xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision b8d312aa)
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * This file contains miscellaneous low-level functions.
4 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
5 *
6 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
7 * and Paul Mackerras.
8 *
9 * kexec bits:
10 * Copyright (C) 2002-2003 Eric Biederman  <ebiederm@xmission.com>
11 * GameCube/ppc32 port Copyright (C) 2004 Albert Herranz
12 * PPC44x port. Copyright (C) 2011,  IBM Corporation
13 * 		Author: Suzuki Poulose <suzuki@in.ibm.com>
14 */
15
16#include <linux/sys.h>
17#include <asm/unistd.h>
18#include <asm/errno.h>
19#include <asm/reg.h>
20#include <asm/page.h>
21#include <asm/cache.h>
22#include <asm/cputable.h>
23#include <asm/mmu.h>
24#include <asm/ppc_asm.h>
25#include <asm/thread_info.h>
26#include <asm/asm-offsets.h>
27#include <asm/processor.h>
28#include <asm/kexec.h>
29#include <asm/bug.h>
30#include <asm/ptrace.h>
31#include <asm/export.h>
32#include <asm/feature-fixups.h>
33
34	.text
35
36/*
37 * We store the saved ksp_limit in the unused part
38 * of the STACK_FRAME_OVERHEAD
39 */
40_GLOBAL(call_do_softirq)
41	mflr	r0
42	stw	r0,4(r1)
43	lwz	r10,THREAD+KSP_LIMIT(r2)
44	stw	r3, THREAD+KSP_LIMIT(r2)
45	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
46	mr	r1,r3
47	stw	r10,8(r1)
48	bl	__do_softirq
49	lwz	r10,8(r1)
50	lwz	r1,0(r1)
51	lwz	r0,4(r1)
52	stw	r10,THREAD+KSP_LIMIT(r2)
53	mtlr	r0
54	blr
55
56/*
57 * void call_do_irq(struct pt_regs *regs, void *sp);
58 */
59_GLOBAL(call_do_irq)
60	mflr	r0
61	stw	r0,4(r1)
62	lwz	r10,THREAD+KSP_LIMIT(r2)
63	stw	r4, THREAD+KSP_LIMIT(r2)
64	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
65	mr	r1,r4
66	stw	r10,8(r1)
67	bl	__do_irq
68	lwz	r10,8(r1)
69	lwz	r1,0(r1)
70	lwz	r0,4(r1)
71	stw	r10,THREAD+KSP_LIMIT(r2)
72	mtlr	r0
73	blr
74
75/*
76 * This returns the high 64 bits of the product of two 64-bit numbers.
77 */
78_GLOBAL(mulhdu)
79	cmpwi	r6,0
80	cmpwi	cr1,r3,0
81	mr	r10,r4
82	mulhwu	r4,r4,r5
83	beq	1f
84	mulhwu	r0,r10,r6
85	mullw	r7,r10,r5
86	addc	r7,r0,r7
87	addze	r4,r4
881:	beqlr	cr1		/* all done if high part of A is 0 */
89	mullw	r9,r3,r5
90	mulhwu	r10,r3,r5
91	beq	2f
92	mullw	r0,r3,r6
93	mulhwu	r8,r3,r6
94	addc	r7,r0,r7
95	adde	r4,r4,r8
96	addze	r10,r10
972:	addc	r4,r4,r9
98	addze	r3,r10
99	blr
100
101/*
102 * reloc_got2 runs through the .got2 section adding an offset
103 * to each entry.
