xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision f677b30b487ca3763c3de3f1b4d8c976c2961cd1)
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	isync
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	blr
452END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
453	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
454	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
455	mtctr	r4
456	mr	r6,r3
4570:	dcbst	0,r3				/* Write line to ram */
458	addi	r3,r3,L1_CACHE_BYTES
459	bdnz	0b
460	sync
461#ifdef CONFIG_44x
462	/* We don't flush the icache on 44x. Those have a virtual icache
463	 * and we don't have access to the virtual address here (it's
464	 * not the page vaddr but where it's mapped in user space). The
465	 * flushing of the icache on these is handled elsewhere, when
466	 * a change in the address space occurs, before returning to
467	 * user space
468	 */
469BEGIN_MMU_FTR_SECTION
470	blr
471END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
472#endif /* CONFIG_44x */
473	mtctr	r4
4741:	icbi	0,r6
475	addi	r6,r6,L1_CACHE_BYTES
476	bdnz	1b
477	sync
478	isync
479	blr
480
481#ifndef CONFIG_BOOKE
482/*
483 * Flush a particular page from the data cache to RAM, identified
484 * by its physical address.  We turn off the MMU so we can just use
485 * the physical address (this may be a highmem page without a kernel
486 * mapping).
487 *
488 *	void __flush_dcache_icache_phys(unsigned long physaddr)
489 */
490_GLOBAL(__flush_dcache_icache_phys)
491BEGIN_FTR_SECTION
492	blr					/* for 601, do nothing */
493END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
494	mfmsr	r10
495	rlwinm	r0,r10,0,28,26			/* clear DR */
496	mtmsr	r0
497	isync
498	rlwinm	r3,r3,0,0,31-PAGE_SHIFT		/* Get page base address */
499	li	r4,PAGE_SIZE/L1_CACHE_BYTES	/* Number of lines in a page */
500	mtctr	r4
501	mr	r6,r3
5020:	dcbst	0,r3				/* Write line to ram */
503	addi	r3,r3,L1_CACHE_BYTES
504	bdnz	0b
505	sync
506	mtctr	r4
5071:	icbi	0,r6
508	addi	r6,r6,L1_CACHE_BYTES
509	bdnz	1b
510	sync
511	mtmsr	r10				/* restore DR */
512	isync
513	blr
514#endif /* CONFIG_BOOKE */
515
516/*
517 * Clear pages using the dcbz instruction, which doesn't cause any
518 * memory traffic (except to write out any cache lines which get
519 * displaced).  This only works on cacheable memory.
520 *
521 * void clear_pages(void *page, int order) ;
522 */
523_GLOBAL(clear_pages)
524	li	r0,PAGE_SIZE/L1_CACHE_BYTES
525	slw	r0,r0,r4
526	mtctr	r0
5271:	dcbz	0,r3
528	addi	r3,r3,L1_CACHE_BYTES
529	bdnz	1b
530	blr
531
532/*
533 * Copy a whole page.  We use the dcbz instruction on the destination
534 * to reduce memory traffic (it eliminates the unnecessary reads of
535 * the destination into cache).  This requires that the destination
536 * is cacheable.
