xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision 05bcf503)
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_GLOBAL(call_do_softirq)
40	mflr	r0
41	stw	r0,4(r1)
42	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
43	mr	r1,r3
44	bl	__do_softirq
45	lwz	r1,0(r1)
46	lwz	r0,4(r1)
47	mtlr	r0
48	blr
49
50_GLOBAL(call_handle_irq)
51	mflr	r0
52	stw	r0,4(r1)
53	mtctr	r6
54	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r5)
55	mr	r1,r5
56	bctrl
57	lwz	r1,0(r1)
58	lwz	r0,4(r1)
59	mtlr	r0
60	blr
61
62/*
63 * This returns the high 64 bits of the product of two 64-bit numbers.
64 */
65_GLOBAL(mulhdu)
66	cmpwi	r6,0
67	cmpwi	cr1,r3,0
68	mr	r10,r4
69	mulhwu	r4,r4,r5
70	beq	1f
71	mulhwu	r0,r10,r6
72	mullw	r7,r10,r5
73	addc	r7,r0,r7
74	addze	r4,r4
751:	beqlr	cr1		/* all done if high part of A is 0 */
76	mr	r10,r3
77	mullw	r9,r3,r5
78	mulhwu	r3,r3,r5
79	beq	2f
80	mullw	r0,r10,r6
81	mulhwu	r8,r10,r6
82	addc	r7,r0,r7
83	adde	r4,r4,r8
84	addze	r3,r3
852:	addc	r4,r4,r9
86	addze	r3,r3
87	blr
88
89/*
90 * sub_reloc_offset(x) returns x - reloc_offset().
91 */
92_GLOBAL(sub_reloc_offset)
93	mflr	r0
94	bl	1f
951:	mflr	r5
96	lis	r4,1b@ha
97	addi	r4,r4,1b@l
98	subf	r5,r4,r5
99	subf	r3,r5,r3
100	mtlr	r0
101	blr
102
103/*
104 * reloc_got2 runs through the .got2 section adding an offset
105 * to each entry.
106 */
107_GLOBAL(reloc_got2)
108	mflr	r11
109	lis	r7,__got2_start@ha
110	addi	r7,r7,__got2_start@l
111	lis	r8,__got2_end@ha
112	addi	r8,r8,__got2_end@l
113	subf	r8,r7,r8
114	srwi.	r8,r8,2
115	beqlr
116	mtctr	r8
117	bl	1f
1181:	mflr	r0
119	lis	r4,1b@ha
120	addi	r4,r4,1b@l
121	subf	r0,r4,r0
122	add	r7,r0,r7
1232:	lwz	r0,0(r7)
124	add	r0,r0,r3
125	stw	r0,0(r7)
126	addi	r7,r7,4
127	bdnz	2b
128	mtlr	r11
129	blr
130
131/*
132 * call_setup_cpu - call the setup_cpu function for this cpu
133 * r3 = data offset, r24 = cpu number
134 *
135 * Setup function is called with:
136 *   r3 = data offset
137 *   r4 = ptr to CPU spec (relocated)
138 */
139_GLOBAL(call_setup_cpu)
140	addis	r4,r3,cur_cpu_spec@ha
141	addi	r4,r4,cur_cpu_spec@l
142	lwz	r4,0(r4)
143	add	r4,r4,r3
144	lwz	r5,CPU_SPEC_SETUP(r4)
145	cmpwi	0,r5,0
146	add	r5,r5,r3
147	beqlr
148	mtctr	r5
149	bctr
150
151#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
152
153/* This gets called by via-pmu.c to switch the PLL selection
154 * on 750fx CPU. This function should really be moved to some
155 * other place (as most of the cpufreq code in via-pmu
156 */
157_GLOBAL(low_choose_750fx_pll)
158	/* Clear MSR:EE */
159	mfmsr	r7
160	rlwinm	r0,r7,0,17,15
161	mtmsr	r0
162
163	/* If switching to PLL1, disable HID0:BTIC */
164	cmplwi	cr0,r3,0
165	beq	1f
166	mfspr	r5,SPRN_HID0
167	rlwinm	r5,r5,0,27,25
168	sync
169	mtspr	SPRN_HID0,r5
170	isync
171	sync
172
1731:
174	/* Calc new HID1 value */
175	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
176	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
177	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
178	or	r4,r4,r5
179	mtspr	SPRN_HID1,r4
180
181	/* Store new HID1 image */
182	CURRENT_THREAD_INFO(r6, r1)
183	lwz	r6,TI_CPU(r6)
