/arch/ia64/kernel/process.c

  1/*
  2 * Architecture-specific setup.
  3 *
  4 * Copyright (C) 1998-2003 Hewlett-Packard Co
  5 *	David Mosberger-Tang <davidm@hpl.hp.com>
  6 */
  7#define __KERNEL_SYSCALLS__	/* see <asm/unistd.h> */
  8#include <linux/config.h>
  9
 10#include <linux/cpu.h>
 11#include <linux/pm.h>
 12#include <linux/elf.h>
 13#include <linux/errno.h>
 14#include <linux/kallsyms.h>
 15#include <linux/kernel.h>
 16#include <linux/mm.h>
 17#include <linux/module.h>
 18#include <linux/notifier.h>
 19#include <linux/personality.h>
 20#include <linux/sched.h>
 21#include <linux/slab.h>
 22#include <linux/smp_lock.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
 30#include <asm/cpu.h>
 31#include <asm/delay.h>
 32#include <asm/elf.h>
 33#include <asm/ia32.h>
 34#include <asm/irq.h>
 35#include <asm/pgalloc.h>
 36#include <asm/processor.h>
 37#include <asm/sal.h>
 38#include <asm/tlbflush.h>
 39#include <asm/uaccess.h>
 40#include <asm/unwind.h>
 41#include <asm/user.h>
 42
 43#include "entry.h"
 44
 45#ifdef CONFIG_PERFMON
 46# include <asm/perfmon.h>
 47#endif
 48
 49#include "sigframe.h"
 50
 51void (*ia64_mark_idle)(int);
 52static cpumask_t cpu_idle_map;
 53
 54unsigned long boot_option_idle_override = 0;
 55EXPORT_SYMBOL(boot_option_idle_override);
 56
 57void
 58ia64_do_show_stack (struct unw_frame_info *info, void *arg)
 59{
 60	unsigned long ip, sp, bsp;
 61	char buf[128];			/* don't make it so big that it overflows the stack! */
 62
 63	printk("\nCall Trace:\n");
 64	do {
 65		unw_get_ip(info, &ip);
 66		if (ip == 0)
 67			break;
 68
 69		unw_get_sp(info, &sp);
 70		unw_get_bsp(info, &bsp);
 71		snprintf(buf, sizeof(buf),
 72			 " [<%016lx>] %%s\n"
 73			 "                                sp=%016lx bsp=%016lx\n",
 74			 ip, sp, bsp);
 75		print_symbol(buf, ip);
 76	} while (unw_unwind(info) >= 0);
 77}
 78
 79void
 80show_stack (struct task_struct *task, unsigned long *sp)
 81{
 82	if (!task)
 83		unw_init_running(ia64_do_show_stack, NULL);
 84	else {
 85		struct unw_frame_info info;
 86
 87		unw_init_from_blocked_task(&info, task);
 88		ia64_do_show_stack(&info, NULL);
 89	}
 90}
 91
 92void
 93dump_stack (void)
 94{
 95	show_stack(NULL, NULL);
 96}
 97
 98EXPORT_SYMBOL(dump_stack);
 99
100void
101show_regs (struct pt_regs *regs)
102{
103	unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
104
105	print_modules();
106	printk("\nPid: %d, CPU %d, comm: %20s\n", current->pid, smp_processor_id(), current->comm);
107	printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s\n",
108	       regs->cr_ipsr, regs->cr_ifs, ip, print_tainted());
109	print_symbol("ip is at %s\n", ip);
110	printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
111	       regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
112	printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
113	       regs->ar_rnat, regs->ar_bspstore, regs->pr);
114	printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
115	       regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
116	printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
117	printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
118	printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
119	       regs->f6.u.bits[1], regs->f6.u.bits[0],
120	       regs->f7.u.bits[1], regs->f7.u.bits[0]);
121	printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
122	       regs->f8.u.bits[1], regs->f8.u.bits[0],
123	       regs->f9.u.bits[1], regs->f9.u.bits[0]);
124	printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
125	       regs->f10.u.bits[1], regs->f10.u.bits[0],
126	       regs->f11.u.bits[1], regs->f11.u.bits[0]);
127
128	printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
129	printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
130	printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
131	printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
132	printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
133	printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
134	printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
135	printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
136	printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
137
138	if (user_mode(regs)) {
139		/* print the stacked registers */
140		unsigned long val, *bsp, ndirty;
141		int i, sof, is_nat = 0;
142
143		sof = regs->cr_ifs & 0x7f;	/* size of frame */
144		ndirty = (regs->loadrs >> 19);
145		bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
146		for (i = 0; i < sof; ++i) {
147			get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
148			printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
149			       ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
150		}
151	} else
152		show_stack(NULL, NULL);
153}
154
155void
156do_notify_resume_user (sigset_t *oldset, struct sigscratch *scr, long in_syscall)
157{
158	if (fsys_mode(current, &scr->pt)) {
159		/* defer signal-handling etc. until we return to privilege-level 0.  */
160		if (!ia64_psr(&scr->pt)->lp)
161			ia64_psr(&scr->pt)->lp = 1;
162		return;
163	}
164
165#ifdef CONFIG_PERFMON
166	if (current->thread.pfm_needs_checking)
167		pfm_handle_work();
168#endif
169
170	/* deal with pending signal delivery */
171	if (test_thread_flag(TIF_SIGPENDING))
172		ia64_do_signal(oldset, scr, in_syscall);
173}
174
175static int pal_halt = 1;
176static int __init nohalt_setup(char * str)
177{
178	pal_halt = 0;
179	return 1;
180}
181__setup("nohalt", nohalt_setup);
182
183/*
184 * We use this if we don't have any better idle routine..
