xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision 6d99a79c)
1/*
2 * This file contains miscellaneous low-level functions.
3 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
4 *
5 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
6 * and Paul Mackerras.
7 *
8 * kexec bits:
9 * Copyright (C) 2002-2003 Eric Biederman  <ebiederm@xmission.com>
10 * GameCube/ppc32 port Copyright (C) 2004 Albert Herranz
11 * PPC44x port. Copyright (C) 2011,  IBM Corporation
12 * 		Author: Suzuki Poulose <suzuki@in.ibm.com>
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 *
19 */
20
21#include <linux/sys.h>
22#include <asm/unistd.h>
23#include <asm/errno.h>
24#include <asm/reg.h>
25#include <asm/page.h>
26#include <asm/cache.h>
27#include <asm/cputable.h>
28#include <asm/mmu.h>
29#include <asm/ppc_asm.h>
30#include <asm/thread_info.h>
31#include <asm/asm-offsets.h>
32#include <asm/processor.h>
33#include <asm/kexec.h>
34#include <asm/bug.h>
35#include <asm/ptrace.h>
36#include <asm/export.h>
37#include <asm/feature-fixups.h>
38
39	.text
40
41/*
42 * We store the saved ksp_limit in the unused part
43 * of the STACK_FRAME_OVERHEAD
44 */
45_GLOBAL(call_do_softirq)
46	mflr	r0
47	stw	r0,4(r1)
48	lwz	r10,THREAD+KSP_LIMIT(r2)
49	addi	r11,r3,THREAD_INFO_GAP
50	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
51	mr	r1,r3
52	stw	r10,8(r1)
53	stw	r11,THREAD+KSP_LIMIT(r2)
54	bl	__do_softirq
55	lwz	r10,8(r1)
56	lwz	r1,0(r1)
57	lwz	r0,4(r1)
58	stw	r10,THREAD+KSP_LIMIT(r2)
59	mtlr	r0
60	blr
61
62/*
63 * void call_do_irq(struct pt_regs *regs, struct thread_info *irqtp);
64 */
65_GLOBAL(call_do_irq)
66	mflr	r0
67	stw	r0,4(r1)
68	lwz	r10,THREAD+KSP_LIMIT(r2)
69	addi	r11,r4,THREAD_INFO_GAP
70	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
71	mr	r1,r4
72	stw	r10,8(r1)
73	stw	r11,THREAD+KSP_LIMIT(r2)
74	bl	__do_irq
75	lwz	r10,8(r1)
76	lwz	r1,0(r1)
77	lwz	r0,4(r1)
78	stw	r10,THREAD+KSP_LIMIT(r2)
79	mtlr	r0
80	blr
81
82/*
83 * This returns the high 64 bits of the product of two 64-bit numbers.
84 */
85_GLOBAL(mulhdu)
86	cmpwi	r6,0
87	cmpwi	cr1,r3,0
88	mr	r10,r4
89	mulhwu	r4,r4,r5
90	beq	1f
91	mulhwu	r0,r10,r6
92	mullw	r7,r10,r5
93	addc	r7,r0,r7
94	addze	r4,r4
951:	beqlr	cr1		/* all done if high part of A is 0 */
96	mullw	r9,r3,r5
97	mulhwu	r10,r3,r5
98	beq	2f
99	mullw	r0,r3,r6
100	mulhwu	r8,r3,r6
101	addc	r7,r0,r7
102	adde	r4,r4,r8
103	addze	r10,r10
1042:	addc	r4,r4,r9
105	addze	r3,r10
106	blr
107
108/*
109 * reloc_got2 runs through the .got2 section adding an offset
110 * to each entry.
