xref: /linux/arch/powerpc/kernel/head_64.S (revision d2912cb15bdda8ba4a5dd73396ad62641af2f520)
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 *  PowerPC version
4 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
5 *
6 *  Rewritten by Cort Dougan (cort@cs.nmt.edu) for PReP
7 *    Copyright (C) 1996 Cort Dougan <cort@cs.nmt.edu>
8 *  Adapted for Power Macintosh by Paul Mackerras.
9 *  Low-level exception handlers and MMU support
10 *  rewritten by Paul Mackerras.
11 *    Copyright (C) 1996 Paul Mackerras.
12 *
13 *  Adapted for 64bit PowerPC by Dave Engebretsen, Peter Bergner, and
14 *    Mike Corrigan {engebret|bergner|mikejc}@us.ibm.com
15 *
16 *  This file contains the entry point for the 64-bit kernel along
17 *  with some early initialization code common to all 64-bit powerpc
18 *  variants.
19 */
20
21#include <linux/threads.h>
22#include <linux/init.h>
23#include <asm/reg.h>
24#include <asm/page.h>
25#include <asm/mmu.h>
26#include <asm/ppc_asm.h>
27#include <asm/head-64.h>
28#include <asm/asm-offsets.h>
29#include <asm/bug.h>
30#include <asm/cputable.h>
31#include <asm/setup.h>
32#include <asm/hvcall.h>
33#include <asm/thread_info.h>
34#include <asm/firmware.h>
35#include <asm/page_64.h>
36#include <asm/irqflags.h>
37#include <asm/kvm_book3s_asm.h>
38#include <asm/ptrace.h>
39#include <asm/hw_irq.h>
40#include <asm/cputhreads.h>
41#include <asm/ppc-opcode.h>
42#include <asm/export.h>
43#include <asm/feature-fixups.h>
44
45/* The physical memory is laid out such that the secondary processor
46 * spin code sits at 0x0000...0x00ff. On server, the vectors follow
47 * using the layout described in exceptions-64s.S
48 */
49
50/*
51 * Entering into this code we make the following assumptions:
52 *
53 *  For pSeries or server processors:
54 *   1. The MMU is off & open firmware is running in real mode.
55 *   2. The primary CPU enters at __start.
56 *   3. If the RTAS supports "query-cpu-stopped-state", then secondary
57 *      CPUs will enter as directed by "start-cpu" RTAS call, which is
58 *      generic_secondary_smp_init, with PIR in r3.
59 *   4. Else the secondary CPUs will enter at secondary_hold (0x60) as
60 *      directed by the "start-cpu" RTS call, with PIR in r3.
61 * -or- For OPAL entry:
62 *   1. The MMU is off, processor in HV mode.
63 *   2. The primary CPU enters at 0 with device-tree in r3, OPAL base
64 *      in r8, and entry in r9 for debugging purposes.
65 *   3. Secondary CPUs enter as directed by OPAL_START_CPU call, which
66 *      is at generic_secondary_smp_init, with PIR in r3.
67 *
68 *  For Book3E processors:
69 *   1. The MMU is on running in AS0 in a state defined in ePAPR
70 *   2. The kernel is entered at __start
71 */
72
73OPEN_FIXED_SECTION(first_256B, 0x0, 0x100)
74USE_FIXED_SECTION(first_256B)
75	/*
76	 * Offsets are relative from the start of fixed section, and
77	 * first_256B starts at 0. Offsets are a bit easier to use here
78	 * than the fixed section entry macros.
79	 */
80	. = 0x0
81_GLOBAL(__start)
82	/* NOP this out unconditionally */
83BEGIN_FTR_SECTION
84	FIXUP_ENDIAN
85	b	__start_initialization_multiplatform
86END_FTR_SECTION(0, 1)
87
88	/* Catch branch to 0 in real mode */
89	trap
90
91	/* Secondary processors spin on this value until it becomes non-zero.
92	 * When non-zero, it contains the real address of the function the cpu
93	 * should jump to.
94	 */
95	.balign 8
96	.globl  __secondary_hold_spinloop
97__secondary_hold_spinloop:
98	.8byte	0x0
99
100	/* Secondary processors write this value with their cpu # */
101	/* after they enter the spin loop immediately below.	  */
102	.globl	__secondary_hold_acknowledge
103__secondary_hold_acknowledge:
104	.8byte	0x0
105
106#ifdef CONFIG_RELOCATABLE
107	/* This flag is set to 1 by a loader if the kernel should run
108	 * at the loaded address instead of the linked address.  This
109	 * is used by kexec-tools to keep the the kdump kernel in the
110	 * crash_kernel region.  The loader is responsible for
111	 * observing the alignment requirement.