185 */
186void
187default_idle (void)
188{
189	unsigned long pmu_active = ia64_getreg(_IA64_REG_PSR) & (IA64_PSR_PP | IA64_PSR_UP);
190
191	while (!need_resched())
192		if (pal_halt && !pmu_active)
193			safe_halt();
194		else
195			cpu_relax();
196}
197
198#ifdef CONFIG_HOTPLUG_CPU
199/* We don't actually take CPU down, just spin without interrupts. */
200static inline void play_dead(void)
201{
202	extern void ia64_cpu_local_tick (void);
203	/* Ack it */
204	__get_cpu_var(cpu_state) = CPU_DEAD;
205
206	/* We shouldn't have to disable interrupts while dead, but
207	 * some interrupts just don't seem to go away, and this makes
208	 * it "work" for testing purposes. */
209	max_xtp();
210	local_irq_disable();
211	/* Death loop */
212	while (__get_cpu_var(cpu_state) != CPU_UP_PREPARE)
213		cpu_relax();
214
215	/*
216	 * Enable timer interrupts from now on
217	 * Not required if we put processor in SAL_BOOT_RENDEZ mode.
218	 */
219	local_flush_tlb_all();
220	cpu_set(smp_processor_id(), cpu_online_map);
221	wmb();
222	ia64_cpu_local_tick ();
223	local_irq_enable();
224}
225#else
226static inline void play_dead(void)
227{
228	BUG();
229}
230#endif /* CONFIG_HOTPLUG_CPU */
231
232
233void cpu_idle_wait(void)
234{
235        int cpu;
236        cpumask_t map;
237
238        for_each_online_cpu(cpu)
239                cpu_set(cpu, cpu_idle_map);
240
241        wmb();
242        do {
243                ssleep(1);
244                cpus_and(map, cpu_idle_map, cpu_online_map);
245        } while (!cpus_empty(map));
246}
247EXPORT_SYMBOL_GPL(cpu_idle_wait);
248
249void __attribute__((noreturn))
250cpu_idle (void)
251{
252	void (*mark_idle)(int) = ia64_mark_idle;
253	int cpu = smp_processor_id();
254
255	/* endless idle loop with no priority at all */
256	while (1) {
257#ifdef CONFIG_SMP
258		if (!need_resched())
259			min_xtp();
260#endif
261		while (!need_resched()) {
262			void (*idle)(void);
263
264			if (mark_idle)
265				(*mark_idle)(1);
266
267			if (cpu_isset(cpu, cpu_idle_map))
268				cpu_clear(cpu, cpu_idle_map);
269			rmb();
270			idle = pm_idle;
271			if (!idle)
272				idle = default_idle;
273			(*idle)();
274		}
275
276		if (mark_idle)
277			(*mark_idle)(0);
278
279#ifdef CONFIG_SMP
280		normal_xtp();
281#endif
282		schedule();
283		check_pgt_cache();
284		if (cpu_is_offline(smp_processor_id()))
285			play_dead();
286	}
287}
288
289void
290ia64_save_extra (struct task_struct *task)
291{
292#ifdef CONFIG_PERFMON
293	unsigned long info;
294#endif
295
296	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
297		ia64_save_debug_regs(&task->thread.dbr[0]);
298
299#ifdef CONFIG_PERFMON
300	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
301		pfm_save_regs(task);
302
303	info = __get_cpu_var(pfm_syst_info);
304	if (info & PFM_CPUINFO_SYST_WIDE)
305		pfm_syst_wide_update_task(task, info, 0);
306#endif
307
308#ifdef CONFIG_IA32_SUPPORT
309	if (IS_IA32_PROCESS(ia64_task_regs(task)))
310		ia32_save_state(task);
311#endif
312}
313
314void
315ia64_load_extra (struct task_struct *task)
316{
317#ifdef CONFIG_PERFMON
318	unsigned long info;
319#endif
320
321	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
322		ia64_load_debug_regs(&task->thread.dbr[0]);
323
324#ifdef CONFIG_PERFMON
325	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
326		pfm_load_regs(task);
327
328	info = __get_cpu_var(pfm_syst_info);
329	if (info & PFM_CPUINFO_SYST_WIDE) 
330		pfm_syst_wide_update_task(task, info, 1);
331#endif
332
333#ifdef CONFIG_IA32_SUPPORT
334	if (IS_IA32_PROCESS(ia64_task_regs(task)))
335		ia32_load_state(task);
336#endif
337}
338
339/*
340 * Copy the state of an ia-64 thread.