104 */
105_GLOBAL(reloc_got2)
106	mflr	r11
107	lis	r7,__got2_start@ha
108	addi	r7,r7,__got2_start@l
109	lis	r8,__got2_end@ha
110	addi	r8,r8,__got2_end@l
111	subf	r8,r7,r8
112	srwi.	r8,r8,2
113	beqlr
114	mtctr	r8
115	bl	1f
1161:	mflr	r0
117	lis	r4,1b@ha
118	addi	r4,r4,1b@l
119	subf	r0,r4,r0
120	add	r7,r0,r7
1212:	lwz	r0,0(r7)
122	add	r0,r0,r3
123	stw	r0,0(r7)
124	addi	r7,r7,4
125	bdnz	2b
126	mtlr	r11
127	blr
128
129/*
130 * call_setup_cpu - call the setup_cpu function for this cpu
131 * r3 = data offset, r24 = cpu number
132 *
133 * Setup function is called with:
134 *   r3 = data offset
135 *   r4 = ptr to CPU spec (relocated)
136 */
137_GLOBAL(call_setup_cpu)
138	addis	r4,r3,cur_cpu_spec@ha
139	addi	r4,r4,cur_cpu_spec@l
140	lwz	r4,0(r4)
141	add	r4,r4,r3
142	lwz	r5,CPU_SPEC_SETUP(r4)
143	cmpwi	0,r5,0
144	add	r5,r5,r3
145	beqlr
146	mtctr	r5
147	bctr
148
149#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_PPC_BOOK3S_32)
150
151/* This gets called by via-pmu.c to switch the PLL selection
152 * on 750fx CPU. This function should really be moved to some
153 * other place (as most of the cpufreq code in via-pmu
154 */
155_GLOBAL(low_choose_750fx_pll)
156	/* Clear MSR:EE */
157	mfmsr	r7
158	rlwinm	r0,r7,0,17,15
159	mtmsr	r0
160
161	/* If switching to PLL1, disable HID0:BTIC */
162	cmplwi	cr0,r3,0
163	beq	1f
164	mfspr	r5,SPRN_HID0
165	rlwinm	r5,r5,0,27,25
166	sync
167	mtspr	SPRN_HID0,r5
168	isync
169	sync
170
1711:
172	/* Calc new HID1 value */
173	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
174	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
175	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
176	or	r4,r4,r5
177	mtspr	SPRN_HID1,r4
178
179#ifdef CONFIG_SMP
180	/* Store new HID1 image */
181	lwz	r6,TASK_CPU(r2)
182	slwi	r6,r6,2
183#else
184	li	r6, 0
185#endif
186	addis	r6,r6,nap_save_hid1@ha
187	stw	r4,nap_save_hid1@l(r6)
188
189	/* If switching to PLL0, enable HID0:BTIC */
190	cmplwi	cr0,r3,0
191	bne	1f
192	mfspr	r5,SPRN_HID0
193	ori	r5,r5,HID0_BTIC
194	sync
195	mtspr	SPRN_HID0,r5
196	isync
197	sync
198
1991:
200	/* Return */
201	mtmsr	r7
202	blr
203
204_GLOBAL(low_choose_7447a_dfs)
205	/* Clear MSR:EE */
206	mfmsr	r7
207	rlwinm	r0,r7,0,17,15
208	mtmsr	r0
209
210	/* Calc new HID1 value */
211	mfspr	r4,SPRN_HID1
212	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
213	sync
214	mtspr	SPRN_HID1,r4
215	sync
216	isync
217
218	/* Return */
219	mtmsr	r7
220	blr
221
222#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_PPC_BOOK3S_32 */
223
224/*
225 * complement mask on the msr then "or" some values on.
226 *     _nmask_and_or_msr(nmask, value_to_or)
227 */
228_GLOBAL(_nmask_and_or_msr)
229	mfmsr	r0		/* Get current msr */
230	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
231	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
232	SYNC			/* Some chip revs have problems here... */
233	mtmsr	r0		/* Update machine state */
234	isync
235	blr			/* Done */
236
237#ifdef CONFIG_40x
238
239/*
240 * Do an IO access in real mode
241 */
242_GLOBAL(real_readb)
243	mfmsr	r7
244	rlwinm	r0,r7,0,~MSR_DR
245	sync
246	mtmsr	r0
247	sync
248	isync
249	lbz	r3,0(r3)
250	sync
251	mtmsr	r7
252	sync
253	isync
254	blr
255
256	/*
257 * Do an IO access in real mode
258 */
259_GLOBAL(real_writeb)
260	mfmsr	r7
261	rlwinm	r0,r7,0,~MSR_DR
262	sync
263	mtmsr	r0
264	sync
265	isync
266	stb	r3,0(r4)
267	sync
268	mtmsr	r7
269	sync
270	isync
271	blr
272
273#endif /* CONFIG_40x */
274
275
276/*
277 * Flush instruction cache.
278 * This is a no-op on the 601.
