Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * linux/arch/i386/traps.c | |
3 | * | |
4 | * Copyright (C) 1991, 1992 Linus Torvalds | |
5 | * | |
6 | * Pentium III FXSR, SSE support | |
7 | * Gareth Hughes <gareth@valinux.com>, May 2000 | |
8 | */ | |
9 | ||
10 | /* | |
11 | * 'Traps.c' handles hardware traps and faults after we have saved some | |
12 | * state in 'asm.s'. | |
13 | */ | |
1da177e4 LT |
14 | #include <linux/sched.h> |
15 | #include <linux/kernel.h> | |
16 | #include <linux/string.h> | |
17 | #include <linux/errno.h> | |
18 | #include <linux/timer.h> | |
19 | #include <linux/mm.h> | |
20 | #include <linux/init.h> | |
21 | #include <linux/delay.h> | |
22 | #include <linux/spinlock.h> | |
23 | #include <linux/interrupt.h> | |
24 | #include <linux/highmem.h> | |
25 | #include <linux/kallsyms.h> | |
26 | #include <linux/ptrace.h> | |
27 | #include <linux/utsname.h> | |
28 | #include <linux/kprobes.h> | |
6e274d14 | 29 | #include <linux/kexec.h> |
176a2718 | 30 | #include <linux/unwind.h> |
1da177e4 LT |
31 | |
32 | #ifdef CONFIG_EISA | |
33 | #include <linux/ioport.h> | |
34 | #include <linux/eisa.h> | |
35 | #endif | |
36 | ||
37 | #ifdef CONFIG_MCA | |
38 | #include <linux/mca.h> | |
39 | #endif | |
40 | ||
41 | #include <asm/processor.h> | |
42 | #include <asm/system.h> | |
43 | #include <asm/uaccess.h> | |
44 | #include <asm/io.h> | |
45 | #include <asm/atomic.h> | |
46 | #include <asm/debugreg.h> | |
47 | #include <asm/desc.h> | |
48 | #include <asm/i387.h> | |
49 | #include <asm/nmi.h> | |
176a2718 | 50 | #include <asm/unwind.h> |
1da177e4 LT |
51 | #include <asm/smp.h> |
52 | #include <asm/arch_hooks.h> | |
53 | #include <asm/kdebug.h> | |
2b14a78c | 54 | #include <asm/stacktrace.h> |
1da177e4 | 55 | |
1da177e4 LT |
56 | #include <linux/module.h> |
57 | ||
58 | #include "mach_traps.h" | |
59 | ||
60 | asmlinkage int system_call(void); | |
61 | ||
62 | struct desc_struct default_ldt[] = { { 0, 0 }, { 0, 0 }, { 0, 0 }, | |
63 | { 0, 0 }, { 0, 0 } }; | |
64 | ||
65 | /* Do we ignore FPU interrupts ? */ | |
66 | char ignore_fpu_irq = 0; | |
67 | ||
68 | /* | |
69 | * The IDT has to be page-aligned to simplify the Pentium | |
70 | * F0 0F bug workaround.. We have a special link segment | |
71 | * for this. | |
72 | */ | |
73 | struct desc_struct idt_table[256] __attribute__((__section__(".data.idt"))) = { {0, 0}, }; | |
74 | ||
75 | asmlinkage void divide_error(void); | |
76 | asmlinkage void debug(void); | |
77 | asmlinkage void nmi(void); | |
78 | asmlinkage void int3(void); | |
79 | asmlinkage void overflow(void); | |
80 | asmlinkage void bounds(void); | |
81 | asmlinkage void invalid_op(void); | |
82 | asmlinkage void device_not_available(void); | |
83 | asmlinkage void coprocessor_segment_overrun(void); | |
84 | asmlinkage void invalid_TSS(void); | |
85 | asmlinkage void segment_not_present(void); | |
86 | asmlinkage void stack_segment(void); | |
87 | asmlinkage void general_protection(void); | |
88 | asmlinkage void page_fault(void); | |
89 | asmlinkage void coprocessor_error(void); | |
90 | asmlinkage void simd_coprocessor_error(void); | |
91 | asmlinkage void alignment_check(void); | |
92 | asmlinkage void spurious_interrupt_bug(void); | |
93 | asmlinkage void machine_check(void); | |
94 | ||
95 | static int kstack_depth_to_print = 24; | |
ea424055 | 96 | #ifdef CONFIG_STACK_UNWIND |
c33bd9aa | 97 | static int call_trace = 1; |
ea424055 JB |
98 | #else |
99 | #define call_trace (-1) | |
100 | #endif | |
e041c683 | 101 | ATOMIC_NOTIFIER_HEAD(i386die_chain); |
1da177e4 LT |
102 | |
103 | int register_die_notifier(struct notifier_block *nb) | |
104 | { | |
101f12af | 105 | vmalloc_sync_all(); |
e041c683 | 106 | return atomic_notifier_chain_register(&i386die_chain, nb); |
1da177e4 | 107 | } |
1454aed9 | 108 | EXPORT_SYMBOL(register_die_notifier); /* used modular by kdb */ |
1da177e4 | 109 | |
e041c683 AS |
110 | int unregister_die_notifier(struct notifier_block *nb) |
111 | { | |
112 | return atomic_notifier_chain_unregister(&i386die_chain, nb); | |
113 | } | |
1454aed9 | 114 | EXPORT_SYMBOL(unregister_die_notifier); /* used modular by kdb */ |
e041c683 | 115 | |
1da177e4 LT |
116 | static inline int valid_stack_ptr(struct thread_info *tinfo, void *p) |
117 | { | |
118 | return p > (void *)tinfo && | |
119 | p < (void *)tinfo + THREAD_SIZE - 3; | |
120 | } | |
121 | ||
122 | static inline unsigned long print_context_stack(struct thread_info *tinfo, | |
7aa89746 | 123 | unsigned long *stack, unsigned long ebp, |
2b14a78c | 124 | struct stacktrace_ops *ops, void *data) |
1da177e4 LT |
125 | { |
126 | unsigned long addr; | |
127 | ||
128 | #ifdef CONFIG_FRAME_POINTER | |
129 | while (valid_stack_ptr(tinfo, (void *)ebp)) { | |
130 | addr = *(unsigned long *)(ebp + 4); | |
2b14a78c | 131 | ops->address(data, addr); |
b88d4f1d IM |
132 | /* |
133 | * break out of recursive entries (such as | |
134 | * end_of_stack_stop_unwind_function): | |
135 | */ | |
136 | if (ebp == *(unsigned long *)ebp) | |
137 | break; | |
1da177e4 LT |
138 | ebp = *(unsigned long *)ebp; |
139 | } | |
140 | #else | |
141 | while (valid_stack_ptr(tinfo, stack)) { | |
142 | addr = *stack++; | |
7aa89746 | 143 | if (__kernel_text_address(addr)) |
2b14a78c | 144 | ops->address(data, addr); |
1da177e4 LT |
145 | } |
146 | #endif | |
147 | return ebp; | |
148 | } | |
149 | ||
2b14a78c AK |
150 | struct ops_and_data { |
151 | struct stacktrace_ops *ops; | |
152 | void *data; | |
153 | }; | |
154 | ||
f0a5c315 | 155 | static asmlinkage int |
2b14a78c | 156 | dump_trace_unwind(struct unwind_frame_info *info, void *data) |
176a2718 | 157 | { |
2b14a78c | 158 | struct ops_and_data *oad = (struct ops_and_data *)data; |
c33bd9aa | 159 | int n = 0; |
176a2718 JB |
160 | |
161 | while (unwind(info) == 0 && UNW_PC(info)) { | |
f0a5c315 | 162 | n++; |
2b14a78c | 163 | oad->ops->address(oad->data, UNW_PC(info)); |
176a2718 JB |
164 | if (arch_unw_user_mode(info)) |
165 | break; | |
166 | } | |
c33bd9aa | 167 | return n; |
176a2718 JB |
168 | } |