341 *
342 * We get here through the following  call chain:
343 *
344 *	from user-level:	from kernel:
345 *
346 *	<clone syscall>	        <some kernel call frames>
347 *	sys_clone		   :
348 *	do_fork			do_fork
349 *	copy_thread		copy_thread
350 *
351 * This means that the stack layout is as follows:
352 *
353 *	+---------------------+ (highest addr)
354 *	|   struct pt_regs    |
355 *	+---------------------+
356 *	| struct switch_stack |
357 *	+---------------------+
358 *	|                     |
359 *	|    memory stack     |
360 *	|                     | <-- sp (lowest addr)
361 *	+---------------------+
362 *
363 * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
364 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
365 * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
366 * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
367 * the stack is page aligned and the page size is at least 4KB, this is always the case,
368 * so there is nothing to worry about.
369 */
370int
371copy_thread (int nr, unsigned long clone_flags,
372	     unsigned long user_stack_base, unsigned long user_stack_size,
373	     struct task_struct *p, struct pt_regs *regs)
374{
375	extern char ia64_ret_from_clone, ia32_ret_from_clone;
376	struct switch_stack *child_stack, *stack;
377	unsigned long rbs, child_rbs, rbs_size;
378	struct pt_regs *child_ptregs;
379	int retval = 0;
380
381#ifdef CONFIG_SMP
382	/*
383	 * For SMP idle threads, fork_by_hand() calls do_fork with
384	 * NULL regs.
385	 */
386	if (!regs)
387		return 0;
388#endif
389
390	stack = ((struct switch_stack *) regs) - 1;
391
392	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
393	child_stack = (struct switch_stack *) child_ptregs - 1;
394
395	/* copy parent's switch_stack & pt_regs to child: */
396	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
397
398	rbs = (unsigned long) current + IA64_RBS_OFFSET;
399	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
400	rbs_size = stack->ar_bspstore - rbs;
401
402	/* copy the parent's register backing store to the child: */
403	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
404
405	if (likely(user_mode(child_ptregs))) {
406		if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
407			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
408		if (user_stack_base) {
409			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
410			child_ptregs->ar_bspstore = user_stack_base;
411			child_ptregs->ar_rnat = 0;
412			child_ptregs->loadrs = 0;
413		}
414	} else {
415		/*
416		 * Note: we simply preserve the relative position of
417		 * the stack pointer here.  There is no need to
418		 * allocate a scratch area here, since that will have
419		 * been taken care of by the caller of sys_clone()
420		 * already.
421		 */
422		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
423		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
424	}
425	child_stack->ar_bspstore = child_rbs + rbs_size;
426	if (IS_IA32_PROCESS(regs))
427		child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
428	else
429		child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
430
431	/* copy parts of thread_struct: */
432	p->thread.ksp = (unsigned long) child_stack - 16;
433
434	/* stop some PSR bits from being inherited.
435	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
436	 * therefore we must specify them explicitly here and not include them in
437	 * IA64_PSR_BITS_TO_CLEAR.
438	 */
439	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
440				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
441
442	/*
443	 * NOTE: The calling convention considers all floating point
444	 * registers in the high partition (fph) to be scratch.  Since
445	 * the only way to get to this point is through a system call,
446	 * we know that the values in fph are all dead.  Hence, there
447	 * is no need to inherit the fph state from the parent to the
448	 * child and all we have to do is to make sure that
449	 * IA64_THREAD_FPH_VALID is cleared in the child.
