xref: /openbmc/linux/arch/powerpc/kernel/misc_32.S (revision 545e4006)
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 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public License
14 * as published by the Free Software Foundation; either version
15 * 2 of the License, or (at your option) any later version.
16 *
17 */
18
19#include <linux/sys.h>
20#include <asm/unistd.h>
21#include <asm/errno.h>
22#include <asm/reg.h>
23#include <asm/page.h>
24#include <asm/cache.h>
25#include <asm/cputable.h>
26#include <asm/mmu.h>
27#include <asm/ppc_asm.h>
28#include <asm/thread_info.h>
29#include <asm/asm-offsets.h>
30#include <asm/processor.h>
31#include <asm/kexec.h>
32
33	.text
34
35#ifdef CONFIG_IRQSTACKS
36_GLOBAL(call_do_softirq)
37	mflr	r0
38	stw	r0,4(r1)
39	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
40	mr	r1,r3
41	bl	__do_softirq
42	lwz	r1,0(r1)
43	lwz	r0,4(r1)
44	mtlr	r0
45	blr
46
47_GLOBAL(call_handle_irq)
48	mflr	r0
49	stw	r0,4(r1)
50	mtctr	r6
51	stwu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r5)
52	mr	r1,r5
53	bctrl
54	lwz	r1,0(r1)
55	lwz	r0,4(r1)
56	mtlr	r0
57	blr
58#endif /* CONFIG_IRQSTACKS */
59
60/*
61 * This returns the high 64 bits of the product of two 64-bit numbers.
62 */
63_GLOBAL(mulhdu)
64	cmpwi	r6,0
65	cmpwi	cr1,r3,0
66	mr	r10,r4
67	mulhwu	r4,r4,r5
68	beq	1f
69	mulhwu	r0,r10,r6
70	mullw	r7,r10,r5
71	addc	r7,r0,r7
72	addze	r4,r4
731:	beqlr	cr1		/* all done if high part of A is 0 */
74	mr	r10,r3
75	mullw	r9,r3,r5
76	mulhwu	r3,r3,r5
77	beq	2f
78	mullw	r0,r10,r6
79	mulhwu	r8,r10,r6
80	addc	r7,r0,r7
81	adde	r4,r4,r8
82	addze	r3,r3
832:	addc	r4,r4,r9
84	addze	r3,r3
85	blr
86
87/*
88 * sub_reloc_offset(x) returns x - reloc_offset().
89 */
90_GLOBAL(sub_reloc_offset)
91	mflr	r0
92	bl	1f
931:	mflr	r5
94	lis	r4,1b@ha
95	addi	r4,r4,1b@l
96	subf	r5,r4,r5
97	subf	r3,r5,r3
98	mtlr	r0
99	blr
100
101/*
102 * reloc_got2 runs through the .got2 section adding an offset
103 * to each entry.
104 */
105_GLOBAL(reloc_got2)
106	mflr	r11
107	lis	r7,__got2_start@ha
108	addi	r7,r7,__got2_start@l
109	lis	r8,__got2_end@ha
110	addi	r8,r8,__got2_end@l
111	subf	r8,r7,r8
112	srwi.	r8,r8,2
113	beqlr
114	mtctr	r8
115	bl	1f
1161:	mflr	r0
117	lis	r4,1b@ha
118	addi	r4,r4,1b@l
119	subf	r0,r4,r0
120	add	r7,r0,r7
1212:	lwz	r0,0(r7)
122	add	r0,r0,r3
123	stw	r0,0(r7)
124	addi	r7,r7,4
125	bdnz	2b
126	mtlr	r11
127	blr
128
129/*
130 * call_setup_cpu - call the setup_cpu function for this cpu
131 * r3 = data offset, r24 = cpu number
132 *
133 * Setup function is called with:
134 *   r3 = data offset
135 *   r4 = ptr to CPU spec (relocated)
136 */
137_GLOBAL(call_setup_cpu)
138	addis	r4,r3,cur_cpu_spec@ha
139	addi	r4,r4,cur_cpu_spec@l
140	lwz	r4,0(r4)
141	add	r4,r4,r3
142	lwz	r5,CPU_SPEC_SETUP(r4)