537 */
538#define COPY_16_BYTES		\
539	lwz	r6,4(r4);	\
540	lwz	r7,8(r4);	\
541	lwz	r8,12(r4);	\
542	lwzu	r9,16(r4);	\
543	stw	r6,4(r3);	\
544	stw	r7,8(r3);	\
545	stw	r8,12(r3);	\
546	stwu	r9,16(r3)
547
548_GLOBAL(copy_page)
549	addi	r3,r3,-4
550	addi	r4,r4,-4
551
552	li	r5,4
553
554#if MAX_COPY_PREFETCH > 1
555	li	r0,MAX_COPY_PREFETCH
556	li	r11,4
557	mtctr	r0
55811:	dcbt	r11,r4
559	addi	r11,r11,L1_CACHE_BYTES
560	bdnz	11b
561#else /* MAX_COPY_PREFETCH == 1 */
562	dcbt	r5,r4
563	li	r11,L1_CACHE_BYTES+4
564#endif /* MAX_COPY_PREFETCH */
565	li	r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
566	crclr	4*cr0+eq
5672:
568	mtctr	r0
5691:
570	dcbt	r11,r4
571	dcbz	r5,r3
572	COPY_16_BYTES
573#if L1_CACHE_BYTES >= 32
574	COPY_16_BYTES
575#if L1_CACHE_BYTES >= 64
576	COPY_16_BYTES
577	COPY_16_BYTES
578#if L1_CACHE_BYTES >= 128
579	COPY_16_BYTES
580	COPY_16_BYTES
581	COPY_16_BYTES
582	COPY_16_BYTES
583#endif
584#endif
585#endif
586	bdnz	1b
587	beqlr
588	crnot	4*cr0+eq,4*cr0+eq
589	li	r0,MAX_COPY_PREFETCH
590	li	r11,4
591	b	2b
592
593/*
594 * void atomic_clear_mask(atomic_t mask, atomic_t *addr)
595 * void atomic_set_mask(atomic_t mask, atomic_t *addr);
596 */
597_GLOBAL(atomic_clear_mask)
59810:	lwarx	r5,0,r4
599	andc	r5,r5,r3
600	PPC405_ERR77(0,r4)
601	stwcx.	r5,0,r4
602	bne-	10b
603	blr
604_GLOBAL(atomic_set_mask)
60510:	lwarx	r5,0,r4
606	or	r5,r5,r3
607	PPC405_ERR77(0,r4)
608	stwcx.	r5,0,r4
609	bne-	10b
610	blr
611
612/*
613 * Extended precision shifts.
614 *
615 * Updated to be valid for shift counts from 0 to 63 inclusive.
616 * -- Gabriel
617 *
618 * R3/R4 has 64 bit value
619 * R5    has shift count
620 * result in R3/R4
621 *
622 *  ashrdi3: arithmetic right shift (sign propagation)
623 *  lshrdi3: logical right shift
624 *  ashldi3: left shift
625 */
626_GLOBAL(__ashrdi3)
627	subfic	r6,r5,32
628	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
629	addi	r7,r5,32	# could be xori, or addi with -32
630	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
631	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
632	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
633	or	r4,r4,r6	# LSW |= t1
634	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
635	sraw	r3,r3,r5	# MSW = MSW >> count
636	or	r4,r4,r7	# LSW |= t2
637	blr
638
639_GLOBAL(__ashldi3)
640	subfic	r6,r5,32
641	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
642	addi	r7,r5,32	# could be xori, or addi with -32
643	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
644	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
645	or	r3,r3,r6	# MSW |= t1
646	slw	r4,r4,r5	# LSW = LSW << count
647	or	r3,r3,r7	# MSW |= t2
648	blr
649
650_GLOBAL(__lshrdi3)
651	subfic	r6,r5,32
652	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
653	addi	r7,r5,32	# could be xori, or addi with -32
654	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
655	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
656	or	r4,r4,r6	# LSW |= t1
657	srw	r3,r3,r5	# MSW = MSW >> count
658	or	r4,r4,r7	# LSW |= t2
659	blr
660
661/*
662 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
663 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
664 */
665_GLOBAL(__cmpdi2)
666	cmpw	r3,r5
667	li	r3,1
668	bne	1f
669	cmplw	r4,r6
670	beqlr
6711:	li	r3,0
672	bltlr
673	li	r3,2
674	blr
675/*
676 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
677 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
678 */
679_GLOBAL(__ucmpdi2)
680	cmplw	r3,r5
681	li	r3,1
682	bne	1f
683	cmplw	r4,r6
684	beqlr
6851:	li	r3,0
686	bltlr
687	li	r3,2
688	blr
689
690_GLOBAL(__bswapdi2)
691	rotlwi  r9,r4,8
692	rotlwi  r10,r3,8
693	rlwimi  r9,r4,24,0,7
694	rlwimi  r10,r3,24,0,7
695	rlwimi  r9,r4,24,16,23
696	rlwimi  r10,r3,24,16,23
697	mr      r3,r9
698	mr      r4,r10
699	blr
700
701_GLOBAL(abs)
702	srawi	r4,r3,31
703	xor	r3,r3,r4
704	sub	r3,r3,r4
705	blr
706
707#ifdef CONFIG_SMP
708_GLOBAL(start_secondary_resume)
709	/* Reset stack */
710	CURRENT_THREAD_INFO(r1, r1)
711	addi	r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
712	li	r3,0
713	stw	r3,0(r1)		/* Zero the stack frame pointer	*/
714	bl	start_secondary
715	b	.
