]>
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 | /* | |
c1d518c8 | 10 | * Handle hardware traps and faults. |
1da177e4 | 11 | */ |
c767a54b JP |
12 | |
13 | #define pr_fmt(fmt) KBUILD_MODNAME ": " fmt | |
14 | ||
56dd9470 | 15 | #include <linux/context_tracking.h> |
b5964405 IM |
16 | #include <linux/interrupt.h> |
17 | #include <linux/kallsyms.h> | |
18 | #include <linux/spinlock.h> | |
b5964405 IM |
19 | #include <linux/kprobes.h> |
20 | #include <linux/uaccess.h> | |
b5964405 | 21 | #include <linux/kdebug.h> |
f503b5ae | 22 | #include <linux/kgdb.h> |
1da177e4 | 23 | #include <linux/kernel.h> |
b5964405 IM |
24 | #include <linux/module.h> |
25 | #include <linux/ptrace.h> | |
b02ef20a | 26 | #include <linux/uprobes.h> |
1da177e4 | 27 | #include <linux/string.h> |
b5964405 | 28 | #include <linux/delay.h> |
1da177e4 | 29 | #include <linux/errno.h> |
b5964405 IM |
30 | #include <linux/kexec.h> |
31 | #include <linux/sched.h> | |
1da177e4 | 32 | #include <linux/timer.h> |
1da177e4 | 33 | #include <linux/init.h> |
91768d6c | 34 | #include <linux/bug.h> |
b5964405 IM |
35 | #include <linux/nmi.h> |
36 | #include <linux/mm.h> | |
c1d518c8 AH |
37 | #include <linux/smp.h> |
38 | #include <linux/io.h> | |
1da177e4 LT |
39 | |
40 | #ifdef CONFIG_EISA | |
41 | #include <linux/ioport.h> | |
42 | #include <linux/eisa.h> | |
43 | #endif | |
44 | ||
c0d12172 DJ |
45 | #if defined(CONFIG_EDAC) |
46 | #include <linux/edac.h> | |
47 | #endif | |
48 | ||
f8561296 | 49 | #include <asm/kmemcheck.h> |
b5964405 | 50 | #include <asm/stacktrace.h> |
1da177e4 | 51 | #include <asm/processor.h> |
1da177e4 | 52 | #include <asm/debugreg.h> |
60063497 | 53 | #include <linux/atomic.h> |
08d636b6 | 54 | #include <asm/ftrace.h> |
c1d518c8 | 55 | #include <asm/traps.h> |
1da177e4 LT |
56 | #include <asm/desc.h> |
57 | #include <asm/i387.h> | |
1361b83a | 58 | #include <asm/fpu-internal.h> |
9e55e44e | 59 | #include <asm/mce.h> |
4eefbe79 | 60 | #include <asm/fixmap.h> |
1164dd00 | 61 | #include <asm/mach_traps.h> |
17f41571 | 62 | #include <asm/alternative.h> |
fe3d197f | 63 | #include <asm/mpx.h> |
c1d518c8 | 64 | |
081f75bb | 65 | #ifdef CONFIG_X86_64 |
428cf902 | 66 | #include <asm/x86_init.h> |
081f75bb AH |
67 | #include <asm/pgalloc.h> |
68 | #include <asm/proto.h> | |
4df05f36 KC |
69 | |
70 | /* No need to be aligned, but done to keep all IDTs defined the same way. */ | |
71 | gate_desc debug_idt_table[NR_VECTORS] __page_aligned_bss; | |
081f75bb | 72 | #else |
c1d518c8 | 73 | #include <asm/processor-flags.h> |
8e6dafd6 | 74 | #include <asm/setup.h> |
1da177e4 | 75 | |
1da177e4 | 76 | asmlinkage int system_call(void); |
081f75bb | 77 | #endif |
1da177e4 | 78 | |
4df05f36 KC |
79 | /* Must be page-aligned because the real IDT is used in a fixmap. */ |
80 | gate_desc idt_table[NR_VECTORS] __page_aligned_bss; | |
81 | ||
b77b881f YL |
82 | DECLARE_BITMAP(used_vectors, NR_VECTORS); |
83 | EXPORT_SYMBOL_GPL(used_vectors); | |
84 | ||
762db434 AH |
85 | static inline void conditional_sti(struct pt_regs *regs) |
86 | { | |
87 | if (regs->flags & X86_EFLAGS_IF) | |
88 | local_irq_enable(); | |
89 | } | |
90 | ||
3d2a71a5 AH |
91 | static inline void preempt_conditional_sti(struct pt_regs *regs) |
92 | { | |
bdb43806 | 93 | preempt_count_inc(); |
3d2a71a5 AH |
94 | if (regs->flags & X86_EFLAGS_IF) |
95 | local_irq_enable(); | |
96 | } | |
97 | ||
be716615 TG |
98 | static inline void conditional_cli(struct pt_regs *regs) |
99 | { | |
100 | if (regs->flags & X86_EFLAGS_IF) | |
101 | local_irq_disable(); | |
102 | } | |
103 | ||
3d2a71a5 AH |
104 | static inline void preempt_conditional_cli(struct pt_regs *regs) |
105 | { | |
106 | if (regs->flags & X86_EFLAGS_IF) | |
107 | local_irq_disable(); | |
bdb43806 | 108 | preempt_count_dec(); |
3d2a71a5 AH |
109 | } |
110 | ||
95927475 AL |
111 | enum ctx_state ist_enter(struct pt_regs *regs) |
112 | { | |
113 | /* | |
114 | * We are atomic because we're on the IST stack (or we're on x86_32, | |
115 | * in which case we still shouldn't schedule. | |
116 | */ | |
117 | preempt_count_add(HARDIRQ_OFFSET); | |
118 | ||
119 | if (user_mode_vm(regs)) { | |
120 | /* Other than that, we're just an exception. */ | |
121 | return exception_enter(); | |
122 | } else { | |
123 | /* | |
124 | * We might have interrupted pretty much anything. In | |
125 | * fact, if we're a machine check, we can even interrupt | |
126 | * NMI processing. We don't want in_nmi() to return true, | |
127 | * but we need to notify RCU. | |
128 | */ | |
129 | rcu_nmi_enter(); | |
130 | return IN_KERNEL; /* the value is irrelevant. */ | |
131 | } | |
132 | } | |
133 | ||
134 | void ist_exit(struct pt_regs *regs, enum ctx_state prev_state) | |
135 | { | |
136 | preempt_count_sub(HARDIRQ_OFFSET); | |
137 | ||
138 | if (user_mode_vm(regs)) | |
139 | return exception_exit(prev_state); | |
140 | else | |
141 | rcu_nmi_exit(); | |
142 | } | |
143 | ||
bced35b6 AL |
144 | /** |
145 | * ist_begin_non_atomic() - begin a non-atomic section in an IST exception | |