143	cmpwi	0,r5,0
144	add	r5,r5,r3
145	beqlr
146	mtctr	r5
147	bctr
148
149#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
150
151/* This gets called by via-pmu.c to switch the PLL selection
152 * on 750fx CPU. This function should really be moved to some
153 * other place (as most of the cpufreq code in via-pmu
154 */
155_GLOBAL(low_choose_750fx_pll)
156	/* Clear MSR:EE */
157	mfmsr	r7
158	rlwinm	r0,r7,0,17,15
159	mtmsr	r0
160
161	/* If switching to PLL1, disable HID0:BTIC */
162	cmplwi	cr0,r3,0
163	beq	1f
164	mfspr	r5,SPRN_HID0
165	rlwinm	r5,r5,0,27,25
166	sync
167	mtspr	SPRN_HID0,r5
168	isync
169	sync
170
1711:
172	/* Calc new HID1 value */
173	mfspr	r4,SPRN_HID1	/* Build a HID1:PS bit from parameter */
174	rlwinm	r5,r3,16,15,15	/* Clear out HID1:PS from value read */
175	rlwinm	r4,r4,0,16,14	/* Could have I used rlwimi here ? */
176	or	r4,r4,r5
177	mtspr	SPRN_HID1,r4
178
179	/* Store new HID1 image */
180	rlwinm	r6,r1,0,0,(31-THREAD_SHIFT)
181	lwz	r6,TI_CPU(r6)
182	slwi	r6,r6,2
183	addis	r6,r6,nap_save_hid1@ha
184	stw	r4,nap_save_hid1@l(r6)
185
186	/* If switching to PLL0, enable HID0:BTIC */
187	cmplwi	cr0,r3,0
188	bne	1f
189	mfspr	r5,SPRN_HID0
190	ori	r5,r5,HID0_BTIC
191	sync
192	mtspr	SPRN_HID0,r5
193	isync
194	sync
195
1961:
197	/* Return */
198	mtmsr	r7
199	blr
200
201_GLOBAL(low_choose_7447a_dfs)
202	/* Clear MSR:EE */
203	mfmsr	r7
204	rlwinm	r0,r7,0,17,15
205	mtmsr	r0
206
207	/* Calc new HID1 value */
208	mfspr	r4,SPRN_HID1
209	insrwi	r4,r3,1,9	/* insert parameter into bit 9 */
210	sync
211	mtspr	SPRN_HID1,r4
212	sync
213	isync
214
215	/* Return */
216	mtmsr	r7
217	blr
218
219#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
220
221/*
222 * complement mask on the msr then "or" some values on.
223 *     _nmask_and_or_msr(nmask, value_to_or)
224 */
225_GLOBAL(_nmask_and_or_msr)
226	mfmsr	r0		/* Get current msr */
227	andc	r0,r0,r3	/* And off the bits set in r3 (first parm) */
228	or	r0,r0,r4	/* Or on the bits in r4 (second parm) */
229	SYNC			/* Some chip revs have problems here... */
230	mtmsr	r0		/* Update machine state */
231	isync
232	blr			/* Done */
233
234#ifdef CONFIG_40x
235
236/*
237 * Do an IO access in real mode
238 */
239_GLOBAL(real_readb)
240	mfmsr	r7
241	ori	r0,r7,MSR_DR
242	xori	r0,r0,MSR_DR
243	sync
244	mtmsr	r0
245	sync
246	isync
247	lbz	r3,0(r3)
248	sync
249	mtmsr	r7
250	sync
251	isync
252	blr
253
254	/*
255 * Do an IO access in real mode
256 */
257_GLOBAL(real_writeb)
258	mfmsr	r7
259	ori	r0,r7,MSR_DR
260	xori	r0,r0,MSR_DR
261	sync
262	mtmsr	r0
263	sync
264	isync
265	stb	r3,0(r4)
266	sync
267	mtmsr	r7
268	sync
269	isync
270	blr
271
272#endif /* CONFIG_40x */
273
274/*
275 * Flush MMU TLB
276 */
277_GLOBAL(_tlbia)
278#if defined(CONFIG_40x)
