xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision 3e26a691)
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
37	.text
38
39/*
40 * We store the saved ksp_limit in the unused part
41 * of the STACK_FRAME_OVERHEAD
42 */
43_GLOBAL(call_do_softirq)
44	mflr	r0
45	stw	r0,4(r1)
46	lwz	r10,THREAD+KSP_LIMIT(r2)
47	addi	r11,r3,THREAD_INFO_GAP
48	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
49	mr	r1,r3
50	stw	r10,8(r1)
51	stw	r11,THREAD+KSP_LIMIT(r2)
52	bl	__do_softirq
53	lwz	r10,8(r1)
54	lwz	r1,0(r1)
55	lwz	r0,4(r1)
56	stw	r10,THREAD+KSP_LIMIT(r2)
57	mtlr	r0
58	blr
59
60/*
61 * void call_do_irq(struct pt_regs *regs, struct thread_info *irqtp);
62 */
63_GLOBAL(call_do_irq)
64	mflr	r0
65	stw	r0,4(r1)
66	lwz	r10,THREAD+KSP_LIMIT(r2)
67	addi	r11,r4,THREAD_INFO_GAP
68	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
69	mr	r1,r4
70	stw	r10,8(r1)
71	stw	r11,THREAD+KSP_LIMIT(r2)
72	bl	__do_irq
73	lwz	r10,8(r1)
74	lwz	r1,0(r1)
75	lwz	r0,4(r1)
76	stw	r10,THREAD+KSP_LIMIT(r2)
77	mtlr	r0
78	blr
79
80/*
81 * This returns the high 64 bits of the product of two 64-bit numbers.
82 */
83_GLOBAL(mulhdu)
84	cmpwi	r6,0
85	cmpwi	cr1,r3,0
86	mr	r10,r4
87	mulhwu	r4,r4,r5
88	beq	1f
89	mulhwu	r0,r10,r6
90	mullw	r7,r10,r5
91	addc	r7,r0,r7
92	addze	r4,r4
931:	beqlr	cr1		/* all done if high part of A is 0 */
94	mullw	r9,r3,r5
95	mulhwu	r10,r3,r5
96	beq	2f
97	mullw	r0,r3,r6
98	mulhwu	r8,r3,r6
99	addc	r7,r0,r7
100	adde	r4,r4,r8
101	addze	r10,r10
1022:	addc	r4,r4,r9
103	addze	r3,r10
104	blr
105
106/*
107 * sub_reloc_offset(x) returns x - reloc_offset().
108 */
109_GLOBAL(sub_reloc_offset)
110	mflr	r0
111	bl	1f
1121:	mflr	r5
113	lis	r4,1b@ha
114	addi	r4,r4,1b@l
115	subf	r5,r4,r5
116	subf	r3,r5,r3
117	mtlr	r0
118	blr
119
120/*
121 * reloc_got2 runs through the .got2 section adding an offset
122 * to each entry.
123 */
124_GLOBAL(reloc_got2)
125	mflr	r11
126	lis	r7,__got2_start@ha
127	addi	r7,r7,__got2_start@l
128	lis	r8,__got2_end@ha
129	addi	r8,r8,__got2_end@l
130	subf	r8,r7,r8
131	srwi.	r8,r8,2
132	beqlr
133	mtctr	r8
134	bl	1f
1351:	mflr	r0
136	lis	r4,1b@ha
137	addi	r4,r4,1b@l
138	subf	r0,r4,r0
139	add	r7,r0,r7
1402:	lwz	r0,0(r7)
141	add	r0,r0,r3
142	stw	r0,0(r7)
143	addi	r7,r7,4
144	bdnz	2b
145	mtlr	r11
146	blr
147
148/*
149 * call_setup_cpu - call the setup_cpu function for this cpu
150 * r3 = data offset, r24 = cpu number
151 *
152 * Setup function is called with:
153 *   r3 = data offset
154 *   r4 = ptr to CPU spec (relocated)
155 */
156_GLOBAL(call_setup_cpu)
157	addis	r4,r3,cur_cpu_spec@ha
158	addi	r4,r4,cur_cpu_spec@l
159	lwz	r4,0(r4)
160	add	r4,r4,r3
161	lwz	r5,CPU_SPEC_SETUP(r4)
162	cmpwi	0,r5,0
163	add	r5,r5,r3
