process_64.c 16.9 KB
Newer Older
L
Linus Torvalds 已提交
1 2 3 4 5
/*
 *  Copyright (C) 1995  Linus Torvalds
 *
 *  Pentium III FXSR, SSE support
 *	Gareth Hughes <gareth@valinux.com>, May 2000
6
 *
L
Linus Torvalds 已提交
7 8
 *  X86-64 port
 *	Andi Kleen.
A
Ashok Raj 已提交
9 10
 *
 *	CPU hotplug support - ashok.raj@intel.com
L
Linus Torvalds 已提交
11 12 13 14 15 16
 */

/*
 * This file handles the architecture-dependent parts of process handling..
 */

A
Ashok Raj 已提交
17
#include <linux/cpu.h>
L
Linus Torvalds 已提交
18 19
#include <linux/errno.h>
#include <linux/sched.h>
20
#include <linux/fs.h>
L
Linus Torvalds 已提交
21 22 23 24 25 26 27 28
#include <linux/kernel.h>
#include <linux/mm.h>
#include <linux/elfcore.h>
#include <linux/smp.h>
#include <linux/slab.h>
#include <linux/user.h>
#include <linux/interrupt.h>
#include <linux/delay.h>
29
#include <linux/export.h>
L
Linus Torvalds 已提交
30
#include <linux/ptrace.h>
A
Andi Kleen 已提交
31
#include <linux/notifier.h>
32
#include <linux/kprobes.h>
33
#include <linux/kdebug.h>
34
#include <linux/prctl.h>
35 36
#include <linux/uaccess.h>
#include <linux/io.h>
37
#include <linux/ftrace.h>
L
Linus Torvalds 已提交
38 39 40

#include <asm/pgtable.h>
#include <asm/processor.h>
41
#include <asm/fpu/internal.h>
L
Linus Torvalds 已提交
42 43 44 45 46
#include <asm/mmu_context.h>
#include <asm/prctl.h>
#include <asm/desc.h>
#include <asm/proto.h>
#include <asm/ia32.h>
A
Andi Kleen 已提交
47
#include <asm/idle.h>
48
#include <asm/syscalls.h>
49
#include <asm/debugreg.h>
50
#include <asm/switch_to.h>
51
#include <asm/xen/hypervisor.h>
52
#include <asm/vdso.h>
L
Linus Torvalds 已提交
53

54
__visible DEFINE_PER_CPU(unsigned long, rsp_scratch);
L
Linus Torvalds 已提交
55

56
/* Prints also some state that isn't saved in the pt_regs */
57
void __show_regs(struct pt_regs *regs, int all)
L
Linus Torvalds 已提交
58 59
{
	unsigned long cr0 = 0L, cr2 = 0L, cr3 = 0L, cr4 = 0L, fs, gs, shadowgs;
60
	unsigned long d0, d1, d2, d3, d6, d7;
61 62
	unsigned int fsindex, gsindex;
	unsigned int ds, cs, es;
63

64
	printk(KERN_DEFAULT "RIP: %04lx:[<%016lx>] ", regs->cs & 0xffff, regs->ip);
65
	printk_address(regs->ip);
66
	printk(KERN_DEFAULT "RSP: %04lx:%016lx  EFLAGS: %08lx\n", regs->ss,
67
			regs->sp, regs->flags);
68
	printk(KERN_DEFAULT "RAX: %016lx RBX: %016lx RCX: %016lx\n",
69
	       regs->ax, regs->bx, regs->cx);
70
	printk(KERN_DEFAULT "RDX: %016lx RSI: %016lx RDI: %016lx\n",
71
	       regs->dx, regs->si, regs->di);
72
	printk(KERN_DEFAULT "RBP: %016lx R08: %016lx R09: %016lx\n",
73
	       regs->bp, regs->r8, regs->r9);
74
	printk(KERN_DEFAULT "R10: %016lx R11: %016lx R12: %016lx\n",
75
	       regs->r10, regs->r11, regs->r12);
76
	printk(KERN_DEFAULT "R13: %016lx R14: %016lx R15: %016lx\n",
77
	       regs->r13, regs->r14, regs->r15);
L
Linus Torvalds 已提交
78