169 | ||
2b14a78c AK |
170 | void dump_trace(struct task_struct *task, struct pt_regs *regs, |
171 | unsigned long *stack, | |
172 | struct stacktrace_ops *ops, void *data) | |
1da177e4 LT |
173 | { |
174 | unsigned long ebp; | |
175 | ||
176 | if (!task) | |
177 | task = current; | |
178 | ||
c33bd9aa JB |
179 | if (call_trace >= 0) { |
180 | int unw_ret = 0; | |
181 | struct unwind_frame_info info; | |
2b14a78c | 182 | struct ops_and_data oad = { .ops = ops, .data = data }; |
c33bd9aa JB |
183 | |
184 | if (regs) { | |
185 | if (unwind_init_frame_info(&info, task, regs) == 0) | |
2b14a78c | 186 | unw_ret = dump_trace_unwind(&info, &oad); |
c33bd9aa | 187 | } else if (task == current) |
2b14a78c | 188 | unw_ret = unwind_init_running(&info, dump_trace_unwind, &oad); |
c33bd9aa JB |
189 | else { |
190 | if (unwind_init_blocked(&info, task) == 0) | |
2b14a78c | 191 | unw_ret = dump_trace_unwind(&info, &oad); |
176a2718 | 192 | } |
ea424055 JB |
193 | if (unw_ret > 0) { |
194 | if (call_trace == 1 && !arch_unw_user_mode(&info)) { | |
2b14a78c | 195 | ops->warning_symbol(data, "DWARF2 unwinder stuck at %s\n", |
ea424055 JB |
196 | UNW_PC(&info)); |
197 | if (UNW_SP(&info) >= PAGE_OFFSET) { | |
2b14a78c | 198 | ops->warning(data, "Leftover inexact backtrace:\n"); |
70583161 | 199 | stack = (void *)UNW_SP(&info); |
950fee84 AK |
200 | if (!stack) |
201 | return; | |
ea424055 | 202 | } else |
2b14a78c | 203 | ops->warning(data, "Full inexact backtrace again:\n"); |
ea424055 | 204 | } else if (call_trace >= 1) |
c33bd9aa | 205 | return; |
c97d20a6 | 206 | else |
2b14a78c | 207 | ops->warning(data, "Full inexact backtrace again:\n"); |
ea424055 | 208 | } else |
2b14a78c AK |
209 | ops->warning(data, "Inexact backtrace:\n"); |
210 | } else if (!stack) { | |
211 | unsigned long dummy; | |
212 | stack = &dummy; | |
213 | if (task && task != current) | |
214 | stack = (unsigned long *)task->thread.esp; | |
176a2718 JB |
215 | } |
216 | ||
1da177e4 LT |
217 | if (task == current) { |
218 | /* Grab ebp right from our regs */ | |
219 | asm ("movl %%ebp, %0" : "=r" (ebp) : ); | |
220 | } else { | |
221 | /* ebp is the last reg pushed by switch_to */ | |
222 | ebp = *(unsigned long *) task->thread.esp; | |
223 | } | |
224 | ||
225 | while (1) { | |
226 | struct thread_info *context; | |
227 | context = (struct thread_info *) | |
228 | ((unsigned long)stack & (~(THREAD_SIZE - 1))); | |
2b14a78c AK |
229 | ebp = print_context_stack(context, stack, ebp, ops, data); |
230 | /* Should be after the line below, but somewhere | |
231 | in early boot context comes out corrupted and we | |
232 | can't reference it -AK */ | |
233 | if (ops->stack(data, "IRQ") < 0) | |
234 | break; | |
1da177e4 LT |
235 | stack = (unsigned long*)context->previous_esp; |
236 | if (!stack) | |
237 | break; | |
1da177e4 LT |
238 | } |
239 | } | |
2b14a78c AK |
240 | EXPORT_SYMBOL(dump_trace); |
241 | ||
242 | static void | |
243 | print_trace_warning_symbol(void *data, char *msg, unsigned long symbol) | |
244 | { | |
245 | printk(data); | |
246 | print_symbol(msg, symbol); | |
247 | printk("\n"); | |
248 | } | |
249 | ||
250 | static void print_trace_warning(void *data, char *msg) | |
251 | { | |
252 | printk("%s%s\n", (char *)data, msg); | |
253 | } | |
254 | ||
255 | static int print_trace_stack(void *data, char *name) | |
256 | { | |
257 | return 0; | |
258 | } | |
259 | ||
260 | /* | |
261 | * Print one address/symbol entries per line. | |
262 | */ | |
263 | static void print_trace_address(void *data, unsigned long addr) | |
264 | { | |
265 | printk("%s [<%08lx>] ", (char *)data, addr); | |
266 | print_symbol("%s\n", addr); | |
267 | } | |
268 | ||
269 | static struct stacktrace_ops print_trace_ops = { | |
270 | .warning = print_trace_warning, | |
271 | .warning_symbol = print_trace_warning_symbol, | |
272 | .stack = print_trace_stack, | |
273 | .address = print_trace_address, | |
274 | }; | |
275 | ||
276 | static void | |
277 | show_trace_log_lvl(struct task_struct *task, struct pt_regs *regs, | |
278 | unsigned long * stack, char *log_lvl) | |
279 | { | |
280 | dump_trace(task, regs, stack, &print_trace_ops, log_lvl); | |
281 | printk("%s =======================\n", log_lvl); | |
282 | } | |
1da177e4 | 283 | |
2b14a78c AK |
284 | void show_trace(struct task_struct *task, struct pt_regs *regs, |
285 | unsigned long * stack) | |
7aa89746 | 286 | { |
176a2718 | 287 | show_trace_log_lvl(task, regs, stack, ""); |
7aa89746 CE |
288 | } |
289 | ||
176a2718 JB |
290 | static void show_stack_log_lvl(struct task_struct *task, struct pt_regs *regs, |
291 | unsigned long *esp, char *log_lvl) | |
1da177e4 LT |
292 | { |
293 | unsigned long *stack; | |
294 | int i; | |
295 | ||
296 | if (esp == NULL) { | |
297 | if (task) | |
298 | esp = (unsigned long*)task->thread.esp; | |
299 | else | |
300 | esp = (unsigned long *)&esp; | |
301 | } | |
302 | ||
303 | stack = esp; | |
304 | for(i = 0; i < kstack_depth_to_print; i++) { | |
305 | if (kstack_end(stack)) | |
306 | break; | |
75874d5c CE |
307 | if (i && ((i % 8) == 0)) |
308 | printk("\n%s ", log_lvl); | |
1da177e4 LT |
309 | printk("%08lx ", *stack++); |
310 | } | |
75874d5c | 311 | printk("\n%sCall Trace:\n", log_lvl); |
176a2718 | 312 | show_trace_log_lvl(task, regs, esp, log_lvl); |
7aa89746 CE |
313 | } |
314 | ||
315 | void show_stack(struct task_struct *task, unsigned long *esp) | |
316 | { | |
75874d5c | 317 | printk(" "); |
176a2718 | 318 | show_stack_log_lvl(task, NULL, esp, ""); |
1da177e4 LT |
319 | } |
320 | ||
321 | /* | |
322 | * The architecture-independent dump_stack generator | |
323 | */ | |
324 | void dump_stack(void) | |
325 | { | |
326 | unsigned long stack; | |
327 | ||
176a2718 | 328 | show_trace(current, NULL, &stack); |
1da177e4 LT |
329 | } |
330 | ||
331 | EXPORT_SYMBOL(dump_stack); | |
332 | ||
333 | void show_registers(struct pt_regs *regs) | |
334 | { | |
335 | int i; | |
336 | int in_kernel = 1; | |
337 | unsigned long esp; | |
338 | unsigned short ss; | |
339 | ||
340 | esp = (unsigned long) (®s->esp); | |