279 */
280#ifndef CONFIG_PPC_8xx
281_GLOBAL(flush_instruction_cache)
282#if defined(CONFIG_4xx)
283#ifdef CONFIG_403GCX
284	li      r3, 512
285	mtctr   r3
286	lis     r4, KERNELBASE@h
2871:	iccci   0, r4
288	addi    r4, r4, 16
289	bdnz    1b
290#else
291	lis	r3, KERNELBASE@h
292	iccci	0,r3
293#endif
294#elif defined(CONFIG_FSL_BOOKE)
295BEGIN_FTR_SECTION
296	mfspr   r3,SPRN_L1CSR0
297	ori     r3,r3,L1CSR0_CFI|L1CSR0_CLFC
298	/* msync; isync recommended here */
299	mtspr   SPRN_L1CSR0,r3
300	isync
301	blr
302END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
303	mfspr	r3,SPRN_L1CSR1
304	ori	r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
305	mtspr	SPRN_L1CSR1,r3
306#else
307	mfspr	r3,SPRN_PVR
308	rlwinm	r3,r3,16,16,31
309	cmpwi	0,r3,1
310	beqlr			/* for 601, do nothing */
311	/* 603/604 processor - use invalidate-all bit in HID0 */
312	mfspr	r3,SPRN_HID0
313	ori	r3,r3,HID0_ICFI
314	mtspr	SPRN_HID0,r3
315#endif /* CONFIG_4xx */
316	isync
317	blr
318EXPORT_SYMBOL(flush_instruction_cache)
319#endif /* CONFIG_PPC_8xx */
320
321/*
322 * Write any modified data cache blocks out to memory
323 * and invalidate the corresponding instruction cache blocks.
324 * This is a no-op on the 601.
325 *
326 * flush_icache_range(unsigned long start, unsigned long stop)
327 */
328_GLOBAL(flush_icache_range)
329BEGIN_FTR_SECTION
330	PURGE_PREFETCHED_INS
331	blr				/* for 601, do nothing */
332END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
333	rlwinm	r3,r3,0,0,31 - L1_CACHE_SHIFT
334	subf	r4,r3,r4
335	addi	r4,r4,L1_CACHE_BYTES - 1
336	srwi.	r4,r4,L1_CACHE_SHIFT
337	beqlr
338	mtctr	r4
339	mr	r6,r3
3401:	dcbst	0,r3
341	addi	r3,r3,L1_CACHE_BYTES
342	bdnz	1b
343	sync				/* wait for dcbst's to get to ram */
344#ifndef CONFIG_44x
345	mtctr	r4
3462:	icbi	0,r6
347	addi	r6,r6,L1_CACHE_BYTES
348	bdnz	2b
349#else
350	/* Flash invalidate on 44x because we are passed kmapped addresses and
351	   this doesn't work for userspace pages due to the virtually tagged
352	   icache.  Sigh. */
353	iccci	0, r0
354#endif
355	sync				/* additional sync needed on g4 */
356	isync
357	blr
358_ASM_NOKPROBE_SYMBOL(flush_icache_range)
359EXPORT_SYMBOL(flush_icache_range)
360
361/*
362 * Flush a particular page from the data cache to RAM.
363 * Note: this is necessary because the instruction cache does *not*
364 * snoop from the data cache.
365 * This is a no-op on the 601 which has a unified cache.
366 *
367 *	void __flush_dcache_icache(void *page)
368 */
369_GLOBAL(__flush_dcache_icache)
370BEGIN_FTR_SECTION
371	PURGE_PREFETCHED_INS
372	blr
373END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
374	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
375	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
376	mtctr	r4
377	mr	r6,r3
3780:	dcbst	0,r3				/* Write line to ram */
379	addi	r3,r3,L1_CACHE_BYTES
380	bdnz	0b
381	sync
382#ifdef CONFIG_44x
383	/* We don't flush the icache on 44x. Those have a virtual icache
384	 * and we don't have access to the virtual address here (it's
385	 * not the page vaddr but where it's mapped in user space). The
386	 * flushing of the icache on these is handled elsewhere, when
387	 * a change in the address space occurs, before returning to
388	 * user space
389	 */
390BEGIN_MMU_FTR_SECTION
391	blr
392END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
393#endif /* CONFIG_44x */
394	mtctr	r4
3951:	icbi	0,r6
396	addi	r6,r6,L1_CACHE_BYTES
397	bdnz	1b
398	sync
399	isync
400	blr
401
402#ifndef CONFIG_BOOKE
403/*
404 * Flush a particular page from the data cache to RAM, identified
405 * by its physical address.  We turn off the MMU so we can just use
406 * the physical address (this may be a highmem page without a kernel
407 * mapping).