716#endif /* CONFIG_SMP */
717
718/*
719 * This routine is just here to keep GCC happy - sigh...
720 */
721_GLOBAL(__main)
722	blr
723
724#ifdef CONFIG_KEXEC
725	/*
726	 * Must be relocatable PIC code callable as a C function.
727	 */
728	.globl relocate_new_kernel
729relocate_new_kernel:
730	/* r3 = page_list   */
731	/* r4 = reboot_code_buffer */
732	/* r5 = start_address      */
733
734#ifdef CONFIG_FSL_BOOKE
735
736	mr	r29, r3
737	mr	r30, r4
738	mr	r31, r5
739
740#define ENTRY_MAPPING_KEXEC_SETUP
741#include "fsl_booke_entry_mapping.S"
742#undef ENTRY_MAPPING_KEXEC_SETUP
743
744	mr      r3, r29
745	mr      r4, r30
746	mr      r5, r31
747
748	li	r0, 0
749#elif defined(CONFIG_44x)
750
751	/* Save our parameters */
752	mr	r29, r3
753	mr	r30, r4
754	mr	r31, r5
755
756#ifdef CONFIG_PPC_47x
757	/* Check for 47x cores */
758	mfspr	r3,SPRN_PVR
759	srwi	r3,r3,16
760	cmplwi	cr0,r3,PVR_476@h
761	beq	setup_map_47x
762	cmplwi	cr0,r3,PVR_476_ISS@h
763	beq	setup_map_47x
764#endif /* CONFIG_PPC_47x */
765
766/*
767 * Code for setting up 1:1 mapping for PPC440x for KEXEC
768 *
769 * We cannot switch off the MMU on PPC44x.
770 * So we:
771 * 1) Invalidate all the mappings except the one we are running from.
772 * 2) Create a tmp mapping for our code in the other address space(TS) and
773 *    jump to it. Invalidate the entry we started in.
774 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
775 * 4) Jump to the 1:1 mapping in original TS.
776 * 5) Invalidate the tmp mapping.
777 *
778 * - Based on the kexec support code for FSL BookE
779 *
780 */
781
782	/*
783	 * Load the PID with kernel PID (0).
784	 * Also load our MSR_IS and TID to MMUCR for TLB search.
785	 */
786	li	r3, 0
787	mtspr	SPRN_PID, r3
788	mfmsr	r4
789	andi.	r4,r4,MSR_IS@l
790	beq	wmmucr
791	oris	r3,r3,PPC44x_MMUCR_STS@h
792wmmucr:
793	mtspr	SPRN_MMUCR,r3
794	sync
795
796	/*
797	 * Invalidate all the TLB entries except the current entry
798	 * where we are running from
799	 */
800	bl	0f				/* Find our address */
8010:	mflr	r5				/* Make it accessible */
802	tlbsx	r23,0,r5			/* Find entry we are in */
803	li	r4,0				/* Start at TLB entry 0 */
804	li	r3,0				/* Set PAGEID inval value */
8051:	cmpw	r23,r4				/* Is this our entry? */
806	beq	skip				/* If so, skip the inval */
807	tlbwe	r3,r4,PPC44x_TLB_PAGEID		/* If not, inval the entry */
808skip:
809	addi	r4,r4,1				/* Increment */
810	cmpwi	r4,64				/* Are we done?	*/
811	bne	1b				/* If not, repeat */
812	isync
813
814	/* Create a temp mapping and jump to it */
815	andi.	r6, r23, 1		/* Find the index to use */
816	addi	r24, r6, 1		/* r24 will contain 1 or 2 */
817
818	mfmsr	r9			/* get the MSR */
819	rlwinm	r5, r9, 27, 31, 31	/* Extract the MSR[IS] */
820	xori	r7, r5, 1		/* Use the other address space */
821
822	/* Read the current mapping entries */
823	tlbre	r3, r23, PPC44x_TLB_PAGEID
824	tlbre	r4, r23, PPC44x_TLB_XLAT
825	tlbre	r5, r23, PPC44x_TLB_ATTRIB