0998e422 | 341 | savesegment(ss, ss); |
db753bdf | 342 | if (user_mode_vm(regs)) { |
1da177e4 LT |
343 | in_kernel = 0; |
344 | esp = regs->esp; | |
345 | ss = regs->xss & 0xffff; | |
346 | } | |
347 | print_modules(); | |
9c107805 | 348 | printk(KERN_EMERG "CPU: %d\nEIP: %04x:[<%08lx>] %s VLI\n" |
b53e8f68 | 349 | "EFLAGS: %08lx (%s %.*s) \n", |
1da177e4 | 350 | smp_processor_id(), 0xffff & regs->xcs, regs->eip, |
b53e8f68 CE |
351 | print_tainted(), regs->eflags, system_utsname.release, |
352 | (int)strcspn(system_utsname.version, " "), | |
353 | system_utsname.version); | |
9c107805 DJ |
354 | print_symbol(KERN_EMERG "EIP is at %s\n", regs->eip); |
355 | printk(KERN_EMERG "eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", | |
1da177e4 | 356 | regs->eax, regs->ebx, regs->ecx, regs->edx); |
9c107805 | 357 | printk(KERN_EMERG "esi: %08lx edi: %08lx ebp: %08lx esp: %08lx\n", |
1da177e4 | 358 | regs->esi, regs->edi, regs->ebp, esp); |
9c107805 | 359 | printk(KERN_EMERG "ds: %04x es: %04x ss: %04x\n", |
1da177e4 | 360 | regs->xds & 0xffff, regs->xes & 0xffff, ss); |
7e04a118 CE |
361 | printk(KERN_EMERG "Process %.*s (pid: %d, ti=%p task=%p task.ti=%p)", |
362 | TASK_COMM_LEN, current->comm, current->pid, | |
363 | current_thread_info(), current, current->thread_info); | |
1da177e4 LT |
364 | /* |
365 | * When in-kernel, we also print out the stack and code at the | |
366 | * time of the fault.. | |
367 | */ | |
368 | if (in_kernel) { | |
3f3ae347 | 369 | u8 __user *eip; |
1da177e4 | 370 | |
9c107805 | 371 | printk("\n" KERN_EMERG "Stack: "); |
176a2718 | 372 | show_stack_log_lvl(NULL, regs, (unsigned long *)esp, KERN_EMERG); |
1da177e4 | 373 | |
9c107805 | 374 | printk(KERN_EMERG "Code: "); |
1da177e4 | 375 | |
3f3ae347 | 376 | eip = (u8 __user *)regs->eip - 43; |
1da177e4 LT |
377 | for (i = 0; i < 64; i++, eip++) { |
378 | unsigned char c; | |
379 | ||
3f3ae347 | 380 | if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) { |
1da177e4 LT |
381 | printk(" Bad EIP value."); |
382 | break; | |
383 | } | |
3f3ae347 | 384 | if (eip == (u8 __user *)regs->eip) |
1da177e4 LT |
385 | printk("<%02x> ", c); |
386 | else | |
387 | printk("%02x ", c); | |
388 | } | |
389 | } | |
390 | printk("\n"); | |
391 | } | |
392 | ||
393 | static void handle_BUG(struct pt_regs *regs) | |
394 | { | |
b7015331 | 395 | unsigned long eip = regs->eip; |
1da177e4 | 396 | unsigned short ud2; |
1da177e4 LT |
397 | |
398 | if (eip < PAGE_OFFSET) | |
b7015331 | 399 | return; |
3f3ae347 | 400 | if (__get_user(ud2, (unsigned short __user *)eip)) |
b7015331 | 401 | return; |
1da177e4 | 402 | if (ud2 != 0x0b0f) |
b7015331 | 403 | return; |
1da177e4 | 404 | |
9c107805 | 405 | printk(KERN_EMERG "------------[ cut here ]------------\n"); |
1da177e4 | 406 | |
b7015331 CE |
407 | #ifdef CONFIG_DEBUG_BUGVERBOSE |
408 | do { | |
409 | unsigned short line; | |
410 | char *file; | |
411 | char c; | |
412 | ||
413 | if (__get_user(line, (unsigned short __user *)(eip + 2))) | |
414 | break; | |
415 | if (__get_user(file, (char * __user *)(eip + 4)) || | |
416 | (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) | |
417 | file = "<bad filename>"; | |
1da177e4 | 418 | |
b7015331 CE |
419 | printk(KERN_EMERG "kernel BUG at %s:%d!\n", file, line); |
420 | return; | |
421 | } while (0); | |
422 | #endif | |
423 | printk(KERN_EMERG "Kernel BUG at [verbose debug info unavailable]\n"); | |
1da177e4 LT |
424 | } |
425 | ||
6e274d14 AN |
426 | /* This is gone through when something in the kernel |
427 | * has done something bad and is about to be terminated. | |
428 | */ | |
1da177e4 LT |
429 | void die(const char * str, struct pt_regs * regs, long err) |
430 | { | |
431 | static struct { | |
432 | spinlock_t lock; | |
433 | u32 lock_owner; | |
434 | int lock_owner_depth; | |
435 | } die = { | |
436 | .lock = SPIN_LOCK_UNLOCKED, | |
437 | .lock_owner = -1, | |
438 | .lock_owner_depth = 0 | |
439 | }; | |
440 | static int die_counter; | |
e43d674f | 441 | unsigned long flags; |
1da177e4 | 442 | |
dd287796 AM |
443 | oops_enter(); |
444 | ||
39c715b7 | 445 | if (die.lock_owner != raw_smp_processor_id()) { |
1da177e4 | 446 | console_verbose(); |
e43d674f | 447 | spin_lock_irqsave(&die.lock, flags); |
1da177e4 LT |
448 | die.lock_owner = smp_processor_id(); |
449 | die.lock_owner_depth = 0; | |
450 | bust_spinlocks(1); | |
451 | } | |
e43d674f JB |
452 | else |
453 | local_save_flags(flags); | |
1da177e4 LT |
454 | |
455 | if (++die.lock_owner_depth < 3) { | |
456 | int nl = 0; | |
7bee5c0f RD |
457 | unsigned long esp; |
458 | unsigned short ss; | |
459 | ||
1da177e4 | 460 | handle_BUG(regs); |
9c107805 | 461 | printk(KERN_EMERG "%s: %04lx [#%d]\n", str, err & 0xffff, ++die_counter); |
1da177e4 | 462 | #ifdef CONFIG_PREEMPT |
9c107805 | 463 | printk(KERN_EMERG "PREEMPT "); |
1da177e4 LT |
464 | nl = 1; |
465 | #endif | |
466 | #ifdef CONFIG_SMP | |
9c107805 DJ |
467 | if (!nl) |
468 | printk(KERN_EMERG); | |
1da177e4 LT |
469 | printk("SMP "); |
470 | nl = 1; | |
471 | #endif | |
472 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
9c107805 DJ |
473 | if (!nl) |
474 | printk(KERN_EMERG); | |
1da177e4 LT |
475 | printk("DEBUG_PAGEALLOC"); |
476 | nl = 1; | |
477 | #endif | |
478 | if (nl) | |
479 | printk("\n"); | |
20c0d2d4 JB |
480 | if (notify_die(DIE_OOPS, str, regs, err, |
481 | current->thread.trap_no, SIGSEGV) != | |
7bee5c0f | 482 | NOTIFY_STOP) { |
20c0d2d4 | 483 | show_registers(regs); |
7bee5c0f RD |
484 | /* Executive summary in case the oops scrolled away */ |
485 | esp = (unsigned long) (®s->esp); | |
486 | savesegment(ss, ss); | |
487 | if (user_mode(regs)) { | |
488 | esp = regs->esp; | |
489 | ss = regs->xss & 0xffff; | |
490 | } | |
491 | printk(KERN_EMERG "EIP: [<%08lx>] ", regs->eip); | |
492 | print_symbol("%s", regs->eip); | |
493 | printk(" SS:ESP %04x:%08lx\n", ss, esp); | |
494 | } | |
20c0d2d4 JB |
495 | else |
496 | regs = NULL; | |
1da177e4 | 497 | } else |
9c107805 | 498 | printk(KERN_EMERG "Recursive die() failure, output suppressed\n"); |
