1 // SPDX-License-Identifier: GPL-2.0
3 * Architecture-specific setup.
5 * Copyright (C) 1998-2003 Hewlett-Packard Co
6 * David Mosberger-Tang <davidm@hpl.hp.com>
7 * 04/11/17 Ashok Raj <ashok.raj@intel.com> Added CPU Hotplug Support
9 * 2005-10-07 Keith Owens <kaos@sgi.com>
10 * Add notify_die() hooks.
12 #include <linux/cpu.h>
14 #include <linux/elf.h>
15 #include <linux/errno.h>
16 #include <linux/kernel.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/sched/debug.h>
24 #include <linux/sched/hotplug.h>
25 #include <linux/sched/task.h>
26 #include <linux/sched/task_stack.h>
27 #include <linux/stddef.h>
28 #include <linux/thread_info.h>
29 #include <linux/unistd.h>
30 #include <linux/efi.h>
31 #include <linux/interrupt.h>
32 #include <linux/delay.h>
33 #include <linux/kdebug.h>
34 #include <linux/utsname.h>
35 #include <linux/tracehook.h>
36 #include <linux/rcupdate.h>
39 #include <asm/delay.h>
42 #include <asm/kexec.h>
43 #include <asm/pgalloc.h>
44 #include <asm/processor.h>
46 #include <asm/switch_to.h>
47 #include <asm/tlbflush.h>
48 #include <linux/uaccess.h>
49 #include <asm/unwind.h>
55 # include <asm/perfmon.h>
60 void (*ia64_mark_idle
)(int);
62 unsigned long boot_option_idle_override
= IDLE_NO_OVERRIDE
;
63 EXPORT_SYMBOL(boot_option_idle_override
);
64 void (*pm_power_off
) (void);
65 EXPORT_SYMBOL(pm_power_off
);
68 ia64_do_show_stack (struct unw_frame_info
*info
, void *arg
)
70 unsigned long ip
, sp
, bsp
;
72 printk("\nCall Trace:\n");
74 unw_get_ip(info
, &ip
);
78 unw_get_sp(info
, &sp
);
79 unw_get_bsp(info
, &bsp
);
80 printk(" [<%016lx>] %pS\n"
81 " sp=%016lx bsp=%016lx\n",
82 ip
, (void *)ip
, sp
, bsp
);
83 } while (unw_unwind(info
) >= 0);
87 show_stack (struct task_struct
*task
, unsigned long *sp
)
90 unw_init_running(ia64_do_show_stack
, NULL
);
92 struct unw_frame_info info
;
94 unw_init_from_blocked_task(&info
, task
);
95 ia64_do_show_stack(&info
, NULL
);
100 show_regs (struct pt_regs
*regs
)
102 unsigned long ip
= regs
->cr_iip
+ ia64_psr(regs
)->ri
;
106 show_regs_print_info(KERN_DEFAULT
);
107 printk("psr : %016lx ifs : %016lx ip : [<%016lx>] %s (%s)\n",
108 regs
->cr_ipsr
, regs
->cr_ifs
, ip
, print_tainted(),
109 init_utsname()->release
);
110 printk("ip is at %pS\n", (void *)ip
);
111 printk("unat: %016lx pfs : %016lx rsc : %016lx\n",
112 regs
->ar_unat
, regs
->ar_pfs
, regs
->ar_rsc
);
113 printk("rnat: %016lx bsps: %016lx pr : %016lx\n",
114 regs
->ar_rnat
, regs
->ar_bspstore
, regs
->pr
);
115 printk("ldrs: %016lx ccv : %016lx fpsr: %016lx\n",
116 regs
->loadrs
, regs
->ar_ccv
, regs
->ar_fpsr
);