826
827	/* Save our current XLAT entry */
828	mr	r25, r4
829
830	/* Extract the TLB PageSize */
831	li	r10, 1 			/* r10 will hold PageSize */
832	rlwinm	r11, r3, 0, 24, 27	/* bits 24-27 */
833
834	/* XXX: As of now we use 256M, 4K pages */
835	cmpwi	r11, PPC44x_TLB_256M
836	bne	tlb_4k
837	rotlwi	r10, r10, 28		/* r10 = 256M */
838	b	write_out
839tlb_4k:
840	cmpwi	r11, PPC44x_TLB_4K
841	bne	default
842	rotlwi	r10, r10, 12		/* r10 = 4K */
843	b	write_out
844default:
845	rotlwi	r10, r10, 10		/* r10 = 1K */
846
847write_out:
848	/*
849	 * Write out the tmp 1:1 mapping for this code in other address space
850	 * Fixup  EPN = RPN , TS=other address space
851	 */
852	insrwi	r3, r7, 1, 23		/* Bit 23 is TS for PAGEID field */
853
854	/* Write out the tmp mapping entries */
855	tlbwe	r3, r24, PPC44x_TLB_PAGEID
856	tlbwe	r4, r24, PPC44x_TLB_XLAT
857	tlbwe	r5, r24, PPC44x_TLB_ATTRIB
858
859	subi	r11, r10, 1		/* PageOffset Mask = PageSize - 1 */
860	not	r10, r11		/* Mask for PageNum */
861
862	/* Switch to other address space in MSR */
863	insrwi	r9, r7, 1, 26		/* Set MSR[IS] = r7 */
864
865	bl	1f
8661:	mflr	r8
867	addi	r8, r8, (2f-1b)		/* Find the target offset */
868
869	/* Jump to the tmp mapping */
870	mtspr	SPRN_SRR0, r8
871	mtspr	SPRN_SRR1, r9
872	rfi
873
8742:
875	/* Invalidate the entry we were executing from */
876	li	r3, 0
877	tlbwe	r3, r23, PPC44x_TLB_PAGEID
878
879	/* attribute fields. rwx for SUPERVISOR mode */
880	li	r5, 0
881	ori	r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
882
883	/* Create 1:1 mapping in 256M pages */
884	xori	r7, r7, 1			/* Revert back to Original TS */
885
886	li	r8, 0				/* PageNumber */
887	li	r6, 3				/* TLB Index, start at 3  */
888
889next_tlb:
890	rotlwi	r3, r8, 28			/* Create EPN (bits 0-3) */
891	mr	r4, r3				/* RPN = EPN  */
892	ori	r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
893	insrwi	r3, r7, 1, 23			/* Set TS from r7 */
894
895	tlbwe	r3, r6, PPC44x_TLB_PAGEID	/* PageID field : EPN, V, SIZE */
896	tlbwe	r4, r6, PPC44x_TLB_XLAT		/* Address translation : RPN   */
897	tlbwe	r5, r6, PPC44x_TLB_ATTRIB	/* Attributes */
898
899	addi	r8, r8, 1			/* Increment PN */
900	addi	r6, r6, 1			/* Increment TLB Index */
901	cmpwi	r8, 8				/* Are we done ? */
902	bne	next_tlb
903	isync
904
905	/* Jump to the new mapping 1:1 */
906	li	r9,0
907	insrwi	r9, r7, 1, 26			/* Set MSR[IS] = r7 */
908
909	bl	1f
9101:	mflr	r8
911	and	r8, r8, r11			/* Get our offset within page */
912	addi	r8, r8, (2f-1b)
913
914	and	r5, r25, r10			/* Get our target PageNum */
915	or	r8, r8, r5			/* Target jump address */
916
917	mtspr	SPRN_SRR0, r8
918	mtspr	SPRN_SRR1, r9
919	rfi
9202:
921	/* Invalidate the tmp entry we used */
922	li	r3, 0
923	tlbwe	r3, r24, PPC44x_TLB_PAGEID
924	sync
925	b	ppc44x_map_done
926
927#ifdef CONFIG_PPC_47x
928
929	/* 1:1 mapping for 47x */
930
931setup_map_47x:
932
933	/*
934	 * Load the kernel pid (0) to PID and also to MMUCR[TID].
