Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
1da177e4 | 2 | * Copyright (C) 1991, 1992 Linus Torvalds |
a8c1be9d | 3 | * Copyright (C) 2000, 2001, 2002 Andi Kleen, SuSE Labs |
1da177e4 LT |
4 | * |
5 | * Pentium III FXSR, SSE support | |
6 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
7 | */ | |
8 | ||
9 | /* | |
10 | * 'Traps.c' handles hardware traps and faults after we have saved some | |
11 | * state in 'asm.s'. | |
12 | */ | |
b5964405 IM |
13 | #include <linux/interrupt.h> |
14 | #include <linux/kallsyms.h> | |
15 | #include <linux/spinlock.h> | |
16 | #include <linux/highmem.h> | |
17 | #include <linux/kprobes.h> | |
18 | #include <linux/uaccess.h> | |
19 | #include <linux/utsname.h> | |
20 | #include <linux/kdebug.h> | |
1da177e4 | 21 | #include <linux/kernel.h> |
b5964405 IM |
22 | #include <linux/module.h> |
23 | #include <linux/ptrace.h> | |
1da177e4 | 24 | #include <linux/string.h> |
b5964405 IM |
25 | #include <linux/unwind.h> |
26 | #include <linux/delay.h> | |
1da177e4 | 27 | #include <linux/errno.h> |
b5964405 IM |
28 | #include <linux/kexec.h> |
29 | #include <linux/sched.h> | |
1da177e4 | 30 | #include <linux/timer.h> |
1da177e4 | 31 | #include <linux/init.h> |
91768d6c | 32 | #include <linux/bug.h> |
b5964405 IM |
33 | #include <linux/nmi.h> |
34 | #include <linux/mm.h> | |
1da177e4 LT |
35 | |
36 | #ifdef CONFIG_EISA | |
37 | #include <linux/ioport.h> | |
38 | #include <linux/eisa.h> | |
39 | #endif | |
40 | ||
41 | #ifdef CONFIG_MCA | |
42 | #include <linux/mca.h> | |
43 | #endif | |
44 | ||
c0d12172 DJ |
45 | #if defined(CONFIG_EDAC) |
46 | #include <linux/edac.h> | |
47 | #endif | |
48 | ||
b5964405 IM |
49 | #include <asm/arch_hooks.h> |
50 | #include <asm/stacktrace.h> | |
1da177e4 | 51 | #include <asm/processor.h> |
1da177e4 | 52 | #include <asm/debugreg.h> |
b5964405 IM |
53 | #include <asm/atomic.h> |
54 | #include <asm/system.h> | |
55 | #include <asm/unwind.h> | |
1da177e4 LT |
56 | #include <asm/desc.h> |
57 | #include <asm/i387.h> | |
58 | #include <asm/nmi.h> | |
1da177e4 | 59 | #include <asm/smp.h> |
b5964405 | 60 | #include <asm/io.h> |
1da177e4 LT |
61 | |
62 | #include "mach_traps.h" | |
63 | ||
dbeb2be2 RR |
64 | DECLARE_BITMAP(used_vectors, NR_VECTORS); |
65 | EXPORT_SYMBOL_GPL(used_vectors); | |
66 | ||
1da177e4 LT |
67 | asmlinkage int system_call(void); |
68 | ||
1da177e4 | 69 | /* Do we ignore FPU interrupts ? */ |
b5964405 | 70 | char ignore_fpu_irq; |
1da177e4 LT |
71 | |
72 | /* | |
73 | * The IDT has to be page-aligned to simplify the Pentium | |
74 | * F0 0F bug workaround.. We have a special link segment | |
75 | * for this. | |
76 | */ | |
010d4f82 | 77 | gate_desc idt_table[256] |
6842ef0e | 78 | __attribute__((__section__(".data.idt"))) = { { { { 0, 0 } } }, }; |
1da177e4 LT |
79 | |
80 | asmlinkage void divide_error(void); | |
81 | asmlinkage void debug(void); | |
82 | asmlinkage void nmi(void); | |
83 | asmlinkage void int3(void); | |
84 | asmlinkage void overflow(void); | |
85 | asmlinkage void bounds(void); | |
86 | asmlinkage void invalid_op(void); | |
87 | asmlinkage void device_not_available(void); | |
88 | asmlinkage void coprocessor_segment_overrun(void); | |
89 | asmlinkage void invalid_TSS(void); | |
90 | asmlinkage void segment_not_present(void); | |
91 | asmlinkage void stack_segment(void); | |
92 | asmlinkage void general_protection(void); | |
93 | asmlinkage void page_fault(void); | |
94 | asmlinkage void coprocessor_error(void); | |
95 | asmlinkage void simd_coprocessor_error(void); | |
96 | asmlinkage void alignment_check(void); | |
97 | asmlinkage void spurious_interrupt_bug(void); | |
98 | asmlinkage void machine_check(void); | |
99 | ||
badc7652 | 100 | int panic_on_unrecovered_nmi; |
0741f4d2 | 101 | int kstack_depth_to_print = 24; |
86c41837 | 102 | static unsigned int code_bytes = 64; |
badc7652 AH |
103 | static int ignore_nmis; |
104 | static int die_counter; | |
e041c683 | 105 | |
a5ff677c HH |
106 | void printk_address(unsigned long address, int reliable) |
107 | { | |
108 | #ifdef CONFIG_KALLSYMS | |
b5964405 IM |
109 | char namebuf[KSYM_NAME_LEN]; |
110 | unsigned long offset = 0; | |
111 | unsigned long symsize; | |
a5ff677c | 112 | const char *symname; |
a5ff677c | 113 | char reliab[4] = ""; |
b5964405 IM |
114 | char *delim = ":"; |
115 | char *modname; | |
a5ff677c HH |
116 | |
117 | symname = kallsyms_lookup(address, &symsize, &offset, | |
118 | &modname, namebuf); | |
119 | if (!symname) { | |
120 | printk(" [<%08lx>]\n", address); | |
121 | return; | |
122 | } | |
123 | if (!reliable) | |
124 | strcpy(reliab, "? "); | |
125 | ||
126 | if (!modname) | |
127 | modname = delim = ""; | |
128 | printk(" [<%08lx>] %s%s%s%s%s+0x%lx/0x%lx\n", | |
129 | address, reliab, delim, modname, delim, symname, offset, symsize); | |
130 | #else | |
131 | printk(" [<%08lx>]\n", address); | |
132 | #endif | |
133 | } | |
134 | ||
a8c1be9d AH |
135 | static inline int valid_stack_ptr(struct thread_info *tinfo, |
136 | void *p, unsigned int size) | |
1da177e4 LT |
137 | { |
138 | return p > (void *)tinfo && | |
36ad4885 | 139 | p <= (void *)tinfo + THREAD_SIZE - size; |
1da177e4 LT |
140 | } |
141 | ||
36ad4885 LT |
142 | /* The form of the top of the frame on the stack */ |
143 | struct stack_frame { | |
a8c1be9d AH |
144 | struct stack_frame *next_frame; |
145 | unsigned long return_address; | |
36ad4885 LT |
146 | }; |
147 | ||
b5964405 IM |
148 | static inline unsigned long |
149 | print_context_stack(struct thread_info *tinfo, | |
a8c1be9d AH |
150 | unsigned long *stack, unsigned long bp, |
151 | const struct stacktrace_ops *ops, void *data) | |
1da177e4 | 152 | { |
65ea5b03 | 153 | struct stack_frame *frame = (struct stack_frame *)bp; |
36ad4885 | 154 | |
36ad4885 LT |
155 | while (valid_stack_ptr(tinfo, stack, sizeof(*stack))) { |
156 | unsigned long addr; | |
157 | ||
e9d4efdd AV |
158 | addr = *stack; |
159 | if (__kernel_text_address(addr)) { | |
160 | if ((unsigned long) stack == bp + 4) { | |
