_code = SEGV_ACCERR;
133 write = 0; 133 write = 0;
134 switch (esr0 & ESR0_ATXC) { 134 switch (esr0 & ESR0_ATXC) {
135 default: 135 default:
136 /* handle write to write protected page */ 136 /* handle write to write protected page */
137 case ESR0_ATXC_WP_EXCEP: 137 case ESR0_ATXC_WP_EXCEP:
138 #ifdef TEST_VERIFY_AREA 138 #ifdef TEST_VERIFY_AREA
139 if (!(user_mode(__frame))) 139 if (!(user_mode(__frame)))
140 printk("WP fault at %08lx\n", __frame->pc); 140 printk("WP fault at %08lx\n", __frame->pc);
141 #endif 141 #endif
142 if (!(vma->vm_flags & VM_WRITE)) 142 if (!(vma->vm_flags & VM_WRITE))
143 goto bad_area; 143 goto bad_area;
144 write = 1; 144 write = 1;
145 break; 145 break;
146 146
147 /* handle read from protected page */ 147 /* handle read from protected page */
148 case ESR0_ATXC_PRIV_EXCEP: 148 case ESR0_ATXC_PRIV_EXCEP:
149 goto bad_area; 149 goto bad_area;
150 150
151 /* handle read, write or exec on absent page 151 /* handle read, write or exec on absent page
152 * - can't support write without permitting read 152 * - can't support write without permitting read
153 * - don't support execute without permitting read and vice-versa 153 * - don't support execute without permitting read and vice-versa
154 */ 154 */
155 case ESR0_ATXC_AMRTLB_MISS: 155 case ESR0_ATXC_AMRTLB_MISS:
156 if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC))) 156 if (!(vma->vm_flags & (VM_READ | VM_WRITE | VM_EXEC)))
157 goto bad_area; 157 goto bad_area;
158 break; 158 break;
159 } 159 }
160 160
161 /* 161 /*
162 * If for any reason at all we couldn't handle the fault, 162 * If for any reason at all we couldn't handle the fault,
163 * make sure we exit gracefully rather than endlessly redo 163 * make sure we exit gracefully rather than endlessly redo
164 * the fault. 164 * the fault.
165 */ 165 */
166 fault = handle_mm_fault(mm, vma, ear0, write ? FAULT_FLAG_WRITE : 0); 166 fault = handle_mm_fault(mm, vma, ear0, write ? FAULT_FLAG_WRITE : 0);
167 if (unlikely(fault & VM_FAULT_ERROR)) { 167 if (unlikely(fault & VM_FAULT_ERROR)) {
168 if (fault & VM_FAULT_OOM) 168 if (fault & VM_FAULT_OOM)
169 goto out_of_memory; 169 goto out_of_memory;
170 else if (fault & VM_FAULT_SIGBUS) 170 else if (fault & VM_FAULT_SIGBUS)
171 goto do_sigbus; 171 goto do_sigbus;
172 BUG(); 172 BUG();
173 } 173 }
174 if (fault & VM_FAULT_MAJOR) 174 if (fault & VM_FAULT_MAJOR)
175 current->maj_flt++; 175 current->maj_flt++;
176 else 176 else
177 current->min_flt++; 177 current->min_flt++;
178 178
179 up_read(&mm->mmap_sem); 179 up_read(&mm->mmap_sem);
180 return; 180 return;
181 181
182 /* 182 /*
183 * Something tried to access memory that isn't in our memory map.. 183 * Something tried to access memory that isn't in our memory map..
184 * Fix it, but check if it's kernel or user first.. 184 * Fix it, but check if it's kernel or user first..
185 */ 185 */
186 bad_area: 186 bad_area:
187 up_read(&mm->mmap_sem); 187 up_read(&mm->mmap_sem);
188 188
189 /* User mode accesses just cause a SIGSEGV */ 189 /* User mode accesses just cause a SIGSEGV */
190 if (user_mode(__frame)) { 190 if (user_mode(__frame)) {
191 info.si_signo = SIGSEGV; 191 info.si_signo = SIGSEGV;
192 info.si_errno = 0; 192 info.si_errno = 0;
193 /* info.si_code has been set above */ 193 /* info.si_code has been set above */
194 info.si_addr = (void *) ear0; 194 info.si_addr = (void *) ear0;
195 force_sig_info(SIGSEGV, &info, current); 195 force_sig_info(SIGSEGV, &info, current);
196 return; 196 return;
197 } 197 }
198 198
199 no_context: 199 no_context:
200 /* are we prepared to handle this kernel fault? */ 200 /* are we prepared to handle this kernel fault? */
201 if ((fixup = search_exception_table(__frame->pc)) != 0) { 201 if ((fixup = search_exception_table(__frame->pc)) != 0) {
202 __frame->pc = fixup; 202 __frame->pc = fixup;
203 return; 203 return;
204 } 204 }
205 205
206 /* 206 /*
207 * Oops. The kernel tried to access some bad page. We'll have to 207 * Oops. The kernel tried to access some bad page. We'll have to
208 * terminate things with extreme prejudice. 208 * terminate things with extreme prejudice.
