xref: /openbmc/linux/arch/powerpc/kernel/head_64.S (revision 750afb08)
1/*
2 *  PowerPC version
3 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
4 *
5 *  Rewritten by Cort Dougan (cort@cs.nmt.edu) for PReP
6 *    Copyright (C) 1996 Cort Dougan <cort@cs.nmt.edu>
7 *  Adapted for Power Macintosh by Paul Mackerras.
8 *  Low-level exception handlers and MMU support
9 *  rewritten by Paul Mackerras.
10 *    Copyright (C) 1996 Paul Mackerras.
11 *
12 *  Adapted for 64bit PowerPC by Dave Engebretsen, Peter Bergner, and
13 *    Mike Corrigan {engebret|bergner|mikejc}@us.ibm.com
14 *
15 *  This file contains the entry point for the 64-bit kernel along
16 *  with some early initialization code common to all 64-bit powerpc
17 *  variants.
18 *
19 *  This program is free software; you can redistribute it and/or
20 *  modify it under the terms of the GNU General Public License
21 *  as published by the Free Software Foundation; either version
22 *  2 of the License, or (at your option) any later version.
23 */
24
25#include <linux/threads.h>
26#include <linux/init.h>
27#include <asm/reg.h>
28#include <asm/page.h>
29#include <asm/mmu.h>
30#include <asm/ppc_asm.h>
31#include <asm/head-64.h>
32#include <asm/asm-offsets.h>
33#include <asm/bug.h>
34#include <asm/cputable.h>
35#include <asm/setup.h>
36#include <asm/hvcall.h>
37#include <asm/thread_info.h>
38#include <asm/firmware.h>
39#include <asm/page_64.h>
40#include <asm/irqflags.h>
41#include <asm/kvm_book3s_asm.h>
42#include <asm/ptrace.h>
43#include <asm/hw_irq.h>
44#include <asm/cputhreads.h>
45#include <asm/ppc-opcode.h>
46#include <asm/export.h>
47#include <asm/feature-fixups.h>
48
49/* The physical memory is laid out such that the secondary processor
50 * spin code sits at 0x0000...0x00ff. On server, the vectors follow
51 * using the layout described in exceptions-64s.S
52 */
53
54/*
55 * Entering into this code we make the following assumptions:
56 *
57 *  For pSeries or server processors:
58 *   1. The MMU is off & open firmware is running in real mode.
59 *   2. The primary CPU enters at __start.
60 *   3. If the RTAS supports "query-cpu-stopped-state", then secondary
61 *      CPUs will enter as directed by "start-cpu" RTAS call, which is
62 *      generic_secondary_smp_init, with PIR in r3.
63 *   4. Else the secondary CPUs will enter at secondary_hold (0x60) as
64 *      directed by the "start-cpu" RTS call, with PIR in r3.
65 * -or- For OPAL entry:
66 *   1. The MMU is off, processor in HV mode.
67 *   2. The primary CPU enters at 0 with device-tree in r3, OPAL base
68 *      in r8, and entry in r9 for debugging purposes.
69 *   3. Secondary CPUs enter as directed by OPAL_START_CPU call, which
70 *      is at generic_secondary_smp_init, with PIR in r3.
71 *
72 *  For Book3E processors:
73 *   1. The MMU is on running in AS0 in a state defined in ePAPR
74 *   2. The kernel is entered at __start
75 */
76
77OPEN_FIXED_SECTION(first_256B, 0x0, 0x100)
78USE_FIXED_SECTION(first_256B)
79	/*
80	 * Offsets are relative from the start of fixed section, and
81	 * first_256B starts at 0. Offsets are a bit easier to use here
82	 * than the fixed section entry macros.
83	 */
84	. = 0x0
85_GLOBAL(__start)
86	/* NOP this out unconditionally */
87BEGIN_FTR_SECTION
88	FIXUP_ENDIAN
89	b	__start_initialization_multiplatform
90END_FTR_SECTION(0, 1)
91
92	/* Catch branch to 0 in real mode */
93	trap
94
95	/* Secondary processors spin on this value until it becomes non-zero.
96	 * When non-zero, it contains the real address of the function the cpu
97	 * should jump to.
98	 */
99	.balign 8
100	.globl  __secondary_hold_spinloop
101__secondary_hold_spinloop:
102	.8byte	0x0
103
104	/* Secondary processors write this value with their cpu # */
105	/* after they enter the spin loop immediately below.	  */
106	.globl	__secondary_hold_acknowledge
107__secondary_hold_acknowledge:
108	.8byte	0x0
109
110#ifdef CONFIG_RELOCATABLE
111	/* This flag is set to 1 by a loader if the kernel should run
112	 * at the loaded address instead of the linked address.  This
113	 * is used by kexec-tools to keep the the kdump kernel in the
114	 * crash_kernel region.  The loader is responsible for
115	 * observing the alignment requirement.
