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 | */ | |
14 | #include <linux/config.h> | |
15 | #include <linux/sched.h> | |
16 | #include <linux/kernel.h> | |
17 | #include <linux/string.h> | |
18 | #include <linux/errno.h> | |
19 | #include <linux/timer.h> | |
20 | #include <linux/mm.h> | |
21 | #include <linux/init.h> | |
22 | #include <linux/delay.h> | |
23 | #include <linux/spinlock.h> | |
24 | #include <linux/interrupt.h> | |
25 | #include <linux/highmem.h> | |
26 | #include <linux/kallsyms.h> | |
27 | #include <linux/ptrace.h> | |
28 | #include <linux/utsname.h> | |
29 | #include <linux/kprobes.h> | |
6e274d14 | 30 | #include <linux/kexec.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> | |
50 | ||
51 | #include <asm/smp.h> | |
52 | #include <asm/arch_hooks.h> | |
53 | #include <asm/kdebug.h> | |
54 | ||
1da177e4 LT |
55 | #include <linux/module.h> |
56 | ||
57 | #include "mach_traps.h" | |
58 | ||
59 | asmlinkage int system_call(void); | |
60 | ||
61 | struct desc_struct default_ldt[] = { { 0, 0 }, { 0, 0 }, { 0, 0 }, | |
62 | { 0, 0 }, { 0, 0 } }; | |
63 | ||
64 | /* Do we ignore FPU interrupts ? */ | |
65 | char ignore_fpu_irq = 0; | |
66 | ||
67 | /* | |
68 | * The IDT has to be page-aligned to simplify the Pentium | |
69 | * F0 0F bug workaround.. We have a special link segment | |
70 | * for this. | |
71 | */ | |
72 | struct desc_struct idt_table[256] __attribute__((__section__(".data.idt"))) = { {0, 0}, }; | |
73 | ||
74 | asmlinkage void divide_error(void); | |
75 | asmlinkage void debug(void); | |
76 | asmlinkage void nmi(void); | |
77 | asmlinkage void int3(void); | |
78 | asmlinkage void overflow(void); | |
79 | asmlinkage void bounds(void); | |
80 | asmlinkage void invalid_op(void); | |
81 | asmlinkage void device_not_available(void); | |
82 | asmlinkage void coprocessor_segment_overrun(void); | |
83 | asmlinkage void invalid_TSS(void); | |
84 | asmlinkage void segment_not_present(void); | |
85 | asmlinkage void stack_segment(void); | |
86 | asmlinkage void general_protection(void); | |
87 | asmlinkage void page_fault(void); | |
88 | asmlinkage void coprocessor_error(void); | |
89 | asmlinkage void simd_coprocessor_error(void); | |
90 | asmlinkage void alignment_check(void); | |
91 | asmlinkage void spurious_interrupt_bug(void); | |
92 | asmlinkage void machine_check(void); | |
93 | ||
94 | static int kstack_depth_to_print = 24; | |
95 | struct notifier_block *i386die_chain; | |
96 | static DEFINE_SPINLOCK(die_notifier_lock); | |
97 | ||
98 | int register_die_notifier(struct notifier_block *nb) | |
99 | { | |
100 | int err = 0; | |
101 | unsigned long flags; | |
101f12af JB |
102 | |
103 | vmalloc_sync_all(); | |
1da177e4 LT |
104 | spin_lock_irqsave(&die_notifier_lock, flags); |
105 | err = notifier_chain_register(&i386die_chain, nb); | |
106 | spin_unlock_irqrestore(&die_notifier_lock, flags); | |
107 | return err; | |
108 | } | |
129f6946 | 109 | EXPORT_SYMBOL(register_die_notifier); |
1da177e4 LT |
110 | |
111 | static inline int valid_stack_ptr(struct thread_info *tinfo, void *p) | |
112 | { | |
113 | return p > (void *)tinfo && | |
114 | p < (void *)tinfo + THREAD_SIZE - 3; | |
115 | } | |
116 | ||
4d7d8c82 CE |
117 | /* |
118 | * Print CONFIG_STACK_BACKTRACE_COLS address/symbol entries per line. | |
119 | */ | |
120 | static inline int print_addr_and_symbol(unsigned long addr, char *log_lvl, | |
121 | int printed) | |
7aa89746 | 122 | { |
4d7d8c82 CE |
123 | if (!printed) |
124 | printk(log_lvl); | |
125 | ||
126 | #if CONFIG_STACK_BACKTRACE_COLS == 1 | |
7aa89746 | 127 | printk(" [<%08lx>] ", addr); |
4d7d8c82 CE |
128 | #else |
129 | printk(" <%08lx> ", addr); | |
130 | #endif | |
7aa89746 | 131 | print_symbol("%s", addr); |
4d7d8c82 CE |
132 | |
133 | printed = (printed + 1) % CONFIG_STACK_BACKTRACE_COLS; | |
134 | ||
135 | if (printed) | |
136 | printk(" "); | |
137 | else | |
138 | printk("\n"); | |
139 | ||
140 | return printed; | |
7aa89746 CE |
141 | } |
142 | ||
1da177e4 | 143 | static inline unsigned long print_context_stack(struct thread_info *tinfo, |
7aa89746 CE |
144 | unsigned long *stack, unsigned long ebp, |
145 | char *log_lvl) | |
1da177e4 LT |
146 | { |
147 | unsigned long addr; | |
4d7d8c82 | 148 | int printed = 0; /* nr of entries already printed on current line */ |
1da177e4 LT |
149 | |
150 | #ifdef CONFIG_FRAME_POINTER | |
151 | while (valid_stack_ptr(tinfo, (void *)ebp)) { | |
152 | addr = *(unsigned long *)(ebp + 4); | |
4d7d8c82 | 153 | printed = print_addr_and_symbol(addr, log_lvl, printed); |
1da177e4 LT |
154 | ebp = *(unsigned long *)ebp; |
155 | } | |
156 | #else | |
157 | while (valid_stack_ptr(tinfo, stack)) { | |
158 | addr = *stack++; | |
7aa89746 | 159 | if (__kernel_text_address(addr)) |
4d7d8c82 | 160 | printed = print_addr_and_symbol(addr, log_lvl, printed); |
1da177e4 LT |
161 | } |
162 | #endif | |
4d7d8c82 CE |
163 | if (printed) |
164 | printk("\n"); | |
165 | ||
1da177e4 LT |
166 | return ebp; |
167 | } | |
168 | ||
7aa89746 CE |
169 | static void show_trace_log_lvl(struct task_struct *task, |
170 | unsigned long *stack, char *log_lvl) | |
1da177e4 LT |
171 | { |
172 | unsigned long ebp; | |
173 | ||
174 | if (!task) | |
175 | task = current; | |
176 | ||
177 | if (task == current) { | |
178 | /* Grab ebp right from our regs */ | |