112	 */
113
114#ifdef CONFIG_RELOCATABLE_TEST
115#define RUN_AT_LOAD_DEFAULT 1		/* Test relocation, do not copy to 0 */
116#else
117#define RUN_AT_LOAD_DEFAULT 0x72756e30  /* "run0" -- relocate to 0 by default */
118#endif
119
120	/* Do not move this variable as kexec-tools knows about it. */
121	. = 0x5c
122	.globl	__run_at_load
123__run_at_load:
124DEFINE_FIXED_SYMBOL(__run_at_load)
125	.long	RUN_AT_LOAD_DEFAULT
126#endif
127
128	. = 0x60
129/*
130 * The following code is used to hold secondary processors
131 * in a spin loop after they have entered the kernel, but
132 * before the bulk of the kernel has been relocated.  This code
133 * is relocated to physical address 0x60 before prom_init is run.
134 * All of it must fit below the first exception vector at 0x100.
135 * Use .globl here not _GLOBAL because we want __secondary_hold
136 * to be the actual text address, not a descriptor.
137 */
138	.globl	__secondary_hold
139__secondary_hold:
140	FIXUP_ENDIAN
141#ifndef CONFIG_PPC_BOOK3E
142	mfmsr	r24
143	ori	r24,r24,MSR_RI
144	mtmsrd	r24			/* RI on */
145#endif
146	/* Grab our physical cpu number */
147	mr	r24,r3
148	/* stash r4 for book3e */
149	mr	r25,r4
150
151	/* Tell the master cpu we're here */
152	/* Relocation is off & we are located at an address less */
153	/* than 0x100, so only need to grab low order offset.    */
154	std	r24,(ABS_ADDR(__secondary_hold_acknowledge))(0)
155	sync
156
157	li	r26,0
158#ifdef CONFIG_PPC_BOOK3E
159	tovirt(r26,r26)
160#endif
161	/* All secondary cpus wait here until told to start. */
162100:	ld	r12,(ABS_ADDR(__secondary_hold_spinloop))(r26)
163	cmpdi	0,r12,0
164	beq	100b
165
166#if defined(CONFIG_SMP) || defined(CONFIG_KEXEC_CORE)
167#ifdef CONFIG_PPC_BOOK3E
168	tovirt(r12,r12)
169#endif
170	mtctr	r12
171	mr	r3,r24
172	/*
173	 * it may be the case that other platforms have r4 right to
174	 * begin with, this gives us some safety in case it is not
175	 */
176#ifdef CONFIG_PPC_BOOK3E
177	mr	r4,r25
178#else
179	li	r4,0
180#endif
181	/* Make sure that patched code is visible */
182	isync
183	bctr
184#else
185	BUG_OPCODE
186#endif
187CLOSE_FIXED_SECTION(first_256B)
188
189/* This value is used to mark exception frames on the stack. */
190	.section ".toc","aw"
191exception_marker:
192	.tc	ID_72656773_68657265[TC],0x7265677368657265
193	.previous
194
195/*
196 * On server, we include the exception vectors code here as it
197 * relies on absolute addressing which is only possible within
198 * this compilation unit
199 */
200#ifdef CONFIG_PPC_BOOK3S
201#include "exceptions-64s.S"
202#else
203OPEN_TEXT_SECTION(0x100)
204#endif
205
206USE_TEXT_SECTION()
207
208#ifdef CONFIG_PPC_BOOK3E
209/*
210 * The booting_thread_hwid holds the thread id we want to boot in cpu
211 * hotplug case. It is set by cpu hotplug code, and is invalid by default.
212 * The thread id is the same as the initial value of SPRN_PIR[THREAD_ID]
213 * bit field.
214 */
215	.globl	booting_thread_hwid
216booting_thread_hwid:
217	.long  INVALID_THREAD_HWID
218	.align 3
219/*
220 * start a thread in the same core
221 * input parameters:
222 * r3 = the thread physical id
223 * r4 = the entry point where thread starts
224 */
225_GLOBAL(book3e_start_thread)
226	LOAD_REG_IMMEDIATE(r5, MSR_KERNEL)
227	cmpwi	r3, 0
228	beq	10f
229	cmpwi	r3, 1
230	beq	11f
231	/* If the thread id is invalid, just exit. */
232	b	13f
23310:
234	MTTMR(TMRN_IMSR0, 5)
235	MTTMR(TMRN_INIA0, 4)
236	b	12f
23711:
238	MTTMR(TMRN_IMSR1, 5)
239	MTTMR(TMRN_INIA1, 4)
24012:
241	isync
242	li	r6, 1
243	sld	r6, r6, r3
244	mtspr	SPRN_TENS, r6
24513:
246	blr
247
248/*
249 * stop a thread in the same core
250 * input parameter:
251 * r3 = the thread physical id
252 */
253_GLOBAL(book3e_stop_thread)
254	cmpwi	r3, 0
255	beq	10f
256	cmpwi	r3, 1
257	beq	10f
258	/* If the thread id is invalid, just exit. */
259	b	13f
26010:
261	li	r4, 1
262	sld	r4, r4, r3
263	mtspr	SPRN_TENC, r4
26413:
265	blr
266
267_GLOBAL(fsl_secondary_thread_init)
268	mfspr	r4,SPRN_BUCSR
269
270	/* Enable branch prediction */
271	lis     r3,BUCSR_INIT@h
272	ori     r3,r3,BUCSR_INIT@l
273	mtspr   SPRN_BUCSR,r3
274	isync
275
276	/*
277	 * Fix PIR to match the linear numbering in the device tree.