161 | ops->address(data, addr, 1); | |
162 | frame = frame->next_frame; | |
163 | bp = (unsigned long) frame; | |
164 | } else { | |
5bc27dc2 | 165 | ops->address(data, addr, bp == 0); |
e9d4efdd AV |
166 | } |
167 | } | |
168 | stack++; | |
1da177e4 | 169 | } |
65ea5b03 | 170 | return bp; |
1da177e4 LT |
171 | } |
172 | ||
2b14a78c | 173 | void dump_trace(struct task_struct *task, struct pt_regs *regs, |
5bc27dc2 | 174 | unsigned long *stack, unsigned long bp, |
9689ba8a | 175 | const struct stacktrace_ops *ops, void *data) |
1da177e4 | 176 | { |
1da177e4 LT |
177 | if (!task) |
178 | task = current; | |
179 | ||
a32cf397 | 180 | if (!stack) { |
2b14a78c AK |
181 | unsigned long dummy; |
182 | stack = &dummy; | |
028a690a | 183 | if (task != current) |
faca6227 | 184 | stack = (unsigned long *)task->thread.sp; |
176a2718 JB |
185 | } |
186 | ||
a32cf397 | 187 | #ifdef CONFIG_FRAME_POINTER |
65ea5b03 | 188 | if (!bp) { |
a32cf397 | 189 | if (task == current) { |
65ea5b03 | 190 | /* Grab bp right from our regs */ |
b5964405 | 191 | asm("movl %%ebp, %0" : "=r" (bp) :); |
a32cf397 | 192 | } else { |
65ea5b03 | 193 | /* bp is the last reg pushed by switch_to */ |
faca6227 | 194 | bp = *(unsigned long *) task->thread.sp; |
a32cf397 | 195 | } |
1da177e4 | 196 | } |
a32cf397 | 197 | #endif |
1da177e4 | 198 | |
a8c1be9d | 199 | for (;;) { |
1da177e4 | 200 | struct thread_info *context; |
b5964405 | 201 | |
1da177e4 LT |
202 | context = (struct thread_info *) |
203 | ((unsigned long)stack & (~(THREAD_SIZE - 1))); | |
65ea5b03 | 204 | bp = print_context_stack(context, stack, bp, ops, data); |
b5964405 IM |
205 | /* |
206 | * Should be after the line below, but somewhere | |
207 | * in early boot context comes out corrupted and we | |
208 | * can't reference it: | |
209 | */ | |
2b14a78c AK |
210 | if (ops->stack(data, "IRQ") < 0) |
211 | break; | |
b5964405 | 212 | stack = (unsigned long *)context->previous_esp; |
1da177e4 LT |
213 | if (!stack) |
214 | break; | |
a36df98a | 215 | touch_nmi_watchdog(); |
1da177e4 LT |
216 | } |
217 | } | |
2b14a78c AK |
218 | EXPORT_SYMBOL(dump_trace); |
219 | ||
220 | static void | |
221 | print_trace_warning_symbol(void *data, char *msg, unsigned long symbol) | |
222 | { | |
223 | printk(data); | |
224 | print_symbol(msg, symbol); | |
225 | printk("\n"); | |
226 | } | |
227 | ||
228 | static void print_trace_warning(void *data, char *msg) | |
229 | { | |
230 | printk("%s%s\n", (char *)data, msg); | |
231 | } | |
232 | ||
233 | static int print_trace_stack(void *data, char *name) | |
234 | { | |
235 | return 0; | |
236 | } | |
237 | ||
238 | /* | |
239 | * Print one address/symbol entries per line. | |
240 | */ | |
bc850d6b | 241 | static void print_trace_address(void *data, unsigned long addr, int reliable) |
2b14a78c AK |
242 | { |
243 | printk("%s [<%08lx>] ", (char *)data, addr); | |
bc850d6b AV |
244 | if (!reliable) |
245 | printk("? "); | |
2b14a78c | 246 | print_symbol("%s\n", addr); |
601e6255 | 247 | touch_nmi_watchdog(); |
2b14a78c AK |
248 | } |
249 | ||
9689ba8a | 250 | static const struct stacktrace_ops print_trace_ops = { |
a8c1be9d AH |
251 | .warning = print_trace_warning, |
252 | .warning_symbol = print_trace_warning_symbol, | |
253 | .stack = print_trace_stack, | |
254 | .address = print_trace_address, | |
2b14a78c AK |
255 | }; |
256 | ||
257 | static void | |
258 | show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, | |
b5964405 | 259 | unsigned long *stack, unsigned long bp, char *log_lvl) |
2b14a78c | 260 | { |
5bc27dc2 | 261 | dump_trace(task, regs, stack, bp, &print_trace_ops, log_lvl); |
2b14a78c AK |
262 | printk("%s =======================\n", log_lvl); |
263 | } | |
1da177e4 | 264 | |
2b14a78c | 265 | void show_trace(struct task_struct *task, struct pt_regs *regs, |
5bc27dc2 | 266 | unsigned long *stack, unsigned long bp) |
7aa89746 | 267 | { |
5bc27dc2 | 268 | show_trace_log_lvl(task, regs, stack, bp, ""); |
7aa89746 CE |
269 | } |
270 | ||
b5964405 IM |
271 | static void |
272 | show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, | |
273 | unsigned long *sp, unsigned long bp, char *log_lvl) | |
1da177e4 LT |
274 | { |
275 | unsigned long *stack; | |
276 | int i; | |
277 | ||
65ea5b03 | 278 | if (sp == NULL) { |
1da177e4 | 279 | if (task) |
b5964405 | 280 | sp = (unsigned long *)task->thread.sp; |
1da177e4 | 281 | else |
65ea5b03 | 282 | sp = (unsigned long *)&sp; |
1da177e4 LT |
283 | } |
284 | ||
65ea5b03 | 285 | stack = sp; |
b5964405 | 286 | for (i = 0; i < kstack_depth_to_print; i++) { |
1da177e4 LT |
287 | if (kstack_end(stack)) |
288 | break; | |
75874d5c CE |
289 | if (i && ((i % 8) == 0)) |
290 | printk("\n%s ", log_lvl); | |
1da177e4 LT |
291 | printk("%08lx ", *stack++); |
292 | } | |
75874d5c | 293 | printk("\n%sCall Trace:\n", log_lvl); |
b5964405 | 294 | |
5bc27dc2 | 295 | show_trace_log_lvl(task, regs, sp, bp, log_lvl); |
7aa89746 CE |
296 | } |
297 | ||
65ea5b03 | 298 | void show_stack(struct task_struct *task, unsigned long *sp) |
7aa89746 | 299 | { |
75874d5c | 300 | printk(" "); |
5bc27dc2 | 301 | show_stack_log_lvl(task, NULL, sp, 0, ""); |
1da177e4 LT |
302 | } |
303 | ||
304 | /* | |
305 | * The architecture-independent dump_stack generator | |
306 | */ | |
307 | void dump_stack(void) | |
308 | { | |
5bc27dc2 | 309 | unsigned long bp = 0; |
b5964405 | 310 | unsigned long stack; |
5bc27dc2 AV |
311 | |
312 | #ifdef CONFIG_FRAME_POINTER | |
313 | if (!bp) | |
314 | asm("movl %%ebp, %0" : "=r" (bp):); | |
315 | #endif | |
1da177e4 | 316 | |
57c351de AV |
317 | printk("Pid: %d, comm: %.20s %s %s %.*s\n", |
318 | current->pid, current->comm, print_tainted(), | |
319 | init_utsname()->release, | |
320 | (int)strcspn(init_utsname()->version, " "), | |
321 | init_utsname()->version); | |
b5964405 | 322 | |
5bc27dc2 | 323 | show_trace(current, NULL, &stack, bp); |
1da177e4 LT |
324 | } |
325 | ||
326 | EXPORT_SYMBOL(dump_stack); | |
327 | ||
328 | void show_registers(struct pt_regs *regs) | |
329 | { | |
330 | int i; | |
9d975ebd | 331 | |