116	 */
117
118#ifdef CONFIG_RELOCATABLE_TEST
119#define RUN_AT_LOAD_DEFAULT 1		/* Test relocation, do not copy to 0 */
120#else
121#define RUN_AT_LOAD_DEFAULT 0x72756e30  /* "run0" -- relocate to 0 by default */
122#endif
123
124	/* Do not move this variable as kexec-tools knows about it. */
125	. = 0x5c
126	.globl	__run_at_load
127__run_at_load:
128DEFINE_FIXED_SYMBOL(__run_at_load)
129	.long	RUN_AT_LOAD_DEFAULT
130#endif
131
132	. = 0x60
133/*
134 * The following code is used to hold secondary processors
135 * in a spin loop after they have entered the kernel, but
136 * before the bulk of the kernel has been relocated.  This code
137 * is relocated to physical address 0x60 before prom_init is run.
138 * All of it must fit below the first exception vector at 0x100.
139 * Use .globl here not _GLOBAL because we want __secondary_hold
140 * to be the actual text address, not a descriptor.
141 */
142	.globl	__secondary_hold
143__secondary_hold:
144	FIXUP_ENDIAN
145#ifndef CONFIG_PPC_BOOK3E
146	mfmsr	r24
147	ori	r24,r24,MSR_RI
148	mtmsrd	r24			/* RI on */
149#endif
150	/* Grab our physical cpu number */
151	mr	r24,r3
152	/* stash r4 for book3e */
153	mr	r25,r4
154
155	/* Tell the master cpu we're here */
156	/* Relocation is off & we are located at an address less */
157	/* than 0x100, so only need to grab low order offset.    */
158	std	r24,(ABS_ADDR(__secondary_hold_acknowledge))(0)
159	sync
160
161	li	r26,0
162#ifdef CONFIG_PPC_BOOK3E
163	tovirt(r26,r26)
164#endif
165	/* All secondary cpus wait here until told to start. */
166100:	ld	r12,(ABS_ADDR(__secondary_hold_spinloop))(r26)
167	cmpdi	0,r12,0
168	beq	100b
169
170#if defined(CONFIG_SMP) || defined(CONFIG_KEXEC_CORE)
171#ifdef CONFIG_PPC_BOOK3E
172	tovirt(r12,r12)
173#endif
174	mtctr	r12
175	mr	r3,r24
176	/*
177	 * it may be the case that other platforms have r4 right to
178	 * begin with, this gives us some safety in case it is not
179	 */
180#ifdef CONFIG_PPC_BOOK3E
181	mr	r4,r25
182#else
183	li	r4,0
184#endif
185	/* Make sure that patched code is visible */
186	isync
187	bctr
188#else
189	BUG_OPCODE
190#endif
191CLOSE_FIXED_SECTION(first_256B)
192
193/* This value is used to mark exception frames on the stack. */
194	.section ".toc","aw"
195exception_marker:
196	.tc	ID_72656773_68657265[TC],0x7265677368657265
197	.previous
198
199/*
200 * On server, we include the exception vectors code here as it
201 * relies on absolute addressing which is only possible within
202 * this compilation unit
203 */
204#ifdef CONFIG_PPC_BOOK3S
205#include "exceptions-64s.S"
206#else
207OPEN_TEXT_SECTION(0x100)
208#endif
209
210USE_TEXT_SECTION()
211
212#ifdef CONFIG_PPC_BOOK3E
213/*
214 * The booting_thread_hwid holds the thread id we want to boot in cpu
215 * hotplug case. It is set by cpu hotplug code, and is invalid by default.
216 * The thread id is the same as the initial value of SPRN_PIR[THREAD_ID]
217 * bit field.