179 | asm ("movl %%ebp, %0" : "=r" (ebp) : ); | |
180 | } else { | |
181 | /* ebp is the last reg pushed by switch_to */ | |
182 | ebp = *(unsigned long *) task->thread.esp; | |
183 | } | |
184 | ||
185 | while (1) { | |
186 | struct thread_info *context; | |
187 | context = (struct thread_info *) | |
188 | ((unsigned long)stack & (~(THREAD_SIZE - 1))); | |
7aa89746 | 189 | ebp = print_context_stack(context, stack, ebp, log_lvl); |
1da177e4 LT |
190 | stack = (unsigned long*)context->previous_esp; |
191 | if (!stack) | |
192 | break; | |
cc04ee9c | 193 | printk("%s =======================\n", log_lvl); |
1da177e4 LT |
194 | } |
195 | } | |
196 | ||
7aa89746 CE |
197 | void show_trace(struct task_struct *task, unsigned long * stack) |
198 | { | |
199 | show_trace_log_lvl(task, stack, ""); | |
200 | } | |
201 | ||
202 | static void show_stack_log_lvl(struct task_struct *task, unsigned long *esp, | |
203 | char *log_lvl) | |
1da177e4 LT |
204 | { |
205 | unsigned long *stack; | |
206 | int i; | |
207 | ||
208 | if (esp == NULL) { | |
209 | if (task) | |
210 | esp = (unsigned long*)task->thread.esp; | |
211 | else | |
212 | esp = (unsigned long *)&esp; | |
213 | } | |
214 | ||
215 | stack = esp; | |
7aa89746 | 216 | printk(log_lvl); |
1da177e4 LT |
217 | for(i = 0; i < kstack_depth_to_print; i++) { |
218 | if (kstack_end(stack)) | |
219 | break; | |
7aa89746 CE |
220 | if (i && ((i % 8) == 0)) { |
221 | printk("\n"); | |
cc04ee9c | 222 | printk("%s ", log_lvl); |
7aa89746 | 223 | } |
1da177e4 LT |
224 | printk("%08lx ", *stack++); |
225 | } | |
7aa89746 | 226 | printk("\n"); |
cc04ee9c | 227 | printk("%sCall Trace:\n", log_lvl); |
7aa89746 CE |
228 | show_trace_log_lvl(task, esp, log_lvl); |
229 | } | |
230 | ||
231 | void show_stack(struct task_struct *task, unsigned long *esp) | |
232 | { | |
233 | show_stack_log_lvl(task, esp, ""); | |
1da177e4 LT |
234 | } |
235 | ||
236 | /* | |
237 | * The architecture-independent dump_stack generator | |
238 | */ | |
239 | void dump_stack(void) | |
240 | { | |
241 | unsigned long stack; | |
242 | ||
243 | show_trace(current, &stack); | |
244 | } | |
245 | ||
246 | EXPORT_SYMBOL(dump_stack); | |
247 | ||
248 | void show_registers(struct pt_regs *regs) | |
249 | { | |
250 | int i; | |
251 | int in_kernel = 1; | |
252 | unsigned long esp; | |
253 | unsigned short ss; | |
254 | ||
255 | esp = (unsigned long) (®s->esp); | |
0998e422 | 256 | savesegment(ss, ss); |
db753bdf | 257 | if (user_mode_vm(regs)) { |
1da177e4 LT |
258 | in_kernel = 0; |
259 | esp = regs->esp; | |
260 | ss = regs->xss & 0xffff; | |
261 | } | |
262 | print_modules(); | |
9c107805 | 263 | printk(KERN_EMERG "CPU: %d\nEIP: %04x:[<%08lx>] %s VLI\n" |
b53e8f68 | 264 | "EFLAGS: %08lx (%s %.*s) \n", |
1da177e4 | 265 | smp_processor_id(), 0xffff & regs->xcs, regs->eip, |
b53e8f68 CE |
266 | print_tainted(), regs->eflags, system_utsname.release, |
267 | (int)strcspn(system_utsname.version, " "), | |
268 | system_utsname.version); | |
9c107805 DJ |
269 | print_symbol(KERN_EMERG "EIP is at %s\n", regs->eip); |
270 | printk(KERN_EMERG "eax: %08lx ebx: %08lx ecx: %08lx edx: %08lx\n", | |
1da177e4 | 271 | regs->eax, regs->ebx, regs->ecx, regs->edx); |
9c107805 | 272 | printk(KERN_EMERG "esi: %08lx edi: %08lx ebp: %08lx esp: %08lx\n", |
1da177e4 | 273 | regs->esi, regs->edi, regs->ebp, esp); |
9c107805 | 274 | printk(KERN_EMERG "ds: %04x es: %04x ss: %04x\n", |
1da177e4 | 275 | regs->xds & 0xffff, regs->xes & 0xffff, ss); |
9c107805 | 276 | printk(KERN_EMERG "Process %s (pid: %d, threadinfo=%p task=%p)", |
1da177e4 LT |
277 | current->comm, current->pid, current_thread_info(), current); |
278 | /* | |
279 | * When in-kernel, we also print out the stack and code at the | |
280 | * time of the fault.. | |
281 | */ | |
282 | if (in_kernel) { | |
3f3ae347 | 283 | u8 __user *eip; |
1da177e4 | 284 | |
9c107805 | 285 | printk("\n" KERN_EMERG "Stack: "); |
7aa89746 | 286 | show_stack_log_lvl(NULL, (unsigned long *)esp, KERN_EMERG); |
1da177e4 | 287 | |
9c107805 | 288 | printk(KERN_EMERG "Code: "); |
1da177e4 | 289 | |
3f3ae347 | 290 | eip = (u8 __user *)regs->eip - 43; |
1da177e4 LT |
291 | for (i = 0; i < 64; i++, eip++) { |
292 | unsigned char c; | |
293 | ||
3f3ae347 | 294 | if (eip < (u8 __user *)PAGE_OFFSET || __get_user(c, eip)) { |
1da177e4 LT |
295 | printk(" Bad EIP value."); |
296 | break; | |
297 | } | |
3f3ae347 | 298 | if (eip == (u8 __user *)regs->eip) |
1da177e4 LT |
299 | printk("<%02x> ", c); |
300 | else | |
301 | printk("%02x ", c); | |
302 | } | |
303 | } | |
304 | printk("\n"); | |
305 | } | |
306 | ||
307 | static void handle_BUG(struct pt_regs *regs) | |
308 | { | |
309 | unsigned short ud2; | |
310 | unsigned short line; | |
311 | char *file; | |
312 | char c; | |
313 | unsigned long eip; | |
314 | ||
1da177e4 LT |
315 | eip = regs->eip; |
316 | ||
317 | if (eip < PAGE_OFFSET) | |
318 | goto no_bug; | |
3f3ae347 | 319 | if (__get_user(ud2, (unsigned short __user *)eip)) |
1da177e4 LT |
320 | goto no_bug; |
321 | if (ud2 != 0x0b0f) | |
322 | goto no_bug; | |
3f3ae347 | 323 | if (__get_user(line, (unsigned short __user *)(eip + 2))) |
1da177e4 | 324 | goto bug; |
3f3ae347 | 325 | if (__get_user(file, (char * __user *)(eip + 4)) || |
1da177e4 LT |
326 | (unsigned long)file < PAGE_OFFSET || __get_user(c, file)) |