111 */
112_GLOBAL(reloc_got2)
113	mflr	r11
114	lis	r7,__got2_start@ha
115	addi	r7,r7,__got2_start@l
116	lis	r8,__got2_end@ha
117	addi	r8,r8,__got2_end@l
118	subf	r8,r7,r8
119	srwi.	r8,r8,2
120	beqlr
121	mtctr	r8
122	bl	1f
1231:	mflr	r0
124	lis	r4,1b@ha
125	addi	r4,r4,1b@l
126	subf	r0,r4,r0
127	add	r7,r0,r7
1282:	lwz	r0,0(r7)
129	add	r0,r0,r3
130	stw	r0,0(r7)
131	addi	r7,r7,4
132	bdnz	2b
133	mtlr	r11
134	blr
135
136/*
137 * call_setup_cpu - call the setup_cpu function for this cpu
138 * r3 = data offset, r24 = cpu number
139 *
140 * Setup function is called with:
141 *   r3 = data offset
142 *   r4 = ptr to CPU spec (relocated)
143 */
144_GLOBAL(call_setup_cpu)
145	addis	r4,r3,cur_cpu_spec@ha
146	addi	r4,r4,cur_cpu_spec@l
147	lwz	r4,0(r4)
148	add	r4,r4,r3
149	lwz	r5,CPU_SPEC_SETUP(r4)
150	cmpwi	0,r5,0
151	add	r5,r5,r3
152	beqlr
153	mtctr	r5
154	bctr
155
156#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
157
158/* This gets called by via-pmu.c to switch the PLL selection
159 * on 750fx CPU. This function should really be moved to some
160 * other place (as most of the cpufreq code in via-pmu
161 */
162_GLOBAL(low_choose_750fx_pll)
163	/* Clear MSR:EE */
164	mfmsr	r7
165	rlwinm	r0,r7,0,17,15
166	mtmsr	r0
167
168	/* If switching to PLL1, disable HID0:BTIC */
169	cmplwi	cr0,r3,0
170	beq	1f
171	mfspr	r5,SPRN_HID0
172	rlwinm	r5,r5,0,27,25
173	sync
174	mtspr	SPRN_HID0,r5
175	isync
176	sync
177
1781:
179	/* Calc new HID1 value */
180	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
181	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
182	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
183	or	r4,r4,r5
184	mtspr	SPRN_HID1,r4
185
186	/* Store new HID1 image */
187	CURRENT_THREAD_INFO(r6, r1)
188	lwz	r6,TI_CPU(r6)
189	slwi	r6,r6,2
190	addis	r6,r6,nap_save_hid1@ha
191	stw	r4,nap_save_hid1@l(r6)
192
193	/* If switching to PLL0, enable HID0:BTIC */
194	cmplwi	cr0,r3,0
195	bne	1f
196	mfspr	r5,SPRN_HID0
197	ori	r5,r5,HID0_BTIC
198	sync
199	mtspr	SPRN_HID0,r5
200	isync
201	sync
202
2031:
204	/* Return */
205	mtmsr	r7
206	blr
207
208_GLOBAL(low_choose_7447a_dfs)
209	/* Clear MSR:EE */
210	mfmsr	r7
211	rlwinm	r0,r7,0,17,15
212	mtmsr	r0
213
214	/* Calc new HID1 value */
215	mfspr	r4,SPRN_HID1
216	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
217	sync
218	mtspr	SPRN_HID1,r4
219	sync
220	isync
221
222	/* Return */
223	mtmsr	r7
224	blr
225
226#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
227
228/*
229 * complement mask on the msr then "or" some values on.
230 *     _nmask_and_or_msr(nmask, value_to_or)
231 */
232_GLOBAL(_nmask_and_or_msr)
233	mfmsr	r0		/* Get current msr */
234	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
235	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
236	SYNC			/* Some chip revs have problems here... */
237	mtmsr	r0		/* Update machine state */
238	isync
239	blr			/* Done */
240
241#ifdef CONFIG_40x
242
243/*
244 * Do an IO access in real mode
245 */
246_GLOBAL(real_readb)
247	mfmsr	r7
248	rlwinm	r0,r7,0,~MSR_DR
249	sync
250	mtmsr	r0
251	sync
252	isync
253	lbz	r3,0(r3)
254	sync
255	mtmsr	r7
256	sync
257	isync
258	blr
259
260	/*
261 * Do an IO access in real mode
262 */
263_GLOBAL(real_writeb)
264	mfmsr	r7
265	rlwinm	r0,r7,0,~MSR_DR
266	sync
267	mtmsr	r0
268	sync
269	isync
270	stb	r3,0(r4)
271	sync
272	mtmsr	r7
273	sync
274	isync
275	blr
276
277#endif /* CONFIG_40x */
278
279
280/*
281 * Flush instruction cache.
282 * This is a no-op on the 601.
