xref: /openbmc/linux/arch/ia64/kernel/process.c (revision b6dcefde)
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 #include <linux/tracehook.h>
32 
33 #include <asm/cpu.h>
34 #include <asm/delay.h>
35 #include <asm/elf.h>
36 #include <asm/ia32.h>
37 #include <asm/irq.h>
38 #include <asm/kexec.h>
39 #include <asm/pgalloc.h>
40 #include <asm/processor.h>
41 #include <asm/sal.h>
42 #include <asm/tlbflush.h>
43 #include <asm/uaccess.h>
44 #include <asm/unwind.h>
45 #include <asm/user.h>
46 
47 #include "entry.h"
48 
49 #ifdef CONFIG_PERFMON
50 # include <asm/perfmon.h>
51 #endif
52 
53 #include "sigframe.h"
54 
55 void (*ia64_mark_idle)(int);
56 
57 unsigned long boot_option_idle_override = 0;
58 EXPORT_SYMBOL(boot_option_idle_override);
59 unsigned long idle_halt;
60 EXPORT_SYMBOL(idle_halt);
61 unsigned long idle_nomwait;
62 EXPORT_SYMBOL(idle_nomwait);
63 
64 void
65 ia64_do_show_stack (struct unw_frame_info *info, void *arg)
66 {
67 	unsigned long ip, sp, bsp;
68 	char buf[128];			/* don't make it so big that it overflows the stack! */
69 
70 	printk("\nCall Trace:\n");
71 	do {
72 		unw_get_ip(info, &ip);
73 		if (ip == 0)
74 			break;
75 
76 		unw_get_sp(info, &sp);
77 		unw_get_bsp(info, &bsp);
78 		snprintf(buf, sizeof(buf),
79 			 " [<%016lx>] %%s\n"
80 			 "                                sp=%016lx bsp=%016lx\n",
81 			 ip, sp, bsp);
82 		print_symbol(buf, ip);
83 	} while (unw_unwind(info) >= 0);
84 }
85 
86 void
87 show_stack (struct task_struct *task, unsigned long *sp)
88 {
89 	if (!task)
90 		unw_init_running(ia64_do_show_stack, NULL);
91 	else {
92 		struct unw_frame_info info;
93 
94 		unw_init_from_blocked_task(&info, task);
95 		ia64_do_show_stack(&info, NULL);
96 	}
97 }
98 
99 void
100 dump_stack (void)
101 {
102 	show_stack(NULL, NULL);
103 }
104 
105 EXPORT_SYMBOL(dump_stack);
106 
107 void
108 show_regs (struct pt_regs *regs)
109 {
110 	unsigned long ip = regs->cr_iip + ia64_psr(regs)->ri;
111 
112 	print_modules();
113 	printk("\nPid: %d, CPU %d, comm: %20s\n", task_pid_nr(current),
114 			smp_processor_id(), current->comm);
115 	printk("psr : %016lx ifs : %016lx ip  : [<%016lx>]    %s (%s)\n",
116 	       regs->cr_ipsr, regs->cr_ifs, ip, print_tainted(),
117 	       init_utsname()->release);
118 	print_symbol("ip is at %s\n", ip);
119 	printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
120 	       regs->ar_unat, regs->ar_pfs, regs->ar_rsc);
121 	printk("rnat: %016lx bsps: %016lx pr  : %016lx\n",
122 	       regs->ar_rnat, regs->ar_bspstore, regs->pr);
123 	printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
124 	       regs->loadrs, regs->ar_ccv, regs->ar_fpsr);
125 	printk("csd : %016lx ssd : %016lx\n", regs->ar_csd, regs->ar_ssd);
126 	printk("b0  : %016lx b6  : %016lx b7  : %016lx\n", regs->b0, regs->b6, regs->b7);
127 	printk("f6  : %05lx%016lx f7  : %05lx%016lx\n",
128 	       regs->f6.u.bits[1], regs->f6.u.bits[0],
129 	       regs->f7.u.bits[1], regs->f7.u.bits[0]);
130 	printk("f8  : %05lx%016lx f9  : %05lx%016lx\n",
131 	       regs->f8.u.bits[1], regs->f8.u.bits[0],
132 	       regs->f9.u.bits[1], regs->f9.u.bits[0]);
133 	printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
134 	       regs->f10.u.bits[1], regs->f10.u.bits[0],
135 	       regs->f11.u.bits[1], regs->f11.u.bits[0]);
136 
137 	printk("r1  : %016lx r2  : %016lx r3  : %016lx\n", regs->r1, regs->r2, regs->r3);
138 	printk("r8  : %016lx r9  : %016lx r10 : %016lx\n", regs->r8, regs->r9, regs->r10);
