xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision 8a10bc9d)
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_GLOBAL(call_do_irq)
61	mflr	r0
62	stw	r0,4(r1)
63	lwz	r10,THREAD+KSP_LIMIT(r2)
64	addi	r11,r3,THREAD_INFO_GAP
65	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
66	mr	r1,r4
67	stw	r10,8(r1)
68	stw	r11,THREAD+KSP_LIMIT(r2)
69	bl	__do_irq
70	lwz	r10,8(r1)
71	lwz	r1,0(r1)
72	lwz	r0,4(r1)
73	stw	r10,THREAD+KSP_LIMIT(r2)
74	mtlr	r0
75	blr
76
77/*
78 * This returns the high 64 bits of the product of two 64-bit numbers.
79 */
80_GLOBAL(mulhdu)
81	cmpwi	r6,0
82	cmpwi	cr1,r3,0
83	mr	r10,r4
84	mulhwu	r4,r4,r5
85	beq	1f
86	mulhwu	r0,r10,r6
87	mullw	r7,r10,r5
88	addc	r7,r0,r7
89	addze	r4,r4
901:	beqlr	cr1		/* all done if high part of A is 0 */
91	mr	r10,r3
92	mullw	r9,r3,r5
93	mulhwu	r3,r3,r5
94	beq	2f
95	mullw	r0,r10,r6
96	mulhwu	r8,r10,r6
97	addc	r7,r0,r7
98	adde	r4,r4,r8
99	addze	r3,r3
1002:	addc	r4,r4,r9
101	addze	r3,r3
102	blr
103
104/*
105 * sub_reloc_offset(x) returns x - reloc_offset().
106 */
107_GLOBAL(sub_reloc_offset)
108	mflr	r0
109	bl	1f
1101:	mflr	r5
111	lis	r4,1b@ha
112	addi	r4,r4,1b@l
113	subf	r5,r4,r5
114	subf	r3,r5,r3
115	mtlr	r0
116	blr
117
118/*
119 * reloc_got2 runs through the .got2 section adding an offset
120 * to each entry.
121 */
122_GLOBAL(reloc_got2)
123	mflr	r11
124	lis	r7,__got2_start@ha
125	addi	r7,r7,__got2_start@l
126	lis	r8,__got2_end@ha
127	addi	r8,r8,__got2_end@l
128	subf	r8,r7,r8
129	srwi.	r8,r8,2
130	beqlr
131	mtctr	r8
132	bl	1f
1331:	mflr	r0
134	lis	r4,1b@ha
135	addi	r4,r4,1b@l
136	subf	r0,r4,r0
137	add	r7,r0,r7
1382:	lwz	r0,0(r7)
139	add	r0,r0,r3
140	stw	r0,0(r7)
141	addi	r7,r7,4
142	bdnz	2b
143	mtlr	r11
144	blr
145
146/*
147 * call_setup_cpu - call the setup_cpu function for this cpu
148 * r3 = data offset, r24 = cpu number
149 *
150 * Setup function is called with:
151 *   r3 = data offset
152 *   r4 = ptr to CPU spec (relocated)
153 */
154_GLOBAL(call_setup_cpu)
155	addis	r4,r3,cur_cpu_spec@ha
156	addi	r4,r4,cur_cpu_spec@l
157	lwz	r4,0(r4)
158	add	r4,r4,r3
159	lwz	r5,CPU_SPEC_SETUP(r4)
160	cmpwi	0,r5,0
161	add	r5,r5,r3
162	beqlr
163	mtctr	r5
164	bctr
165
166#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
167
168/* This gets called by via-pmu.c to switch the PLL selection
169 * on 750fx CPU. This function should really be moved to some
170 * other place (as most of the cpufreq code in via-pmu
171 */
172_GLOBAL(low_choose_750fx_pll)
173	/* Clear MSR:EE */
174	mfmsr	r7
175	rlwinm	r0,r7,0,17,15
176	mtmsr	r0
177
178	/* If switching to PLL1, disable HID0:BTIC */
179	cmplwi	cr0,r3,0
180	beq	1f
181	mfspr	r5,SPRN_HID0
182	rlwinm	r5,r5,0,27,25
183	sync
184	mtspr	SPRN_HID0,r5
185	isync
186	sync
187
1881:
189	/* Calc new HID1 value */
190	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
191	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
192	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
193	or	r4,r4,r5
194	mtspr	SPRN_HID1,r4
195
196	/* Store new HID1 image */
197	CURRENT_THREAD_INFO(r6, r1)
198	lwz	r6,TI_CPU(r6)
199	slwi	r6,r6,2
200	addis	r6,r6,nap_save_hid1@ha
201	stw	r4,nap_save_hid1@l(r6)
202
203	/* If switching to PLL0, enable HID0:BTIC */
204	cmplwi	cr0,r3,0
205	bne	1f
206	mfspr	r5,SPRN_HID0
207	ori	r5,r5,HID0_BTIC
208	sync
209	mtspr	SPRN_HID0,r5
210	isync
211	sync
212
2131:
214	/* Return */
215	mtmsr	r7
216	blr
217
218_GLOBAL(low_choose_7447a_dfs)
219	/* Clear MSR:EE */
220	mfmsr	r7
221	rlwinm	r0,r7,0,17,15
222	mtmsr	r0
223
224	/* Calc new HID1 value */
225	mfspr	r4,SPRN_HID1
226	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
227	sync
228	mtspr	SPRN_HID1,r4
229	sync
230	isync
231
232	/* Return */
233	mtmsr	r7
234	blr
235
236#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
237
238/*
239 * complement mask on the msr then "or" some values on.
