xref: /openbmc/linux/arch/ia64/kernel/process.c (revision baa7eb025ab14f3cba2e35c0a8648f9c9f01d24f)
1 /*
2  * Architecture-specific setup.
3  *
4  * Copyright (C) 1998-2003 Hewlett-Packard Co
5  *	David Mosberger-Tang <davidm@hpl.hp.com>
6  * 04/11/17 Ashok Raj	<ashok.raj@intel.com> Added CPU Hotplug Support
7  *
8  * 2005-10-07 Keith Owens <kaos@sgi.com>
9  *	      Add notify_die() hooks.
10  */
11 #include <linux/cpu.h>
12 #include <linux/pm.h>
13 #include <linux/elf.h>
14 #include <linux/errno.h>
15 #include <linux/kallsyms.h>
16 #include <linux/kernel.h>
17 #include <linux/mm.h>
18 #include <linux/slab.h>
19 #include <linux/module.h>
20 #include <linux/notifier.h>
21 #include <linux/personality.h>
22 #include <linux/sched.h>
23 #include <linux/stddef.h>
24 #include <linux/thread_info.h>
25 #include <linux/unistd.h>
26 #include <linux/efi.h>
27 #include <linux/interrupt.h>
28 #include <linux/delay.h>
29 #include <linux/kdebug.h>
30 #include <linux/utsname.h>
31 #include <linux/tracehook.h>
32 
33 #include <asm/cpu.h>
34 #include <asm/delay.h>
35 #include <asm/elf.h>
36 #include <asm/irq.h>
37 #include <asm/kexec.h>
38 #include <asm/pgalloc.h>
39 #include <asm/processor.h>
40 #include <asm/sal.h>
41 #include <asm/tlbflush.h>
42 #include <asm/uaccess.h>
43 #include <asm/unwind.h>
44 #include <asm/user.h>
45 
46 #include "entry.h"
47 
48 #ifdef CONFIG_PERFMON
49 # include <asm/perfmon.h>
50 #endif
51 
52 #include "sigframe.h"
53 
54 void (*ia64_mark_idle)(int);
55 
56 unsigned long boot_option_idle_override = 0;
57 EXPORT_SYMBOL(boot_option_idle_override);
58 unsigned long idle_halt;
59 EXPORT_SYMBOL(idle_halt);
60 unsigned long idle_nomwait;
61 EXPORT_SYMBOL(idle_nomwait);
62 void (*pm_idle) (void);
63 EXPORT_SYMBOL(pm_idle);
64 void (*pm_power_off) (void);
65 EXPORT_SYMBOL(pm_power_off);
66 
67 void
68 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
69 {
70 	unsigned long ip, sp, bsp;
71 	char buf[128];			/* don't make it so big that it overflows the stack! */
72 
73 	printk("\nCall Trace:\n");
74 	do {
75 		unw_get_ip(info, &ip);
76 		if (ip == 0)
77 			break;
78 
79 		unw_get_sp(info, &sp);
80 		unw_get_bsp(info, &bsp);
81 		snprintf(buf, sizeof(buf),
82 			 " [<%016lx>] %%s\n"
83 			 "                                sp=%016lx bsp=%016lx\n",
84 			 ip, sp, bsp);
85 		print_symbol(buf, ip);
86 	} while (unw_unwind(info) >= 0);
87 }
88 
89 void
90 show_stack (struct task_struct *task, unsigned long *sp)
91 {
92 	if (!task)
93 		unw_init_running(ia64_do_show_stack, NULL);
94 	else {
95 		struct unw_frame_info info;
96 
97 		unw_init_from_blocked_task(&info, task);
98 		ia64_do_show_stack(&info, NULL);
99 	}
100 }
101 
102 void
103 dump_stack (void)
104 {
105 	show_stack(NULL, NULL);
106 }
107 
108 EXPORT_SYMBOL(dump_stack);
109 
110 void
111 show_regs (struct pt_regs *regs)
112 {
113 	unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
114 
115 	print_modules();
116 	printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
117 			smp_processor_id(), current->comm);
118 	printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s (%s)\n",
119 	       regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
120 	       init_utsname()->release);
121 	print_symbol("ip is at %s\n", ip);
122 	printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
123 	       regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
124 	printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
125 	       regs->ar_rnat, regs->ar_bspstore, regs->pr);
126 	printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
127 	       regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