139 	printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs->r11, regs->r12, regs->r13);
140 	printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs->r14, regs->r15, regs->r16);
141 	printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs->r17, regs->r18, regs->r19);
142 	printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs->r20, regs->r21, regs->r22);
143 	printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs->r23, regs->r24, regs->r25);
144 	printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs->r26, regs->r27, regs->r28);
145 	printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs->r29, regs->r30, regs->r31);
146 
147 	if (user_mode(regs)) {
148 		/* print the stacked registers */
149 		unsigned long val, *bsp, ndirty;
150 		int i, sof, is_nat = 0;
151 
152 		sof = regs->cr_ifs & 0x7f;	/* size of frame */
153 		ndirty = (regs->loadrs >> 19);
154 		bsp = ia64_rse_skip_regs((unsigned long *) regs->ar_bspstore, ndirty);
155 		for (i = 0; i < sof; ++i) {
156 			get_user(val, (unsigned long __user *) ia64_rse_skip_regs(bsp, i));
157 			printk("r%-3u:%c%016lx%s", 32 + i, is_nat ? '*' : ' ', val,
158 			       ((i == sof - 1) || (i % 3) == 2) ? "\n" : " ");
159 		}
160 	} else
161 		show_stack(NULL, NULL);
162 }
163 
164 /* local support for deprecated console_print */
165 void
166 console_print(const char *s)
167 {
168 	printk(KERN_EMERG "%s", s);
169 }
170 
171 void
172 do_notify_resume_user(sigset_t *unused, struct sigscratch *scr, long in_syscall)
173 {
174 	if (fsys_mode(current, &scr->pt)) {
175 		/*
176 		 * defer signal-handling etc. until we return to
177 		 * privilege-level 0.
178 		 */
179 		if (!ia64_psr(&scr->pt)->lp)
180 			ia64_psr(&scr->pt)->lp = 1;
181 		return;
182 	}
183 
184 #ifdef CONFIG_PERFMON
185 	if (current->thread.pfm_needs_checking)
186 		/*
187 		 * Note: pfm_handle_work() allow us to call it with interrupts
188 		 * disabled, and may enable interrupts within the function.
189 		 */
190 		pfm_handle_work();
191 #endif
192 
193 	/* deal with pending signal delivery */
194 	if (test_thread_flag(TIF_SIGPENDING)) {
195 		local_irq_enable();	/* force interrupt enable */
196 		ia64_do_signal(scr, in_syscall);
197 	}
198 
199 	if (test_thread_flag(TIF_NOTIFY_RESUME)) {
200 		clear_thread_flag(TIF_NOTIFY_RESUME);
201 		tracehook_notify_resume(&scr->pt);
202 		if (current->replacement_session_keyring)
203 			key_replace_session_keyring();
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 	unsigned int this_cpu = smp_processor_id();
255 
256 	/* Ack it */
257 	__get_cpu_var(cpu_state) = CPU_DEAD;
258 
259 	max_xtp();
260 	local_irq_disable();
261 	idle_task_exit();
262 	ia64_jump_to_sal(&sal_boot_rendez_state[this_cpu]);
263 	/*
264 	 * The above is a point of no-return, the processor is
265 	 * expected to be in SAL loop now.
266 	 */
267 	BUG();
268 }
269 #else
270 static inline void play_dead(void)
271 {
272 	BUG();
273 }
274 #endif /* CONFIG_HOTPLUG_CPU */
275 
276 static void do_nothing(void *unused)
277 {
278 }
279 
280 /*
281  * cpu_idle_wait - Used to ensure that all the CPUs discard old value of
282  * pm_idle and update to new pm_idle value. Required while changing pm_idle
283  * handler on SMP systems.
284  *
285  * Caller must have changed pm_idle to the new value before the call. Old
286  * pm_idle value will not be used by any CPU after the return of this function.
