xref: /linux/arch/powerpc/kernel/misc_32.S (revision bcb63314e2c23f1ed622418b65f9409512659c73)
1/*
2 * This file contains miscellaneous low-level functions.
3 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
4 *
5 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
6 * and Paul Mackerras.
7 *
8 * kexec bits:
9 * Copyright (C) 2002-2003 Eric Biederman  <ebiederm@xmission.com>
10 * GameCube/ppc32 port Copyright (C) 2004 Albert Herranz
11 * PPC44x port. Copyright (C) 2011,  IBM Corporation
12 * 		Author: Suzuki Poulose <suzuki@in.ibm.com>
13 *
14 * This program is free software; you can redistribute it and/or
15 * modify it under the terms of the GNU General Public License
16 * as published by the Free Software Foundation; either version
17 * 2 of the License, or (at your option) any later version.
18 *
19 */
20
21#include <linux/sys.h>
22#include <asm/unistd.h>
23#include <asm/errno.h>
24#include <asm/reg.h>
25#include <asm/page.h>
26#include <asm/cache.h>
27#include <asm/cputable.h>
28#include <asm/mmu.h>
29#include <asm/ppc_asm.h>
30#include <asm/thread_info.h>
31#include <asm/asm-offsets.h>
32#include <asm/processor.h>
33#include <asm/kexec.h>
34#include <asm/bug.h>
35#include <asm/ptrace.h>
36#include <asm/export.h>
37
38	.text
39
40/*
41 * We store the saved ksp_limit in the unused part
42 * of the STACK_FRAME_OVERHEAD
43 */
44_GLOBAL(call_do_softirq)
45	mflr	r0
46	stw	r0,4(r1)
47	lwz	r10,THREAD+KSP_LIMIT(r2)
48	addi	r11,r3,THREAD_INFO_GAP
49	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
50	mr	r1,r3
51	stw	r10,8(r1)
52	stw	r11,THREAD+KSP_LIMIT(r2)
53	bl	__do_softirq
54	lwz	r10,8(r1)
55	lwz	r1,0(r1)
56	lwz	r0,4(r1)
57	stw	r10,THREAD+KSP_LIMIT(r2)
58	mtlr	r0
59	blr
60
61/*
62 * void call_do_irq(struct pt_regs *regs, struct thread_info *irqtp);
63 */
64_GLOBAL(call_do_irq)
65	mflr	r0
66	stw	r0,4(r1)
67	lwz	r10,THREAD+KSP_LIMIT(r2)
68	addi	r11,r4,THREAD_INFO_GAP
69	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
70	mr	r1,r4
71	stw	r10,8(r1)
72	stw	r11,THREAD+KSP_LIMIT(r2)
73	bl	__do_irq
74	lwz	r10,8(r1)
75	lwz	r1,0(r1)
76	lwz	r0,4(r1)
77	stw	r10,THREAD+KSP_LIMIT(r2)
78	mtlr	r0
79	blr
80
81/*
82 * This returns the high 64 bits of the product of two 64-bit numbers.
83 */
84_GLOBAL(mulhdu)
85	cmpwi	r6,0
86	cmpwi	cr1,r3,0
87	mr	r10,r4
88	mulhwu	r4,r4,r5
89	beq	1f
90	mulhwu	r0,r10,r6
91	mullw	r7,r10,r5
92	addc	r7,r0,r7
93	addze	r4,r4
941:	beqlr	cr1		/* all done if high part of A is 0 */
95	mullw	r9,r3,r5
96	mulhwu	r10,r3,r5
97	beq	2f
98	mullw	r0,r3,r6
99	mulhwu	r8,r3,r6
100	addc	r7,r0,r7
101	adde	r4,r4,r8
102	addze	r10,r10
1032:	addc	r4,r4,r9
104	addze	r3,r10
105	blr
106
107/*
108 * reloc_got2 runs through the .got2 section adding an offset
109 * to each entry.