327 | file = "<bad filename>"; | |
328 | ||
9c107805 DJ |
329 | printk(KERN_EMERG "------------[ cut here ]------------\n"); |
330 | printk(KERN_EMERG "kernel BUG at %s:%d!\n", file, line); | |
1da177e4 LT |
331 | |
332 | no_bug: | |
333 | return; | |
334 | ||
335 | /* Here we know it was a BUG but file-n-line is unavailable */ | |
336 | bug: | |
9c107805 | 337 | printk(KERN_EMERG "Kernel BUG\n"); |
1da177e4 LT |
338 | } |
339 | ||
6e274d14 AN |
340 | /* This is gone through when something in the kernel |
341 | * has done something bad and is about to be terminated. | |
342 | */ | |
1da177e4 LT |
343 | void die(const char * str, struct pt_regs * regs, long err) |
344 | { | |
345 | static struct { | |
346 | spinlock_t lock; | |
347 | u32 lock_owner; | |
348 | int lock_owner_depth; | |
349 | } die = { | |
350 | .lock = SPIN_LOCK_UNLOCKED, | |
351 | .lock_owner = -1, | |
352 | .lock_owner_depth = 0 | |
353 | }; | |
354 | static int die_counter; | |
e43d674f | 355 | unsigned long flags; |
1da177e4 | 356 | |
39c715b7 | 357 | if (die.lock_owner != raw_smp_processor_id()) { |
1da177e4 | 358 | console_verbose(); |
e43d674f | 359 | spin_lock_irqsave(&die.lock, flags); |
1da177e4 LT |
360 | die.lock_owner = smp_processor_id(); |
361 | die.lock_owner_depth = 0; | |
362 | bust_spinlocks(1); | |
363 | } | |
e43d674f JB |
364 | else |
365 | local_save_flags(flags); | |
1da177e4 LT |
366 | |
367 | if (++die.lock_owner_depth < 3) { | |
368 | int nl = 0; | |
369 | handle_BUG(regs); | |
9c107805 | 370 | printk(KERN_EMERG "%s: %04lx [#%d]\n", str, err & 0xffff, ++die_counter); |
1da177e4 | 371 | #ifdef CONFIG_PREEMPT |
9c107805 | 372 | printk(KERN_EMERG "PREEMPT "); |
1da177e4 LT |
373 | nl = 1; |
374 | #endif | |
375 | #ifdef CONFIG_SMP | |
9c107805 DJ |
376 | if (!nl) |
377 | printk(KERN_EMERG); | |
1da177e4 LT |
378 | printk("SMP "); |
379 | nl = 1; | |
380 | #endif | |
381 | #ifdef CONFIG_DEBUG_PAGEALLOC | |
9c107805 DJ |
382 | if (!nl) |
383 | printk(KERN_EMERG); | |
1da177e4 LT |
384 | printk("DEBUG_PAGEALLOC"); |
385 | nl = 1; | |
386 | #endif | |
387 | if (nl) | |
388 | printk("\n"); | |
389 | notify_die(DIE_OOPS, (char *)str, regs, err, 255, SIGSEGV); | |
390 | show_registers(regs); | |
391 | } else | |
9c107805 | 392 | printk(KERN_EMERG "Recursive die() failure, output suppressed\n"); |
1da177e4 LT |
393 | |
394 | bust_spinlocks(0); | |
395 | die.lock_owner = -1; | |
e43d674f | 396 | spin_unlock_irqrestore(&die.lock, flags); |
6e274d14 AN |
397 | |
398 | if (kexec_should_crash(current)) | |
399 | crash_kexec(regs); | |
400 | ||
1da177e4 LT |
401 | if (in_interrupt()) |
402 | panic("Fatal exception in interrupt"); | |
403 | ||
404 | if (panic_on_oops) { | |
405 | printk(KERN_EMERG "Fatal exception: panic in 5 seconds\n"); | |
406 | ssleep(5); | |
407 | panic("Fatal exception"); | |
408 | } | |
409 | do_exit(SIGSEGV); | |
410 | } | |
411 | ||
412 | static inline void die_if_kernel(const char * str, struct pt_regs * regs, long err) | |
413 | { | |
717b594a | 414 | if (!user_mode_vm(regs)) |
1da177e4 LT |
415 | die(str, regs, err); |
416 | } | |
417 | ||
3d97ae5b PP |
418 | static void __kprobes do_trap(int trapnr, int signr, char *str, int vm86, |
419 | struct pt_regs * regs, long error_code, | |
420 | siginfo_t *info) | |
1da177e4 | 421 | { |
4f339ecb AN |
422 | struct task_struct *tsk = current; |
423 | tsk->thread.error_code = error_code; | |
424 | tsk->thread.trap_no = trapnr; | |
425 | ||
1da177e4 LT |
426 | if (regs->eflags & VM_MASK) { |
427 | if (vm86) | |
428 | goto vm86_trap; | |
429 | goto trap_signal; | |
430 | } | |
431 | ||
717b594a | 432 | if (!user_mode(regs)) |
1da177e4 LT |
433 | goto kernel_trap; |
434 | ||
435 | trap_signal: { | |
1da177e4 LT |
436 | if (info) |
437 | force_sig_info(signr, info, tsk); | |
438 | else | |
439 | force_sig(signr, tsk); | |
440 | return; | |
441 | } | |
442 | ||
443 | kernel_trap: { | |
444 | if (!fixup_exception(regs)) | |
445 | die(str, regs, error_code); | |
446 | return; | |
447 | } | |
448 | ||
449 | vm86_trap: { | |
450 | int ret = handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, trapnr); | |
451 | if (ret) goto trap_signal; | |
452 | return; | |
453 | } | |
454 | } | |
455 | ||
456 | #define DO_ERROR(trapnr, signr, str, name) \ | |
457 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
458 | { \ | |
459 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
460 | == NOTIFY_STOP) \ | |
461 | return; \ | |
462 | do_trap(trapnr, signr, str, 0, regs, error_code, NULL); \ | |
463 | } | |
464 | ||
465 | #define DO_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ | |
466 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
467 | { \ | |
468 | siginfo_t info; \ | |
469 | info.si_signo = signr; \ | |
470 | info.si_errno = 0; \ | |
471 | info.si_code = sicode; \ | |
472 | info.si_addr = (void __user *)siaddr; \ | |
473 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
474 | == NOTIFY_STOP) \ | |
475 | return; \ | |
476 | do_trap(trapnr, signr, str, 0, regs, error_code, &info); \ | |
477 | } | |
478 | ||
479 | #define DO_VM86_ERROR(trapnr, signr, str, name) \ | |
480 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
481 | { \ | |
482 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
483 | == NOTIFY_STOP) \ | |
484 | return; \ | |
485 | do_trap(trapnr, signr, str, 1, regs, error_code, NULL); \ | |