278	 *
279	 * On e6500, the reset value of PIR uses the low three bits for
280	 * the thread within a core, and the upper bits for the core
281	 * number.  There are two threads per core, so shift everything
282	 * but the low bit right by two bits so that the cpu numbering is
283	 * continuous.
284	 *
285	 * If the old value of BUCSR is non-zero, this thread has run
286	 * before.  Thus, we assume we are coming from kexec or a similar
287	 * scenario, and PIR is already set to the correct value.  This
288	 * is a bit of a hack, but there are limited opportunities for
289	 * getting information into the thread and the alternatives
290	 * seemed like they'd be overkill.  We can't tell just by looking
291	 * at the old PIR value which state it's in, since the same value
292	 * could be valid for one thread out of reset and for a different
293	 * thread in Linux.
294	 */
295
296	mfspr	r3, SPRN_PIR
297	cmpwi	r4,0
298	bne	1f
299	rlwimi	r3, r3, 30, 2, 30
300	mtspr	SPRN_PIR, r3
3011:
302#endif
303
304_GLOBAL(generic_secondary_thread_init)
305	mr	r24,r3
306
307	/* turn on 64-bit mode */
308	bl	enable_64b_mode
309
310	/* get a valid TOC pointer, wherever we're mapped at */
311	bl	relative_toc
312	tovirt(r2,r2)
313
314#ifdef CONFIG_PPC_BOOK3E
315	/* Book3E initialization */
316	mr	r3,r24
317	bl	book3e_secondary_thread_init
318#endif
319	b	generic_secondary_common_init
320
321/*
322 * On pSeries and most other platforms, secondary processors spin
323 * in the following code.
324 * At entry, r3 = this processor's number (physical cpu id)
325 *
326 * On Book3E, r4 = 1 to indicate that the initial TLB entry for
327 * this core already exists (setup via some other mechanism such
328 * as SCOM before entry).
329 */
330_GLOBAL(generic_secondary_smp_init)
331	FIXUP_ENDIAN
332	mr	r24,r3
333	mr	r25,r4
334
335	/* turn on 64-bit mode */
336	bl	enable_64b_mode
337
338	/* get a valid TOC pointer, wherever we're mapped at */
339	bl	relative_toc
340	tovirt(r2,r2)
341
342#ifdef CONFIG_PPC_BOOK3E
343	/* Book3E initialization */
344	mr	r3,r24
345	mr	r4,r25
346	bl	book3e_secondary_core_init
347
348/*
349 * After common core init has finished, check if the current thread is the
350 * one we wanted to boot. If not, start the specified thread and stop the
351 * current thread.
352 */
353	LOAD_REG_ADDR(r4, booting_thread_hwid)
354	lwz     r3, 0(r4)
355	li	r5, INVALID_THREAD_HWID
356	cmpw	r3, r5
357	beq	20f
358
359	/*
360	 * The value of booting_thread_hwid has been stored in r3,
361	 * so make it invalid.
362	 */
363	stw	r5, 0(r4)
364
365	/*
366	 * Get the current thread id and check if it is the one we wanted.
367	 * If not, start the one specified in booting_thread_hwid and stop
368	 * the current thread.
369	 */
370	mfspr	r8, SPRN_TIR
371	cmpw	r3, r8
372	beq	20f
373
374	/* start the specified thread */
375	LOAD_REG_ADDR(r5, fsl_secondary_thread_init)
376	ld	r4, 0(r5)
377	bl	book3e_start_thread
378
379	/* stop the current thread */
380	mr	r3, r8
381	bl	book3e_stop_thread
38210:
383	b	10b
38420:
385#endif
386
387generic_secondary_common_init:
388	/* Set up a paca value for this processor. Since we have the
389	 * physical cpu id in r24, we need to search the pacas to find
390	 * which logical id maps to our physical one.