110 */
111_GLOBAL(reloc_got2)
112	mflr	r11
113	lis	r7,__got2_start@ha
114	addi	r7,r7,__got2_start@l
115	lis	r8,__got2_end@ha
116	addi	r8,r8,__got2_end@l
117	subf	r8,r7,r8
118	srwi.	r8,r8,2
119	beqlr
120	mtctr	r8
121	bl	1f
1221:	mflr	r0
123	lis	r4,1b@ha
124	addi	r4,r4,1b@l
125	subf	r0,r4,r0
126	add	r7,r0,r7
1272:	lwz	r0,0(r7)
128	add	r0,r0,r3
129	stw	r0,0(r7)
130	addi	r7,r7,4
131	bdnz	2b
132	mtlr	r11
133	blr
134
135/*
136 * call_setup_cpu - call the setup_cpu function for this cpu
137 * r3 = data offset, r24 = cpu number
138 *
139 * Setup function is called with:
140 *   r3 = data offset
141 *   r4 = ptr to CPU spec (relocated)
142 */
143_GLOBAL(call_setup_cpu)
144	addis	r4,r3,cur_cpu_spec@ha
145	addi	r4,r4,cur_cpu_spec@l
146	lwz	r4,0(r4)
147	add	r4,r4,r3
148	lwz	r5,CPU_SPEC_SETUP(r4)
149	cmpwi	0,r5,0
150	add	r5,r5,r3
151	beqlr
152	mtctr	r5
153	bctr
154
155#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
156
157/* This gets called by via-pmu.c to switch the PLL selection
158 * on 750fx CPU. This function should really be moved to some
159 * other place (as most of the cpufreq code in via-pmu
160 */
161_GLOBAL(low_choose_750fx_pll)
162	/* Clear MSR:EE */
163	mfmsr	r7
164	rlwinm	r0,r7,0,17,15
165	mtmsr	r0
166
167	/* If switching to PLL1, disable HID0:BTIC */
168	cmplwi	cr0,r3,0
169	beq	1f
170	mfspr	r5,SPRN_HID0
171	rlwinm	r5,r5,0,27,25
172	sync
173	mtspr	SPRN_HID0,r5
174	isync
175	sync
176
1771:
178	/* Calc new HID1 value */
179	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
180	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
181	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
182	or	r4,r4,r5
183	mtspr	SPRN_HID1,r4
184
185	/* Store new HID1 image */
186	CURRENT_THREAD_INFO(r6, r1)
187	lwz	r6,TI_CPU(r6)
188	slwi	r6,r6,2
189	addis	r6,r6,nap_save_hid1@ha
190	stw	r4,nap_save_hid1@l(r6)
191
192	/* If switching to PLL0, enable HID0:BTIC */
193	cmplwi	cr0,r3,0
194	bne	1f
195	mfspr	r5,SPRN_HID0
196	ori	r5,r5,HID0_BTIC
197	sync
198	mtspr	SPRN_HID0,r5
199	isync
200	sync
201
2021:
203	/* Return */
204	mtmsr	r7
205	blr
206
207_GLOBAL(low_choose_7447a_dfs)
208	/* Clear MSR:EE */
209	mfmsr	r7
210	rlwinm	r0,r7,0,17,15
211	mtmsr	r0
212
213	/* Calc new HID1 value */
214	mfspr	r4,SPRN_HID1
215	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
216	sync
217	mtspr	SPRN_HID1,r4
218	sync
219	isync
220
221	/* Return */
222	mtmsr	r7
223	blr
224
225#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
226
227/*
228 * complement mask on the msr then "or" some values on.
229 *     _nmask_and_or_msr(nmask, value_to_or)
230 */
231_GLOBAL(_nmask_and_or_msr)
232	mfmsr	r0		/* Get current msr */
233	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
234	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
235	SYNC			/* Some chip revs have problems here... */
236	mtmsr	r0		/* Update machine state */
237	isync
238	blr			/* Done */
239
240#ifdef CONFIG_40x
241
242/*
243 * Do an IO access in real mode
244 */
245_GLOBAL(real_readb)
246	mfmsr	r7
247	ori	r0,r7,MSR_DR
248	xori	r0,r0,MSR_DR
249	sync
250	mtmsr	r0
251	sync
252	isync
253	lbz	r3,0(r3)
254	sync
255	mtmsr	r7
256	sync
257	isync
258	blr
259
260	/*
261 * Do an IO access in real mode
262 */
263_GLOBAL(real_writeb)
264	mfmsr	r7
265	ori	r0,r7,MSR_DR
266	xori	r0,r0,MSR_DR
267	sync
268	mtmsr	r0
269	sync
270	isync
271	stb	r3,0(r4)
272	sync
273	mtmsr	r7
274	sync
275	isync
276	blr
277
278#endif /* CONFIG_40x */
279
280
281/*
282 * Flush instruction cache.