146 | * @regs: regs passed to the IST exception handler | |
147 | * | |
148 | * IST exception handlers normally cannot schedule. As a special | |
149 | * exception, if the exception interrupted userspace code (i.e. | |
150 | * user_mode_vm(regs) would return true) and the exception was not | |
151 | * a double fault, it can be safe to schedule. ist_begin_non_atomic() | |
152 | * begins a non-atomic section within an ist_enter()/ist_exit() region. | |
153 | * Callers are responsible for enabling interrupts themselves inside | |
154 | * the non-atomic section, and callers must call is_end_non_atomic() | |
155 | * before ist_exit(). | |
156 | */ | |
157 | void ist_begin_non_atomic(struct pt_regs *regs) | |
158 | { | |
159 | BUG_ON(!user_mode_vm(regs)); | |
160 | ||
161 | /* | |
162 | * Sanity check: we need to be on the normal thread stack. This | |
163 | * will catch asm bugs and any attempt to use ist_preempt_enable | |
164 | * from double_fault. | |
165 | */ | |
166 | BUG_ON(((current_stack_pointer() ^ this_cpu_read_stable(kernel_stack)) | |
167 | & ~(THREAD_SIZE - 1)) != 0); | |
168 | ||
169 | preempt_count_sub(HARDIRQ_OFFSET); | |
170 | } | |
171 | ||
172 | /** | |
173 | * ist_end_non_atomic() - begin a non-atomic section in an IST exception | |
174 | * | |
175 | * Ends a non-atomic section started with ist_begin_non_atomic(). | |
176 | */ | |
177 | void ist_end_non_atomic(void) | |
178 | { | |
179 | preempt_count_add(HARDIRQ_OFFSET); | |
180 | } | |
181 | ||
9326638c | 182 | static nokprobe_inline int |
c416ddf5 FW |
183 | do_trap_no_signal(struct task_struct *tsk, int trapnr, char *str, |
184 | struct pt_regs *regs, long error_code) | |
1da177e4 | 185 | { |
081f75bb | 186 | #ifdef CONFIG_X86_32 |
6b6891f9 | 187 | if (regs->flags & X86_VM_MASK) { |
3c1326f8 | 188 | /* |
c416ddf5 | 189 | * Traps 0, 1, 3, 4, and 5 should be forwarded to vm86. |
3c1326f8 AH |
190 | * On nmi (interrupt 2), do_trap should not be called. |
191 | */ | |
c416ddf5 FW |
192 | if (trapnr < X86_TRAP_UD) { |
193 | if (!handle_vm86_trap((struct kernel_vm86_regs *) regs, | |
194 | error_code, trapnr)) | |
195 | return 0; | |
196 | } | |
197 | return -1; | |
1da177e4 | 198 | } |
081f75bb | 199 | #endif |
c416ddf5 FW |
200 | if (!user_mode(regs)) { |
201 | if (!fixup_exception(regs)) { | |
202 | tsk->thread.error_code = error_code; | |
203 | tsk->thread.trap_nr = trapnr; | |
204 | die(str, regs, error_code); | |
205 | } | |
206 | return 0; | |
207 | } | |
1da177e4 | 208 | |
c416ddf5 FW |
209 | return -1; |
210 | } | |
1da177e4 | 211 | |
1c326c4d ON |
212 | static siginfo_t *fill_trap_info(struct pt_regs *regs, int signr, int trapnr, |
213 | siginfo_t *info) | |
958d3d72 ON |
214 | { |
215 | unsigned long siaddr; | |
216 | int sicode; | |
217 | ||
218 | switch (trapnr) { | |
1c326c4d ON |
219 | default: |
220 | return SEND_SIG_PRIV; | |
221 | ||
958d3d72 ON |
222 | case X86_TRAP_DE: |
223 | sicode = FPE_INTDIV; | |
b02ef20a | 224 | siaddr = uprobe_get_trap_addr(regs); |
958d3d72 ON |
225 | break; |
226 | case X86_TRAP_UD: | |
227 | sicode = ILL_ILLOPN; | |
b02ef20a | 228 | siaddr = uprobe_get_trap_addr(regs); |
958d3d72 ON |
229 | break; |
230 | case X86_TRAP_AC: | |
231 | sicode = BUS_ADRALN; | |
232 | siaddr = 0; | |
233 | break; | |
234 | } | |
235 | ||
236 | info->si_signo = signr; | |
237 | info->si_errno = 0; | |
238 | info->si_code = sicode; | |
239 | info->si_addr = (void __user *)siaddr; | |
1c326c4d | 240 | return info; |
958d3d72 ON |
241 | } |
242 | ||
9326638c | 243 | static void |
c416ddf5 FW |
244 | do_trap(int trapnr, int signr, char *str, struct pt_regs *regs, |
245 | long error_code, siginfo_t *info) | |
246 | { | |
247 | struct task_struct *tsk = current; | |
248 | ||
249 | ||
250 | if (!do_trap_no_signal(tsk, trapnr, str, regs, error_code)) | |
251 | return; | |
b5964405 | 252 | /* |
51e7dc70 | 253 | * We want error_code and trap_nr set for userspace faults and |
b5964405 IM |
254 | * kernelspace faults which result in die(), but not |
255 | * kernelspace faults which are fixed up. die() gives the | |
256 | * process no chance to handle the signal and notice the | |
257 | * kernel fault information, so that won't result in polluting | |
258 | * the information about previously queued, but not yet | |
259 | * delivered, faults. See also do_general_protection below. | |
260 | */ | |
261 | tsk->thread.error_code = error_code; | |
51e7dc70 | 262 | tsk->thread.trap_nr = trapnr; |
d1895183 | 263 | |
081f75bb AH |
264 | #ifdef CONFIG_X86_64 |
265 | if (show_unhandled_signals && unhandled_signal(tsk, signr) && | |
266 | printk_ratelimit()) { | |
c767a54b JP |
267 | pr_info("%s[%d] trap %s ip:%lx sp:%lx error:%lx", |
268 | tsk->comm, tsk->pid, str, | |
269 | regs->ip, regs->sp, error_code); | |
081f75bb | 270 | print_vma_addr(" in ", regs->ip); |
c767a54b | 271 | pr_cont("\n"); |
081f75bb AH |
272 | } |
273 | #endif | |
274 | ||
38cad57b | 275 | force_sig_info(signr, info ?: SEND_SIG_PRIV, tsk); |
1da177e4 | 276 | } |
9326638c | 277 | NOKPROBE_SYMBOL(do_trap); |