279	sync			/* Flush to memory before changing mapping */
280	tlbia
281	isync			/* Flush shadow TLB */
282#elif defined(CONFIG_44x)
283	li	r3,0
284	sync
285
286	/* Load high watermark */
287	lis	r4,tlb_44x_hwater@ha
288	lwz	r5,tlb_44x_hwater@l(r4)
289
2901:	tlbwe	r3,r3,PPC44x_TLB_PAGEID
291	addi	r3,r3,1
292	cmpw	0,r3,r5
293	ble	1b
294
295	isync
296#elif defined(CONFIG_FSL_BOOKE)
297	/* Invalidate all entries in TLB0 */
298	li	r3, 0x04
299	tlbivax	0,3
300	/* Invalidate all entries in TLB1 */
301	li	r3, 0x0c
302	tlbivax	0,3
303	msync
304#ifdef CONFIG_SMP
305	tlbsync
306#endif /* CONFIG_SMP */
307#else /* !(CONFIG_40x || CONFIG_44x || CONFIG_FSL_BOOKE) */
308#if defined(CONFIG_SMP)
309	rlwinm	r8,r1,0,0,(31-THREAD_SHIFT)
310	lwz	r8,TI_CPU(r8)
311	oris	r8,r8,10
312	mfmsr	r10
313	SYNC
314	rlwinm	r0,r10,0,17,15		/* clear bit 16 (MSR_EE) */
315	rlwinm	r0,r0,0,28,26		/* clear DR */
316	mtmsr	r0
317	SYNC_601
318	isync
319	lis	r9,mmu_hash_lock@h
320	ori	r9,r9,mmu_hash_lock@l
321	tophys(r9,r9)
32210:	lwarx	r7,0,r9
323	cmpwi	0,r7,0
324	bne-	10b
325	stwcx.	r8,0,r9
326	bne-	10b
327	sync
328	tlbia
329	sync
330	TLBSYNC
331	li	r0,0
332	stw	r0,0(r9)		/* clear mmu_hash_lock */
333	mtmsr	r10
334	SYNC_601
335	isync
336#else /* CONFIG_SMP */
337	sync
338	tlbia
339	sync
340#endif /* CONFIG_SMP */
341#endif /* ! defined(CONFIG_40x) */
342	blr
343
344/*
345 * Flush MMU TLB for a particular address
346 */
347_GLOBAL(_tlbie)
348#if defined(CONFIG_40x)
349	/* We run the search with interrupts disabled because we have to change
350	 * the PID and I don't want to preempt when that happens.
351	 */
352	mfmsr	r5
353	mfspr	r6,SPRN_PID
354	wrteei	0
355	mtspr	SPRN_PID,r4
356	tlbsx.	r3, 0, r3
357	mtspr	SPRN_PID,r6
358	wrtee	r5
359	bne	10f
360	sync
361	/* There are only 64 TLB entries, so r3 < 64, which means bit 25 is clear.
362	 * Since 25 is the V bit in the TLB_TAG, loading this value will invalidate
363	 * the TLB entry. */
364	tlbwe	r3, r3, TLB_TAG
365	isync
36610:
367
368#elif defined(CONFIG_44x)
369	mfspr	r5,SPRN_MMUCR
370	rlwimi	r5,r4,0,24,31			/* Set TID */
371
372	/* We have to run the search with interrupts disabled, even critical
373	 * and debug interrupts (in fact the only critical exceptions we have
374	 * are debug and machine check).  Otherwise  an interrupt which causes
375	 * a TLB miss can clobber the MMUCR between the mtspr and the tlbsx. */
376	mfmsr	r4
377	lis	r6,(MSR_EE|MSR_CE|MSR_ME|MSR_DE)@ha
378	addi	r6,r6,(MSR_EE|MSR_CE|MSR_ME|MSR_DE)@l
379	andc	r6,r4,r6
380	mtmsr	r6
381	mtspr	SPRN_MMUCR,r5
382	tlbsx.	r3, 0, r3
383	mtmsr	r4
384	bne	10f
385	sync
386	/* There are only 64 TLB entries, so r3 < 64,
387	 * which means bit 22, is clear.  Since 22 is
388	 * the V bit in the TLB_PAGEID, loading this
389	 * value will invalidate the TLB entry.
