xref: /linux/arch/powerpc/kernel/misc_32.S (revision b6ebbac51bedf9e98e837688bc838f400196da5e)
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 * reloc_got2 runs through the .got2 section adding an offset
108 * to each entry.
109 */
110_GLOBAL(reloc_got2)
111	mflr	r11
112	lis	r7,__got2_start@ha
113	addi	r7,r7,__got2_start@l
114	lis	r8,__got2_end@ha
115	addi	r8,r8,__got2_end@l
116	subf	r8,r7,r8
117	srwi.	r8,r8,2
118	beqlr
119	mtctr	r8
120	bl	1f
1211:	mflr	r0
122	lis	r4,1b@ha
123	addi	r4,r4,1b@l
124	subf	r0,r4,r0
125	add	r7,r0,r7
1262:	lwz	r0,0(r7)
127	add	r0,r0,r3
128	stw	r0,0(r7)
129	addi	r7,r7,4
130	bdnz	2b
131	mtlr	r11
132	blr
133
134/*
135 * call_setup_cpu - call the setup_cpu function for this cpu
136 * r3 = data offset, r24 = cpu number
137 *
138 * Setup function is called with:
139 *   r3 = data offset
140 *   r4 = ptr to CPU spec (relocated)
141 */
142_GLOBAL(call_setup_cpu)
143	addis	r4,r3,cur_cpu_spec@ha
144	addi	r4,r4,cur_cpu_spec@l
145	lwz	r4,0(r4)
146	add	r4,r4,r3
147	lwz	r5,CPU_SPEC_SETUP(r4)
148	cmpwi	0,r5,0
149	add	r5,r5,r3
150	beqlr
151	mtctr	r5
152	bctr
153
154#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
155
156/* This gets called by via-pmu.c to switch the PLL selection
157 * on 750fx CPU. This function should really be moved to some
158 * other place (as most of the cpufreq code in via-pmu
159 */
160_GLOBAL(low_choose_750fx_pll)
161	/* Clear MSR:EE */
162	mfmsr	r7
163	rlwinm	r0,r7,0,17,15
164	mtmsr	r0
165
166	/* If switching to PLL1, disable HID0:BTIC */
167	cmplwi	cr0,r3,0
168	beq	1f
169	mfspr	r5,SPRN_HID0
170	rlwinm	r5,r5,0,27,25
171	sync
172	mtspr	SPRN_HID0,r5
173	isync
174	sync
175
1761:
177	/* Calc new HID1 value */
178	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
179	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
180	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
181	or	r4,r4,r5
182	mtspr	SPRN_HID1,r4
183
184	/* Store new HID1 image */
185	CURRENT_THREAD_INFO(r6, r1)
186	lwz	r6,TI_CPU(r6)
187	slwi	r6,r6,2
188	addis	r6,r6,nap_save_hid1@ha
189	stw	r4,nap_save_hid1@l(r6)
190
191	/* If switching to PLL0, enable HID0:BTIC */
192	cmplwi	cr0,r3,0
193	bne	1f
194	mfspr	r5,SPRN_HID0
195	ori	r5,r5,HID0_BTIC
196	sync
197	mtspr	SPRN_HID0,r5
198	isync
199	sync
200
2011:
202	/* Return */
203	mtmsr	r7
204	blr
205
206_GLOBAL(low_choose_7447a_dfs)
207	/* Clear MSR:EE */
208	mfmsr	r7
209	rlwinm	r0,r7,0,17,15
210	mtmsr	r0
211
212	/* Calc new HID1 value */
213	mfspr	r4,SPRN_HID1
214	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
215	sync
216	mtspr	SPRN_HID1,r4
217	sync
218	isync
219
220	/* Return */
221	mtmsr	r7
222	blr
223
224#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
225
226/*
227 * complement mask on the msr then "or" some values on.