935	 * Also set the MSR IS->MMUCR STS
936	 */
937	li	r3, 0
938	mtspr	SPRN_PID, r3			/* Set PID */
939	mfmsr	r4				/* Get MSR */
940	andi.	r4, r4, MSR_IS@l		/* TS=1? */
941	beq	1f				/* If not, leave STS=0 */
942	oris	r3, r3, PPC47x_MMUCR_STS@h	/* Set STS=1 */
9431:	mtspr	SPRN_MMUCR, r3			/* Put MMUCR */
944	sync
945
946	/* Find the entry we are running from */
947	bl	2f
9482:	mflr	r23
949	tlbsx	r23, 0, r23
950	tlbre	r24, r23, 0			/* TLB Word 0 */
951	tlbre	r25, r23, 1			/* TLB Word 1 */
952	tlbre	r26, r23, 2			/* TLB Word 2 */
953
954
955	/*
956	 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
957	 * of 4k page size in all  4 ways (0-3 in r3).
958	 * This would invalidate the entire UTLB including the one we are
959	 * running from. However the shadow TLB entries would help us
960	 * to continue the execution, until we flush them (rfi/isync).
961	 */
962	addis	r3, 0, 0x8000			/* specify the way */
963	addi	r4, 0, 0			/* TLB Word0 = (EPN=0, VALID = 0) */
964	addi	r5, 0, 0
965	b	clear_utlb_entry
966
967	/* Align the loop to speed things up. from head_44x.S */
968	.align	6
969
970clear_utlb_entry:
971
972	tlbwe	r4, r3, 0
973	tlbwe	r5, r3, 1
974	tlbwe	r5, r3, 2
975	addis	r3, r3, 0x2000			/* Increment the way */
976	cmpwi	r3, 0
977	bne	clear_utlb_entry
978	addis	r3, 0, 0x8000
979	addis	r4, r4, 0x100			/* Increment the EPN */
980	cmpwi	r4, 0
981	bne	clear_utlb_entry
982
983	/* Create the entries in the other address space */
984	mfmsr	r5
985	rlwinm	r7, r5, 27, 31, 31		/* Get the TS (Bit 26) from MSR */
986	xori	r7, r7, 1			/* r7 = !TS */
987
988	insrwi	r24, r7, 1, 21			/* Change the TS in the saved TLB word 0 */
989
990	/*
991	 * write out the TLB entries for the tmp mapping
992	 * Use way '0' so that we could easily invalidate it later.
993	 */
994	lis	r3, 0x8000			/* Way '0' */
995
996	tlbwe	r24, r3, 0
997	tlbwe	r25, r3, 1
998	tlbwe	r26, r3, 2
999
1000	/* Update the msr to the new TS */
1001	insrwi	r5, r7, 1, 26
1002
1003	bl	1f
10041:	mflr	r6
1005	addi	r6, r6, (2f-1b)
1006
1007	mtspr	SPRN_SRR0, r6
1008	mtspr	SPRN_SRR1, r5
1009	rfi
1010
1011	/*
1012	 * Now we are in the tmp address space.
1013	 * Create a 1:1 mapping for 0-2GiB in the original TS.
1014	 */
10152:
1016	li	r3, 0
1017	li	r4, 0				/* TLB Word 0 */
1018	li	r5, 0				/* TLB Word 1 */
1019	li	r6, 0
1020	ori	r6, r6, PPC47x_TLB2_S_RWX	/* TLB word 2 */
1021
1022	li	r8, 0				/* PageIndex */
1023
1024	xori	r7, r7, 1			/* revert back to original TS */
1025
1026write_utlb:
1027	rotlwi	r5, r8, 28			/* RPN = PageIndex * 256M */
1028						/* ERPN = 0 as we don't use memory above 2G */
1029
1030	mr	r4, r5				/* EPN = RPN */
1031	ori	r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
1032	insrwi	r4, r7, 1, 21			/* Insert the TS to Word 0 */
1033
1034	tlbwe	r4, r3, 0			/* Write out the entries */
1035	tlbwe	r5, r3, 1
1036	tlbwe	r6, r3, 2
1037	addi	r8, r8, 1
1038	cmpwi	r8, 8				/* Have we completed ? */
1039	bne	write_utlb
1040
1041	/* make sure we complete the TLB write up */
1042	isync
1043
1044	/*
1045	 * Prepare to jump to the 1:1 mapping.
1046	 * 1) Extract page size of the tmp mapping
1047	 *    DSIZ = TLB_Word0[22:27]
1048	 * 2) Calculate the physical address of the address
1049	 *    to jump to.