391	 */
392#ifndef CONFIG_SMP
393	b	kexec_wait		/* wait for next kernel if !SMP	 */
394#else
395	LOAD_REG_ADDR(r8, paca_ptrs)	/* Load paca_ptrs pointe	 */
396	ld	r8,0(r8)		/* Get base vaddr of array	 */
397	LOAD_REG_ADDR(r7, nr_cpu_ids)	/* Load nr_cpu_ids address       */
398	lwz	r7,0(r7)		/* also the max paca allocated 	 */
399	li	r5,0			/* logical cpu id                */
4001:
401	sldi	r9,r5,3			/* get paca_ptrs[] index from cpu id */
402	ldx	r13,r9,r8		/* r13 = paca_ptrs[cpu id]       */
403	lhz	r6,PACAHWCPUID(r13)	/* Load HW procid from paca      */
404	cmpw	r6,r24			/* Compare to our id             */
405	beq	2f
406	addi	r5,r5,1
407	cmpw	r5,r7			/* Check if more pacas exist     */
408	blt	1b
409
410	mr	r3,r24			/* not found, copy phys to r3	 */
411	b	kexec_wait		/* next kernel might do better	 */
412
4132:	SET_PACA(r13)
414#ifdef CONFIG_PPC_BOOK3E
415	addi	r12,r13,PACA_EXTLB	/* and TLB exc frame in another  */
416	mtspr	SPRN_SPRG_TLB_EXFRAME,r12
417#endif
418
419	/* From now on, r24 is expected to be logical cpuid */
420	mr	r24,r5
421
422	/* See if we need to call a cpu state restore handler */
423	LOAD_REG_ADDR(r23, cur_cpu_spec)
424	ld	r23,0(r23)
425	ld	r12,CPU_SPEC_RESTORE(r23)
426	cmpdi	0,r12,0
427	beq	3f
428#ifdef PPC64_ELF_ABI_v1
429	ld	r12,0(r12)
430#endif
431	mtctr	r12
432	bctrl
433
4343:	LOAD_REG_ADDR(r3, spinning_secondaries) /* Decrement spinning_secondaries */
435	lwarx	r4,0,r3
436	subi	r4,r4,1
437	stwcx.	r4,0,r3
438	bne	3b
439	isync
440
4414:	HMT_LOW
442	lbz	r23,PACAPROCSTART(r13)	/* Test if this processor should */
443					/* start.			 */
444	cmpwi	0,r23,0
445	beq	4b			/* Loop until told to go	 */
446
447	sync				/* order paca.run and cur_cpu_spec */
448	isync				/* In case code patching happened */
449
450	/* Create a temp kernel stack for use before relocation is on.	*/
451	ld	r1,PACAEMERGSP(r13)
452	subi	r1,r1,STACK_FRAME_OVERHEAD
453
454	b	__secondary_start
455#endif /* SMP */
456
457/*
458 * Turn the MMU off.
459 * Assumes we're mapped EA == RA if the MMU is on.
460 */
461#ifdef CONFIG_PPC_BOOK3S
462__mmu_off:
463	mfmsr	r3
464	andi.	r0,r3,MSR_IR|MSR_DR
465	beqlr
466	mflr	r4
467	andc	r3,r3,r0
468	mtspr	SPRN_SRR0,r4
469	mtspr	SPRN_SRR1,r3
470	sync
471	rfid
472	b	.	/* prevent speculative execution */
473#endif
474
475
476/*
477 * Here is our main kernel entry point. We support currently 2 kind of entries
478 * depending on the value of r5.
479 *
480 *   r5 != NULL -> OF entry, we go to prom_init, "legacy" parameter content
481 *                 in r3...r7
482 *
483 *   r5 == NULL -> kexec style entry. r3 is a physical pointer to the
484 *                 DT block, r4 is a physical pointer to the kernel itself
485 *
486 */
487__start_initialization_multiplatform:
488	/* Make sure we are running in 64 bits mode */
489	bl	enable_64b_mode
490
491	/* Get TOC pointer (current runtime address) */
492	bl	relative_toc
493
494	/* find out where we are now */
495	bcl	20,31,$+4
4960:	mflr	r26			/* r26 = runtime addr here */
497	addis	r26,r26,(_stext - 0b)@ha
498	addi	r26,r26,(_stext - 0b)@l	/* current runtime base addr */
499
500	/*
501	 * Are we booted from a PROM Of-type client-interface ?