228 *     _nmask_and_or_msr(nmask, value_to_or)
229 */
230_GLOBAL(_nmask_and_or_msr)
231	mfmsr	r0		/* Get current msr */
232	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
233	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
234	SYNC			/* Some chip revs have problems here... */
235	mtmsr	r0		/* Update machine state */
236	isync
237	blr			/* Done */
238
239#ifdef CONFIG_40x
240
241/*
242 * Do an IO access in real mode
243 */
244_GLOBAL(real_readb)
245	mfmsr	r7
246	ori	r0,r7,MSR_DR
247	xori	r0,r0,MSR_DR
248	sync
249	mtmsr	r0
250	sync
251	isync
252	lbz	r3,0(r3)
253	sync
254	mtmsr	r7
255	sync
256	isync
257	blr
258
259	/*
260 * Do an IO access in real mode
261 */
262_GLOBAL(real_writeb)
263	mfmsr	r7
264	ori	r0,r7,MSR_DR
265	xori	r0,r0,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 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
322#endif /* CONFIG_PPC_8xx */
323
324/*
325 * Write any modified data cache blocks out to memory
326 * and invalidate the corresponding instruction cache blocks.
327 * This is a no-op on the 601.
328 *
329 * flush_icache_range(unsigned long start, unsigned long stop)
330 */
331_KPROBE(flush_icache_range)
332BEGIN_FTR_SECTION
333	PURGE_PREFETCHED_INS
334	blr				/* for 601, do nothing */
335END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
336	rlwinm	r3,r3,0,0,31 - L1_CACHE_SHIFT
337	subf	r4,r3,r4
338	addi	r4,r4,L1_CACHE_BYTES - 1
339	srwi.	r4,r4,L1_CACHE_SHIFT
340	beqlr
341	mtctr	r4
342	mr	r6,r3
3431:	dcbst	0,r3
344	addi	r3,r3,L1_CACHE_BYTES
345	bdnz	1b
346	sync				/* wait for dcbst's to get to ram */
347#ifndef CONFIG_44x
348	mtctr	r4
3492:	icbi	0,r6
350	addi	r6,r6,L1_CACHE_BYTES
351	bdnz	2b
352#else
353	/* Flash invalidate on 44x because we are passed kmapped addresses and
354	   this doesn't work for userspace pages due to the virtually tagged
355	   icache.  Sigh. */
356	iccci	0, r0
357#endif
358	sync				/* additional sync needed on g4 */
359	isync
360	blr
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
498
499/*
500 * Extended precision shifts.
501 *
502 * Updated to be valid for shift counts from 0 to 63 inclusive.
503 * -- Gabriel
504 *
505 * R3/R4 has 64 bit value
506 * R5    has shift count
507 * result in R3/R4
508 *
509 *  ashrdi3: arithmetic right shift (sign propagation)
510 *  lshrdi3: logical right shift
511 *  ashldi3: left shift
512 */
513_GLOBAL(__ashrdi3)
514	subfic	r6,r5,32
515	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
516	addi	r7,r5,32	# could be xori, or addi with -32
517	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
518	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
519	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
520	or	r4,r4,r6	# LSW |= t1
521	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
522	sraw	r3,r3,r5	# MSW = MSW >> count
523	or	r4,r4,r7	# LSW |= t2
524	blr
525
526_GLOBAL(__ashldi3)
527	subfic	r6,r5,32
528	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
529	addi	r7,r5,32	# could be xori, or addi with -32
530	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
531	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
532	or	r3,r3,r6	# MSW |= t1
533	slw	r4,r4,r5	# LSW = LSW << count
534	or	r3,r3,r7	# MSW |= t2
535	blr
536
537_GLOBAL(__lshrdi3)
538	subfic	r6,r5,32
539	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
540	addi	r7,r5,32	# could be xori, or addi with -32
541	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
542	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
543	or	r4,r4,r6	# LSW |= t1
544	srw	r3,r3,r5	# MSW = MSW >> count
545	or	r4,r4,r7	# LSW |= t2
546	blr
547
548/*
549 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
550 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
551 */
552_GLOBAL(__cmpdi2)
553	cmpw	r3,r5
554	li	r3,1
555	bne	1f
556	cmplw	r4,r6
557	beqlr
5581:	li	r3,0
559	bltlr
560	li	r3,2
561	blr
562/*
563 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
564 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
565 */
566_GLOBAL(__ucmpdi2)
567	cmplw	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_GLOBAL(__bswapdi2)
578	rotlwi  r9,r4,8
579	rotlwi  r10,r3,8
580	rlwimi  r9,r4,24,0,7
581	rlwimi  r10,r3,24,0,7
582	rlwimi  r9,r4,24,16,23
583	rlwimi  r10,r3,24,16,23
584	mr      r3,r9
585	mr      r4,r10
586	blr
587
588#ifdef CONFIG_SMP
589_GLOBAL(start_secondary_resume)
590	/* Reset stack */
591	CURRENT_THREAD_INFO(r1, r1)
592	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
593	li	r3,0
594	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
595	bl	start_secondary
596	b	.