218 */
219	.globl	booting_thread_hwid
220booting_thread_hwid:
221	.long  INVALID_THREAD_HWID
222	.align 3
223/*
224 * start a thread in the same core
225 * input parameters:
226 * r3 = the thread physical id
227 * r4 = the entry point where thread starts
228 */
229_GLOBAL(book3e_start_thread)
230	LOAD_REG_IMMEDIATE(r5, MSR_KERNEL)
231	cmpwi	r3, 0
232	beq	10f
233	cmpwi	r3, 1
234	beq	11f
235	/* If the thread id is invalid, just exit. */
236	b	13f
23710:
238	MTTMR(TMRN_IMSR0, 5)
239	MTTMR(TMRN_INIA0, 4)
240	b	12f
24111:
242	MTTMR(TMRN_IMSR1, 5)
243	MTTMR(TMRN_INIA1, 4)
24412:
245	isync
246	li	r6, 1
247	sld	r6, r6, r3
248	mtspr	SPRN_TENS, r6
24913:
250	blr
251
252/*
253 * stop a thread in the same core
254 * input parameter:
255 * r3 = the thread physical id
256 */
257_GLOBAL(book3e_stop_thread)
258	cmpwi	r3, 0
259	beq	10f
260	cmpwi	r3, 1
261	beq	10f
262	/* If the thread id is invalid, just exit. */
263	b	13f
26410:
265	li	r4, 1
266	sld	r4, r4, r3
267	mtspr	SPRN_TENC, r4
26813:
269	blr
270
271_GLOBAL(fsl_secondary_thread_init)
272	mfspr	r4,SPRN_BUCSR
273
274	/* Enable branch prediction */
275	lis     r3,BUCSR_INIT@h
276	ori     r3,r3,BUCSR_INIT@l
277	mtspr   SPRN_BUCSR,r3
278	isync
279
280	/*
281	 * Fix PIR to match the linear numbering in the device tree.
282	 *
283	 * On e6500, the reset value of PIR uses the low three bits for
284	 * the thread within a core, and the upper bits for the core
285	 * number.  There are two threads per core, so shift everything
286	 * but the low bit right by two bits so that the cpu numbering is
287	 * continuous.
288	 *
289	 * If the old value of BUCSR is non-zero, this thread has run
290	 * before.  Thus, we assume we are coming from kexec or a similar
291	 * scenario, and PIR is already set to the correct value.  This
292	 * is a bit of a hack, but there are limited opportunities for
293	 * getting information into the thread and the alternatives
294	 * seemed like they'd be overkill.  We can't tell just by looking
295	 * at the old PIR value which state it's in, since the same value
296	 * could be valid for one thread out of reset and for a different
297	 * thread in Linux.
298	 */
299
300	mfspr	r3, SPRN_PIR
301	cmpwi	r4,0
302	bne	1f
303	rlwimi	r3, r3, 30, 2, 30
304	mtspr	SPRN_PIR, r3
3051:
306#endif
307
308_GLOBAL(generic_secondary_thread_init)
309	mr	r24,r3
310
311	/* turn on 64-bit mode */
312	bl	enable_64b_mode
313
314	/* get a valid TOC pointer, wherever we're mapped at */
315	bl	relative_toc
316	tovirt(r2,r2)
317
318#ifdef CONFIG_PPC_BOOK3E
319	/* Book3E initialization */
320	mr	r3,r24
321	bl	book3e_secondary_thread_init
322#endif
323	b	generic_secondary_common_init
324
325/*
326 * On pSeries and most other platforms, secondary processors spin
327 * in the following code.
328 * At entry, r3 = this processor's number (physical cpu id)
329 *
330 * On Book3E, r4 = 1 to indicate that the initial TLB entry for
331 * this core already exists (setup via some other mechanism such
332 * as SCOM before entry).
333 */
334_GLOBAL(generic_secondary_smp_init)
335	FIXUP_ENDIAN
336	mr	r24,r3
337	mr	r25,r4
338
339	/* turn on 64-bit mode */
340	bl	enable_64b_mode
341
342	/* get a valid TOC pointer, wherever we're mapped at */
343	bl	relative_toc
344	tovirt(r2,r2)
345
346#ifdef CONFIG_PPC_BOOK3E
347	/* Book3E initialization */
348	mr	r3,r24
349	mr	r4,r25
350	bl	book3e_secondary_core_init
351
352/*
353 * After common core init has finished, check if the current thread is the
354 * one we wanted to boot. If not, start the specified thread and stop the
355 * current thread.
356 */
357	LOAD_REG_ADDR(r4, booting_thread_hwid)
358	lwz     r3, 0(r4)
359	li	r5, INVALID_THREAD_HWID
360	cmpw	r3, r5
361	beq	20f
362
363	/*
364	 * The value of booting_thread_hwid has been stored in r3,
365	 * so make it invalid.
366	 */
367	stw	r5, 0(r4)
368
369	/*
370	 * Get the current thread id and check if it is the one we wanted.
371	 * If not, start the one specified in booting_thread_hwid and stop
372	 * the current thread.
373	 */
374	mfspr	r8, SPRN_TIR
375	cmpw	r3, r8
376	beq	20f
377
378	/* start the specified thread */
379	LOAD_REG_ADDR(r5, fsl_secondary_thread_init)
380	ld	r4, 0(r5)
381	bl	book3e_start_thread
382
383	/* stop the current thread */
384	mr	r3, r8
385	bl	book3e_stop_thread
38610:
387	b	10b
38820:
389#endif
390
391generic_secondary_common_init:
392	/* Set up a paca value for this processor. Since we have the
393	 * physical cpu id in r24, we need to search the pacas to find
394	 * which logical id maps to our physical one.