1da177e4 LT |
499 | |
500 | bust_spinlocks(0); | |
501 | die.lock_owner = -1; | |
e43d674f | 502 | spin_unlock_irqrestore(&die.lock, flags); |
6e274d14 | 503 | |
20c0d2d4 JB |
504 | if (!regs) |
505 | return; | |
506 | ||
6e274d14 AN |
507 | if (kexec_should_crash(current)) |
508 | crash_kexec(regs); | |
509 | ||
1da177e4 LT |
510 | if (in_interrupt()) |
511 | panic("Fatal exception in interrupt"); | |
512 | ||
cea6a4ba | 513 | if (panic_on_oops) |
012c437d | 514 | panic("Fatal exception"); |
cea6a4ba | 515 | |
dd287796 | 516 | oops_exit(); |
1da177e4 LT |
517 | do_exit(SIGSEGV); |
518 | } | |
519 | ||
520 | static inline void die_if_kernel(const char * str, struct pt_regs * regs, long err) | |
521 | { | |
717b594a | 522 | if (!user_mode_vm(regs)) |
1da177e4 LT |
523 | die(str, regs, err); |
524 | } | |
525 | ||
3d97ae5b PP |
526 | static void __kprobes do_trap(int trapnr, int signr, char *str, int vm86, |
527 | struct pt_regs * regs, long error_code, | |
528 | siginfo_t *info) | |
1da177e4 | 529 | { |
4f339ecb AN |
530 | struct task_struct *tsk = current; |
531 | tsk->thread.error_code = error_code; | |
532 | tsk->thread.trap_no = trapnr; | |
533 | ||
1da177e4 LT |
534 | if (regs->eflags & VM_MASK) { |
535 | if (vm86) | |
536 | goto vm86_trap; | |
537 | goto trap_signal; | |
538 | } | |
539 | ||
717b594a | 540 | if (!user_mode(regs)) |
1da177e4 LT |
541 | goto kernel_trap; |
542 | ||
543 | trap_signal: { | |
1da177e4 LT |
544 | if (info) |
545 | force_sig_info(signr, info, tsk); | |
546 | else | |
547 | force_sig(signr, tsk); | |
548 | return; | |
549 | } | |
550 | ||
551 | kernel_trap: { | |
552 | if (!fixup_exception(regs)) | |
553 | die(str, regs, error_code); | |
554 | return; | |
555 | } | |
556 | ||
557 | vm86_trap: { | |
558 | int ret = handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, trapnr); | |
559 | if (ret) goto trap_signal; | |
560 | return; | |
561 | } | |
562 | } | |
563 | ||
564 | #define DO_ERROR(trapnr, signr, str, name) \ | |
565 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
566 | { \ | |
567 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
568 | == NOTIFY_STOP) \ | |
569 | return; \ | |
570 | do_trap(trapnr, signr, str, 0, regs, error_code, NULL); \ | |
571 | } | |
572 | ||
573 | #define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ | |
574 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
575 | { \ | |
576 | siginfo_t info; \ | |
577 | info.si_signo = signr; \ | |
578 | info.si_errno = 0; \ | |
579 | info.si_code = sicode; \ | |
580 | info.si_addr = (void __user *)siaddr; \ | |
581 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
582 | == NOTIFY_STOP) \ | |
583 | return; \ | |
584 | do_trap(trapnr, signr, str, 0, regs, error_code, &info); \ | |
585 | } | |
586 | ||
587 | #define DO_VM86_ERROR(trapnr, signr, str, name) \ | |
588 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
589 | { \ | |
590 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
591 | == NOTIFY_STOP) \ | |
592 | return; \ | |
593 | do_trap(trapnr, signr, str, 1, regs, error_code, NULL); \ | |
594 | } | |
595 | ||
596 | #define DO_VM86_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ | |
597 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
598 | { \ | |
599 | siginfo_t info; \ | |
600 | info.si_signo = signr; \ | |
601 | info.si_errno = 0; \ | |
602 | info.si_code = sicode; \ | |
603 | info.si_addr = (void __user *)siaddr; \ | |
604 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
605 | == NOTIFY_STOP) \ | |
606 | return; \ | |
607 | do_trap(trapnr, signr, str, 1, regs, error_code, &info); \ | |
608 | } | |
609 | ||
610 | DO_VM86_ERROR_INFO( 0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->eip) | |
611 | #ifndef CONFIG_KPROBES | |
612 | DO_VM86_ERROR( 3, SIGTRAP, "int3", int3) | |
613 | #endif | |
614 | DO_VM86_ERROR( 4, SIGSEGV, "overflow", overflow) | |
615 | DO_VM86_ERROR( 5, SIGSEGV, "bounds", bounds) | |
631b0347 | 616 | DO_ERROR_INFO( 6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->eip) |
1da177e4 LT |
617 | DO_ERROR( 9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun) |
618 | DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS) | |
619 | DO_ERROR(11, SIGBUS, "segment not present", segment_not_present) | |
620 | DO_ERROR(12, SIGBUS, "stack segment", stack_segment) | |
621 | DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0) | |
a879cbbb | 622 | DO_ERROR_INFO(32, SIGSEGV, "iret exception", iret_error, ILL_BADSTK, 0) |
1da177e4 | 623 | |
3d97ae5b PP |
624 | fastcall void __kprobes do_general_protection(struct pt_regs * regs, |
625 | long error_code) | |
1da177e4 LT |
626 | { |
627 | int cpu = get_cpu(); | |
628 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
629 | struct thread_struct *thread = ¤t->thread; | |
630 | ||
631 | /* | |
632 | * Perform the lazy TSS's I/O bitmap copy. If the TSS has an | |
633 | * invalid offset set (the LAZY one) and the faulting thread has | |
634 | * a valid I/O bitmap pointer, we copy the I/O bitmap in the TSS | |
635 | * and we set the offset field correctly. Then we let the CPU to | |
636 | * restart the faulting instruction. | |
637 | */ | |
638 | if (tss->io_bitmap_base == INVALID_IO_BITMAP_OFFSET_LAZY && | |
639 | thread->io_bitmap_ptr) { | |
640 | memcpy(tss->io_bitmap, thread->io_bitmap_ptr, | |
641 | thread->io_bitmap_max); | |
642 | /* | |
643 | * If the previously set map was extending to higher ports | |
644 | * than the current one, pad extra space with 0xff (no access). | |
645 | */ | |
646 | if (thread->io_bitmap_max < tss->io_bitmap_max) | |
647 | memset((char *) tss->io_bitmap + | |
648 | thread->io_bitmap_max, 0xff, | |
649 | tss->io_bitmap_max - thread->io_bitmap_max); | |
650 | tss->io_bitmap_max = thread->io_bitmap_max; | |
651 | tss->io_bitmap_base = IO_BITMAP_OFFSET; | |
d5cd4aad | 652 | tss->io_bitmap_owner = thread; |
1da177e4 LT |
653 | put_cpu(); |
654 | return; | |
655 | } | |
656 | put_cpu(); | |
657 | ||
4f339ecb AN |
658 | current->thread.error_code = error_code; |