184	slwi	r6,r6,2
185	addis	r6,r6,nap_save_hid1@ha
186	stw	r4,nap_save_hid1@l(r6)
187
188	/* If switching to PLL0, enable HID0:BTIC */
189	cmplwi	cr0,r3,0
190	bne	1f
191	mfspr	r5,SPRN_HID0
192	ori	r5,r5,HID0_BTIC
193	sync
194	mtspr	SPRN_HID0,r5
195	isync
196	sync
197
1981:
199	/* Return */
200	mtmsr	r7
201	blr
202
203_GLOBAL(low_choose_7447a_dfs)
204	/* Clear MSR:EE */
205	mfmsr	r7
206	rlwinm	r0,r7,0,17,15
207	mtmsr	r0
208
209	/* Calc new HID1 value */
210	mfspr	r4,SPRN_HID1
211	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
212	sync
213	mtspr	SPRN_HID1,r4
214	sync
215	isync
216
217	/* Return */
218	mtmsr	r7
219	blr
220
221#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
222
223/*
224 * complement mask on the msr then "or" some values on.
225 *     _nmask_and_or_msr(nmask, value_to_or)
226 */
227_GLOBAL(_nmask_and_or_msr)
228	mfmsr	r0		/* Get current msr */
229	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
230	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
231	SYNC			/* Some chip revs have problems here... */
232	mtmsr	r0		/* Update machine state */
233	isync
234	blr			/* Done */
235
236#ifdef CONFIG_40x
237
238/*
239 * Do an IO access in real mode
240 */
241_GLOBAL(real_readb)
242	mfmsr	r7
243	ori	r0,r7,MSR_DR
244	xori	r0,r0,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	ori	r0,r7,MSR_DR
262	xori	r0,r0,MSR_DR
263	sync
264	mtmsr	r0
265	sync
266	isync
267	stb	r3,0(r4)
268	sync
269	mtmsr	r7
270	sync
271	isync
272	blr
273
274#endif /* CONFIG_40x */
275
276
277/*
278 * Flush instruction cache.
279 * This is a no-op on the 601.
280 */
281_GLOBAL(flush_instruction_cache)
282#if defined(CONFIG_8xx)
283	isync
284	lis	r5, IDC_INVALL@h
285	mtspr	SPRN_IC_CST, r5
286#elif 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 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_8xx/4xx */
320	isync
321	blr
322
323/*
324 * Write any modified data cache blocks out to memory
325 * and invalidate the corresponding instruction cache blocks.
326 * This is a no-op on the 601.
327 *
328 * flush_icache_range(unsigned long start, unsigned long stop)
329 */
330_KPROBE(__flush_icache_range)
331BEGIN_FTR_SECTION
332	blr				/* for 601, do nothing */
333END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
334	li	r5,L1_CACHE_BYTES-1
335	andc	r3,r3,r5
336	subf	r4,r3,r4
337	add	r4,r4,r5
338	srwi.	r4,r4,L1_CACHE_SHIFT
339	beqlr
340	mtctr	r4
341	mr	r6,r3
3421:	dcbst	0,r3
343	addi	r3,r3,L1_CACHE_BYTES
344	bdnz	1b
345	sync				/* wait for dcbst's to get to ram */
346#ifndef CONFIG_44x
347	mtctr	r4
3482:	icbi	0,r6
349	addi	r6,r6,L1_CACHE_BYTES
350	bdnz	2b
351#else
352	/* Flash invalidate on 44x because we are passed kmapped addresses and
353	   this doesn't work for userspace pages due to the virtually tagged
354	   icache.  Sigh. */
355	iccci	0, r0
356#endif
357	sync				/* additional sync needed on g4 */
358	isync
359	blr
360/*
361 * Write any modified data cache blocks out to memory.
362 * Does not invalidate the corresponding cache lines (especially for
363 * any corresponding instruction cache).