283 */
284#ifndef CONFIG_PPC_8xx
285_GLOBAL(flush_instruction_cache)
286#if defined(CONFIG_4xx)
287#ifdef CONFIG_403GCX
288	li      r3, 512
289	mtctr   r3
290	lis     r4, KERNELBASE@h
2911:	iccci   0, r4
292	addi    r4, r4, 16
293	bdnz    1b
294#else
295	lis	r3, KERNELBASE@h
296	iccci	0,r3
297#endif
298#elif defined(CONFIG_FSL_BOOKE)
299BEGIN_FTR_SECTION
300	mfspr   r3,SPRN_L1CSR0
301	ori     r3,r3,L1CSR0_CFI|L1CSR0_CLFC
302	/* msync; isync recommended here */
303	mtspr   SPRN_L1CSR0,r3
304	isync
305	blr
306END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
307	mfspr	r3,SPRN_L1CSR1
308	ori	r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
309	mtspr	SPRN_L1CSR1,r3
310#else
311	mfspr	r3,SPRN_PVR
312	rlwinm	r3,r3,16,16,31
313	cmpwi	0,r3,1
314	beqlr			/* for 601, do nothing */
315	/* 603/604 processor - use invalidate-all bit in HID0 */
316	mfspr	r3,SPRN_HID0
317	ori	r3,r3,HID0_ICFI
318	mtspr	SPRN_HID0,r3
319#endif /* CONFIG_4xx */
320	isync
321	blr
322EXPORT_SYMBOL(flush_instruction_cache)
323#endif /* CONFIG_PPC_8xx */
324
325/*
326 * Write any modified data cache blocks out to memory
327 * and invalidate the corresponding instruction cache blocks.
328 * This is a no-op on the 601.
329 *
330 * flush_icache_range(unsigned long start, unsigned long stop)
331 */
332_GLOBAL(flush_icache_range)
333BEGIN_FTR_SECTION
334	PURGE_PREFETCHED_INS
335	blr				/* for 601, do nothing */
336END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
337	rlwinm	r3,r3,0,0,31 - L1_CACHE_SHIFT
338	subf	r4,r3,r4
339	addi	r4,r4,L1_CACHE_BYTES - 1
340	srwi.	r4,r4,L1_CACHE_SHIFT
341	beqlr
342	mtctr	r4
343	mr	r6,r3
3441:	dcbst	0,r3
345	addi	r3,r3,L1_CACHE_BYTES
346	bdnz	1b
347	sync				/* wait for dcbst's to get to ram */
348#ifndef CONFIG_44x
349	mtctr	r4
3502:	icbi	0,r6
351	addi	r6,r6,L1_CACHE_BYTES
352	bdnz	2b
353#else
354	/* Flash invalidate on 44x because we are passed kmapped addresses and
355	   this doesn't work for userspace pages due to the virtually tagged
356	   icache.  Sigh. */
357	iccci	0, r0
358#endif
359	sync				/* additional sync needed on g4 */
360	isync
361	blr
362_ASM_NOKPROBE_SYMBOL(flush_icache_range)
363EXPORT_SYMBOL(flush_icache_range)
364
365/*
366 * Flush a particular page from the data cache to RAM.
367 * Note: this is necessary because the instruction cache does *not*
368 * snoop from the data cache.
369 * This is a no-op on the 601 which has a unified cache.
370 *
371 *	void __flush_dcache_icache(void *page)
372 */
373_GLOBAL(__flush_dcache_icache)
374BEGIN_FTR_SECTION
375	PURGE_PREFETCHED_INS
376	blr
377END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
378	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
379	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
380	mtctr	r4
381	mr	r6,r3
3820:	dcbst	0,r3				/* Write line to ram */
383	addi	r3,r3,L1_CACHE_BYTES
384	bdnz	0b
385	sync
386#ifdef CONFIG_44x
387	/* We don't flush the icache on 44x. Those have a virtual icache
388	 * and we don't have access to the virtual address here (it's
389	 * not the page vaddr but where it's mapped in user space). The
390	 * flushing of the icache on these is handled elsewhere, when
391	 * a change in the address space occurs, before returning to
392	 * user space
393	 */
394BEGIN_MMU_FTR_SECTION
395	blr
396END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
397#endif /* CONFIG_44x */
398	mtctr	r4
3991:	icbi	0,r6
400	addi	r6,r6,L1_CACHE_BYTES
401	bdnz	1b
402	sync
403	isync
404	blr
405
406#ifndef CONFIG_BOOKE
407/*
408 * Flush a particular page from the data cache to RAM, identified
409 * by its physical address.  We turn off the MMU so we can just use
410 * the physical address (this may be a highmem page without a kernel
411 * mapping).