209 */ 209 */
210 210
211 bust_spinlocks(1); 211 bust_spinlocks(1);
212 212
213 if (ear0 < PAGE_SIZE) 213 if (ear0 < PAGE_SIZE)
214 printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference"); 214 printk(KERN_ALERT "Unable to handle kernel NULL pointer dereference");
215 else 215 else
216 printk(KERN_ALERT "Unable to handle kernel paging request"); 216 printk(KERN_ALERT "Unable to handle kernel paging request");
217 printk(" at virtual addr %08lx\n", ear0); 217 printk(" at virtual addr %08lx\n", ear0);
218 printk(" PC : %08lx\n", __frame->pc); 218 printk(" PC : %08lx\n", __frame->pc);
219 printk(" EXC : esr0=%08lx ear0=%08lx\n", esr0, ear0); 219 printk(" EXC : esr0=%08lx ear0=%08lx\n", esr0, ear0);
220 220
221 asm("lrai %1,%0,#1,#0,#0" : "=&r"(lrai) : "r"(ear0)); 221 asm("lrai %1,%0,#1,#0,#0" : "=&r"(lrai) : "r"(ear0));
222 asm("lrad %1,%0,#1,#0,#0" : "=&r"(lrad) : "r"(ear0)); 222 asm("lrad %1,%0,#1,#0,#0" : "=&r"(lrad) : "r"(ear0));
223 223
224 printk(KERN_ALERT " LRAI: %08lx\n", lrai); 224 printk(KERN_ALERT " LRAI: %08lx\n", lrai);
225 printk(KERN_ALERT " LRAD: %08lx\n", lrad); 225 printk(KERN_ALERT " LRAD: %08lx\n", lrad);
226 226
227 __break_hijack_kernel_event(); 227 __break_hijack_kernel_event();
228 228
229 pge = pgd_offset(current->mm, ear0); 229 pge = pgd_offset(current->mm, ear0);
230 pue = pud_offset(pge, ear0); 230 pue = pud_offset(pge, ear0);
231 _pme = pue->pue[0].ste[0]; 231 _pme = pue->pue[0].ste[0];
232 232
233 printk(KERN_ALERT " PGE : %8p { PME %08lx }\n", pge, _pme); 233 printk(KERN_ALERT " PGE : %8p { PME %08lx }\n", pge, _pme);
234 234
235 if (_pme & xAMPRx_V) { 235 if (_pme & xAMPRx_V) {
236 unsigned long dampr, damlr, val; 236 unsigned long dampr, damlr, val;
237 237
238 asm volatile("movsg dampr2,%0 ! movgs %2,dampr2 ! movsg damlr2,%1" 238 asm volatile("movsg dampr2,%0 ! movgs %2,dampr2 ! movsg damlr2,%1"
239 : "=&r"(dampr), "=r"(damlr) 239 : "=&r"(dampr), "=r"(damlr)
240 : "r" (_pme | xAMPRx_L|xAMPRx_SS_16Kb|xAMPRx_S|xAMPRx_C|xAMPRx_V) 240 : "r" (_pme | xAMPRx_L|xAMPRx_SS_16Kb|xAMPRx_S|xAMPRx_C|xAMPRx_V)
241 ); 241 );
242 242
243 pte = (pte_t *) damlr + __pte_index(ear0); 243 pte = (pte_t *) damlr + __pte_index(ear0);
244 val = pte_val(*pte); 244 val = pte_val(*pte);
245 245
246 asm volatile("movgs %0,dampr2" :: "r" (dampr)); 246 asm volatile("movgs %0,dampr2" :: "r" (dampr));
247 247
248 printk(KERN_ALERT " PTE : %8p { %08lx }\n", pte, val); 248 printk(KERN_ALERT " PTE : %8p { %08lx }\n", pte, val);
249 } 249 }
250 250
251 die_if_kernel("Oops\n"); 251 die_if_kernel("Oops\n");
252 do_exit(SIGKILL); 252 do_exit(SIGKILL);
253 253
254 /* 254 /*
255 * We ran out of memory, or some other thing happened to us that made 255 * We ran out of memory, or some other thing happened to us that made
256 * us unable to handle the page fault gracefully. 256 * us unable to handle the page fault gracefully.
