xref: /openbmc/linux/arch/ia64/kernel/process.c (revision 545e4006)
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/module.h>
19 #include <linux/notifier.h>
20 #include <linux/personality.h>
21 #include <linux/sched.h>
22 #include <linux/slab.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 
32 #include <asm/cpu.h>
33 #include <asm/delay.h>
34 #include <asm/elf.h>
35 #include <asm/ia32.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 
63 void
64 ia64_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 
85 void
86 show_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 
98 void
99 dump_stack (void)
100 {
101 	show_stack(NULL, NULL);
102 }
103 
104 EXPORT_SYMBOL(dump_stack);
105 
106 void
107 show_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 void tsk_clear_notify_resume(struct task_struct *tsk)
164 {
165 #ifdef CONFIG_PERFMON
166 	if (tsk->thread.pfm_needs_checking)
167 		return;
168 #endif
169 	if (test_ti_thread_flag(task_thread_info(tsk), TIF_RESTORE_RSE))
170 		return;
171 	clear_ti_thread_flag(task_thread_info(tsk), TIF_NOTIFY_RESUME);
172 }
173 
174 /*
175  * do_notify_resume_user():
176  *	Called from notify_resume_user at entry.S, with interrupts disabled.
177  */
178 void
179 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
180 {
181 	if (fsys_mode(current, &scr->pt)) {
182 		/*
183 		 * defer signal-handling etc. until we return to
184 		 * privilege-level 0.
185 		 */
186 		if (!ia64_psr(&scr->pt)->lp)
187 			ia64_psr(&scr->pt)->lp = 1;
188 		return;
189 	}
190 
191 #ifdef CONFIG_PERFMON
192 	if (current->thread.pfm_needs_checking)
193 		/*
194 		 * Note: pfm_handle_work() allow us to call it with interrupts
195 		 * disabled, and may enable interrupts within the function.
196 		 */
197 		pfm_handle_work();
198 #endif
199 
200 	/* deal with pending signal delivery */
201 	if (test_thread_flag(TIF_SIGPENDING)) {
202 		local_irq_enable();	/* force interrupt enable */
203 		ia64_do_signal(scr, in_syscall);
204 	}
205 
206 	/* copy user rbs to kernel rbs */
207 	if (unlikely(test_thread_flag(TIF_RESTORE_RSE))) {
208 		local_irq_enable();	/* force interrupt enable */
209 		ia64_sync_krbs();
210 	}
211 
212 	local_irq_disable();	/* force interrupt disable */
213 }
214 
215 static int pal_halt        = 1;
216 static int can_do_pal_halt = 1;
217 
218 static int __init nohalt_setup(char * str)
219 {
220 	pal_halt = can_do_pal_halt = 0;
221 	return 1;
222 }
223 __setup("nohalt", nohalt_setup);
224 
225 void
226 update_pal_halt_status(int status)
227 {
228 	can_do_pal_halt = pal_halt && status;
229 }
230 
231 /*
232  * We use this if we don't have any better idle routine..
233  */
234 void
235 default_idle (void)
236 {
237 	local_irq_enable();
238 	while (!need_resched()) {
239 		if (can_do_pal_halt) {
240 			local_irq_disable();
241 			if (!need_resched()) {
242 				safe_halt();
243 			}
244 			local_irq_enable();
245 		} else
246 			cpu_relax();
247 	}
248 }
249 
250 #ifdef CONFIG_HOTPLUG_CPU
251 /* We don't actually take CPU down, just spin without interrupts. */
252 static inline void play_dead(void)
253 {
254 	extern void ia64_cpu_local_tick (void);
255 	unsigned int this_cpu = smp_processor_id();
256 
257 	/* Ack it */
258 	__get_cpu_var(cpu_state) = CPU_DEAD;
259 
260 	max_xtp();
261 	local_irq_disable();
262 	idle_task_exit();
263 	ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
264 	/*
265 	 * The above is a point of no-return, the processor is
266 	 * expected to be in SAL loop now.
267 	 */
268 	BUG();
269 }
270 #else
271 static inline void play_dead(void)
272 {
273 	BUG();
274 }
275 #endif /* CONFIG_HOTPLUG_CPU */
276 
277 static void do_nothing(void *unused)
278 {
279 }
280 
281 /*
282  * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
283  * pm_idle and update to new pm_idle value. Required while changing pm_idle
284  * handler on SMP systems.