390	 */
391	tlbwe	r3, r3, PPC44x_TLB_PAGEID
392	isync
39310:
394#elif defined(CONFIG_FSL_BOOKE)
395	rlwinm	r4, r3, 0, 0, 19
396	ori	r5, r4, 0x08	/* TLBSEL = 1 */
397	tlbivax	0, r4
398	tlbivax	0, r5
399	msync
400#if defined(CONFIG_SMP)
401	tlbsync
402#endif /* CONFIG_SMP */
403#else /* !(CONFIG_40x || CONFIG_44x || CONFIG_FSL_BOOKE) */
404#if defined(CONFIG_SMP)
405	rlwinm	r8,r1,0,0,(31-THREAD_SHIFT)
406	lwz	r8,TI_CPU(r8)
407	oris	r8,r8,11
408	mfmsr	r10
409	SYNC
410	rlwinm	r0,r10,0,17,15		/* clear bit 16 (MSR_EE) */
411	rlwinm	r0,r0,0,28,26		/* clear DR */
412	mtmsr	r0
413	SYNC_601
414	isync
415	lis	r9,mmu_hash_lock@h
416	ori	r9,r9,mmu_hash_lock@l
417	tophys(r9,r9)
41810:	lwarx	r7,0,r9
419	cmpwi	0,r7,0
420	bne-	10b
421	stwcx.	r8,0,r9
422	bne-	10b
423	eieio
424	tlbie	r3
425	sync
426	TLBSYNC
427	li	r0,0
428	stw	r0,0(r9)		/* clear mmu_hash_lock */
429	mtmsr	r10
430	SYNC_601
431	isync
432#else /* CONFIG_SMP */
433	tlbie	r3
434	sync
435#endif /* CONFIG_SMP */
436#endif /* ! CONFIG_40x */
437	blr
438
439/*
440 * Flush instruction cache.
441 * This is a no-op on the 601.
442 */
443_GLOBAL(flush_instruction_cache)
444#if defined(CONFIG_8xx)
445	isync
446	lis	r5, IDC_INVALL@h
447	mtspr	SPRN_IC_CST, r5
448#elif defined(CONFIG_4xx)
449#ifdef CONFIG_403GCX
450	li      r3, 512
451	mtctr   r3
452	lis     r4, KERNELBASE@h
4531:	iccci   0, r4
454	addi    r4, r4, 16
455	bdnz    1b
456#else
457	lis	r3, KERNELBASE@h
458	iccci	0,r3
459#endif
460#elif CONFIG_FSL_BOOKE
461BEGIN_FTR_SECTION
462	mfspr   r3,SPRN_L1CSR0
463	ori     r3,r3,L1CSR0_CFI|L1CSR0_CLFC
464	/* msync; isync recommended here */
465	mtspr   SPRN_L1CSR0,r3
466	isync
467	blr
468END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
469	mfspr	r3,SPRN_L1CSR1
470	ori	r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
471	mtspr	SPRN_L1CSR1,r3
472#else
473	mfspr	r3,SPRN_PVR
474	rlwinm	r3,r3,16,16,31
475	cmpwi	0,r3,1
476	beqlr			/* for 601, do nothing */
477	/* 603/604 processor - use invalidate-all bit in HID0 */
478	mfspr	r3,SPRN_HID0
479	ori	r3,r3,HID0_ICFI
480	mtspr	SPRN_HID0,r3
481#endif /* CONFIG_8xx/4xx */
482	isync
483	blr
484
485/*
486 * Write any modified data cache blocks out to memory
487 * and invalidate the corresponding instruction cache blocks.
488 * This is a no-op on the 601.