486 | } | |
487 | ||
488 | #define DO_VM86_ERROR_INFO(trapnr, signr, str, name, sicode, siaddr) \ | |
489 | fastcall void do_##name(struct pt_regs * regs, long error_code) \ | |
490 | { \ | |
491 | siginfo_t info; \ | |
492 | info.si_signo = signr; \ | |
493 | info.si_errno = 0; \ | |
494 | info.si_code = sicode; \ | |
495 | info.si_addr = (void __user *)siaddr; \ | |
496 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) \ | |
497 | == NOTIFY_STOP) \ | |
498 | return; \ | |
499 | do_trap(trapnr, signr, str, 1, regs, error_code, &info); \ | |
500 | } | |
501 | ||
502 | DO_VM86_ERROR_INFO( 0, SIGFPE, "divide error", divide_error, FPE_INTDIV, regs->eip) | |
503 | #ifndef CONFIG_KPROBES | |
504 | DO_VM86_ERROR( 3, SIGTRAP, "int3", int3) | |
505 | #endif | |
506 | DO_VM86_ERROR( 4, SIGSEGV, "overflow", overflow) | |
507 | DO_VM86_ERROR( 5, SIGSEGV, "bounds", bounds) | |
631b0347 | 508 | DO_ERROR_INFO( 6, SIGILL, "invalid opcode", invalid_op, ILL_ILLOPN, regs->eip) |
1da177e4 LT |
509 | DO_ERROR( 9, SIGFPE, "coprocessor segment overrun", coprocessor_segment_overrun) |
510 | DO_ERROR(10, SIGSEGV, "invalid TSS", invalid_TSS) | |
511 | DO_ERROR(11, SIGBUS, "segment not present", segment_not_present) | |
512 | DO_ERROR(12, SIGBUS, "stack segment", stack_segment) | |
513 | DO_ERROR_INFO(17, SIGBUS, "alignment check", alignment_check, BUS_ADRALN, 0) | |
a879cbbb | 514 | DO_ERROR_INFO(32, SIGSEGV, "iret exception", iret_error, ILL_BADSTK, 0) |
1da177e4 | 515 | |
3d97ae5b PP |
516 | fastcall void __kprobes do_general_protection(struct pt_regs * regs, |
517 | long error_code) | |
1da177e4 LT |
518 | { |
519 | int cpu = get_cpu(); | |
520 | struct tss_struct *tss = &per_cpu(init_tss, cpu); | |
521 | struct thread_struct *thread = ¤t->thread; | |
522 | ||
523 | /* | |
524 | * Perform the lazy TSS's I/O bitmap copy. If the TSS has an | |
525 | * invalid offset set (the LAZY one) and the faulting thread has | |
526 | * a valid I/O bitmap pointer, we copy the I/O bitmap in the TSS | |
527 | * and we set the offset field correctly. Then we let the CPU to | |
528 | * restart the faulting instruction. | |
529 | */ | |
530 | if (tss->io_bitmap_base == INVALID_IO_BITMAP_OFFSET_LAZY && | |
531 | thread->io_bitmap_ptr) { | |
532 | memcpy(tss->io_bitmap, thread->io_bitmap_ptr, | |
533 | thread->io_bitmap_max); | |
534 | /* | |
535 | * If the previously set map was extending to higher ports | |
536 | * than the current one, pad extra space with 0xff (no access). | |
537 | */ | |
538 | if (thread->io_bitmap_max < tss->io_bitmap_max) | |
539 | memset((char *) tss->io_bitmap + | |
540 | thread->io_bitmap_max, 0xff, | |
541 | tss->io_bitmap_max - thread->io_bitmap_max); | |
542 | tss->io_bitmap_max = thread->io_bitmap_max; | |
543 | tss->io_bitmap_base = IO_BITMAP_OFFSET; | |
d5cd4aad | 544 | tss->io_bitmap_owner = thread; |
1da177e4 LT |
545 | put_cpu(); |
546 | return; | |
547 | } | |
548 | put_cpu(); | |
549 | ||
4f339ecb AN |
550 | current->thread.error_code = error_code; |
551 | current->thread.trap_no = 13; | |
552 | ||
1da177e4 LT |
553 | if (regs->eflags & VM_MASK) |
554 | goto gp_in_vm86; | |
555 | ||
717b594a | 556 | if (!user_mode(regs)) |
1da177e4 LT |
557 | goto gp_in_kernel; |
558 | ||
559 | current->thread.error_code = error_code; | |
560 | current->thread.trap_no = 13; | |
561 | force_sig(SIGSEGV, current); | |
562 | return; | |
563 | ||
564 | gp_in_vm86: | |
565 | local_irq_enable(); | |
566 | handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code); | |
567 | return; | |
568 | ||
569 | gp_in_kernel: | |
570 | if (!fixup_exception(regs)) { | |
571 | if (notify_die(DIE_GPF, "general protection fault", regs, | |
572 | error_code, 13, SIGSEGV) == NOTIFY_STOP) | |
573 | return; | |
574 | die("general protection fault", regs, error_code); | |
575 | } | |
576 | } | |
577 | ||
578 | static void mem_parity_error(unsigned char reason, struct pt_regs * regs) | |
579 | { | |
9c107805 DJ |
580 | printk(KERN_EMERG "Uhhuh. NMI received. Dazed and confused, but trying " |
581 | "to continue\n"); | |
582 | printk(KERN_EMERG "You probably have a hardware problem with your RAM " | |
583 | "chips\n"); | |
1da177e4 LT |
584 | |
585 | /* Clear and disable the memory parity error line. */ | |
586 | clear_mem_error(reason); | |
587 | } | |
588 | ||
589 | static void io_check_error(unsigned char reason, struct pt_regs * regs) | |
590 | { | |
591 | unsigned long i; | |
592 | ||
9c107805 | 593 | printk(KERN_EMERG "NMI: IOCK error (debug interrupt?)\n"); |
1da177e4 LT |
594 | show_registers(regs); |
595 | ||
596 | /* Re-enable the IOCK line, wait for a few seconds */ | |
597 | reason = (reason & 0xf) | 8; | |
598 | outb(reason, 0x61); | |
599 | i = 2000; | |
600 | while (--i) udelay(1000); | |
601 | reason &= ~8; | |
602 | outb(reason, 0x61); | |
603 | } | |
604 | ||
605 | static void unknown_nmi_error(unsigned char reason, struct pt_regs * regs) | |
606 | { | |
607 | #ifdef CONFIG_MCA | |
608 | /* Might actually be able to figure out what the guilty party | |
609 | * is. */ | |
610 | if( MCA_bus ) { | |
611 | mca_handle_nmi(); | |
612 | return; | |
613 | } | |
614 | #endif | |
615 | printk("Uhhuh. NMI received for unknown reason %02x on CPU %d.\n", | |
616 | reason, smp_processor_id()); | |
617 | printk("Dazed and confused, but trying to continue\n"); | |