79 80 81
	asm("movl %%ds,%0" : "=r" (ds));
	asm("movl %%cs,%0" : "=r" (cs));
	asm("movl %%es,%0" : "=r" (es));
L
Linus Torvalds 已提交
82 83 84 85
	asm("movl %%fs,%0" : "=r" (fsindex));
	asm("movl %%gs,%0" : "=r" (gsindex));

	rdmsrl(MSR_FS_BASE, fs);
86 87
	rdmsrl(MSR_GS_BASE, gs);
	rdmsrl(MSR_KERNEL_GS_BASE, shadowgs);
L
Linus Torvalds 已提交
88

89 90
	if (!all)
		return;
L
Linus Torvalds 已提交
91

92 93 94
	cr0 = read_cr0();
	cr2 = read_cr2();
	cr3 = read_cr3();
95
	cr4 = __read_cr4();
L
Linus Torvalds 已提交
96

97
	printk(KERN_DEFAULT "FS:  %016lx(%04x) GS:%016lx(%04x) knlGS:%016lx\n",
98
	       fs, fsindex, gs, gsindex, shadowgs);
99
	printk(KERN_DEFAULT "CS:  %04x DS: %04x ES: %04x CR0: %016lx\n", cs, ds,
100
			es, cr0);
101
	printk(KERN_DEFAULT "CR2: %016lx CR3: %016lx CR4: %016lx\n", cr2, cr3,
102
			cr4);
103 104 105 106 107 108 109

	get_debugreg(d0, 0);
	get_debugreg(d1, 1);
	get_debugreg(d2, 2);
	get_debugreg(d3, 3);
	get_debugreg(d6, 6);
	get_debugreg(d7, 7);
110 111 112 113 114 115 116

	/* Only print out debug registers if they are in their non-default state. */
	if ((d0 == 0) && (d1 == 0) && (d2 == 0) && (d3 == 0) &&
	    (d6 == DR6_RESERVED) && (d7 == 0x400))
		return;

	printk(KERN_DEFAULT "DR0: %016lx DR1: %016lx DR2: %016lx\n", d0, d1, d2);
117
	printk(KERN_DEFAULT "DR3: %016lx DR6: %016lx DR7: %016lx\n", d3, d6, d7);
118

119 120
	if (boot_cpu_has(X86_FEATURE_OSPKE))
		printk(KERN_DEFAULT "PKRU: %08x\n", read_pkru());
L
Linus Torvalds 已提交
121 122 123 124 125
}

void release_thread(struct task_struct *dead_task)
{
	if (dead_task->mm) {
126
#ifdef CONFIG_MODIFY_LDT_SYSCALL
127
		if (dead_task->mm->context.ldt) {
128
			pr_warn("WARNING: dead process %s still has LDT? <%p/%d>\n",
129
				dead_task->comm,
130
				dead_task->mm->context.ldt->entries,
131
				dead_task->mm->context.ldt->size);
L
Linus Torvalds 已提交
132 133
			BUG();
		}
134
#endif
L
Linus Torvalds 已提交
135 136 137
	}
}

138 139
int copy_thread_tls(unsigned long clone_flags, unsigned long sp,
		unsigned long arg, struct task_struct *p, unsigned long tls)
L
Linus Torvalds 已提交
140 141
{
	int err;
142
	struct pt_regs *childregs;
143 144
	struct fork_frame *fork_frame;
	struct inactive_task_frame *frame;
L
Linus Torvalds 已提交
145 146
	struct task_struct *me = current;

A
Al Viro 已提交
147 148
	p->thread.sp0 = (unsigned long)task_stack_page(p) + THREAD_SIZE;
	childregs = task_pt_regs(p);
149 150 151 152 153
	fork_frame = container_of(childregs, struct fork_frame, regs);
	frame = &fork_frame->frame;
	frame->bp = 0;
	frame->ret_addr = (unsigned long) ret_from_fork;
	p->thread.sp = (unsigned long) fork_frame;
154
	p->thread.io_bitmap_ptr = NULL;
L
Linus Torvalds 已提交
155