502	 */
503	cmpldi	cr0,r5,0
504	beq	1f
505	b	__boot_from_prom		/* yes -> prom */
5061:
507	/* Save parameters */
508	mr	r31,r3
509	mr	r30,r4
510#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
511	/* Save OPAL entry */
512	mr	r28,r8
513	mr	r29,r9
514#endif
515
516#ifdef CONFIG_PPC_BOOK3E
517	bl	start_initialization_book3e
518	b	__after_prom_start
519#else
520	/* Setup some critical 970 SPRs before switching MMU off */
521	mfspr	r0,SPRN_PVR
522	srwi	r0,r0,16
523	cmpwi	r0,0x39		/* 970 */
524	beq	1f
525	cmpwi	r0,0x3c		/* 970FX */
526	beq	1f
527	cmpwi	r0,0x44		/* 970MP */
528	beq	1f
529	cmpwi	r0,0x45		/* 970GX */
530	bne	2f
5311:	bl	__cpu_preinit_ppc970
5322:
533
534	/* Switch off MMU if not already off */
535	bl	__mmu_off
536	b	__after_prom_start
537#endif /* CONFIG_PPC_BOOK3E */
538
539__boot_from_prom:
540#ifdef CONFIG_PPC_OF_BOOT_TRAMPOLINE
541	/* Save parameters */
542	mr	r31,r3
543	mr	r30,r4
544	mr	r29,r5
545	mr	r28,r6
546	mr	r27,r7
547
548	/*
549	 * Align the stack to 16-byte boundary
550	 * Depending on the size and layout of the ELF sections in the initial
551	 * boot binary, the stack pointer may be unaligned on PowerMac
552	 */
553	rldicr	r1,r1,0,59
554
555#ifdef CONFIG_RELOCATABLE
556	/* Relocate code for where we are now */
557	mr	r3,r26
558	bl	relocate
559#endif
560
561	/* Restore parameters */
562	mr	r3,r31
563	mr	r4,r30
564	mr	r5,r29
565	mr	r6,r28
566	mr	r7,r27
567
568	/* Do all of the interaction with OF client interface */
569	mr	r8,r26
570	bl	prom_init
571#endif /* #CONFIG_PPC_OF_BOOT_TRAMPOLINE */
572
573	/* We never return. We also hit that trap if trying to boot
574	 * from OF while CONFIG_PPC_OF_BOOT_TRAMPOLINE isn't selected */
575	trap
576
577__after_prom_start:
578#ifdef CONFIG_RELOCATABLE
579	/* process relocations for the final address of the kernel */
580	lis	r25,PAGE_OFFSET@highest	/* compute virtual base of kernel */
581	sldi	r25,r25,32
582#if defined(CONFIG_PPC_BOOK3E)
583	tovirt(r26,r26)		/* on booke, we already run at PAGE_OFFSET */
584#endif
585	lwz	r7,(FIXED_SYMBOL_ABS_ADDR(__run_at_load))(r26)
586#if defined(CONFIG_PPC_BOOK3E)
587	tophys(r26,r26)
588#endif
589	cmplwi	cr0,r7,1	/* flagged to stay where we are ? */
590	bne	1f
591	add	r25,r25,r26
5921:	mr	r3,r25
593	bl	relocate
594#if defined(CONFIG_PPC_BOOK3E)
595	/* IVPR needs to be set after relocation. */
596	bl	init_core_book3e
597#endif
598#endif
599
600/*
601 * We need to run with _stext at physical address PHYSICAL_START.
602 * This will leave some code in the first 256B of
603 * real memory, which are reserved for software use.
604 *
605 * Note: This process overwrites the OF exception vectors.
606 */
607	li	r3,0			/* target addr */
608#ifdef CONFIG_PPC_BOOK3E
609	tovirt(r3,r3)		/* on booke, we already run at PAGE_OFFSET */
610#endif
611	mr.	r4,r26			/* In some cases the loader may  */
612#if defined(CONFIG_PPC_BOOK3E)
613	tovirt(r4,r4)
614#endif
615	beq	9f			/* have already put us at zero */
616	li	r6,0x100		/* Start offset, the first 0x100 */
617					/* bytes were copied earlier.	 */
618
619#ifdef CONFIG_RELOCATABLE
620/*
621 * Check if the kernel has to be running as relocatable kernel based on the
622 * variable __run_at_load, if it is set the kernel is treated as relocatable
623 * kernel, otherwise it will be moved to PHYSICAL_START
624 */
625#if defined(CONFIG_PPC_BOOK3E)
626	tovirt(r26,r26)		/* on booke, we already run at PAGE_OFFSET */
627#endif
628	lwz	r7,(FIXED_SYMBOL_ABS_ADDR(__run_at_load))(r26)
629	cmplwi	cr0,r7,1
630	bne	3f
631
632#ifdef CONFIG_PPC_BOOK3E
633	LOAD_REG_ADDR(r5, __end_interrupts)
634	LOAD_REG_ADDR(r11, _stext)
635	sub	r5,r5,r11
636#else
637	/* just copy interrupts */
638	LOAD_REG_IMMEDIATE(r5, FIXED_SYMBOL_ABS_ADDR(__end_interrupts))
639#endif
640	b	5f
6413:
642#endif
643	/* # bytes of memory to copy */
644	lis	r5,(ABS_ADDR(copy_to_here))@ha
645	addi	r5,r5,(ABS_ADDR(copy_to_here))@l
646
647	bl	copy_and_flush		/* copy the first n bytes	 */
648					/* this includes the code being	 */
649					/* executed here.		 */
650	/* Jump to the copy of this code that we just made */
651	addis	r8,r3,(ABS_ADDR(4f))@ha
652	addi	r12,r8,(ABS_ADDR(4f))@l
653	mtctr	r12
654	bctr
655
656.balign 8
657p_end: .8byte _end - copy_to_here
658
6594:
660	/*
661	 * Now copy the rest of the kernel up to _end, add
662	 * _end - copy_to_here to the copy limit and run again.