164	beqlr
165	mtctr	r5
166	bctr
167
168#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
169
170/* This gets called by via-pmu.c to switch the PLL selection
171 * on 750fx CPU. This function should really be moved to some
172 * other place (as most of the cpufreq code in via-pmu
173 */
174_GLOBAL(low_choose_750fx_pll)
175	/* Clear MSR:EE */
176	mfmsr	r7
177	rlwinm	r0,r7,0,17,15
178	mtmsr	r0
179
180	/* If switching to PLL1, disable HID0:BTIC */
181	cmplwi	cr0,r3,0
182	beq	1f
183	mfspr	r5,SPRN_HID0
184	rlwinm	r5,r5,0,27,25
185	sync
186	mtspr	SPRN_HID0,r5
187	isync
188	sync
189
1901:
191	/* Calc new HID1 value */
192	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
193	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
194	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
195	or	r4,r4,r5
196	mtspr	SPRN_HID1,r4
197
198	/* Store new HID1 image */
199	CURRENT_THREAD_INFO(r6, r1)
200	lwz	r6,TI_CPU(r6)
201	slwi	r6,r6,2
202	addis	r6,r6,nap_save_hid1@ha
203	stw	r4,nap_save_hid1@l(r6)
204
205	/* If switching to PLL0, enable HID0:BTIC */
206	cmplwi	cr0,r3,0
207	bne	1f
208	mfspr	r5,SPRN_HID0
209	ori	r5,r5,HID0_BTIC
210	sync
211	mtspr	SPRN_HID0,r5
212	isync
213	sync
214
2151:
216	/* Return */
217	mtmsr	r7
218	blr
219
220_GLOBAL(low_choose_7447a_dfs)
221	/* Clear MSR:EE */
222	mfmsr	r7
223	rlwinm	r0,r7,0,17,15
224	mtmsr	r0
225
226	/* Calc new HID1 value */
227	mfspr	r4,SPRN_HID1
228	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
229	sync
230	mtspr	SPRN_HID1,r4
231	sync
232	isync
233
234	/* Return */
235	mtmsr	r7
236	blr
237
238#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
239
240/*
241 * complement mask on the msr then "or" some values on.
242 *     _nmask_and_or_msr(nmask, value_to_or)
243 */
244_GLOBAL(_nmask_and_or_msr)
245	mfmsr	r0		/* Get current msr */
246	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
247	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
248	SYNC			/* Some chip revs have problems here... */
249	mtmsr	r0		/* Update machine state */
250	isync
251	blr			/* Done */
252
253#ifdef CONFIG_40x
254
255/*
256 * Do an IO access in real mode
257 */
258_GLOBAL(real_readb)
259	mfmsr	r7
260	ori	r0,r7,MSR_DR
261	xori	r0,r0,MSR_DR
262	sync
263	mtmsr	r0
264	sync
265	isync
266	lbz	r3,0(r3)
267	sync
268	mtmsr	r7
269	sync
270	isync
271	blr
272
273	/*
274 * Do an IO access in real mode
275 */
276_GLOBAL(real_writeb)
277	mfmsr	r7
278	ori	r0,r7,MSR_DR
279	xori	r0,r0,MSR_DR
280	sync
281	mtmsr	r0
282	sync
283	isync
284	stb	r3,0(r4)
285	sync
286	mtmsr	r7
287	sync
288	isync
289	blr
290
291#endif /* CONFIG_40x */
292
293
294/*
295 * Flush instruction cache.
296 * This is a no-op on the 601.