156
	savesegment(gs, p->thread.gsindex);
157
	p->thread.gsbase = p->thread.gsindex ? 0 : me->thread.gsbase;
158
	savesegment(fs, p->thread.fsindex);
159
	p->thread.fsbase = p->thread.fsindex ? 0 : me->thread.fsbase;
160 161
	savesegment(es, p->thread.es);
	savesegment(ds, p->thread.ds);
A
Al Viro 已提交
162 163
	memset(p->thread.ptrace_bps, 0, sizeof(p->thread.ptrace_bps));

164
	if (unlikely(p->flags & PF_KTHREAD)) {
A
Al Viro 已提交
165 166
		/* kernel thread */
		memset(childregs, 0, sizeof(struct pt_regs));
167 168
		frame->bx = sp;		/* function */
		frame->r12 = arg;
A
Al Viro 已提交
169 170
		return 0;
	}
171
	frame->bx = 0;
172
	*childregs = *current_pt_regs();
A
Al Viro 已提交
173 174

	childregs->ax = 0;
175 176
	if (sp)
		childregs->sp = sp;
L
Linus Torvalds 已提交
177

178
	err = -ENOMEM;
179
	if (unlikely(test_tsk_thread_flag(me, TIF_IO_BITMAP))) {
180 181
		p->thread.io_bitmap_ptr = kmemdup(me->thread.io_bitmap_ptr,
						  IO_BITMAP_BYTES, GFP_KERNEL);
L
Linus Torvalds 已提交
182 183 184 185
		if (!p->thread.io_bitmap_ptr) {
			p->thread.io_bitmap_max = 0;
			return -ENOMEM;
		}
186
		set_tsk_thread_flag(p, TIF_IO_BITMAP);
187
	}
L
Linus Torvalds 已提交
188 189 190 191 192 193

	/*
	 * Set a new TLS for the child thread?
	 */
	if (clone_flags & CLONE_SETTLS) {
#ifdef CONFIG_IA32_EMULATION
194
		if (in_ia32_syscall())
R
Roland McGrath 已提交
195
			err = do_set_thread_area(p, -1,
196
				(struct user_desc __user *)tls, 0);
197 198
		else
#endif
199
			err = do_arch_prctl(p, ARCH_SET_FS, tls);
200
		if (err)
L
Linus Torvalds 已提交
201 202 203 204 205 206 207 208
			goto out;
	}
	err = 0;
out:
	if (err && p->thread.io_bitmap_ptr) {
		kfree(p->thread.io_bitmap_ptr);
		p->thread.io_bitmap_max = 0;
	}
209

L
Linus Torvalds 已提交
210 211 212
	return err;
}

213 214 215 216
static void
start_thread_common(struct pt_regs *regs, unsigned long new_ip,
		    unsigned long new_sp,
		    unsigned int _cs, unsigned int _ss, unsigned int _ds)
I
Ingo Molnar 已提交
217
{
218
	loadsegment(fs, 0);
219 220
	loadsegment(es, _ds);
	loadsegment(ds, _ds);
I
Ingo Molnar 已提交
221 222 223
	load_gs_index(0);
	regs->ip		= new_ip;
	regs->sp		= new_sp;
224 225
	regs->cs		= _cs;
	regs->ss		= _ss;
226
	regs->flags		= X86_EFLAGS_IF;
227
	force_iret();
I
Ingo Molnar 已提交
228
}
229 230 231 232 233 234 235

void
start_thread(struct pt_regs *regs, unsigned long new_ip, unsigned long new_sp)
{
	start_thread_common(regs, new_ip, new_sp,
			    __USER_CS, __USER_DS, 0);
}
I
Ingo Molnar 已提交
236

237 238
#ifdef CONFIG_COMPAT
void compat_start_thread(struct pt_regs *regs, u32 new_ip, u32 new_sp)
239
{
240
	start_thread_common(regs, new_ip, new_sp,
H
H. Peter Anvin 已提交
241 242 243
			    test_thread_flag(TIF_X32)
			    ? __USER_CS : __USER32_CS,
			    __USER_DS, __USER_DS);
244 245
}
#endif
I
Ingo Molnar 已提交
246