395	 */
396#ifndef CONFIG_SMP
397	b	kexec_wait		/* wait for next kernel if !SMP	 */
398#else
399	LOAD_REG_ADDR(r8, paca_ptrs)	/* Load paca_ptrs pointe	 */
400	ld	r8,0(r8)		/* Get base vaddr of array	 */
401	LOAD_REG_ADDR(r7, nr_cpu_ids)	/* Load nr_cpu_ids address       */
402	lwz	r7,0(r7)		/* also the max paca allocated 	 */
403	li	r5,0			/* logical cpu id                */
4041:
405	sldi	r9,r5,3			/* get paca_ptrs[] index from cpu id */
406	ldx	r13,r9,r8		/* r13 = paca_ptrs[cpu id]       */
407	lhz	r6,PACAHWCPUID(r13)	/* Load HW procid from paca      */
408	cmpw	r6,r24			/* Compare to our id             */
409	beq	2f
410	addi	r5,r5,1
411	cmpw	r5,r7			/* Check if more pacas exist     */
412	blt	1b
413
414	mr	r3,r24			/* not found, copy phys to r3	 */
415	b	kexec_wait		/* next kernel might do better	 */
416
4172:	SET_PACA(r13)
418#ifdef CONFIG_PPC_BOOK3E
419	addi	r12,r13,PACA_EXTLB	/* and TLB exc frame in another  */
420	mtspr	SPRN_SPRG_TLB_EXFRAME,r12
421#endif
422
423	/* From now on, r24 is expected to be logical cpuid */
424	mr	r24,r5
425
426	/* See if we need to call a cpu state restore handler */
427	LOAD_REG_ADDR(r23, cur_cpu_spec)
428	ld	r23,0(r23)
429	ld	r12,CPU_SPEC_RESTORE(r23)
430	cmpdi	0,r12,0
431	beq	3f
432#ifdef PPC64_ELF_ABI_v1
433	ld	r12,0(r12)
434#endif
435	mtctr	r12
436	bctrl
437
4383:	LOAD_REG_ADDR(r3, spinning_secondaries) /* Decrement spinning_secondaries */
439	lwarx	r4,0,r3
440	subi	r4,r4,1
441	stwcx.	r4,0,r3
442	bne	3b
443	isync
444
4454:	HMT_LOW
446	lbz	r23,PACAPROCSTART(r13)	/* Test if this processor should */
447					/* start.			 */
448	cmpwi	0,r23,0
449	beq	4b			/* Loop until told to go	 */
450
451	sync				/* order paca.run and cur_cpu_spec */
452	isync				/* In case code patching happened */
453
454	/* Create a temp kernel stack for use before relocation is on.	*/
455	ld	r1,PACAEMERGSP(r13)
456	subi	r1,r1,STACK_FRAME_OVERHEAD
457
458	b	__secondary_start
459#endif /* SMP */
460
461/*
462 * Turn the MMU off.
463 * Assumes we're mapped EA == RA if the MMU is on.
464 */
465#ifdef CONFIG_PPC_BOOK3S
466__mmu_off:
467	mfmsr	r3
468	andi.	r0,r3,MSR_IR|MSR_DR
469	beqlr
470	mflr	r4
471	andc	r3,r3,r0
472	mtspr	SPRN_SRR0,r4
473	mtspr	SPRN_SRR1,r3
474	sync
475	rfid
476	b	.	/* prevent speculative execution */
477#endif
478
479
480/*
481 * Here is our main kernel entry point. We support currently 2 kind of entries
482 * depending on the value of r5.
483 *
484 *   r5 != NULL -> OF entry, we go to prom_init, "legacy" parameter content
485 *                 in r3...r7
486 *
487 *   r5 == NULL -> kexec style entry. r3 is a physical pointer to the
488 *                 DT block, r4 is a physical pointer to the kernel itself
489 *
490 */
491__start_initialization_multiplatform:
492	/* Make sure we are running in 64 bits mode */
493	bl	enable_64b_mode
494
495	/* Get TOC pointer (current runtime address) */
496	bl	relative_toc
497
498	/* find out where we are now */
499	bcl	20,31,$+4
5000:	mflr	r26			/* r26 = runtime addr here */
501	addis	r26,r26,(_stext - 0b)@ha
502	addi	r26,r26,(_stext - 0b)@l	/* current runtime base addr */
503
504	/*
505	 * Are we booted from a PROM Of-type client-interface ?