297 */
298#ifndef CONFIG_PPC_8xx
299_GLOBAL(flush_instruction_cache)
300#if defined(CONFIG_4xx)
301#ifdef CONFIG_403GCX
302	li      r3, 512
303	mtctr   r3
304	lis     r4, KERNELBASE@h
3051:	iccci   0, r4
306	addi    r4, r4, 16
307	bdnz    1b
308#else
309	lis	r3, KERNELBASE@h
310	iccci	0,r3
311#endif
312#elif CONFIG_FSL_BOOKE
313BEGIN_FTR_SECTION
314	mfspr   r3,SPRN_L1CSR0
315	ori     r3,r3,L1CSR0_CFI|L1CSR0_CLFC
316	/* msync; isync recommended here */
317	mtspr   SPRN_L1CSR0,r3
318	isync
319	blr
320END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
321	mfspr	r3,SPRN_L1CSR1
322	ori	r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
323	mtspr	SPRN_L1CSR1,r3
324#else
325	mfspr	r3,SPRN_PVR
326	rlwinm	r3,r3,16,16,31
327	cmpwi	0,r3,1
328	beqlr			/* for 601, do nothing */
329	/* 603/604 processor - use invalidate-all bit in HID0 */
330	mfspr	r3,SPRN_HID0
331	ori	r3,r3,HID0_ICFI
332	mtspr	SPRN_HID0,r3
333#endif /* CONFIG_4xx */
334	isync
335	blr
336#endif /* CONFIG_PPC_8xx */
337
338/*
339 * Write any modified data cache blocks out to memory
340 * and invalidate the corresponding instruction cache blocks.
341 * This is a no-op on the 601.
342 *
343 * flush_icache_range(unsigned long start, unsigned long stop)
344 */
345_KPROBE(flush_icache_range)
346BEGIN_FTR_SECTION
347	PURGE_PREFETCHED_INS
348	blr				/* for 601, do nothing */
349END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
350	rlwinm	r3,r3,0,0,31 - L1_CACHE_SHIFT
351	subf	r4,r3,r4
352	addi	r4,r4,L1_CACHE_BYTES - 1
353	srwi.	r4,r4,L1_CACHE_SHIFT
354	beqlr
355	mtctr	r4
356	mr	r6,r3
3571:	dcbst	0,r3
358	addi	r3,r3,L1_CACHE_BYTES
359	bdnz	1b
360	sync				/* wait for dcbst's to get to ram */
361#ifndef CONFIG_44x
362	mtctr	r4
3632:	icbi	0,r6
364	addi	r6,r6,L1_CACHE_BYTES
365	bdnz	2b
366#else
367	/* Flash invalidate on 44x because we are passed kmapped addresses and
368	   this doesn't work for userspace pages due to the virtually tagged
369	   icache.  Sigh. */
370	iccci	0, r0
371#endif
372	sync				/* additional sync needed on g4 */
373	isync
374	blr
375/*
376 * Flush a particular page from the data cache to RAM.
377 * Note: this is necessary because the instruction cache does *not*
378 * snoop from the data cache.
379 * This is a no-op on the 601 which has a unified cache.
380 *
381 *	void __flush_dcache_icache(void *page)
382 */
383_GLOBAL(__flush_dcache_icache)
384BEGIN_FTR_SECTION
385	PURGE_PREFETCHED_INS
386	blr
387END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
388	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
389	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
390	mtctr	r4
391	mr	r6,r3
3920:	dcbst	0,r3				/* Write line to ram */
393	addi	r3,r3,L1_CACHE_BYTES
394	bdnz	0b
395	sync
396#ifdef CONFIG_44x
397	/* We don't flush the icache on 44x. Those have a virtual icache
398	 * and we don't have access to the virtual address here (it's
399	 * not the page vaddr but where it's mapped in user space). The
400	 * flushing of the icache on these is handled elsewhere, when
401	 * a change in the address space occurs, before returning to
402	 * user space
403	 */
404BEGIN_MMU_FTR_SECTION
405	blr
406END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
407#endif /* CONFIG_44x */
408	mtctr	r4
4091:	icbi	0,r6
410	addi	r6,r6,L1_CACHE_BYTES
411	bdnz	1b
412	sync
413	isync
414	blr
415
416#ifndef CONFIG_BOOKE
417/*
418 * Flush a particular page from the data cache to RAM, identified
419 * by its physical address.  We turn off the MMU so we can just use
420 * the physical address (this may be a highmem page without a kernel
421 * mapping).