364 *
365 * clean_dcache_range(unsigned long start, unsigned long stop)
366 */
367_GLOBAL(clean_dcache_range)
368	li	r5,L1_CACHE_BYTES-1
369	andc	r3,r3,r5
370	subf	r4,r3,r4
371	add	r4,r4,r5
372	srwi.	r4,r4,L1_CACHE_SHIFT
373	beqlr
374	mtctr	r4
375
3761:	dcbst	0,r3
377	addi	r3,r3,L1_CACHE_BYTES
378	bdnz	1b
379	sync				/* wait for dcbst's to get to ram */
380	blr
381
382/*
383 * Write any modified data cache blocks out to memory and invalidate them.
384 * Does not invalidate the corresponding instruction cache blocks.
385 *
386 * flush_dcache_range(unsigned long start, unsigned long stop)
387 */
388_GLOBAL(flush_dcache_range)
389	li	r5,L1_CACHE_BYTES-1
390	andc	r3,r3,r5
391	subf	r4,r3,r4
392	add	r4,r4,r5
393	srwi.	r4,r4,L1_CACHE_SHIFT
394	beqlr
395	mtctr	r4
396
3971:	dcbf	0,r3
398	addi	r3,r3,L1_CACHE_BYTES
399	bdnz	1b
400	sync				/* wait for dcbst's to get to ram */
401	blr
402
403/*
404 * Like above, but invalidate the D-cache.  This is used by the 8xx
405 * to invalidate the cache so the PPC core doesn't get stale data
406 * from the CPM (no cache snooping here :-).
407 *
408 * invalidate_dcache_range(unsigned long start, unsigned long stop)
409 */
410_GLOBAL(invalidate_dcache_range)
411	li	r5,L1_CACHE_BYTES-1
412	andc	r3,r3,r5
413	subf	r4,r3,r4
414	add	r4,r4,r5
415	srwi.	r4,r4,L1_CACHE_SHIFT
416	beqlr
417	mtctr	r4
418
4191:	dcbi	0,r3
420	addi	r3,r3,L1_CACHE_BYTES
421	bdnz	1b
422	sync				/* wait for dcbi's to get to ram */
423	blr
424
425/*
426 * Flush a particular page from the data cache to RAM.
427 * Note: this is necessary because the instruction cache does *not*
428 * snoop from the data cache.
429 * This is a no-op on the 601 which has a unified cache.
430 *
431 *	void __flush_dcache_icache(void *page)
432 */
433_GLOBAL(__flush_dcache_icache)
434BEGIN_FTR_SECTION
435	blr
436END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
437	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
438	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
439	mtctr	r4
440	mr	r6,r3
4410:	dcbst	0,r3				/* Write line to ram */
442	addi	r3,r3,L1_CACHE_BYTES
443	bdnz	0b
444	sync
445#ifdef CONFIG_44x
446	/* We don't flush the icache on 44x. Those have a virtual icache
447	 * and we don't have access to the virtual address here (it's
448	 * not the page vaddr but where it's mapped in user space). The
449	 * flushing of the icache on these is handled elsewhere, when
450	 * a change in the address space occurs, before returning to
451	 * user space
452	 */
453BEGIN_MMU_FTR_SECTION
454	blr
455END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
456#endif /* CONFIG_44x */
457	mtctr	r4
4581:	icbi	0,r6
459	addi	r6,r6,L1_CACHE_BYTES
460	bdnz	1b
461	sync
462	isync
463	blr
464
465#ifndef CONFIG_BOOKE
466/*
467 * Flush a particular page from the data cache to RAM, identified
468 * by its physical address.  We turn off the MMU so we can just use
469 * the physical address (this may be a highmem page without a kernel
470 * mapping).
471 *
472 *	void __flush_dcache_icache_phys(unsigned long physaddr)
473 */
474_GLOBAL(__flush_dcache_icache_phys)
475BEGIN_FTR_SECTION
476	blr					/* for 601, do nothing */
477END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
478	mfmsr	r10
479	rlwinm	r0,r10,0,28,26			/* clear DR */
480	mtmsr	r0
481	isync
482	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
483	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
484	mtctr	r4
485	mr	r6,r3
4860:	dcbst	0,r3				/* Write line to ram */
487	addi	r3,r3,L1_CACHE_BYTES
488	bdnz	0b
489	sync
490	mtctr	r4
4911:	icbi	0,r6
492	addi	r6,r6,L1_CACHE_BYTES
493	bdnz	1b
494	sync
495	mtmsr	r10				/* restore DR */
496	isync
497	blr
498#endif /* CONFIG_BOOKE */
499
500/*
501 * Clear pages using the dcbz instruction, which doesn't cause any
502 * memory traffic (except to write out any cache lines which get
503 * displaced).  This only works on cacheable memory.