287  */
288 void cpu_idle_wait(void)
289 {
290 	smp_mb();
291 	/* kick all the CPUs so that they exit out of pm_idle */
292 	smp_call_function(do_nothing, NULL, 1);
293 }
294 EXPORT_SYMBOL_GPL(cpu_idle_wait);
295 
296 void __attribute__((noreturn))
297 cpu_idle (void)
298 {
299 	void (*mark_idle)(int) = ia64_mark_idle;
300   	int cpu = smp_processor_id();
301 
302 	/* endless idle loop with no priority at all */
303 	while (1) {
304 		if (can_do_pal_halt) {
305 			current_thread_info()->status &= ~TS_POLLING;
306 			/*
307 			 * TS_POLLING-cleared state must be visible before we
308 			 * test NEED_RESCHED:
309 			 */
310 			smp_mb();
311 		} else {
312 			current_thread_info()->status |= TS_POLLING;
313 		}
314 
315 		if (!need_resched()) {
316 			void (*idle)(void);
317 #ifdef CONFIG_SMP
318 			min_xtp();
319 #endif
320 			rmb();
321 			if (mark_idle)
322 				(*mark_idle)(1);
323 
324 			idle = pm_idle;
325 			if (!idle)
326 				idle = default_idle;
327 			(*idle)();
328 			if (mark_idle)
329 				(*mark_idle)(0);
330 #ifdef CONFIG_SMP
331 			normal_xtp();
332 #endif
333 		}
334 		preempt_enable_no_resched();
335 		schedule();
336 		preempt_disable();
337 		check_pgt_cache();
338 		if (cpu_is_offline(cpu))
339 			play_dead();
340 	}
341 }
342 
343 void
344 ia64_save_extra (struct task_struct *task)
345 {
346 #ifdef CONFIG_PERFMON
347 	unsigned long info;
348 #endif
349 
350 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
351 		ia64_save_debug_regs(&task->thread.dbr[0]);
352 
353 #ifdef CONFIG_PERFMON
354 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
355 		pfm_save_regs(task);
356 
357 	info = __get_cpu_var(pfm_syst_info);
358 	if (info & PFM_CPUINFO_SYST_WIDE)
359 		pfm_syst_wide_update_task(task, info, 0);
360 #endif
361 
362 #ifdef CONFIG_IA32_SUPPORT
363 	if (IS_IA32_PROCESS(task_pt_regs(task)))
364 		ia32_save_state(task);
365 #endif
366 }
367 
368 void
369 ia64_load_extra (struct task_struct *task)
370 {
371 #ifdef CONFIG_PERFMON
372 	unsigned long info;
373 #endif
374 
375 	if ((task->thread.flags & IA64_THREAD_DBG_VALID) != 0)
376 		ia64_load_debug_regs(&task->thread.dbr[0]);
377 
378 #ifdef CONFIG_PERFMON
379 	if ((task->thread.flags & IA64_THREAD_PM_VALID) != 0)
380 		pfm_load_regs(task);
381 
382 	info = __get_cpu_var(pfm_syst_info);
383 	if (info & PFM_CPUINFO_SYST_WIDE)
384 		pfm_syst_wide_update_task(task, info, 1);
385 #endif
386 
387 #ifdef CONFIG_IA32_SUPPORT
388 	if (IS_IA32_PROCESS(task_pt_regs(task)))
389 		ia32_load_state(task);
390 #endif
391 }
392 
393 /*
394  * Copy the state of an ia-64 thread.
395  *
396  * We get here through the following  call chain:
397  *
398  *	from user-level:	from kernel:
399  *
400  *	<clone syscall>	        <some kernel call frames>
401  *	sys_clone		   :
402  *	do_fork			do_fork
403  *	copy_thread		copy_thread
404  *
405  * This means that the stack layout is as follows:
406  *
407  *	+---------------------+ (highest addr)
408  *	|   struct pt_regs    |
409  *	+---------------------+
410  *	| struct switch_stack |
411  *	+---------------------+
412  *	|                     |
413  *	|    memory stack     |
414  *	|                     | <-- sp (lowest addr)
415  *	+---------------------+
416  *
417  * Observe that we copy the unat values that are in pt_regs and switch_stack.  Spilling an
418  * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
419  * with N=(X & 0x1ff)/8.  Thus, copying the unat value preserves the NaT bits ONLY if the
420  * pt_regs structure in the parent is congruent to that of the child, modulo 512.  Since
421  * the stack is page aligned and the page size is at least 4KB, this is always the case,
422  * so there is nothing to worry about.
