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