489 *
490 * flush_icache_range(unsigned long start, unsigned long stop)
491 */
492_KPROBE(__flush_icache_range)
493BEGIN_FTR_SECTION
494	blr				/* for 601, do nothing */
495END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
496	li	r5,L1_CACHE_BYTES-1
497	andc	r3,r3,r5
498	subf	r4,r3,r4
499	add	r4,r4,r5
500	srwi.	r4,r4,L1_CACHE_SHIFT
501	beqlr
502	mtctr	r4
503	mr	r6,r3
5041:	dcbst	0,r3
505	addi	r3,r3,L1_CACHE_BYTES
506	bdnz	1b
507	sync				/* wait for dcbst's to get to ram */
508	mtctr	r4
5092:	icbi	0,r6
510	addi	r6,r6,L1_CACHE_BYTES
511	bdnz	2b
512	sync				/* additional sync needed on g4 */
513	isync
514	blr
515/*
516 * Write any modified data cache blocks out to memory.
517 * Does not invalidate the corresponding cache lines (especially for
518 * any corresponding instruction cache).
519 *
520 * clean_dcache_range(unsigned long start, unsigned long stop)
521 */
522_GLOBAL(clean_dcache_range)
523	li	r5,L1_CACHE_BYTES-1
524	andc	r3,r3,r5
525	subf	r4,r3,r4
526	add	r4,r4,r5
527	srwi.	r4,r4,L1_CACHE_SHIFT
528	beqlr
529	mtctr	r4
530
5311:	dcbst	0,r3
532	addi	r3,r3,L1_CACHE_BYTES
533	bdnz	1b
534	sync				/* wait for dcbst's to get to ram */
535	blr
536
537/*
538 * Write any modified data cache blocks out to memory and invalidate them.
539 * Does not invalidate the corresponding instruction cache blocks.
540 *
541 * flush_dcache_range(unsigned long start, unsigned long stop)
542 */
543_GLOBAL(flush_dcache_range)
544	li	r5,L1_CACHE_BYTES-1
545	andc	r3,r3,r5
546	subf	r4,r3,r4
547	add	r4,r4,r5
548	srwi.	r4,r4,L1_CACHE_SHIFT
549	beqlr
550	mtctr	r4
551
5521:	dcbf	0,r3
553	addi	r3,r3,L1_CACHE_BYTES
554	bdnz	1b
555	sync				/* wait for dcbst's to get to ram */
556	blr
557
558/*
559 * Like above, but invalidate the D-cache.  This is used by the 8xx
560 * to invalidate the cache so the PPC core doesn't get stale data
561 * from the CPM (no cache snooping here :-).
562 *
563 * invalidate_dcache_range(unsigned long start, unsigned long stop)
564 */
565_GLOBAL(invalidate_dcache_range)
566	li	r5,L1_CACHE_BYTES-1
567	andc	r3,r3,r5
568	subf	r4,r3,r4
569	add	r4,r4,r5
570	srwi.	r4,r4,L1_CACHE_SHIFT
571	beqlr
572	mtctr	r4
573
5741:	dcbi	0,r3
575	addi	r3,r3,L1_CACHE_BYTES
576	bdnz	1b
577	sync				/* wait for dcbi's to get to ram */
578	blr
579
580/*
581 * Flush a particular page from the data cache to RAM.
582 * Note: this is necessary because the instruction cache does *not*
583 * snoop from the data cache.
584 * This is a no-op on the 601 which has a unified cache.
585 *
586 *	void __flush_dcache_icache(void *page)
587 */
588_GLOBAL(__flush_dcache_icache)
589BEGIN_FTR_SECTION
590	blr
591END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
592	rlwinm	r3,r3,0,0,19			/* Get page base address */
593	li	r4,4096/L1_CACHE_BYTES	/* Number of lines in a page */
594	mtctr	r4
595	mr	r6,r3
5960:	dcbst	0,r3				/* Write line to ram */
597	addi	r3,r3,L1_CACHE_BYTES
598	bdnz	0b
599	sync
600#ifndef CONFIG_44x
601	/* We don't flush the icache on 44x. Those have a virtual icache
602	 * and we don't have access to the virtual address here (it's
603	 * not the page vaddr but where it's mapped in user space). The
604	 * flushing of the icache on these is handled elsewhere, when
605	 * a change in the address space occurs, before returning to
606	 * user space
607	 */
608	mtctr	r4
6091:	icbi	0,r6
610	addi	r6,r6,L1_CACHE_BYTES
611	bdnz	1b
612	sync
613	isync
614#endif /* CONFIG_44x */
615	blr
616
617/*
618 * Flush a particular page from the data cache to RAM, identified
619 * by its physical address.  We turn off the MMU so we can just use
620 * the physical address (this may be a highmem page without a kernel
621 * mapping).