506	 */
507	cmpldi	cr0,r5,0
508	beq	1f
509	b	__boot_from_prom		/* yes -> prom */
5101:
511	/* Save parameters */
512	mr	r31,r3
513	mr	r30,r4
514#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
515	/* Save OPAL entry */
516	mr	r28,r8
517	mr	r29,r9
518#endif
519
520#ifdef CONFIG_PPC_BOOK3E
521	bl	start_initialization_book3e
522	b	__after_prom_start
523#else
524	/* Setup some critical 970 SPRs before switching MMU off */
525	mfspr	r0,SPRN_PVR
526	srwi	r0,r0,16
527	cmpwi	r0,0x39		/* 970 */
528	beq	1f
529	cmpwi	r0,0x3c		/* 970FX */
530	beq	1f
531	cmpwi	r0,0x44		/* 970MP */
532	beq	1f
533	cmpwi	r0,0x45		/* 970GX */
534	bne	2f
5351:	bl	__cpu_preinit_ppc970
5362:
537
538	/* Switch off MMU if not already off */
539	bl	__mmu_off
540	b	__after_prom_start
541#endif /* CONFIG_PPC_BOOK3E */
542
543__boot_from_prom:
544#ifdef CONFIG_PPC_OF_BOOT_TRAMPOLINE
545	/* Save parameters */
546	mr	r31,r3
547	mr	r30,r4
548	mr	r29,r5
549	mr	r28,r6
550	mr	r27,r7
551
552	/*
553	 * Align the stack to 16-byte boundary
554	 * Depending on the size and layout of the ELF sections in the initial
555	 * boot binary, the stack pointer may be unaligned on PowerMac
556	 */
557	rldicr	r1,r1,0,59
558
559#ifdef CONFIG_RELOCATABLE
560	/* Relocate code for where we are now */
561	mr	r3,r26
562	bl	relocate
563#endif
564
565	/* Restore parameters */
566	mr	r3,r31
567	mr	r4,r30
568	mr	r5,r29
569	mr	r6,r28
570	mr	r7,r27
571
572	/* Do all of the interaction with OF client interface */
573	mr	r8,r26
574	bl	prom_init
575#endif /* #CONFIG_PPC_OF_BOOT_TRAMPOLINE */
576
577	/* We never return. We also hit that trap if trying to boot
578	 * from OF while CONFIG_PPC_OF_BOOT_TRAMPOLINE isn't selected */
579	trap
580
581__after_prom_start:
582#ifdef CONFIG_RELOCATABLE
583	/* process relocations for the final address of the kernel */
584	lis	r25,PAGE_OFFSET@highest	/* compute virtual base of kernel */
585	sldi	r25,r25,32
586#if defined(CONFIG_PPC_BOOK3E)
587	tovirt(r26,r26)		/* on booke, we already run at PAGE_OFFSET */
588#endif
589	lwz	r7,(FIXED_SYMBOL_ABS_ADDR(__run_at_load))(r26)
590#if defined(CONFIG_PPC_BOOK3E)
591	tophys(r26,r26)
592#endif
593	cmplwi	cr0,r7,1	/* flagged to stay where we are ? */
594	bne	1f
595	add	r25,r25,r26
5961:	mr	r3,r25
597	bl	relocate
598#if defined(CONFIG_PPC_BOOK3E)
599	/* IVPR needs to be set after relocation. */
600	bl	init_core_book3e
601#endif
602#endif
603
604/*
605 * We need to run with _stext at physical address PHYSICAL_START.
606 * This will leave some code in the first 256B of
607 * real memory, which are reserved for software use.
608 *
609 * Note: This process overwrites the OF exception vectors.
610 */
611	li	r3,0			/* target addr */
612#ifdef CONFIG_PPC_BOOK3E
613	tovirt(r3,r3)		/* on booke, we already run at PAGE_OFFSET */
614#endif
615	mr.	r4,r26			/* In some cases the loader may  */
616#if defined(CONFIG_PPC_BOOK3E)
617	tovirt(r4,r4)
618#endif
619	beq	9f			/* have already put us at zero */
620	li	r6,0x100		/* Start offset, the first 0x100 */
621					/* bytes were copied earlier.	 */
622
623#ifdef CONFIG_RELOCATABLE
624/*
625 * Check if the kernel has to be running as relocatable kernel based on the
626 * variable __run_at_load, if it is set the kernel is treated as relocatable
627 * kernel, otherwise it will be moved to PHYSICAL_START
628 */
629#if defined(CONFIG_PPC_BOOK3E)
630	tovirt(r26,r26)		/* on booke, we already run at PAGE_OFFSET */
631#endif
632	lwz	r7,(FIXED_SYMBOL_ABS_ADDR(__run_at_load))(r26)
633	cmplwi	cr0,r7,1
634	bne	3f
635
636#ifdef CONFIG_PPC_BOOK3E
637	LOAD_REG_ADDR(r5, __end_interrupts)
638	LOAD_REG_ADDR(r11, _stext)
639	sub	r5,r5,r11
640#else
641	/* just copy interrupts */
642	LOAD_REG_IMMEDIATE(r5, FIXED_SYMBOL_ABS_ADDR(__end_interrupts))
643#endif
644	b	5f
6453:
646#endif
647	/* # bytes of memory to copy */
648	lis	r5,(ABS_ADDR(copy_to_here))@ha
649	addi	r5,r5,(ABS_ADDR(copy_to_here))@l
650
651	bl	copy_and_flush		/* copy the first n bytes	 */
652					/* this includes the code being	 */
653					/* executed here.		 */
654	/* Jump to the copy of this code that we just made */
655	addis	r8,r3,(ABS_ADDR(4f))@ha
656	addi	r12,r8,(ABS_ADDR(4f))@l
657	mtctr	r12
658	bctr
659
660.balign 8
661p_end: .8byte _end - copy_to_here
662
6634:
664	/*
665	 * Now copy the rest of the kernel up to _end, add
666	 * _end - copy_to_here to the copy limit and run again.