663	 */
664	addis   r8,r26,(ABS_ADDR(p_end))@ha
665	ld      r8,(ABS_ADDR(p_end))@l(r8)
666	add	r5,r5,r8
6675:	bl	copy_and_flush		/* copy the rest */
668
6699:	b	start_here_multiplatform
670
671/*
672 * Copy routine used to copy the kernel to start at physical address 0
673 * and flush and invalidate the caches as needed.
674 * r3 = dest addr, r4 = source addr, r5 = copy limit, r6 = start offset
675 * on exit, r3, r4, r5 are unchanged, r6 is updated to be >= r5.
676 *
677 * Note: this routine *only* clobbers r0, r6 and lr
678 */
679_GLOBAL(copy_and_flush)
680	addi	r5,r5,-8
681	addi	r6,r6,-8
6824:	li	r0,8			/* Use the smallest common	*/
683					/* denominator cache line	*/
684					/* size.  This results in	*/
685					/* extra cache line flushes	*/
686					/* but operation is correct.	*/
687					/* Can't get cache line size	*/
688					/* from NACA as it is being	*/
689					/* moved too.			*/
690
691	mtctr	r0			/* put # words/line in ctr	*/
6923:	addi	r6,r6,8			/* copy a cache line		*/
693	ldx	r0,r6,r4
694	stdx	r0,r6,r3
695	bdnz	3b
696	dcbst	r6,r3			/* write it to memory		*/
697	sync
698	icbi	r6,r3			/* flush the icache line	*/
699	cmpld	0,r6,r5
700	blt	4b
701	sync
702	addi	r5,r5,8
703	addi	r6,r6,8
704	isync
705	blr
706
707.align 8
708copy_to_here:
709
710#ifdef CONFIG_SMP
711#ifdef CONFIG_PPC_PMAC
712/*
713 * On PowerMac, secondary processors starts from the reset vector, which
714 * is temporarily turned into a call to one of the functions below.
715 */
716	.section ".text";
717	.align 2 ;
718
719	.globl	__secondary_start_pmac_0
720__secondary_start_pmac_0:
721	/* NB the entries for cpus 0, 1, 2 must each occupy 8 bytes. */
722	li	r24,0
723	b	1f
724	li	r24,1
725	b	1f
726	li	r24,2
727	b	1f
728	li	r24,3
7291:
730
731_GLOBAL(pmac_secondary_start)
732	/* turn on 64-bit mode */
733	bl	enable_64b_mode
734
735	li	r0,0
736	mfspr	r3,SPRN_HID4
737	rldimi	r3,r0,40,23	/* clear bit 23 (rm_ci) */
738	sync
739	mtspr	SPRN_HID4,r3
740	isync
741	sync
742	slbia
743
744	/* get TOC pointer (real address) */
745	bl	relative_toc
746	tovirt(r2,r2)
747
748	/* Copy some CPU settings from CPU 0 */
749	bl	__restore_cpu_ppc970
750
751	/* pSeries do that early though I don't think we really need it */
752	mfmsr	r3
753	ori	r3,r3,MSR_RI
754	mtmsrd	r3			/* RI on */
755
756	/* Set up a paca value for this processor. */
757	LOAD_REG_ADDR(r4,paca_ptrs)	/* Load paca pointer		*/
758	ld	r4,0(r4)		/* Get base vaddr of paca_ptrs array */
759	sldi	r5,r24,3		/* get paca_ptrs[] index from cpu id */
760	ldx	r13,r5,r4		/* r13 = paca_ptrs[cpu id]       */
761	SET_PACA(r13)			/* Save vaddr of paca in an SPRG*/
762
763	/* Mark interrupts soft and hard disabled (they might be enabled
764	 * in the PACA when doing hotplug)
765	 */
766	li	r0,IRQS_DISABLED
767	stb	r0,PACAIRQSOFTMASK(r13)
768	li	r0,PACA_IRQ_HARD_DIS
769	stb	r0,PACAIRQHAPPENED(r13)
770
771	/* Create a temp kernel stack for use before relocation is on.	*/
772	ld	r1,PACAEMERGSP(r13)
773	subi	r1,r1,STACK_FRAME_OVERHEAD
774
775	b	__secondary_start
776
777#endif /* CONFIG_PPC_PMAC */
778
779/*
780 * This function is called after the master CPU has released the
781 * secondary processors.  The execution environment is relocation off.