283 * This is a no-op on the 601.
284 */
285#ifndef CONFIG_PPC_8xx
286_GLOBAL(flush_instruction_cache)
287#if defined(CONFIG_4xx)
288#ifdef CONFIG_403GCX
289	li      r3, 512
290	mtctr   r3
291	lis     r4, KERNELBASE@h
2921:	iccci   0, r4
293	addi    r4, r4, 16
294	bdnz    1b
295#else
296	lis	r3, KERNELBASE@h
297	iccci	0,r3
298#endif
299#elif defined(CONFIG_FSL_BOOKE)
300BEGIN_FTR_SECTION
301	mfspr   r3,SPRN_L1CSR0
302	ori     r3,r3,L1CSR0_CFI|L1CSR0_CLFC
303	/* msync; isync recommended here */
304	mtspr   SPRN_L1CSR0,r3
305	isync
306	blr
307END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
308	mfspr	r3,SPRN_L1CSR1
309	ori	r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
310	mtspr	SPRN_L1CSR1,r3
311#else
312	mfspr	r3,SPRN_PVR
313	rlwinm	r3,r3,16,16,31
314	cmpwi	0,r3,1
315	beqlr			/* for 601, do nothing */
316	/* 603/604 processor - use invalidate-all bit in HID0 */
317	mfspr	r3,SPRN_HID0
318	ori	r3,r3,HID0_ICFI
319	mtspr	SPRN_HID0,r3
320#endif /* CONFIG_4xx */
321	isync
322	blr
323EXPORT_SYMBOL(flush_instruction_cache)
324#endif /* CONFIG_PPC_8xx */
325
326/*
327 * Write any modified data cache blocks out to memory
328 * and invalidate the corresponding instruction cache blocks.
329 * This is a no-op on the 601.
330 *
331 * flush_icache_range(unsigned long start, unsigned long stop)
332 */
333_GLOBAL(flush_icache_range)
334BEGIN_FTR_SECTION
335	PURGE_PREFETCHED_INS
336	blr				/* for 601, do nothing */
337END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
338	rlwinm	r3,r3,0,0,31 - L1_CACHE_SHIFT
339	subf	r4,r3,r4
340	addi	r4,r4,L1_CACHE_BYTES - 1
341	srwi.	r4,r4,L1_CACHE_SHIFT
342	beqlr
343	mtctr	r4
344	mr	r6,r3
3451:	dcbst	0,r3
346	addi	r3,r3,L1_CACHE_BYTES
347	bdnz	1b
348	sync				/* wait for dcbst's to get to ram */
349#ifndef CONFIG_44x
350	mtctr	r4
3512:	icbi	0,r6
352	addi	r6,r6,L1_CACHE_BYTES
353	bdnz	2b
354#else
355	/* Flash invalidate on 44x because we are passed kmapped addresses and
356	   this doesn't work for userspace pages due to the virtually tagged
357	   icache.  Sigh. */
358	iccci	0, r0
359#endif
360	sync				/* additional sync needed on g4 */
361	isync
362	blr
363_ASM_NOKPROBE_SYMBOL(flush_icache_range)
364EXPORT_SYMBOL(flush_icache_range)
365
366/*
367 * Flush a particular page from the data cache to RAM.
368 * Note: this is necessary because the instruction cache does *not*
369 * snoop from the data cache.
370 * This is a no-op on the 601 which has a unified cache.
371 *
372 *	void __flush_dcache_icache(void *page)
373 */
374_GLOBAL(__flush_dcache_icache)
375BEGIN_FTR_SECTION
376	PURGE_PREFETCHED_INS
377	blr
378END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
379	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
380	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
381	mtctr	r4
382	mr	r6,r3
3830:	dcbst	0,r3				/* Write line to ram */
384	addi	r3,r3,L1_CACHE_BYTES
385	bdnz	0b
386	sync
387#ifdef CONFIG_44x
388	/* We don't flush the icache on 44x. Those have a virtual icache
389	 * and we don't have access to the virtual address here (it's
390	 * not the page vaddr but where it's mapped in user space). The
391	 * flushing of the icache on these is handled elsewhere, when
392	 * a change in the address space occurs, before returning to
393	 * user space
394	 */
395BEGIN_MMU_FTR_SECTION
396	blr
397END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
398#endif /* CONFIG_44x */
399	mtctr	r4
4001:	icbi	0,r6
401	addi	r6,r6,L1_CACHE_BYTES
402	bdnz	1b
403	sync
404	isync
405	blr
406
407#ifndef CONFIG_BOOKE
408/*
409 * Flush a particular page from the data cache to RAM, identified
410 * by its physical address.  We turn off the MMU so we can just use
411 * the physical address (this may be a highmem page without a kernel
412 * mapping).