667	 */
668	addis   r8,r26,(ABS_ADDR(p_end))@ha
669	ld      r8,(ABS_ADDR(p_end))@l(r8)
670	add	r5,r5,r8
6715:	bl	copy_and_flush		/* copy the rest */
672
6739:	b	start_here_multiplatform
674
675/*
676 * Copy routine used to copy the kernel to start at physical address 0
677 * and flush and invalidate the caches as needed.
678 * r3 = dest addr, r4 = source addr, r5 = copy limit, r6 = start offset
679 * on exit, r3, r4, r5 are unchanged, r6 is updated to be >= r5.
680 *
681 * Note: this routine *only* clobbers r0, r6 and lr
682 */
683_GLOBAL(copy_and_flush)
684	addi	r5,r5,-8
685	addi	r6,r6,-8
6864:	li	r0,8			/* Use the smallest common	*/
687					/* denominator cache line	*/
688					/* size.  This results in	*/
689					/* extra cache line flushes	*/
690					/* but operation is correct.	*/
691					/* Can't get cache line size	*/
692					/* from NACA as it is being	*/
693					/* moved too.			*/
694
695	mtctr	r0			/* put # words/line in ctr	*/
6963:	addi	r6,r6,8			/* copy a cache line		*/
697	ldx	r0,r6,r4
698	stdx	r0,r6,r3
699	bdnz	3b
700	dcbst	r6,r3			/* write it to memory		*/
701	sync
702	icbi	r6,r3			/* flush the icache line	*/
703	cmpld	0,r6,r5
704	blt	4b
705	sync
706	addi	r5,r5,8
707	addi	r6,r6,8
708	isync
709	blr
710
711.align 8
712copy_to_here:
713
714#ifdef CONFIG_SMP
715#ifdef CONFIG_PPC_PMAC
716/*
717 * On PowerMac, secondary processors starts from the reset vector, which
718 * is temporarily turned into a call to one of the functions below.
719 */
720	.section ".text";
721	.align 2 ;
722
723	.globl	__secondary_start_pmac_0
724__secondary_start_pmac_0:
725	/* NB the entries for cpus 0, 1, 2 must each occupy 8 bytes. */
726	li	r24,0
727	b	1f
728	li	r24,1
729	b	1f
730	li	r24,2
731	b	1f
732	li	r24,3
7331:
734
735_GLOBAL(pmac_secondary_start)
736	/* turn on 64-bit mode */
737	bl	enable_64b_mode
738
739	li	r0,0
740	mfspr	r3,SPRN_HID4
741	rldimi	r3,r0,40,23	/* clear bit 23 (rm_ci) */
742	sync
743	mtspr	SPRN_HID4,r3
744	isync
745	sync
746	slbia
747
748	/* get TOC pointer (real address) */
749	bl	relative_toc
750	tovirt(r2,r2)
751
752	/* Copy some CPU settings from CPU 0 */
753	bl	__restore_cpu_ppc970
754
755	/* pSeries do that early though I don't think we really need it */
756	mfmsr	r3
757	ori	r3,r3,MSR_RI
758	mtmsrd	r3			/* RI on */
759
760	/* Set up a paca value for this processor. */
761	LOAD_REG_ADDR(r4,paca_ptrs)	/* Load paca pointer		*/
762	ld	r4,0(r4)		/* Get base vaddr of paca_ptrs array */
763	sldi	r5,r24,3		/* get paca_ptrs[] index from cpu id */
764	ldx	r13,r5,r4		/* r13 = paca_ptrs[cpu id]       */
765	SET_PACA(r13)			/* Save vaddr of paca in an SPRG*/
766
767	/* Mark interrupts soft and hard disabled (they might be enabled
768	 * in the PACA when doing hotplug)
769	 */
770	li	r0,IRQS_DISABLED
771	stb	r0,PACAIRQSOFTMASK(r13)
772	li	r0,PACA_IRQ_HARD_DIS
773	stb	r0,PACAIRQHAPPENED(r13)
774
775	/* Create a temp kernel stack for use before relocation is on.	*/
776	ld	r1,PACAEMERGSP(r13)
777	subi	r1,r1,STACK_FRAME_OVERHEAD
778
779	b	__secondary_start
780
781#endif /* CONFIG_PPC_PMAC */
782
783/*
784 * This function is called after the master CPU has released the
785 * secondary processors.  The execution environment is relocation off.