408 *
409 *	void __flush_dcache_icache_phys(unsigned long physaddr)
410 */
411_GLOBAL(__flush_dcache_icache_phys)
412BEGIN_FTR_SECTION
413	PURGE_PREFETCHED_INS
414	blr					/* for 601, do nothing */
415END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
416	mfmsr	r10
417	rlwinm	r0,r10,0,28,26			/* clear DR */
418	mtmsr	r0
419	isync
420	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
421	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
422	mtctr	r4
423	mr	r6,r3
4240:	dcbst	0,r3				/* Write line to ram */
425	addi	r3,r3,L1_CACHE_BYTES
426	bdnz	0b
427	sync
428	mtctr	r4
4291:	icbi	0,r6
430	addi	r6,r6,L1_CACHE_BYTES
431	bdnz	1b
432	sync
433	mtmsr	r10				/* restore DR */
434	isync
435	blr
436#endif /* CONFIG_BOOKE */
437
438/*
439 * Copy a whole page.  We use the dcbz instruction on the destination
440 * to reduce memory traffic (it eliminates the unnecessary reads of
441 * the destination into cache).  This requires that the destination
442 * is cacheable.
443 */
444#define COPY_16_BYTES		\
445	lwz	r6,4(r4);	\
446	lwz	r7,8(r4);	\
447	lwz	r8,12(r4);	\
448	lwzu	r9,16(r4);	\
449	stw	r6,4(r3);	\
450	stw	r7,8(r3);	\
451	stw	r8,12(r3);	\
452	stwu	r9,16(r3)
453
454_GLOBAL(copy_page)
455	addi	r3,r3,-4
456	addi	r4,r4,-4
457
458	li	r5,4
459
460#if MAX_COPY_PREFETCH > 1
461	li	r0,MAX_COPY_PREFETCH
462	li	r11,4
463	mtctr	r0
46411:	dcbt	r11,r4
465	addi	r11,r11,L1_CACHE_BYTES
466	bdnz	11b
467#else /* MAX_COPY_PREFETCH == 1 */
468	dcbt	r5,r4
469	li	r11,L1_CACHE_BYTES+4
470#endif /* MAX_COPY_PREFETCH */
471	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
472	crclr	4*cr0+eq
4732:
474	mtctr	r0
4751:
476	dcbt	r11,r4
477	dcbz	r5,r3
478	COPY_16_BYTES
479#if L1_CACHE_BYTES >= 32
480	COPY_16_BYTES
481#if L1_CACHE_BYTES >= 64
482	COPY_16_BYTES
483	COPY_16_BYTES
484#if L1_CACHE_BYTES >= 128
485	COPY_16_BYTES
486	COPY_16_BYTES
487	COPY_16_BYTES
488	COPY_16_BYTES
489#endif
490#endif
491#endif
492	bdnz	1b
493	beqlr
494	crnot	4*cr0+eq,4*cr0+eq
495	li	r0,MAX_COPY_PREFETCH
496	li	r11,4
497	b	2b
498EXPORT_SYMBOL(copy_page)
499
500/*
501 * Extended precision shifts.
502 *
503 * Updated to be valid for shift counts from 0 to 63 inclusive.