413 *
414 *	void __flush_dcache_icache_phys(unsigned long physaddr)
415 */
416_GLOBAL(__flush_dcache_icache_phys)
417BEGIN_FTR_SECTION
418	PURGE_PREFETCHED_INS
419	blr					/* for 601, do nothing */
420END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
421	mfmsr	r10
422	rlwinm	r0,r10,0,28,26			/* clear DR */
423	mtmsr	r0
424	isync
425	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
426	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
427	mtctr	r4
428	mr	r6,r3
4290:	dcbst	0,r3				/* Write line to ram */
430	addi	r3,r3,L1_CACHE_BYTES
431	bdnz	0b
432	sync
433	mtctr	r4
4341:	icbi	0,r6
435	addi	r6,r6,L1_CACHE_BYTES
436	bdnz	1b
437	sync
438	mtmsr	r10				/* restore DR */
439	isync
440	blr
441#endif /* CONFIG_BOOKE */
442
443/*
444 * Copy a whole page.  We use the dcbz instruction on the destination
445 * to reduce memory traffic (it eliminates the unnecessary reads of
446 * the destination into cache).  This requires that the destination
447 * is cacheable.
448 */
449#define COPY_16_BYTES		\
450	lwz	r6,4(r4);	\
451	lwz	r7,8(r4);	\
452	lwz	r8,12(r4);	\
453	lwzu	r9,16(r4);	\
454	stw	r6,4(r3);	\
455	stw	r7,8(r3);	\
456	stw	r8,12(r3);	\
457	stwu	r9,16(r3)
458
459_GLOBAL(copy_page)
460	addi	r3,r3,-4
461	addi	r4,r4,-4
462
463	li	r5,4
464
465#if MAX_COPY_PREFETCH > 1
466	li	r0,MAX_COPY_PREFETCH
467	li	r11,4
468	mtctr	r0
46911:	dcbt	r11,r4
470	addi	r11,r11,L1_CACHE_BYTES
471	bdnz	11b
472#else /* MAX_COPY_PREFETCH == 1 */
473	dcbt	r5,r4
474	li	r11,L1_CACHE_BYTES+4
475#endif /* MAX_COPY_PREFETCH */
476	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
477	crclr	4*cr0+eq
4782:
479	mtctr	r0
4801:
481	dcbt	r11,r4
482	dcbz	r5,r3
483	COPY_16_BYTES
484#if L1_CACHE_BYTES >= 32
485	COPY_16_BYTES
486#if L1_CACHE_BYTES >= 64
487	COPY_16_BYTES
488	COPY_16_BYTES
489#if L1_CACHE_BYTES >= 128
490	COPY_16_BYTES
491	COPY_16_BYTES
492	COPY_16_BYTES
493	COPY_16_BYTES
494#endif
495#endif
496#endif
497	bdnz	1b
498	beqlr
499	crnot	4*cr0+eq,4*cr0+eq
500	li	r0,MAX_COPY_PREFETCH
501	li	r11,4
502	b	2b
503EXPORT_SYMBOL(copy_page)
504
505/*
506 * Extended precision shifts.
507 *
508 * Updated to be valid for shift counts from 0 to 63 inclusive.