618 | printk("Do you have a strange power saving mode enabled?\n"); | |
619 | } | |
620 | ||
621 | static DEFINE_SPINLOCK(nmi_print_lock); | |
622 | ||
623 | void die_nmi (struct pt_regs *regs, const char *msg) | |
624 | { | |
748f2edb GA |
625 | if (notify_die(DIE_NMIWATCHDOG, msg, regs, 0, 0, SIGINT) == |
626 | NOTIFY_STOP) | |
627 | return; | |
628 | ||
1da177e4 LT |
629 | spin_lock(&nmi_print_lock); |
630 | /* | |
631 | * We are in trouble anyway, lets at least try | |
632 | * to get a message out. | |
633 | */ | |
634 | bust_spinlocks(1); | |
9c107805 | 635 | printk(KERN_EMERG "%s", msg); |
1da177e4 LT |
636 | printk(" on CPU%d, eip %08lx, registers:\n", |
637 | smp_processor_id(), regs->eip); | |
638 | show_registers(regs); | |
9c107805 | 639 | printk(KERN_EMERG "console shuts up ...\n"); |
1da177e4 LT |
640 | console_silent(); |
641 | spin_unlock(&nmi_print_lock); | |
642 | bust_spinlocks(0); | |
6e274d14 AN |
643 | |
644 | /* If we are in kernel we are probably nested up pretty bad | |
645 | * and might aswell get out now while we still can. | |
646 | */ | |
db753bdf | 647 | if (!user_mode_vm(regs)) { |
6e274d14 AN |
648 | current->thread.trap_no = 2; |
649 | crash_kexec(regs); | |
650 | } | |
651 | ||
1da177e4 LT |
652 | do_exit(SIGSEGV); |
653 | } | |
654 | ||
655 | static void default_do_nmi(struct pt_regs * regs) | |
656 | { | |
657 | unsigned char reason = 0; | |
658 | ||
659 | /* Only the BSP gets external NMIs from the system. */ | |
660 | if (!smp_processor_id()) | |
661 | reason = get_nmi_reason(); | |
662 | ||
663 | if (!(reason & 0xc0)) { | |
664 | if (notify_die(DIE_NMI_IPI, "nmi_ipi", regs, reason, 0, SIGINT) | |
665 | == NOTIFY_STOP) | |
666 | return; | |
667 | #ifdef CONFIG_X86_LOCAL_APIC | |
668 | /* | |
669 | * Ok, so this is none of the documented NMI sources, | |
670 | * so it must be the NMI watchdog. | |
671 | */ | |
672 | if (nmi_watchdog) { | |
673 | nmi_watchdog_tick(regs); | |
674 | return; | |
675 | } | |
676 | #endif | |
677 | unknown_nmi_error(reason, regs); | |
678 | return; | |
679 | } | |
680 | if (notify_die(DIE_NMI, "nmi", regs, reason, 0, SIGINT) == NOTIFY_STOP) | |
681 | return; | |
682 | if (reason & 0x80) | |
683 | mem_parity_error(reason, regs); | |
684 | if (reason & 0x40) | |
685 | io_check_error(reason, regs); | |
686 | /* | |
687 | * Reassert NMI in case it became active meanwhile | |
688 | * as it's edge-triggered. | |
689 | */ | |
690 | reassert_nmi(); | |
691 | } | |
692 | ||
693 | static int dummy_nmi_callback(struct pt_regs * regs, int cpu) | |
694 | { | |
695 | return 0; | |
696 | } | |
697 | ||
698 | static nmi_callback_t nmi_callback = dummy_nmi_callback; | |
699 | ||
700 | fastcall void do_nmi(struct pt_regs * regs, long error_code) | |
701 | { | |
702 | int cpu; | |
703 | ||
704 | nmi_enter(); | |
705 | ||
706 | cpu = smp_processor_id(); | |
f3705136 | 707 | |
1da177e4 LT |
708 | ++nmi_count(cpu); |
709 | ||
19306059 | 710 | if (!rcu_dereference(nmi_callback)(regs, cpu)) |
1da177e4 LT |
711 | default_do_nmi(regs); |
712 | ||
713 | nmi_exit(); | |
714 | } | |
715 | ||
716 | void set_nmi_callback(nmi_callback_t callback) | |
717 | { | |
101f12af | 718 | vmalloc_sync_all(); |
19306059 | 719 | rcu_assign_pointer(nmi_callback, callback); |
1da177e4 | 720 | } |
129f6946 | 721 | EXPORT_SYMBOL_GPL(set_nmi_callback); |
1da177e4 LT |
722 | |
723 | void unset_nmi_callback(void) | |
724 | { | |
725 | nmi_callback = dummy_nmi_callback; | |
726 | } | |
129f6946 | 727 | EXPORT_SYMBOL_GPL(unset_nmi_callback); |
1da177e4 LT |
728 | |
729 | #ifdef CONFIG_KPROBES | |
3d97ae5b | 730 | fastcall void __kprobes do_int3(struct pt_regs *regs, long error_code) |
1da177e4 LT |
731 | { |
732 | if (notify_die(DIE_INT3, "int3", regs, error_code, 3, SIGTRAP) | |
733 | == NOTIFY_STOP) | |
48c88211 | 734 | return; |
1da177e4 LT |
735 | /* This is an interrupt gate, because kprobes wants interrupts |
736 | disabled. Normal trap handlers don't. */ | |
737 | restore_interrupts(regs); | |
738 | do_trap(3, SIGTRAP, "int3", 1, regs, error_code, NULL); | |
1da177e4 LT |
739 | } |
740 | #endif | |
741 | ||
742 | /* | |
743 | * Our handling of the processor debug registers is non-trivial. | |
744 | * We do not clear them on entry and exit from the kernel. Therefore | |
745 | * it is possible to get a watchpoint trap here from inside the kernel. | |
746 | * However, the code in ./ptrace.c has ensured that the user can | |
747 | * only set watchpoints on userspace addresses. Therefore the in-kernel | |
748 | * watchpoint trap can only occur in code which is reading/writing | |
749 | * from user space. Such code must not hold kernel locks (since it | |
750 | * can equally take a page fault), therefore it is safe to call | |
751 | * force_sig_info even though that claims and releases locks. | |
752 | * | |
753 | * Code in ./signal.c ensures that the debug control register | |
754 | * is restored before we deliver any signal, and therefore that | |
755 | * user code runs with the correct debug control register even though | |
756 | * we clear it here. | |
757 | * | |
758 | * Being careful here means that we don't have to be as careful in a | |
759 | * lot of more complicated places (task switching can be a bit lazy | |
760 | * about restoring all the debug state, and ptrace doesn't have to | |