504 *
505 * void clear_pages(void *page, int order) ;
506 */
507_GLOBAL(clear_pages)
508	li	r0,PAGE_SIZE/L1_CACHE_BYTES
509	slw	r0,r0,r4
510	mtctr	r0
5111:	dcbz	0,r3
512	addi	r3,r3,L1_CACHE_BYTES
513	bdnz	1b
514	blr
515
516/*
517 * Copy a whole page.  We use the dcbz instruction on the destination
518 * to reduce memory traffic (it eliminates the unnecessary reads of
519 * the destination into cache).  This requires that the destination
520 * is cacheable.
521 */
522#define COPY_16_BYTES		\
523	lwz	r6,4(r4);	\
524	lwz	r7,8(r4);	\
525	lwz	r8,12(r4);	\
526	lwzu	r9,16(r4);	\
527	stw	r6,4(r3);	\
528	stw	r7,8(r3);	\
529	stw	r8,12(r3);	\
530	stwu	r9,16(r3)
531
532_GLOBAL(copy_page)
533	addi	r3,r3,-4
534	addi	r4,r4,-4
535
536	li	r5,4
537
538#if MAX_COPY_PREFETCH > 1
539	li	r0,MAX_COPY_PREFETCH
540	li	r11,4
541	mtctr	r0
54211:	dcbt	r11,r4
543	addi	r11,r11,L1_CACHE_BYTES
544	bdnz	11b
545#else /* MAX_COPY_PREFETCH == 1 */
546	dcbt	r5,r4
547	li	r11,L1_CACHE_BYTES+4
548#endif /* MAX_COPY_PREFETCH */
549	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
550	crclr	4*cr0+eq
5512:
552	mtctr	r0
5531:
554	dcbt	r11,r4
555	dcbz	r5,r3
556	COPY_16_BYTES
557#if L1_CACHE_BYTES >= 32
558	COPY_16_BYTES
559#if L1_CACHE_BYTES >= 64
560	COPY_16_BYTES
561	COPY_16_BYTES
562#if L1_CACHE_BYTES >= 128
563	COPY_16_BYTES
564	COPY_16_BYTES
565	COPY_16_BYTES
566	COPY_16_BYTES
567#endif
568#endif
569#endif
570	bdnz	1b
571	beqlr
572	crnot	4*cr0+eq,4*cr0+eq
573	li	r0,MAX_COPY_PREFETCH
574	li	r11,4
575	b	2b
576
577/*
578 * void atomic_clear_mask(atomic_t mask, atomic_t *addr)
579 * void atomic_set_mask(atomic_t mask, atomic_t *addr);
580 */
581_GLOBAL(atomic_clear_mask)
58210:	lwarx	r5,0,r4
583	andc	r5,r5,r3
584	PPC405_ERR77(0,r4)
585	stwcx.	r5,0,r4
586	bne-	10b
587	blr
588_GLOBAL(atomic_set_mask)
58910:	lwarx	r5,0,r4
590	or	r5,r5,r3
591	PPC405_ERR77(0,r4)
592	stwcx.	r5,0,r4
593	bne-	10b
594	blr
595
596/*
597 * Extended precision shifts.
598 *
599 * Updated to be valid for shift counts from 0 to 63 inclusive.