422 *
423 *	void __flush_dcache_icache_phys(unsigned long physaddr)
424 */
425_GLOBAL(__flush_dcache_icache_phys)
426BEGIN_FTR_SECTION
427	PURGE_PREFETCHED_INS
428	blr					/* for 601, do nothing */
429END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
430	mfmsr	r10
431	rlwinm	r0,r10,0,28,26			/* clear DR */
432	mtmsr	r0
433	isync
434	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
435	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
436	mtctr	r4
437	mr	r6,r3
4380:	dcbst	0,r3				/* Write line to ram */
439	addi	r3,r3,L1_CACHE_BYTES
440	bdnz	0b
441	sync
442	mtctr	r4
4431:	icbi	0,r6
444	addi	r6,r6,L1_CACHE_BYTES
445	bdnz	1b
446	sync
447	mtmsr	r10				/* restore DR */
448	isync
449	blr
450#endif /* CONFIG_BOOKE */
451
452/*
453 * Copy a whole page.  We use the dcbz instruction on the destination
454 * to reduce memory traffic (it eliminates the unnecessary reads of
455 * the destination into cache).  This requires that the destination
456 * is cacheable.
457 */
458#define COPY_16_BYTES		\
459	lwz	r6,4(r4);	\
460	lwz	r7,8(r4);	\
461	lwz	r8,12(r4);	\
462	lwzu	r9,16(r4);	\
463	stw	r6,4(r3);	\
464	stw	r7,8(r3);	\
465	stw	r8,12(r3);	\
466	stwu	r9,16(r3)
467
468_GLOBAL(copy_page)
469	addi	r3,r3,-4
470	addi	r4,r4,-4
471
472	li	r5,4
473
474#if MAX_COPY_PREFETCH > 1
475	li	r0,MAX_COPY_PREFETCH
476	li	r11,4
477	mtctr	r0
47811:	dcbt	r11,r4
479	addi	r11,r11,L1_CACHE_BYTES
480	bdnz	11b
481#else /* MAX_COPY_PREFETCH == 1 */
482	dcbt	r5,r4
483	li	r11,L1_CACHE_BYTES+4
484#endif /* MAX_COPY_PREFETCH */
485	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
486	crclr	4*cr0+eq
4872:
488	mtctr	r0
4891:
490	dcbt	r11,r4
491	dcbz	r5,r3
492	COPY_16_BYTES
493#if L1_CACHE_BYTES >= 32
494	COPY_16_BYTES
495#if L1_CACHE_BYTES >= 64
496	COPY_16_BYTES
497	COPY_16_BYTES
498#if L1_CACHE_BYTES >= 128
499	COPY_16_BYTES
500	COPY_16_BYTES
501	COPY_16_BYTES
502	COPY_16_BYTES
503#endif
504#endif
505#endif
506	bdnz	1b
507	beqlr
508	crnot	4*cr0+eq,4*cr0+eq
509	li	r0,MAX_COPY_PREFETCH
510	li	r11,4
511	b	2b
512
513/*
514 * Extended precision shifts.
515 *
516 * Updated to be valid for shift counts from 0 to 63 inclusive.
517 * -- Gabriel
518 *
519 * R3/R4 has 64 bit value
520 * R5    has shift count
521 * result in R3/R4
522 *
523 *  ashrdi3: arithmetic right shift (sign propagation)
524 *  lshrdi3: logical right shift
525 *  ashldi3: left shift
526 */
527_GLOBAL(__ashrdi3)
528	subfic	r6,r5,32
529	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
530	addi	r7,r5,32	# could be xori, or addi with -32
531	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
532	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
533	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
534	or	r4,r4,r6	# LSW |= t1
535	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
536	sraw	r3,r3,r5	# MSW = MSW >> count
537	or	r4,r4,r7	# LSW |= t2
538	blr
539
540_GLOBAL(__ashldi3)
541	subfic	r6,r5,32
542	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
543	addi	r7,r5,32	# could be xori, or addi with -32
544	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
545	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
546	or	r3,r3,r6	# MSW |= t1
547	slw	r4,r4,r5	# LSW = LSW << count
548	or	r3,r3,r7	# MSW |= t2