117 printk("csd : %016lx ssd : %016lx\n", regs
->ar_csd
, regs
->ar_ssd
);
118 printk("b0 : %016lx b6 : %016lx b7 : %016lx\n", regs
->b0
, regs
->b6
, regs
->b7
);
119 printk("f6 : %05lx%016lx f7 : %05lx%016lx\n",
120 regs
->f6
.u
.bits
[1], regs
->f6
.u
.bits
[0],
121 regs
->f7
.u
.bits
[1], regs
->f7
.u
.bits
[0]);
122 printk("f8 : %05lx%016lx f9 : %05lx%016lx\n",
123 regs
->f8
.u
.bits
[1], regs
->f8
.u
.bits
[0],
124 regs
->f9
.u
.bits
[1], regs
->f9
.u
.bits
[0]);
125 printk("f10 : %05lx%016lx f11 : %05lx%016lx\n",
126 regs
->f10
.u
.bits
[1], regs
->f10
.u
.bits
[0],
127 regs
->f11
.u
.bits
[1], regs
->f11
.u
.bits
[0]);
129 printk("r1 : %016lx r2 : %016lx r3 : %016lx\n", regs
->r1
, regs
->r2
, regs
->r3
);
130 printk("r8 : %016lx r9 : %016lx r10 : %016lx\n", regs
->r8
, regs
->r9
, regs
->r10
);
131 printk("r11 : %016lx r12 : %016lx r13 : %016lx\n", regs
->r11
, regs
->r12
, regs
->r13
);
132 printk("r14 : %016lx r15 : %016lx r16 : %016lx\n", regs
->r14
, regs
->r15
, regs
->r16
);
133 printk("r17 : %016lx r18 : %016lx r19 : %016lx\n", regs
->r17
, regs
->r18
, regs
->r19
);
134 printk("r20 : %016lx r21 : %016lx r22 : %016lx\n", regs
->r20
, regs
->r21
, regs
->r22
);
135 printk("r23 : %016lx r24 : %016lx r25 : %016lx\n", regs
->r23
, regs
->r24
, regs
->r25
);
136 printk("r26 : %016lx r27 : %016lx r28 : %016lx\n", regs
->r26
, regs
->r27
, regs
->r28
);
137 printk("r29 : %016lx r30 : %016lx r31 : %016lx\n", regs
->r29
, regs
->r30
, regs
->r31
);
139 if (user_mode(regs
)) {
140 /* print the stacked registers */
141 unsigned long val
, *bsp
, ndirty
;
142 int i
, sof
, is_nat
= 0;
144 sof
= regs
->cr_ifs
& 0x7f; /* size of frame */
145 ndirty
= (regs
->loadrs
>> 19);
146 bsp
= ia64_rse_skip_regs((unsigned long *) regs
->ar_bspstore
, ndirty
);
147 for (i
= 0; i
< sof
; ++i
) {
148 get_user(val
, (unsigned long __user
*) ia64_rse_skip_regs(bsp
, i
));
149 printk("r%-3u:%c%016lx%s", 32 + i
, is_nat
? '*' : ' ', val
,
150 ((i
== sof
- 1) || (i
% 3) == 2) ? "\n" : " ");
153 show_stack(NULL
, NULL
);
156 /* local support for deprecated console_print */
158 console_print(const char *s
)
160 printk(KERN_EMERG
"%s", s
);
164 do_notify_resume_user(sigset_t
*unused
, struct sigscratch
*scr
, long in_syscall
)
166 if (fsys_mode(current
, &scr
->pt
)) {
168 * defer signal-handling etc. until we return to
171 if (!ia64_psr(&scr
->pt
)->lp
)
172 ia64_psr(&scr
->pt
)->lp
= 1;
176 #ifdef CONFIG_PERFMON
177 if (current
->thread
.pfm_needs_checking
)
179 * Note: pfm_handle_work() allow us to call it with interrupts
180 * disabled, and may enable interrupts within the function.