782 * The paca for this processor has the following fields initialized at
783 * this point:
784 *   1. Processor number
785 *   2. Segment table pointer (virtual address)
786 * On entry the following are set:
787 *   r1	       = stack pointer (real addr of temp stack)
788 *   r24       = cpu# (in Linux terms)
789 *   r13       = paca virtual address
790 *   SPRG_PACA = paca virtual address
791 */
792	.section ".text";
793	.align 2 ;
794
795	.globl	__secondary_start
796__secondary_start:
797	/* Set thread priority to MEDIUM */
798	HMT_MEDIUM
799
800	/*
801	 * Do early setup for this CPU, in particular initialising the MMU so we
802	 * can turn it on below. This is a call to C, which is OK, we're still
803	 * running on the emergency stack.
804	 */
805	bl	early_setup_secondary
806
807	/*
808	 * The primary has initialized our kernel stack for us in the paca, grab
809	 * it and put it in r1. We must *not* use it until we turn on the MMU
810	 * below, because it may not be inside the RMO.
811	 */
812	ld	r1, PACAKSAVE(r13)
813
814	/* Clear backchain so we get nice backtraces */
815	li	r7,0
816	mtlr	r7
817
818	/* Mark interrupts soft and hard disabled (they might be enabled
819	 * in the PACA when doing hotplug)
820	 */
821	li	r7,IRQS_DISABLED
822	stb	r7,PACAIRQSOFTMASK(r13)
823	li	r0,PACA_IRQ_HARD_DIS
824	stb	r0,PACAIRQHAPPENED(r13)
825
826	/* enable MMU and jump to start_secondary */
827	LOAD_REG_ADDR(r3, start_secondary_prolog)
828	LOAD_REG_IMMEDIATE(r4, MSR_KERNEL)
829
830	mtspr	SPRN_SRR0,r3
831	mtspr	SPRN_SRR1,r4
832	RFI
833	b	.	/* prevent speculative execution */
834
835/*
836 * Running with relocation on at this point.  All we want to do is
837 * zero the stack back-chain pointer and get the TOC virtual address
838 * before going into C code.
839 */
840start_secondary_prolog:
841	ld	r2,PACATOC(r13)
842	li	r3,0
843	std	r3,0(r1)		/* Zero the stack frame pointer	*/
844	bl	start_secondary
845	b	.
846/*
847 * Reset stack pointer and call start_secondary
848 * to continue with online operation when woken up
849 * from cede in cpu offline.
850 */
851_GLOBAL(start_secondary_resume)
852	ld	r1,PACAKSAVE(r13)	/* Reload kernel stack pointer */
853	li	r3,0
854	std	r3,0(r1)		/* Zero the stack frame pointer	*/
855	bl	start_secondary
856	b	.
857#endif
858
859/*
860 * This subroutine clobbers r11 and r12
861 */
862enable_64b_mode:
863	mfmsr	r11			/* grab the current MSR */
864#ifdef CONFIG_PPC_BOOK3E
865	oris	r11,r11,0x8000		/* CM bit set, we'll set ICM later */
866	mtmsr	r11
867#else /* CONFIG_PPC_BOOK3E */
868	li	r12,(MSR_64BIT | MSR_ISF)@highest
869	sldi	r12,r12,48
870	or	r11,r11,r12
871	mtmsrd	r11
872	isync
873#endif
874	blr
875
876/*
877 * This puts the TOC pointer into r2, offset by 0x8000 (as expected
878 * by the toolchain).  It computes the correct value for wherever we
879 * are running at the moment, using position-independent code.
880 *
881 * Note: The compiler constructs pointers using offsets from the
882 * TOC in -mcmodel=medium mode. After we relocate to 0 but before
883 * the MMU is on we need our TOC to be a virtual address otherwise
884 * these pointers will be real addresses which may get stored and
885 * accessed later with the MMU on. We use tovirt() at the call
886 * sites to handle this.