412 *
413 *	void __flush_dcache_icache_phys(unsigned long physaddr)
414 */
415_GLOBAL(__flush_dcache_icache_phys)
416BEGIN_FTR_SECTION
417	PURGE_PREFETCHED_INS
418	blr					/* for 601, do nothing */
419END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
420	mfmsr	r10
421	rlwinm	r0,r10,0,28,26			/* clear DR */
422	mtmsr	r0
423	isync
424	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
425	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
426	mtctr	r4
427	mr	r6,r3
4280:	dcbst	0,r3				/* Write line to ram */
429	addi	r3,r3,L1_CACHE_BYTES
430	bdnz	0b
431	sync
432	mtctr	r4
4331:	icbi	0,r6
434	addi	r6,r6,L1_CACHE_BYTES
435	bdnz	1b
436	sync
437	mtmsr	r10				/* restore DR */
438	isync
439	blr
440#endif /* CONFIG_BOOKE */
441
442/*
443 * Copy a whole page.  We use the dcbz instruction on the destination
444 * to reduce memory traffic (it eliminates the unnecessary reads of
445 * the destination into cache).  This requires that the destination
446 * is cacheable.
447 */
448#define COPY_16_BYTES		\
449	lwz	r6,4(r4);	\
450	lwz	r7,8(r4);	\
451	lwz	r8,12(r4);	\
452	lwzu	r9,16(r4);	\
453	stw	r6,4(r3);	\
454	stw	r7,8(r3);	\
455	stw	r8,12(r3);	\
456	stwu	r9,16(r3)
457
458_GLOBAL(copy_page)
459	addi	r3,r3,-4
460	addi	r4,r4,-4
461
462	li	r5,4
463
464#if MAX_COPY_PREFETCH > 1
465	li	r0,MAX_COPY_PREFETCH
466	li	r11,4
467	mtctr	r0
46811:	dcbt	r11,r4
469	addi	r11,r11,L1_CACHE_BYTES
470	bdnz	11b
471#else /* MAX_COPY_PREFETCH == 1 */
472	dcbt	r5,r4
473	li	r11,L1_CACHE_BYTES+4
474#endif /* MAX_COPY_PREFETCH */
475	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
476	crclr	4*cr0+eq
4772:
478	mtctr	r0
4791:
480	dcbt	r11,r4
481	dcbz	r5,r3
482	COPY_16_BYTES
483#if L1_CACHE_BYTES >= 32
484	COPY_16_BYTES
485#if L1_CACHE_BYTES >= 64
486	COPY_16_BYTES
487	COPY_16_BYTES
488#if L1_CACHE_BYTES >= 128
489	COPY_16_BYTES
490	COPY_16_BYTES
491	COPY_16_BYTES
492	COPY_16_BYTES
493#endif
494#endif
495#endif
496	bdnz	1b
497	beqlr
498	crnot	4*cr0+eq,4*cr0+eq
499	li	r0,MAX_COPY_PREFETCH
500	li	r11,4
501	b	2b
502EXPORT_SYMBOL(copy_page)
503
504/*
505 * Extended precision shifts.
506 *
507 * Updated to be valid for shift counts from 0 to 63 inclusive.