285  *
286  * Caller must have changed pm_idle to the new value before the call. Old
287  * pm_idle value will not be used by any CPU after the return of this function.
288  */
289 void cpu_idle_wait(void)
290 {
291 	smp_mb();
292 	/* kick all the CPUs so that they exit out of pm_idle */
293 	smp_call_function(do_nothing, NULL, 1);
294 }
295 EXPORT_SYMBOL_GPL(cpu_idle_wait);
296 
297 void __attribute__((noreturn))
298 cpu_idle (void)
299 {
300 	void (*mark_idle)(int) = ia64_mark_idle;
301   	int cpu = smp_processor_id();
302 
303 	/* endless idle loop with no priority at all */
304 	while (1) {
305 		if (can_do_pal_halt) {
306 			current_thread_info()->status &= ~TS_POLLING;
307 			/*
308 			 * TS_POLLING-cleared state must be visible before we
309 			 * test NEED_RESCHED:
310 			 */
311 			smp_mb();
312 		} else {
313 			current_thread_info()->status |= TS_POLLING;
314 		}
315 
316 		if (!need_resched()) {
317 			void (*idle)(void);
318 #ifdef CONFIG_SMP
319 			min_xtp();
320 #endif
321 			rmb();
322 			if (mark_idle)
323 				(*mark_idle)(1);
324 
325 			idle = pm_idle;
326 			if (!idle)
327 				idle = default_idle;
328 			(*idle)();
329 			if (mark_idle)
330 				(*mark_idle)(0);
331 #ifdef CONFIG_SMP
332 			normal_xtp();
333 #endif
334 		}
335 		preempt_enable_no_resched();
336 		schedule();
337 		preempt_disable();
338 		check_pgt_cache();
339 		if (cpu_is_offline(cpu))
340 			play_dead();
341 	}
342 }
343 
344 void
345 ia64_save_extra (struct task_struct *task)
346 {
347 #ifdef CONFIG_PERFMON
348 	unsigned long info;
349 #endif
350 
351 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
352 		ia64_save_debug_regs(&task->thread.dbr[0]);
353 
354 #ifdef CONFIG_PERFMON
355 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
356 		pfm_save_regs(task);
357 
358 	info = __get_cpu_var(pfm_syst_info);
359 	if (info & PFM_CPUINFO_SYST_WIDE)
360 		pfm_syst_wide_update_task(task, info, 0);
361 #endif
362 
363 #ifdef CONFIG_IA32_SUPPORT
364 	if (IS_IA32_PROCESS(task_pt_regs(task)))
365 		ia32_save_state(task);
366 #endif
367 }
368 
369 void
370 ia64_load_extra (struct task_struct *task)
371 {
372 #ifdef CONFIG_PERFMON
373 	unsigned long info;
374 #endif
375 
376 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
377 		ia64_load_debug_regs(&task->thread.dbr[0]);
378 
379 #ifdef CONFIG_PERFMON
380 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
381 		pfm_load_regs(task);
382 
383 	info = __get_cpu_var(pfm_syst_info);
384 	if (info & PFM_CPUINFO_SYST_WIDE)
385 		pfm_syst_wide_update_task(task, info, 1);
386 #endif
387 
388 #ifdef CONFIG_IA32_SUPPORT
389 	if (IS_IA32_PROCESS(task_pt_regs(task)))
390 		ia32_load_state(task);
391 #endif
392 }
393 
394 /*
395  * Copy the state of an ia-64 thread.
396  *
397  * We get here through the following  call chain:
398  *
399  *	from user-level:	from kernel:
400  *
401  *	<clone syscall>	        <some kernel call frames>
402  *	sys_clone		   :
403  *	do_fork			do_fork
404  *	copy_thread		copy_thread
405  *
406  * This means that the stack layout is as follows:
407  *
408  *	+---------------------+ (highest addr)
409  *	|   struct pt_regs    |
410  *	+---------------------+
411  *	| struct switch_stack |
412  *	+---------------------+
413  *	|                     |
414  *	|    memory stack     |
415  *	|                     | <-- sp (lowest addr)
416  *	+---------------------+
417  *
418  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
419  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
420  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
421  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
422  * the stack is page aligned and the page size is at least 4KB, this is always the case,
423  * so there is nothing to worry about.