600 * -- Gabriel
601 *
602 * R3/R4 has 64 bit value
603 * R5    has shift count
604 * result in R3/R4
605 *
606 *  ashrdi3: arithmetic right shift (sign propagation)
607 *  lshrdi3: logical right shift
608 *  ashldi3: left shift
609 */
610_GLOBAL(__ashrdi3)
611	subfic	r6,r5,32
612	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
613	addi	r7,r5,32	# could be xori, or addi with -32
614	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
615	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
616	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
617	or	r4,r4,r6	# LSW |= t1
618	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
619	sraw	r3,r3,r5	# MSW = MSW >> count
620	or	r4,r4,r7	# LSW |= t2
621	blr
622
623_GLOBAL(__ashldi3)
624	subfic	r6,r5,32
625	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
626	addi	r7,r5,32	# could be xori, or addi with -32
627	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
628	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
629	or	r3,r3,r6	# MSW |= t1
630	slw	r4,r4,r5	# LSW = LSW << count
631	or	r3,r3,r7	# MSW |= t2
632	blr
633
634_GLOBAL(__lshrdi3)
635	subfic	r6,r5,32
636	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
637	addi	r7,r5,32	# could be xori, or addi with -32
638	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
639	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
640	or	r4,r4,r6	# LSW |= t1
641	srw	r3,r3,r5	# MSW = MSW >> count
642	or	r4,r4,r7	# LSW |= t2
643	blr
644
645/*
646 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
647 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
648 */
649_GLOBAL(__ucmpdi2)
650	cmplw	r3,r5
651	li	r3,1
652	bne	1f
653	cmplw	r4,r6
654	beqlr
6551:	li	r3,0
656	bltlr
657	li	r3,2
658	blr
659
660_GLOBAL(abs)
661	srawi	r4,r3,31
662	xor	r3,r3,r4
663	sub	r3,r3,r4
664	blr
665
666#ifdef CONFIG_SMP
667_GLOBAL(start_secondary_resume)
668	/* Reset stack */
669	CURRENT_THREAD_INFO(r1, r1)
670	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
671	li	r3,0
672	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
673	bl	start_secondary
674	b	.
675#endif /* CONFIG_SMP */
676
677/*
678 * This routine is just here to keep GCC happy - sigh...
679 */
680_GLOBAL(__main)
681	blr
682
683#ifdef CONFIG_KEXEC
684	/*
685	 * Must be relocatable PIC code callable as a C function.
686	 */
687	.globl relocate_new_kernel
688relocate_new_kernel:
689	/* r3 = page_list   */
690	/* r4 = reboot_code_buffer */
691	/* r5 = start_address      */
692
693#ifdef CONFIG_FSL_BOOKE
694
695	mr	r29, r3
696	mr	r30, r4
697	mr	r31, r5
698
699#define ENTRY_MAPPING_KEXEC_SETUP
700#include "fsl_booke_entry_mapping.S"
701#undef ENTRY_MAPPING_KEXEC_SETUP
702
703	mr      r3, r29
704	mr      r4, r30
705	mr      r5, r31
706
707	li	r0, 0
708#elif defined(CONFIG_44x)
709
710	/* Save our parameters */
711	mr	r29, r3
712	mr	r30, r4
713	mr	r31, r5
714
715#ifdef CONFIG_PPC_47x
716	/* Check for 47x cores */
717	mfspr	r3,SPRN_PVR
718	srwi	r3,r3,16
719	cmplwi	cr0,r3,PVR_476@h
720	beq	setup_map_47x
721	cmplwi	cr0,r3,PVR_476_ISS@h
722	beq	setup_map_47x
723#endif /* CONFIG_PPC_47x */
724
725/*
726 * Code for setting up 1:1 mapping for PPC440x for KEXEC
727 *
728 * We cannot switch off the MMU on PPC44x.
729 * So we:
730 * 1) Invalidate all the mappings except the one we are running from.
731 * 2) Create a tmp mapping for our code in the other address space(TS) and
732 *    jump to it. Invalidate the entry we started in.
733 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
734 * 4) Jump to the 1:1 mapping in original TS.
735 * 5) Invalidate the tmp mapping.
736 *
737 * - Based on the kexec support code for FSL BookE
738 *
739 */
740
741	/*
742	 * Load the PID with kernel PID (0).
743	 * Also load our MSR_IS and TID to MMUCR for TLB search.