185 /* deal with pending signal delivery */
186 if (test_thread_flag(TIF_SIGPENDING
)) {
187 local_irq_enable(); /* force interrupt enable */
188 ia64_do_signal(scr
, in_syscall
);
191 if (test_and_clear_thread_flag(TIF_NOTIFY_RESUME
)) {
192 local_irq_enable(); /* force interrupt enable */
193 tracehook_notify_resume(&scr
->pt
);
196 /* copy user rbs to kernel rbs */
197 if (unlikely(test_thread_flag(TIF_RESTORE_RSE
))) {
198 local_irq_enable(); /* force interrupt enable */
202 local_irq_disable(); /* force interrupt disable */
205 static int __init
nohalt_setup(char * str
)
207 cpu_idle_poll_ctrl(true);
210 __setup("nohalt", nohalt_setup
);
212 #ifdef CONFIG_HOTPLUG_CPU
213 /* We don't actually take CPU down, just spin without interrupts. */
214 static inline void play_dead(void)
216 unsigned int this_cpu
= smp_processor_id();
219 __this_cpu_write(cpu_state
, CPU_DEAD
);
224 ia64_jump_to_sal(&sal_boot_rendez_state
[this_cpu
]);
226 * The above is a point of no-return, the processor is
227 * expected to be in SAL loop now.
232 static inline void play_dead(void)
236 #endif /* CONFIG_HOTPLUG_CPU */
238 void arch_cpu_idle_dead(void)
243 void arch_cpu_idle(void)
245 void (*mark_idle
)(int) = ia64_mark_idle
;
264 ia64_save_extra (struct task_struct
*task
)
266 #ifdef CONFIG_PERFMON
270 if ((task
->thread
.flags
& IA64_THREAD_DBG_VALID
) != 0)
271 ia64_save_debug_regs(&task
->thread
.dbr
[0]);
273 #ifdef CONFIG_PERFMON
274 if ((task
->thread
.flags
& IA64_THREAD_PM_VALID
) != 0)
277 info
= __this_cpu_read(pfm_syst_info
);
278 if (info
& PFM_CPUINFO_SYST_WIDE
)
279 pfm_syst_wide_update_task(task
, info
, 0);
284 ia64_load_extra (struct task_struct
*task
)
286 #ifdef CONFIG_PERFMON
290 if ((task
->thread
.flags
& IA64_THREAD_DBG_VALID
) != 0)
291 ia64_load_debug_regs(&task
->thread
.dbr
[0]);
293 #ifdef CONFIG_PERFMON
294 if ((task
->thread
.flags
& IA64_THREAD_PM_VALID
) != 0)
297 info
= __this_cpu_read(pfm_syst_info
);
298 if (info
& PFM_CPUINFO_SYST_WIDE
)
299 pfm_syst_wide_update_task(task
, info
, 1);
304 * Copy the state of an ia-64 thread.
306 * We get here through the following call chain:
308 * from user-level: from kernel:
310 * <clone syscall> <some kernel call frames>
313 * copy_thread copy_thread
315 * This means that the stack layout is as follows:
317 * +---------------------+ (highest addr)
319 * +---------------------+
320 * | struct switch_stack |
321 * +---------------------+
324 * | | <-- sp (lowest addr)
325 * +---------------------+
327 * Observe that we copy the unat values that are in pt_regs and switch_stack. Spilling an
328 * integer to address X causes bit N in ar.unat to be set to the NaT bit of the register,
329 * with N=(X & 0x1ff)/8. Thus, copying the unat value preserves the NaT bits ONLY if the
330 * pt_regs structure in the parent is congruent to that of the child, modulo 512. Since
331 * the stack is page aligned and the page size is at least 4KB, this is always the case,
332 * so there is nothing to worry about.