786 * The paca for this processor has the following fields initialized at
787 * this point:
788 *   1. Processor number
789 *   2. Segment table pointer (virtual address)
790 * On entry the following are set:
791 *   r1	       = stack pointer (real addr of temp stack)
792 *   r24       = cpu# (in Linux terms)
793 *   r13       = paca virtual address
794 *   SPRG_PACA = paca virtual address
795 */
796	.section ".text";
797	.align 2 ;
798
799	.globl	__secondary_start
800__secondary_start:
801	/* Set thread priority to MEDIUM */
802	HMT_MEDIUM
803
804	/* Initialize the kernel stack */
805	LOAD_REG_ADDR(r3, current_set)
806	sldi	r28,r24,3		/* get current_set[cpu#]	 */
807	ldx	r14,r3,r28
808	addi	r14,r14,THREAD_SIZE-STACK_FRAME_OVERHEAD
809	std	r14,PACAKSAVE(r13)
810
811	/* Do early setup for that CPU (SLB and hash table pointer) */
812	bl	early_setup_secondary
813
814	/*
815	 * setup the new stack pointer, but *don't* use this until
816	 * translation is on.
817	 */
818	mr	r1, r14
819
820	/* Clear backchain so we get nice backtraces */
821	li	r7,0
822	mtlr	r7
823
824	/* Mark interrupts soft and hard disabled (they might be enabled
825	 * in the PACA when doing hotplug)
826	 */
827	li	r7,IRQS_DISABLED
828	stb	r7,PACAIRQSOFTMASK(r13)
829	li	r0,PACA_IRQ_HARD_DIS
830	stb	r0,PACAIRQHAPPENED(r13)
831
832	/* enable MMU and jump to start_secondary */
833	LOAD_REG_ADDR(r3, start_secondary_prolog)
834	LOAD_REG_IMMEDIATE(r4, MSR_KERNEL)
835
836	mtspr	SPRN_SRR0,r3
837	mtspr	SPRN_SRR1,r4
838	RFI
839	b	.	/* prevent speculative execution */
840
841/*
842 * Running with relocation on at this point.  All we want to do is
843 * zero the stack back-chain pointer and get the TOC virtual address
844 * before going into C code.
845 */
846start_secondary_prolog:
847	ld	r2,PACATOC(r13)
848	li	r3,0
849	std	r3,0(r1)		/* Zero the stack frame pointer	*/
850	bl	start_secondary
851	b	.
852/*
853 * Reset stack pointer and call start_secondary
854 * to continue with online operation when woken up
855 * from cede in cpu offline.
856 */
857_GLOBAL(start_secondary_resume)
858	ld	r1,PACAKSAVE(r13)	/* Reload kernel stack pointer */
859	li	r3,0
860	std	r3,0(r1)		/* Zero the stack frame pointer	*/
861	bl	start_secondary
862	b	.
863#endif
864
865/*
866 * This subroutine clobbers r11 and r12
867 */
868enable_64b_mode:
869	mfmsr	r11			/* grab the current MSR */
870#ifdef CONFIG_PPC_BOOK3E
871	oris	r11,r11,0x8000		/* CM bit set, we'll set ICM later */
872	mtmsr	r11
873#else /* CONFIG_PPC_BOOK3E */
874	li	r12,(MSR_64BIT | MSR_ISF)@highest
875	sldi	r12,r12,48
876	or	r11,r11,r12
877	mtmsrd	r11
878	isync
879#endif
880	blr
881
882/*
883 * This puts the TOC pointer into r2, offset by 0x8000 (as expected
884 * by the toolchain).  It computes the correct value for wherever we
885 * are running at the moment, using position-independent code.
886 *
887 * Note: The compiler constructs pointers using offsets from the
888 * TOC in -mcmodel=medium mode. After we relocate to 0 but before
889 * the MMU is on we need our TOC to be a virtual address otherwise
890 * these pointers will be real addresses which may get stored and
891 * accessed later with the MMU on. We use tovirt() at the call
892 * sites to handle this.