549	blr
550
551_GLOBAL(__lshrdi3)
552	subfic	r6,r5,32
553	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
554	addi	r7,r5,32	# could be xori, or addi with -32
555	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
556	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
557	or	r4,r4,r6	# LSW |= t1
558	srw	r3,r3,r5	# MSW = MSW >> count
559	or	r4,r4,r7	# LSW |= t2
560	blr
561
562/*
563 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
564 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
565 */
566_GLOBAL(__cmpdi2)
567	cmpw	r3,r5
568	li	r3,1
569	bne	1f
570	cmplw	r4,r6
571	beqlr
5721:	li	r3,0
573	bltlr
574	li	r3,2
575	blr
576/*
577 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
578 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
579 */
580_GLOBAL(__ucmpdi2)
581	cmplw	r3,r5
582	li	r3,1
583	bne	1f
584	cmplw	r4,r6
585	beqlr
5861:	li	r3,0
587	bltlr
588	li	r3,2
589	blr
590
591_GLOBAL(__bswapdi2)
592	rotlwi  r9,r4,8
593	rotlwi  r10,r3,8
594	rlwimi  r9,r4,24,0,7
595	rlwimi  r10,r3,24,0,7
596	rlwimi  r9,r4,24,16,23
597	rlwimi  r10,r3,24,16,23
598	mr      r3,r9
599	mr      r4,r10
600	blr
601
602_GLOBAL(abs)
603	srawi	r4,r3,31
604	xor	r3,r3,r4
605	sub	r3,r3,r4
606	blr
607
608#ifdef CONFIG_SMP
609_GLOBAL(start_secondary_resume)
610	/* Reset stack */
611	CURRENT_THREAD_INFO(r1, r1)
612	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
613	li	r3,0
614	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
615	bl	start_secondary
616	b	.
617#endif /* CONFIG_SMP */
618
619/*
620 * This routine is just here to keep GCC happy - sigh...
621 */
622_GLOBAL(__main)
623	blr
624
625#ifdef CONFIG_KEXEC
626	/*
627	 * Must be relocatable PIC code callable as a C function.
628	 */
629	.globl relocate_new_kernel
630relocate_new_kernel:
631	/* r3 = page_list   */
632	/* r4 = reboot_code_buffer */
633	/* r5 = start_address      */
634
635#ifdef CONFIG_FSL_BOOKE
636
637	mr	r29, r3
638	mr	r30, r4
639	mr	r31, r5
640
641#define ENTRY_MAPPING_KEXEC_SETUP
642#include "fsl_booke_entry_mapping.S"
643#undef ENTRY_MAPPING_KEXEC_SETUP
644
645	mr      r3, r29
646	mr      r4, r30
647	mr      r5, r31
648
649	li	r0, 0
650#elif defined(CONFIG_44x)
651
652	/* Save our parameters */
653	mr	r29, r3
654	mr	r30, r4
655	mr	r31, r5
656
657#ifdef CONFIG_PPC_47x
658	/* Check for 47x cores */
659	mfspr	r3,SPRN_PVR
660	srwi	r3,r3,16
661	cmplwi	cr0,r3,PVR_476FPE@h
662	beq	setup_map_47x
663	cmplwi	cr0,r3,PVR_476@h
664	beq	setup_map_47x
665	cmplwi	cr0,r3,PVR_476_ISS@h
666	beq	setup_map_47x
667#endif /* CONFIG_PPC_47x */
668
669/*
670 * Code for setting up 1:1 mapping for PPC440x for KEXEC
671 *
672 * We cannot switch off the MMU on PPC44x.
673 * So we:
674 * 1) Invalidate all the mappings except the one we are running from.
675 * 2) Create a tmp mapping for our code in the other address space(TS) and
676 *    jump to it. Invalidate the entry we started in.
677 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
678 * 4) Jump to the 1:1 mapping in original TS.
679 * 5) Invalidate the tmp mapping.
680 *
681 * - Based on the kexec support code for FSL BookE
682 *
683 */
684
685	/*
686	 * Load the PID with kernel PID (0).
687	 * Also load our MSR_IS and TID to MMUCR for TLB search.