335 copy_thread(unsigned long clone_flags
,
336 unsigned long user_stack_base
, unsigned long user_stack_size
,
337 struct task_struct
*p
)
339 extern char ia64_ret_from_clone
;
340 struct switch_stack
*child_stack
, *stack
;
341 unsigned long rbs
, child_rbs
, rbs_size
;
342 struct pt_regs
*child_ptregs
;
343 struct pt_regs
*regs
= current_pt_regs();
346 child_ptregs
= (struct pt_regs
*) ((unsigned long) p
+ IA64_STK_OFFSET
) - 1;
347 child_stack
= (struct switch_stack
*) child_ptregs
- 1;
349 rbs
= (unsigned long) current
+ IA64_RBS_OFFSET
;
350 child_rbs
= (unsigned long) p
+ IA64_RBS_OFFSET
;
352 /* copy parts of thread_struct: */
353 p
->thread
.ksp
= (unsigned long) child_stack
- 16;
356 * NOTE: The calling convention considers all floating point
357 * registers in the high partition (fph) to be scratch. Since
358 * the only way to get to this point is through a system call,
359 * we know that the values in fph are all dead. Hence, there
360 * is no need to inherit the fph state from the parent to the
361 * child and all we have to do is to make sure that
362 * IA64_THREAD_FPH_VALID is cleared in the child.
364 * XXX We could push this optimization a bit further by
365 * clearing IA64_THREAD_FPH_VALID on ANY system call.
366 * However, it's not clear this is worth doing. Also, it
367 * would be a slight deviation from the normal Linux system
368 * call behavior where scratch registers are preserved across
369 * system calls (unless used by the system call itself).
371 # define THREAD_FLAGS_TO_CLEAR (IA64_THREAD_FPH_VALID | IA64_THREAD_DBG_VALID \
372 | IA64_THREAD_PM_VALID)
373 # define THREAD_FLAGS_TO_SET 0
374 p
->thread
.flags
= ((current
->thread
.flags
& ~THREAD_FLAGS_TO_CLEAR
)
375 | THREAD_FLAGS_TO_SET
);
377 ia64_drop_fpu(p
); /* don't pick up stale state from a CPU's fph */
379 if (unlikely(p
->flags
& PF_KTHREAD
)) {
380 if (unlikely(!user_stack_base
)) {
381 /* fork_idle() called us */
384 memset(child_stack
, 0, sizeof(*child_ptregs
) + sizeof(*child_stack
));
385 child_stack
->r4
= user_stack_base
; /* payload */
386 child_stack
->r5
= user_stack_size
; /* argument */
388 * Preserve PSR bits, except for bits 32-34 and 37-45,
389 * which we can't read.
391 child_ptregs
->cr_ipsr
= ia64_getreg(_IA64_REG_PSR
) | IA64_PSR_BN
;
392 /* mark as valid, empty frame */
393 child_ptregs
->cr_ifs
= 1UL << 63;
394 child_stack
->ar_fpsr
= child_ptregs
->ar_fpsr
395 = ia64_getreg(_IA64_REG_AR_FPSR
);
396 child_stack
->pr
= (1 << PRED_KERNEL_STACK
);
397 child_stack
->ar_bspstore
= child_rbs
;
398 child_stack
->b0
= (unsigned long) &ia64_ret_from_clone
;
400 /* stop some PSR bits from being inherited.
401 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
402 * therefore we must specify them explicitly here and not include them in
403 * IA64_PSR_BITS_TO_CLEAR.
405 child_ptregs
->cr_ipsr
= ((child_ptregs
->cr_ipsr
| IA64_PSR_BITS_TO_SET
)
406 & ~(IA64_PSR_BITS_TO_CLEAR
| IA64_PSR_PP
| IA64_PSR_UP
));
410 stack
= ((struct switch_stack
*) regs
) - 1;
411 /* copy parent's switch_stack & pt_regs to child: */
412 memcpy(child_stack
, stack
, sizeof(*child_ptregs
) + sizeof(*child_stack
));
414 /* copy the parent's register backing store to the child: */
415 rbs_size
= stack
->ar_bspstore
- rbs
;
416 memcpy((void *) child_rbs
, (void *) rbs
, rbs_size
);
417 if (clone_flags
& CLONE_SETTLS
)
418 child_ptregs
->r13
= regs
->r16
; /* see sys_clone2() in entry.S */
419 if (user_stack_base
) {
420 child_ptregs
->r12
= user_stack_base
+ user_stack_size
- 16;
421 child_ptregs
->ar_bspstore
= user_stack_base
;
422 child_ptregs
->ar_rnat
= 0;
423 child_ptregs
->loadrs
= 0;
425 child_stack
->ar_bspstore
= child_rbs
+ rbs_size
;
426 child_stack
->b0
= (unsigned long) &ia64_ret_from_clone
;
428 /* stop some PSR bits from being inherited.