659 | current->thread.trap_no = 13; | |
660 | ||
1da177e4 LT |
661 | if (regs->eflags & VM_MASK) |
662 | goto gp_in_vm86; | |
663 | ||
717b594a | 664 | if (!user_mode(regs)) |
1da177e4 LT |
665 | goto gp_in_kernel; |
666 | ||
667 | current->thread.error_code = error_code; | |
668 | current->thread.trap_no = 13; | |
669 | force_sig(SIGSEGV, current); | |
670 | return; | |
671 | ||
672 | gp_in_vm86: | |
673 | local_irq_enable(); | |
674 | handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code); | |
675 | return; | |
676 | ||
677 | gp_in_kernel: | |
678 | if (!fixup_exception(regs)) { | |
679 | if (notify_die(DIE_GPF, "general protection fault", regs, | |
680 | error_code, 13, SIGSEGV) == NOTIFY_STOP) | |
681 | return; | |
682 | die("general protection fault", regs, error_code); | |
683 | } | |
684 | } | |
685 | ||
686 | static void mem_parity_error(unsigned char reason, struct pt_regs * regs) | |
687 | { | |
c41c5cd3 DZ |
688 | printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " |
689 | "CPU %d.\n", reason, smp_processor_id()); | |
9c107805 DJ |
690 | printk(KERN_EMERG "You probably have a hardware problem with your RAM " |
691 | "chips\n"); | |
8da5adda DZ |
692 | if (panic_on_unrecovered_nmi) |
693 | panic("NMI: Not continuing"); | |
1da177e4 | 694 | |
c41c5cd3 DZ |
695 | printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
696 | ||
1da177e4 LT |
697 | /* Clear and disable the memory parity error line. */ |
698 | clear_mem_error(reason); | |
699 | } | |
700 | ||
701 | static void io_check_error(unsigned char reason, struct pt_regs * regs) | |
702 | { | |
703 | unsigned long i; | |
704 | ||
9c107805 | 705 | printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); |
1da177e4 LT |
706 | show_registers(regs); |
707 | ||
708 | /* Re-enable the IOCK line, wait for a few seconds */ | |
709 | reason = (reason & 0xf) | 8; | |
710 | outb(reason, 0x61); | |
711 | i = 2000; | |
712 | while (--i) udelay(1000); | |
713 | reason &= ~8; | |
714 | outb(reason, 0x61); | |
715 | } | |
716 | ||
717 | static void unknown_nmi_error(unsigned char reason, struct pt_regs * regs) | |
718 | { | |
719 | #ifdef CONFIG_MCA | |
720 | /* Might actually be able to figure out what the guilty party | |
721 | * is. */ | |
722 | if( MCA_bus ) { | |
723 | mca_handle_nmi(); | |
724 | return; | |
725 | } | |
726 | #endif | |
c41c5cd3 DZ |
727 | printk(KERN_EMERG "Uhhuh. NMI received for unknown reason %02x on " |
728 | "CPU %d.\n", reason, smp_processor_id()); | |
729 | printk(KERN_EMERG "Do you have a strange power saving mode enabled?\n"); | |
8da5adda DZ |
730 | if (panic_on_unrecovered_nmi) |
731 | panic("NMI: Not continuing"); | |
732 | ||
c41c5cd3 | 733 | printk(KERN_EMERG "Dazed and confused, but trying to continue\n"); |
1da177e4 LT |
734 | } |
735 | ||
736 | static DEFINE_SPINLOCK(nmi_print_lock); | |
737 | ||
738 | void die_nmi (struct pt_regs *regs, const char *msg) | |
739 | { | |
20c0d2d4 | 740 | if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 2, SIGINT) == |
748f2edb GA |
741 | NOTIFY_STOP) |
742 | return; | |
743 | ||
1da177e4 LT |
744 | spin_lock(&nmi_print_lock); |
745 | /* | |
746 | * We are in trouble anyway, lets at least try | |
747 | * to get a message out. | |
748 | */ | |
749 | bust_spinlocks(1); | |
9c107805 | 750 | printk(KERN_EMERG "%s", msg); |
1da177e4 LT |
751 | printk(" on CPU%d, eip %08lx, registers:\n", |
752 | smp_processor_id(), regs->eip); | |
753 | show_registers(regs); | |
9c107805 | 754 | printk(KERN_EMERG "console shuts up ...\n"); |
1da177e4 LT |
755 | console_silent(); |
756 | spin_unlock(&nmi_print_lock); | |
757 | bust_spinlocks(0); | |
6e274d14 AN |
758 | |
759 | /* If we are in kernel we are probably nested up pretty bad | |
760 | * and might aswell get out now while we still can. | |
761 | */ | |
db753bdf | 762 | if (!user_mode_vm(regs)) { |
6e274d14 AN |
763 | current->thread.trap_no = 2; |
764 | crash_kexec(regs); | |
765 | } | |
766 | ||
1da177e4 LT |
767 | do_exit(SIGSEGV); |
768 | } | |
769 | ||
770 | static void default_do_nmi(struct pt_regs * regs) | |
771 | { | |
772 | unsigned char reason = 0; | |
773 | ||
774 | /* Only the BSP gets external NMIs from the system. */ | |
775 | if (!smp_processor_id()) | |
776 | reason = get_nmi_reason(); | |
777 | ||
778 | if (!(reason & 0xc0)) { | |
20c0d2d4 | 779 | if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 2, SIGINT) |
1da177e4 LT |
780 | == NOTIFY_STOP) |
781 | return; | |
782 | #ifdef CONFIG_X86_LOCAL_APIC | |
783 | /* | |
784 | * Ok, so this is none of the documented NMI sources, | |
785 | * so it must be the NMI watchdog. | |
786 | */ | |
3adbbcce | 787 | if (nmi_watchdog_tick(regs, reason)) |
1da177e4 | 788 | return; |
2fbe7b25 | 789 | if (!do_nmi_callback(regs, smp_processor_id())) |
1da177e4 | 790 | #endif |
3adbbcce | 791 | unknown_nmi_error(reason, regs); |
2fbe7b25 | 792 | |
1da177e4 LT |
793 | return; |
794 | } | |
20c0d2d4 | 795 | if (notify_die(DIE_NMI, "nmi", regs, reason, 2, SIGINT) == NOTIFY_STOP) |
1da177e4 LT |
796 | return; |
797 | if (reason & 0x80) | |
798 | mem_parity_error(reason, regs); | |
799 | if (reason & 0x40) | |
800 | io_check_error(reason, regs); | |
801 | /* | |
802 | * Reassert NMI in case it became active meanwhile | |
803 | * as it's edge-triggered. | |
804 | */ | |
805 | reassert_nmi(); | |
806 | } | |
807 | ||
1da177e4 LT |
808 | fastcall void do_nmi(struct pt_regs * regs, long error_code) |
809 | { | |
810 | int cpu; | |
811 | ||
812 | nmi_enter(); | |
813 | ||
814 | cpu = smp_processor_id(); | |
f3705136 | 815 | |
1da177e4 LT |
816 | ++nmi_count(cpu); |
817 | ||
3adbbcce | 818 | default_do_nmi(regs); |
1da177e4 LT |
819 | |
820 | nmi_exit(); | |
821 | } | |
822 | ||
1da177e4 | 823 | #ifdef CONFIG_KPROBES |
3d97ae5b | 824 | fastcall void __kprobes do_int3(struct pt_regs *regs, long error_code) |
1da177e4 LT |
825 | { |
826 | if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP) | |
827 | == NOTIFY_STOP) | |
48c88211 | 828 | return; |
1da177e4 LT |
829 | /* This is an interrupt gate, because kprobes wants interrupts |