128 	printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
129 	printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
130 	printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
131 	       regs->f6.u.bits[1], regs->f6.u.bits[0],
132 	       regs->f7.u.bits[1], regs->f7.u.bits[0]);
133 	printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
134 	       regs->f8.u.bits[1], regs->f8.u.bits[0],
135 	       regs->f9.u.bits[1], regs->f9.u.bits[0]);
136 	printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
137 	       regs->f10.u.bits[1], regs->f10.u.bits[0],
138 	       regs->f11.u.bits[1], regs->f11.u.bits[0]);
139 
140 	printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
141 	printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
142 	printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
143 	printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
144 	printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
145 	printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
146 	printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
147 	printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
148 	printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
149 
150 	if (user_mode(regs)) {
151 		/* print the stacked registers */
152 		unsigned long val, *bsp, ndirty;
153 		int i, sof, is_nat = 0;
154 
155 		sof = regs->cr_ifs & 0x7f;	/* size of frame */
156 		ndirty = (regs->loadrs >> 19);
157 		bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
158 		for (i = 0; i < sof; ++i) {
159 			get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
160 			printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
161 			       ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
162 		}
163 	} else
164 		show_stack(NULL, NULL);
165 }
166 
167 /* local support for deprecated console_print */
168 void
169 console_print(const char *s)
170 {
171 	printk(KERN_EMERG "%s", s);
172 }
173 
174 void
175 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
176 {
177 	if (fsys_mode(current, &scr->pt)) {
178 		/*
179 		 * defer signal-handling etc. until we return to
180 		 * privilege-level 0.
181 		 */
182 		if (!ia64_psr(&scr->pt)->lp)
183 			ia64_psr(&scr->pt)->lp = 1;
184 		return;
185 	}
186 
187 #ifdef CONFIG_PERFMON
188 	if (current->thread.pfm_needs_checking)
189 		/*
190 		 * Note: pfm_handle_work() allow us to call it with interrupts
191 		 * disabled, and may enable interrupts within the function.
192 		 */
193 		pfm_handle_work();
194 #endif
195 
196 	/* deal with pending signal delivery */
197 	if (test_thread_flag(TIF_SIGPENDING)) {
198 		local_irq_enable();	/* force interrupt enable */
199 		ia64_do_signal(scr, in_syscall);
200 	}
201 
202 	if (test_thread_flag(TIF_NOTIFY_RESUME)) {
203 		clear_thread_flag(TIF_NOTIFY_RESUME);
204 		tracehook_notify_resume(&scr->pt);
205 		if (current->replacement_session_keyring)
206 			key_replace_session_keyring();
207 	}
208 
209 	/* copy user rbs to kernel rbs */
210 	if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
211 		local_irq_enable();	/* force interrupt enable */
212 		ia64_sync_krbs();
213 	}
214 
215 	local_irq_disable();	/* force interrupt disable */
216 }
217 
218 static int pal_halt        = 1;
219 static int can_do_pal_halt = 1;
220 
221 static int __init nohalt_setup(char * str)
222 {
223 	pal_halt = can_do_pal_halt = 0;
224 	return 1;
225 }
226 __setup("nohalt", nohalt_setup);
227 
228 void
229 update_pal_halt_status(int status)
230 {
231 	can_do_pal_halt = pal_halt && status;
232 }
233 
234 /*
235  * We use this if we don't have any better idle routine..