L
Linus Torvalds 已提交
247 248 249
/*
 *	switch_to(x,y) should switch tasks from x to y.
 *
250
 * This could still be optimized:
L
Linus Torvalds 已提交
251 252
 * - fold all the options into a flag word and test it with a single test.
 * - could test fs/gs bitsliced
253 254
 *
 * Kprobes not supported here. Set the probe on schedule instead.
255
 * Function graph tracer not supported too.
L
Linus Torvalds 已提交
256
 */
257
__visible __notrace_funcgraph struct task_struct *
258
__switch_to(struct task_struct *prev_p, struct task_struct *next_p)
L
Linus Torvalds 已提交
259
{
260 261
	struct thread_struct *prev = &prev_p->thread;
	struct thread_struct *next = &next_p->thread;
262 263
	struct fpu *prev_fpu = &prev->fpu;
	struct fpu *next_fpu = &next->fpu;
264
	int cpu = smp_processor_id();
265
	struct tss_struct *tss = &per_cpu(cpu_tss, cpu);
266
	unsigned prev_fsindex, prev_gsindex;
267
	fpu_switch_t fpu_switch;
268

269
	fpu_switch = switch_fpu_prepare(prev_fpu, next_fpu, cpu);
270

271 272 273 274 275
	/* We must save %fs and %gs before load_TLS() because
	 * %fs and %gs may be cleared by load_TLS().
	 *
	 * (e.g. xen_load_tls())
	 */
276 277
	savesegment(fs, prev_fsindex);
	savesegment(gs, prev_gsindex);
278

279 280 281 282
	/*
	 * Load TLS before restoring any segments so that segment loads
	 * reference the correct GDT entries.
	 */
L
Linus Torvalds 已提交
283 284
	load_TLS(next, cpu);

285
	/*
286 287 288
	 * Leave lazy mode, flushing any hypercalls made here.  This
	 * must be done after loading TLS entries in the GDT but before
	 * loading segments that might reference them, and and it must
289
	 * be done before fpu__restore(), so the TS bit is up to
290
	 * date.
291
	 */
292
	arch_end_context_switch(next_p);
293

294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315
	/* Switch DS and ES.
	 *
	 * Reading them only returns the selectors, but writing them (if
	 * nonzero) loads the full descriptor from the GDT or LDT.  The
	 * LDT for next is loaded in switch_mm, and the GDT is loaded
	 * above.
	 *
	 * We therefore need to write new values to the segment
	 * registers on every context switch unless both the new and old
	 * values are zero.
	 *
	 * Note that we don't need to do anything for CS and SS, as
	 * those are saved and restored as part of pt_regs.
	 */
	savesegment(es, prev->es);
	if (unlikely(next->es | prev->es))
		loadsegment(es, next->es);

	savesegment(ds, prev->ds);
	if (unlikely(next->ds | prev->ds))
		loadsegment(ds, next->ds);