509 * -- Gabriel
510 *
511 * R3/R4 has 64 bit value
512 * R5    has shift count
513 * result in R3/R4
514 *
515 *  ashrdi3: arithmetic right shift (sign propagation)
516 *  lshrdi3: logical right shift
517 *  ashldi3: left shift
518 */
519_GLOBAL(__ashrdi3)
520	subfic	r6,r5,32
521	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
522	addi	r7,r5,32	# could be xori, or addi with -32
523	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
524	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
525	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
526	or	r4,r4,r6	# LSW |= t1
527	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
528	sraw	r3,r3,r5	# MSW = MSW >> count
529	or	r4,r4,r7	# LSW |= t2
530	blr
531EXPORT_SYMBOL(__ashrdi3)
532
533_GLOBAL(__ashldi3)
534	subfic	r6,r5,32
535	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
536	addi	r7,r5,32	# could be xori, or addi with -32
537	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
538	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
539	or	r3,r3,r6	# MSW |= t1
540	slw	r4,r4,r5	# LSW = LSW << count
541	or	r3,r3,r7	# MSW |= t2
542	blr
543EXPORT_SYMBOL(__ashldi3)
544
545_GLOBAL(__lshrdi3)
546	subfic	r6,r5,32
547	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
548	addi	r7,r5,32	# could be xori, or addi with -32
549	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
550	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
551	or	r4,r4,r6	# LSW |= t1
552	srw	r3,r3,r5	# MSW = MSW >> count
553	or	r4,r4,r7	# LSW |= t2
554	blr
555EXPORT_SYMBOL(__lshrdi3)
556
557/*
558 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
559 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
560 */
561_GLOBAL(__cmpdi2)
562	cmpw	r3,r5
563	li	r3,1
564	bne	1f
565	cmplw	r4,r6
566	beqlr
5671:	li	r3,0
568	bltlr
569	li	r3,2
570	blr
571EXPORT_SYMBOL(__cmpdi2)
572/*
573 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
574 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
575 */
576_GLOBAL(__ucmpdi2)
577	cmplw	r3,r5
578	li	r3,1
579	bne	1f
580	cmplw	r4,r6
581	beqlr
5821:	li	r3,0
583	bltlr
584	li	r3,2
585	blr
586EXPORT_SYMBOL(__ucmpdi2)
587
588_GLOBAL(__bswapdi2)
589	rotlwi  r9,r4,8
590	rotlwi  r10,r3,8
591	rlwimi  r9,r4,24,0,7
592	rlwimi  r10,r3,24,0,7
593	rlwimi  r9,r4,24,16,23
594	rlwimi  r10,r3,24,16,23
595	mr      r3,r9
596	mr      r4,r10
597	blr
598EXPORT_SYMBOL(__bswapdi2)
599
600#ifdef CONFIG_SMP
601_GLOBAL(start_secondary_resume)
602	/* Reset stack */
603	CURRENT_THREAD_INFO(r1, r1)
604	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
605	li	r3,0
606	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
607	bl	start_secondary
608	b	.
609#endif /* CONFIG_SMP */
610
611/*
612 * This routine is just here to keep GCC happy - sigh...
613 */
614_GLOBAL(__main)
615	blr
616
617#ifdef CONFIG_KEXEC_CORE
618	/*
619	 * Must be relocatable PIC code callable as a C function.
620	 */
621	.globl relocate_new_kernel
622relocate_new_kernel:
623	/* r3 = page_list   */
624	/* r4 = reboot_code_buffer */
625	/* r5 = start_address      */
626
627#ifdef CONFIG_FSL_BOOKE
628
629	mr	r29, r3
630	mr	r30, r4
631	mr	r31, r5
632
633#define ENTRY_MAPPING_KEXEC_SETUP
634#include "fsl_booke_entry_mapping.S"
635#undef ENTRY_MAPPING_KEXEC_SETUP
636
637	mr      r3, r29
638	mr      r4, r30
639	mr      r5, r31
640
641	li	r0, 0
642#elif defined(CONFIG_44x)
643
644	/* Save our parameters */
645	mr	r29, r3
646	mr	r30, r4
647	mr	r31, r5
648
649#ifdef CONFIG_PPC_47x
650	/* Check for 47x cores */
651	mfspr	r3,SPRN_PVR
652	srwi	r3,r3,16
653	cmplwi	cr0,r3,PVR_476FPE@h
654	beq	setup_map_47x
655	cmplwi	cr0,r3,PVR_476@h
656	beq	setup_map_47x
657	cmplwi	cr0,r3,PVR_476_ISS@h
658	beq	setup_map_47x
659#endif /* CONFIG_PPC_47x */
660
661/*
662 * Code for setting up 1:1 mapping for PPC440x for KEXEC
663 *
664 * We cannot switch off the MMU on PPC44x.
665 * So we:
666 * 1) Invalidate all the mappings except the one we are running from.
667 * 2) Create a tmp mapping for our code in the other address space(TS) and
668 *    jump to it. Invalidate the entry we started in.
669 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
670 * 4) Jump to the 1:1 mapping in original TS.