424  */
425 int
426 copy_thread (int nr, unsigned long clone_flags,
427 	     unsigned long user_stack_base, unsigned long user_stack_size,
428 	     struct task_struct *p, struct pt_regs *regs)
429 {
430 	extern char ia64_ret_from_clone, ia32_ret_from_clone;
431 	struct switch_stack *child_stack, *stack;
432 	unsigned long rbs, child_rbs, rbs_size;
433 	struct pt_regs *child_ptregs;
434 	int retval = 0;
435 
436 #ifdef CONFIG_SMP
437 	/*
438 	 * For SMP idle threads, fork_by_hand() calls do_fork with
439 	 * NULL regs.
440 	 */
441 	if (!regs)
442 		return 0;
443 #endif
444 
445 	stack = ((struct switch_stack *) regs) - 1;
446 
447 	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
448 	child_stack = (struct switch_stack *) child_ptregs - 1;
449 
450 	/* copy parent's switch_stack & pt_regs to child: */
451 	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
452 
453 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
454 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
455 	rbs_size = stack->ar_bspstore - rbs;
456 
457 	/* copy the parent's register backing store to the child: */
458 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
459 
460 	if (likely(user_mode(child_ptregs))) {
461 		if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
462 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
463 		if (user_stack_base) {
464 			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
465 			child_ptregs->ar_bspstore = user_stack_base;
466 			child_ptregs->ar_rnat = 0;
467 			child_ptregs->loadrs = 0;
468 		}
469 	} else {
470 		/*
471 		 * Note: we simply preserve the relative position of
472 		 * the stack pointer here.  There is no need to
473 		 * allocate a scratch area here, since that will have
474 		 * been taken care of by the caller of sys_clone()
475 		 * already.
476 		 */
477 		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
478 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
479 	}
480 	child_stack->ar_bspstore = child_rbs + rbs_size;
481 	if (IS_IA32_PROCESS(regs))
482 		child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
483 	else
484 		child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
485 
486 	/* copy parts of thread_struct: */
487 	p->thread.ksp = (unsigned long) child_stack - 16;
488 
489 	/* stop some PSR bits from being inherited.
490 	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
491 	 * therefore we must specify them explicitly here and not include them in
492 	 * IA64_PSR_BITS_TO_CLEAR.
493 	 */
494 	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
495 				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
496 
497 	/*
498 	 * NOTE: The calling convention considers all floating point
499 	 * registers in the high partition (fph) to be scratch.  Since
500 	 * the only way to get to this point is through a system call,
501 	 * we know that the values in fph are all dead.  Hence, there
502 	 * is no need to inherit the fph state from the parent to the
503 	 * child and all we have to do is to make sure that
504 	 * IA64_THREAD_FPH_VALID is cleared in the child.
505 	 *
506 	 * XXX We could push this optimization a bit further by
507 	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
508 	 * However, it's not clear this is worth doing.  Also, it
509 	 * would be a slight deviation from the normal Linux system
510 	 * call behavior where scratch registers are preserved across
511 	 * system calls (unless used by the system call itself).