1da177e4 | 278 | |
dff0796e | 279 | static void do_error_trap(struct pt_regs *regs, long error_code, char *str, |
1c326c4d | 280 | unsigned long trapnr, int signr) |
dff0796e ON |
281 | { |
282 | enum ctx_state prev_state = exception_enter(); | |
1c326c4d | 283 | siginfo_t info; |
dff0796e ON |
284 | |
285 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, signr) != | |
286 | NOTIFY_STOP) { | |
287 | conditional_sti(regs); | |
1c326c4d ON |
288 | do_trap(trapnr, signr, str, regs, error_code, |
289 | fill_trap_info(regs, signr, trapnr, &info)); | |
dff0796e ON |
290 | } |
291 | ||
292 | exception_exit(prev_state); | |
293 | } | |
294 | ||
b5964405 | 295 | #define DO_ERROR(trapnr, signr, str, name) \ |
e407d620 | 296 | dotraplinkage void do_##name(struct pt_regs *regs, long error_code) \ |
b5964405 | 297 | { \ |
1c326c4d | 298 | do_error_trap(regs, error_code, str, trapnr, signr); \ |
1da177e4 LT |
299 | } |
300 | ||
0eb14833 ON |
301 | DO_ERROR(X86_TRAP_DE, SIGFPE, "divide error", divide_error) |
302 | DO_ERROR(X86_TRAP_OF, SIGSEGV, "overflow", overflow) | |
0eb14833 ON |
303 | DO_ERROR(X86_TRAP_UD, SIGILL, "invalid opcode", invalid_op) |
304 | DO_ERROR(X86_TRAP_OLD_MF, SIGFPE, "coprocessor segment overrun",coprocessor_segment_overrun) | |
305 | DO_ERROR(X86_TRAP_TS, SIGSEGV, "invalid TSS", invalid_TSS) | |
306 | DO_ERROR(X86_TRAP_NP, SIGBUS, "segment not present", segment_not_present) | |
0eb14833 | 307 | DO_ERROR(X86_TRAP_SS, SIGBUS, "stack segment", stack_segment) |
0eb14833 | 308 | DO_ERROR(X86_TRAP_AC, SIGBUS, "alignment check", alignment_check) |
1da177e4 | 309 | |
081f75bb AH |
310 | #ifdef CONFIG_X86_64 |
311 | /* Runs on IST stack */ | |
081f75bb AH |
312 | dotraplinkage void do_double_fault(struct pt_regs *regs, long error_code) |
313 | { | |
314 | static const char str[] = "double fault"; | |
315 | struct task_struct *tsk = current; | |
316 | ||
af726f21 AL |
317 | #ifdef CONFIG_X86_ESPFIX64 |
318 | extern unsigned char native_irq_return_iret[]; | |
319 | ||
320 | /* | |
321 | * If IRET takes a non-IST fault on the espfix64 stack, then we | |
322 | * end up promoting it to a doublefault. In that case, modify | |
323 | * the stack to make it look like we just entered the #GP | |
324 | * handler from user space, similar to bad_iret. | |
95927475 AL |
325 | * |
326 | * No need for ist_enter here because we don't use RCU. | |
af726f21 AL |
327 | */ |
328 | if (((long)regs->sp >> PGDIR_SHIFT) == ESPFIX_PGD_ENTRY && | |
329 | regs->cs == __KERNEL_CS && | |
330 | regs->ip == (unsigned long)native_irq_return_iret) | |
331 | { | |
332 | struct pt_regs *normal_regs = task_pt_regs(current); | |
333 | ||
334 | /* Fake a #GP(0) from userspace. */ | |
335 | memmove(&normal_regs->ip, (void *)regs->sp, 5*8); | |
336 | normal_regs->orig_ax = 0; /* Missing (lost) #GP error code */ | |
337 | regs->ip = (unsigned long)general_protection; | |
338 | regs->sp = (unsigned long)&normal_regs->orig_ax; | |
95927475 | 339 | |
af726f21 AL |
340 | return; |
341 | } | |
342 | #endif | |
343 | ||
95927475 | 344 | ist_enter(regs); /* Discard prev_state because we won't return. */ |
c9408265 | 345 | notify_die(DIE_TRAP, str, regs, error_code, X86_TRAP_DF, SIGSEGV); |
081f75bb AH |
346 | |
347 | tsk->thread.error_code = error_code; | |
51e7dc70 | 348 | tsk->thread.trap_nr = X86_TRAP_DF; |
081f75bb | 349 | |
4d067d8e BP |
350 | #ifdef CONFIG_DOUBLEFAULT |
351 | df_debug(regs, error_code); | |
352 | #endif | |
bd8b96df IM |
353 | /* |
354 | * This is always a kernel trap and never fixable (and thus must | |
355 | * never return). | |
356 | */ | |
081f75bb AH |
357 | for (;;) |
358 | die(str, regs, error_code); | |
359 | } | |
360 | #endif | |
361 | ||
fe3d197f DH |
362 | dotraplinkage void do_bounds(struct pt_regs *regs, long error_code) |
363 | { | |
364 | struct task_struct *tsk = current; | |
365 | struct xsave_struct *xsave_buf; | |
366 | enum ctx_state prev_state; | |
367 | struct bndcsr *bndcsr; | |
368 | siginfo_t *info; | |
369 | ||
370 | prev_state = exception_enter(); | |
371 | if (notify_die(DIE_TRAP, "bounds", regs, error_code, | |
372 | X86_TRAP_BR, SIGSEGV) == NOTIFY_STOP) | |
373 | goto exit; | |
374 | conditional_sti(regs); | |
375 | ||
376 | if (!user_mode(regs)) | |
377 | die("bounds", regs, error_code); | |
378 | ||
379 | if (!cpu_feature_enabled(X86_FEATURE_MPX)) { | |
380 | /* The exception is not from Intel MPX */ | |
381 | goto exit_trap; | |
382 | } | |
383 | ||
384 | /* | |
385 | * We need to look at BNDSTATUS to resolve this exception. | |
386 | * It is not directly accessible, though, so we need to | |
387 | * do an xsave and then pull it out of the xsave buffer. | |
388 | */ | |
389 | fpu_save_init(&tsk->thread.fpu); | |
390 | xsave_buf = &(tsk->thread.fpu.state->xsave); | |
391 | bndcsr = get_xsave_addr(xsave_buf, XSTATE_BNDCSR); | |
392 | if (!bndcsr) | |
393 | goto exit_trap; | |
394 | ||
395 | /* | |
396 | * The error code field of the BNDSTATUS register communicates status | |