830 | disabled. Normal trap handlers don't. */ | |
831 | restore_interrupts(regs); | |
832 | do_trap(3, SIGTRAP, "int3", 1, regs, error_code, NULL); | |
1da177e4 LT |
833 | } |
834 | #endif | |
835 | ||
836 | /* | |
837 | * Our handling of the processor debug registers is non-trivial. | |
838 | * We do not clear them on entry and exit from the kernel. Therefore | |
839 | * it is possible to get a watchpoint trap here from inside the kernel. | |
840 | * However, the code in ./ptrace.c has ensured that the user can | |
841 | * only set watchpoints on userspace addresses. Therefore the in-kernel | |
842 | * watchpoint trap can only occur in code which is reading/writing | |
843 | * from user space. Such code must not hold kernel locks (since it | |
844 | * can equally take a page fault), therefore it is safe to call | |
845 | * force_sig_info even though that claims and releases locks. | |
846 | * | |
847 | * Code in ./signal.c ensures that the debug control register | |
848 | * is restored before we deliver any signal, and therefore that | |
849 | * user code runs with the correct debug control register even though | |
850 | * we clear it here. | |
851 | * | |
852 | * Being careful here means that we don't have to be as careful in a | |
853 | * lot of more complicated places (task switching can be a bit lazy | |
854 | * about restoring all the debug state, and ptrace doesn't have to | |
855 | * find every occurrence of the TF bit that could be saved away even | |
856 | * by user code) | |
857 | */ | |
3d97ae5b | 858 | fastcall void __kprobes do_debug(struct pt_regs * regs, long error_code) |
1da177e4 LT |
859 | { |
860 | unsigned int condition; | |
861 | struct task_struct *tsk = current; | |
862 | ||
1cc6f12e | 863 | get_debugreg(condition, 6); |
1da177e4 LT |
864 | |
865 | if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code, | |
866 | SIGTRAP) == NOTIFY_STOP) | |
867 | return; | |
868 | /* It's safe to allow irq's after DR6 has been saved */ | |
869 | if (regs->eflags & X86_EFLAGS_IF) | |
870 | local_irq_enable(); | |
871 | ||
872 | /* Mask out spurious debug traps due to lazy DR7 setting */ | |
873 | if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) { | |
874 | if (!tsk->thread.debugreg[7]) | |
875 | goto clear_dr7; | |
876 | } | |
877 | ||
878 | if (regs->eflags & VM_MASK) | |
879 | goto debug_vm86; | |
880 | ||
881 | /* Save debug status register where ptrace can see it */ | |
882 | tsk->thread.debugreg[6] = condition; | |
883 | ||
884 | /* | |
885 | * Single-stepping through TF: make sure we ignore any events in | |
886 | * kernel space (but re-enable TF when returning to user mode). | |
887 | */ | |
888 | if (condition & DR_STEP) { | |
889 | /* | |
890 | * We already checked v86 mode above, so we can | |
891 | * check for kernel mode by just checking the CPL | |
892 | * of CS. | |
893 | */ | |
717b594a | 894 | if (!user_mode(regs)) |
1da177e4 LT |
895 | goto clear_TF_reenable; |
896 | } | |
897 | ||
898 | /* Ok, finally something we can handle */ | |
899 | send_sigtrap(tsk, regs, error_code); | |
900 | ||
901 | /* Disable additional traps. They'll be re-enabled when | |
902 | * the signal is delivered. | |
903 | */ | |
904 | clear_dr7: | |
1cc6f12e | 905 | set_debugreg(0, 7); |
1da177e4 LT |
906 | return; |
907 | ||
908 | debug_vm86: | |
909 | handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, 1); | |
910 | return; | |
911 | ||
912 | clear_TF_reenable: | |
913 | set_tsk_thread_flag(tsk, TIF_SINGLESTEP); | |
914 | regs->eflags &= ~TF_MASK; | |
915 | return; | |
916 | } | |
917 | ||
918 | /* | |
919 | * Note that we play around with the 'TS' bit in an attempt to get | |
920 | * the correct behaviour even in the presence of the asynchronous | |
921 | * IRQ13 behaviour | |
922 | */ | |
923 | void math_error(void __user *eip) | |
924 | { | |
925 | struct task_struct * task; | |
926 | siginfo_t info; | |
927 | unsigned short cwd, swd; | |
928 | ||
929 | /* | |
930 | * Save the info for the exception handler and clear the error. | |
931 | */ | |
932 | task = current; | |
933 | save_init_fpu(task); | |
934 | task->thread.trap_no = 16; | |
935 | task->thread.error_code = 0; | |
936 | info.si_signo = SIGFPE; | |
937 | info.si_errno = 0; | |
938 | info.si_code = __SI_FAULT; | |
939 | info.si_addr = eip; | |
940 | /* | |
941 | * (~cwd & swd) will mask out exceptions that are not set to unmasked | |
942 | * status. 0x3f is the exception bits in these regs, 0x200 is the | |
943 | * C1 reg you need in case of a stack fault, 0x040 is the stack | |
944 | * fault bit. We should only be taking one exception at a time, | |
945 | * so if this combination doesn't produce any single exception, | |
946 | * then we have a bad program that isn't syncronizing its FPU usage | |
947 | * and it will suffer the consequences since we won't be able to | |
948 | * fully reproduce the context of the exception | |
949 | */ | |
950 | cwd = get_fpu_cwd(task); | |
951 | swd = get_fpu_swd(task); | |
b1daec30 | 952 | switch (swd & ~cwd & 0x3f) { |
33333373 CE |
953 | case 0x000: /* No unmasked exception */ |
954 | return; | |
955 | default: /* Multiple exceptions */ | |
1da177e4 LT |
956 | break; |
957 | case 0x001: /* Invalid Op */ | |
b1daec30 CE |
958 | /* |
959 | * swd & 0x240 == 0x040: Stack Underflow | |
960 | * swd & 0x240 == 0x240: Stack Overflow | |
961 | * User must clear the SF bit (0x40) if set | |
962 | */ | |
1da177e4 | 963 | info.si_code = FPE_FLTINV; |
1da177e4 LT |
964 | break; |
965 | case 0x002: /* Denormalize */ | |
966 | case 0x010: /* Underflow */ | |
967 | info.si_code = FPE_FLTUND; | |
968 | break; | |
969 | case 0x004: /* Zero Divide */ | |
970 | info.si_code = FPE_FLTDIV; | |
971 | break; | |
972 | case 0x008: /* Overflow */ | |
973 | info.si_code = FPE_FLTOVF; | |
974 | break; | |
975 | case 0x020: /* Precision */ | |
976 | info.si_code = FPE_FLTRES; | |
977 | break; | |
978 | } | |
979 | force_sig_info(SIGFPE, &info, task); | |
980 | } | |
981 | ||