257 */ 257 */
258 out_of_memory: 258 out_of_memory:
259 up_read(&mm->mmap_sem); 259 up_read(&mm->mmap_sem);
260 printk("VM: killing process %s\n", current->comm); 260 if (!user_mode(__frame))
261 if (user_mode(__frame)) 261 goto no_context;
262 do_group_exit(SIGKILL); 262 pagefault_out_of_memory();
263 goto no_context; 263 return;
264 264
265 do_sigbus: 265 do_sigbus:
266 up_read(&mm->mmap_sem); 266 up_read(&mm->mmap_sem);
267 267
268 /* 268 /*
269 * Send a sigbus, regardless of whether we were in kernel 269 * Send a sigbus, regardless of whether we were in kernel
270 * or user mode. 270 * or user mode.
271 */ 271 */
272 info.si_signo = SIGBUS; 272 info.si_signo = SIGBUS;
273 info.si_errno = 0; 273 info.si_errno = 0;
274 info.si_code = BUS_ADRERR; 274 info.si_code = BUS_ADRERR;
275 info.si_addr = (void *) ear0; 275 info.si_addr = (void *) ear0;
276 force_sig_info(SIGBUS, &info, current); 276 force_sig_info(SIGBUS, &info, current);
277 277
278 /* Kernel mode? Handle exceptions or die */ 278 /* Kernel mode? Handle exceptions or die */
279 if (!user_mode(__frame)) 279 if (!user_mode(__frame))
280 goto no_context; 280 goto no_context;
281 return; 281 return;
282 282
283 /* 283 /*
284 * The fault was caused by a kernel PTE (such as installed by vmalloc or kmap) 284 * The fault was caused by a kernel PTE (such as installed by vmalloc or kmap)
285 */ 285 */
286 kernel_pte_fault: 286 kernel_pte_fault:
287 { 287 {
288 /* 288 /*
289 * Synchronize this task's top level page-table 289 * Synchronize this task's top level page-table
290 * with the 'reference' page table. 290 * with the 'reference' page table.
291 * 291 *
292 * Do _not_ use "tsk" here. We might be inside 292 * Do _not_ use "tsk" here. We might be inside
293 * an interrupt in the middle of a task switch.. 293 * an interrupt in the middle of a task switch..
294 */ 294 */
295 int index = pgd_index(ear0); 295 int index = pgd_index(ear0);
296 pgd_t *pgd, *pgd_k; 296 pgd_t *pgd, *pgd_k;
297 pud_t *pud, *pud_k; 297 pud_t *pud, *pud_k;
298 pmd_t *pmd, *pmd_k; 298 pmd_t *pmd, *pmd_k;
299 pte_t *pte_k; 299 pte_t *pte_k;
300 300
301 pgd = (pgd_t *) __get_TTBR(); 301 pgd = (pgd_t *) __get_TTBR();
302 pgd = (pgd_t *)__va(pgd) + index; 302 pgd = (pgd_t *)__va(pgd) + index;
303 pgd_k = ((pgd_t *)(init_mm.pgd)) + index; 303 pgd_k = ((pgd_t *)(init_mm.pgd)) + index;
304 304
305 if (!pgd_present(*pgd_k)) 305 if (!pgd_present(*pgd_k))
306 goto no_context; 306 goto no_context;
307 //set_pgd(pgd, *pgd_k); /////// gcc ICE's on this line 307 //set_pgd(pgd, *pgd_k); /////// gcc ICE's on this line
308 308
309 pud_k = pud_offset(pgd_k, ear0); 309 pud_k = pud_offset(pgd_k, ear0);
310 if (!pud_present(*pud_k)) 310 if (!pud_present(*pud_k))
311 goto no_context; 311 goto no_context;
312 312
313 pmd_k = pmd_offset(pud_k, ear0); 313 pmd_k = pmd_offset(pud_k, ear0);
314 if (!pmd_present(*pmd_k)) 314 if (!pmd_present(*pmd_k))
315 goto no_context; 315 goto no_context;
316 316
317 pud = pud_offset(pgd, ear0); 317 pud = pud_offset(pgd, ear0);
318 pmd = pmd_offset(pud, ear0); 318 pmd = pmd_offset(pud, ear0);
319 set_pmd(pmd, *pmd_k); 319 set_pmd(pmd, *pmd_k);
320 320
321 pte_k = pte_offset_kernel(pmd_k, ear0); 321 pte_k = pte_offset_kernel(pmd_k, ear0);
322 if (!pte_present(*pte_k)) 322 if (!pte_present(*pte_k))
323 goto no_context; 323 goto no_context;
324 return; 324 return;
325 } 325 }
326 } /* end do_page_fault() */ 326 } /* end do_page_fault() */
327 327