423  */
424 int
425 copy_thread(unsigned long clone_flags,
426 	     unsigned long user_stack_base, unsigned long user_stack_size,
427 	     struct task_struct *p, struct pt_regs *regs)
428 {
429 	extern char ia64_ret_from_clone, ia32_ret_from_clone;
430 	struct switch_stack *child_stack, *stack;
431 	unsigned long rbs, child_rbs, rbs_size;
432 	struct pt_regs *child_ptregs;
433 	int retval = 0;
434 
435 #ifdef CONFIG_SMP
436 	/*
437 	 * For SMP idle threads, fork_by_hand() calls do_fork with
438 	 * NULL regs.
439 	 */
440 	if (!regs)
441 		return 0;
442 #endif
443 
444 	stack = ((struct switch_stack *) regs) - 1;
445 
446 	child_ptregs = (struct pt_regs *) ((unsigned long) p + IA64_STK_OFFSET) - 1;
447 	child_stack = (struct switch_stack *) child_ptregs - 1;
448 
449 	/* copy parent's switch_stack & pt_regs to child: */
450 	memcpy(child_stack, stack, sizeof(*child_ptregs) + sizeof(*child_stack));
451 
452 	rbs = (unsigned long) current + IA64_RBS_OFFSET;
453 	child_rbs = (unsigned long) p + IA64_RBS_OFFSET;
454 	rbs_size = stack->ar_bspstore - rbs;
455 
456 	/* copy the parent's register backing store to the child: */
457 	memcpy((void *) child_rbs, (void *) rbs, rbs_size);
458 
459 	if (likely(user_mode(child_ptregs))) {
460 		if ((clone_flags & CLONE_SETTLS) && !IS_IA32_PROCESS(regs))
461 			child_ptregs->r13 = regs->r16;	/* see sys_clone2() in entry.S */
462 		if (user_stack_base) {
463 			child_ptregs->r12 = user_stack_base + user_stack_size - 16;
464 			child_ptregs->ar_bspstore = user_stack_base;
465 			child_ptregs->ar_rnat = 0;
466 			child_ptregs->loadrs = 0;
467 		}
468 	} else {
469 		/*
470 		 * Note: we simply preserve the relative position of
471 		 * the stack pointer here.  There is no need to
472 		 * allocate a scratch area here, since that will have
473 		 * been taken care of by the caller of sys_clone()
474 		 * already.
475 		 */
476 		child_ptregs->r12 = (unsigned long) child_ptregs - 16; /* kernel sp */
477 		child_ptregs->r13 = (unsigned long) p;		/* set `current' pointer */
478 	}
479 	child_stack->ar_bspstore = child_rbs + rbs_size;
480 	if (IS_IA32_PROCESS(regs))
481 		child_stack->b0 = (unsigned long) &ia32_ret_from_clone;
482 	else
483 		child_stack->b0 = (unsigned long) &ia64_ret_from_clone;
484 
485 	/* copy parts of thread_struct: */
486 	p->thread.ksp = (unsigned long) child_stack - 16;
487 
488 	/* stop some PSR bits from being inherited.
489 	 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
490 	 * therefore we must specify them explicitly here and not include them in
491 	 * IA64_PSR_BITS_TO_CLEAR.
492 	 */
493 	child_ptregs->cr_ipsr = ((child_ptregs->cr_ipsr | IA64_PSR_BITS_TO_SET)
494 				 & ~(IA64_PSR_BITS_TO_CLEAR | IA64_PSR_PP | IA64_PSR_UP));
495 
496 	/*
497 	 * NOTE: The calling convention considers all floating point
498 	 * registers in the high partition (fph) to be scratch.  Since
499 	 * the only way to get to this point is through a system call,
500 	 * we know that the values in fph are all dead.  Hence, there
501 	 * is no need to inherit the fph state from the parent to the
502 	 * child and all we have to do is to make sure that
503 	 * IA64_THREAD_FPH_VALID is cleared in the child.
504 	 *
505 	 * XXX We could push this optimization a bit further by
506 	 * clearing IA64_THREAD_FPH_VALID on ANY system call.
507 	 * However, it's not clear this is worth doing.  Also, it
508 	 * would be a slight deviation from the normal Linux system
509 	 * call behavior where scratch registers are preserved across
510 	 * system calls (unless used by the system call itself).