240 *     _nmask_and_or_msr(nmask, value_to_or)
241 */
242_GLOBAL(_nmask_and_or_msr)
243	mfmsr	r0		/* Get current msr */
244	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
245	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
246	SYNC			/* Some chip revs have problems here... */
247	mtmsr	r0		/* Update machine state */
248	isync
249	blr			/* Done */
250
251#ifdef CONFIG_40x
252
253/*
254 * Do an IO access in real mode
255 */
256_GLOBAL(real_readb)
257	mfmsr	r7
258	ori	r0,r7,MSR_DR
259	xori	r0,r0,MSR_DR
260	sync
261	mtmsr	r0
262	sync
263	isync
264	lbz	r3,0(r3)
265	sync
266	mtmsr	r7
267	sync
268	isync
269	blr
270
271	/*
272 * Do an IO access in real mode
273 */
274_GLOBAL(real_writeb)
275	mfmsr	r7
276	ori	r0,r7,MSR_DR
277	xori	r0,r0,MSR_DR
278	sync
279	mtmsr	r0
280	sync
281	isync
282	stb	r3,0(r4)
283	sync
284	mtmsr	r7
285	sync
286	isync
287	blr
288
289#endif /* CONFIG_40x */
290
291
292/*
293 * Flush instruction cache.
294 * This is a no-op on the 601.
295 */
296_GLOBAL(flush_instruction_cache)
297#if defined(CONFIG_8xx)
298	isync
299	lis	r5, IDC_INVALL@h
300	mtspr	SPRN_IC_CST, r5
301#elif defined(CONFIG_4xx)
302#ifdef CONFIG_403GCX
303	li      r3, 512
304	mtctr   r3
305	lis     r4, KERNELBASE@h
3061:	iccci   0, r4
307	addi    r4, r4, 16
308	bdnz    1b
309#else
310	lis	r3, KERNELBASE@h
311	iccci	0,r3
312#endif
313#elif CONFIG_FSL_BOOKE
314BEGIN_FTR_SECTION
315	mfspr   r3,SPRN_L1CSR0
316	ori     r3,r3,L1CSR0_CFI|L1CSR0_CLFC
317	/* msync; isync recommended here */
318	mtspr   SPRN_L1CSR0,r3
319	isync
320	blr
321END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
322	mfspr	r3,SPRN_L1CSR1
323	ori	r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
324	mtspr	SPRN_L1CSR1,r3
325#else
326	mfspr	r3,SPRN_PVR
327	rlwinm	r3,r3,16,16,31
328	cmpwi	0,r3,1
329	beqlr			/* for 601, do nothing */
330	/* 603/604 processor - use invalidate-all bit in HID0 */
331	mfspr	r3,SPRN_HID0
332	ori	r3,r3,HID0_ICFI
333	mtspr	SPRN_HID0,r3
334#endif /* CONFIG_8xx/4xx */
335	isync
336	blr
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	li	r5,L1_CACHE_BYTES-1
351	andc	r3,r3,r5
352	subf	r4,r3,r4
353	add	r4,r4,r5
354	srwi.	r4,r4,L1_CACHE_SHIFT
355	beqlr
356	mtctr	r4
357	mr	r6,r3
3581:	dcbst	0,r3
359	addi	r3,r3,L1_CACHE_BYTES
360	bdnz	1b
361	sync				/* wait for dcbst's to get to ram */
362#ifndef CONFIG_44x
363	mtctr	r4
3642:	icbi	0,r6
365	addi	r6,r6,L1_CACHE_BYTES
366	bdnz	2b
367#else
368	/* Flash invalidate on 44x because we are passed kmapped addresses and
369	   this doesn't work for userspace pages due to the virtually tagged
370	   icache.  Sigh. */
371	iccci	0, r0
372#endif
373	sync				/* additional sync needed on g4 */
374	isync
375	blr
376/*
377 * Write any modified data cache blocks out to memory.