744	 */
745	li	r3, 0
746	mtspr	SPRN_PID, r3
747	mfmsr	r4
748	andi.	r4,r4,MSR_IS@l
749	beq	wmmucr
750	oris	r3,r3,PPC44x_MMUCR_STS@h
751wmmucr:
752	mtspr	SPRN_MMUCR,r3
753	sync
754
755	/*
756	 * Invalidate all the TLB entries except the current entry
757	 * where we are running from
758	 */
759	bl	0f				/* Find our address */
7600:	mflr	r5				/* Make it accessible */
761	tlbsx	r23,0,r5			/* Find entry we are in */
762	li	r4,0				/* Start at TLB entry 0 */
763	li	r3,0				/* Set PAGEID inval value */
7641:	cmpw	r23,r4				/* Is this our entry? */
765	beq	skip				/* If so, skip the inval */
766	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
767skip:
768	addi	r4,r4,1				/* Increment */
769	cmpwi	r4,64				/* Are we done?	*/
770	bne	1b				/* If not, repeat */
771	isync
772
773	/* Create a temp mapping and jump to it */
774	andi.	r6, r23, 1		/* Find the index to use */
775	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
776
777	mfmsr	r9			/* get the MSR */
778	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
779	xori	r7, r5, 1		/* Use the other address space */
780
781	/* Read the current mapping entries */
782	tlbre	r3, r23, PPC44x_TLB_PAGEID
783	tlbre	r4, r23, PPC44x_TLB_XLAT
784	tlbre	r5, r23, PPC44x_TLB_ATTRIB
785
786	/* Save our current XLAT entry */
787	mr	r25, r4
788
789	/* Extract the TLB PageSize */
790	li	r10, 1 			/* r10 will hold PageSize */
791	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
792
793	/* XXX: As of now we use 256M, 4K pages */
794	cmpwi	r11, PPC44x_TLB_256M
795	bne	tlb_4k
796	rotlwi	r10, r10, 28		/* r10 = 256M */
797	b	write_out
798tlb_4k:
799	cmpwi	r11, PPC44x_TLB_4K
800	bne	default
801	rotlwi	r10, r10, 12		/* r10 = 4K */
802	b	write_out
803default:
804	rotlwi	r10, r10, 10		/* r10 = 1K */
805
806write_out:
807	/*
808	 * Write out the tmp 1:1 mapping for this code in other address space
809	 * Fixup  EPN = RPN , TS=other address space
810	 */
811	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
812
813	/* Write out the tmp mapping entries */
814	tlbwe	r3, r24, PPC44x_TLB_PAGEID
815	tlbwe	r4, r24, PPC44x_TLB_XLAT
816	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
817
818	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
819	not	r10, r11		/* Mask for PageNum */
820
821	/* Switch to other address space in MSR */
822	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
823
824	bl	1f
8251:	mflr	r8
826	addi	r8, r8, (2f-1b)		/* Find the target offset */
827
828	/* Jump to the tmp mapping */
829	mtspr	SPRN_SRR0, r8
830	mtspr	SPRN_SRR1, r9
831	rfi
832
8332:
834	/* Invalidate the entry we were executing from */
835	li	r3, 0
836	tlbwe	r3, r23, PPC44x_TLB_PAGEID
837
838	/* attribute fields. rwx for SUPERVISOR mode */
839	li	r5, 0
840	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
841
842	/* Create 1:1 mapping in 256M pages */
843	xori	r7, r7, 1			/* Revert back to Original TS */
844
845	li	r8, 0				/* PageNumber */
846	li	r6, 3				/* TLB Index, start at 3  */
847
848next_tlb:
849	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
850	mr	r4, r3				/* RPN = EPN  */
851	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
852	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
853
854	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
855	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
856	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
857
858	addi	r8, r8, 1			/* Increment PN */
859	addi	r6, r6, 1			/* Increment TLB Index */
860	cmpwi	r8, 8				/* Are we done ? */
861	bne	next_tlb
862	isync
863
864	/* Jump to the new mapping 1:1 */
865	li	r9,0
866	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
867
868	bl	1f
8691:	mflr	r8
870	and	r8, r8, r11			/* Get our offset within page */
871	addi	r8, r8, (2f-1b)
872
873	and	r5, r25, r10			/* Get our target PageNum */
874	or	r8, r8, r5			/* Target jump address */
875
876	mtspr	SPRN_SRR0, r8
877	mtspr	SPRN_SRR1, r9
878	rfi
8792:
880	/* Invalidate the tmp entry we used */
881	li	r3, 0
882	tlbwe	r3, r24, PPC44x_TLB_PAGEID
883	sync
884	b	ppc44x_map_done
885
886#ifdef CONFIG_PPC_47x
887
888	/* 1:1 mapping for 47x */
889
890setup_map_47x:
891
892	/*
893	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
894	 * Also set the MSR IS->MMUCR STS
895	 */
896	li	r3, 0
897	mtspr	SPRN_PID, r3			/* Set PID */
898	mfmsr	r4				/* Get MSR */
899	andi.	r4, r4, MSR_IS@l		/* TS=1? */
900	beq	1f				/* If not, leave STS=0 */
901	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
9021:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
903	sync
904
905	/* Find the entry we are running from */
906	bl	2f
9072:	mflr	r23
908	tlbsx	r23, 0, r23
909	tlbre	r24, r23, 0			/* TLB Word 0 */
910	tlbre	r25, r23, 1			/* TLB Word 1 */
911	tlbre	r26, r23, 2			/* TLB Word 2 */
912
913
914	/*
915	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
916	 * of 4k page size in all  4 ways (0-3 in r3).