688	 */
689	li	r3, 0
690	mtspr	SPRN_PID, r3
691	mfmsr	r4
692	andi.	r4,r4,MSR_IS@l
693	beq	wmmucr
694	oris	r3,r3,PPC44x_MMUCR_STS@h
695wmmucr:
696	mtspr	SPRN_MMUCR,r3
697	sync
698
699	/*
700	 * Invalidate all the TLB entries except the current entry
701	 * where we are running from
702	 */
703	bl	0f				/* Find our address */
7040:	mflr	r5				/* Make it accessible */
705	tlbsx	r23,0,r5			/* Find entry we are in */
706	li	r4,0				/* Start at TLB entry 0 */
707	li	r3,0				/* Set PAGEID inval value */
7081:	cmpw	r23,r4				/* Is this our entry? */
709	beq	skip				/* If so, skip the inval */
710	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
711skip:
712	addi	r4,r4,1				/* Increment */
713	cmpwi	r4,64				/* Are we done?	*/
714	bne	1b				/* If not, repeat */
715	isync
716
717	/* Create a temp mapping and jump to it */
718	andi.	r6, r23, 1		/* Find the index to use */
719	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
720
721	mfmsr	r9			/* get the MSR */
722	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
723	xori	r7, r5, 1		/* Use the other address space */
724
725	/* Read the current mapping entries */
726	tlbre	r3, r23, PPC44x_TLB_PAGEID
727	tlbre	r4, r23, PPC44x_TLB_XLAT
728	tlbre	r5, r23, PPC44x_TLB_ATTRIB
729
730	/* Save our current XLAT entry */
731	mr	r25, r4
732
733	/* Extract the TLB PageSize */
734	li	r10, 1 			/* r10 will hold PageSize */
735	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
736
737	/* XXX: As of now we use 256M, 4K pages */
738	cmpwi	r11, PPC44x_TLB_256M
739	bne	tlb_4k
740	rotlwi	r10, r10, 28		/* r10 = 256M */
741	b	write_out
742tlb_4k:
743	cmpwi	r11, PPC44x_TLB_4K
744	bne	default
745	rotlwi	r10, r10, 12		/* r10 = 4K */
746	b	write_out
747default:
748	rotlwi	r10, r10, 10		/* r10 = 1K */
749
750write_out:
751	/*
752	 * Write out the tmp 1:1 mapping for this code in other address space
753	 * Fixup  EPN = RPN , TS=other address space
754	 */
755	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
756
757	/* Write out the tmp mapping entries */
758	tlbwe	r3, r24, PPC44x_TLB_PAGEID
759	tlbwe	r4, r24, PPC44x_TLB_XLAT
760	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
761
762	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
763	not	r10, r11		/* Mask for PageNum */
764
765	/* Switch to other address space in MSR */
766	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
767
768	bl	1f
7691:	mflr	r8
770	addi	r8, r8, (2f-1b)		/* Find the target offset */
771
772	/* Jump to the tmp mapping */
773	mtspr	SPRN_SRR0, r8
774	mtspr	SPRN_SRR1, r9
775	rfi
776
7772:
778	/* Invalidate the entry we were executing from */
779	li	r3, 0
780	tlbwe	r3, r23, PPC44x_TLB_PAGEID
781
782	/* attribute fields. rwx for SUPERVISOR mode */
783	li	r5, 0
784	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
785
786	/* Create 1:1 mapping in 256M pages */
787	xori	r7, r7, 1			/* Revert back to Original TS */
788
789	li	r8, 0				/* PageNumber */
790	li	r6, 3				/* TLB Index, start at 3  */
791
792next_tlb:
793	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
794	mr	r4, r3				/* RPN = EPN  */
795	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
796	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
797
798	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
799	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
800	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
801
802	addi	r8, r8, 1			/* Increment PN */
803	addi	r6, r6, 1			/* Increment TLB Index */
804	cmpwi	r8, 8				/* Are we done ? */
805	bne	next_tlb
806	isync
807
808	/* Jump to the new mapping 1:1 */
809	li	r9,0
810	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
811
812	bl	1f
8131:	mflr	r8
814	and	r8, r8, r11			/* Get our offset within page */
815	addi	r8, r8, (2f-1b)
816
817	and	r5, r25, r10			/* Get our target PageNum */
818	or	r8, r8, r5			/* Target jump address */
819
820	mtspr	SPRN_SRR0, r8
821	mtspr	SPRN_SRR1, r9
822	rfi
8232:
824	/* Invalidate the tmp entry we used */
825	li	r3, 0
826	tlbwe	r3, r24, PPC44x_TLB_PAGEID
827	sync
828	b	ppc44x_map_done
829
830#ifdef CONFIG_PPC_47x
831
832	/* 1:1 mapping for 47x */
833
834setup_map_47x:
835
836	/*
837	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
838	 * Also set the MSR IS->MMUCR STS
839	 */
840	li	r3, 0
841	mtspr	SPRN_PID, r3			/* Set PID */
842	mfmsr	r4				/* Get MSR */
843	andi.	r4, r4, MSR_IS@l		/* TS=1? */
844	beq	1f				/* If not, leave STS=0 */
845	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
8461:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
847	sync
848
849	/* Find the entry we are running from */
850	bl	2f
8512:	mflr	r23
852	tlbsx	r23, 0, r23
853	tlbre	r24, r23, 0			/* TLB Word 0 */
854	tlbre	r25, r23, 1			/* TLB Word 1 */
855	tlbre	r26, r23, 2			/* TLB Word 2 */
856
857
858	/*
859	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
860	 * of 4k page size in all  4 ways (0-3 in r3).