511 	 */
512 #	define THREAD_FLAGS_TO_CLEAR	(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
513 					 | IA64_THREAD_PM_VALID)
514 #	define THREAD_FLAGS_TO_SET	0
515 	p->thread.flags = ((current->thread.flags & ~THREAD_FLAGS_TO_CLEAR)
516 			   | THREAD_FLAGS_TO_SET);
517 	ia64_drop_fpu(p);	/* don't pick up stale state from a CPU's fph */
518 #ifdef CONFIG_IA32_SUPPORT
519 	/*
520 	 * If we're cloning an IA32 task then save the IA32 extra
521 	 * state from the current task to the new task
522 	 */
523 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
524 		ia32_save_state(p);
525 		if (clone_flags & CLONE_SETTLS)
526 			retval = ia32_clone_tls(p, child_ptregs);
527 
528 		/* Copy partially mapped page list */
529 		if (!retval)
530 			retval = ia32_copy_ia64_partial_page_list(p,
531 								clone_flags);
532 	}
533 #endif
534 
535 #ifdef CONFIG_PERFMON
536 	if (current->thread.pfm_context)
537 		pfm_inherit(p, child_ptregs);
538 #endif
539 	return retval;
540 }
541 
542 static void
543 do_copy_task_regs (struct task_struct *task, struct unw_frame_info *info, void *arg)
544 {
545 	unsigned long mask, sp, nat_bits = 0, ar_rnat, urbs_end, cfm;
546 	unsigned long uninitialized_var(ip);	/* GCC be quiet */
547 	elf_greg_t *dst = arg;
548 	struct pt_regs *pt;
549 	char nat;
550 	int i;
551 
552 	memset(dst, 0, sizeof(elf_gregset_t));	/* don't leak any kernel bits to user-level */
553 
554 	if (unw_unwind_to_user(info) < 0)
555 		return;
556 
557 	unw_get_sp(info, &sp);
558 	pt = (struct pt_regs *) (sp + 16);
559 
560 	urbs_end = ia64_get_user_rbs_end(task, pt, &cfm);
561 
562 	if (ia64_sync_user_rbs(task, info->sw, pt->ar_bspstore, urbs_end) < 0)
563 		return;
564 
565 	ia64_peek(task, info->sw, urbs_end, (long) ia64_rse_rnat_addr((long *) urbs_end),
566 		  &ar_rnat);
567 
568 	/*
569 	 * coredump format:
570 	 *	r0-r31
571 	 *	NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
572 	 *	predicate registers (p0-p63)
573 	 *	b0-b7
574 	 *	ip cfm user-mask
575 	 *	ar.rsc ar.bsp ar.bspstore ar.rnat
576 	 *	ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
577 	 */
578 
579 	/* r0 is zero */
580 	for (i = 1, mask = (1UL << i); i < 32; ++i) {
581 		unw_get_gr(info, i, &dst[i], &nat);
582 		if (nat)
583 			nat_bits |= mask;
584 		mask <<= 1;
585 	}
586 	dst[32] = nat_bits;
587 	unw_get_pr(info, &dst[33]);
588 
589 	for (i = 0; i < 8; ++i)
590 		unw_get_br(info, i, &dst[34 + i]);
591 
592 	unw_get_rp(info, &ip);
593 	dst[42] = ip + ia64_psr(pt)->ri;
594 	dst[43] = cfm;
595 	dst[44] = pt->cr_ipsr & IA64_PSR_UM;
596 
597 	unw_get_ar(info, UNW_AR_RSC, &dst[45]);
598 	/*
599 	 * For bsp and bspstore, unw_get_ar() would return the kernel
600 	 * addresses, but we need the user-level addresses instead:
601 	 */
602 	dst[46] = urbs_end;	/* note: by convention PT_AR_BSP points to the end of the urbs! */
603 	dst[47] = pt->ar_bspstore;
604 	dst[48] = ar_rnat;
605 	unw_get_ar(info, UNW_AR_CCV, &dst[49]);
606 	unw_get_ar(info, UNW_AR_UNAT, &dst[50]);
607 	unw_get_ar(info, UNW_AR_FPSR, &dst[51]);
608 	dst[52] = pt->ar_pfs;	/* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
609 	unw_get_ar(info, UNW_AR_LC, &dst[53]);
610 	unw_get_ar(info, UNW_AR_EC, &dst[54]);
611 	unw_get_ar(info, UNW_AR_CSD, &dst[55]);
612 	unw_get_ar(info, UNW_AR_SSD, &dst[56]);