597#endif /* CONFIG_SMP */
598
599/*
600 * This routine is just here to keep GCC happy - sigh...
601 */
602_GLOBAL(__main)
603	blr
604
605#ifdef CONFIG_KEXEC
606	/*
607	 * Must be relocatable PIC code callable as a C function.
608	 */
609	.globl relocate_new_kernel
610relocate_new_kernel:
611	/* r3 = page_list   */
612	/* r4 = reboot_code_buffer */
613	/* r5 = start_address      */
614
615#ifdef CONFIG_FSL_BOOKE
616
617	mr	r29, r3
618	mr	r30, r4
619	mr	r31, r5
620
621#define ENTRY_MAPPING_KEXEC_SETUP
622#include "fsl_booke_entry_mapping.S"
623#undef ENTRY_MAPPING_KEXEC_SETUP
624
625	mr      r3, r29
626	mr      r4, r30
627	mr      r5, r31
628
629	li	r0, 0
630#elif defined(CONFIG_44x)
631
632	/* Save our parameters */
633	mr	r29, r3
634	mr	r30, r4
635	mr	r31, r5
636
637#ifdef CONFIG_PPC_47x
638	/* Check for 47x cores */
639	mfspr	r3,SPRN_PVR
640	srwi	r3,r3,16
641	cmplwi	cr0,r3,PVR_476FPE@h
642	beq	setup_map_47x
643	cmplwi	cr0,r3,PVR_476@h
644	beq	setup_map_47x
645	cmplwi	cr0,r3,PVR_476_ISS@h
646	beq	setup_map_47x
647#endif /* CONFIG_PPC_47x */
648
649/*
650 * Code for setting up 1:1 mapping for PPC440x for KEXEC
651 *
652 * We cannot switch off the MMU on PPC44x.
653 * So we:
654 * 1) Invalidate all the mappings except the one we are running from.
655 * 2) Create a tmp mapping for our code in the other address space(TS) and
656 *    jump to it. Invalidate the entry we started in.
657 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
658 * 4) Jump to the 1:1 mapping in original TS.
659 * 5) Invalidate the tmp mapping.
660 *
661 * - Based on the kexec support code for FSL BookE
662 *
663 */
664
665	/*
666	 * Load the PID with kernel PID (0).
667	 * Also load our MSR_IS and TID to MMUCR for TLB search.