622 *
623 *	void __flush_dcache_icache_phys(unsigned long physaddr)
624 */
625_GLOBAL(__flush_dcache_icache_phys)
626BEGIN_FTR_SECTION
627	blr					/* for 601, do nothing */
628END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
629	mfmsr	r10
630	rlwinm	r0,r10,0,28,26			/* clear DR */
631	mtmsr	r0
632	isync
633	rlwinm	r3,r3,0,0,19			/* Get page base address */
634	li	r4,4096/L1_CACHE_BYTES	/* Number of lines in a page */
635	mtctr	r4
636	mr	r6,r3
6370:	dcbst	0,r3				/* Write line to ram */
638	addi	r3,r3,L1_CACHE_BYTES
639	bdnz	0b
640	sync
641	mtctr	r4
6421:	icbi	0,r6
643	addi	r6,r6,L1_CACHE_BYTES
644	bdnz	1b
645	sync
646	mtmsr	r10				/* restore DR */
647	isync
648	blr
649
650/*
651 * Clear pages using the dcbz instruction, which doesn't cause any
652 * memory traffic (except to write out any cache lines which get
653 * displaced).  This only works on cacheable memory.
654 *
655 * void clear_pages(void *page, int order) ;
656 */
657_GLOBAL(clear_pages)
658	li	r0,4096/L1_CACHE_BYTES
659	slw	r0,r0,r4
660	mtctr	r0
661#ifdef CONFIG_8xx
662	li	r4, 0
6631:	stw	r4, 0(r3)
664	stw	r4, 4(r3)
665	stw	r4, 8(r3)
666	stw	r4, 12(r3)
667#else
6681:	dcbz	0,r3
669#endif
670	addi	r3,r3,L1_CACHE_BYTES
671	bdnz	1b
672	blr
673
674/*
675 * Copy a whole page.  We use the dcbz instruction on the destination
676 * to reduce memory traffic (it eliminates the unnecessary reads of
677 * the destination into cache).  This requires that the destination
678 * is cacheable.
679 */
680#define COPY_16_BYTES		\
681	lwz	r6,4(r4);	\
682	lwz	r7,8(r4);	\
683	lwz	r8,12(r4);	\
684	lwzu	r9,16(r4);	\
685	stw	r6,4(r3);	\
686	stw	r7,8(r3);	\
687	stw	r8,12(r3);	\
688	stwu	r9,16(r3)
689
690_GLOBAL(copy_page)
691	addi	r3,r3,-4
692	addi	r4,r4,-4
693
694#ifdef CONFIG_8xx
695	/* don't use prefetch on 8xx */
696    	li	r0,4096/L1_CACHE_BYTES
697	mtctr	r0
6981:	COPY_16_BYTES
699	bdnz	1b
700	blr
701
702#else	/* not 8xx, we can prefetch */
703	li	r5,4
704
705#if MAX_COPY_PREFETCH > 1
706	li	r0,MAX_COPY_PREFETCH
707	li	r11,4
708	mtctr	r0
70911:	dcbt	r11,r4
710	addi	r11,r11,L1_CACHE_BYTES
711	bdnz	11b
712#else /* MAX_COPY_PREFETCH == 1 */
713	dcbt	r5,r4
714	li	r11,L1_CACHE_BYTES+4
715#endif /* MAX_COPY_PREFETCH */
716	li	r0,4096/L1_CACHE_BYTES - MAX_COPY_PREFETCH
717	crclr	4*cr0+eq
7182:
719	mtctr	r0
7201:
721	dcbt	r11,r4
722	dcbz	r5,r3
723	COPY_16_BYTES
724#if L1_CACHE_BYTES >= 32
725	COPY_16_BYTES
726#if L1_CACHE_BYTES >= 64
727	COPY_16_BYTES
728	COPY_16_BYTES
729#if L1_CACHE_BYTES >= 128
730	COPY_16_BYTES
731	COPY_16_BYTES
732	COPY_16_BYTES
733	COPY_16_BYTES
734#endif
735#endif
736#endif
737	bdnz	1b
738	beqlr
739	crnot	4*cr0+eq,4*cr0+eq
740	li	r0,MAX_COPY_PREFETCH
741	li	r11,4
742	b	2b
743#endif	/* CONFIG_8xx */
744
745/*
746 * void atomic_clear_mask(atomic_t mask, atomic_t *addr)
747 * void atomic_set_mask(atomic_t mask, atomic_t *addr);
748 */
749_GLOBAL(atomic_clear_mask)
75010:	lwarx	r5,0,r4
751	andc	r5,r5,r3
752	PPC405_ERR77(0,r4)
753	stwcx.	r5,0,r4
754	bne-	10b
755	blr
756_GLOBAL(atomic_set_mask)
75710:	lwarx	r5,0,r4
758	or	r5,r5,r3
759	PPC405_ERR77(0,r4)
760	stwcx.	r5,0,r4
761	bne-	10b
762	blr
763
764/*
765 * Extended precision shifts.