671 * 5) Invalidate the tmp mapping.
672 *
673 * - Based on the kexec support code for FSL BookE
674 *
675 */
676
677	/*
678	 * Load the PID with kernel PID (0).
679	 * Also load our MSR_IS and TID to MMUCR for TLB search.
680	 */
681	li	r3, 0
682	mtspr	SPRN_PID, r3
683	mfmsr	r4
684	andi.	r4,r4,MSR_IS@l
685	beq	wmmucr
686	oris	r3,r3,PPC44x_MMUCR_STS@h
687wmmucr:
688	mtspr	SPRN_MMUCR,r3
689	sync
690
691	/*
692	 * Invalidate all the TLB entries except the current entry
693	 * where we are running from
694	 */
695	bl	0f				/* Find our address */
6960:	mflr	r5				/* Make it accessible */
697	tlbsx	r23,0,r5			/* Find entry we are in */
698	li	r4,0				/* Start at TLB entry 0 */
699	li	r3,0				/* Set PAGEID inval value */
7001:	cmpw	r23,r4				/* Is this our entry? */
701	beq	skip				/* If so, skip the inval */
702	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
703skip:
704	addi	r4,r4,1				/* Increment */
705	cmpwi	r4,64				/* Are we done?	*/
706	bne	1b				/* If not, repeat */
707	isync
708
709	/* Create a temp mapping and jump to it */
710	andi.	r6, r23, 1		/* Find the index to use */
711	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
712
713	mfmsr	r9			/* get the MSR */
714	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
715	xori	r7, r5, 1		/* Use the other address space */
716
717	/* Read the current mapping entries */
718	tlbre	r3, r23, PPC44x_TLB_PAGEID
719	tlbre	r4, r23, PPC44x_TLB_XLAT
720	tlbre	r5, r23, PPC44x_TLB_ATTRIB
721
722	/* Save our current XLAT entry */
723	mr	r25, r4
724
725	/* Extract the TLB PageSize */
726	li	r10, 1 			/* r10 will hold PageSize */
727	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
728
729	/* XXX: As of now we use 256M, 4K pages */
730	cmpwi	r11, PPC44x_TLB_256M
731	bne	tlb_4k
732	rotlwi	r10, r10, 28		/* r10 = 256M */
733	b	write_out
734tlb_4k:
735	cmpwi	r11, PPC44x_TLB_4K
736	bne	default
737	rotlwi	r10, r10, 12		/* r10 = 4K */
738	b	write_out
739default:
740	rotlwi	r10, r10, 10		/* r10 = 1K */
741
742write_out:
743	/*
744	 * Write out the tmp 1:1 mapping for this code in other address space
745	 * Fixup  EPN = RPN , TS=other address space
746	 */
747	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
748
749	/* Write out the tmp mapping entries */
750	tlbwe	r3, r24, PPC44x_TLB_PAGEID
751	tlbwe	r4, r24, PPC44x_TLB_XLAT
752	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
753
754	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
755	not	r10, r11		/* Mask for PageNum */
756
757	/* Switch to other address space in MSR */
758	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
759
760	bl	1f
7611:	mflr	r8
762	addi	r8, r8, (2f-1b)		/* Find the target offset */
763
764	/* Jump to the tmp mapping */
765	mtspr	SPRN_SRR0, r8
766	mtspr	SPRN_SRR1, r9
767	rfi
768
7692:
770	/* Invalidate the entry we were executing from */
771	li	r3, 0
772	tlbwe	r3, r23, PPC44x_TLB_PAGEID
773
774	/* attribute fields. rwx for SUPERVISOR mode */
775	li	r5, 0
776	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
777
778	/* Create 1:1 mapping in 256M pages */
779	xori	r7, r7, 1			/* Revert back to Original TS */
780
781	li	r8, 0				/* PageNumber */
782	li	r6, 3				/* TLB Index, start at 3  */
783
784next_tlb:
785	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
786	mr	r4, r3				/* RPN = EPN  */
787	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
788	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
789
790	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
791	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
792	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
793
794	addi	r8, r8, 1			/* Increment PN */
795	addi	r6, r6, 1			/* Increment TLB Index */
796	cmpwi	r8, 8				/* Are we done ? */