236  */
237 void
238 default_idle (void)
239 {
240 	local_irq_enable();
241 	while (!need_resched()) {
242 		if (can_do_pal_halt) {
243 			local_irq_disable();
244 			if (!need_resched()) {
245 				safe_halt();
246 			}
247 			local_irq_enable();
248 		} else
249 			cpu_relax();
250 	}
251 }
252 
253 #ifdef CONFIG_HOTPLUG_CPU
254 /* We don't actually take CPU down, just spin without interrupts. */
255 static inline void play_dead(void)
256 {
257 	unsigned int this_cpu = smp_processor_id();
258 
259 	/* Ack it */
260 	__get_cpu_var(cpu_state) = CPU_DEAD;
261 
262 	max_xtp();
263 	local_irq_disable();
264 	idle_task_exit();
265 	ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
266 	/*
267 	 * The above is a point of no-return, the processor is
268 	 * expected to be in SAL loop now.
269 	 */
270 	BUG();
271 }
272 #else
273 static inline void play_dead(void)
274 {
275 	BUG();
276 }
277 #endif /* CONFIG_HOTPLUG_CPU */
278 
279 static void do_nothing(void *unused)
280 {
281 }
282 
283 /*
284  * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
285  * pm_idle and update to new pm_idle value. Required while changing pm_idle
286  * handler on SMP systems.
287  *
288  * Caller must have changed pm_idle to the new value before the call. Old
289  * pm_idle value will not be used by any CPU after the return of this function.
290  */
291 void cpu_idle_wait(void)
292 {
293 	smp_mb();
294 	/* kick all the CPUs so that they exit out of pm_idle */
295 	smp_call_function(do_nothing, NULL, 1);
296 }
297 EXPORT_SYMBOL_GPL(cpu_idle_wait);
298 
299 void __attribute__((noreturn))
300 cpu_idle (void)
301 {
302 	void (*mark_idle)(int) = ia64_mark_idle;
303   	int cpu = smp_processor_id();
304 
305 	/* endless idle loop with no priority at all */
306 	while (1) {
307 		if (can_do_pal_halt) {
308 			current_thread_info()->status &= ~TS_POLLING;
309 			/*
310 			 * TS_POLLING-cleared state must be visible before we
311 			 * test NEED_RESCHED:
312 			 */
313 			smp_mb();
314 		} else {
315 			current_thread_info()->status |= TS_POLLING;
316 		}
317 
318 		if (!need_resched()) {
319 			void (*idle)(void);
320 #ifdef CONFIG_SMP
321 			min_xtp();
322 #endif
323 			rmb();
324 			if (mark_idle)
325 				(*mark_idle)(1);
326 
327 			idle = pm_idle;
328 			if (!idle)
329 				idle = default_idle;
330 			(*idle)();
331 			if (mark_idle)
332 				(*mark_idle)(0);
333 #ifdef CONFIG_SMP
334 			normal_xtp();
335 #endif
336 		}
337 		preempt_enable_no_resched();
338 		schedule();
339 		preempt_disable();
340 		check_pgt_cache();
341 		if (cpu_is_offline(cpu))
342 			play_dead();
343 	}
344 }
345 
346 void
347 ia64_save_extra (struct task_struct *task)
348 {
349 #ifdef CONFIG_PERFMON
350 	unsigned long info;
351 #endif
352 
353 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
354 		ia64_save_debug_regs(&task->thread.dbr[0]);
355 
356 #ifdef CONFIG_PERFMON
357 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
358 		pfm_save_regs(task);
359 
360 	info = __get_cpu_var(pfm_syst_info);
361 	if (info & PFM_CPUINFO_SYST_WIDE)
362 		pfm_syst_wide_update_task(task, info, 0);
363 #endif
364 }
365 
366 void
367 ia64_load_extra (struct task_struct *task)
368 {
369 #ifdef CONFIG_PERFMON
370 	unsigned long info;
371 #endif
372 
373 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
374 		ia64_load_debug_regs(&task->thread.dbr[0]);
375 
376 #ifdef CONFIG_PERFMON
377 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
378 		pfm_load_regs(task);
379 
380 	info = __get_cpu_var(pfm_syst_info);
381 	if (info & PFM_CPUINFO_SYST_WIDE)
382 		pfm_syst_wide_update_task(task, info, 1);
383 #endif
384 }
385 
386 /*
387  * Copy the state of an ia-64 thread.