450	 *
451	 * XXX We could push this optimization a bit further by
452	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
453	 * However, it's not clear this is worth doing.  Also, it
454	 * would be a slight deviation from the normal Linux system
455	 * call behavior where scratch registers are preserved across
456	 * system calls (unless used by the system call itself).
457	 */
458#	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
459					 | IA64_THREAD_PM_VALID)
460#	define THREAD_FLAGS_TO_SET	0
461	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
462			   | THREAD_FLAGS_TO_SET);
463	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
464#ifdef CONFIG_IA32_SUPPORT
465	/*
466	 * If we're cloning an IA32 task then save the IA32 extra
467	 * state from the current task to the new task
468	 */
469	if (IS_IA32_PROCESS(ia64_task_regs(current))) {
470		ia32_save_state(p);
471		if (clone_flags & CLONE_SETTLS)
472			retval = ia32_clone_tls(p, child_ptregs);
473
474		/* Copy partially mapped page list */
475		if (!retval)
476			retval = ia32_copy_partial_page_list(p, clone_flags);
477	}
478#endif
479
480#ifdef CONFIG_PERFMON
481	if (current->thread.pfm_context)
482		pfm_inherit(p, child_ptregs);
483#endif
484	return retval;
485}
486
487static void
488do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
489{
490	unsigned long mask, sp, nat_bits = 0, ip, ar_rnat, urbs_end, cfm;
491	elf_greg_t *dst = arg;
492	struct pt_regs *pt;
493	char nat;
494	int i;
495
496	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
497
498	if (unw_unwind_to_user(info) < 0)
499		return;
500
501	unw_get_sp(info, &sp);
502	pt = (struct pt_regs *) (sp + 16);
503
504	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
505
506	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
507		return;
508
509	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
510		  &ar_rnat);
511
512	/*
513	 * coredump format:
514	 *	r0-r31
515	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
516	 *	predicate registers (p0-p63)
517	 *	b0-b7
518	 *	ip cfm user-mask
519	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
520	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
521	 */
522
523	/* r0 is zero */
524	for (i = 1, mask = (1UL << i); i < 32; ++i) {
525		unw_get_gr(info, i, &dst[i], &nat);
526		if (nat)
527			nat_bits |= mask;
528		mask <<= 1;
529	}
530	dst[32] = nat_bits;
531	unw_get_pr(info, &dst[33]);
532
533	for (i = 0; i < 8; ++i)
534		unw_get_br(info, i, &dst[34 + i]);
535
536	unw_get_rp(info, &ip);
537	dst[42] = ip + ia64_psr(pt)->ri;
538	dst[43] = cfm;
539	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
540
541	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
542	/*
543	 * For bsp and bspstore, unw_get_ar() would return the kernel
544	 * addresses, but we need the user-level addresses instead:
545	 */
546	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
547	dst[47] = pt->ar_bspstore;
548	dst[48] = ar_rnat;
549	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
550	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
551	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
552	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
553	unw_get_ar(info, UNW_AR_LC, &dst[53]);
554	unw_get_ar(info, UNW_AR_EC, &dst[54]);
555	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
556	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
557}
558
559void
560do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
561{
562	elf_fpreg_t *dst = arg;
563	int i;
564
565	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
566
567	if (unw_unwind_to_user(info) < 0)
568		return;
569
570	/* f0 is 0.0, f1 is 1.0 */
571
572	for (i = 2; i < 32; ++i)
573		unw_get_fr(info, i, dst + i);
574
575	ia64_flush_fph(task);
576	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
577		memcpy(dst + 32, task->thread.fph, 96*16);
578}
579
580void
581do_copy_regs (struct unw_frame_info *info, void *arg)
582{
583	do_copy_task_regs(current, info, arg);
584}
585
586void
587do_dump_fpu (struct unw_frame_info *info, void *arg)
588{
589	do_dump_task_fpu(current, info, arg);
590}
591
592int
593dump_task_regs(struct task_struct *task, elf_gregset_t *regs)
594{
595	struct unw_frame_info tcore_info;
596
597	if (current == task) {
598		unw_init_running(do_copy_regs, regs);
599	} else {
600		memset(&tcore_info, 0, sizeof(tcore_info));
601		unw_init_from_blocked_task(&tcore_info, task);
602		do_copy_task_regs(task, &tcore_info, regs);
603	}
604	return 1;
605}
606
607void
608ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
609{
610	unw_init_running(do_copy_regs, dst);
611}
612
613int
614dump_task_fpu (struct task_struct *task, elf_fpregset_t *dst)
615{
616	struct unw_frame_info tcore_info;
617
618	if (current == task) {