512 	 */
513 #	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
514 					 | IA64_THREAD_PM_VALID)
515 #	define THREAD_FLAGS_TO_SET	0
516 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
517 			   | THREAD_FLAGS_TO_SET);
518 	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
519 #ifdef CONFIG_IA32_SUPPORT
520 	/*
521 	 * If we're cloning an IA32 task then save the IA32 extra
522 	 * state from the current task to the new task
523 	 */
524 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
525 		ia32_save_state(p);
526 		if (clone_flags & CLONE_SETTLS)
527 			retval = ia32_clone_tls(p, child_ptregs);
528 
529 		/* Copy partially mapped page list */
530 		if (!retval)
531 			retval = ia32_copy_ia64_partial_page_list(p,
532 								clone_flags);
533 	}
534 #endif
535 
536 #ifdef CONFIG_PERFMON
537 	if (current->thread.pfm_context)
538 		pfm_inherit(p, child_ptregs);
539 #endif
540 	return retval;
541 }
542 
543 static void
544 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
545 {
546 	unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
547 	unsigned long uninitialized_var(ip);	/* GCC be quiet */
548 	elf_greg_t *dst = arg;
549 	struct pt_regs *pt;
550 	char nat;
551 	int i;
552 
553 	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
554 
555 	if (unw_unwind_to_user(info) < 0)
556 		return;
557 
558 	unw_get_sp(info, &sp);
559 	pt = (struct pt_regs *) (sp + 16);
560 
561 	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
562 
563 	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
564 		return;
565 
566 	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
567 		  &ar_rnat);
568 
569 	/*
570 	 * coredump format:
571 	 *	r0-r31
572 	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
573 	 *	predicate registers (p0-p63)
574 	 *	b0-b7
575 	 *	ip cfm user-mask
576 	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
577 	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
578 	 */
579 
580 	/* r0 is zero */
581 	for (i = 1, mask = (1UL << i); i < 32; ++i) {
582 		unw_get_gr(info, i, &dst[i], &nat);
583 		if (nat)
584 			nat_bits |= mask;
585 		mask <<= 1;
586 	}
587 	dst[32] = nat_bits;
588 	unw_get_pr(info, &dst[33]);
589 
590 	for (i = 0; i < 8; ++i)
591 		unw_get_br(info, i, &dst[34 + i]);
592 
593 	unw_get_rp(info, &ip);
594 	dst[42] = ip + ia64_psr(pt)->ri;
595 	dst[43] = cfm;
596 	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
597 
598 	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
599 	/*
600 	 * For bsp and bspstore, unw_get_ar() would return the kernel
601 	 * addresses, but we need the user-level addresses instead:
602 	 */
603 	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
604 	dst[47] = pt->ar_bspstore;
605 	dst[48] = ar_rnat;
606 	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
607 	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
608 	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
609 	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
610 	unw_get_ar(info, UNW_AR_LC, &dst[53]);
611 	unw_get_ar(info, UNW_AR_EC, &dst[54]);
612 	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
613 	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
614 }
615 
616 void
617 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
618 {
619 	elf_fpreg_t *dst = arg;
620 	int i;
621 
622 	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
623 
624 	if (unw_unwind_to_user(info) < 0)
625 		return;
626 
627 	/* f0 is 0.0, f1 is 1.0 */
628 
629 	for (i = 2; i < 32; ++i)
630 		unw_get_fr(info, i, dst + i);
631 
632 	ia64_flush_fph(task);
633 	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
634 		memcpy(dst + 32, task->thread.fph, 96*16);
635 }
636 
637 void
638 do_copy_regs (struct unw_frame_info *info, void *arg)
639 {
640 	do_copy_task_regs(current, info, arg);
641 }
642 
643 void
644 do_dump_fpu (struct unw_frame_info *info, void *arg)
645 {
646 	do_dump_task_fpu(current, info, arg);
647 }
648 
649 void
650 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
651 {
652 	unw_init_running(do_copy_regs, dst);
653 }
654 
655 int
656 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
657 {
658 	unw_init_running(do_dump_fpu, dst);
659 	return 1;	/* f0-f31 are always valid so we always return 1 */
660 }
661 
662 long
663 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
664 	    struct pt_regs *regs)
665 {
666 	char *fname;
667 	int error;
668 
669 	fname = getname(filename);
670 	error = PTR_ERR(fname);
671 	if (IS_ERR(fname))
672 		goto out;
673 	error = do_execve(fname, argv, envp, regs);
674 	putname(fname);
675 out:
676 	return error;
677 }
678 
679 pid_t
680 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
681 {
682 	extern void start_kernel_thread (void);
683 	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
684 	struct {
685 		struct switch_stack sw;
686 		struct pt_regs pt;
687 	} regs;
688 
689 	memset(&regs, 0, sizeof(regs));
690 	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
691 	regs.pt.r1 = helper_fptr[1];		/* set GP */
692 	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
693 	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
694 	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
695 	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
696 	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
697 	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
698 	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
699 	regs.sw.pr = (1 << PRED_KERNEL_STACK);
700 	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
701 }
702 EXPORT_SYMBOL(kernel_thread);
703 
704 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
705 int
706 kernel_thread_helper (int (*fn)(void *), void *arg)
707 {
708 #ifdef CONFIG_IA32_SUPPORT
709 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
710 		/* A kernel thread is always a 64-bit process. */
711 		current->thread.map_base  = DEFAULT_MAP_BASE;
712 		current->thread.task_size = DEFAULT_TASK_SIZE;
713 		ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
714 		ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
715 	}
716 #endif
717 	return (*fn)(arg);
718 }
719 
720 /*
721  * Flush thread state.  This is called when a thread does an execve().