1da177e4 | 332 | print_modules(); |
9d975ebd | 333 | __show_registers(regs, 0); |
b5964405 | 334 | |
7e04a118 | 335 | printk(KERN_EMERG "Process %.*s (pid: %d, ti=%p task=%p task.ti=%p)", |
19c5870c | 336 | TASK_COMM_LEN, current->comm, task_pid_nr(current), |
c9f4f06d | 337 | current_thread_info(), current, task_thread_info(current)); |
1da177e4 LT |
338 | /* |
339 | * When in-kernel, we also print out the stack and code at the | |
340 | * time of the fault.. | |
341 | */ | |
9d975ebd | 342 | if (!user_mode_vm(regs)) { |
86c41837 CE |
343 | unsigned int code_prologue = code_bytes * 43 / 64; |
344 | unsigned int code_len = code_bytes; | |
99325326 | 345 | unsigned char c; |
b5964405 | 346 | u8 *ip; |
1da177e4 | 347 | |
9c107805 | 348 | printk("\n" KERN_EMERG "Stack: "); |
5bc27dc2 | 349 | show_stack_log_lvl(NULL, regs, ®s->sp, 0, KERN_EMERG); |
1da177e4 | 350 | |
9c107805 | 351 | printk(KERN_EMERG "Code: "); |
1da177e4 | 352 | |
65ea5b03 | 353 | ip = (u8 *)regs->ip - code_prologue; |
a8c1be9d | 354 | if (ip < (u8 *)PAGE_OFFSET || probe_kernel_address(ip, c)) { |
99325326 | 355 | /* try starting at EIP */ |
65ea5b03 | 356 | ip = (u8 *)regs->ip; |
86c41837 | 357 | code_len = code_len - code_prologue + 1; |
99325326 | 358 | } |
65ea5b03 PA |
359 | for (i = 0; i < code_len; i++, ip++) { |
360 | if (ip < (u8 *)PAGE_OFFSET || | |
a8c1be9d | 361 | probe_kernel_address(ip, c)) { |
1da177e4 LT |
362 | printk(" Bad EIP value."); |
363 | break; | |
364 | } | |
65ea5b03 | 365 | if (ip == (u8 *)regs->ip) |
1da177e4 LT |
366 | printk("<%02x> ", c); |
367 | else | |
368 | printk("%02x ", c); | |
369 | } | |
370 | } | |
371 | printk("\n"); | |
b5964405 | 372 | } |
1da177e4 | 373 | |
65ea5b03 | 374 | int is_valid_bugaddr(unsigned long ip) |
1da177e4 LT |
375 | { |
376 | unsigned short ud2; | |
1da177e4 | 377 | |
65ea5b03 | 378 | if (ip < PAGE_OFFSET) |
91768d6c | 379 | return 0; |
65ea5b03 | 380 | if (probe_kernel_address((unsigned short *)ip, ud2)) |
91768d6c | 381 | return 0; |
1da177e4 | 382 | |
91768d6c | 383 | return ud2 == 0x0b0f; |
1da177e4 LT |
384 | } |
385 | ||
b5964405 | 386 | int __kprobes __die(const char *str, struct pt_regs *regs, long err) |
a604b380 | 387 | { |
a604b380 | 388 | unsigned short ss; |
b5964405 | 389 | unsigned long sp; |
a604b380 HH |
390 | |
391 | printk(KERN_EMERG "%s: %04lx [#%d] ", str, err & 0xffff, ++die_counter); | |
392 | #ifdef CONFIG_PREEMPT | |
393 | printk("PREEMPT "); | |
394 | #endif | |
395 | #ifdef CONFIG_SMP | |
396 | printk("SMP "); | |
397 | #endif | |
398 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
399 | printk("DEBUG_PAGEALLOC"); | |
400 | #endif | |
401 | printk("\n"); | |
a604b380 | 402 | if (notify_die(DIE_OOPS, str, regs, err, |
e423f49f AH |
403 | current->thread.trap_no, SIGSEGV) == NOTIFY_STOP) |
404 | return 1; | |
405 | ||
406 | show_registers(regs); | |
407 | /* Executive summary in case the oops scrolled away */ | |
408 | sp = (unsigned long) (®s->sp); | |
409 | savesegment(ss, ss); | |
410 | if (user_mode(regs)) { | |
411 | sp = regs->sp; | |
412 | ss = regs->ss & 0xffff; | |
a604b380 | 413 | } |
e423f49f AH |
414 | printk(KERN_EMERG "EIP: [<%08lx>] ", regs->ip); |
415 | print_symbol("%s", regs->ip); | |
416 | printk(" SS:ESP %04x:%08lx\n", ss, sp); | |
417 | return 0; | |
a604b380 HH |
418 | } |
419 | ||
91768d6c | 420 | /* |
b5964405 IM |
421 | * This is gone through when something in the kernel has done something bad |
422 | * and is about to be terminated: | |
91768d6c | 423 | */ |
b5964405 | 424 | void die(const char *str, struct pt_regs *regs, long err) |
1da177e4 LT |
425 | { |
426 | static struct { | |
39743c9e | 427 | raw_spinlock_t lock; |
1da177e4 LT |
428 | u32 lock_owner; |
429 | int lock_owner_depth; | |
430 | } die = { | |
39743c9e | 431 | .lock = __RAW_SPIN_LOCK_UNLOCKED, |
1da177e4 LT |
432 | .lock_owner = -1, |
433 | .lock_owner_depth = 0 | |
434 | }; | |
e43d674f | 435 | unsigned long flags; |
1da177e4 | 436 | |
dd287796 AM |
437 | oops_enter(); |
438 | ||
39c715b7 | 439 | if (die.lock_owner != raw_smp_processor_id()) { |
1da177e4 | 440 | console_verbose(); |
c0a698b7 | 441 | raw_local_irq_save(flags); |
39743c9e | 442 | __raw_spin_lock(&die.lock); |
1da177e4 LT |
443 | die.lock_owner = smp_processor_id(); |
444 | die.lock_owner_depth = 0; | |
445 | bust_spinlocks(1); | |
b5964405 | 446 | } else { |
c0a698b7 | 447 | raw_local_irq_save(flags); |
b5964405 | 448 | } |
1da177e4 LT |
449 | |
450 | if (++die.lock_owner_depth < 3) { | |
65ea5b03 | 451 | report_bug(regs->ip, regs); |
91768d6c | 452 | |
a604b380 | 453 | if (__die(str, regs, err)) |
20c0d2d4 | 454 | regs = NULL; |
a604b380 | 455 | } else { |
9c107805 | 456 | printk(KERN_EMERG "Recursive die() failure, output suppressed\n"); |
a604b380 | 457 | } |
1da177e4 LT |
458 | |
459 | bust_spinlocks(0); | |
460 | die.lock_owner = -1; | |
bcdcd8e7 | 461 | add_taint(TAINT_DIE); |
39743c9e AK |
462 | __raw_spin_unlock(&die.lock); |
463 | raw_local_irq_restore(flags); | |
6e274d14 | 464 | |
20c0d2d4 JB |
465 | if (!regs) |
466 | return; | |
467 | ||
6e274d14 AN |
468 | if (kexec_should_crash(current)) |
469 | crash_kexec(regs); | |
470 | ||
1da177e4 LT |
471 | if (in_interrupt()) |
472 | panic("Fatal exception in interrupt"); | |
473 | ||
cea6a4ba | 474 | if (panic_on_oops) |
012c437d | 475 | panic("Fatal exception"); |
cea6a4ba | 476 | |
dd287796 | 477 | oops_exit(); |
1da177e4 LT |
478 | do_exit(SIGSEGV); |
479 | } | |
480 | ||
b5964405 IM |
481 | static inline void |
482 | die_if_kernel(const char *str, struct pt_regs *regs, long err) | |
1da177e4 | 483 | { |
717b594a | 484 | if (!user_mode_vm(regs)) |
1da177e4 LT |
485 | die(str, regs, err); |
486 | } | |
487 | ||
b5964405 IM |
488 | static void __kprobes |
489 | do_trap(int trapnr, int signr, char *str, int vm86, struct pt_regs *regs, | |
490 | long error_code, siginfo_t *info) | |
1da177e4 | 491 | { |
4f339ecb | 492 | struct task_struct *tsk = current; |
4f339ecb | 493 | |
6b6891f9 | 494 | if (regs->flags & X86_VM_MASK) { |