797	bne	next_tlb
798	isync
799
800	/* Jump to the new mapping 1:1 */
801	li	r9,0
802	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
803
804	bl	1f
8051:	mflr	r8
806	and	r8, r8, r11			/* Get our offset within page */
807	addi	r8, r8, (2f-1b)
808
809	and	r5, r25, r10			/* Get our target PageNum */
810	or	r8, r8, r5			/* Target jump address */
811
812	mtspr	SPRN_SRR0, r8
813	mtspr	SPRN_SRR1, r9
814	rfi
8152:
816	/* Invalidate the tmp entry we used */
817	li	r3, 0
818	tlbwe	r3, r24, PPC44x_TLB_PAGEID
819	sync
820	b	ppc44x_map_done
821
822#ifdef CONFIG_PPC_47x
823
824	/* 1:1 mapping for 47x */
825
826setup_map_47x:
827
828	/*
829	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
830	 * Also set the MSR IS->MMUCR STS
831	 */
832	li	r3, 0
833	mtspr	SPRN_PID, r3			/* Set PID */
834	mfmsr	r4				/* Get MSR */
835	andi.	r4, r4, MSR_IS@l		/* TS=1? */
836	beq	1f				/* If not, leave STS=0 */
837	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
8381:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
839	sync
840
841	/* Find the entry we are running from */
842	bl	2f
8432:	mflr	r23
844	tlbsx	r23, 0, r23
845	tlbre	r24, r23, 0			/* TLB Word 0 */
846	tlbre	r25, r23, 1			/* TLB Word 1 */
847	tlbre	r26, r23, 2			/* TLB Word 2 */
848
849
850	/*
851	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
852	 * of 4k page size in all  4 ways (0-3 in r3).
853	 * This would invalidate the entire UTLB including the one we are
854	 * running from. However the shadow TLB entries would help us
855	 * to continue the execution, until we flush them (rfi/isync).
856	 */
857	addis	r3, 0, 0x8000			/* specify the way */
858	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
859	addi	r5, 0, 0
860	b	clear_utlb_entry
861
862	/* Align the loop to speed things up. from head_44x.S */
863	.align	6
864
865clear_utlb_entry:
866
867	tlbwe	r4, r3, 0
868	tlbwe	r5, r3, 1
869	tlbwe	r5, r3, 2
870	addis	r3, r3, 0x2000			/* Increment the way */
871	cmpwi	r3, 0
872	bne	clear_utlb_entry
873	addis	r3, 0, 0x8000
874	addis	r4, r4, 0x100			/* Increment the EPN */
875	cmpwi	r4, 0
876	bne	clear_utlb_entry
877
878	/* Create the entries in the other address space */
879	mfmsr	r5
880	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
881	xori	r7, r7, 1			/* r7 = !TS */
882
883	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
884
885	/*
886	 * write out the TLB entries for the tmp mapping
887	 * Use way '0' so that we could easily invalidate it later.
888	 */
889	lis	r3, 0x8000			/* Way '0' */
890
891	tlbwe	r24, r3, 0
892	tlbwe	r25, r3, 1
893	tlbwe	r26, r3, 2
894
895	/* Update the msr to the new TS */
896	insrwi	r5, r7, 1, 26
897
898	bl	1f
8991:	mflr	r6
900	addi	r6, r6, (2f-1b)
901
902	mtspr	SPRN_SRR0, r6
903	mtspr	SPRN_SRR1, r5
904	rfi
905
906	/*
907	 * Now we are in the tmp address space.
908	 * Create a 1:1 mapping for 0-2GiB in the original TS.
909	 */
9102:
911	li	r3, 0
912	li	r4, 0				/* TLB Word 0 */
913	li	r5, 0				/* TLB Word 1 */
914	li	r6, 0
915	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
916
917	li	r8, 0				/* PageIndex */
918
919	xori	r7, r7, 1			/* revert back to original TS */
920
921write_utlb:
922	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
923						/* ERPN = 0 as we don't use memory above 2G */
924
925	mr	r4, r5				/* EPN = RPN */
926	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
927	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
928
929	tlbwe	r4, r3, 0			/* Write out the entries */
930	tlbwe	r5, r3, 1
931	tlbwe	r6, r3, 2
932	addi	r8, r8, 1
933	cmpwi	r8, 8				/* Have we completed ? */
934	bne	write_utlb
935
936	/* make sure we complete the TLB write up */
937	isync
938
939	/*
940	 * Prepare to jump to the 1:1 mapping.