378 * Does not invalidate the corresponding cache lines (especially for
379 * any corresponding instruction cache).
380 *
381 * clean_dcache_range(unsigned long start, unsigned long stop)
382 */
383_GLOBAL(clean_dcache_range)
384	li	r5,L1_CACHE_BYTES-1
385	andc	r3,r3,r5
386	subf	r4,r3,r4
387	add	r4,r4,r5
388	srwi.	r4,r4,L1_CACHE_SHIFT
389	beqlr
390	mtctr	r4
391
3921:	dcbst	0,r3
393	addi	r3,r3,L1_CACHE_BYTES
394	bdnz	1b
395	sync				/* wait for dcbst's to get to ram */
396	blr
397
398/*
399 * Write any modified data cache blocks out to memory and invalidate them.
400 * Does not invalidate the corresponding instruction cache blocks.
401 *
402 * flush_dcache_range(unsigned long start, unsigned long stop)
403 */
404_GLOBAL(flush_dcache_range)
405	li	r5,L1_CACHE_BYTES-1
406	andc	r3,r3,r5
407	subf	r4,r3,r4
408	add	r4,r4,r5
409	srwi.	r4,r4,L1_CACHE_SHIFT
410	beqlr
411	mtctr	r4
412
4131:	dcbf	0,r3
414	addi	r3,r3,L1_CACHE_BYTES
415	bdnz	1b
416	sync				/* wait for dcbst's to get to ram */
417	blr
418
419/*
420 * Like above, but invalidate the D-cache.  This is used by the 8xx
421 * to invalidate the cache so the PPC core doesn't get stale data
422 * from the CPM (no cache snooping here :-).
423 *
424 * invalidate_dcache_range(unsigned long start, unsigned long stop)
425 */
426_GLOBAL(invalidate_dcache_range)
427	li	r5,L1_CACHE_BYTES-1
428	andc	r3,r3,r5
429	subf	r4,r3,r4
430	add	r4,r4,r5
431	srwi.	r4,r4,L1_CACHE_SHIFT
432	beqlr
433	mtctr	r4
434
4351:	dcbi	0,r3
436	addi	r3,r3,L1_CACHE_BYTES
437	bdnz	1b
438	sync				/* wait for dcbi's to get to ram */
439	blr
440
441/*
442 * Flush a particular page from the data cache to RAM.
443 * Note: this is necessary because the instruction cache does *not*
444 * snoop from the data cache.
445 * This is a no-op on the 601 which has a unified cache.
446 *
447 *	void __flush_dcache_icache(void *page)
448 */
449_GLOBAL(__flush_dcache_icache)
450BEGIN_FTR_SECTION
451	PURGE_PREFETCHED_INS
452	blr
453END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
454	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
455	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
456	mtctr	r4
457	mr	r6,r3
4580:	dcbst	0,r3				/* Write line to ram */
459	addi	r3,r3,L1_CACHE_BYTES
460	bdnz	0b
461	sync
462#ifdef CONFIG_44x
463	/* We don't flush the icache on 44x. Those have a virtual icache
464	 * and we don't have access to the virtual address here (it's
465	 * not the page vaddr but where it's mapped in user space). The
466	 * flushing of the icache on these is handled elsewhere, when
467	 * a change in the address space occurs, before returning to
468	 * user space
469	 */
470BEGIN_MMU_FTR_SECTION
471	blr
472END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
473#endif /* CONFIG_44x */
474	mtctr	r4
4751:	icbi	0,r6
476	addi	r6,r6,L1_CACHE_BYTES
477	bdnz	1b
478	sync
479	isync
480	blr
481
482#ifndef CONFIG_BOOKE
483/*
484 * Flush a particular page from the data cache to RAM, identified
485 * by its physical address.  We turn off the MMU so we can just use
486 * the physical address (this may be a highmem page without a kernel
487 * mapping).