508 * -- Gabriel
509 *
510 * R3/R4 has 64 bit value
511 * R5    has shift count
512 * result in R3/R4
513 *
514 *  ashrdi3: arithmetic right shift (sign propagation)
515 *  lshrdi3: logical right shift
516 *  ashldi3: left shift
517 */
518_GLOBAL(__ashrdi3)
519	subfic	r6,r5,32
520	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
521	addi	r7,r5,32	# could be xori, or addi with -32
522	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
523	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
524	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
525	or	r4,r4,r6	# LSW |= t1
526	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
527	sraw	r3,r3,r5	# MSW = MSW >> count
528	or	r4,r4,r7	# LSW |= t2
529	blr
530EXPORT_SYMBOL(__ashrdi3)
531
532_GLOBAL(__ashldi3)
533	subfic	r6,r5,32
534	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
535	addi	r7,r5,32	# could be xori, or addi with -32
536	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
537	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
538	or	r3,r3,r6	# MSW |= t1
539	slw	r4,r4,r5	# LSW = LSW << count
540	or	r3,r3,r7	# MSW |= t2
541	blr
542EXPORT_SYMBOL(__ashldi3)
543
544_GLOBAL(__lshrdi3)
545	subfic	r6,r5,32
546	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
547	addi	r7,r5,32	# could be xori, or addi with -32
548	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
549	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
550	or	r4,r4,r6	# LSW |= t1
551	srw	r3,r3,r5	# MSW = MSW >> count
552	or	r4,r4,r7	# LSW |= t2
553	blr
554EXPORT_SYMBOL(__lshrdi3)
555
556/*
557 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
558 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
559 */
560_GLOBAL(__cmpdi2)
561	cmpw	r3,r5
562	li	r3,1
563	bne	1f
564	cmplw	r4,r6
565	beqlr
5661:	li	r3,0
567	bltlr
568	li	r3,2
569	blr
570EXPORT_SYMBOL(__cmpdi2)
571/*
572 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
573 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
574 */
575_GLOBAL(__ucmpdi2)
576	cmplw	r3,r5
577	li	r3,1
578	bne	1f
579	cmplw	r4,r6
580	beqlr
5811:	li	r3,0
582	bltlr
583	li	r3,2
584	blr
585EXPORT_SYMBOL(__ucmpdi2)
586
587_GLOBAL(__bswapdi2)
588	rotlwi  r9,r4,8
589	rotlwi  r10,r3,8
590	rlwimi  r9,r4,24,0,7
591	rlwimi  r10,r3,24,0,7
592	rlwimi  r9,r4,24,16,23
593	rlwimi  r10,r3,24,16,23
594	mr      r3,r9
595	mr      r4,r10
596	blr
597EXPORT_SYMBOL(__bswapdi2)
598
599#ifdef CONFIG_SMP
600_GLOBAL(start_secondary_resume)
601	/* Reset stack */
602	CURRENT_THREAD_INFO(r1, r1)
603	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
604	li	r3,0
605	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
606	bl	start_secondary
607	b	.
608#endif /* CONFIG_SMP */
609
610/*
611 * This routine is just here to keep GCC happy - sigh...
612 */
613_GLOBAL(__main)
614	blr
615
616#ifdef CONFIG_KEXEC_CORE
617	/*
618	 * Must be relocatable PIC code callable as a C function.
619	 */
620	.globl relocate_new_kernel
621relocate_new_kernel:
622	/* r3 = page_list   */
623	/* r4 = reboot_code_buffer */
624	/* r5 = start_address      */
625
626#ifdef CONFIG_FSL_BOOKE
627
628	mr	r29, r3
629	mr	r30, r4
630	mr	r31, r5
631
632#define ENTRY_MAPPING_KEXEC_SETUP
633#include "fsl_booke_entry_mapping.S"
634#undef ENTRY_MAPPING_KEXEC_SETUP
635
636	mr      r3, r29
637	mr      r4, r30
638	mr      r5, r31
639
640	li	r0, 0
641#elif defined(CONFIG_44x)
642
643	/* Save our parameters */
644	mr	r29, r3
645	mr	r30, r4
646	mr	r31, r5
647
648#ifdef CONFIG_PPC_47x
649	/* Check for 47x cores */
650	mfspr	r3,SPRN_PVR
651	srwi	r3,r3,16
652	cmplwi	cr0,r3,PVR_476FPE@h
653	beq	setup_map_47x
654	cmplwi	cr0,r3,PVR_476@h
655	beq	setup_map_47x
656	cmplwi	cr0,r3,PVR_476_ISS@h
657	beq	setup_map_47x
658#endif /* CONFIG_PPC_47x */
659
660/*
661 * Code for setting up 1:1 mapping for PPC440x for KEXEC
662 *
663 * We cannot switch off the MMU on PPC44x.
664 * So we:
665 * 1) Invalidate all the mappings except the one we are running from.
666 * 2) Create a tmp mapping for our code in the other address space(TS) and
667 *    jump to it. Invalidate the entry we started in.
668 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
669 * 4) Jump to the 1:1 mapping in original TS.
670 * 5) Invalidate the tmp mapping.
671 *
672 * - Based on the kexec support code for FSL BookE
673 *
674 */
675
676	/*
677	 * Load the PID with kernel PID (0).