1da177e4 LT |
495 | if (vm86) |
496 | goto vm86_trap; | |
497 | goto trap_signal; | |
498 | } | |
499 | ||
717b594a | 500 | if (!user_mode(regs)) |
1da177e4 LT |
501 | goto kernel_trap; |
502 | ||
b5964405 IM |
503 | trap_signal: |
504 | /* | |
505 | * We want error_code and trap_no set for userspace faults and | |
506 | * kernelspace faults which result in die(), but not | |
507 | * kernelspace faults which are fixed up. die() gives the | |
508 | * process no chance to handle the signal and notice the | |
509 | * kernel fault information, so that won't result in polluting | |
510 | * the information about previously queued, but not yet | |
511 | * delivered, faults. See also do_general_protection below. | |
512 | */ | |
513 | tsk->thread.error_code = error_code; | |
514 | tsk->thread.trap_no = trapnr; | |
d1895183 | 515 | |
b5964405 IM |
516 | if (info) |
517 | force_sig_info(signr, info, tsk); | |
518 | else | |
519 | force_sig(signr, tsk); | |
520 | return; | |
1da177e4 | 521 | |
b5964405 IM |
522 | kernel_trap: |
523 | if (!fixup_exception(regs)) { | |
524 | tsk->thread.error_code = error_code; | |
525 | tsk->thread.trap_no = trapnr; | |
526 | die(str, regs, error_code); | |
1da177e4 | 527 | } |
b5964405 | 528 | return; |
1da177e4 | 529 | |
b5964405 IM |
530 | vm86_trap: |
531 | if (handle_vm86_trap((struct kernel_vm86_regs *) regs, | |
532 | error_code, trapnr)) | |
533 | goto trap_signal; | |
534 | return; | |
1da177e4 LT |
535 | } |
536 | ||
b5964405 IM |
537 | #define DO_ERROR(trapnr, signr, str, name) \ |
538 | void do_##name(struct pt_regs *regs, long error_code) \ | |
539 | { \ | |
4461145e | 540 | trace_hardirqs_fixup(); \ |
b5964405 | 541 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ |
a8c1be9d | 542 | == NOTIFY_STOP) \ |
b5964405 IM |
543 | return; \ |
544 | do_trap(trapnr, signr, str, 0, regs, error_code, NULL); \ | |
1da177e4 LT |
545 | } |
546 | ||
b5964405 IM |
547 | #define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr, irq) \ |
548 | void do_##name(struct pt_regs *regs, long error_code) \ | |
549 | { \ | |
550 | siginfo_t info; \ | |
551 | if (irq) \ | |
552 | local_irq_enable(); \ | |
553 | info.si_signo = signr; \ | |
554 | info.si_errno = 0; \ | |
555 | info.si_code = sicode; \ | |
556 | info.si_addr = (void __user *)siaddr; \ | |
557 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
a8c1be9d | 558 | == NOTIFY_STOP) \ |
b5964405 IM |
559 | return; \ |
560 | do_trap(trapnr, signr, str, 0, regs, error_code, &info); \ | |
1da177e4 LT |
561 | } |
562 | ||
b5964405 IM |
563 | #define DO_VM86_ERROR(trapnr, signr, str, name) \ |
564 | void do_##name(struct pt_regs *regs, long error_code) \ | |
565 | { \ | |
566 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
a8c1be9d | 567 | == NOTIFY_STOP) \ |
b5964405 IM |
568 | return; \ |
569 | do_trap(trapnr, signr, str, 1, regs, error_code, NULL); \ | |
1da177e4 LT |
570 | } |
571 | ||
b5964405 IM |
572 | #define DO_VM86_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ |
573 | void do_##name(struct pt_regs *regs, long error_code) \ | |
574 | { \ | |
575 | siginfo_t info; \ | |
576 | info.si_signo = signr; \ | |
577 | info.si_errno = 0; \ | |
578 | info.si_code = sicode; \ | |
579 | info.si_addr = (void __user *)siaddr; \ | |
580 | trace_hardirqs_fixup(); \ | |
581 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
a8c1be9d | 582 | == NOTIFY_STOP) \ |
b5964405 IM |
583 | return; \ |
584 | do_trap(trapnr, signr, str, 1, regs, error_code, &info); \ | |
1da177e4 LT |
585 | } |
586 | ||
a8c1be9d | 587 | DO_VM86_ERROR_INFO(0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->ip) |
1da177e4 | 588 | #ifndef CONFIG_KPROBES |
b5964405 | 589 | DO_VM86_ERROR(3, SIGTRAP, "int3", int3) |
1da177e4 | 590 | #endif |
b5964405 IM |
591 | DO_VM86_ERROR(4, SIGSEGV, "overflow", overflow) |
592 | DO_VM86_ERROR(5, SIGSEGV, "bounds", bounds) | |
a8c1be9d AH |
593 | DO_ERROR_INFO(6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->ip, 0) |
594 | DO_ERROR(9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun) | |
1da177e4 | 595 | DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS) |
a8c1be9d AH |
596 | DO_ERROR(11, SIGBUS, "segment not present", segment_not_present) |
597 | DO_ERROR(12, SIGBUS, "stack segment", stack_segment) | |
a10d9a71 | 598 | DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0, 0) |
79bf0e03 | 599 | DO_ERROR_INFO(32, SIGILL, "iret exception", iret_error, ILL_BADSTK, 0, 1) |
1da177e4 | 600 | |
b5964405 | 601 | void __kprobes do_general_protection(struct pt_regs *regs, long error_code) |
1da177e4 | 602 | { |
b5964405 IM |
603 | struct thread_struct *thread; |
604 | struct tss_struct *tss; | |
605 | int cpu; | |
606 | ||
607 | cpu = get_cpu(); | |
608 | tss = &per_cpu(init_tss, cpu); | |
609 | thread = ¤t->thread; | |
1da177e4 LT |
610 | |
611 | /* | |
612 | * Perform the lazy TSS's I/O bitmap copy. If the TSS has an | |
613 | * invalid offset set (the LAZY one) and the faulting thread has | |
614 | * a valid I/O bitmap pointer, we copy the I/O bitmap in the TSS | |
615 | * and we set the offset field correctly. Then we let the CPU to | |
616 | * restart the faulting instruction. | |
617 | */ | |
a75c54f9 | 618 | if (tss->x86_tss.io_bitmap_base == INVALID_IO_BITMAP_OFFSET_LAZY && |
1da177e4 LT |
619 | thread->io_bitmap_ptr) { |
620 | memcpy(tss->io_bitmap, thread->io_bitmap_ptr, | |
621 | thread->io_bitmap_max); | |
622 | /* | |
623 | * If the previously set map was extending to higher ports | |
624 | * than the current one, pad extra space with 0xff (no access). | |
625 | */ | |
b5964405 | 626 | if (thread->io_bitmap_max < tss->io_bitmap_max) { |
1da177e4 LT |
627 | memset((char *) tss->io_bitmap + |
628 | thread->io_bitmap_max, 0xff, | |
629 | tss->io_bitmap_max - thread->io_bitmap_max); | |
b5964405 | 630 | } |
1da177e4 | 631 | tss->io_bitmap_max = thread->io_bitmap_max; |
a75c54f9 | 632 | tss->x86_tss.io_bitmap_base = IO_BITMAP_OFFSET; |
d5cd4aad | 633 | tss->io_bitmap_owner = thread; |