488 *
489 *	void __flush_dcache_icache_phys(unsigned long physaddr)
490 */
491_GLOBAL(__flush_dcache_icache_phys)
492BEGIN_FTR_SECTION
493	PURGE_PREFETCHED_INS
494	blr					/* for 601, do nothing */
495END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
496	mfmsr	r10
497	rlwinm	r0,r10,0,28,26			/* clear DR */
498	mtmsr	r0
499	isync
500	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
501	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
502	mtctr	r4
503	mr	r6,r3
5040:	dcbst	0,r3				/* Write line to ram */
505	addi	r3,r3,L1_CACHE_BYTES
506	bdnz	0b
507	sync
508	mtctr	r4
5091:	icbi	0,r6
510	addi	r6,r6,L1_CACHE_BYTES
511	bdnz	1b
512	sync
513	mtmsr	r10				/* restore DR */
514	isync
515	blr
516#endif /* CONFIG_BOOKE */
517
518/*
519 * Clear pages using the dcbz instruction, which doesn't cause any
520 * memory traffic (except to write out any cache lines which get
521 * displaced).  This only works on cacheable memory.
522 *
523 * void clear_pages(void *page, int order) ;
524 */
525_GLOBAL(clear_pages)
526	li	r0,PAGE_SIZE/L1_CACHE_BYTES
527	slw	r0,r0,r4
528	mtctr	r0
5291:	dcbz	0,r3
530	addi	r3,r3,L1_CACHE_BYTES
531	bdnz	1b
532	blr
533
534/*
535 * Copy a whole page.  We use the dcbz instruction on the destination
536 * to reduce memory traffic (it eliminates the unnecessary reads of
537 * the destination into cache).  This requires that the destination
538 * is cacheable.
539 */
540#define COPY_16_BYTES		\
541	lwz	r6,4(r4);	\
542	lwz	r7,8(r4);	\
543	lwz	r8,12(r4);	\
544	lwzu	r9,16(r4);	\
545	stw	r6,4(r3);	\
546	stw	r7,8(r3);	\
547	stw	r8,12(r3);	\
548	stwu	r9,16(r3)
549
550_GLOBAL(copy_page)
551	addi	r3,r3,-4
552	addi	r4,r4,-4
553
554	li	r5,4
555
556#if MAX_COPY_PREFETCH > 1
557	li	r0,MAX_COPY_PREFETCH
558	li	r11,4
559	mtctr	r0
56011:	dcbt	r11,r4
561	addi	r11,r11,L1_CACHE_BYTES
562	bdnz	11b
563#else /* MAX_COPY_PREFETCH == 1 */
564	dcbt	r5,r4
565	li	r11,L1_CACHE_BYTES+4
566#endif /* MAX_COPY_PREFETCH */
567	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
568	crclr	4*cr0+eq
5692:
570	mtctr	r0
5711:
572	dcbt	r11,r4
573	dcbz	r5,r3
574	COPY_16_BYTES
575#if L1_CACHE_BYTES >= 32
576	COPY_16_BYTES
577#if L1_CACHE_BYTES >= 64
578	COPY_16_BYTES
579	COPY_16_BYTES
580#if L1_CACHE_BYTES >= 128
581	COPY_16_BYTES
582	COPY_16_BYTES
583	COPY_16_BYTES
584	COPY_16_BYTES
585#endif
586#endif
587#endif
588	bdnz	1b
589	beqlr
590	crnot	4*cr0+eq,4*cr0+eq
591	li	r0,MAX_COPY_PREFETCH
592	li	r11,4
593	b	2b
594
595/*
596 * void atomic_clear_mask(atomic_t mask, atomic_t *addr)
597 * void atomic_set_mask(atomic_t mask, atomic_t *addr);
598 */
599_GLOBAL(atomic_clear_mask)
60010:	lwarx	r5,0,r4
601	andc	r5,r5,r3
602	PPC405_ERR77(0,r4)
603	stwcx.	r5,0,r4
604	bne-	10b
605	blr
606_GLOBAL(atomic_set_mask)
60710:	lwarx	r5,0,r4
608	or	r5,r5,r3
609	PPC405_ERR77(0,r4)
610	stwcx.	r5,0,r4
611	bne-	10b
612	blr
613
614/*
615 * Extended precision shifts.
616 *
617 * Updated to be valid for shift counts from 0 to 63 inclusive.