761 | * find every occurrence of the TF bit that could be saved away even | |
762 | * by user code) | |
763 | */ | |
3d97ae5b | 764 | fastcall void __kprobes do_debug(struct pt_regs * regs, long error_code) |
1da177e4 LT |
765 | { |
766 | unsigned int condition; | |
767 | struct task_struct *tsk = current; | |
768 | ||
1cc6f12e | 769 | get_debugreg(condition, 6); |
1da177e4 LT |
770 | |
771 | if (notify_die(DIE_DEBUG, "debug", regs, condition, error_code, | |
772 | SIGTRAP) == NOTIFY_STOP) | |
773 | return; | |
774 | /* It's safe to allow irq's after DR6 has been saved */ | |
775 | if (regs->eflags & X86_EFLAGS_IF) | |
776 | local_irq_enable(); | |
777 | ||
778 | /* Mask out spurious debug traps due to lazy DR7 setting */ | |
779 | if (condition & (DR_TRAP0|DR_TRAP1|DR_TRAP2|DR_TRAP3)) { | |
780 | if (!tsk->thread.debugreg[7]) | |
781 | goto clear_dr7; | |
782 | } | |
783 | ||
784 | if (regs->eflags & VM_MASK) | |
785 | goto debug_vm86; | |
786 | ||
787 | /* Save debug status register where ptrace can see it */ | |
788 | tsk->thread.debugreg[6] = condition; | |
789 | ||
790 | /* | |
791 | * Single-stepping through TF: make sure we ignore any events in | |
792 | * kernel space (but re-enable TF when returning to user mode). | |
793 | */ | |
794 | if (condition & DR_STEP) { | |
795 | /* | |
796 | * We already checked v86 mode above, so we can | |
797 | * check for kernel mode by just checking the CPL | |
798 | * of CS. | |
799 | */ | |
717b594a | 800 | if (!user_mode(regs)) |
1da177e4 LT |
801 | goto clear_TF_reenable; |
802 | } | |
803 | ||
804 | /* Ok, finally something we can handle */ | |
805 | send_sigtrap(tsk, regs, error_code); | |
806 | ||
807 | /* Disable additional traps. They'll be re-enabled when | |
808 | * the signal is delivered. | |
809 | */ | |
810 | clear_dr7: | |
1cc6f12e | 811 | set_debugreg(0, 7); |
1da177e4 LT |
812 | return; |
813 | ||
814 | debug_vm86: | |
815 | handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, 1); | |
816 | return; | |
817 | ||
818 | clear_TF_reenable: | |
819 | set_tsk_thread_flag(tsk, TIF_SINGLESTEP); | |
820 | regs->eflags &= ~TF_MASK; | |
821 | return; | |
822 | } | |
823 | ||
824 | /* | |
825 | * Note that we play around with the 'TS' bit in an attempt to get | |
826 | * the correct behaviour even in the presence of the asynchronous | |
827 | * IRQ13 behaviour | |
828 | */ | |
829 | void math_error(void __user *eip) | |
830 | { | |
831 | struct task_struct * task; | |
832 | siginfo_t info; | |
833 | unsigned short cwd, swd; | |
834 | ||
835 | /* | |
836 | * Save the info for the exception handler and clear the error. | |
837 | */ | |
838 | task = current; | |
839 | save_init_fpu(task); | |
840 | task->thread.trap_no = 16; | |
841 | task->thread.error_code = 0; | |
842 | info.si_signo = SIGFPE; | |
843 | info.si_errno = 0; | |
844 | info.si_code = __SI_FAULT; | |
845 | info.si_addr = eip; | |
846 | /* | |
847 | * (~cwd & swd) will mask out exceptions that are not set to unmasked | |
848 | * status. 0x3f is the exception bits in these regs, 0x200 is the | |
849 | * C1 reg you need in case of a stack fault, 0x040 is the stack | |
850 | * fault bit. We should only be taking one exception at a time, | |
851 | * so if this combination doesn't produce any single exception, | |
852 | * then we have a bad program that isn't syncronizing its FPU usage | |
853 | * and it will suffer the consequences since we won't be able to | |
854 | * fully reproduce the context of the exception | |
855 | */ | |
856 | cwd = get_fpu_cwd(task); | |
857 | swd = get_fpu_swd(task); | |
b1daec30 | 858 | switch (swd & ~cwd & 0x3f) { |
33333373 CE |
859 | case 0x000: /* No unmasked exception */ |
860 | return; | |
861 | default: /* Multiple exceptions */ | |
1da177e4 LT |
862 | break; |
863 | case 0x001: /* Invalid Op */ | |
b1daec30 CE |
864 | /* |
865 | * swd & 0x240 == 0x040: Stack Underflow | |
866 | * swd & 0x240 == 0x240: Stack Overflow | |
867 | * User must clear the SF bit (0x40) if set | |
868 | */ | |
1da177e4 | 869 | info.si_code = FPE_FLTINV; |
1da177e4 LT |
870 | break; |
871 | case 0x002: /* Denormalize */ | |
872 | case 0x010: /* Underflow */ | |
873 | info.si_code = FPE_FLTUND; | |
874 | break; | |
875 | case 0x004: /* Zero Divide */ | |
876 | info.si_code = FPE_FLTDIV; | |
877 | break; | |
878 | case 0x008: /* Overflow */ | |
879 | info.si_code = FPE_FLTOVF; | |
880 | break; | |
881 | case 0x020: /* Precision */ | |
882 | info.si_code = FPE_FLTRES; | |
883 | break; | |
884 | } | |
885 | force_sig_info(SIGFPE, &info, task); | |
886 | } | |
887 | ||
888 | fastcall void do_coprocessor_error(struct pt_regs * regs, long error_code) | |
889 | { | |
890 | ignore_fpu_irq = 1; | |
891 | math_error((void __user *)regs->eip); | |
892 | } | |
893 | ||
894 | static void simd_math_error(void __user *eip) | |
895 | { | |
896 | struct task_struct * task; | |
897 | siginfo_t info; | |
898 | unsigned short mxcsr; | |
899 | ||
900 | /* | |
901 | * Save the info for the exception handler and clear the error. | |
902 | */ | |
903 | task = current; | |
904 | save_init_fpu(task); | |
905 | task->thread.trap_no = 19; | |
906 | task->thread.error_code = 0; | |
907 | info.si_signo = SIGFPE; | |
908 | info.si_errno = 0; | |
909 | info.si_code = __SI_FAULT; | |
910 | info.si_addr = eip; | |
911 | /* | |