397 | * information of a bound range exception #BR or operation involving | |
398 | * bound directory. | |
399 | */ | |
400 | switch (bndcsr->bndstatus & MPX_BNDSTA_ERROR_CODE) { | |
401 | case 2: /* Bound directory has invalid entry. */ | |
402 | if (mpx_handle_bd_fault(xsave_buf)) | |
403 | goto exit_trap; | |
404 | break; /* Success, it was handled */ | |
405 | case 1: /* Bound violation. */ | |
406 | info = mpx_generate_siginfo(regs, xsave_buf); | |
e10abb2f | 407 | if (IS_ERR(info)) { |
fe3d197f DH |
408 | /* |
409 | * We failed to decode the MPX instruction. Act as if | |
410 | * the exception was not caused by MPX. | |
411 | */ | |
412 | goto exit_trap; | |
413 | } | |
414 | /* | |
415 | * Success, we decoded the instruction and retrieved | |
416 | * an 'info' containing the address being accessed | |
417 | * which caused the exception. This information | |
418 | * allows and application to possibly handle the | |
419 | * #BR exception itself. | |
420 | */ | |
421 | do_trap(X86_TRAP_BR, SIGSEGV, "bounds", regs, error_code, info); | |
422 | kfree(info); | |
423 | break; | |
424 | case 0: /* No exception caused by Intel MPX operations. */ | |
425 | goto exit_trap; | |
426 | default: | |
427 | die("bounds", regs, error_code); | |
428 | } | |
429 | ||
430 | exit: | |
431 | exception_exit(prev_state); | |
432 | return; | |
433 | exit_trap: | |
434 | /* | |
435 | * This path out is for all the cases where we could not | |
436 | * handle the exception in some way (like allocating a | |
437 | * table or telling userspace about it. We will also end | |
438 | * up here if the kernel has MPX turned off at compile | |
439 | * time.. | |
440 | */ | |
441 | do_trap(X86_TRAP_BR, SIGSEGV, "bounds", regs, error_code, NULL); | |
442 | exception_exit(prev_state); | |
443 | } | |
444 | ||
9326638c | 445 | dotraplinkage void |
13485ab5 | 446 | do_general_protection(struct pt_regs *regs, long error_code) |
1da177e4 | 447 | { |
13485ab5 | 448 | struct task_struct *tsk; |
6c1e0256 | 449 | enum ctx_state prev_state; |
b5964405 | 450 | |
6c1e0256 | 451 | prev_state = exception_enter(); |
c6df0d71 AH |
452 | conditional_sti(regs); |
453 | ||
081f75bb | 454 | #ifdef CONFIG_X86_32 |
ef3f6288 FW |
455 | if (regs->flags & X86_VM_MASK) { |
456 | local_irq_enable(); | |
457 | handle_vm86_fault((struct kernel_vm86_regs *) regs, error_code); | |
6ba3c97a | 458 | goto exit; |
ef3f6288 | 459 | } |
081f75bb | 460 | #endif |
1da177e4 | 461 | |
13485ab5 | 462 | tsk = current; |
ef3f6288 FW |
463 | if (!user_mode(regs)) { |
464 | if (fixup_exception(regs)) | |
6ba3c97a | 465 | goto exit; |
ef3f6288 FW |
466 | |
467 | tsk->thread.error_code = error_code; | |
468 | tsk->thread.trap_nr = X86_TRAP_GP; | |
6ba3c97a FW |
469 | if (notify_die(DIE_GPF, "general protection fault", regs, error_code, |
470 | X86_TRAP_GP, SIGSEGV) != NOTIFY_STOP) | |
ef3f6288 | 471 | die("general protection fault", regs, error_code); |
6ba3c97a | 472 | goto exit; |
ef3f6288 | 473 | } |
1da177e4 | 474 | |
13485ab5 | 475 | tsk->thread.error_code = error_code; |
51e7dc70 | 476 | tsk->thread.trap_nr = X86_TRAP_GP; |
b5964405 | 477 | |
13485ab5 AH |
478 | if (show_unhandled_signals && unhandled_signal(tsk, SIGSEGV) && |
479 | printk_ratelimit()) { | |
c767a54b | 480 | pr_info("%s[%d] general protection ip:%lx sp:%lx error:%lx", |
13485ab5 AH |
481 | tsk->comm, task_pid_nr(tsk), |
482 | regs->ip, regs->sp, error_code); | |
03252919 | 483 | print_vma_addr(" in ", regs->ip); |
c767a54b | 484 | pr_cont("\n"); |
03252919 | 485 | } |
abd4f750 | 486 | |
38cad57b | 487 | force_sig_info(SIGSEGV, SEND_SIG_PRIV, tsk); |
6ba3c97a | 488 | exit: |
6c1e0256 | 489 | exception_exit(prev_state); |
1da177e4 | 490 | } |
9326638c | 491 | NOKPROBE_SYMBOL(do_general_protection); |
1da177e4 | 492 | |
c1d518c8 | 493 | /* May run on IST stack. */ |
9326638c | 494 | dotraplinkage void notrace do_int3(struct pt_regs *regs, long error_code) |
1da177e4 | 495 | { |
6c1e0256 FW |
496 | enum ctx_state prev_state; |
497 | ||
08d636b6 | 498 | #ifdef CONFIG_DYNAMIC_FTRACE |
a192cd04 SR |
499 | /* |
500 | * ftrace must be first, everything else may cause a recursive crash. | |
501 | * See note by declaration of modifying_ftrace_code in ftrace.c | |
502 | */ | |
503 | if (unlikely(atomic_read(&modifying_ftrace_code)) && | |
504 | ftrace_int3_handler(regs)) | |
08d636b6 SR |
505 | return; |
506 | #endif | |
17f41571 JK |
507 | if (poke_int3_handler(regs)) |
508 | return; | |
509 | ||
95927475 | 510 | prev_state = ist_enter(regs); |
f503b5ae | 511 | #ifdef CONFIG_KGDB_LOW_LEVEL_TRAP |
c9408265 KC |
512 | if (kgdb_ll_trap(DIE_INT3, "int3", regs, error_code, X86_TRAP_BP, |
513 | SIGTRAP) == NOTIFY_STOP) | |
6ba3c97a | 514 | goto exit; |
f503b5ae | 515 | #endif /* CONFIG_KGDB_LOW_LEVEL_TRAP */ |
cc3a1bf5 | 516 | |
6f6343f5 MH |
517 | #ifdef CONFIG_KPROBES |
518 | if (kprobe_int3_handler(regs)) | |
4cdf77a8 | 519 | goto exit; |
6f6343f5 MH |
520 | #endif |
521 | ||
c9408265 KC |
522 | if (notify_die(DIE_INT3, "int3", regs, error_code, X86_TRAP_BP, |
523 | SIGTRAP) == NOTIFY_STOP) | |