618 * -- Gabriel
619 *
620 * R3/R4 has 64 bit value
621 * R5    has shift count
622 * result in R3/R4
623 *
624 *  ashrdi3: arithmetic right shift (sign propagation)
625 *  lshrdi3: logical right shift
626 *  ashldi3: left shift
627 */
628_GLOBAL(__ashrdi3)
629	subfic	r6,r5,32
630	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
631	addi	r7,r5,32	# could be xori, or addi with -32
632	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
633	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
634	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
635	or	r4,r4,r6	# LSW |= t1
636	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
637	sraw	r3,r3,r5	# MSW = MSW >> count
638	or	r4,r4,r7	# LSW |= t2
639	blr
640
641_GLOBAL(__ashldi3)
642	subfic	r6,r5,32
643	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
644	addi	r7,r5,32	# could be xori, or addi with -32
645	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
646	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
647	or	r3,r3,r6	# MSW |= t1
648	slw	r4,r4,r5	# LSW = LSW << count
649	or	r3,r3,r7	# MSW |= t2
650	blr
651
652_GLOBAL(__lshrdi3)
653	subfic	r6,r5,32
654	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
655	addi	r7,r5,32	# could be xori, or addi with -32
656	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
657	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
658	or	r4,r4,r6	# LSW |= t1
659	srw	r3,r3,r5	# MSW = MSW >> count
660	or	r4,r4,r7	# LSW |= t2
661	blr
662
663/*
664 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
665 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
666 */
667_GLOBAL(__cmpdi2)
668	cmpw	r3,r5
669	li	r3,1
670	bne	1f
671	cmplw	r4,r6
672	beqlr
6731:	li	r3,0
674	bltlr
675	li	r3,2
676	blr
677/*
678 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
679 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
680 */
681_GLOBAL(__ucmpdi2)
682	cmplw	r3,r5
683	li	r3,1
684	bne	1f
685	cmplw	r4,r6
686	beqlr
6871:	li	r3,0
688	bltlr
689	li	r3,2
690	blr
691
692_GLOBAL(__bswapdi2)
693	rotlwi  r9,r4,8
694	rotlwi  r10,r3,8
695	rlwimi  r9,r4,24,0,7
696	rlwimi  r10,r3,24,0,7
697	rlwimi  r9,r4,24,16,23
698	rlwimi  r10,r3,24,16,23
699	mr      r3,r9
700	mr      r4,r10
701	blr
702
703_GLOBAL(abs)
704	srawi	r4,r3,31
705	xor	r3,r3,r4
706	sub	r3,r3,r4
707	blr
708
709#ifdef CONFIG_SMP
710_GLOBAL(start_secondary_resume)
711	/* Reset stack */
712	CURRENT_THREAD_INFO(r1, r1)
713	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
714	li	r3,0
715	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
716	bl	start_secondary
717	b	.
718#endif /* CONFIG_SMP */
719
720/*
721 * This routine is just here to keep GCC happy - sigh...
722 */
723_GLOBAL(__main)
724	blr
725
726#ifdef CONFIG_KEXEC
727	/*
728	 * Must be relocatable PIC code callable as a C function.
729	 */
730	.globl relocate_new_kernel
731relocate_new_kernel:
732	/* r3 = page_list   */
733	/* r4 = reboot_code_buffer */
734	/* r5 = start_address      */
735
736#ifdef CONFIG_FSL_BOOKE
737
738	mr	r29, r3
739	mr	r30, r4
740	mr	r31, r5
741
742#define ENTRY_MAPPING_KEXEC_SETUP
743#include "fsl_booke_entry_mapping.S"
744#undef ENTRY_MAPPING_KEXEC_SETUP
745
746	mr      r3, r29
747	mr      r4, r30
748	mr      r5, r31
749
750	li	r0, 0
751#elif defined(CONFIG_44x)
752
753	/* Save our parameters */
754	mr	r29, r3
755	mr	r30, r4
756	mr	r31, r5
757
758#ifdef CONFIG_PPC_47x
759	/* Check for 47x cores */
760	mfspr	r3,SPRN_PVR
761	srwi	r3,r3,16
762	cmplwi	cr0,r3,PVR_476@h
763	beq	setup_map_47x
764	cmplwi	cr0,r3,PVR_476_ISS@h
765	beq	setup_map_47x
766#endif /* CONFIG_PPC_47x */
767
768/*
769 * Code for setting up 1:1 mapping for PPC440x for KEXEC
770 *
771 * We cannot switch off the MMU on PPC44x.
772 * So we:
773 * 1) Invalidate all the mappings except the one we are running from.
774 * 2) Create a tmp mapping for our code in the other address space(TS) and
775 *    jump to it. Invalidate the entry we started in.