1050	 */
1051	rlwinm	r10, r24, 0, 22, 27
1052
1053	cmpwi	r10, PPC47x_TLB0_4K
1054	bne	0f
1055	li	r10, 0x1000			/* r10 = 4k */
1056	bl	1f
1057
10580:
1059	/* Defaults to 256M */
1060	lis	r10, 0x1000
1061
1062	bl	1f
10631:	mflr	r4
1064	addi	r4, r4, (2f-1b)			/* virtual address  of 2f */
1065
1066	subi	r11, r10, 1			/* offsetmask = Pagesize - 1 */
1067	not	r10, r11			/* Pagemask = ~(offsetmask) */
1068
1069	and	r5, r25, r10			/* Physical page */
1070	and	r6, r4, r11			/* offset within the current page */
1071
1072	or	r5, r5, r6			/* Physical address for 2f */
1073
1074	/* Switch the TS in MSR to the original one */
1075	mfmsr	r8
1076	insrwi	r8, r7, 1, 26
1077
1078	mtspr	SPRN_SRR1, r8
1079	mtspr	SPRN_SRR0, r5
1080	rfi
1081
10822:
1083	/* Invalidate the tmp mapping */
1084	lis	r3, 0x8000			/* Way '0' */
1085
1086	clrrwi	r24, r24, 12			/* Clear the valid bit */
1087	tlbwe	r24, r3, 0
1088	tlbwe	r25, r3, 1
1089	tlbwe	r26, r3, 2
1090
1091	/* Make sure we complete the TLB write and flush the shadow TLB */
1092	isync
1093
1094#endif
1095
1096ppc44x_map_done:
1097
1098
1099	/* Restore the parameters */
1100	mr	r3, r29
1101	mr	r4, r30
1102	mr	r5, r31
1103
1104	li	r0, 0
1105#else
1106	li	r0, 0
1107
1108	/*
1109	 * Set Machine Status Register to a known status,
1110	 * switch the MMU off and jump to 1: in a single step.
1111	 */
1112
1113	mr	r8, r0
1114	ori     r8, r8, MSR_RI|MSR_ME
1115	mtspr	SPRN_SRR1, r8
1116	addi	r8, r4, 1f - relocate_new_kernel
1117	mtspr	SPRN_SRR0, r8
1118	sync
1119	rfi
1120
11211:
1122#endif
1123	/* from this point address translation is turned off */
1124	/* and interrupts are disabled */
1125
1126	/* set a new stack at the bottom of our page... */
1127	/* (not really needed now) */
1128	addi	r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1129	stw	r0, 0(r1)
1130
1131	/* Do the copies */
1132	li	r6, 0 /* checksum */
1133	mr	r0, r3
1134	b	1f
1135
11360:	/* top, read another word for the indirection page */
1137	lwzu	r0, 4(r3)
1138
11391:
1140	/* is it a destination page? (r8) */
1141	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1142	beq	2f
1143
1144	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1145	b	0b
1146
11472:	/* is it an indirection page? (r3) */
1148	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1149	beq	2f
1150
1151	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1152	subi	r3, r3, 4
1153	b	0b
1154
11552:	/* are we done? */
1156	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1157	beq	2f
1158	b	3f
1159
11602:	/* is it a source page? (r9) */
1161	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1162	beq	0b
1163
1164	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1165
1166	li	r7, PAGE_SIZE / 4
1167	mtctr   r7
1168	subi    r9, r9, 4
1169	subi    r8, r8, 4
11709:
1171	lwzu    r0, 4(r9)  /* do the copy */
1172	xor	r6, r6, r0
1173	stwu    r0, 4(r8)
1174	dcbst	0, r8
1175	sync
1176	icbi	0, r8
1177	bdnz    9b
1178
1179	addi    r9, r9, 4
1180	addi    r8, r8, 4
1181	b	0b
1182
11833:
1184
1185	/* To be certain of avoiding problems with self-modifying code
1186	 * execute a serializing instruction here.
1187	 */
1188	isync
1189	sync
1190
1191	mfspr	r3, SPRN_PIR /* current core we are running on */
1192	mr	r4, r5 /* load physical address of chunk called */
1193
1194	/* jump to the entry point, usually the setup routine */
1195	mtlr	r5
1196	blrl
1197
11981:	b	1b
1199
1200relocate_new_kernel_end:
1201
1202	.globl relocate_new_kernel_size
1203relocate_new_kernel_size:
1204	.long relocate_new_kernel_end - relocate_new_kernel
1205#endif
1206