678	 * Also load our MSR_IS and TID to MMUCR for TLB search.
679	 */
680	li	r3, 0
681	mtspr	SPRN_PID, r3
682	mfmsr	r4
683	andi.	r4,r4,MSR_IS@l
684	beq	wmmucr
685	oris	r3,r3,PPC44x_MMUCR_STS@h
686wmmucr:
687	mtspr	SPRN_MMUCR,r3
688	sync
689
690	/*
691	 * Invalidate all the TLB entries except the current entry
692	 * where we are running from
693	 */
694	bl	0f				/* Find our address */
6950:	mflr	r5				/* Make it accessible */
696	tlbsx	r23,0,r5			/* Find entry we are in */
697	li	r4,0				/* Start at TLB entry 0 */
698	li	r3,0				/* Set PAGEID inval value */
6991:	cmpw	r23,r4				/* Is this our entry? */
700	beq	skip				/* If so, skip the inval */
701	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
702skip:
703	addi	r4,r4,1				/* Increment */
704	cmpwi	r4,64				/* Are we done?	*/
705	bne	1b				/* If not, repeat */
706	isync
707
708	/* Create a temp mapping and jump to it */
709	andi.	r6, r23, 1		/* Find the index to use */
710	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
711
712	mfmsr	r9			/* get the MSR */
713	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
714	xori	r7, r5, 1		/* Use the other address space */
715
716	/* Read the current mapping entries */
717	tlbre	r3, r23, PPC44x_TLB_PAGEID
718	tlbre	r4, r23, PPC44x_TLB_XLAT
719	tlbre	r5, r23, PPC44x_TLB_ATTRIB
720
721	/* Save our current XLAT entry */
722	mr	r25, r4
723
724	/* Extract the TLB PageSize */
725	li	r10, 1 			/* r10 will hold PageSize */
726	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
727
728	/* XXX: As of now we use 256M, 4K pages */
729	cmpwi	r11, PPC44x_TLB_256M
730	bne	tlb_4k
731	rotlwi	r10, r10, 28		/* r10 = 256M */
732	b	write_out
733tlb_4k:
734	cmpwi	r11, PPC44x_TLB_4K
735	bne	default
736	rotlwi	r10, r10, 12		/* r10 = 4K */
737	b	write_out
738default:
739	rotlwi	r10, r10, 10		/* r10 = 1K */
740
741write_out:
742	/*
743	 * Write out the tmp 1:1 mapping for this code in other address space
744	 * Fixup  EPN = RPN , TS=other address space
745	 */
746	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
747
748	/* Write out the tmp mapping entries */
749	tlbwe	r3, r24, PPC44x_TLB_PAGEID
750	tlbwe	r4, r24, PPC44x_TLB_XLAT
751	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
752
753	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
754	not	r10, r11		/* Mask for PageNum */
755
756	/* Switch to other address space in MSR */
757	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
758
759	bl	1f
7601:	mflr	r8
761	addi	r8, r8, (2f-1b)		/* Find the target offset */
762
763	/* Jump to the tmp mapping */
764	mtspr	SPRN_SRR0, r8
765	mtspr	SPRN_SRR1, r9
766	rfi
767
7682:
769	/* Invalidate the entry we were executing from */
770	li	r3, 0
771	tlbwe	r3, r23, PPC44x_TLB_PAGEID
772
773	/* attribute fields. rwx for SUPERVISOR mode */
774	li	r5, 0
775	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
776
777	/* Create 1:1 mapping in 256M pages */
778	xori	r7, r7, 1			/* Revert back to Original TS */
779
780	li	r8, 0				/* PageNumber */
781	li	r6, 3				/* TLB Index, start at 3  */
782
783next_tlb:
784	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
785	mr	r4, r3				/* RPN = EPN  */
786	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
787	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
788
789	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
790	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
791	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
792
793	addi	r8, r8, 1			/* Increment PN */
794	addi	r6, r6, 1			/* Increment TLB Index */
795	cmpwi	r8, 8				/* Are we done ? */
796	bne	next_tlb
797	isync
798
799	/* Jump to the new mapping 1:1 */