388  *
389  * We get here through the following  call chain:
390  *
391  *	from user-level:	from kernel:
392  *
393  *	<clone syscall>	        <some kernel call frames>
394  *	sys_clone		   :
395  *	do_fork			do_fork
396  *	copy_thread		copy_thread
397  *
398  * This means that the stack layout is as follows:
399  *
400  *	+---------------------+ (highest addr)
401  *	|   struct pt_regs    |
402  *	+---------------------+
403  *	| struct switch_stack |
404  *	+---------------------+
405  *	|                     |
406  *	|    memory stack     |
407  *	|                     | <-- sp (lowest addr)
408  *	+---------------------+
409  *
410  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
411  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
412  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
413  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
414  * the stack is page aligned and the page size is at least 4KB, this is always the case,
415  * so there is nothing to worry about.
416  */
417 int
418 copy_thread(unsigned long clone_flags,
419 	     unsigned long user_stack_base, unsigned long user_stack_size,
420 	     struct task_struct *p, struct pt_regs *regs)
421 {
422 	extern char ia64_ret_from_clone;
423 	struct switch_stack *child_stack, *stack;
424 	unsigned long rbs, child_rbs, rbs_size;
425 	struct pt_regs *child_ptregs;
426 	int retval = 0;
427 
428 #ifdef CONFIG_SMP
429 	/*
430 	 * For SMP idle threads, fork_by_hand() calls do_fork with
431 	 * NULL regs.
432 	 */
433 	if (!regs)
434 		return 0;
435 #endif
436 
437 	stack = ((struct switch_stack *) regs) - 1;
438 
439 	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
440 	child_stack = (struct switch_stack *) child_ptregs - 1;
441 
442 	/* copy parent's switch_stack & pt_regs to child: */
443 	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
444 
445 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
446 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
447 	rbs_size = stack->ar_bspstore - rbs;
448 
449 	/* copy the parent's register backing store to the child: */
450 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
451 
452 	if (likely(user_mode(child_ptregs))) {
453 		if (clone_flags & CLONE_SETTLS)
454 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
455 		if (user_stack_base) {
456 			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
457 			child_ptregs->ar_bspstore = user_stack_base;
458 			child_ptregs->ar_rnat = 0;
459 			child_ptregs->loadrs = 0;
460 		}
461 	} else {
462 		/*
463 		 * Note: we simply preserve the relative position of
464 		 * the stack pointer here.  There is no need to
465 		 * allocate a scratch area here, since that will have
466 		 * been taken care of by the caller of sys_clone()
467 		 * already.
468 		 */
469 		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
470 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
471 	}
472 	child_stack->ar_bspstore = child_rbs + rbs_size;
473 	child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
474 
475 	/* copy parts of thread_struct: */
476 	p->thread.ksp = (unsigned long) child_stack - 16;
477 
478 	/* stop some PSR bits from being inherited.
479 	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
480 	 * therefore we must specify them explicitly here and not include them in
481 	 * IA64_PSR_BITS_TO_CLEAR.
482 	 */
483 	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
484 				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
485 
486 	/*
487 	 * NOTE: The calling convention considers all floating point
488 	 * registers in the high partition (fph) to be scratch.  Since
489 	 * the only way to get to this point is through a system call,
490 	 * we know that the values in fph are all dead.  Hence, there
491 	 * is no need to inherit the fph state from the parent to the
492 	 * child and all we have to do is to make sure that
493 	 * IA64_THREAD_FPH_VALID is cleared in the child.
494 	 *
495 	 * XXX We could push this optimization a bit further by
496 	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
497 	 * However, it's not clear this is worth doing.  Also, it
498 	 * would be a slight deviation from the normal Linux system
499 	 * call behavior where scratch registers are preserved across
500 	 * system calls (unless used by the system call itself).