429 * the psr.up/psr.pp bits must be cleared on fork but inherited on execve()
430 * therefore we must specify them explicitly here and not include them in
431 * IA64_PSR_BITS_TO_CLEAR.
433 child_ptregs
->cr_ipsr
= ((child_ptregs
->cr_ipsr
| IA64_PSR_BITS_TO_SET
)
434 & ~(IA64_PSR_BITS_TO_CLEAR
| IA64_PSR_PP
| IA64_PSR_UP
));
436 #ifdef CONFIG_PERFMON
437 if (current
->thread
.pfm_context
)
438 pfm_inherit(p
, child_ptregs
);
444 do_copy_task_regs (struct task_struct
*task
, struct unw_frame_info
*info
, void *arg
)
446 unsigned long mask
, sp
, nat_bits
= 0, ar_rnat
, urbs_end
, cfm
;
447 unsigned long uninitialized_var(ip
); /* GCC be quiet */
448 elf_greg_t
*dst
= arg
;
453 memset(dst
, 0, sizeof(elf_gregset_t
)); /* don't leak any kernel bits to user-level */
455 if (unw_unwind_to_user(info
) < 0)
458 unw_get_sp(info
, &sp
);
459 pt
= (struct pt_regs
*) (sp
+ 16);
461 urbs_end
= ia64_get_user_rbs_end(task
, pt
, &cfm
);
463 if (ia64_sync_user_rbs(task
, info
->sw
, pt
->ar_bspstore
, urbs_end
) < 0)
466 ia64_peek(task
, info
->sw
, urbs_end
, (long) ia64_rse_rnat_addr((long *) urbs_end
),
472 * NaT bits (for r0-r31; bit N == 1 iff rN is a NaT)
473 * predicate registers (p0-p63)
476 * ar.rsc ar.bsp ar.bspstore ar.rnat
477 * ar.ccv ar.unat ar.fpsr ar.pfs ar.lc ar.ec
481 for (i
= 1, mask
= (1UL << i
); i
< 32; ++i
) {
482 unw_get_gr(info
, i
, &dst
[i
], &nat
);
488 unw_get_pr(info
, &dst
[33]);
490 for (i
= 0; i
< 8; ++i
)
491 unw_get_br(info
, i
, &dst
[34 + i
]);
493 unw_get_rp(info
, &ip
);
494 dst
[42] = ip
+ ia64_psr(pt
)->ri
;
496 dst
[44] = pt
->cr_ipsr
& IA64_PSR_UM
;
498 unw_get_ar(info
, UNW_AR_RSC
, &dst
[45]);
500 * For bsp and bspstore, unw_get_ar() would return the kernel
501 * addresses, but we need the user-level addresses instead:
503 dst
[46] = urbs_end
; /* note: by convention PT_AR_BSP points to the end of the urbs! */
504 dst
[47] = pt
->ar_bspstore
;
506 unw_get_ar(info
, UNW_AR_CCV
, &dst
[49]);
507 unw_get_ar(info
, UNW_AR_UNAT
, &dst
[50]);
508 unw_get_ar(info
, UNW_AR_FPSR
, &dst
[51]);
509 dst
[52] = pt
->ar_pfs
; /* UNW_AR_PFS is == to pt->cr_ifs for interrupt frames */
510 unw_get_ar(info
, UNW_AR_LC
, &dst
[53]);
511 unw_get_ar(info
, UNW_AR_EC
, &dst
[54]);
512 unw_get_ar(info
, UNW_AR_CSD
, &dst
[55]);
513 unw_get_ar(info
, UNW_AR_SSD
, &dst
[56]);
517 do_dump_task_fpu (struct task_struct
*task
, struct unw_frame_info
*info
, void *arg
)
519 elf_fpreg_t
*dst
= arg
;
522 memset(dst
, 0, sizeof(elf_fpregset_t
)); /* don't leak any "random" bits */
524 if (unw_unwind_to_user(info
) < 0)
527 /* f0 is 0.0, f1 is 1.0 */
529 for (i
= 2; i
< 32; ++i
)
530 unw_get_fr(info
, i
, dst
+ i
);
532 ia64_flush_fph(task
);
533 if ((task
->thread
.flags
& IA64_THREAD_FPH_VALID
) != 0)
534 memcpy(dst