504 * -- Gabriel
505 *
506 * R3/R4 has 64 bit value
507 * R5    has shift count
508 * result in R3/R4
509 *
510 *  ashrdi3: arithmetic right shift (sign propagation)
511 *  lshrdi3: logical right shift
512 *  ashldi3: left shift
513 */
514_GLOBAL(__ashrdi3)
515	subfic	r6,r5,32
516	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
517	addi	r7,r5,32	# could be xori, or addi with -32
518	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
519	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
520	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
521	or	r4,r4,r6	# LSW |= t1
522	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
523	sraw	r3,r3,r5	# MSW = MSW >> count
524	or	r4,r4,r7	# LSW |= t2
525	blr
526EXPORT_SYMBOL(__ashrdi3)
527
528_GLOBAL(__ashldi3)
529	subfic	r6,r5,32
530	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
531	addi	r7,r5,32	# could be xori, or addi with -32
532	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
533	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
534	or	r3,r3,r6	# MSW |= t1
535	slw	r4,r4,r5	# LSW = LSW << count
536	or	r3,r3,r7	# MSW |= t2
537	blr
538EXPORT_SYMBOL(__ashldi3)
539
540_GLOBAL(__lshrdi3)
541	subfic	r6,r5,32
542	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
543	addi	r7,r5,32	# could be xori, or addi with -32
544	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
545	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
546	or	r4,r4,r6	# LSW |= t1
547	srw	r3,r3,r5	# MSW = MSW >> count
548	or	r4,r4,r7	# LSW |= t2
549	blr
550EXPORT_SYMBOL(__lshrdi3)
551
552/*
553 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
554 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
555 */
556_GLOBAL(__cmpdi2)
557	cmpw	r3,r5
558	li	r3,1
559	bne	1f
560	cmplw	r4,r6
561	beqlr
5621:	li	r3,0
563	bltlr
564	li	r3,2
565	blr
566EXPORT_SYMBOL(__cmpdi2)
567/*
568 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
569 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
570 */
571_GLOBAL(__ucmpdi2)
572	cmplw	r3,r5
573	li	r3,1
574	bne	1f
575	cmplw	r4,r6
576	beqlr
5771:	li	r3,0
578	bltlr
579	li	r3,2
580	blr
581EXPORT_SYMBOL(__ucmpdi2)
582
583_GLOBAL(__bswapdi2)
584	rotlwi  r9,r4,8
585	rotlwi  r10,r3,8
586	rlwimi  r9,r4,24,0,7
587	rlwimi  r10,r3,24,0,7
588	rlwimi  r9,r4,24,16,23
589	rlwimi  r10,r3,24,16,23
590	mr      r3,r9
591	mr      r4,r10
592	blr
593EXPORT_SYMBOL(__bswapdi2)
594
595#ifdef CONFIG_SMP
596_GLOBAL(start_secondary_resume)
597	/* Reset stack */
598	rlwinm	r1, r1, 0, 0, 31 - THREAD_SHIFT
599	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
600	li	r3,0
601	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
602	bl	start_secondary
603	b	.
604#endif /* CONFIG_SMP */
605
606/*
607 * This routine is just here to keep GCC happy - sigh...
608 */
609_GLOBAL(__main)
610	blr
611
612#ifdef CONFIG_KEXEC_CORE
613	/*
614	 * Must be relocatable PIC code callable as a C function.
615	 */
616	.globl relocate_new_kernel
617relocate_new_kernel:
618	/* r3 = page_list   */
619	/* r4 = reboot_code_buffer */
620	/* r5 = start_address      */
621
622#ifdef CONFIG_FSL_BOOKE
623
624	mr	r29, r3
625	mr	r30, r4
626	mr	r31, r5
627
628#define ENTRY_MAPPING_KEXEC_SETUP
629#include "fsl_booke_entry_mapping.S"
630#undef ENTRY_MAPPING_KEXEC_SETUP
631
632	mr      r3, r29
633	mr      r4, r30
634	mr      r5, r31
635
636	li	r0, 0
637#elif defined(CONFIG_44x)
638
639	/* Save our parameters */
640	mr	r29, r3
641	mr	r30, r4
642	mr	r31, r5
643
644#ifdef CONFIG_PPC_47x
645	/* Check for 47x cores */
646	mfspr	r3,SPRN_PVR
647	srwi	r3,r3,16
648	cmplwi	cr0,r3,PVR_476FPE@h
649	beq	setup_map_47x
650	cmplwi	cr0,r3,PVR_476@h
651	beq	setup_map_47x
652	cmplwi	cr0,r3,PVR_476_ISS@h
653	beq	setup_map_47x
654#endif /* CONFIG_PPC_47x */
655
656/*
657 * Code for setting up 1:1 mapping for PPC440x for KEXEC
658 *
659 * We cannot switch off the MMU on PPC44x.
660 * So we:
661 * 1) Invalidate all the mappings except the one we are running from.
662 * 2) Create a tmp mapping for our code in the other address space(TS) and
663 *    jump to it. Invalidate the entry we started in.
664 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
665 * 4) Jump to the 1:1 mapping in original TS.
666 * 5) Invalidate the tmp mapping.
667 *
668 * - Based on the kexec support code for FSL BookE
669 *
670 */
671
672	/*
673	 * Load the PID with kernel PID (0).