800	li	r9,0
801	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
802
803	bl	1f
8041:	mflr	r8
805	and	r8, r8, r11			/* Get our offset within page */
806	addi	r8, r8, (2f-1b)
807
808	and	r5, r25, r10			/* Get our target PageNum */
809	or	r8, r8, r5			/* Target jump address */
810
811	mtspr	SPRN_SRR0, r8
812	mtspr	SPRN_SRR1, r9
813	rfi
8142:
815	/* Invalidate the tmp entry we used */
816	li	r3, 0
817	tlbwe	r3, r24, PPC44x_TLB_PAGEID
818	sync
819	b	ppc44x_map_done
820
821#ifdef CONFIG_PPC_47x
822
823	/* 1:1 mapping for 47x */
824
825setup_map_47x:
826
827	/*
828	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
829	 * Also set the MSR IS->MMUCR STS
830	 */
831	li	r3, 0
832	mtspr	SPRN_PID, r3			/* Set PID */
833	mfmsr	r4				/* Get MSR */
834	andi.	r4, r4, MSR_IS@l		/* TS=1? */
835	beq	1f				/* If not, leave STS=0 */
836	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
8371:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
838	sync
839
840	/* Find the entry we are running from */
841	bl	2f
8422:	mflr	r23
843	tlbsx	r23, 0, r23
844	tlbre	r24, r23, 0			/* TLB Word 0 */
845	tlbre	r25, r23, 1			/* TLB Word 1 */
846	tlbre	r26, r23, 2			/* TLB Word 2 */
847
848
849	/*
850	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
851	 * of 4k page size in all  4 ways (0-3 in r3).
852	 * This would invalidate the entire UTLB including the one we are
853	 * running from. However the shadow TLB entries would help us
854	 * to continue the execution, until we flush them (rfi/isync).
855	 */
856	addis	r3, 0, 0x8000			/* specify the way */
857	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
858	addi	r5, 0, 0
859	b	clear_utlb_entry
860
861	/* Align the loop to speed things up. from head_44x.S */
862	.align	6
863
864clear_utlb_entry:
865
866	tlbwe	r4, r3, 0
867	tlbwe	r5, r3, 1
868	tlbwe	r5, r3, 2
869	addis	r3, r3, 0x2000			/* Increment the way */
870	cmpwi	r3, 0
871	bne	clear_utlb_entry
872	addis	r3, 0, 0x8000
873	addis	r4, r4, 0x100			/* Increment the EPN */
874	cmpwi	r4, 0
875	bne	clear_utlb_entry
876
877	/* Create the entries in the other address space */
878	mfmsr	r5
879	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
880	xori	r7, r7, 1			/* r7 = !TS */
881
882	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
883
884	/*
885	 * write out the TLB entries for the tmp mapping
886	 * Use way '0' so that we could easily invalidate it later.
887	 */
888	lis	r3, 0x8000			/* Way '0' */
889
890	tlbwe	r24, r3, 0
891	tlbwe	r25, r3, 1
892	tlbwe	r26, r3, 2
893
894	/* Update the msr to the new TS */
895	insrwi	r5, r7, 1, 26
896
897	bl	1f
8981:	mflr	r6
899	addi	r6, r6, (2f-1b)
900
901	mtspr	SPRN_SRR0, r6
902	mtspr	SPRN_SRR1, r5
903	rfi
904
905	/*
906	 * Now we are in the tmp address space.
907	 * Create a 1:1 mapping for 0-2GiB in the original TS.
908	 */
9092:
910	li	r3, 0
911	li	r4, 0				/* TLB Word 0 */
912	li	r5, 0				/* TLB Word 1 */
913	li	r6, 0
914	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
915
916	li	r8, 0				/* PageIndex */
917
918	xori	r7, r7, 1			/* revert back to original TS */
919
920write_utlb:
921	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
922						/* ERPN = 0 as we don't use memory above 2G */
923
924	mr	r4, r5				/* EPN = RPN */
925	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
926	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
927
928	tlbwe	r4, r3, 0			/* Write out the entries */
929	tlbwe	r5, r3, 1
930	tlbwe	r6, r3, 2
931	addi	r8, r8, 1
932	cmpwi	r8, 8				/* Have we completed ? */
933	bne	write_utlb
934
935	/* make sure we complete the TLB write up */
936	isync
937
938	/*
939	 * Prepare to jump to the 1:1 mapping.