668	 */
669	li	r3, 0
670	mtspr	SPRN_PID, r3
671	mfmsr	r4
672	andi.	r4,r4,MSR_IS@l
673	beq	wmmucr
674	oris	r3,r3,PPC44x_MMUCR_STS@h
675wmmucr:
676	mtspr	SPRN_MMUCR,r3
677	sync
678
679	/*
680	 * Invalidate all the TLB entries except the current entry
681	 * where we are running from
682	 */
683	bl	0f				/* Find our address */
6840:	mflr	r5				/* Make it accessible */
685	tlbsx	r23,0,r5			/* Find entry we are in */
686	li	r4,0				/* Start at TLB entry 0 */
687	li	r3,0				/* Set PAGEID inval value */
6881:	cmpw	r23,r4				/* Is this our entry? */
689	beq	skip				/* If so, skip the inval */
690	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
691skip:
692	addi	r4,r4,1				/* Increment */
693	cmpwi	r4,64				/* Are we done?	*/
694	bne	1b				/* If not, repeat */
695	isync
696
697	/* Create a temp mapping and jump to it */
698	andi.	r6, r23, 1		/* Find the index to use */
699	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
700
701	mfmsr	r9			/* get the MSR */
702	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
703	xori	r7, r5, 1		/* Use the other address space */
704
705	/* Read the current mapping entries */
706	tlbre	r3, r23, PPC44x_TLB_PAGEID
707	tlbre	r4, r23, PPC44x_TLB_XLAT
708	tlbre	r5, r23, PPC44x_TLB_ATTRIB
709
710	/* Save our current XLAT entry */
711	mr	r25, r4
712
713	/* Extract the TLB PageSize */
714	li	r10, 1 			/* r10 will hold PageSize */
715	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
716
717	/* XXX: As of now we use 256M, 4K pages */
718	cmpwi	r11, PPC44x_TLB_256M
719	bne	tlb_4k
720	rotlwi	r10, r10, 28		/* r10 = 256M */
721	b	write_out
722tlb_4k:
723	cmpwi	r11, PPC44x_TLB_4K
724	bne	default
725	rotlwi	r10, r10, 12		/* r10 = 4K */
726	b	write_out
727default:
728	rotlwi	r10, r10, 10		/* r10 = 1K */
729
730write_out:
731	/*
732	 * Write out the tmp 1:1 mapping for this code in other address space
733	 * Fixup  EPN = RPN , TS=other address space
734	 */
735	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
736
737	/* Write out the tmp mapping entries */
738	tlbwe	r3, r24, PPC44x_TLB_PAGEID
739	tlbwe	r4, r24, PPC44x_TLB_XLAT
740	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
741
742	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
743	not	r10, r11		/* Mask for PageNum */
744
745	/* Switch to other address space in MSR */
746	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
747
748	bl	1f
7491:	mflr	r8
750	addi	r8, r8, (2f-1b)		/* Find the target offset */
751
752	/* Jump to the tmp mapping */
753	mtspr	SPRN_SRR0, r8
754	mtspr	SPRN_SRR1, r9
755	rfi
756
7572:
758	/* Invalidate the entry we were executing from */
759	li	r3, 0
760	tlbwe	r3, r23, PPC44x_TLB_PAGEID
761
762	/* attribute fields. rwx for SUPERVISOR mode */
763	li	r5, 0
764	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
765
766	/* Create 1:1 mapping in 256M pages */
767	xori	r7, r7, 1			/* Revert back to Original TS */
768
769	li	r8, 0				/* PageNumber */
770	li	r6, 3				/* TLB Index, start at 3  */
771
772next_tlb:
773	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
774	mr	r4, r3				/* RPN = EPN  */
775	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
776	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
777
778	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
779	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
780	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
781
782	addi	r8, r8, 1			/* Increment PN */
783	addi	r6, r6, 1			/* Increment TLB Index */
784	cmpwi	r8, 8				/* Are we done ? */
785	bne	next_tlb
786	isync
787
788	/* Jump to the new mapping 1:1 */
789	li	r9,0
790	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
791
792	bl	1f
7931:	mflr	r8
794	and	r8, r8, r11			/* Get our offset within page */
795	addi	r8, r8, (2f-1b)
796
797	and	r5, r25, r10			/* Get our target PageNum */
798	or	r8, r8, r5			/* Target jump address */
799
800	mtspr	SPRN_SRR0, r8
801	mtspr	SPRN_SRR1, r9
802	rfi
8032:
804	/* Invalidate the tmp entry we used */
805	li	r3, 0
806	tlbwe	r3, r24, PPC44x_TLB_PAGEID
807	sync
808	b	ppc44x_map_done
809
810#ifdef CONFIG_PPC_47x
811
812	/* 1:1 mapping for 47x */
813
814setup_map_47x:
815
816	/*
817	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
818	 * Also set the MSR IS->MMUCR STS
819	 */
820	li	r3, 0
821	mtspr	SPRN_PID, r3			/* Set PID */
822	mfmsr	r4				/* Get MSR */
823	andi.	r4, r4, MSR_IS@l		/* TS=1? */
824	beq	1f				/* If not, leave STS=0 */
825	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
8261:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
827	sync
828
829	/* Find the entry we are running from */
830	bl	2f
8312:	mflr	r23
832	tlbsx	r23, 0, r23
833	tlbre	r24, r23, 0			/* TLB Word 0 */
834	tlbre	r25, r23, 1			/* TLB Word 1 */
835	tlbre	r26, r23, 2			/* TLB Word 2 */
836
837
838	/*
839	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
840	 * of 4k page size in all  4 ways (0-3 in r3).