912 | * The SIMD FPU exceptions are handled a little differently, as there | |
913 | * is only a single status/control register. Thus, to determine which | |
914 | * unmasked exception was caught we must mask the exception mask bits | |
915 | * at 0x1f80, and then use these to mask the exception bits at 0x3f. | |
916 | */ | |
917 | mxcsr = get_fpu_mxcsr(task); | |
918 | switch (~((mxcsr & 0x1f80) >> 7) & (mxcsr & 0x3f)) { | |
919 | case 0x000: | |
920 | default: | |
921 | break; | |
922 | case 0x001: /* Invalid Op */ | |
923 | info.si_code = FPE_FLTINV; | |
924 | break; | |
925 | case 0x002: /* Denormalize */ | |
926 | case 0x010: /* Underflow */ | |
927 | info.si_code = FPE_FLTUND; | |
928 | break; | |
929 | case 0x004: /* Zero Divide */ | |
930 | info.si_code = FPE_FLTDIV; | |
931 | break; | |
932 | case 0x008: /* Overflow */ | |
933 | info.si_code = FPE_FLTOVF; | |
934 | break; | |
935 | case 0x020: /* Precision */ | |
936 | info.si_code = FPE_FLTRES; | |
937 | break; | |
938 | } | |
939 | force_sig_info(SIGFPE, &info, task); | |
940 | } | |
941 | ||
942 | fastcall void do_simd_coprocessor_error(struct pt_regs * regs, | |
943 | long error_code) | |
944 | { | |
945 | if (cpu_has_xmm) { | |
946 | /* Handle SIMD FPU exceptions on PIII+ processors. */ | |
947 | ignore_fpu_irq = 1; | |
948 | simd_math_error((void __user *)regs->eip); | |
949 | } else { | |
950 | /* | |
951 | * Handle strange cache flush from user space exception | |
952 | * in all other cases. This is undocumented behaviour. | |
953 | */ | |
954 | if (regs->eflags & VM_MASK) { | |
955 | handle_vm86_fault((struct kernel_vm86_regs *)regs, | |
956 | error_code); | |
957 | return; | |
958 | } | |
1da177e4 LT |
959 | current->thread.trap_no = 19; |
960 | current->thread.error_code = error_code; | |
4f339ecb | 961 | die_if_kernel("cache flush denied", regs, error_code); |
1da177e4 LT |
962 | force_sig(SIGSEGV, current); |
963 | } | |
964 | } | |
965 | ||
966 | fastcall void do_spurious_interrupt_bug(struct pt_regs * regs, | |
967 | long error_code) | |
968 | { | |
969 | #if 0 | |
970 | /* No need to warn about this any longer. */ | |
971 | printk("Ignoring P6 Local APIC Spurious Interrupt Bug...\n"); | |
972 | #endif | |
973 | } | |
974 | ||
975 | fastcall void setup_x86_bogus_stack(unsigned char * stk) | |
976 | { | |
977 | unsigned long *switch16_ptr, *switch32_ptr; | |
978 | struct pt_regs *regs; | |
979 | unsigned long stack_top, stack_bot; | |
980 | unsigned short iret_frame16_off; | |
981 | int cpu = smp_processor_id(); | |
982 | /* reserve the space on 32bit stack for the magic switch16 pointer */ | |
983 | memmove(stk, stk + 8, sizeof(struct pt_regs)); | |
984 | switch16_ptr = (unsigned long *)(stk + sizeof(struct pt_regs)); | |
985 | regs = (struct pt_regs *)stk; | |
986 | /* now the switch32 on 16bit stack */ | |
987 | stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); | |
988 | stack_top = stack_bot + CPU_16BIT_STACK_SIZE; | |
989 | switch32_ptr = (unsigned long *)(stack_top - 8); | |
990 | iret_frame16_off = CPU_16BIT_STACK_SIZE - 8 - 20; | |
991 | /* copy iret frame on 16bit stack */ | |
992 | memcpy((void *)(stack_bot + iret_frame16_off), ®s->eip, 20); | |
993 | /* fill in the switch pointers */ | |
994 | switch16_ptr[0] = (regs->esp & 0xffff0000) | iret_frame16_off; | |
995 | switch16_ptr[1] = __ESPFIX_SS; | |
996 | switch32_ptr[0] = (unsigned long)stk + sizeof(struct pt_regs) + | |
997 | 8 - CPU_16BIT_STACK_SIZE; | |
998 | switch32_ptr[1] = __KERNEL_DS; | |
999 | } | |
1000 | ||
1001 | fastcall unsigned char * fixup_x86_bogus_stack(unsigned short sp) | |
1002 | { | |
1003 | unsigned long *switch32_ptr; | |
1004 | unsigned char *stack16, *stack32; | |
1005 | unsigned long stack_top, stack_bot; | |
1006 | int len; | |
1007 | int cpu = smp_processor_id(); | |
1008 | stack_bot = (unsigned long)&per_cpu(cpu_16bit_stack, cpu); | |
1009 | stack_top = stack_bot + CPU_16BIT_STACK_SIZE; | |
1010 | switch32_ptr = (unsigned long *)(stack_top - 8); | |
1011 | /* copy the data from 16bit stack to 32bit stack */ | |
1012 | len = CPU_16BIT_STACK_SIZE - 8 - sp; | |
1013 | stack16 = (unsigned char *)(stack_bot + sp); | |
1014 | stack32 = (unsigned char *) | |
1015 | (switch32_ptr[0] + CPU_16BIT_STACK_SIZE - 8 - len); | |
1016 | memcpy(stack32, stack16, len); | |
1017 | return stack32; | |
1018 | } | |
1019 | ||
1020 | /* | |
1021 | * 'math_state_restore()' saves the current math information in the | |
1022 | * old math state array, and gets the new ones from the current task | |
1023 | * | |
1024 | * Careful.. There are problems with IBM-designed IRQ13 behaviour. | |
1025 | * Don't touch unless you *really* know how it works. | |
1026 | * | |
1027 | * Must be called with kernel preemption disabled (in this case, | |
1028 | * local interrupts are disabled at the call-site in entry.S). | |
1029 | */ | |
1030 | asmlinkage void math_state_restore(struct pt_regs regs) | |
1031 | { | |
1032 | struct thread_info *thread = current_thread_info(); | |
1033 | struct task_struct *tsk = thread->task; | |
1034 | ||
1035 | clts(); /* Allow maths ops (or we recurse) */ | |
1036 | if (!tsk_used_math(tsk)) | |
1037 | init_fpu(tsk); | |
1038 | restore_fpu(tsk); | |
1039 | thread->status |= TS_USEDFPU; /* So we fnsave on switch_to() */ | |
1040 | } | |
1041 | ||