316
	/*
L
Linus Torvalds 已提交
317
	 * Switch FS and GS.
318
	 *
319
	 * These are even more complicated than DS and ES: they have
320 321 322
	 * 64-bit bases are that controlled by arch_prctl.  The bases
	 * don't necessarily match the selectors, as user code can do
	 * any number of things to cause them to be inconsistent.
323
	 *
324 325 326 327 328 329
	 * We don't promise to preserve the bases if the selectors are
	 * nonzero.  We also don't promise to preserve the base if the
	 * selector is zero and the base doesn't match whatever was
	 * most recently passed to ARCH_SET_FS/GS.  (If/when the
	 * FSGSBASE instructions are enabled, we'll need to offer
	 * stronger guarantees.)
330
	 *
331
	 * As an invariant,
332
	 * (fsbase != 0 && fsindex != 0) || (gsbase != 0 && gsindex != 0) is
333
	 * impossible.
L
Linus Torvalds 已提交
334
	 */
335 336
	if (next->fsindex) {
		/* Loading a nonzero value into FS sets the index and base. */
337
		loadsegment(fs, next->fsindex);
338
	} else {
339
		if (next->fsbase) {
340 341 342
			/* Next index is zero but next base is nonzero. */
			if (prev_fsindex)
				loadsegment(fs, 0);
343
			wrmsrl(MSR_FS_BASE, next->fsbase);
344 345 346 347 348 349 350 351 352 353 354 355 356 357 358
		} else {
			/* Next base and index are both zero. */
			if (static_cpu_has_bug(X86_BUG_NULL_SEG)) {
				/*
				 * We don't know the previous base and can't
				 * find out without RDMSR.  Forcibly clear it.
				 */
				loadsegment(fs, __USER_DS);
				loadsegment(fs, 0);
			} else {
				/*
				 * If the previous index is zero and ARCH_SET_FS
				 * didn't change the base, then the base is
				 * also zero and we don't need to do anything.
				 */
359
				if (prev->fsbase || prev_fsindex)
360 361 362
					loadsegment(fs, 0);
			}
		}
L
Linus Torvalds 已提交
363
	}
364 365 366 367 368 369 370 371
	/*
	 * Save the old state and preserve the invariant.
	 * NB: if prev_fsindex == 0, then we can't reliably learn the base
	 * without RDMSR because Intel user code can zero it without telling
	 * us and AMD user code can program any 32-bit value without telling
	 * us.
	 */
	if (prev_fsindex)
372
		prev->fsbase = 0;
373
	prev->fsindex = prev_fsindex;
374

375 376
	if (next->gsindex) {
		/* Loading a nonzero value into GS sets the index and base. */
377
		load_gs_index(next->gsindex);
378
	} else {
379
		if (next->gsbase) {
380 381 382
			/* Next index is zero but next base is nonzero. */
			if (prev_gsindex)
				load_gs_index(0);
383
			wrmsrl(MSR_KERNEL_GS_BASE, next->gsbase);
384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402
		} else {
			/* Next base and index are both zero. */
			if (static_cpu_has_bug(X86_BUG_NULL_SEG)) {
				/*
				 * We don't know the previous base and can't
				 * find out without RDMSR.  Forcibly clear it.
				 *
				 * This contains a pointless SWAPGS pair.
				 * Fixing it would involve an explicit check
				 * for Xen or a new pvop.
				 */
				load_gs_index(__USER_DS);
				load_gs_index(0);
			} else {
				/*
				 * If the previous index is zero and ARCH_SET_GS
				 * didn't change the base, then the base is
				 * also zero and we don't need to do anything.
				 */
403
				if (prev->gsbase || prev_gsindex)
404 405 406
					load_gs_index(0);
			}
		}
L
Linus Torvalds 已提交
407
	}
408 409 410 411 412 413 414 415
	/*
	 * Save the old state and preserve the invariant.
	 * NB: if prev_gsindex == 0, then we can't reliably learn the base
	 * without RDMSR because Intel user code can zero it without telling
	 * us and AMD user code can program any 32-bit value without telling
	 * us.
	 */
	if (prev_gsindex)
416
		prev->gsbase = 0;
417
	prev->gsindex = prev_gsindex;
L
Linus Torvalds 已提交
418

419
	switch_fpu_finish(next_fpu, fpu_switch);
420

421
	/*
422
	 * Switch the PDA and FPU contexts.
L
Linus Torvalds 已提交
423
	 */
424
	this_cpu_write(current_task, next_p);
425

426 427 428
	/* Reload esp0 and ss1.  This changes current_thread_info(). */
	load_sp0(tss, next);

L
Linus Torvalds 已提交
429
	/*
430
	 * Now maybe reload the debug registers and handle I/O bitmaps
L
Linus Torvalds 已提交
431
	 */
432 433
	if (unlikely(task_thread_info(next_p)->flags & _TIF_WORK_CTXSW_NEXT ||
		     task_thread_info(prev_p)->flags & _TIF_WORK_CTXSW_PREV))
434
		__switch_to_xtra(prev_p, next_p, tss);
L
Linus Torvalds 已提交
435