1da177e4 | 634 | put_cpu(); |
b5964405 | 635 | |
1da177e4 LT |
636 | return; |
637 | } | |
638 | put_cpu(); | |
639 | ||
6b6891f9 | 640 | if (regs->flags & X86_VM_MASK) |
1da177e4 LT |
641 | goto gp_in_vm86; |
642 | ||
717b594a | 643 | if (!user_mode(regs)) |
1da177e4 LT |
644 | goto gp_in_kernel; |
645 | ||
646 | current->thread.error_code = error_code; | |
647 | current->thread.trap_no = 13; | |
b5964405 | 648 | |
abd4f750 | 649 | if (show_unhandled_signals && unhandled_signal(current, SIGSEGV) && |
03252919 | 650 | printk_ratelimit()) { |
abd4f750 | 651 | printk(KERN_INFO |
03252919 | 652 | "%s[%d] general protection ip:%lx sp:%lx error:%lx", |
19c5870c | 653 | current->comm, task_pid_nr(current), |
65ea5b03 | 654 | regs->ip, regs->sp, error_code); |
03252919 AK |
655 | print_vma_addr(" in ", regs->ip); |
656 | printk("\n"); | |
657 | } | |
abd4f750 | 658 | |
1da177e4 LT |
659 | force_sig(SIGSEGV, current); |
660 | return; | |
661 | ||
662 | gp_in_vm86: | |
663 | local_irq_enable(); | |
664 | handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code); | |
665 | return; | |
666 | ||
667 | gp_in_kernel: | |
668 | if (!fixup_exception(regs)) { | |
d1895183 AK |
669 | current->thread.error_code = error_code; |
670 | current->thread.trap_no = 13; | |
1da177e4 LT |
671 | if (notify_die(DIE_GPF, "general protection fault", regs, |
672 | error_code, 13, SIGSEGV) == NOTIFY_STOP) | |
673 | return; | |
674 | die("general protection fault", regs, error_code); | |
675 | } | |
676 | } | |
677 | ||
5deb45e3 | 678 | static notrace __kprobes void |
b5964405 | 679 | mem_parity_error(unsigned char reason, struct pt_regs *regs) |
1da177e4 | 680 | { |
b5964405 IM |
681 | printk(KERN_EMERG |
682 | "Uhhuh. NMI received for unknown reason %02x on CPU %d.\n", | |
683 | reason, smp_processor_id()); | |
684 | ||
685 | printk(KERN_EMERG | |
686 | "You have some hardware problem, likely on the PCI bus.\n"); | |
c0d12172 DJ |
687 | |
688 | #if defined(CONFIG_EDAC) | |
b5964405 | 689 | if (edac_handler_set()) { |
c0d12172 DJ |
690 | edac_atomic_assert_error(); |
691 | return; | |
692 | } | |
693 | #endif | |
694 | ||
8da5adda | 695 | if (panic_on_unrecovered_nmi) |
b5964405 | 696 | panic("NMI: Not continuing"); |
1da177e4 | 697 | |
c41c5cd3 | 698 | printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
1da177e4 LT |
699 | |
700 | /* Clear and disable the memory parity error line. */ | |
701 | clear_mem_error(reason); | |
702 | } | |
703 | ||
5deb45e3 | 704 | static notrace __kprobes void |
b5964405 | 705 | io_check_error(unsigned char reason, struct pt_regs *regs) |
1da177e4 LT |
706 | { |
707 | unsigned long i; | |
708 | ||
9c107805 | 709 | printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); |
1da177e4 LT |
710 | show_registers(regs); |
711 | ||
712 | /* Re-enable the IOCK line, wait for a few seconds */ | |
713 | reason = (reason & 0xf) | 8; | |
714 | outb(reason, 0x61); | |
b5964405 | 715 | |
1da177e4 | 716 | i = 2000; |
b5964405 IM |
717 | while (--i) |
718 | udelay(1000); | |
719 | ||
1da177e4 LT |
720 | reason &= ~8; |
721 | outb(reason, 0x61); | |
722 | } | |
723 | ||
5deb45e3 | 724 | static notrace __kprobes void |
b5964405 | 725 | unknown_nmi_error(unsigned char reason, struct pt_regs *regs) |
1da177e4 | 726 | { |
d3597524 JW |
727 | if (notify_die(DIE_NMIUNKNOWN, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) |
728 | return; | |
1da177e4 | 729 | #ifdef CONFIG_MCA |
b5964405 IM |
730 | /* |
731 | * Might actually be able to figure out what the guilty party | |
732 | * is: | |
733 | */ | |
734 | if (MCA_bus) { | |
1da177e4 LT |
735 | mca_handle_nmi(); |
736 | return; | |
737 | } | |
738 | #endif | |
b5964405 IM |
739 | printk(KERN_EMERG |
740 | "Uhhuh. NMI received for unknown reason %02x on CPU %d.\n", | |
741 | reason, smp_processor_id()); | |
742 | ||
c41c5cd3 | 743 | printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); |
8da5adda | 744 | if (panic_on_unrecovered_nmi) |
b5964405 | 745 | panic("NMI: Not continuing"); |
8da5adda | 746 | |
c41c5cd3 | 747 | printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
1da177e4 LT |
748 | } |
749 | ||
750 | static DEFINE_SPINLOCK(nmi_print_lock); | |
751 | ||
ddca03c9 | 752 | void notrace __kprobes die_nmi(char *str, struct pt_regs *regs, int do_panic) |
1da177e4 | 753 | { |
ddca03c9 | 754 | if (notify_die(DIE_NMIWATCHDOG, str, regs, 0, 2, SIGINT) == NOTIFY_STOP) |
748f2edb GA |
755 | return; |
756 | ||
1da177e4 LT |
757 | spin_lock(&nmi_print_lock); |
758 | /* | |
759 | * We are in trouble anyway, lets at least try | |
b5964405 | 760 | * to get a message out: |
1da177e4 LT |
761 | */ |
762 | bust_spinlocks(1); | |
ddca03c9 | 763 | printk(KERN_EMERG "%s", str); |
65ea5b03 PA |
764 | printk(" on CPU%d, ip %08lx, registers:\n", |
765 | smp_processor_id(), regs->ip); | |
1da177e4 | 766 | show_registers(regs); |
ddca03c9 CG |
767 | if (do_panic) |
768 | panic("Non maskable interrupt"); | |
1da177e4 LT |
769 | console_silent(); |
770 | spin_unlock(&nmi_print_lock); | |
771 | bust_spinlocks(0); | |
6e274d14 | 772 | |
b5964405 IM |
773 | /* |
774 | * If we are in kernel we are probably nested up pretty bad | |
775 | * and might aswell get out now while we still can: | |
776 | */ | |
db753bdf | 777 | if (!user_mode_vm(regs)) { |
6e274d14 AN |
778 | current->thread.trap_no = 2; |
779 | crash_kexec(regs); | |
780 | } | |
781 | ||
1da177e4 LT |
782 | do_exit(SIGSEGV); |
783 | } | |
784 | ||
5deb45e3 | 785 | static notrace __kprobes void default_do_nmi(struct pt_regs *regs) |
1da177e4 LT |
786 | { |
787 | unsigned char reason = 0; | |
788 | ||
b5964405 | 789 | /* Only the BSP gets external NMIs from the system: */ |
1da177e4 LT |
790 | if (!smp_processor_id()) |
791 | reason = get_nmi_reason(); | |
b5964405 | 792 | |
1da177e4 | 793 | if (!(reason & 0xc0)) { |
20c0d2d4 | 794 | if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT) |
a8c1be9d | 795 | == NOTIFY_STOP) |
1da177e4 LT |
796 | return; |
797 | #ifdef CONFIG_X86_LOCAL_APIC | |
798 | /* | |