722  */
723 void
724 flush_thread (void)
725 {
726 	/* drop floating-point and debug-register state if it exists: */
727 	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
728 	ia64_drop_fpu(current);
729 #ifdef CONFIG_IA32_SUPPORT
730 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
731 		ia32_drop_ia64_partial_page_list(current);
732 		current->thread.task_size = IA32_PAGE_OFFSET;
733 		set_fs(USER_DS);
734 		memset(current->thread.tls_array, 0, sizeof(current->thread.tls_array));
735 	}
736 #endif
737 }
738 
739 /*
740  * Clean up state associated with current thread.  This is called when
741  * the thread calls exit().
742  */
743 void
744 exit_thread (void)
745 {
746 
747 	ia64_drop_fpu(current);
748 #ifdef CONFIG_PERFMON
749        /* if needed, stop monitoring and flush state to perfmon context */
750 	if (current->thread.pfm_context)
751 		pfm_exit_thread(current);
752 
753 	/* free debug register resources */
754 	if (current->thread.flags & IA64_THREAD_DBG_VALID)
755 		pfm_release_debug_registers(current);
756 #endif
757 	if (IS_IA32_PROCESS(task_pt_regs(current)))
758 		ia32_drop_ia64_partial_page_list(current);
759 }
760 
761 unsigned long
762 get_wchan (struct task_struct *p)
763 {
764 	struct unw_frame_info info;
765 	unsigned long ip;
766 	int count = 0;
767 
768 	if (!p || p == current || p->state == TASK_RUNNING)
769 		return 0;
770 
771 	/*
772 	 * Note: p may not be a blocked task (it could be current or
773 	 * another process running on some other CPU.  Rather than
774 	 * trying to determine if p is really blocked, we just assume
775 	 * it's blocked and rely on the unwind routines to fail
776 	 * gracefully if the process wasn't really blocked after all.
777 	 * --davidm 99/12/15
778 	 */
779 	unw_init_from_blocked_task(&info, p);
780 	do {
781 		if (p->state == TASK_RUNNING)
782 			return 0;
783 		if (unw_unwind(&info) < 0)
784 			return 0;
785 		unw_get_ip(&info, &ip);
786 		if (!in_sched_functions(ip))
787 			return ip;
788 	} while (count++ < 16);
789 	return 0;
790 }
791 
792 void
793 cpu_halt (void)
794 {
795 	pal_power_mgmt_info_u_t power_info[8];
796 	unsigned long min_power;
797 	int i, min_power_state;
798 
799 	if (ia64_pal_halt_info(power_info) != 0)
800 		return;
801 
802 	min_power_state = 0;
803 	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
804 	for (i = 1; i < 8; ++i)
805 		if (power_info[i].pal_power_mgmt_info_s.im
806 		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
807 			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
808 			min_power_state = i;
809 		}
810 
811 	while (1)
812 		ia64_pal_halt(min_power_state);
813 }
814 
815 void machine_shutdown(void)
816 {
817 #ifdef CONFIG_HOTPLUG_CPU
818 	int cpu;
819 
820 	for_each_online_cpu(cpu) {
821 		if (cpu != smp_processor_id())
822 			cpu_down(cpu);
823 	}
824 #endif
825 #ifdef CONFIG_KEXEC
826 	kexec_disable_iosapic();
827 #endif
828 }
829 
830 void
831 machine_restart (char *restart_cmd)
832 {
833 	(void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
834 	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
835 }
836 
837 void
838 machine_halt (void)
839 {
840 	(void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
841 	cpu_halt();
842 }
843 
844 void
845 machine_power_off (void)
846 {
847 	if (pm_power_off)
848 		pm_power_off();
849 	machine_halt();
850 }
851 
852