6ba3c97a | 524 | goto exit; |
b5964405 | 525 | |
42181186 SR |
526 | /* |
527 | * Let others (NMI) know that the debug stack is in use | |
528 | * as we may switch to the interrupt stack. | |
529 | */ | |
530 | debug_stack_usage_inc(); | |
4915a35e | 531 | preempt_conditional_sti(regs); |
c9408265 | 532 | do_trap(X86_TRAP_BP, SIGTRAP, "int3", regs, error_code, NULL); |
4915a35e | 533 | preempt_conditional_cli(regs); |
42181186 | 534 | debug_stack_usage_dec(); |
6ba3c97a | 535 | exit: |
95927475 | 536 | ist_exit(regs, prev_state); |
1da177e4 | 537 | } |
9326638c | 538 | NOKPROBE_SYMBOL(do_int3); |
1da177e4 | 539 | |
081f75bb | 540 | #ifdef CONFIG_X86_64 |
bd8b96df | 541 | /* |
48e08d0f AL |
542 | * Help handler running on IST stack to switch off the IST stack if the |
543 | * interrupted code was in user mode. The actual stack switch is done in | |
544 | * entry_64.S | |
bd8b96df | 545 | */ |
7ddc6a21 | 546 | asmlinkage __visible notrace struct pt_regs *sync_regs(struct pt_regs *eregs) |
081f75bb | 547 | { |
48e08d0f AL |
548 | struct pt_regs *regs = task_pt_regs(current); |
549 | *regs = *eregs; | |
081f75bb AH |
550 | return regs; |
551 | } | |
9326638c | 552 | NOKPROBE_SYMBOL(sync_regs); |
b645af2d AL |
553 | |
554 | struct bad_iret_stack { | |
555 | void *error_entry_ret; | |
556 | struct pt_regs regs; | |
557 | }; | |
558 | ||
7ddc6a21 | 559 | asmlinkage __visible notrace |
b645af2d AL |
560 | struct bad_iret_stack *fixup_bad_iret(struct bad_iret_stack *s) |
561 | { | |
562 | /* | |
563 | * This is called from entry_64.S early in handling a fault | |
564 | * caused by a bad iret to user mode. To handle the fault | |
565 | * correctly, we want move our stack frame to task_pt_regs | |
566 | * and we want to pretend that the exception came from the | |
567 | * iret target. | |
568 | */ | |
569 | struct bad_iret_stack *new_stack = | |
570 | container_of(task_pt_regs(current), | |
571 | struct bad_iret_stack, regs); | |
572 | ||
573 | /* Copy the IRET target to the new stack. */ | |
574 | memmove(&new_stack->regs.ip, (void *)s->regs.sp, 5*8); | |
575 | ||
576 | /* Copy the remainder of the stack from the current stack. */ | |
577 | memmove(new_stack, s, offsetof(struct bad_iret_stack, regs.ip)); | |
578 | ||
579 | BUG_ON(!user_mode_vm(&new_stack->regs)); | |
580 | return new_stack; | |
581 | } | |
7ddc6a21 | 582 | NOKPROBE_SYMBOL(fixup_bad_iret); |
081f75bb AH |
583 | #endif |
584 | ||
1da177e4 LT |
585 | /* |
586 | * Our handling of the processor debug registers is non-trivial. | |
587 | * We do not clear them on entry and exit from the kernel. Therefore | |
588 | * it is possible to get a watchpoint trap here from inside the kernel. | |
589 | * However, the code in ./ptrace.c has ensured that the user can | |
590 | * only set watchpoints on userspace addresses. Therefore the in-kernel | |
591 | * watchpoint trap can only occur in code which is reading/writing | |
592 | * from user space. Such code must not hold kernel locks (since it | |
593 | * can equally take a page fault), therefore it is safe to call | |
594 | * force_sig_info even though that claims and releases locks. | |
b5964405 | 595 | * |
1da177e4 LT |
596 | * Code in ./signal.c ensures that the debug control register |
597 | * is restored before we deliver any signal, and therefore that | |
598 | * user code runs with the correct debug control register even though | |
599 | * we clear it here. | |
600 | * | |
601 | * Being careful here means that we don't have to be as careful in a | |
602 | * lot of more complicated places (task switching can be a bit lazy | |
603 | * about restoring all the debug state, and ptrace doesn't have to | |
604 | * find every occurrence of the TF bit that could be saved away even | |
605 | * by user code) | |
c1d518c8 AH |
606 | * |
607 | * May run on IST stack. | |
1da177e4 | 608 | */ |
9326638c | 609 | dotraplinkage void do_debug(struct pt_regs *regs, long error_code) |
1da177e4 | 610 | { |
1da177e4 | 611 | struct task_struct *tsk = current; |
6c1e0256 | 612 | enum ctx_state prev_state; |
a1e80faf | 613 | int user_icebp = 0; |
08d68323 | 614 | unsigned long dr6; |
da654b74 | 615 | int si_code; |
1da177e4 | 616 | |
95927475 | 617 | prev_state = ist_enter(regs); |
4cdf77a8 | 618 | |
08d68323 | 619 | get_debugreg(dr6, 6); |
1da177e4 | 620 | |
40f9249a P |
621 | /* Filter out all the reserved bits which are preset to 1 */ |
622 | dr6 &= ~DR6_RESERVED; | |
623 | ||
a1e80faf FW |
624 | /* |
625 | * If dr6 has no reason to give us about the origin of this trap, | |
626 | * then it's very likely the result of an icebp/int01 trap. | |
627 | * User wants a sigtrap for that. | |
628 | */ | |
629 | if (!dr6 && user_mode(regs)) | |
630 | user_icebp = 1; | |
631 | ||
f8561296 | 632 | /* Catch kmemcheck conditions first of all! */ |
eadb8a09 | 633 | if ((dr6 & DR_STEP) && kmemcheck_trap(regs)) |
6ba3c97a | 634 | goto exit; |
f8561296 | 635 | |
08d68323 P |
636 | /* DR6 may or may not be cleared by the CPU */ |
637 | set_debugreg(0, 6); | |
10faa81e | 638 | |