887 */
888_GLOBAL(relative_toc)
889	mflr	r0
890	bcl	20,31,$+4
8910:	mflr	r11
892	ld	r2,(p_toc - 0b)(r11)
893	add	r2,r2,r11
894	mtlr	r0
895	blr
896
897.balign 8
898p_toc:	.8byte	__toc_start + 0x8000 - 0b
899
900/*
901 * This is where the main kernel code starts.
902 */
903start_here_multiplatform:
904	/* set up the TOC */
905	bl      relative_toc
906	tovirt(r2,r2)
907
908	/* Clear out the BSS. It may have been done in prom_init,
909	 * already but that's irrelevant since prom_init will soon
910	 * be detached from the kernel completely. Besides, we need
911	 * to clear it now for kexec-style entry.
912	 */
913	LOAD_REG_ADDR(r11,__bss_stop)
914	LOAD_REG_ADDR(r8,__bss_start)
915	sub	r11,r11,r8		/* bss size			*/
916	addi	r11,r11,7		/* round up to an even double word */
917	srdi.	r11,r11,3		/* shift right by 3		*/
918	beq	4f
919	addi	r8,r8,-8
920	li	r0,0
921	mtctr	r11			/* zero this many doublewords	*/
9223:	stdu	r0,8(r8)
923	bdnz	3b
9244:
925
926#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
927	/* Setup OPAL entry */
928	LOAD_REG_ADDR(r11, opal)
929	std	r28,0(r11);
930	std	r29,8(r11);
931#endif
932
933#ifndef CONFIG_PPC_BOOK3E
934	mfmsr	r6
935	ori	r6,r6,MSR_RI
936	mtmsrd	r6			/* RI on */
937#endif
938
939#ifdef CONFIG_RELOCATABLE
940	/* Save the physical address we're running at in kernstart_addr */
941	LOAD_REG_ADDR(r4, kernstart_addr)
942	clrldi	r0,r25,2
943	std	r0,0(r4)
944#endif
945
946	/* The following gets the stack set up with the regs */
947	/* pointing to the real addr of the kernel stack.  This is   */
948	/* all done to support the C function call below which sets  */
949	/* up the htab.  This is done because we have relocated the  */
950	/* kernel but are still running in real mode. */
951
952	LOAD_REG_ADDR(r3,init_thread_union)
953
954	/* set up a stack pointer */
955	LOAD_REG_IMMEDIATE(r1,THREAD_SIZE)
956	add	r1,r3,r1
957	li	r0,0
958	stdu	r0,-STACK_FRAME_OVERHEAD(r1)
959
960	/*
961	 * Do very early kernel initializations, including initial hash table
962	 * and SLB setup before we turn on relocation.
963	 */
964
965	/* Restore parameters passed from prom_init/kexec */
966	mr	r3,r31
967	LOAD_REG_ADDR(r12, DOTSYM(early_setup))
968	mtctr	r12
969	bctrl		/* also sets r13 and SPRG_PACA */
970
971	LOAD_REG_ADDR(r3, start_here_common)
972	ld	r4,PACAKMSR(r13)
973	mtspr	SPRN_SRR0,r3
974	mtspr	SPRN_SRR1,r4
975	RFI
976	b	.	/* prevent speculative execution */
977
978	/* This is where all platforms converge execution */
979
980start_here_common:
981	/* relocation is on at this point */
982	std	r1,PACAKSAVE(r13)
983
984	/* Load the TOC (virtual address) */
985	ld	r2,PACATOC(r13)
986
987	/* Mark interrupts soft and hard disabled (they might be enabled
988	 * in the PACA when doing hotplug)
989	 */
990	li	r0,IRQS_DISABLED
991	stb	r0,PACAIRQSOFTMASK(r13)
992	li	r0,PACA_IRQ_HARD_DIS
993	stb	r0,PACAIRQHAPPENED(r13)
994
995	/* Generic kernel entry */
996	bl	start_kernel
997
998	/* Not reached */
999	BUG_OPCODE
1000
1001/*
1002 * We put a few things here that have to be page-aligned.
1003 * This stuff goes at the beginning of the bss, which is page-aligned.
1004 */
1005	.section ".bss"
1006/*
1007 * pgd dir should be aligned to PGD_TABLE_SIZE which is 64K.
1008 * We will need to find a better way to fix this
1009 */
1010	.align	16
1011
1012	.globl	swapper_pg_dir
1013swapper_pg_dir:
1014	.space	PGD_TABLE_SIZE
1015
1016	.globl	empty_zero_page
1017empty_zero_page:
1018	.space	PAGE_SIZE
1019EXPORT_SYMBOL(empty_zero_page)
1020