674	 * Also load our MSR_IS and TID to MMUCR for TLB search.
675	 */
676	li	r3, 0
677	mtspr	SPRN_PID, r3
678	mfmsr	r4
679	andi.	r4,r4,MSR_IS@l
680	beq	wmmucr
681	oris	r3,r3,PPC44x_MMUCR_STS@h
682wmmucr:
683	mtspr	SPRN_MMUCR,r3
684	sync
685
686	/*
687	 * Invalidate all the TLB entries except the current entry
688	 * where we are running from
689	 */
690	bl	0f				/* Find our address */
6910:	mflr	r5				/* Make it accessible */
692	tlbsx	r23,0,r5			/* Find entry we are in */
693	li	r4,0				/* Start at TLB entry 0 */
694	li	r3,0				/* Set PAGEID inval value */
6951:	cmpw	r23,r4				/* Is this our entry? */
696	beq	skip				/* If so, skip the inval */
697	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
698skip:
699	addi	r4,r4,1				/* Increment */
700	cmpwi	r4,64				/* Are we done?	*/
701	bne	1b				/* If not, repeat */
702	isync
703
704	/* Create a temp mapping and jump to it */
705	andi.	r6, r23, 1		/* Find the index to use */
706	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
707
708	mfmsr	r9			/* get the MSR */
709	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
710	xori	r7, r5, 1		/* Use the other address space */
711
712	/* Read the current mapping entries */
713	tlbre	r3, r23, PPC44x_TLB_PAGEID
714	tlbre	r4, r23, PPC44x_TLB_XLAT
715	tlbre	r5, r23, PPC44x_TLB_ATTRIB
716
717	/* Save our current XLAT entry */
718	mr	r25, r4
719
720	/* Extract the TLB PageSize */
721	li	r10, 1 			/* r10 will hold PageSize */
722	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
723
724	/* XXX: As of now we use 256M, 4K pages */
725	cmpwi	r11, PPC44x_TLB_256M
726	bne	tlb_4k
727	rotlwi	r10, r10, 28		/* r10 = 256M */
728	b	write_out
729tlb_4k:
730	cmpwi	r11, PPC44x_TLB_4K
731	bne	default
732	rotlwi	r10, r10, 12		/* r10 = 4K */
733	b	write_out
734default:
735	rotlwi	r10, r10, 10		/* r10 = 1K */
736
737write_out:
738	/*
739	 * Write out the tmp 1:1 mapping for this code in other address space
740	 * Fixup  EPN = RPN , TS=other address space
741	 */
742	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
743
744	/* Write out the tmp mapping entries */
745	tlbwe	r3, r24, PPC44x_TLB_PAGEID
746	tlbwe	r4, r24, PPC44x_TLB_XLAT
747	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
748
749	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
750	not	r10, r11		/* Mask for PageNum */
751
752	/* Switch to other address space in MSR */
753	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
754
755	bl	1f
7561:	mflr	r8
757	addi	r8, r8, (2f-1b)		/* Find the target offset */
758
759	/* Jump to the tmp mapping */
760	mtspr	SPRN_SRR0, r8
761	mtspr	SPRN_SRR1, r9
762	rfi
763
7642:
765	/* Invalidate the entry we were executing from */
766	li	r3, 0
767	tlbwe	r3, r23, PPC44x_TLB_PAGEID
768
769	/* attribute fields. rwx for SUPERVISOR mode */
770	li	r5, 0
771	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
772
773	/* Create 1:1 mapping in 256M pages */
774	xori	r7, r7, 1			/* Revert back to Original TS */
775
776	li	r8, 0				/* PageNumber */
777	li	r6, 3				/* TLB Index, start at 3  */
778
779next_tlb:
780	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
781	mr	r4, r3				/* RPN = EPN  */
782	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
783	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
784
785	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
786	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
787	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
788
789	addi	r8, r8, 1			/* Increment PN */
790	addi	r6, r6, 1			/* Increment TLB Index */
791	cmpwi	r8, 8				/* Are we done ? */
792	bne	next_tlb
793	isync
794
795	/* Jump to the new mapping 1:1 */