ea8e61b7 PZ |
639 | /* |
640 | * The processor cleared BTF, so don't mark that we need it set. | |
641 | */ | |
642 | clear_tsk_thread_flag(tsk, TIF_BLOCKSTEP); | |
643 | ||
08d68323 P |
644 | /* Store the virtualized DR6 value */ |
645 | tsk->thread.debugreg6 = dr6; | |
646 | ||
6f6343f5 MH |
647 | #ifdef CONFIG_KPROBES |
648 | if (kprobe_debug_handler(regs)) | |
649 | goto exit; | |
650 | #endif | |
651 | ||
5a802e15 | 652 | if (notify_die(DIE_DEBUG, "debug", regs, (long)&dr6, error_code, |
62edab90 | 653 | SIGTRAP) == NOTIFY_STOP) |
6ba3c97a | 654 | goto exit; |
3d2a71a5 | 655 | |
42181186 SR |
656 | /* |
657 | * Let others (NMI) know that the debug stack is in use | |
658 | * as we may switch to the interrupt stack. | |
659 | */ | |
660 | debug_stack_usage_inc(); | |
661 | ||
1da177e4 | 662 | /* It's safe to allow irq's after DR6 has been saved */ |
3d2a71a5 | 663 | preempt_conditional_sti(regs); |
1da177e4 | 664 | |
08d68323 | 665 | if (regs->flags & X86_VM_MASK) { |
c9408265 KC |
666 | handle_vm86_trap((struct kernel_vm86_regs *) regs, error_code, |
667 | X86_TRAP_DB); | |
6554287b | 668 | preempt_conditional_cli(regs); |
42181186 | 669 | debug_stack_usage_dec(); |
6ba3c97a | 670 | goto exit; |
1da177e4 LT |
671 | } |
672 | ||
1da177e4 | 673 | /* |
08d68323 P |
674 | * Single-stepping through system calls: ignore any exceptions in |
675 | * kernel space, but re-enable TF when returning to user mode. | |
676 | * | |
677 | * We already checked v86 mode above, so we can check for kernel mode | |
678 | * by just checking the CPL of CS. | |
1da177e4 | 679 | */ |
08d68323 P |
680 | if ((dr6 & DR_STEP) && !user_mode(regs)) { |
681 | tsk->thread.debugreg6 &= ~DR_STEP; | |
682 | set_tsk_thread_flag(tsk, TIF_SINGLESTEP); | |
683 | regs->flags &= ~X86_EFLAGS_TF; | |
1da177e4 | 684 | } |
08d68323 | 685 | si_code = get_si_code(tsk->thread.debugreg6); |
a1e80faf | 686 | if (tsk->thread.debugreg6 & (DR_STEP | DR_TRAP_BITS) || user_icebp) |
08d68323 | 687 | send_sigtrap(tsk, regs, error_code, si_code); |
3d2a71a5 | 688 | preempt_conditional_cli(regs); |
42181186 | 689 | debug_stack_usage_dec(); |
1da177e4 | 690 | |
6ba3c97a | 691 | exit: |
95927475 | 692 | ist_exit(regs, prev_state); |
1da177e4 | 693 | } |
9326638c | 694 | NOKPROBE_SYMBOL(do_debug); |
1da177e4 LT |
695 | |
696 | /* | |
697 | * Note that we play around with the 'TS' bit in an attempt to get | |
698 | * the correct behaviour even in the presence of the asynchronous | |
699 | * IRQ13 behaviour | |
700 | */ | |
5e1b05be | 701 | static void math_error(struct pt_regs *regs, int error_code, int trapnr) |
1da177e4 | 702 | { |
e2e75c91 | 703 | struct task_struct *task = current; |
1da177e4 | 704 | siginfo_t info; |
9b6dba9e | 705 | unsigned short err; |
c9408265 KC |
706 | char *str = (trapnr == X86_TRAP_MF) ? "fpu exception" : |
707 | "simd exception"; | |
e2e75c91 BG |
708 | |
709 | if (notify_die(DIE_TRAP, str, regs, error_code, trapnr, SIGFPE) == NOTIFY_STOP) | |
710 | return; | |
711 | conditional_sti(regs); | |
712 | ||
713 | if (!user_mode_vm(regs)) | |
714 | { | |
715 | if (!fixup_exception(regs)) { | |
716 | task->thread.error_code = error_code; | |
51e7dc70 | 717 | task->thread.trap_nr = trapnr; |
e2e75c91 BG |
718 | die(str, regs, error_code); |
719 | } | |
720 | return; | |
721 | } | |
1da177e4 LT |
722 | |
723 | /* | |
724 | * Save the info for the exception handler and clear the error. | |
725 | */ | |
1da177e4 | 726 | save_init_fpu(task); |
51e7dc70 | 727 | task->thread.trap_nr = trapnr; |
9b6dba9e | 728 | task->thread.error_code = error_code; |
1da177e4 LT |
729 | info.si_signo = SIGFPE; |
730 | info.si_errno = 0; | |
b02ef20a | 731 | info.si_addr = (void __user *)uprobe_get_trap_addr(regs); |
c9408265 | 732 | if (trapnr == X86_TRAP_MF) { |
9b6dba9e BG |
733 | unsigned short cwd, swd; |
734 | /* | |
735 | * (~cwd & swd) will mask out exceptions that are not set to unmasked | |
736 | * status. 0x3f is the exception bits in these regs, 0x200 is the | |
737 | * C1 reg you need in case of a stack fault, 0x040 is the stack | |
738 | * fault bit. We should only be taking one exception at a time, | |
739 | * so if this combination doesn't produce any single exception, | |
740 | * then we have a bad program that isn't synchronizing its FPU usage | |
741 | * and it will suffer the consequences since we won't be able to | |
742 | * fully reproduce the context of the exception | |
743 | */ | |
744 | cwd = get_fpu_cwd(task); | |
745 | swd = get_fpu_swd(task); | |
adf77bac | 746 | |
9b6dba9e BG |
747 | err = swd & ~cwd; |
748 | } else { | |
749 | /* | |
750 | * The SIMD FPU exceptions are handled a little differently, as there | |
751 | * is only a single status/control register. Thus, to determine which | |
752 | * unmasked exception was caught we must mask the exception mask bits | |
753 | * at 0x1f80, and then use these to mask the exception bits at 0x3f. | |
754 | */ | |
755 | unsigned short mxcsr = get_fpu_mxcsr(task); | |
756 | err = ~(mxcsr >> 7) & mxcsr; | |
757 | } | |