776 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
777 * 4) Jump to the 1:1 mapping in original TS.
778 * 5) Invalidate the tmp mapping.
779 *
780 * - Based on the kexec support code for FSL BookE
781 *
782 */
783
784	/*
785	 * Load the PID with kernel PID (0).
786	 * Also load our MSR_IS and TID to MMUCR for TLB search.
787	 */
788	li	r3, 0
789	mtspr	SPRN_PID, r3
790	mfmsr	r4
791	andi.	r4,r4,MSR_IS@l
792	beq	wmmucr
793	oris	r3,r3,PPC44x_MMUCR_STS@h
794wmmucr:
795	mtspr	SPRN_MMUCR,r3
796	sync
797
798	/*
799	 * Invalidate all the TLB entries except the current entry
800	 * where we are running from
801	 */
802	bl	0f				/* Find our address */
8030:	mflr	r5				/* Make it accessible */
804	tlbsx	r23,0,r5			/* Find entry we are in */
805	li	r4,0				/* Start at TLB entry 0 */
806	li	r3,0				/* Set PAGEID inval value */
8071:	cmpw	r23,r4				/* Is this our entry? */
808	beq	skip				/* If so, skip the inval */
809	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
810skip:
811	addi	r4,r4,1				/* Increment */
812	cmpwi	r4,64				/* Are we done?	*/
813	bne	1b				/* If not, repeat */
814	isync
815
816	/* Create a temp mapping and jump to it */
817	andi.	r6, r23, 1		/* Find the index to use */
818	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
819
820	mfmsr	r9			/* get the MSR */
821	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
822	xori	r7, r5, 1		/* Use the other address space */
823
824	/* Read the current mapping entries */
825	tlbre	r3, r23, PPC44x_TLB_PAGEID
826	tlbre	r4, r23, PPC44x_TLB_XLAT
827	tlbre	r5, r23, PPC44x_TLB_ATTRIB
828
829	/* Save our current XLAT entry */
830	mr	r25, r4
831
832	/* Extract the TLB PageSize */
833	li	r10, 1 			/* r10 will hold PageSize */
834	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
835
836	/* XXX: As of now we use 256M, 4K pages */
837	cmpwi	r11, PPC44x_TLB_256M
838	bne	tlb_4k
839	rotlwi	r10, r10, 28		/* r10 = 256M */
840	b	write_out
841tlb_4k:
842	cmpwi	r11, PPC44x_TLB_4K
843	bne	default
844	rotlwi	r10, r10, 12		/* r10 = 4K */
845	b	write_out
846default:
847	rotlwi	r10, r10, 10		/* r10 = 1K */
848
849write_out:
850	/*
851	 * Write out the tmp 1:1 mapping for this code in other address space
852	 * Fixup  EPN = RPN , TS=other address space
853	 */
854	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
855
856	/* Write out the tmp mapping entries */
857	tlbwe	r3, r24, PPC44x_TLB_PAGEID
858	tlbwe	r4, r24, PPC44x_TLB_XLAT
859	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
860
861	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
862	not	r10, r11		/* Mask for PageNum */
863
864	/* Switch to other address space in MSR */
865	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
866
867	bl	1f
8681:	mflr	r8
869	addi	r8, r8, (2f-1b)		/* Find the target offset */
870
871	/* Jump to the tmp mapping */
872	mtspr	SPRN_SRR0, r8
873	mtspr	SPRN_SRR1, r9
874	rfi
875
8762:
877	/* Invalidate the entry we were executing from */
878	li	r3, 0
879	tlbwe	r3, r23, PPC44x_TLB_PAGEID
880
881	/* attribute fields. rwx for SUPERVISOR mode */
882	li	r5, 0
883	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
884
885	/* Create 1:1 mapping in 256M pages */
886	xori	r7, r7, 1			/* Revert back to Original TS */
887
888	li	r8, 0				/* PageNumber */
889	li	r6, 3				/* TLB Index, start at 3  */
890
891next_tlb:
892	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
893	mr	r4, r3				/* RPN = EPN  */
894	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
895	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
896
897	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
898	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
899	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
900
901	addi	r8, r8, 1			/* Increment PN */
902	addi	r6, r6, 1			/* Increment TLB Index */