796	li	r9,0
797	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
798
799	bl	1f
8001:	mflr	r8
801	and	r8, r8, r11			/* Get our offset within page */
802	addi	r8, r8, (2f-1b)
803
804	and	r5, r25, r10			/* Get our target PageNum */
805	or	r8, r8, r5			/* Target jump address */
806
807	mtspr	SPRN_SRR0, r8
808	mtspr	SPRN_SRR1, r9
809	rfi
8102:
811	/* Invalidate the tmp entry we used */
812	li	r3, 0
813	tlbwe	r3, r24, PPC44x_TLB_PAGEID
814	sync
815	b	ppc44x_map_done
816
817#ifdef CONFIG_PPC_47x
818
819	/* 1:1 mapping for 47x */
820
821setup_map_47x:
822
823	/*
824	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
825	 * Also set the MSR IS->MMUCR STS
826	 */
827	li	r3, 0
828	mtspr	SPRN_PID, r3			/* Set PID */
829	mfmsr	r4				/* Get MSR */
830	andi.	r4, r4, MSR_IS@l		/* TS=1? */
831	beq	1f				/* If not, leave STS=0 */
832	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
8331:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
834	sync
835
836	/* Find the entry we are running from */
837	bl	2f
8382:	mflr	r23
839	tlbsx	r23, 0, r23
840	tlbre	r24, r23, 0			/* TLB Word 0 */
841	tlbre	r25, r23, 1			/* TLB Word 1 */
842	tlbre	r26, r23, 2			/* TLB Word 2 */
843
844
845	/*
846	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
847	 * of 4k page size in all  4 ways (0-3 in r3).
848	 * This would invalidate the entire UTLB including the one we are
849	 * running from. However the shadow TLB entries would help us
850	 * to continue the execution, until we flush them (rfi/isync).
851	 */
852	addis	r3, 0, 0x8000			/* specify the way */
853	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
854	addi	r5, 0, 0
855	b	clear_utlb_entry
856
857	/* Align the loop to speed things up. from head_44x.S */
858	.align	6
859
860clear_utlb_entry:
861
862	tlbwe	r4, r3, 0
863	tlbwe	r5, r3, 1
864	tlbwe	r5, r3, 2
865	addis	r3, r3, 0x2000			/* Increment the way */
866	cmpwi	r3, 0
867	bne	clear_utlb_entry
868	addis	r3, 0, 0x8000
869	addis	r4, r4, 0x100			/* Increment the EPN */
870	cmpwi	r4, 0
871	bne	clear_utlb_entry
872
873	/* Create the entries in the other address space */
874	mfmsr	r5
875	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
876	xori	r7, r7, 1			/* r7 = !TS */
877
878	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
879
880	/*
881	 * write out the TLB entries for the tmp mapping
882	 * Use way '0' so that we could easily invalidate it later.
883	 */
884	lis	r3, 0x8000			/* Way '0' */
885
886	tlbwe	r24, r3, 0
887	tlbwe	r25, r3, 1
888	tlbwe	r26, r3, 2
889
890	/* Update the msr to the new TS */
891	insrwi	r5, r7, 1, 26
892
893	bl	1f
8941:	mflr	r6
895	addi	r6, r6, (2f-1b)
896
897	mtspr	SPRN_SRR0, r6
898	mtspr	SPRN_SRR1, r5
899	rfi
900
901	/*
902	 * Now we are in the tmp address space.
903	 * Create a 1:1 mapping for 0-2GiB in the original TS.
904	 */
9052:
906	li	r3, 0
907	li	r4, 0				/* TLB Word 0 */
908	li	r5, 0				/* TLB Word 1 */
909	li	r6, 0
910	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
911
912	li	r8, 0				/* PageIndex */
913
914	xori	r7, r7, 1			/* revert back to original TS */
915
916write_utlb:
917	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
918						/* ERPN = 0 as we don't use memory above 2G */
919
920	mr	r4, r5				/* EPN = RPN */
921	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
922	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
923
924	tlbwe	r4, r3, 0			/* Write out the entries */
925	tlbwe	r5, r3, 1
926	tlbwe	r6, r3, 2
927	addi	r8, r8, 1
928	cmpwi	r8, 8				/* Have we completed ? */
929	bne	write_utlb
930
931	/* make sure we complete the TLB write up */
932	isync
933
934	/*
935	 * Prepare to jump to the 1:1 mapping.