766 *
767 * Updated to be valid for shift counts from 0 to 63 inclusive.
768 * -- Gabriel
769 *
770 * R3/R4 has 64 bit value
771 * R5    has shift count
772 * result in R3/R4
773 *
774 *  ashrdi3: arithmetic right shift (sign propagation)
775 *  lshrdi3: logical right shift
776 *  ashldi3: left shift
777 */
778_GLOBAL(__ashrdi3)
779	subfic	r6,r5,32
780	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
781	addi	r7,r5,32	# could be xori, or addi with -32
782	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
783	rlwinm	r8,r7,0,32	# t3 = (count < 32) ? 32 : 0
784	sraw	r7,r3,r7	# t2 = MSW >> (count-32)
785	or	r4,r4,r6	# LSW |= t1
786	slw	r7,r7,r8	# t2 = (count < 32) ? 0 : t2
787	sraw	r3,r3,r5	# MSW = MSW >> count
788	or	r4,r4,r7	# LSW |= t2
789	blr
790
791_GLOBAL(__ashldi3)
792	subfic	r6,r5,32
793	slw	r3,r3,r5	# MSW = count > 31 ? 0 : MSW << count
794	addi	r7,r5,32	# could be xori, or addi with -32
795	srw	r6,r4,r6	# t1 = count > 31 ? 0 : LSW >> (32-count)
796	slw	r7,r4,r7	# t2 = count < 32 ? 0 : LSW << (count-32)
797	or	r3,r3,r6	# MSW |= t1
798	slw	r4,r4,r5	# LSW = LSW << count
799	or	r3,r3,r7	# MSW |= t2
800	blr
801
802_GLOBAL(__lshrdi3)
803	subfic	r6,r5,32
804	srw	r4,r4,r5	# LSW = count > 31 ? 0 : LSW >> count
805	addi	r7,r5,32	# could be xori, or addi with -32
806	slw	r6,r3,r6	# t1 = count > 31 ? 0 : MSW << (32-count)
807	srw	r7,r3,r7	# t2 = count < 32 ? 0 : MSW >> (count-32)
808	or	r4,r4,r6	# LSW |= t1
809	srw	r3,r3,r5	# MSW = MSW >> count
810	or	r4,r4,r7	# LSW |= t2
811	blr
812
813/*
814 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
815 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
816 */
817_GLOBAL(__ucmpdi2)
818	cmplw	r3,r5
819	li	r3,1
820	bne	1f
821	cmplw	r4,r6
822	beqlr
8231:	li	r3,0
824	bltlr
825	li	r3,2
826	blr
827
828_GLOBAL(abs)
829	srawi	r4,r3,31
830	xor	r3,r3,r4
831	sub	r3,r3,r4
832	blr
833
834/*
835 * Create a kernel thread
836 *   kernel_thread(fn, arg, flags)
837 */
838_GLOBAL(kernel_thread)
839	stwu	r1,-16(r1)
840	stw	r30,8(r1)
841	stw	r31,12(r1)
842	mr	r30,r3		/* function */
843	mr	r31,r4		/* argument */
844	ori	r3,r5,CLONE_VM	/* flags */
845	oris	r3,r3,CLONE_UNTRACED>>16
846	li	r4,0		/* new sp (unused) */
847	li	r0,__NR_clone
848	sc
849	cmpwi	0,r3,0		/* parent or child? */
850	bne	1f		/* return if parent */
851	li	r0,0		/* make top-level stack frame */
852	stwu	r0,-16(r1)
853	mtlr	r30		/* fn addr in lr */
854	mr	r3,r31		/* load arg and call fn */
855	PPC440EP_ERR42
856	blrl
857	li	r0,__NR_exit	/* exit if function returns */
858	li	r3,0
859	sc
8601:	lwz	r30,8(r1)
861	lwz	r31,12(r1)
862	addi	r1,r1,16
863	blr
864
865/*
866 * This routine is just here to keep GCC happy - sigh...