903	cmpwi	r8, 8				/* Are we done ? */
904	bne	next_tlb
905	isync
906
907	/* Jump to the new mapping 1:1 */
908	li	r9,0
909	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
910
911	bl	1f
9121:	mflr	r8
913	and	r8, r8, r11			/* Get our offset within page */
914	addi	r8, r8, (2f-1b)
915
916	and	r5, r25, r10			/* Get our target PageNum */
917	or	r8, r8, r5			/* Target jump address */
918
919	mtspr	SPRN_SRR0, r8
920	mtspr	SPRN_SRR1, r9
921	rfi
9222:
923	/* Invalidate the tmp entry we used */
924	li	r3, 0
925	tlbwe	r3, r24, PPC44x_TLB_PAGEID
926	sync
927	b	ppc44x_map_done
928
929#ifdef CONFIG_PPC_47x
930
931	/* 1:1 mapping for 47x */
932
933setup_map_47x:
934
935	/*
936	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
937	 * Also set the MSR IS->MMUCR STS
938	 */
939	li	r3, 0
940	mtspr	SPRN_PID, r3			/* Set PID */
941	mfmsr	r4				/* Get MSR */
942	andi.	r4, r4, MSR_IS@l		/* TS=1? */
943	beq	1f				/* If not, leave STS=0 */
944	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
9451:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
946	sync
947
948	/* Find the entry we are running from */
949	bl	2f
9502:	mflr	r23
951	tlbsx	r23, 0, r23
952	tlbre	r24, r23, 0			/* TLB Word 0 */
953	tlbre	r25, r23, 1			/* TLB Word 1 */
954	tlbre	r26, r23, 2			/* TLB Word 2 */
955
956
957	/*
958	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
959	 * of 4k page size in all  4 ways (0-3 in r3).
960	 * This would invalidate the entire UTLB including the one we are
961	 * running from. However the shadow TLB entries would help us
962	 * to continue the execution, until we flush them (rfi/isync).
963	 */
964	addis	r3, 0, 0x8000			/* specify the way */
965	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
966	addi	r5, 0, 0
967	b	clear_utlb_entry
968
969	/* Align the loop to speed things up. from head_44x.S */
970	.align	6
971
972clear_utlb_entry:
973
974	tlbwe	r4, r3, 0
975	tlbwe	r5, r3, 1
976	tlbwe	r5, r3, 2
977	addis	r3, r3, 0x2000			/* Increment the way */
978	cmpwi	r3, 0
979	bne	clear_utlb_entry
980	addis	r3, 0, 0x8000
981	addis	r4, r4, 0x100			/* Increment the EPN */
982	cmpwi	r4, 0
983	bne	clear_utlb_entry
984
985	/* Create the entries in the other address space */
986	mfmsr	r5
987	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
988	xori	r7, r7, 1			/* r7 = !TS */
989
990	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
991
992	/*
993	 * write out the TLB entries for the tmp mapping
994	 * Use way '0' so that we could easily invalidate it later.
995	 */
996	lis	r3, 0x8000			/* Way '0' */
997
998	tlbwe	r24, r3, 0
999	tlbwe	r25, r3, 1
1000	tlbwe	r26, r3, 2
1001
1002	/* Update the msr to the new TS */
1003	insrwi	r5, r7, 1, 26
1004
1005	bl	1f
10061:	mflr	r6
1007	addi	r6, r6, (2f-1b)
1008
1009	mtspr	SPRN_SRR0, r6
1010	mtspr	SPRN_SRR1, r5
1011	rfi
1012
1013	/*
1014	 * Now we are in the tmp address space.
1015	 * Create a 1:1 mapping for 0-2GiB in the original TS.
1016	 */
10172:
1018	li	r3, 0
1019	li	r4, 0				/* TLB Word 0 */
1020	li	r5, 0				/* TLB Word 1 */
1021	li	r6, 0
1022	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
1023
1024	li	r8, 0				/* PageIndex */
1025
1026	xori	r7, r7, 1			/* revert back to original TS */
1027
1028write_utlb:
1029	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
1030						/* ERPN = 0 as we don't use memory above 2G */
1031
1032	mr	r4, r5				/* EPN = RPN */
1033	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
1034	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
1035
1036	tlbwe	r4, r3, 0			/* Write out the entries */
1037	tlbwe	r5, r3, 1
1038	tlbwe	r6, r3, 2
1039	addi	r8, r8, 1
1040	cmpwi	r8, 8				/* Have we completed ? */
1041	bne	write_utlb
1042
1043	/* make sure we complete the TLB write up */
1044	isync
1045
1046	/*
1047	 * Prepare to jump to the 1:1 mapping.