619		unw_init_running(do_dump_fpu, dst);
620	} else {
621		memset(&tcore_info, 0, sizeof(tcore_info));
622		unw_init_from_blocked_task(&tcore_info, task);
623		do_dump_task_fpu(task, &tcore_info, dst);
624	}
625	return 1;
626}
627
628int
629dump_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
635long
636sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
637	    struct pt_regs *regs)
638{
639	char *fname;
640	int error;
641
642	fname = getname(filename);
643	error = PTR_ERR(fname);
644	if (IS_ERR(fname))
645		goto out;
646	error = do_execve(fname, argv, envp, regs);
647	putname(fname);
648out:
649	return error;
650}
651
652pid_t
653kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
654{
655	extern void start_kernel_thread (void);
656	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
657	struct {
658		struct switch_stack sw;
659		struct pt_regs pt;
660	} regs;
661
662	memset(&regs, 0, sizeof(regs));
663	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
664	regs.pt.r1 = helper_fptr[1];		/* set GP */
665	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
666	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
667	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
668	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
669	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
670	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
671	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
672	regs.sw.pr = (1 << PRED_KERNEL_STACK);
673	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
674}
675EXPORT_SYMBOL(kernel_thread);
676
677/* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
678int
679kernel_thread_helper (int (*fn)(void *), void *arg)
680{
681#ifdef CONFIG_IA32_SUPPORT
682	if (IS_IA32_PROCESS(ia64_task_regs(current))) {
683		/* A kernel thread is always a 64-bit process. */
684		current->thread.map_base  = DEFAULT_MAP_BASE;
685		current->thread.task_size = DEFAULT_TASK_SIZE;
686		ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
687		ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
688	}
689#endif
690	return (*fn)(arg);
691}
692
693/*
694 * Flush thread state.  This is called when a thread does an execve().
695 */
696void
697flush_thread (void)
698{
699	/* drop floating-point and debug-register state if it exists: */
700	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
701	ia64_drop_fpu(current);
702	if (IS_IA32_PROCESS(ia64_task_regs(current)))
703		ia32_drop_partial_page_list(current);
704}
705
706/*
707 * Clean up state associated with current thread.  This is called when
708 * the thread calls exit().
709 */
710void
711exit_thread (void)
712{
713	ia64_drop_fpu(current);
714#ifdef CONFIG_PERFMON
715       /* if needed, stop monitoring and flush state to perfmon context */
716	if (current->thread.pfm_context)
717		pfm_exit_thread(current);
718
719	/* free debug register resources */
720	if (current->thread.flags & IA64_THREAD_DBG_VALID)
721		pfm_release_debug_registers(current);
722#endif
723	if (IS_IA32_PROCESS(ia64_task_regs(current)))
724		ia32_drop_partial_page_list(current);
725}
726
727unsigned long
728get_wchan (struct task_struct *p)
729{
730	struct unw_frame_info info;
731	unsigned long ip;
732	int count = 0;
733
734	/*
735	 * Note: p may not be a blocked task (it could be current or
736	 * another process running on some other CPU.  Rather than
737	 * trying to determine if p is really blocked, we just assume
738	 * it's blocked and rely on the unwind routines to fail
739	 * gracefully if the process wasn't really blocked after all.
740	 * --davidm 99/12/15
741	 */
742	unw_init_from_blocked_task(&info, p);
743	do {
744		if (unw_unwind(&info) < 0)
745			return 0;
746		unw_get_ip(&info, &ip);
747		if (!in_sched_functions(ip))
748			return ip;
749	} while (count++ < 16);
750	return 0;
751}
752
753void
754cpu_halt (void)
755{
756	pal_power_mgmt_info_u_t power_info[8];
757	unsigned long min_power;
758	int i, min_power_state;
759
760	if (ia64_pal_halt_info(power_info) != 0)
761		return;
762
763	min_power_state = 0;
764	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
765	for (i = 1; i < 8; ++i)
766		if (power_info[i].pal_power_mgmt_info_s.im
767		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
768			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
769			min_power_state = i;
770		}
771
772	while (1)
773		ia64_pal_halt(min_power_state);
774}
775
776void
777machine_restart (char *restart_cmd)
778{
779	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
780}
781
782EXPORT_SYMBOL(machine_restart);
783
784void
785machine_halt (void)
786{
787	cpu_halt();
788}
789
790EXPORT_SYMBOL(machine_halt);
791
792void
793machine_power_off (void)
794{
795	if (pm_power_off)
796		pm_power_off();
797	machine_halt();
798}
799
800EXPORT_SYMBOL(machine_power_off);