436 437 438 439 440 441 442 443 444 445 446
#ifdef CONFIG_XEN
	/*
	 * On Xen PV, IOPL bits in pt_regs->flags have no effect, and
	 * current_pt_regs()->flags may not match the current task's
	 * intended IOPL.  We need to switch it manually.
	 */
	if (unlikely(static_cpu_has(X86_FEATURE_XENPV) &&
		     prev->iopl != next->iopl))
		xen_set_iopl_mask(next->iopl);
#endif

447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474
	if (static_cpu_has_bug(X86_BUG_SYSRET_SS_ATTRS)) {
		/*
		 * AMD CPUs have a misfeature: SYSRET sets the SS selector but
		 * does not update the cached descriptor.  As a result, if we
		 * do SYSRET while SS is NULL, we'll end up in user mode with
		 * SS apparently equal to __USER_DS but actually unusable.
		 *
		 * The straightforward workaround would be to fix it up just
		 * before SYSRET, but that would slow down the system call
		 * fast paths.  Instead, we ensure that SS is never NULL in
		 * system call context.  We do this by replacing NULL SS
		 * selectors at every context switch.  SYSCALL sets up a valid
		 * SS, so the only way to get NULL is to re-enter the kernel
		 * from CPL 3 through an interrupt.  Since that can't happen
		 * in the same task as a running syscall, we are guaranteed to
		 * context switch between every interrupt vector entry and a
		 * subsequent SYSRET.
		 *
		 * We read SS first because SS reads are much faster than
		 * writes.  Out of caution, we force SS to __KERNEL_DS even if
		 * it previously had a different non-NULL value.
		 */
		unsigned short ss_sel;
		savesegment(ss, ss_sel);
		if (ss_sel != __KERNEL_DS)
			loadsegment(ss, __KERNEL_DS);
	}

L
Linus Torvalds 已提交
475 476 477 478 479 480 481 482
	return prev_p;
}

void set_personality_64bit(void)
{
	/* inherit personality from parent */

	/* Make sure to be in 64bit mode */
483
	clear_thread_flag(TIF_IA32);
484
	clear_thread_flag(TIF_ADDR32);
485
	clear_thread_flag(TIF_X32);
L
Linus Torvalds 已提交
486

487 488 489 490
	/* Ensure the corresponding mm is not marked. */
	if (current->mm)
		current->mm->context.ia32_compat = 0;

L
Linus Torvalds 已提交
491 492 493
	/* TBD: overwrites user setup. Should have two bits.
	   But 64bit processes have always behaved this way,
	   so it's not too bad. The main problem is just that
494
	   32bit childs are affected again. */
L
Linus Torvalds 已提交
495 496 497
	current->personality &= ~READ_IMPLIES_EXEC;
}

H
H. Peter Anvin 已提交
498
void set_personality_ia32(bool x32)
499 500 501 502
{
	/* inherit personality from parent */

	/* Make sure to be in 32bit mode */
503
	set_thread_flag(TIF_ADDR32);
504

505
	/* Mark the associated mm as containing 32-bit tasks. */
H
H. Peter Anvin 已提交
506 507 508
	if (x32) {
		clear_thread_flag(TIF_IA32);
		set_thread_flag(TIF_X32);
509 510
		if (current->mm)
			current->mm->context.ia32_compat = TIF_X32;
H
H. Peter Anvin 已提交
511
		current->personality &= ~READ_IMPLIES_EXEC;
512
		/* in_compat_syscall() uses the presence of the x32
513
		   syscall bit flag to determine compat status */
514
		current->thread.status &= ~TS_COMPAT;
H
H. Peter Anvin 已提交
515 516 517
	} else {
		set_thread_flag(TIF_IA32);
		clear_thread_flag(TIF_X32);
518 519
		if (current->mm)
			current->mm->context.ia32_compat = TIF_IA32;
H
H. Peter Anvin 已提交
520 521
		current->personality |= force_personality32;
		/* Prepare the first "return" to user space */
522
		current->thread.status |= TS_COMPAT;
H
H. Peter Anvin 已提交
523
	}
524
}
525
EXPORT_SYMBOL_GPL(set_personality_ia32);
526