501 	 */
502 #	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
503 					 | IA64_THREAD_PM_VALID)
504 #	define THREAD_FLAGS_TO_SET	0
505 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
506 			   | THREAD_FLAGS_TO_SET);
507 	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
508 
509 #ifdef CONFIG_PERFMON
510 	if (current->thread.pfm_context)
511 		pfm_inherit(p, child_ptregs);
512 #endif
513 	return retval;
514 }
515 
516 static void
517 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
518 {
519 	unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
520 	unsigned long uninitialized_var(ip);	/* GCC be quiet */
521 	elf_greg_t *dst = arg;
522 	struct pt_regs *pt;
523 	char nat;
524 	int i;
525 
526 	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
527 
528 	if (unw_unwind_to_user(info) < 0)
529 		return;
530 
531 	unw_get_sp(info, &sp);
532 	pt = (struct pt_regs *) (sp + 16);
533 
534 	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
535 
536 	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
537 		return;
538 
539 	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
540 		  &ar_rnat);
541 
542 	/*
543 	 * coredump format:
544 	 *	r0-r31
545 	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
546 	 *	predicate registers (p0-p63)
547 	 *	b0-b7
548 	 *	ip cfm user-mask
549 	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
550 	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
551 	 */
552 
553 	/* r0 is zero */
554 	for (i = 1, mask = (1UL << i); i < 32; ++i) {
555 		unw_get_gr(info, i, &dst[i], &nat);
556 		if (nat)
557 			nat_bits |= mask;
558 		mask <<= 1;
559 	}
560 	dst[32] = nat_bits;
561 	unw_get_pr(info, &dst[33]);
562 
563 	for (i = 0; i < 8; ++i)
564 		unw_get_br(info, i, &dst[34 + i]);
565 
566 	unw_get_rp(info, &ip);
567 	dst[42] = ip + ia64_psr(pt)->ri;
568 	dst[43] = cfm;
569 	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
570 
571 	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
572 	/*
573 	 * For bsp and bspstore, unw_get_ar() would return the kernel
574 	 * addresses, but we need the user-level addresses instead:
575 	 */
576 	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
577 	dst[47] = pt->ar_bspstore;
578 	dst[48] = ar_rnat;
579 	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
580 	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
581 	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
582 	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
583 	unw_get_ar(info, UNW_AR_LC, &dst[53]);
584 	unw_get_ar(info, UNW_AR_EC, &dst[54]);
585 	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
586 	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
587 }
588 
589 void
590 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
591 {
592 	elf_fpreg_t *dst = arg;
593 	int i;
594 
595 	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
596 
597 	if (unw_unwind_to_user(info) < 0)
598 		return;
599 
600 	/* f0 is 0.0, f1 is 1.0 */
601 
602 	for (i = 2; i < 32; ++i)
603 		unw_get_fr(info, i, dst + i);
604 
605 	ia64_flush_fph(task);
606 	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
607 		memcpy(dst + 32, task->thread.fph, 96*16);
608 }
609 
610 void
611 do_copy_regs (struct unw_frame_info *info, void *arg)
612 {
613 	do_copy_task_regs(current, info, arg);
614 }
615 
616 void
617 do_dump_fpu (struct unw_frame_info *info, void *arg)
618 {
619 	do_dump_task_fpu(current, info, arg);
620 }
621 
622 void
623 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
624 {
625 	unw_init_running(do_copy_regs, dst);
626 }
627 
628 int
629 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
630 {
631 	unw_init_running(do_dump_fpu, dst);
632 	return 1;	/* f0-f31 are always valid so we always return 1 */
633 }
634 
635 long
636 sys_execve (const char __user *filename,
637 	    const char __user *const __user *argv,
638 	    const char __user *const __user *envp,
639 	    struct pt_regs *regs)
640 {
641 	char *fname;
642 	int error;
643 
644 	fname = getname(filename);
645 	error = PTR_ERR(fname);
646 	if (IS_ERR(fname))
647 		goto out;
648 	error = do_execve(fname, argv, envp, regs);
649 	putname(fname);
650 out:
651 	return error;
652 }
653 
654 pid_t
655 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
656 {
657 	extern void start_kernel_thread (void);
658 	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
659 	struct {
660 		struct switch_stack sw;
661 		struct pt_regs pt;
662 	} regs;
663 
664 	memset(&regs, 0, sizeof(regs));
665 	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
666 	regs.pt.r1 = helper_fptr[1];		/* set GP */
667 	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
668 	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
669 	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
670 	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
671 	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
672 	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
673 	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
674 	regs.sw.pr = (1 << PRED_KERNEL_STACK);
675 	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
676 }
677 EXPORT_SYMBOL(kernel_thread);
678 
679 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
680 int
681 kernel_thread_helper (int (*fn)(void *), void *arg)
682 {
683 	return (*fn)(arg);
684 }
685 
686 /*
687  * Flush thread state.  This is called when a thread does an execve().