917	 * This would invalidate the entire UTLB including the one we are
918	 * running from. However the shadow TLB entries would help us
919	 * to continue the execution, until we flush them (rfi/isync).
920	 */
921	addis	r3, 0, 0x8000			/* specify the way */
922	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
923	addi	r5, 0, 0
924	b	clear_utlb_entry
925
926	/* Align the loop to speed things up. from head_44x.S */
927	.align	6
928
929clear_utlb_entry:
930
931	tlbwe	r4, r3, 0
932	tlbwe	r5, r3, 1
933	tlbwe	r5, r3, 2
934	addis	r3, r3, 0x2000			/* Increment the way */
935	cmpwi	r3, 0
936	bne	clear_utlb_entry
937	addis	r3, 0, 0x8000
938	addis	r4, r4, 0x100			/* Increment the EPN */
939	cmpwi	r4, 0
940	bne	clear_utlb_entry
941
942	/* Create the entries in the other address space */
943	mfmsr	r5
944	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
945	xori	r7, r7, 1			/* r7 = !TS */
946
947	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
948
949	/*
950	 * write out the TLB entries for the tmp mapping
951	 * Use way '0' so that we could easily invalidate it later.
952	 */
953	lis	r3, 0x8000			/* Way '0' */
954
955	tlbwe	r24, r3, 0
956	tlbwe	r25, r3, 1
957	tlbwe	r26, r3, 2
958
959	/* Update the msr to the new TS */
960	insrwi	r5, r7, 1, 26
961
962	bl	1f
9631:	mflr	r6
964	addi	r6, r6, (2f-1b)
965
966	mtspr	SPRN_SRR0, r6
967	mtspr	SPRN_SRR1, r5
968	rfi
969
970	/*
971	 * Now we are in the tmp address space.
972	 * Create a 1:1 mapping for 0-2GiB in the original TS.
973	 */
9742:
975	li	r3, 0
976	li	r4, 0				/* TLB Word 0 */
977	li	r5, 0				/* TLB Word 1 */
978	li	r6, 0
979	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
980
981	li	r8, 0				/* PageIndex */
982
983	xori	r7, r7, 1			/* revert back to original TS */
984
985write_utlb:
986	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
987						/* ERPN = 0 as we don't use memory above 2G */
988
989	mr	r4, r5				/* EPN = RPN */
990	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
991	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
992
993	tlbwe	r4, r3, 0			/* Write out the entries */
994	tlbwe	r5, r3, 1
995	tlbwe	r6, r3, 2
996	addi	r8, r8, 1
997	cmpwi	r8, 8				/* Have we completed ? */
998	bne	write_utlb
999
1000	/* make sure we complete the TLB write up */
1001	isync
1002
1003	/*
1004	 * Prepare to jump to the 1:1 mapping.
1005	 * 1) Extract page size of the tmp mapping
1006	 *    DSIZ = TLB_Word0[22:27]
1007	 * 2) Calculate the physical address of the address
1008	 *    to jump to.