527
#ifdef CONFIG_CHECKPOINT_RESTORE
528 529 530 531 532 533 534 535 536 537
static long prctl_map_vdso(const struct vdso_image *image, unsigned long addr)
{
	int ret;

	ret = map_vdso_once(image, addr);
	if (ret)
		return ret;

	return (long)image->size;
}
538
#endif
539

L
Linus Torvalds 已提交
540
long do_arch_prctl(struct task_struct *task, int code, unsigned long addr)
541 542
{
	int ret = 0;
L
Linus Torvalds 已提交
543 544 545
	int doit = task == current;
	int cpu;

546
	switch (code) {
L
Linus Torvalds 已提交
547
	case ARCH_SET_GS:
548
		if (addr >= TASK_SIZE_MAX)
549
			return -EPERM;
L
Linus Torvalds 已提交
550
		cpu = get_cpu();
551
		task->thread.gsindex = 0;
552
		task->thread.gsbase = addr;
553 554 555
		if (doit) {
			load_gs_index(0);
			ret = wrmsrl_safe(MSR_KERNEL_GS_BASE, addr);
L
Linus Torvalds 已提交
556
		}
557
		put_cpu();
L
Linus Torvalds 已提交
558 559 560 561
		break;
	case ARCH_SET_FS:
		/* Not strictly needed for fs, but do it for symmetry
		   with gs */
562
		if (addr >= TASK_SIZE_MAX)
563
			return -EPERM;
L
Linus Torvalds 已提交
564
		cpu = get_cpu();
565
		task->thread.fsindex = 0;
566
		task->thread.fsbase = addr;
567 568 569 570
		if (doit) {
			/* set the selector to 0 to not confuse __switch_to */
			loadsegment(fs, 0);
			ret = wrmsrl_safe(MSR_FS_BASE, addr);
L
Linus Torvalds 已提交
571 572 573
		}
		put_cpu();
		break;
574 575
	case ARCH_GET_FS: {
		unsigned long base;
576
		if (doit)
L
Linus Torvalds 已提交
577
			rdmsrl(MSR_FS_BASE, base);
578
		else
579
			base = task->thread.fsbase;
580 581
		ret = put_user(base, (unsigned long __user *)addr);
		break;
L
Linus Torvalds 已提交
582
	}
583
	case ARCH_GET_GS: {
L
Linus Torvalds 已提交
584
		unsigned long base;
585 586 587
		if (doit)
			rdmsrl(MSR_KERNEL_GS_BASE, base);
		else
588
			base = task->thread.gsbase;
589
		ret = put_user(base, (unsigned long __user *)addr);
L
Linus Torvalds 已提交
590 591 592
		break;
	}

593
#ifdef CONFIG_CHECKPOINT_RESTORE
594
# ifdef CONFIG_X86_X32_ABI
595 596
	case ARCH_MAP_VDSO_X32:
		return prctl_map_vdso(&vdso_image_x32, addr);
597 598
# endif
# if defined CONFIG_X86_32 || defined CONFIG_IA32_EMULATION
599 600
	case ARCH_MAP_VDSO_32:
		return prctl_map_vdso(&vdso_image_32, addr);
601
# endif
602 603 604 605
	case ARCH_MAP_VDSO_64:
		return prctl_map_vdso(&vdso_image_64, addr);
#endif

L
Linus Torvalds 已提交
606 607 608
	default:
		ret = -EINVAL;
		break;
609
	}
L
Linus Torvalds 已提交
610

611 612
	return ret;
}
L
Linus Torvalds 已提交
613 614 615 616 617 618

long sys_arch_prctl(int code, unsigned long addr)
{
	return do_arch_prctl(current, code, addr);
}

619 620
unsigned long KSTK_ESP(struct task_struct *task)
{
621
	return task_pt_regs(task)->sp;
622
}