adf77bac PA |
758 | |
759 | if (err & 0x001) { /* Invalid op */ | |
b5964405 IM |
760 | /* |
761 | * swd & 0x240 == 0x040: Stack Underflow | |
762 | * swd & 0x240 == 0x240: Stack Overflow | |
763 | * User must clear the SF bit (0x40) if set | |
764 | */ | |
765 | info.si_code = FPE_FLTINV; | |
adf77bac | 766 | } else if (err & 0x004) { /* Divide by Zero */ |
b5964405 | 767 | info.si_code = FPE_FLTDIV; |
adf77bac | 768 | } else if (err & 0x008) { /* Overflow */ |
b5964405 | 769 | info.si_code = FPE_FLTOVF; |
adf77bac PA |
770 | } else if (err & 0x012) { /* Denormal, Underflow */ |
771 | info.si_code = FPE_FLTUND; | |
772 | } else if (err & 0x020) { /* Precision */ | |
b5964405 | 773 | info.si_code = FPE_FLTRES; |
adf77bac | 774 | } else { |
bd8b96df | 775 | /* |
c9408265 KC |
776 | * If we're using IRQ 13, or supposedly even some trap |
777 | * X86_TRAP_MF implementations, it's possible | |
778 | * we get a spurious trap, which is not an error. | |
bd8b96df | 779 | */ |
c9408265 | 780 | return; |
1da177e4 LT |
781 | } |
782 | force_sig_info(SIGFPE, &info, task); | |
783 | } | |
784 | ||
e407d620 | 785 | dotraplinkage void do_coprocessor_error(struct pt_regs *regs, long error_code) |
1da177e4 | 786 | { |
6c1e0256 FW |
787 | enum ctx_state prev_state; |
788 | ||
789 | prev_state = exception_enter(); | |
c9408265 | 790 | math_error(regs, error_code, X86_TRAP_MF); |
6c1e0256 | 791 | exception_exit(prev_state); |
1da177e4 LT |
792 | } |
793 | ||
e407d620 AH |
794 | dotraplinkage void |
795 | do_simd_coprocessor_error(struct pt_regs *regs, long error_code) | |
1da177e4 | 796 | { |
6c1e0256 FW |
797 | enum ctx_state prev_state; |
798 | ||
799 | prev_state = exception_enter(); | |
c9408265 | 800 | math_error(regs, error_code, X86_TRAP_XF); |
6c1e0256 | 801 | exception_exit(prev_state); |
1da177e4 LT |
802 | } |
803 | ||
e407d620 AH |
804 | dotraplinkage void |
805 | do_spurious_interrupt_bug(struct pt_regs *regs, long error_code) | |
1da177e4 | 806 | { |
cf81978d | 807 | conditional_sti(regs); |
1da177e4 LT |
808 | #if 0 |
809 | /* No need to warn about this any longer. */ | |
c767a54b | 810 | pr_info("Ignoring P6 Local APIC Spurious Interrupt Bug...\n"); |
1da177e4 LT |
811 | #endif |
812 | } | |
813 | ||
2605fc21 | 814 | asmlinkage __visible void __attribute__((weak)) smp_thermal_interrupt(void) |
1da177e4 | 815 | { |
1da177e4 | 816 | } |
4efc0670 | 817 | |
2605fc21 | 818 | asmlinkage __visible void __attribute__((weak)) smp_threshold_interrupt(void) |
081f75bb AH |
819 | { |
820 | } | |
821 | ||
1da177e4 | 822 | /* |
b5964405 | 823 | * 'math_state_restore()' saves the current math information in the |
1da177e4 LT |
824 | * old math state array, and gets the new ones from the current task |
825 | * | |
826 | * Careful.. There are problems with IBM-designed IRQ13 behaviour. | |
827 | * Don't touch unless you *really* know how it works. | |
828 | * | |
be98c2cd LT |
829 | * Must be called with kernel preemption disabled (eg with local |
830 | * local interrupts as in the case of do_device_not_available). | |
1da177e4 | 831 | */ |
be98c2cd | 832 | void math_state_restore(void) |
1da177e4 | 833 | { |
f94edacf | 834 | struct task_struct *tsk = current; |
1da177e4 | 835 | |
aa283f49 SS |
836 | if (!tsk_used_math(tsk)) { |
837 | local_irq_enable(); | |
838 | /* | |
839 | * does a slab alloc which can sleep | |
840 | */ | |
841 | if (init_fpu(tsk)) { | |
842 | /* | |
843 | * ran out of memory! | |
844 | */ | |
845 | do_group_exit(SIGKILL); | |
846 | return; | |
847 | } | |
848 | local_irq_disable(); | |
849 | } | |
850 | ||
f94edacf | 851 | __thread_fpu_begin(tsk); |
304bceda | 852 | |
80ab6f1e LT |
853 | /* |
854 | * Paranoid restore. send a SIGSEGV if we fail to restore the state. | |
855 | */ | |
856 | if (unlikely(restore_fpu_checking(tsk))) { | |
304bceda | 857 | drop_init_fpu(tsk); |
38cad57b | 858 | force_sig_info(SIGSEGV, SEND_SIG_PRIV, tsk); |
80ab6f1e LT |
859 | return; |
860 | } | |
b3b0870e | 861 | |
c375f15a | 862 | tsk->thread.fpu_counter++; |
1da177e4 | 863 | } |
5992b6da | 864 | EXPORT_SYMBOL_GPL(math_state_restore); |
1da177e4 | 865 | |
9326638c | 866 | dotraplinkage void |
aa78bcfa | 867 | do_device_not_available(struct pt_regs *regs, long error_code) |
7643e9b9 | 868 | { |
6c1e0256 FW |
869 | enum ctx_state prev_state; |
870 | ||
871 | prev_state = exception_enter(); | |
5d2bd700 | 872 | BUG_ON(use_eager_fpu()); |
304bceda | 873 | |
a334fe43 | 874 | #ifdef CONFIG_MATH_EMULATION |
7643e9b9 | 875 | if (read_cr0() & X86_CR0_EM) { |
d315760f TH |
876 | struct math_emu_info info = { }; |
877 | ||
7643e9b9 | 878 | conditional_sti(regs); |
d315760f | 879 | |
aa78bcfa | 880 | info.regs = regs; |
d315760f | 881 | math_emulate(&info); |
6c1e0256 | 882 | exception_exit(prev_state); |
a334fe43 | 883 | return; |
7643e9b9 | 884 | } |
a334fe43 BG |
885 | #endif |
886 | math_state_restore(); /* interrupts still off */ | |
887 | #ifdef CONFIG_X86_32 | |
888 | conditional_sti(regs); | |
081f75bb | 889 | #endif |
6c1e0256 | 890 | exception_exit(prev_state); |