982 | fastcall void do_coprocessor_error(struct pt_regs * regs, long error_code) | |
983 | { | |
984 | ignore_fpu_irq = 1; | |
985 | math_error((void __user *)regs->eip); | |
986 | } | |
987 | ||
988 | static void simd_math_error(void __user *eip) | |
989 | { | |
990 | struct task_struct * task; | |
991 | siginfo_t info; | |
992 | unsigned short mxcsr; | |
993 | ||
994 | /* | |
995 | * Save the info for the exception handler and clear the error. | |
996 | */ | |
997 | task = current; | |
998 | save_init_fpu(task); | |
999 | task->thread.trap_no = 19; | |
1000 | task->thread.error_code = 0; | |
1001 | info.si_signo = SIGFPE; | |
1002 | info.si_errno = 0; | |
1003 | info.si_code = __SI_FAULT; | |
1004 | info.si_addr = eip; | |
1005 | /* | |
1006 | * The SIMD FPU exceptions are handled a little differently, as there | |
1007 | * is only a single status/control register. Thus, to determine which | |
1008 | * unmasked exception was caught we must mask the exception mask bits | |
1009 | * at 0x1f80, and then use these to mask the exception bits at 0x3f. | |
1010 | */ | |
1011 | mxcsr = get_fpu_mxcsr(task); | |
1012 | switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) { | |
1013 | case 0x000: | |
1014 | default: | |
1015 | break; | |
1016 | case 0x001: /* Invalid Op */ | |
1017 | info.si_code = FPE_FLTINV; | |
1018 | break; | |
1019 | case 0x002: /* Denormalize */ | |
1020 | case 0x010: /* Underflow */ | |
1021 | info.si_code = FPE_FLTUND; | |
1022 | break; | |
1023 | case 0x004: /* Zero Divide */ | |
1024 | info.si_code = FPE_FLTDIV; | |
1025 | break; | |
1026 | case 0x008: /* Overflow */ | |
1027 | info.si_code = FPE_FLTOVF; | |
1028 | break; | |
1029 | case 0x020: /* Precision */ | |
1030 | info.si_code = FPE_FLTRES; | |
1031 | break; | |
1032 | } | |
1033 | force_sig_info(SIGFPE, &info, task); | |
1034 | } | |
1035 | ||
1036 | fastcall void do_simd_coprocessor_error(struct pt_regs * regs, | |
1037 | long error_code) | |
1038 | { | |
1039 | if (cpu_has_xmm) { | |
1040 | /* Handle SIMD FPU exceptions on PIII+ processors. */ | |
1041 | ignore_fpu_irq = 1; | |
1042 | simd_math_error((void __user *)regs->eip); | |
1043 | } else { | |
1044 | /* | |
1045 | * Handle strange cache flush from user space exception | |
1046 | * in all other cases. This is undocumented behaviour. | |
1047 | */ | |
1048 | if (regs->eflags & VM_MASK) { | |
1049 | handle_vm86_fault((struct kernel_vm86_regs *)regs, | |
1050 | error_code); | |
1051 | return; | |
1052 | } | |
1da177e4 LT |
1053 | current->thread.trap_no = 19; |
1054 | current->thread.error_code = error_code; | |
4f339ecb | 1055 | die_if_kernel("cache flush denied", regs, error_code); |
1da177e4 LT |
1056 | force_sig(SIGSEGV, current); |
1057 | } | |
1058 | } | |
1059 | ||
1060 | fastcall void do_spurious_interrupt_bug(struct pt_regs * regs, | |
1061 | long error_code) | |
1062 | { | |
1063 | #if 0 | |
1064 | /* No need to warn about this any longer. */ | |
1065 | printk("Ignoring P6 Local APIC Spurious Interrupt Bug...\n"); | |
1066 | #endif | |
1067 | } | |
1068 | ||
1069 | fastcall void setup_x86_bogus_stack(unsigned char * stk) | |
1070 | { | |
1071 | unsigned long *switch16_ptr, *switch32_ptr; | |
1072 | struct pt_regs *regs; | |
1073 | unsigned long stack_top, stack_bot; | |
1074 | unsigned short iret_frame16_off; | |
1075 | int cpu = smp_processor_id(); | |
1076 | /* reserve the space on 32bit stack for the magic switch16 pointer */ | |
1077 | memmove(stk, stk + 8, sizeof(struct pt_regs)); | |
1078 | switch16_ptr = (unsigned long *)(stk + sizeof(struct pt_regs)); | |
1079 | regs = (struct pt_regs *)stk; | |
1080 | /* now the switch32 on 16bit stack */ | |
1081 | stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); | |
1082 | stack_top = stack_bot + CPU_16BIT_STACK_SIZE; | |
1083 | switch32_ptr = (unsigned long *)(stack_top - 8); | |
1084 | iret_frame16_off = CPU_16BIT_STACK_SIZE - 8 - 20; | |
1085 | /* copy iret frame on 16bit stack */ | |
1086 | memcpy((void *)(stack_bot + iret_frame16_off), ®s->eip, 20); | |
1087 | /* fill in the switch pointers */ | |
1088 | switch16_ptr[0] = (regs->esp & 0xffff0000) | iret_frame16_off; | |
1089 | switch16_ptr[1] = __ESPFIX_SS; | |
1090 | switch32_ptr[0] = (unsigned long)stk + sizeof(struct pt_regs) + | |
1091 | 8 - CPU_16BIT_STACK_SIZE; | |
1092 | switch32_ptr[1] = __KERNEL_DS; | |
1093 | } | |
1094 | ||
1095 | fastcall unsigned char * fixup_x86_bogus_stack(unsigned short sp) | |
1096 | { | |
1097 | unsigned long *switch32_ptr; | |
1098 | unsigned char *stack16, *stack32; | |
1099 | unsigned long stack_top, stack_bot; | |
1100 | int len; | |
1101 | int cpu = smp_processor_id(); | |
1102 | stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); | |
1103 | stack_top = stack_bot + CPU_16BIT_STACK_SIZE; | |
1104 | switch32_ptr = (unsigned long *)(stack_top - 8); | |
1105 | /* copy the data from 16bit stack to 32bit stack */ | |
1106 | len = CPU_16BIT_STACK_SIZE - 8 - sp; | |
1107 | stack16 = (unsigned char *)(stack_bot + sp); | |
1108 | stack32 = (unsigned char *) | |
1109 | (switch32_ptr[0] + CPU_16BIT_STACK_SIZE - 8 - len); | |
1110 | memcpy(stack32, stack16, len); | |
1111 | return stack32; | |
1112 | } | |
1113 | ||
1114 | /* | |
1115 | * 'math_state_restore()' saves the current math information in the | |
1116 | * old math state array, and gets the new ones from the current task | |
1117 | * | |
1118 | * Careful.. There are problems with IBM-designed IRQ13 behaviour. | |
1119 | * Don't touch unless you *really* know how it works. | |
1120 | * | |
1121 | * Must be called with kernel preemption disabled (in this case, | |
1122 | * local interrupts are disabled at the call-site in entry.S). | |
1123 | */ | |
1124 | asmlinkage void math_state_restore(struct pt_regs regs) | |
1125 | { | |
1126 | struct thread_info *thread = current_thread_info(); | |
1127 | struct task_struct *tsk = thread->task; | |
1128 | ||
1129 | clts(); /* Allow maths ops (or we recurse) */ | |
1130 | if (!tsk_used_math(tsk)) | |
1131 | init_fpu(tsk); | |
1132 | restore_fpu(tsk); | |