893 */
894_GLOBAL(relative_toc)
895	mflr	r0
896	bcl	20,31,$+4
8970:	mflr	r11
898	ld	r2,(p_toc - 0b)(r11)
899	add	r2,r2,r11
900	mtlr	r0
901	blr
902
903.balign 8
904p_toc:	.8byte	__toc_start + 0x8000 - 0b
905
906/*
907 * This is where the main kernel code starts.
908 */
909start_here_multiplatform:
910	/* set up the TOC */
911	bl      relative_toc
912	tovirt(r2,r2)
913
914	/* Clear out the BSS. It may have been done in prom_init,
915	 * already but that's irrelevant since prom_init will soon
916	 * be detached from the kernel completely. Besides, we need
917	 * to clear it now for kexec-style entry.
918	 */
919	LOAD_REG_ADDR(r11,__bss_stop)
920	LOAD_REG_ADDR(r8,__bss_start)
921	sub	r11,r11,r8		/* bss size			*/
922	addi	r11,r11,7		/* round up to an even double word */
923	srdi.	r11,r11,3		/* shift right by 3		*/
924	beq	4f
925	addi	r8,r8,-8
926	li	r0,0
927	mtctr	r11			/* zero this many doublewords	*/
9283:	stdu	r0,8(r8)
929	bdnz	3b
9304:
931
932#ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
933	/* Setup OPAL entry */
934	LOAD_REG_ADDR(r11, opal)
935	std	r28,0(r11);
936	std	r29,8(r11);
937#endif
938
939#ifndef CONFIG_PPC_BOOK3E
940	mfmsr	r6
941	ori	r6,r6,MSR_RI
942	mtmsrd	r6			/* RI on */
943#endif
944
945#ifdef CONFIG_RELOCATABLE
946	/* Save the physical address we're running at in kernstart_addr */
947	LOAD_REG_ADDR(r4, kernstart_addr)
948	clrldi	r0,r25,2
949	std	r0,0(r4)
950#endif
951
952	/* The following gets the stack set up with the regs */
953	/* pointing to the real addr of the kernel stack.  This is   */
954	/* all done to support the C function call below which sets  */
955	/* up the htab.  This is done because we have relocated the  */
956	/* kernel but are still running in real mode. */
957
958	LOAD_REG_ADDR(r3,init_thread_union)
959
960	/* set up a stack pointer */
961	LOAD_REG_IMMEDIATE(r1,THREAD_SIZE)
962	add	r1,r3,r1
963	li	r0,0
964	stdu	r0,-STACK_FRAME_OVERHEAD(r1)
965
966	/*
967	 * Do very early kernel initializations, including initial hash table
968	 * and SLB setup before we turn on relocation.
969	 */
970
971	/* Restore parameters passed from prom_init/kexec */
972	mr	r3,r31
973	bl	early_setup		/* also sets r13 and SPRG_PACA */
974
975	LOAD_REG_ADDR(r3, start_here_common)
976	ld	r4,PACAKMSR(r13)
977	mtspr	SPRN_SRR0,r3
978	mtspr	SPRN_SRR1,r4
979	RFI
980	b	.	/* prevent speculative execution */
981
982	/* This is where all platforms converge execution */
983
984start_here_common:
985	/* relocation is on at this point */
986	std	r1,PACAKSAVE(r13)
987
988	/* Load the TOC (virtual address) */
989	ld	r2,PACATOC(r13)
990
991	/* Mark interrupts soft and hard disabled (they might be enabled
992	 * in the PACA when doing hotplug)
993	 */
994	li	r0,IRQS_DISABLED
995	stb	r0,PACAIRQSOFTMASK(r13)
996	li	r0,PACA_IRQ_HARD_DIS
997	stb	r0,PACAIRQHAPPENED(r13)
998
999	/* Generic kernel entry */
1000	bl	start_kernel
1001
1002	/* Not reached */
1003	BUG_OPCODE
1004
1005/*
1006 * We put a few things here that have to be page-aligned.
1007 * This stuff goes at the beginning of the bss, which is page-aligned.
1008 */
1009	.section ".bss"
1010/*
1011 * pgd dir should be aligned to PGD_TABLE_SIZE which is 64K.
1012 * We will need to find a better way to fix this
1013 */
1014	.align	16
1015
1016	.globl	swapper_pg_dir
1017swapper_pg_dir:
1018	.space	PGD_TABLE_SIZE
1019
1020	.globl	empty_zero_page
1021empty_zero_page:
1022	.space	PAGE_SIZE
1023EXPORT_SYMBOL(empty_zero_page)
1024