+ 32, task
->thread
.fph
, 96*16);
538 do_copy_regs (struct unw_frame_info
*info
, void *arg
)
540 do_copy_task_regs(current
, info
, arg
);
544 do_dump_fpu (struct unw_frame_info
*info
, void *arg
)
546 do_dump_task_fpu(current
, info
, arg
);
550 ia64_elf_core_copy_regs (struct pt_regs
*pt
, elf_gregset_t dst
)
552 unw_init_running(do_copy_regs
, dst
);
556 dump_fpu (struct pt_regs
*pt
, elf_fpregset_t dst
)
558 unw_init_running(do_dump_fpu
, dst
);
559 return 1; /* f0-f31 are always valid so we always return 1 */
563 * Flush thread state. This is called when a thread does an execve().
568 /* drop floating-point and debug-register state if it exists: */
569 current
->thread
.flags
&= ~(IA64_THREAD_FPH_VALID
| IA64_THREAD_DBG_VALID
);
570 ia64_drop_fpu(current
);
574 * Clean up state associated with a thread. This is called when
575 * the thread calls exit().
578 exit_thread (struct task_struct
*tsk
)
582 #ifdef CONFIG_PERFMON
583 /* if needed, stop monitoring and flush state to perfmon context */
584 if (tsk
->thread
.pfm_context
)
585 pfm_exit_thread(tsk
);
587 /* free debug register resources */
588 if (tsk
->thread
.flags
& IA64_THREAD_DBG_VALID
)
589 pfm_release_debug_registers(tsk
);
594 get_wchan (struct task_struct
*p
)
596 struct unw_frame_info info
;
600 if (!p
|| p
== current
|| p
->state
== TASK_RUNNING
)
604 * Note: p may not be a blocked task (it could be current or
605 * another process running on some other CPU. Rather than
606 * trying to determine if p is really blocked, we just assume
607 * it's blocked and rely on the unwind routines to fail
608 * gracefully if the process wasn't really blocked after all.
611 unw_init_from_blocked_task(&info
, p
);
613 if (p
->state
== TASK_RUNNING
)
615 if (unw_unwind(&info
) < 0)
617 unw_get_ip(&info
, &ip
);
618 if (!in_sched_functions(ip
))
620 } while (count
++ < 16);
627 pal_power_mgmt_info_u_t power_info
[8];
628 unsigned long min_power
;
629 int i
, min_power_state
;
631 if (ia64_pal_halt_info(power_info
) != 0)
635 min_power
= power_info
[0].pal_power_mgmt_info_s
.power_consumption
;
636 for (i
= 1; i
< 8; ++i
)
637 if (power_info
[i
].pal_power_mgmt_info_s
.im
638 && power_info
[i
].pal_power_mgmt_info_s
.power_consumption
< min_power
) {
639 min_power
= power_info
[i
].pal_power_mgmt_info_s
.power_consumption
;
644 ia64_pal_halt(min_power_state
);
647 void machine_shutdown(void)
649 #ifdef CONFIG_HOTPLUG_CPU
652 for_each_online_cpu(cpu
) {
653 if (cpu
!= smp_processor_id())
658 kexec_disable_iosapic();
663 machine_restart (char *restart_cmd
)
665 (void) notify_die(DIE_MACHINE_RESTART
, restart_cmd
, NULL
, 0, 0, 0);
666 efi_reboot(REBOOT_WARM
, NULL
);
672 (void) notify_die(DIE_MACHINE_HALT
, "", NULL
, 0, 0, 0);
677 machine_power_off (void)