940	 * 1) Extract page size of the tmp mapping
941	 *    DSIZ = TLB_Word0[22:27]
942	 * 2) Calculate the physical address of the address
943	 *    to jump to.
944	 */
945	rlwinm	r10, r24, 0, 22, 27
946
947	cmpwi	r10, PPC47x_TLB0_4K
948	bne	0f
949	li	r10, 0x1000			/* r10 = 4k */
950	bl	1f
951
9520:
953	/* Defaults to 256M */
954	lis	r10, 0x1000
955
956	bl	1f
9571:	mflr	r4
958	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
959
960	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
961	not	r10, r11			/* Pagemask = ~(offsetmask) */
962
963	and	r5, r25, r10			/* Physical page */
964	and	r6, r4, r11			/* offset within the current page */
965
966	or	r5, r5, r6			/* Physical address for 2f */
967
968	/* Switch the TS in MSR to the original one */
969	mfmsr	r8
970	insrwi	r8, r7, 1, 26
971
972	mtspr	SPRN_SRR1, r8
973	mtspr	SPRN_SRR0, r5
974	rfi
975
9762:
977	/* Invalidate the tmp mapping */
978	lis	r3, 0x8000			/* Way '0' */
979
980	clrrwi	r24, r24, 12			/* Clear the valid bit */
981	tlbwe	r24, r3, 0
982	tlbwe	r25, r3, 1
983	tlbwe	r26, r3, 2
984
985	/* Make sure we complete the TLB write and flush the shadow TLB */
986	isync
987
988#endif
989
990ppc44x_map_done:
991
992
993	/* Restore the parameters */
994	mr	r3, r29
995	mr	r4, r30
996	mr	r5, r31
997
998	li	r0, 0
999#else
1000	li	r0, 0
1001
1002	/*
1003	 * Set Machine Status Register to a known status,
1004	 * switch the MMU off and jump to 1: in a single step.
1005	 */
1006
1007	mr	r8, r0
1008	ori     r8, r8, MSR_RI|MSR_ME
1009	mtspr	SPRN_SRR1, r8
1010	addi	r8, r4, 1f - relocate_new_kernel
1011	mtspr	SPRN_SRR0, r8
1012	sync
1013	rfi
1014
10151:
1016#endif
1017	/* from this point address translation is turned off */
1018	/* and interrupts are disabled */
1019
1020	/* set a new stack at the bottom of our page... */
1021	/* (not really needed now) */
1022	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1023	stw	r0, 0(r1)
1024
1025	/* Do the copies */
1026	li	r6, 0 /* checksum */
1027	mr	r0, r3
1028	b	1f
1029
10300:	/* top, read another word for the indirection page */
1031	lwzu	r0, 4(r3)
1032
10331:
1034	/* is it a destination page? (r8) */
1035	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1036	beq	2f
1037
1038	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1039	b	0b
1040
10412:	/* is it an indirection page? (r3) */
1042	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1043	beq	2f
1044
1045	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1046	subi	r3, r3, 4
1047	b	0b
1048
10492:	/* are we done? */
1050	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1051	beq	2f
1052	b	3f
1053
10542:	/* is it a source page? (r9) */
1055	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1056	beq	0b
1057
1058	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1059
1060	li	r7, PAGE_SIZE / 4
1061	mtctr   r7
1062	subi    r9, r9, 4
1063	subi    r8, r8, 4
10649:
1065	lwzu    r0, 4(r9)  /* do the copy */
1066	xor	r6, r6, r0
1067	stwu    r0, 4(r8)
1068	dcbst	0, r8
1069	sync
1070	icbi	0, r8
1071	bdnz    9b
1072
1073	addi    r9, r9, 4
1074	addi    r8, r8, 4
1075	b	0b
1076
10773:
1078
1079	/* To be certain of avoiding problems with self-modifying code
1080	 * execute a serializing instruction here.
1081	 */
1082	isync
1083	sync
1084
1085	mfspr	r3, SPRN_PIR /* current core we are running on */
1086	mr	r4, r5 /* load physical address of chunk called */
1087
1088	/* jump to the entry point, usually the setup routine */
1089	mtlr	r5
1090	blrl
1091
10921:	b	1b
1093
1094relocate_new_kernel_end:
1095
1096	.globl relocate_new_kernel_size
1097relocate_new_kernel_size:
1098	.long relocate_new_kernel_end - relocate_new_kernel
1099#endif
1100