799 | * Ok, so this is none of the documented NMI sources, | |
800 | * so it must be the NMI watchdog. | |
801 | */ | |
3adbbcce | 802 | if (nmi_watchdog_tick(regs, reason)) |
1da177e4 | 803 | return; |
2fbe7b25 | 804 | if (!do_nmi_callback(regs, smp_processor_id())) |
3adbbcce | 805 | unknown_nmi_error(reason, regs); |
b5964405 IM |
806 | #else |
807 | unknown_nmi_error(reason, regs); | |
808 | #endif | |
2fbe7b25 | 809 | |
1da177e4 LT |
810 | return; |
811 | } | |
20c0d2d4 | 812 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) |
1da177e4 | 813 | return; |
a8c1be9d AH |
814 | |
815 | /* AK: following checks seem to be broken on modern chipsets. FIXME */ | |
1da177e4 LT |
816 | if (reason & 0x80) |
817 | mem_parity_error(reason, regs); | |
818 | if (reason & 0x40) | |
819 | io_check_error(reason, regs); | |
820 | /* | |
821 | * Reassert NMI in case it became active meanwhile | |
b5964405 | 822 | * as it's edge-triggered: |
1da177e4 LT |
823 | */ |
824 | reassert_nmi(); | |
825 | } | |
826 | ||
5deb45e3 | 827 | notrace __kprobes void do_nmi(struct pt_regs *regs, long error_code) |
1da177e4 LT |
828 | { |
829 | int cpu; | |
830 | ||
831 | nmi_enter(); | |
832 | ||
833 | cpu = smp_processor_id(); | |
f3705136 | 834 | |
1da177e4 LT |
835 | ++nmi_count(cpu); |
836 | ||
8f4e956b AK |
837 | if (!ignore_nmis) |
838 | default_do_nmi(regs); | |
1da177e4 LT |
839 | |
840 | nmi_exit(); | |
841 | } | |
842 | ||
8f4e956b AK |
843 | void stop_nmi(void) |
844 | { | |
845 | acpi_nmi_disable(); | |
846 | ignore_nmis++; | |
847 | } | |
848 | ||
849 | void restart_nmi(void) | |
850 | { | |
851 | ignore_nmis--; | |
852 | acpi_nmi_enable(); | |
853 | } | |
854 | ||
1da177e4 | 855 | #ifdef CONFIG_KPROBES |
75604d7f | 856 | void __kprobes do_int3(struct pt_regs *regs, long error_code) |
1da177e4 | 857 | { |
143a5d32 PZ |
858 | trace_hardirqs_fixup(); |
859 | ||
1da177e4 LT |
860 | if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP) |
861 | == NOTIFY_STOP) | |
48c88211 | 862 | return; |
b5964405 IM |
863 | /* |
864 | * This is an interrupt gate, because kprobes wants interrupts | |
865 | * disabled. Normal trap handlers don't. | |
866 | */ | |
1da177e4 | 867 | restore_interrupts(regs); |
b5964405 | 868 | |
1da177e4 | 869 | do_trap(3, SIGTRAP, "int3", 1, regs, error_code, NULL); |
1da177e4 LT |
870 | } |
871 | #endif | |
872 | ||
873 | /* | |
874 | * Our handling of the processor debug registers is non-trivial. | |
875 | * We do not clear them on entry and exit from the kernel. Therefore | |
876 | * it is possible to get a watchpoint trap here from inside the kernel. | |
877 | * However, the code in ./ptrace.c has ensured that the user can | |
878 | * only set watchpoints on userspace addresses. Therefore the in-kernel | |
879 | * watchpoint trap can only occur in code which is reading/writing | |
880 | * from user space. Such code must not hold kernel locks (since it | |
881 | * can equally take a page fault), therefore it is safe to call | |
882 | * force_sig_info even though that claims and releases locks. | |
b5964405 | 883 | * |
1da177e4 LT |
884 | * Code in ./signal.c ensures that the debug control register |
885 | * is restored before we deliver any signal, and therefore that | |
886 | * user code runs with the correct debug control register even though | |
887 | * we clear it here. | |
888 | * | |
889 | * Being careful here means that we don't have to be as careful in a | |
890 | * lot of more complicated places (task switching can be a bit lazy | |
891 | * about restoring all the debug state, and ptrace doesn't have to | |
892 | * find every occurrence of the TF bit that could be saved away even | |
893 | * by user code) | |
894 | */ | |
b5964405 | 895 | void __kprobes do_debug(struct pt_regs *regs, long error_code) |
1da177e4 | 896 | { |
1da177e4 | 897 | struct task_struct *tsk = current; |
b5964405 | 898 | unsigned int condition; |
1da177e4 | 899 | |
000f4a9e PZ |
900 | trace_hardirqs_fixup(); |
901 | ||
1cc6f12e | 902 | get_debugreg(condition, 6); |
1da177e4 | 903 | |
10faa81e RM |
904 | /* |
905 | * The processor cleared BTF, so don't mark that we need it set. | |
906 | */ | |
907 | clear_tsk_thread_flag(tsk, TIF_DEBUGCTLMSR); | |
908 | tsk->thread.debugctlmsr = 0; | |
909 | ||
1da177e4 | 910 | if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code, |
a8c1be9d | 911 | SIGTRAP) == NOTIFY_STOP) |
1da177e4 LT |
912 | return; |
913 | /* It's safe to allow irq's after DR6 has been saved */ | |
65ea5b03 | 914 | if (regs->flags & X86_EFLAGS_IF) |
1da177e4 LT |
915 | local_irq_enable(); |
916 | ||
917 | /* Mask out spurious debug traps due to lazy DR7 setting */ | |
918 | if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) { | |
0f534093 | 919 | if (!tsk->thread.debugreg7) |
1da177e4 LT |
920 | goto clear_dr7; |
921 | } | |
922 | ||
6b6891f9 | 923 | if (regs->flags & X86_VM_MASK) |
1da177e4 LT |
924 | goto debug_vm86; |
925 | ||
926 | /* Save debug status register where ptrace can see it */ | |
0f534093 | 927 | tsk->thread.debugreg6 = condition; |
1da177e4 LT |
928 | |
929 | /* | |
930 | * Single-stepping through TF: make sure we ignore any events in | |
931 | * kernel space (but re-enable TF when returning to user mode). | |
932 | */ | |
933 | if (condition & DR_STEP) { | |
934 | /* | |
935 | * We already checked v86 mode above, so we can | |
936 | * check for kernel mode by just checking the CPL | |
937 | * of CS. | |
938 | */ | |
717b594a | 939 | if (!user_mode(regs)) |
1da177e4 LT |
940 | goto clear_TF_reenable; |
941 | } | |
942 | ||
943 | /* Ok, finally something we can handle */ | |
944 | send_sigtrap(tsk, regs, error_code); | |
945 | ||
b5964405 IM |
946 | /* |
947 | * Disable additional traps. They'll be re-enabled when | |
1da177e4 LT |
948 | * the signal is delivered. |
949 | */ | |
950 | clear_dr7: | |
1cc6f12e | 951 | set_debugreg(0, 7); |
1da177e4 LT |
952 | return; |
953 | ||
954 | debug_vm86: | |
955 | handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, 1); | |
956 | return; | |
957 | ||