861	 * This would invalidate the entire UTLB including the one we are
862	 * running from. However the shadow TLB entries would help us
863	 * to continue the execution, until we flush them (rfi/isync).
864	 */
865	addis	r3, 0, 0x8000			/* specify the way */
866	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
867	addi	r5, 0, 0
868	b	clear_utlb_entry
869
870	/* Align the loop to speed things up. from head_44x.S */
871	.align	6
872
873clear_utlb_entry:
874
875	tlbwe	r4, r3, 0
876	tlbwe	r5, r3, 1
877	tlbwe	r5, r3, 2
878	addis	r3, r3, 0x2000			/* Increment the way */
879	cmpwi	r3, 0
880	bne	clear_utlb_entry
881	addis	r3, 0, 0x8000
882	addis	r4, r4, 0x100			/* Increment the EPN */
883	cmpwi	r4, 0
884	bne	clear_utlb_entry
885
886	/* Create the entries in the other address space */
887	mfmsr	r5
888	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
889	xori	r7, r7, 1			/* r7 = !TS */
890
891	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
892
893	/*
894	 * write out the TLB entries for the tmp mapping
895	 * Use way '0' so that we could easily invalidate it later.
896	 */
897	lis	r3, 0x8000			/* Way '0' */
898
899	tlbwe	r24, r3, 0
900	tlbwe	r25, r3, 1
901	tlbwe	r26, r3, 2
902
903	/* Update the msr to the new TS */
904	insrwi	r5, r7, 1, 26
905
906	bl	1f
9071:	mflr	r6
908	addi	r6, r6, (2f-1b)
909
910	mtspr	SPRN_SRR0, r6
911	mtspr	SPRN_SRR1, r5
912	rfi
913
914	/*
915	 * Now we are in the tmp address space.
916	 * Create a 1:1 mapping for 0-2GiB in the original TS.
917	 */
9182:
919	li	r3, 0
920	li	r4, 0				/* TLB Word 0 */
921	li	r5, 0				/* TLB Word 1 */
922	li	r6, 0
923	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
924
925	li	r8, 0				/* PageIndex */
926
927	xori	r7, r7, 1			/* revert back to original TS */
928
929write_utlb:
930	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
931						/* ERPN = 0 as we don't use memory above 2G */
932
933	mr	r4, r5				/* EPN = RPN */
934	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
935	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
936
937	tlbwe	r4, r3, 0			/* Write out the entries */
938	tlbwe	r5, r3, 1
939	tlbwe	r6, r3, 2
940	addi	r8, r8, 1
941	cmpwi	r8, 8				/* Have we completed ? */
942	bne	write_utlb
943
944	/* make sure we complete the TLB write up */
945	isync
946
947	/*
948	 * Prepare to jump to the 1:1 mapping.
949	 * 1) Extract page size of the tmp mapping
950	 *    DSIZ = TLB_Word0[22:27]
951	 * 2) Calculate the physical address of the address
952	 *    to jump to.