613 }
614 
615 void
616 do_dump_task_fpu (struct task_struct *task, struct unw_frame_info *info, void *arg)
617 {
618 	elf_fpreg_t *dst = arg;
619 	int i;
620 
621 	memset(dst, 0, sizeof(elf_fpregset_t));	/* don't leak any "random" bits */
622 
623 	if (unw_unwind_to_user(info) < 0)
624 		return;
625 
626 	/* f0 is 0.0, f1 is 1.0 */
627 
628 	for (i = 2; i < 32; ++i)
629 		unw_get_fr(info, i, dst + i);
630 
631 	ia64_flush_fph(task);
632 	if ((task->thread.flags & IA64_THREAD_FPH_VALID) != 0)
633 		memcpy(dst + 32, task->thread.fph, 96*16);
634 }
635 
636 void
637 do_copy_regs (struct unw_frame_info *info, void *arg)
638 {
639 	do_copy_task_regs(current, info, arg);
640 }
641 
642 void
643 do_dump_fpu (struct unw_frame_info *info, void *arg)
644 {
645 	do_dump_task_fpu(current, info, arg);
646 }
647 
648 void
649 ia64_elf_core_copy_regs (struct pt_regs *pt, elf_gregset_t dst)
650 {
651 	unw_init_running(do_copy_regs, dst);
652 }
653 
654 int
655 dump_fpu (struct pt_regs *pt, elf_fpregset_t dst)
656 {
657 	unw_init_running(do_dump_fpu, dst);
658 	return 1;	/* f0-f31 are always valid so we always return 1 */
659 }
660 
661 long
662 sys_execve (char __user *filename, char __user * __user *argv, char __user * __user *envp,
663 	    struct pt_regs *regs)
664 {
665 	char *fname;
666 	int error;
667 
668 	fname = getname(filename);
669 	error = PTR_ERR(fname);
670 	if (IS_ERR(fname))
671 		goto out;
672 	error = do_execve(fname, argv, envp, regs);
673 	putname(fname);
674 out:
675 	return error;
676 }
677 
678 pid_t
679 kernel_thread (int (*fn)(void *), void *arg, unsigned long flags)
680 {
681 	extern void start_kernel_thread (void);
682 	unsigned long *helper_fptr = (unsigned long *) &start_kernel_thread;
683 	struct {
684 		struct switch_stack sw;
685 		struct pt_regs pt;
686 	} regs;
687 
688 	memset(&regs, 0, sizeof(regs));
689 	regs.pt.cr_iip = helper_fptr[0];	/* set entry point (IP) */
690 	regs.pt.r1 = helper_fptr[1];		/* set GP */
691 	regs.pt.r9 = (unsigned long) fn;	/* 1st argument */
692 	regs.pt.r11 = (unsigned long) arg;	/* 2nd argument */
693 	/* Preserve PSR bits, except for bits 32-34 and 37-45, which we can't read.  */
694 	regs.pt.cr_ipsr = ia64_getreg(_IA64_REG_PSR) | IA64_PSR_BN;
695 	regs.pt.cr_ifs = 1UL << 63;		/* mark as valid, empty frame */
696 	regs.sw.ar_fpsr = regs.pt.ar_fpsr = ia64_getreg(_IA64_REG_AR_FPSR);
697 	regs.sw.ar_bspstore = (unsigned long) current + IA64_RBS_OFFSET;
698 	regs.sw.pr = (1 << PRED_KERNEL_STACK);
699 	return do_fork(flags | CLONE_VM | CLONE_UNTRACED, 0, &regs.pt, 0, NULL, NULL);
700 }
701 EXPORT_SYMBOL(kernel_thread);
702 
703 /* This gets called from kernel_thread() via ia64_invoke_thread_helper().  */
704 int
705 kernel_thread_helper (int (*fn)(void *), void *arg)
706 {
707 #ifdef CONFIG_IA32_SUPPORT
708 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
709 		/* A kernel thread is always a 64-bit process. */
710 		current->thread.map_base  = DEFAULT_MAP_BASE;
711 		current->thread.task_size = DEFAULT_TASK_SIZE;
712 		ia64_set_kr(IA64_KR_IO_BASE, current->thread.old_iob);
713 		ia64_set_kr(IA64_KR_TSSD, current->thread.old_k1);
714 	}
715 #endif
716 	return (*fn)(arg);
717 }
718 
719 /*
720  * Flush thread state.  This is called when a thread does an execve().