867 */
868_GLOBAL(__main)
869	blr
870
871#ifdef CONFIG_KEXEC
872	/*
873	 * Must be relocatable PIC code callable as a C function.
874	 */
875	.globl relocate_new_kernel
876relocate_new_kernel:
877	/* r3 = page_list   */
878	/* r4 = reboot_code_buffer */
879	/* r5 = start_address      */
880
881	li	r0, 0
882
883	/*
884	 * Set Machine Status Register to a known status,
885	 * switch the MMU off and jump to 1: in a single step.
886	 */
887
888	mr	r8, r0
889	ori     r8, r8, MSR_RI|MSR_ME
890	mtspr	SPRN_SRR1, r8
891	addi	r8, r4, 1f - relocate_new_kernel
892	mtspr	SPRN_SRR0, r8
893	sync
894	rfi
895
8961:
897	/* from this point address translation is turned off */
898	/* and interrupts are disabled */
899
900	/* set a new stack at the bottom of our page... */
901	/* (not really needed now) */
902	addi	r1, r4, KEXEC_CONTROL_CODE_SIZE - 8 /* for LR Save+Back Chain */
903	stw	r0, 0(r1)
904
905	/* Do the copies */
906	li	r6, 0 /* checksum */
907	mr	r0, r3
908	b	1f
909
9100:	/* top, read another word for the indirection page */
911	lwzu	r0, 4(r3)
912
9131:
914	/* is it a destination page? (r8) */
915	rlwinm.	r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
916	beq	2f
917
918	rlwinm	r8, r0, 0, 0, 19 /* clear kexec flags, page align */
919	b	0b
920
9212:	/* is it an indirection page? (r3) */
922	rlwinm.	r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
923	beq	2f
924
925	rlwinm	r3, r0, 0, 0, 19 /* clear kexec flags, page align */
926	subi	r3, r3, 4
927	b	0b
928
9292:	/* are we done? */
930	rlwinm.	r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
931	beq	2f
932	b	3f
933
9342:	/* is it a source page? (r9) */
935	rlwinm.	r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
936	beq	0b
937
938	rlwinm	r9, r0, 0, 0, 19 /* clear kexec flags, page align */
939
940	li	r7, PAGE_SIZE / 4
941	mtctr   r7
942	subi    r9, r9, 4
943	subi    r8, r8, 4
9449:
945	lwzu    r0, 4(r9)  /* do the copy */
946	xor	r6, r6, r0
947	stwu    r0, 4(r8)
948	dcbst	0, r8
949	sync
950	icbi	0, r8
951	bdnz    9b
952
953	addi    r9, r9, 4
954	addi    r8, r8, 4
955	b	0b
956
9573:
958
959	/* To be certain of avoiding problems with self-modifying code
960	 * execute a serializing instruction here.
961	 */
962	isync
963	sync
964
965	/* jump to the entry point, usually the setup routine */
966	mtlr	r5
967	blrl
968
9691:	b	1b
970
971relocate_new_kernel_end:
972
973	.globl relocate_new_kernel_size
974relocate_new_kernel_size:
975	.long relocate_new_kernel_end - relocate_new_kernel
976#endif
977