1048	 * 1) Extract page size of the tmp mapping
1049	 *    DSIZ = TLB_Word0[22:27]
1050	 * 2) Calculate the physical address of the address
1051	 *    to jump to.
1052	 */
1053	rlwinm	r10, r24, 0, 22, 27
1054
1055	cmpwi	r10, PPC47x_TLB0_4K
1056	bne	0f
1057	li	r10, 0x1000			/* r10 = 4k */
1058	bl	1f
1059
10600:
1061	/* Defaults to 256M */
1062	lis	r10, 0x1000
1063
1064	bl	1f
10651:	mflr	r4
1066	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
1067
1068	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
1069	not	r10, r11			/* Pagemask = ~(offsetmask) */
1070
1071	and	r5, r25, r10			/* Physical page */
1072	and	r6, r4, r11			/* offset within the current page */
1073
1074	or	r5, r5, r6			/* Physical address for 2f */
1075
1076	/* Switch the TS in MSR to the original one */
1077	mfmsr	r8
1078	insrwi	r8, r7, 1, 26
1079
1080	mtspr	SPRN_SRR1, r8
1081	mtspr	SPRN_SRR0, r5
1082	rfi
1083
10842:
1085	/* Invalidate the tmp mapping */
1086	lis	r3, 0x8000			/* Way '0' */
1087
1088	clrrwi	r24, r24, 12			/* Clear the valid bit */
1089	tlbwe	r24, r3, 0
1090	tlbwe	r25, r3, 1
1091	tlbwe	r26, r3, 2
1092
1093	/* Make sure we complete the TLB write and flush the shadow TLB */
1094	isync
1095
1096#endif
1097
1098ppc44x_map_done:
1099
1100
1101	/* Restore the parameters */
1102	mr	r3, r29
1103	mr	r4, r30
1104	mr	r5, r31
1105
1106	li	r0, 0
1107#else
1108	li	r0, 0
1109
1110	/*
1111	 * Set Machine Status Register to a known status,
1112	 * switch the MMU off and jump to 1: in a single step.
1113	 */
1114
1115	mr	r8, r0
1116	ori     r8, r8, MSR_RI|MSR_ME
1117	mtspr	SPRN_SRR1, r8
1118	addi	r8, r4, 1f - relocate_new_kernel
1119	mtspr	SPRN_SRR0, r8
1120	sync
1121	rfi
1122
11231:
1124#endif
1125	/* from this point address translation is turned off */
1126	/* and interrupts are disabled */
1127
1128	/* set a new stack at the bottom of our page... */
1129	/* (not really needed now) */
1130	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1131	stw	r0, 0(r1)
1132
1133	/* Do the copies */
1134	li	r6, 0 /* checksum */
1135	mr	r0, r3
1136	b	1f
1137
11380:	/* top, read another word for the indirection page */
1139	lwzu	r0, 4(r3)
1140
11411:
1142	/* is it a destination page? (r8) */
1143	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1144	beq	2f
1145
1146	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1147	b	0b
1148
11492:	/* is it an indirection page? (r3) */
1150	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1151	beq	2f
1152
1153	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1154	subi	r3, r3, 4
1155	b	0b
1156
11572:	/* are we done? */
1158	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1159	beq	2f
1160	b	3f
1161
11622:	/* is it a source page? (r9) */
1163	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1164	beq	0b
1165
1166	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1167
1168	li	r7, PAGE_SIZE / 4
1169	mtctr   r7
1170	subi    r9, r9, 4
1171	subi    r8, r8, 4
11729:
1173	lwzu    r0, 4(r9)  /* do the copy */
1174	xor	r6, r6, r0
1175	stwu    r0, 4(r8)
1176	dcbst	0, r8
1177	sync
1178	icbi	0, r8
1179	bdnz    9b
1180
1181	addi    r9, r9, 4
1182	addi    r8, r8, 4
1183	b	0b
1184
11853:
1186
1187	/* To be certain of avoiding problems with self-modifying code
1188	 * execute a serializing instruction here.
1189	 */
1190	isync
1191	sync
1192
1193	mfspr	r3, SPRN_PIR /* current core we are running on */
1194	mr	r4, r5 /* load physical address of chunk called */
1195
1196	/* jump to the entry point, usually the setup routine */
1197	mtlr	r5
1198	blrl
1199
12001:	b	1b
1201
1202relocate_new_kernel_end:
1203
1204	.globl relocate_new_kernel_size
1205relocate_new_kernel_size:
1206	.long relocate_new_kernel_end - relocate_new_kernel
1207#endif
1208