936	 * 1) Extract page size of the tmp mapping
937	 *    DSIZ = TLB_Word0[22:27]
938	 * 2) Calculate the physical address of the address
939	 *    to jump to.
940	 */
941	rlwinm	r10, r24, 0, 22, 27
942
943	cmpwi	r10, PPC47x_TLB0_4K
944	bne	0f
945	li	r10, 0x1000			/* r10 = 4k */
946	bl	1f
947
9480:
949	/* Defaults to 256M */
950	lis	r10, 0x1000
951
952	bl	1f
9531:	mflr	r4
954	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
955
956	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
957	not	r10, r11			/* Pagemask = ~(offsetmask) */
958
959	and	r5, r25, r10			/* Physical page */
960	and	r6, r4, r11			/* offset within the current page */
961
962	or	r5, r5, r6			/* Physical address for 2f */
963
964	/* Switch the TS in MSR to the original one */
965	mfmsr	r8
966	insrwi	r8, r7, 1, 26
967
968	mtspr	SPRN_SRR1, r8
969	mtspr	SPRN_SRR0, r5
970	rfi
971
9722:
973	/* Invalidate the tmp mapping */
974	lis	r3, 0x8000			/* Way '0' */
975
976	clrrwi	r24, r24, 12			/* Clear the valid bit */
977	tlbwe	r24, r3, 0
978	tlbwe	r25, r3, 1
979	tlbwe	r26, r3, 2
980
981	/* Make sure we complete the TLB write and flush the shadow TLB */
982	isync
983
984#endif
985
986ppc44x_map_done:
987
988
989	/* Restore the parameters */
990	mr	r3, r29
991	mr	r4, r30
992	mr	r5, r31
993
994	li	r0, 0
995#else
996	li	r0, 0
997
998	/*
999	 * Set Machine Status Register to a known status,
1000	 * switch the MMU off and jump to 1: in a single step.
1001	 */
1002
1003	mr	r8, r0
1004	ori     r8, r8, MSR_RI|MSR_ME
1005	mtspr	SPRN_SRR1, r8
1006	addi	r8, r4, 1f - relocate_new_kernel
1007	mtspr	SPRN_SRR0, r8
1008	sync
1009	rfi
1010
10111:
1012#endif
1013	/* from this point address translation is turned off */
1014	/* and interrupts are disabled */
1015
1016	/* set a new stack at the bottom of our page... */
1017	/* (not really needed now) */
1018	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1019	stw	r0, 0(r1)
1020
1021	/* Do the copies */
1022	li	r6, 0 /* checksum */
1023	mr	r0, r3
1024	b	1f
1025
10260:	/* top, read another word for the indirection page */
1027	lwzu	r0, 4(r3)
1028
10291:
1030	/* is it a destination page? (r8) */
1031	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1032	beq	2f
1033
1034	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1035	b	0b
1036
10372:	/* is it an indirection page? (r3) */
1038	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1039	beq	2f
1040
1041	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1042	subi	r3, r3, 4
1043	b	0b
1044
10452:	/* are we done? */
1046	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1047	beq	2f
1048	b	3f
1049
10502:	/* is it a source page? (r9) */
1051	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1052	beq	0b
1053
1054	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1055
1056	li	r7, PAGE_SIZE / 4
1057	mtctr   r7
1058	subi    r9, r9, 4
1059	subi    r8, r8, 4
10609:
1061	lwzu    r0, 4(r9)  /* do the copy */
1062	xor	r6, r6, r0
1063	stwu    r0, 4(r8)
1064	dcbst	0, r8
1065	sync
1066	icbi	0, r8
1067	bdnz    9b
1068
1069	addi    r9, r9, 4
1070	addi    r8, r8, 4
1071	b	0b
1072
10733:
1074
1075	/* To be certain of avoiding problems with self-modifying code
1076	 * execute a serializing instruction here.
1077	 */
1078	isync
1079	sync
1080
1081	mfspr	r3, SPRN_PIR /* current core we are running on */
1082	mr	r4, r5 /* load physical address of chunk called */
1083
1084	/* jump to the entry point, usually the setup routine */
1085	mtlr	r5
1086	blrl
1087
10881:	b	1b
1089
1090relocate_new_kernel_end:
1091
1092	.globl relocate_new_kernel_size
1093relocate_new_kernel_size:
1094	.long relocate_new_kernel_end - relocate_new_kernel
1095#endif
1096