841	 * This would invalidate the entire UTLB including the one we are
842	 * running from. However the shadow TLB entries would help us
843	 * to continue the execution, until we flush them (rfi/isync).
844	 */
845	addis	r3, 0, 0x8000			/* specify the way */
846	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
847	addi	r5, 0, 0
848	b	clear_utlb_entry
849
850	/* Align the loop to speed things up. from head_44x.S */
851	.align	6
852
853clear_utlb_entry:
854
855	tlbwe	r4, r3, 0
856	tlbwe	r5, r3, 1
857	tlbwe	r5, r3, 2
858	addis	r3, r3, 0x2000			/* Increment the way */
859	cmpwi	r3, 0
860	bne	clear_utlb_entry
861	addis	r3, 0, 0x8000
862	addis	r4, r4, 0x100			/* Increment the EPN */
863	cmpwi	r4, 0
864	bne	clear_utlb_entry
865
866	/* Create the entries in the other address space */
867	mfmsr	r5
868	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
869	xori	r7, r7, 1			/* r7 = !TS */
870
871	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
872
873	/*
874	 * write out the TLB entries for the tmp mapping
875	 * Use way '0' so that we could easily invalidate it later.
876	 */
877	lis	r3, 0x8000			/* Way '0' */
878
879	tlbwe	r24, r3, 0
880	tlbwe	r25, r3, 1
881	tlbwe	r26, r3, 2
882
883	/* Update the msr to the new TS */
884	insrwi	r5, r7, 1, 26
885
886	bl	1f
8871:	mflr	r6
888	addi	r6, r6, (2f-1b)
889
890	mtspr	SPRN_SRR0, r6
891	mtspr	SPRN_SRR1, r5
892	rfi
893
894	/*
895	 * Now we are in the tmp address space.
896	 * Create a 1:1 mapping for 0-2GiB in the original TS.
897	 */
8982:
899	li	r3, 0
900	li	r4, 0				/* TLB Word 0 */
901	li	r5, 0				/* TLB Word 1 */
902	li	r6, 0
903	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
904
905	li	r8, 0				/* PageIndex */
906
907	xori	r7, r7, 1			/* revert back to original TS */
908
909write_utlb:
910	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
911						/* ERPN = 0 as we don't use memory above 2G */
912
913	mr	r4, r5				/* EPN = RPN */
914	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
915	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
916
917	tlbwe	r4, r3, 0			/* Write out the entries */
918	tlbwe	r5, r3, 1
919	tlbwe	r6, r3, 2
920	addi	r8, r8, 1
921	cmpwi	r8, 8				/* Have we completed ? */
922	bne	write_utlb
923
924	/* make sure we complete the TLB write up */
925	isync
926
927	/*
928	 * Prepare to jump to the 1:1 mapping.
929	 * 1) Extract page size of the tmp mapping
930	 *    DSIZ = TLB_Word0[22:27]
931	 * 2) Calculate the physical address of the address
932	 *    to jump to.