941	 * 1) Extract page size of the tmp mapping
942	 *    DSIZ = TLB_Word0[22:27]
943	 * 2) Calculate the physical address of the address
944	 *    to jump to.
945	 */
946	rlwinm	r10, r24, 0, 22, 27
947
948	cmpwi	r10, PPC47x_TLB0_4K
949	bne	0f
950	li	r10, 0x1000			/* r10 = 4k */
951	bl	1f
952
9530:
954	/* Defaults to 256M */
955	lis	r10, 0x1000
956
957	bl	1f
9581:	mflr	r4
959	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
960
961	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
962	not	r10, r11			/* Pagemask = ~(offsetmask) */
963
964	and	r5, r25, r10			/* Physical page */
965	and	r6, r4, r11			/* offset within the current page */
966
967	or	r5, r5, r6			/* Physical address for 2f */
968
969	/* Switch the TS in MSR to the original one */
970	mfmsr	r8
971	insrwi	r8, r7, 1, 26
972
973	mtspr	SPRN_SRR1, r8
974	mtspr	SPRN_SRR0, r5
975	rfi
976
9772:
978	/* Invalidate the tmp mapping */
979	lis	r3, 0x8000			/* Way '0' */
980
981	clrrwi	r24, r24, 12			/* Clear the valid bit */
982	tlbwe	r24, r3, 0
983	tlbwe	r25, r3, 1
984	tlbwe	r26, r3, 2
985
986	/* Make sure we complete the TLB write and flush the shadow TLB */
987	isync
988
989#endif
990
991ppc44x_map_done:
992
993
994	/* Restore the parameters */
995	mr	r3, r29
996	mr	r4, r30
997	mr	r5, r31
998
999	li	r0, 0
1000#else
1001	li	r0, 0
1002
1003	/*
1004	 * Set Machine Status Register to a known status,
1005	 * switch the MMU off and jump to 1: in a single step.
1006	 */
1007
1008	mr	r8, r0
1009	ori     r8, r8, MSR_RI|MSR_ME
1010	mtspr	SPRN_SRR1, r8
1011	addi	r8, r4, 1f - relocate_new_kernel
1012	mtspr	SPRN_SRR0, r8
1013	sync
1014	rfi
1015
10161:
1017#endif
1018	/* from this point address translation is turned off */
1019	/* and interrupts are disabled */
1020
1021	/* set a new stack at the bottom of our page... */
1022	/* (not really needed now) */
1023	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1024	stw	r0, 0(r1)
1025
1026	/* Do the copies */
1027	li	r6, 0 /* checksum */
1028	mr	r0, r3
1029	b	1f
1030
10310:	/* top, read another word for the indirection page */
1032	lwzu	r0, 4(r3)
1033
10341:
1035	/* is it a destination page? (r8) */
1036	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1037	beq	2f
1038
1039	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1040	b	0b
1041
10422:	/* is it an indirection page? (r3) */
1043	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1044	beq	2f
1045
1046	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1047	subi	r3, r3, 4
1048	b	0b
1049
10502:	/* are we done? */
1051	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1052	beq	2f
1053	b	3f
1054
10552:	/* is it a source page? (r9) */
1056	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1057	beq	0b
1058
1059	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1060
1061	li	r7, PAGE_SIZE / 4
1062	mtctr   r7
1063	subi    r9, r9, 4
1064	subi    r8, r8, 4
10659:
1066	lwzu    r0, 4(r9)  /* do the copy */
1067	xor	r6, r6, r0
1068	stwu    r0, 4(r8)
1069	dcbst	0, r8
1070	sync
1071	icbi	0, r8
1072	bdnz    9b
1073
1074	addi    r9, r9, 4
1075	addi    r8, r8, 4
1076	b	0b
1077
10783:
1079
1080	/* To be certain of avoiding problems with self-modifying code
1081	 * execute a serializing instruction here.
1082	 */
1083	isync
1084	sync
1085
1086	mfspr	r3, SPRN_PIR /* current core we are running on */
1087	mr	r4, r5 /* load physical address of chunk called */
1088
1089	/* jump to the entry point, usually the setup routine */
1090	mtlr	r5
1091	blrl
1092
10931:	b	1b
1094
1095relocate_new_kernel_end:
1096
1097	.globl relocate_new_kernel_size
1098relocate_new_kernel_size:
1099	.long relocate_new_kernel_end - relocate_new_kernel
1100#endif
1101