688  */
689 void
690 flush_thread (void)
691 {
692 	/* drop floating-point and debug-register state if it exists: */
693 	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
694 	ia64_drop_fpu(current);
695 }
696 
697 /*
698  * Clean up state associated with current thread.  This is called when
699  * the thread calls exit().
700  */
701 void
702 exit_thread (void)
703 {
704 
705 	ia64_drop_fpu(current);
706 #ifdef CONFIG_PERFMON
707        /* if needed, stop monitoring and flush state to perfmon context */
708 	if (current->thread.pfm_context)
709 		pfm_exit_thread(current);
710 
711 	/* free debug register resources */
712 	if (current->thread.flags & IA64_THREAD_DBG_VALID)
713 		pfm_release_debug_registers(current);
714 #endif
715 }
716 
717 unsigned long
718 get_wchan (struct task_struct *p)
719 {
720 	struct unw_frame_info info;
721 	unsigned long ip;
722 	int count = 0;
723 
724 	if (!p || p == current || p->state == TASK_RUNNING)
725 		return 0;
726 
727 	/*
728 	 * Note: p may not be a blocked task (it could be current or
729 	 * another process running on some other CPU.  Rather than
730 	 * trying to determine if p is really blocked, we just assume
731 	 * it's blocked and rely on the unwind routines to fail
732 	 * gracefully if the process wasn't really blocked after all.
733 	 * --davidm 99/12/15
734 	 */
735 	unw_init_from_blocked_task(&info, p);
736 	do {
737 		if (p->state == TASK_RUNNING)
738 			return 0;
739 		if (unw_unwind(&info) < 0)
740 			return 0;
741 		unw_get_ip(&info, &ip);
742 		if (!in_sched_functions(ip))
743 			return ip;
744 	} while (count++ < 16);
745 	return 0;
746 }
747 
748 void
749 cpu_halt (void)
750 {
751 	pal_power_mgmt_info_u_t power_info[8];
752 	unsigned long min_power;
753 	int i, min_power_state;
754 
755 	if (ia64_pal_halt_info(power_info) != 0)
756 		return;
757 
758 	min_power_state = 0;
759 	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
760 	for (i = 1; i < 8; ++i)
761 		if (power_info[i].pal_power_mgmt_info_s.im
762 		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
763 			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
764 			min_power_state = i;
765 		}
766 
767 	while (1)
768 		ia64_pal_halt(min_power_state);
769 }
770 
771 void machine_shutdown(void)
772 {
773 #ifdef CONFIG_HOTPLUG_CPU
774 	int cpu;
775 
776 	for_each_online_cpu(cpu) {
777 		if (cpu != smp_processor_id())
778 			cpu_down(cpu);
779 	}
780 #endif
781 #ifdef CONFIG_KEXEC
782 	kexec_disable_iosapic();
783 #endif
784 }
785 
786 void
787 machine_restart (char *restart_cmd)
788 {
789 	(void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
790 	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
791 }
792 
793 void
794 machine_halt (void)
795 {
796 	(void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
797 	cpu_halt();
798 }
799 
800 void
801 machine_power_off (void)
802 {
803 	if (pm_power_off)
804 		pm_power_off();
805 	machine_halt();
806 }
807 
808