933	 */
934	rlwinm	r10, r24, 0, 22, 27
935
936	cmpwi	r10, PPC47x_TLB0_4K
937	bne	0f
938	li	r10, 0x1000			/* r10 = 4k */
939	bl	1f
940
9410:
942	/* Defaults to 256M */
943	lis	r10, 0x1000
944
945	bl	1f
9461:	mflr	r4
947	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
948
949	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
950	not	r10, r11			/* Pagemask = ~(offsetmask) */
951
952	and	r5, r25, r10			/* Physical page */
953	and	r6, r4, r11			/* offset within the current page */
954
955	or	r5, r5, r6			/* Physical address for 2f */
956
957	/* Switch the TS in MSR to the original one */
958	mfmsr	r8
959	insrwi	r8, r7, 1, 26
960
961	mtspr	SPRN_SRR1, r8
962	mtspr	SPRN_SRR0, r5
963	rfi
964
9652:
966	/* Invalidate the tmp mapping */
967	lis	r3, 0x8000			/* Way '0' */
968
969	clrrwi	r24, r24, 12			/* Clear the valid bit */
970	tlbwe	r24, r3, 0
971	tlbwe	r25, r3, 1
972	tlbwe	r26, r3, 2
973
974	/* Make sure we complete the TLB write and flush the shadow TLB */
975	isync
976
977#endif
978
979ppc44x_map_done:
980
981
982	/* Restore the parameters */
983	mr	r3, r29
984	mr	r4, r30
985	mr	r5, r31
986
987	li	r0, 0
988#else
989	li	r0, 0
990
991	/*
992	 * Set Machine Status Register to a known status,
993	 * switch the MMU off and jump to 1: in a single step.
994	 */
995
996	mr	r8, r0
997	ori     r8, r8, MSR_RI|MSR_ME
998	mtspr	SPRN_SRR1, r8
999	addi	r8, r4, 1f - relocate_new_kernel
1000	mtspr	SPRN_SRR0, r8
1001	sync
1002	rfi
1003
10041:
1005#endif
1006	/* from this point address translation is turned off */
1007	/* and interrupts are disabled */
1008
1009	/* set a new stack at the bottom of our page... */
1010	/* (not really needed now) */
1011	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1012	stw	r0, 0(r1)
1013
1014	/* Do the copies */
1015	li	r6, 0 /* checksum */
1016	mr	r0, r3
1017	b	1f
1018
10190:	/* top, read another word for the indirection page */
1020	lwzu	r0, 4(r3)
1021
10221:
1023	/* is it a destination page? (r8) */
1024	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1025	beq	2f
1026
1027	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1028	b	0b
1029
10302:	/* is it an indirection page? (r3) */
1031	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1032	beq	2f
1033
1034	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1035	subi	r3, r3, 4
1036	b	0b
1037
10382:	/* are we done? */
1039	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1040	beq	2f
1041	b	3f
1042
10432:	/* is it a source page? (r9) */
1044	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1045	beq	0b
1046
1047	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1048
1049	li	r7, PAGE_SIZE / 4
1050	mtctr   r7
1051	subi    r9, r9, 4
1052	subi    r8, r8, 4
10539:
1054	lwzu    r0, 4(r9)  /* do the copy */
1055	xor	r6, r6, r0
1056	stwu    r0, 4(r8)
1057	dcbst	0, r8
1058	sync
1059	icbi	0, r8
1060	bdnz    9b
1061
1062	addi    r9, r9, 4
1063	addi    r8, r8, 4
1064	b	0b
1065
10663:
1067
1068	/* To be certain of avoiding problems with self-modifying code
1069	 * execute a serializing instruction here.
1070	 */
1071	isync
1072	sync
1073
1074	mfspr	r3, SPRN_PIR /* current core we are running on */
1075	mr	r4, r5 /* load physical address of chunk called */
1076
1077	/* jump to the entry point, usually the setup routine */
1078	mtlr	r5
1079	blrl
1080
10811:	b	1b
1082
1083relocate_new_kernel_end:
1084
1085	.globl relocate_new_kernel_size
1086relocate_new_kernel_size:
1087	.long relocate_new_kernel_end - relocate_new_kernel
1088#endif
1089