1133 | thread->status |= TS_USEDFPU; /* So we fnsave on switch_to() */ | |
1134 | } | |
1135 | ||
1136 | #ifndef CONFIG_MATH_EMULATION | |
1137 | ||
1138 | asmlinkage void math_emulate(long arg) | |
1139 | { | |
9c107805 DJ |
1140 | printk(KERN_EMERG "math-emulation not enabled and no coprocessor found.\n"); |
1141 | printk(KERN_EMERG "killing %s.\n",current->comm); | |
1da177e4 LT |
1142 | force_sig(SIGFPE,current); |
1143 | schedule(); | |
1144 | } | |
1145 | ||
1146 | #endif /* CONFIG_MATH_EMULATION */ | |
1147 | ||
1148 | #ifdef CONFIG_X86_F00F_BUG | |
1149 | void __init trap_init_f00f_bug(void) | |
1150 | { | |
1151 | __set_fixmap(FIX_F00F_IDT, __pa(&idt_table), PAGE_KERNEL_RO); | |
1152 | ||
1153 | /* | |
1154 | * Update the IDT descriptor and reload the IDT so that | |
1155 | * it uses the read-only mapped virtual address. | |
1156 | */ | |
1157 | idt_descr.address = fix_to_virt(FIX_F00F_IDT); | |
4d37e7e3 | 1158 | load_idt(&idt_descr); |
1da177e4 LT |
1159 | } |
1160 | #endif | |
1161 | ||
1162 | #define _set_gate(gate_addr,type,dpl,addr,seg) \ | |
1163 | do { \ | |
1164 | int __d0, __d1; \ | |
1165 | __asm__ __volatile__ ("movw %%dx,%%ax\n\t" \ | |
1166 | "movw %4,%%dx\n\t" \ | |
1167 | "movl %%eax,%0\n\t" \ | |
1168 | "movl %%edx,%1" \ | |
1169 | :"=m" (*((long *) (gate_addr))), \ | |
1170 | "=m" (*(1+(long *) (gate_addr))), "=&a" (__d0), "=&d" (__d1) \ | |
1171 | :"i" ((short) (0x8000+(dpl<<13)+(type<<8))), \ | |
1172 | "3" ((char *) (addr)),"2" ((seg) << 16)); \ | |
1173 | } while (0) | |
1174 | ||
1175 | ||
1176 | /* | |
1177 | * This needs to use 'idt_table' rather than 'idt', and | |
1178 | * thus use the _nonmapped_ version of the IDT, as the | |
1179 | * Pentium F0 0F bugfix can have resulted in the mapped | |
1180 | * IDT being write-protected. | |
1181 | */ | |
1182 | void set_intr_gate(unsigned int n, void *addr) | |
1183 | { | |
1184 | _set_gate(idt_table+n,14,0,addr,__KERNEL_CS); | |
1185 | } | |
1186 | ||
1187 | /* | |
1188 | * This routine sets up an interrupt gate at directory privilege level 3. | |
1189 | */ | |
1190 | static inline void set_system_intr_gate(unsigned int n, void *addr) | |
1191 | { | |
1192 | _set_gate(idt_table+n, 14, 3, addr, __KERNEL_CS); | |
1193 | } | |
1194 | ||
1195 | static void __init set_trap_gate(unsigned int n, void *addr) | |
1196 | { | |
1197 | _set_gate(idt_table+n,15,0,addr,__KERNEL_CS); | |
1198 | } | |
1199 | ||
1200 | static void __init set_system_gate(unsigned int n, void *addr) | |
1201 | { | |
1202 | _set_gate(idt_table+n,15,3,addr,__KERNEL_CS); | |
1203 | } | |
1204 | ||
1205 | static void __init set_task_gate(unsigned int n, unsigned int gdt_entry) | |
1206 | { | |
1207 | _set_gate(idt_table+n,5,0,0,(gdt_entry<<3)); | |
1208 | } | |
1209 | ||
1210 | ||
1211 | void __init trap_init(void) | |
1212 | { | |
1213 | #ifdef CONFIG_EISA | |
1214 | void __iomem *p = ioremap(0x0FFFD9, 4); | |
1215 | if (readl(p) == 'E'+('I'<<8)+('S'<<16)+('A'<<24)) { | |
1216 | EISA_bus = 1; | |
1217 | } | |
1218 | iounmap(p); | |
1219 | #endif | |
1220 | ||
1221 | #ifdef CONFIG_X86_LOCAL_APIC | |
1222 | init_apic_mappings(); | |
1223 | #endif | |
1224 | ||
1225 | set_trap_gate(0,÷_error); | |
1226 | set_intr_gate(1,&debug); | |
1227 | set_intr_gate(2,&nmi); | |
eb05c324 | 1228 | set_system_intr_gate(3, &int3); /* int3/4 can be called from all */ |
1da177e4 | 1229 | set_system_gate(4,&overflow); |
eb05c324 | 1230 | set_trap_gate(5,&bounds); |
1da177e4 LT |
1231 | set_trap_gate(6,&invalid_op); |
1232 | set_trap_gate(7,&device_not_available); | |
1233 | set_task_gate(8,GDT_ENTRY_DOUBLEFAULT_TSS); | |
1234 | set_trap_gate(9,&coprocessor_segment_overrun); | |
1235 | set_trap_gate(10,&invalid_TSS); | |
1236 | set_trap_gate(11,&segment_not_present); | |
1237 | set_trap_gate(12,&stack_segment); | |
1238 | set_trap_gate(13,&general_protection); | |
1239 | set_intr_gate(14,&page_fault); | |
1240 | set_trap_gate(15,&spurious_interrupt_bug); | |
1241 | set_trap_gate(16,&coprocessor_error); | |
1242 | set_trap_gate(17,&alignment_check); | |
1243 | #ifdef CONFIG_X86_MCE | |
1244 | set_trap_gate(18,&machine_check); | |
1245 | #endif | |
1246 | set_trap_gate(19,&simd_coprocessor_error); | |
1247 | ||
d43c6e80 JB |
1248 | if (cpu_has_fxsr) { |
1249 | /* | |
1250 | * Verify that the FXSAVE/FXRSTOR data will be 16-byte aligned. | |
1251 | * Generates a compile-time "error: zero width for bit-field" if | |
1252 | * the alignment is wrong. | |
1253 | */ | |
1254 | struct fxsrAlignAssert { | |
1255 | int _:!(offsetof(struct task_struct, | |
1256 | thread.i387.fxsave) & 15); | |
1257 | }; | |
1258 | ||
1259 | printk(KERN_INFO "Enabling fast FPU save and restore... "); | |
1260 | set_in_cr4(X86_CR4_OSFXSR); | |
1261 | printk("done.\n"); | |
1262 | } | |
1263 | if (cpu_has_xmm) { | |
1264 | printk(KERN_INFO "Enabling unmasked SIMD FPU exception " | |
1265 | "support... "); | |
1266 | set_in_cr4(X86_CR4_OSXMMEXCPT); | |
1267 | printk("done.\n"); | |
1268 | } | |
1269 | ||
1da177e4 LT |
1270 | set_system_gate(SYSCALL_VECTOR,&system_call); |
1271 | ||
1272 | /* | |
1273 | * Should be a barrier for any external CPU state. | |
1274 | */ | |
1275 | cpu_init(); | |
1276 | ||
1277 | trap_init_hook(); | |
1278 | } | |
1279 | ||
1280 | static int __init kstack_setup(char *s) | |
1281 | { | |
1282 | kstack_depth_to_print = simple_strtoul(s, NULL, 0); | |
9b41046c | 1283 | return 1; |
1da177e4 LT |
1284 | } |
1285 | __setup("kstack=", kstack_setup); | |
c33bd9aa | 1286 | |
ea424055 | 1287 | #ifdef CONFIG_STACK_UNWIND |
c33bd9aa JB |
1288 | static int __init call_trace_setup(char *s) |
1289 | { | |
1290 | if (strcmp(s, "old") == 0) | |
1291 | call_trace = -1; | |
1292 | else if (strcmp(s, "both") == 0) | |
1293 | call_trace = 0; | |
70583161 | 1294 | else if (strcmp(s, "newfallback") == 0) |
c33bd9aa | 1295 | call_trace = 1; |
70583161 AK |
1296 | else if (strcmp(s, "new") == 2) |
1297 | call_trace = 2; | |
c33bd9aa JB |
1298 | return 1; |
1299 | } | |
1300 | __setup("call_trace=", call_trace_setup); | |
ea424055 | 1301 | #endif |