1009	 */
1010	rlwinm	r10, r24, 0, 22, 27
1011
1012	cmpwi	r10, PPC47x_TLB0_4K
1013	bne	0f
1014	li	r10, 0x1000			/* r10 = 4k */
1015	bl	1f
1016
10170:
1018	/* Defaults to 256M */
1019	lis	r10, 0x1000
1020
1021	bl	1f
10221:	mflr	r4
1023	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
1024
1025	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
1026	not	r10, r11			/* Pagemask = ~(offsetmask) */
1027
1028	and	r5, r25, r10			/* Physical page */
1029	and	r6, r4, r11			/* offset within the current page */
1030
1031	or	r5, r5, r6			/* Physical address for 2f */
1032
1033	/* Switch the TS in MSR to the original one */
1034	mfmsr	r8
1035	insrwi	r8, r7, 1, 26
1036
1037	mtspr	SPRN_SRR1, r8
1038	mtspr	SPRN_SRR0, r5
1039	rfi
1040
10412:
1042	/* Invalidate the tmp mapping */
1043	lis	r3, 0x8000			/* Way '0' */
1044
1045	clrrwi	r24, r24, 12			/* Clear the valid bit */
1046	tlbwe	r24, r3, 0
1047	tlbwe	r25, r3, 1
1048	tlbwe	r26, r3, 2
1049
1050	/* Make sure we complete the TLB write and flush the shadow TLB */
1051	isync
1052
1053#endif
1054
1055ppc44x_map_done:
1056
1057
1058	/* Restore the parameters */
1059	mr	r3, r29
1060	mr	r4, r30
1061	mr	r5, r31
1062
1063	li	r0, 0
1064#else
1065	li	r0, 0
1066
1067	/*
1068	 * Set Machine Status Register to a known status,
1069	 * switch the MMU off and jump to 1: in a single step.
1070	 */
1071
1072	mr	r8, r0
1073	ori     r8, r8, MSR_RI|MSR_ME
1074	mtspr	SPRN_SRR1, r8
1075	addi	r8, r4, 1f - relocate_new_kernel
1076	mtspr	SPRN_SRR0, r8
1077	sync
1078	rfi
1079
10801:
1081#endif
1082	/* from this point address translation is turned off */
1083	/* and interrupts are disabled */
1084
1085	/* set a new stack at the bottom of our page... */
1086	/* (not really needed now) */
1087	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1088	stw	r0, 0(r1)
1089
1090	/* Do the copies */
1091	li	r6, 0 /* checksum */
1092	mr	r0, r3
1093	b	1f
1094
10950:	/* top, read another word for the indirection page */
1096	lwzu	r0, 4(r3)
1097
10981:
1099	/* is it a destination page? (r8) */
1100	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1101	beq	2f
1102
1103	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1104	b	0b
1105
11062:	/* is it an indirection page? (r3) */
1107	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1108	beq	2f
1109
1110	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1111	subi	r3, r3, 4
1112	b	0b
1113
11142:	/* are we done? */
1115	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1116	beq	2f
1117	b	3f
1118
11192:	/* is it a source page? (r9) */
1120	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1121	beq	0b
1122
1123	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1124
1125	li	r7, PAGE_SIZE / 4
1126	mtctr   r7
1127	subi    r9, r9, 4
1128	subi    r8, r8, 4
11299:
1130	lwzu    r0, 4(r9)  /* do the copy */
1131	xor	r6, r6, r0
1132	stwu    r0, 4(r8)
1133	dcbst	0, r8
1134	sync
1135	icbi	0, r8
1136	bdnz    9b
1137
1138	addi    r9, r9, 4
1139	addi    r8, r8, 4
1140	b	0b
1141
11423:
1143
1144	/* To be certain of avoiding problems with self-modifying code
1145	 * execute a serializing instruction here.
1146	 */
1147	isync
1148	sync
1149
1150	mfspr	r3, SPRN_PIR /* current core we are running on */
1151	mr	r4, r5 /* load physical address of chunk called */
1152
1153	/* jump to the entry point, usually the setup routine */
1154	mtlr	r5
1155	blrl
1156
11571:	b	1b
1158
1159relocate_new_kernel_end:
1160
1161	.globl relocate_new_kernel_size
1162relocate_new_kernel_size:
1163	.long relocate_new_kernel_end - relocate_new_kernel
1164#endif
1165