721  */
722 void
723 flush_thread (void)
724 {
725 	/* drop floating-point and debug-register state if it exists: */
726 	current->thread.flags &= ~(IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID);
727 	ia64_drop_fpu(current);
728 #ifdef CONFIG_IA32_SUPPORT
729 	if (IS_IA32_PROCESS(task_pt_regs(current))) {
730 		ia32_drop_ia64_partial_page_list(current);
731 		current->thread.task_size = IA32_PAGE_OFFSET;
732 		set_fs(USER_DS);
733 		memset(current->thread.tls_array, 0, sizeof(current->thread.tls_array));
734 	}
735 #endif
736 }
737 
738 /*
739  * Clean up state associated with current thread.  This is called when
740  * the thread calls exit().
741  */
742 void
743 exit_thread (void)
744 {
745 
746 	ia64_drop_fpu(current);
747 #ifdef CONFIG_PERFMON
748        /* if needed, stop monitoring and flush state to perfmon context */
749 	if (current->thread.pfm_context)
750 		pfm_exit_thread(current);
751 
752 	/* free debug register resources */
753 	if (current->thread.flags & IA64_THREAD_DBG_VALID)
754 		pfm_release_debug_registers(current);
755 #endif
756 	if (IS_IA32_PROCESS(task_pt_regs(current)))
757 		ia32_drop_ia64_partial_page_list(current);
758 }
759 
760 unsigned long
761 get_wchan (struct task_struct *p)
762 {
763 	struct unw_frame_info info;
764 	unsigned long ip;
765 	int count = 0;
766 
767 	if (!p || p == current || p->state == TASK_RUNNING)
768 		return 0;
769 
770 	/*
771 	 * Note: p may not be a blocked task (it could be current or
772 	 * another process running on some other CPU.  Rather than
773 	 * trying to determine if p is really blocked, we just assume
774 	 * it's blocked and rely on the unwind routines to fail
775 	 * gracefully if the process wasn't really blocked after all.
776 	 * --davidm 99/12/15
777 	 */
778 	unw_init_from_blocked_task(&info, p);
779 	do {
780 		if (p->state == TASK_RUNNING)
781 			return 0;
782 		if (unw_unwind(&info) < 0)
783 			return 0;
784 		unw_get_ip(&info, &ip);
785 		if (!in_sched_functions(ip))
786 			return ip;
787 	} while (count++ < 16);
788 	return 0;
789 }
790 
791 void
792 cpu_halt (void)
793 {
794 	pal_power_mgmt_info_u_t power_info[8];
795 	unsigned long min_power;
796 	int i, min_power_state;
797 
798 	if (ia64_pal_halt_info(power_info) != 0)
799 		return;
800 
801 	min_power_state = 0;
802 	min_power = power_info[0].pal_power_mgmt_info_s.power_consumption;
803 	for (i = 1; i < 8; ++i)
804 		if (power_info[i].pal_power_mgmt_info_s.im
805 		    && power_info[i].pal_power_mgmt_info_s.power_consumption < min_power) {
806 			min_power = power_info[i].pal_power_mgmt_info_s.power_consumption;
807 			min_power_state = i;
808 		}
809 
810 	while (1)
811 		ia64_pal_halt(min_power_state);
812 }
813 
814 void machine_shutdown(void)
815 {
816 #ifdef CONFIG_HOTPLUG_CPU
817 	int cpu;
818 
819 	for_each_online_cpu(cpu) {
820 		if (cpu != smp_processor_id())
821 			cpu_down(cpu);
822 	}
823 #endif
824 #ifdef CONFIG_KEXEC
825 	kexec_disable_iosapic();
826 #endif
827 }
828 
829 void
830 machine_restart (char *restart_cmd)
831 {
832 	(void) notify_die(DIE_MACHINE_RESTART, restart_cmd, NULL, 0, 0, 0);
833 	(*efi.reset_system)(EFI_RESET_WARM, 0, 0, NULL);
834 }
835 
836 void
837 machine_halt (void)
838 {
839 	(void) notify_die(DIE_MACHINE_HALT, "", NULL, 0, 0, 0);
840 	cpu_halt();
841 }
842 
843 void
844 machine_power_off (void)
845 {
846 	if (pm_power_off)
847 		pm_power_off();
848 	machine_halt();
849 }
850 
851