7643e9b9 | 891 | } |
9326638c | 892 | NOKPROBE_SYMBOL(do_device_not_available); |
7643e9b9 | 893 | |
081f75bb | 894 | #ifdef CONFIG_X86_32 |
e407d620 | 895 | dotraplinkage void do_iret_error(struct pt_regs *regs, long error_code) |
f8e0870f AH |
896 | { |
897 | siginfo_t info; | |
6c1e0256 | 898 | enum ctx_state prev_state; |
6ba3c97a | 899 | |
6c1e0256 | 900 | prev_state = exception_enter(); |
f8e0870f AH |
901 | local_irq_enable(); |
902 | ||
903 | info.si_signo = SIGILL; | |
904 | info.si_errno = 0; | |
905 | info.si_code = ILL_BADSTK; | |
fc6fcdfb | 906 | info.si_addr = NULL; |
c9408265 | 907 | if (notify_die(DIE_TRAP, "iret exception", regs, error_code, |
6ba3c97a FW |
908 | X86_TRAP_IRET, SIGILL) != NOTIFY_STOP) { |
909 | do_trap(X86_TRAP_IRET, SIGILL, "iret exception", regs, error_code, | |
910 | &info); | |
911 | } | |
6c1e0256 | 912 | exception_exit(prev_state); |
f8e0870f | 913 | } |
081f75bb | 914 | #endif |
f8e0870f | 915 | |
29c84391 JK |
916 | /* Set of traps needed for early debugging. */ |
917 | void __init early_trap_init(void) | |
918 | { | |
c9408265 | 919 | set_intr_gate_ist(X86_TRAP_DB, &debug, DEBUG_STACK); |
29c84391 | 920 | /* int3 can be called from all */ |
c9408265 | 921 | set_system_intr_gate_ist(X86_TRAP_BP, &int3, DEBUG_STACK); |
8170e6be | 922 | #ifdef CONFIG_X86_32 |
25c74b10 | 923 | set_intr_gate(X86_TRAP_PF, page_fault); |
8170e6be | 924 | #endif |
29c84391 JK |
925 | load_idt(&idt_descr); |
926 | } | |
927 | ||
8170e6be PA |
928 | void __init early_trap_pf_init(void) |
929 | { | |
930 | #ifdef CONFIG_X86_64 | |
25c74b10 | 931 | set_intr_gate(X86_TRAP_PF, page_fault); |
8170e6be PA |
932 | #endif |
933 | } | |
934 | ||
1da177e4 LT |
935 | void __init trap_init(void) |
936 | { | |
dbeb2be2 RR |
937 | int i; |
938 | ||
1da177e4 | 939 | #ifdef CONFIG_EISA |
927222b1 | 940 | void __iomem *p = early_ioremap(0x0FFFD9, 4); |
b5964405 IM |
941 | |
942 | if (readl(p) == 'E' + ('I'<<8) + ('S'<<16) + ('A'<<24)) | |
1da177e4 | 943 | EISA_bus = 1; |
927222b1 | 944 | early_iounmap(p, 4); |
1da177e4 LT |
945 | #endif |
946 | ||
25c74b10 | 947 | set_intr_gate(X86_TRAP_DE, divide_error); |
c9408265 | 948 | set_intr_gate_ist(X86_TRAP_NMI, &nmi, NMI_STACK); |
699d2937 | 949 | /* int4 can be called from all */ |
c9408265 | 950 | set_system_intr_gate(X86_TRAP_OF, &overflow); |
25c74b10 SA |
951 | set_intr_gate(X86_TRAP_BR, bounds); |
952 | set_intr_gate(X86_TRAP_UD, invalid_op); | |
953 | set_intr_gate(X86_TRAP_NM, device_not_available); | |
081f75bb | 954 | #ifdef CONFIG_X86_32 |
c9408265 | 955 | set_task_gate(X86_TRAP_DF, GDT_ENTRY_DOUBLEFAULT_TSS); |
081f75bb | 956 | #else |
c9408265 | 957 | set_intr_gate_ist(X86_TRAP_DF, &double_fault, DOUBLEFAULT_STACK); |
081f75bb | 958 | #endif |
25c74b10 SA |
959 | set_intr_gate(X86_TRAP_OLD_MF, coprocessor_segment_overrun); |
960 | set_intr_gate(X86_TRAP_TS, invalid_TSS); | |
961 | set_intr_gate(X86_TRAP_NP, segment_not_present); | |
6f442be2 | 962 | set_intr_gate(X86_TRAP_SS, stack_segment); |
25c74b10 SA |
963 | set_intr_gate(X86_TRAP_GP, general_protection); |
964 | set_intr_gate(X86_TRAP_SPURIOUS, spurious_interrupt_bug); | |
965 | set_intr_gate(X86_TRAP_MF, coprocessor_error); | |
966 | set_intr_gate(X86_TRAP_AC, alignment_check); | |
1da177e4 | 967 | #ifdef CONFIG_X86_MCE |
c9408265 | 968 | set_intr_gate_ist(X86_TRAP_MC, &machine_check, MCE_STACK); |
1da177e4 | 969 | #endif |
25c74b10 | 970 | set_intr_gate(X86_TRAP_XF, simd_coprocessor_error); |
1da177e4 | 971 | |
bb3f0b59 YL |
972 | /* Reserve all the builtin and the syscall vector: */ |
973 | for (i = 0; i < FIRST_EXTERNAL_VECTOR; i++) | |
974 | set_bit(i, used_vectors); | |
975 | ||
081f75bb AH |
976 | #ifdef CONFIG_IA32_EMULATION |
977 | set_system_intr_gate(IA32_SYSCALL_VECTOR, ia32_syscall); | |
bb3f0b59 | 978 | set_bit(IA32_SYSCALL_VECTOR, used_vectors); |
081f75bb AH |
979 | #endif |
980 | ||
981 | #ifdef CONFIG_X86_32 | |
699d2937 | 982 | set_system_trap_gate(SYSCALL_VECTOR, &system_call); |
dbeb2be2 | 983 | set_bit(SYSCALL_VECTOR, used_vectors); |
081f75bb | 984 | #endif |
bb3f0b59 | 985 | |
4eefbe79 KC |
986 | /* |
987 | * Set the IDT descriptor to a fixed read-only location, so that the | |
988 | * "sidt" instruction will not leak the location of the kernel, and | |
989 | * to defend the IDT against arbitrary memory write vulnerabilities. | |
990 | * It will be reloaded in cpu_init() */ | |
991 | __set_fixmap(FIX_RO_IDT, __pa_symbol(idt_table), PAGE_KERNEL_RO); | |
992 | idt_descr.address = fix_to_virt(FIX_RO_IDT); | |
993 | ||
1da177e4 | 994 | /* |
b5964405 | 995 | * Should be a barrier for any external CPU state: |
1da177e4 LT |
996 | */ |
997 | cpu_init(); | |
998 | ||
428cf902 | 999 | x86_init.irqs.trap_init(); |
228bdaa9 SR |
1000 | |
1001 | #ifdef CONFIG_X86_64 | |
629f4f9d | 1002 | memcpy(&debug_idt_table, &idt_table, IDT_ENTRIES * 16); |
c9408265 KC |
1003 | set_nmi_gate(X86_TRAP_DB, &debug); |
1004 | set_nmi_gate(X86_TRAP_BP, &int3); | |
228bdaa9 | 1005 | #endif |
1da177e4 | 1006 | } |