1042 | #ifndef CONFIG_MATH_EMULATION | |
1043 | ||
1044 | asmlinkage void math_emulate(long arg) | |
1045 | { | |
9c107805 DJ |
1046 | printk(KERN_EMERG "math-emulation not enabled and no coprocessor found.\n"); |
1047 | printk(KERN_EMERG "killing %s.\n",current->comm); | |
1da177e4 LT |
1048 | force_sig(SIGFPE,current); |
1049 | schedule(); | |
1050 | } | |
1051 | ||
1052 | #endif /* CONFIG_MATH_EMULATION */ | |
1053 | ||
1054 | #ifdef CONFIG_X86_F00F_BUG | |
1055 | void __init trap_init_f00f_bug(void) | |
1056 | { | |
1057 | __set_fixmap(FIX_F00F_IDT, __pa(&idt_table), PAGE_KERNEL_RO); | |
1058 | ||
1059 | /* | |
1060 | * Update the IDT descriptor and reload the IDT so that | |
1061 | * it uses the read-only mapped virtual address. | |
1062 | */ | |
1063 | idt_descr.address = fix_to_virt(FIX_F00F_IDT); | |
4d37e7e3 | 1064 | load_idt(&idt_descr); |
1da177e4 LT |
1065 | } |
1066 | #endif | |
1067 | ||
1068 | #define _set_gate(gate_addr,type,dpl,addr,seg) \ | |
1069 | do { \ | |
1070 | int __d0, __d1; \ | |
1071 | __asm__ __volatile__ ("movw %%dx,%%ax\n\t" \ | |
1072 | "movw %4,%%dx\n\t" \ | |
1073 | "movl %%eax,%0\n\t" \ | |
1074 | "movl %%edx,%1" \ | |
1075 | :"=m" (*((long *) (gate_addr))), \ | |
1076 | "=m" (*(1+(long *) (gate_addr))), "=&a" (__d0), "=&d" (__d1) \ | |
1077 | :"i" ((short) (0x8000+(dpl<<13)+(type<<8))), \ | |
1078 | "3" ((char *) (addr)),"2" ((seg) << 16)); \ | |
1079 | } while (0) | |
1080 | ||
1081 | ||
1082 | /* | |
1083 | * This needs to use 'idt_table' rather than 'idt', and | |
1084 | * thus use the _nonmapped_ version of the IDT, as the | |
1085 | * Pentium F0 0F bugfix can have resulted in the mapped | |
1086 | * IDT being write-protected. | |
1087 | */ | |
1088 | void set_intr_gate(unsigned int n, void *addr) | |
1089 | { | |
1090 | _set_gate(idt_table+n,14,0,addr,__KERNEL_CS); | |
1091 | } | |
1092 | ||
1093 | /* | |
1094 | * This routine sets up an interrupt gate at directory privilege level 3. | |
1095 | */ | |
1096 | static inline void set_system_intr_gate(unsigned int n, void *addr) | |
1097 | { | |
1098 | _set_gate(idt_table+n, 14, 3, addr, __KERNEL_CS); | |
1099 | } | |
1100 | ||
1101 | static void __init set_trap_gate(unsigned int n, void *addr) | |
1102 | { | |
1103 | _set_gate(idt_table+n,15,0,addr,__KERNEL_CS); | |
1104 | } | |
1105 | ||
1106 | static void __init set_system_gate(unsigned int n, void *addr) | |
1107 | { | |
1108 | _set_gate(idt_table+n,15,3,addr,__KERNEL_CS); | |
1109 | } | |
1110 | ||
1111 | static void __init set_task_gate(unsigned int n, unsigned int gdt_entry) | |
1112 | { | |
1113 | _set_gate(idt_table+n,5,0,0,(gdt_entry<<3)); | |
1114 | } | |
1115 | ||
1116 | ||
1117 | void __init trap_init(void) | |
1118 | { | |
1119 | #ifdef CONFIG_EISA | |
1120 | void __iomem *p = ioremap(0x0FFFD9, 4); | |
1121 | if (readl(p) == 'E'+('I'<<8)+('S'<<16)+('A'<<24)) { | |
1122 | EISA_bus = 1; | |
1123 | } | |
1124 | iounmap(p); | |
1125 | #endif | |
1126 | ||
1127 | #ifdef CONFIG_X86_LOCAL_APIC | |
1128 | init_apic_mappings(); | |
1129 | #endif | |
1130 | ||
1131 | set_trap_gate(0,÷_error); | |
1132 | set_intr_gate(1,&debug); | |
1133 | set_intr_gate(2,&nmi); | |
eb05c324 | 1134 | set_system_intr_gate(3, &int3); /* int3/4 can be called from all */ |
1da177e4 | 1135 | set_system_gate(4,&overflow); |
eb05c324 | 1136 | set_trap_gate(5,&bounds); |
1da177e4 LT |
1137 | set_trap_gate(6,&invalid_op); |
1138 | set_trap_gate(7,&device_not_available); | |
1139 | set_task_gate(8,GDT_ENTRY_DOUBLEFAULT_TSS); | |
1140 | set_trap_gate(9,&coprocessor_segment_overrun); | |
1141 | set_trap_gate(10,&invalid_TSS); | |
1142 | set_trap_gate(11,&segment_not_present); | |
1143 | set_trap_gate(12,&stack_segment); | |
1144 | set_trap_gate(13,&general_protection); | |
1145 | set_intr_gate(14,&page_fault); | |
1146 | set_trap_gate(15,&spurious_interrupt_bug); | |
1147 | set_trap_gate(16,&coprocessor_error); | |
1148 | set_trap_gate(17,&alignment_check); | |
1149 | #ifdef CONFIG_X86_MCE | |
1150 | set_trap_gate(18,&machine_check); | |
1151 | #endif | |
1152 | set_trap_gate(19,&simd_coprocessor_error); | |
1153 | ||
d43c6e80 JB |
1154 | if (cpu_has_fxsr) { |
1155 | /* | |
1156 | * Verify that the FXSAVE/FXRSTOR data will be 16-byte aligned. | |
1157 | * Generates a compile-time "error: zero width for bit-field" if | |
1158 | * the alignment is wrong. | |
1159 | */ | |
1160 | struct fxsrAlignAssert { | |
1161 | int _:!(offsetof(struct task_struct, | |
1162 | thread.i387.fxsave) & 15); | |
1163 | }; | |
1164 | ||
1165 | printk(KERN_INFO "Enabling fast FPU save and restore... "); | |
1166 | set_in_cr4(X86_CR4_OSFXSR); | |
1167 | printk("done.\n"); | |
1168 | } | |
1169 | if (cpu_has_xmm) { | |
1170 | printk(KERN_INFO "Enabling unmasked SIMD FPU exception " | |
1171 | "support... "); | |
1172 | set_in_cr4(X86_CR4_OSXMMEXCPT); | |
1173 | printk("done.\n"); | |
1174 | } | |
1175 | ||
1da177e4 LT |
1176 | set_system_gate(SYSCALL_VECTOR,&system_call); |
1177 | ||
1178 | /* | |
1179 | * Should be a barrier for any external CPU state. | |
1180 | */ | |
1181 | cpu_init(); | |
1182 | ||
1183 | trap_init_hook(); | |
1184 | } | |
1185 | ||
1186 | static int __init kstack_setup(char *s) | |
1187 | { | |
1188 | kstack_depth_to_print = simple_strtoul(s, NULL, 0); | |
1189 | return 0; | |
1190 | } | |
1191 | __setup("kstack=", kstack_setup); |