953	 */
954	rlwinm	r10, r24, 0, 22, 27
955
956	cmpwi	r10, PPC47x_TLB0_4K
957	bne	0f
958	li	r10, 0x1000			/* r10 = 4k */
959	bl	1f
960
9610:
962	/* Defaults to 256M */
963	lis	r10, 0x1000
964
965	bl	1f
9661:	mflr	r4
967	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
968
969	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
970	not	r10, r11			/* Pagemask = ~(offsetmask) */
971
972	and	r5, r25, r10			/* Physical page */
973	and	r6, r4, r11			/* offset within the current page */
974
975	or	r5, r5, r6			/* Physical address for 2f */
976
977	/* Switch the TS in MSR to the original one */
978	mfmsr	r8
979	insrwi	r8, r7, 1, 26
980
981	mtspr	SPRN_SRR1, r8
982	mtspr	SPRN_SRR0, r5
983	rfi
984
9852:
986	/* Invalidate the tmp mapping */
987	lis	r3, 0x8000			/* Way '0' */
988
989	clrrwi	r24, r24, 12			/* Clear the valid bit */
990	tlbwe	r24, r3, 0
991	tlbwe	r25, r3, 1
992	tlbwe	r26, r3, 2
993
994	/* Make sure we complete the TLB write and flush the shadow TLB */
995	isync
996
997#endif
998
999ppc44x_map_done:
1000
1001
1002	/* Restore the parameters */
1003	mr	r3, r29
1004	mr	r4, r30
1005	mr	r5, r31
1006
1007	li	r0, 0
1008#else
1009	li	r0, 0
1010
1011	/*
1012	 * Set Machine Status Register to a known status,
1013	 * switch the MMU off and jump to 1: in a single step.
1014	 */
1015
1016	mr	r8, r0
1017	ori     r8, r8, MSR_RI|MSR_ME
1018	mtspr	SPRN_SRR1, r8
1019	addi	r8, r4, 1f - relocate_new_kernel
1020	mtspr	SPRN_SRR0, r8
1021	sync
1022	rfi
1023
10241:
1025#endif
1026	/* from this point address translation is turned off */
1027	/* and interrupts are disabled */
1028
1029	/* set a new stack at the bottom of our page... */
1030	/* (not really needed now) */
1031	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1032	stw	r0, 0(r1)
1033
1034	/* Do the copies */
1035	li	r6, 0 /* checksum */
1036	mr	r0, r3
1037	b	1f
1038
10390:	/* top, read another word for the indirection page */
1040	lwzu	r0, 4(r3)
1041
10421:
1043	/* is it a destination page? (r8) */
1044	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1045	beq	2f
1046
1047	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1048	b	0b
1049
10502:	/* is it an indirection page? (r3) */
1051	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1052	beq	2f
1053
1054	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1055	subi	r3, r3, 4
1056	b	0b
1057
10582:	/* are we done? */
1059	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1060	beq	2f
1061	b	3f
1062
10632:	/* is it a source page? (r9) */
1064	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1065	beq	0b
1066
1067	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1068
1069	li	r7, PAGE_SIZE / 4
1070	mtctr   r7
1071	subi    r9, r9, 4
1072	subi    r8, r8, 4
10739:
1074	lwzu    r0, 4(r9)  /* do the copy */
1075	xor	r6, r6, r0
1076	stwu    r0, 4(r8)
1077	dcbst	0, r8
1078	sync
1079	icbi	0, r8
1080	bdnz    9b
1081
1082	addi    r9, r9, 4
1083	addi    r8, r8, 4
1084	b	0b
1085
10863:
1087
1088	/* To be certain of avoiding problems with self-modifying code
1089	 * execute a serializing instruction here.
1090	 */
1091	isync
1092	sync
1093
1094	mfspr	r3, SPRN_PIR /* current core we are running on */
1095	mr	r4, r5 /* load physical address of chunk called */
1096
1097	/* jump to the entry point, usually the setup routine */
1098	mtlr	r5
1099	blrl
1100
11011:	b	1b
1102
1103relocate_new_kernel_end:
1104
1105	.globl relocate_new_kernel_size
1106relocate_new_kernel_size:
1107	.long relocate_new_kernel_end - relocate_new_kernel
1108#endif
1109