958 | clear_TF_reenable: | |
959 | set_tsk_thread_flag(tsk, TIF_SINGLESTEP); | |
6093015d | 960 | regs->flags &= ~X86_EFLAGS_TF; |
1da177e4 LT |
961 | return; |
962 | } | |
963 | ||
964 | /* | |
965 | * Note that we play around with the 'TS' bit in an attempt to get | |
966 | * the correct behaviour even in the presence of the asynchronous | |
967 | * IRQ13 behaviour | |
968 | */ | |
65ea5b03 | 969 | void math_error(void __user *ip) |
1da177e4 | 970 | { |
b5964405 IM |
971 | struct task_struct *task; |
972 | unsigned short cwd; | |
973 | unsigned short swd; | |
1da177e4 | 974 | siginfo_t info; |
1da177e4 LT |
975 | |
976 | /* | |
977 | * Save the info for the exception handler and clear the error. | |
978 | */ | |
979 | task = current; | |
980 | save_init_fpu(task); | |
981 | task->thread.trap_no = 16; | |
982 | task->thread.error_code = 0; | |
983 | info.si_signo = SIGFPE; | |
984 | info.si_errno = 0; | |
985 | info.si_code = __SI_FAULT; | |
65ea5b03 | 986 | info.si_addr = ip; |
1da177e4 LT |
987 | /* |
988 | * (~cwd & swd) will mask out exceptions that are not set to unmasked | |
989 | * status. 0x3f is the exception bits in these regs, 0x200 is the | |
990 | * C1 reg you need in case of a stack fault, 0x040 is the stack | |
991 | * fault bit. We should only be taking one exception at a time, | |
992 | * so if this combination doesn't produce any single exception, | |
a8c1be9d | 993 | * then we have a bad program that isn't synchronizing its FPU usage |
1da177e4 LT |
994 | * and it will suffer the consequences since we won't be able to |
995 | * fully reproduce the context of the exception | |
996 | */ | |
997 | cwd = get_fpu_cwd(task); | |
998 | swd = get_fpu_swd(task); | |
b1daec30 | 999 | switch (swd & ~cwd & 0x3f) { |
b5964405 IM |
1000 | case 0x000: /* No unmasked exception */ |
1001 | return; | |
a8c1be9d | 1002 | default: /* Multiple exceptions */ |
b5964405 IM |
1003 | break; |
1004 | case 0x001: /* Invalid Op */ | |
1005 | /* | |
1006 | * swd & 0x240 == 0x040: Stack Underflow | |
1007 | * swd & 0x240 == 0x240: Stack Overflow | |
1008 | * User must clear the SF bit (0x40) if set | |
1009 | */ | |
1010 | info.si_code = FPE_FLTINV; | |
1011 | break; | |
1012 | case 0x002: /* Denormalize */ | |
1013 | case 0x010: /* Underflow */ | |
1014 | info.si_code = FPE_FLTUND; | |
1015 | break; | |
1016 | case 0x004: /* Zero Divide */ | |
1017 | info.si_code = FPE_FLTDIV; | |
1018 | break; | |
1019 | case 0x008: /* Overflow */ | |
1020 | info.si_code = FPE_FLTOVF; | |
1021 | break; | |
1022 | case 0x020: /* Precision */ | |
1023 | info.si_code = FPE_FLTRES; | |
1024 | break; | |
1da177e4 LT |
1025 | } |
1026 | force_sig_info(SIGFPE, &info, task); | |
1027 | } | |
1028 | ||
b5964405 | 1029 | void do_coprocessor_error(struct pt_regs *regs, long error_code) |
1da177e4 LT |
1030 | { |
1031 | ignore_fpu_irq = 1; | |
65ea5b03 | 1032 | math_error((void __user *)regs->ip); |
1da177e4 LT |
1033 | } |
1034 | ||
65ea5b03 | 1035 | static void simd_math_error(void __user *ip) |
1da177e4 | 1036 | { |
b5964405 | 1037 | struct task_struct *task; |
1da177e4 | 1038 | unsigned short mxcsr; |
b5964405 | 1039 | siginfo_t info; |
1da177e4 LT |
1040 | |
1041 | /* | |
1042 | * Save the info for the exception handler and clear the error. | |
1043 | */ | |
1044 | task = current; | |
1045 | save_init_fpu(task); | |
1046 | task->thread.trap_no = 19; | |
1047 | task->thread.error_code = 0; | |
1048 | info.si_signo = SIGFPE; | |
1049 | info.si_errno = 0; | |
1050 | info.si_code = __SI_FAULT; | |
65ea5b03 | 1051 | info.si_addr = ip; |
1da177e4 LT |
1052 | /* |
1053 | * The SIMD FPU exceptions are handled a little differently, as there | |
1054 | * is only a single status/control register. Thus, to determine which | |
1055 | * unmasked exception was caught we must mask the exception mask bits | |
1056 | * at 0x1f80, and then use these to mask the exception bits at 0x3f. | |
1057 | */ | |
1058 | mxcsr = get_fpu_mxcsr(task); | |
1059 | switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) { | |
b5964405 IM |
1060 | case 0x000: |
1061 | default: | |
1062 | break; | |
1063 | case 0x001: /* Invalid Op */ | |
1064 | info.si_code = FPE_FLTINV; | |
1065 | break; | |
1066 | case 0x002: /* Denormalize */ | |
1067 | case 0x010: /* Underflow */ | |
1068 | info.si_code = FPE_FLTUND; | |
1069 | break; | |
1070 | case 0x004: /* Zero Divide */ | |
1071 | info.si_code = FPE_FLTDIV; | |
1072 | break; | |
1073 | case 0x008: /* Overflow */ | |
1074 | info.si_code = FPE_FLTOVF; | |
1075 | break; | |
1076 | case 0x020: /* Precision */ | |
1077 | info.si_code = FPE_FLTRES; | |
1078 | break; | |
1da177e4 LT |
1079 | } |
1080 | force_sig_info(SIGFPE, &info, task); | |
1081 | } | |
1082 | ||
b5964405 | 1083 | void do_simd_coprocessor_error(struct pt_regs *regs, long error_code) |
1da177e4 LT |
1084 | { |
1085 | if (cpu_has_xmm) { | |
1086 | /* Handle SIMD FPU exceptions on PIII+ processors. */ | |
1087 | ignore_fpu_irq = 1; | |
65ea5b03 | 1088 | simd_math_error((void __user *)regs->ip); |
b5964405 IM |
1089 | return; |
1090 | } | |
1091 | /* | |
1092 | * Handle strange cache flush from user space exception | |
1093 | * in all other cases. This is undocumented behaviour. | |
1094 | */ | |
6b6891f9 | 1095 | if (regs->flags & X86_VM_MASK) { |
b5964405 IM |
1096 | handle_vm86_fault((struct kernel_vm86_regs *)regs, error_code); |
1097 | return; | |
1da177e4 | 1098 | } |
b5964405 IM |
1099 | current->thread.trap_no = 19; |
1100 | current->thread.error_code = error_code; | |
1101 | die_if_kernel("cache flush denied", regs, error_code); | |
1102 | force_sig(SIGSEGV, current); | |
1da177e4 LT |
1103 | } |
1104 | ||
b5964405 | 1105 | void do_spurious_interrupt_bug(struct pt_regs *regs, long error_code) |
1da177e4 LT |
1106 | { |
1107 | #if 0 | |
1108 | /* No need to warn about this any longer. */ | |
b5964405 | 1109 | printk(KERN_INFO "Ignoring P6 Local APIC Spurious Interrupt Bug...\n"); |
1da177e4 LT |
1110 | #endif |
1111 | } | |
1112 | ||
b5964405 | 1113 | unsigned long patch_espfix_desc(unsigned long uesp, unsigned long kesp) |
1da177e4 | 1114 | { |
736f12bf | 1115 | struct desc_struct *gdt = get_cpu_gdt_table(smp_processor_id()); |
be44d2aa SS |
1116 | unsigned long base = (kesp - uesp) & -THREAD_SIZE; |
1117 | unsigned long new_kesp = kesp - base; | |
1118 | unsigned long lim_pages = (new_kesp | (THREAD_SIZE - 1)) >> PAGE_SHIFT; | |
1119 | __u64 desc = *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS]; | |
b5964405 | 1120 | |
be44d2aa | 1121 | /* Set up base for espfix segment */ |
b5964405 IM |
1122 | desc &= 0x00f0ff0000000000ULL; |
1123 | desc |= ((((__u64)base) << 16) & 0x000000ffffff0000ULL) | | |
be44d2aa SS |
1124 | ((((__u64)base) << 32) & 0xff00000000000000ULL) | |
1125 | ((((__u64)lim_pages) << 32) & 0x000f000000000000ULL) | | |
1126 | (lim_pages & 0xffff); | |
1127 | *(__u64 *)&gdt[GDT_ENTRY_ESPFIX_SS] = desc; | |
b5964405 | 1128 | |
be44d2aa | 1129 | return new_kesp; |
1da177e4 LT |
1130 | } |
1131 | ||
1132 | /* | |
b5964405 | 1133 | * 'math_state_restore()' saves the current math information in the |
1da177e4 LT |
1134 | * old math state array, and gets the new ones from the current task |
1135 | * | |
1136 | * Careful.. There are problems with IBM-designed IRQ13 behaviour. | |
1137 | * Don't touch unless you *really* know how it works. | |
1138 | * | |
1139 | * Must be called with kernel preemption disabled (in this case, | |
1140 | * local interrupts are disabled at the call-site in entry.S). | |
1141 | */ | |
acc20761 | 1142 | asmlinkage void math_state_restore(void) |
1da177e4 LT |
1143 | { |
1144 | struct thread_info *thread = current_thread_info(); | |
1145 | struct task_struct *tsk = thread->task; | |
1146 | ||
aa283f49 SS |
1147 | if (!tsk_used_math(tsk)) { |
1148 | local_irq_enable(); | |
1149 | /* | |
1150 | * does a slab alloc which can sleep | |
1151 | */ | |
1152 | if (init_fpu(tsk)) { | |
1153 | /* | |
1154 | * ran out of memory! | |
1155 | */ | |
1156 | do_group_exit(SIGKILL); | |
1157 | return; | |
1158 | } | |
1159 | local_irq_disable(); | |
1160 | } | |
1161 | ||
b5964405 | 1162 | clts(); /* Allow maths ops (or we recurse) */ |
1da177e4 LT |
1163 | restore_fpu(tsk); |
1164 | thread->status |= TS_USEDFPU; /* So we fnsave on switch_to() */ | |
acc20761 | 1165 | tsk->fpu_counter++; |
1da177e4 | 1166 | } |
5992b6da | 1167 | EXPORT_SYMBOL_GPL(math_state_restore); |
1da177e4 LT |
1168 | |
1169 | #ifndef CONFIG_MATH_EMULATION | |
1170 | ||
1171 | asmlinkage void math_emulate(long arg) | |
1172 | { | |
b5964405 IM |
1173 | printk(KERN_EMERG |
1174 | "math-emulation not enabled and no coprocessor found.\n"); | |
1175 | printk(KERN_EMERG "killing %s.\n", current->comm); | |
1176 | force_sig(SIGFPE, current); | |
1da177e4 LT |
1177 | schedule(); |
1178 | } | |
1179 | ||
1180 | #endif /* CONFIG_MATH_EMULATION */ | |
1181 | ||
1da177e4 LT |
1182 | void __init trap_init(void) |
1183 | { | |
dbeb2be2 RR |
1184 | int i; |
1185 | ||
1da177e4 | 1186 | #ifdef CONFIG_EISA |
927222b1 | 1187 | void __iomem *p = early_ioremap(0x0FFFD9, 4); |
b5964405 IM |
1188 | |
1189 | if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24)) | |
1da177e4 | 1190 | EISA_bus = 1; |
927222b1 | 1191 | early_iounmap(p, 4); |
1da177e4 LT |
1192 | #endif |
1193 | ||
a8c1be9d AH |
1194 | set_trap_gate(0, ÷_error); |
1195 | set_intr_gate(1, &debug); | |
1196 | set_intr_gate(2, &nmi); | |
1197 | set_system_intr_gate(3, &int3); /* int3 can be called from all */ | |
1198 | set_system_gate(4, &overflow); /* int4 can be called from all */ | |
1199 | set_trap_gate(5, &bounds); | |
1200 | set_trap_gate(6, &invalid_op); | |
1201 | set_trap_gate(7, &device_not_available); | |
1202 | set_task_gate(8, GDT_ENTRY_DOUBLEFAULT_TSS); | |
1203 | set_trap_gate(9, &coprocessor_segment_overrun); | |
b5964405 IM |
1204 | set_trap_gate(10, &invalid_TSS); |
1205 | set_trap_gate(11, &segment_not_present); | |
1206 | set_trap_gate(12, &stack_segment); | |
1207 | set_trap_gate(13, &general_protection); | |
1208 | set_intr_gate(14, &page_fault); | |
1209 | set_trap_gate(15, &spurious_interrupt_bug); | |
1210 | set_trap_gate(16, &coprocessor_error); | |
1211 | set_trap_gate(17, &alignment_check); | |
1da177e4 | 1212 | #ifdef CONFIG_X86_MCE |
b5964405 | 1213 | set_trap_gate(18, &machine_check); |
1da177e4 | 1214 | #endif |
b5964405 | 1215 | set_trap_gate(19, &simd_coprocessor_error); |
1da177e4 | 1216 | |
d43c6e80 | 1217 | if (cpu_has_fxsr) { |
d43c6e80 JB |
1218 | printk(KERN_INFO "Enabling fast FPU save and restore... "); |
1219 | set_in_cr4(X86_CR4_OSFXSR); | |
1220 | printk("done.\n"); | |
1221 | } | |
1222 | if (cpu_has_xmm) { | |
b5964405 IM |
1223 | printk(KERN_INFO |
1224 | "Enabling unmasked SIMD FPU exception support... "); | |
d43c6e80 JB |
1225 | set_in_cr4(X86_CR4_OSXMMEXCPT); |
1226 | printk("done.\n"); | |
1227 | } | |
1228 | ||
b5964405 | 1229 | set_system_gate(SYSCALL_VECTOR, &system_call); |
1da177e4 | 1230 | |
b5964405 | 1231 | /* Reserve all the builtin and the syscall vector: */ |
dbeb2be2 RR |
1232 | for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++) |
1233 | set_bit(i, used_vectors); | |
b5964405 | 1234 | |
dbeb2be2 RR |
1235 | set_bit(SYSCALL_VECTOR, used_vectors); |
1236 | ||
61c4628b | 1237 | init_thread_xstate(); |
1da177e4 | 1238 | /* |
b5964405 | 1239 | * Should be a barrier for any external CPU state: |
1da177e4 LT |
1240 | */ |
1241 | cpu_init(); | |
1242 | ||
1243 | trap_init_hook(); | |
1244 | } | |
1245 | ||
1246 | static int __init kstack_setup(char *s) | |
1247 | { | |
1248 | kstack_depth_to_print = simple_strtoul(s, NULL, 0); | |
b5964405 | 1249 | |
9b41046c | 1250 | return 1; |
1da177e4 LT |
1251 | } |
1252 | __setup("kstack=", kstack_setup); | |
86c41837 CE |
1253 | |
1254 | static int __init code_bytes_setup(char *s) | |
1255 | { | |
1256 | code_bytes = simple_strtoul(s, NULL, 0); | |
1257 | if (code_bytes > 8192) | |
1258 | code_bytes = 8192; | |
1259 | ||
1260 | return 1; | |
1261 | } | |
1262 | __setup("code_bytes=", code_bytes_setup); |