xref: /openbmc/linux/arch/powerpc/kernel/misc_64.S (revision 44ce0cd3)
1/* SPDX-License-Identifier: GPL-2.0-or-later */
2/*
3 * This file contains miscellaneous low-level functions.
4 *    Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
5 *
6 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
7 * and Paul Mackerras.
8 * Adapted for iSeries by Mike Corrigan (mikejc@us.ibm.com)
9 * PPC64 updates by Dave Engebretsen (engebret@us.ibm.com)
10 */
11
12#include <linux/sys.h>
13#include <asm/unistd.h>
14#include <asm/errno.h>
15#include <asm/processor.h>
16#include <asm/page.h>
17#include <asm/cache.h>
18#include <asm/ppc_asm.h>
19#include <asm/asm-offsets.h>
20#include <asm/cputable.h>
21#include <asm/thread_info.h>
22#include <asm/kexec.h>
23#include <asm/ptrace.h>
24#include <asm/mmu.h>
25#include <asm/export.h>
26#include <asm/feature-fixups.h>
27
28	.text
29
30_GLOBAL(call_do_softirq)
31	mflr	r0
32	std	r0,16(r1)
33	stdu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
34	mr	r1,r3
35	bl	__do_softirq
36	ld	r1,0(r1)
37	ld	r0,16(r1)
38	mtlr	r0
39	blr
40
41_GLOBAL(call_do_irq)
42	mflr	r0
43	std	r0,16(r1)
44	stdu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
45	mr	r1,r4
46	bl	__do_irq
47	ld	r1,0(r1)
48	ld	r0,16(r1)
49	mtlr	r0
50	blr
51
52	.section	".toc","aw"
53PPC64_CACHES:
54	.tc		ppc64_caches[TC],ppc64_caches
55	.section	".text"
56
57/*
58 * Write any modified data cache blocks out to memory
59 * and invalidate the corresponding instruction cache blocks.
60 *
61 * flush_icache_range(unsigned long start, unsigned long stop)
62 *
63 *   flush all bytes from start through stop-1 inclusive
64 */
65
66_GLOBAL_TOC(flush_icache_range)
67BEGIN_FTR_SECTION
68	PURGE_PREFETCHED_INS
69	blr
70END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
71/*
72 * Flush the data cache to memory
73 *
74 * Different systems have different cache line sizes
75 * and in some cases i-cache and d-cache line sizes differ from
76 * each other.
77 */
78 	ld	r10,PPC64_CACHES@toc(r2)
79	lwz	r7,DCACHEL1BLOCKSIZE(r10)/* Get cache block size */
80	addi	r5,r7,-1
81	andc	r6,r3,r5		/* round low to line bdy */
82	subf	r8,r6,r4		/* compute length */
83	add	r8,r8,r5		/* ensure we get enough */
84	lwz	r9,DCACHEL1LOGBLOCKSIZE(r10)	/* Get log-2 of cache block size */
85	srw.	r8,r8,r9		/* compute line count */
86	beqlr				/* nothing to do? */
87	mtctr	r8
881:	dcbst	0,r6
89	add	r6,r6,r7
90	bdnz	1b
91	sync
92
93/* Now invalidate the instruction cache */
94
95	lwz	r7,ICACHEL1BLOCKSIZE(r10)	/* Get Icache block size */
96	addi	r5,r7,-1
97	andc	r6,r3,r5		/* round low to line bdy */
98	subf	r8,r6,r4		/* compute length */
99	add	r8,r8,r5
100	lwz	r9,ICACHEL1LOGBLOCKSIZE(r10)	/* Get log-2 of Icache block size */
101	srw.	r8,r8,r9		/* compute line count */
102	beqlr				/* nothing to do? */
103	mtctr	r8
1042:	icbi	0,r6
105	add	r6,r6,r7
106	bdnz	2b
107	isync
108	blr
109_ASM_NOKPROBE_SYMBOL(flush_icache_range)
110EXPORT_SYMBOL(flush_icache_range)
111
112/*
113 * Flush a particular page from the data cache to RAM.
114 * Note: this is necessary because the instruction cache does *not*
115 * snoop from the data cache.
116 *
117 *	void __flush_dcache_icache(void *page)
118 */
119_GLOBAL(__flush_dcache_icache)
120/*
121 * Flush the data cache to memory
122 *
123 * Different systems have different cache line sizes
124 */
125
126BEGIN_FTR_SECTION
127	PURGE_PREFETCHED_INS
128	blr
129END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
130
131/* Flush the dcache */
132 	ld	r7,PPC64_CACHES@toc(r2)
133	clrrdi	r3,r3,PAGE_SHIFT           	    /* Page align */
134	lwz	r4,DCACHEL1BLOCKSPERPAGE(r7)	/* Get # dcache blocks per page */
135	lwz	r5,DCACHEL1BLOCKSIZE(r7)	/* Get dcache block size */
136	mr	r6,r3
137	mtctr	r4
1380:	dcbst	0,r6
139	add	r6,r6,r5
140	bdnz	0b
141	sync
142
143/* Now invalidate the icache */
144
145	lwz	r4,ICACHEL1BLOCKSPERPAGE(r7)	/* Get # icache blocks per page */
146	lwz	r5,ICACHEL1BLOCKSIZE(r7)	/* Get icache block size */
147	mtctr	r4
1481:	icbi	0,r3
149	add	r3,r3,r5
150	bdnz	1b
151	isync
152	blr
153
154_GLOBAL(__bswapdi2)
155EXPORT_SYMBOL(__bswapdi2)
156	srdi	r8,r3,32
157	rlwinm	r7,r3,8,0xffffffff
158	rlwimi	r7,r3,24,0,7
159	rlwinm	r9,r8,8,0xffffffff
160	rlwimi	r7,r3,24,16,23
161	rlwimi	r9,r8,24,0,7
162	rlwimi	r9,r8,24,16,23
163	sldi	r7,r7,32
164	or	r3,r7,r9
165	blr
166
167
168#ifdef CONFIG_PPC_EARLY_DEBUG_BOOTX
169_GLOBAL(rmci_on)
170	sync
171	isync
172	li	r3,0x100
173	rldicl	r3,r3,32,0
174	mfspr	r5,SPRN_HID4
175	or	r5,r5,r3
176	sync
177	mtspr	SPRN_HID4,r5
178	isync
179	slbia
180	isync
181	sync
182	blr
183
184_GLOBAL(rmci_off)
185	sync
186	isync
187	li	r3,0x100
188	rldicl	r3,r3,32,0
189	mfspr	r5,SPRN_HID4
190	andc	r5,r5,r3
191	sync
192	mtspr	SPRN_HID4,r5
193	isync
194	slbia
195	isync
196	sync
197	blr
198#endif /* CONFIG_PPC_EARLY_DEBUG_BOOTX */
199
200#if defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE)
201
202/*
203 * Do an IO access in real mode
204 */
205_GLOBAL(real_readb)
206	mfmsr	r7
207	ori	r0,r7,MSR_DR
208	xori	r0,r0,MSR_DR
209	sync
210	mtmsrd	r0
211	sync
212	isync
213	mfspr	r6,SPRN_HID4
214	rldicl	r5,r6,32,0
215	ori	r5,r5,0x100
216	rldicl	r5,r5,32,0
217	sync
218	mtspr	SPRN_HID4,r5
219	isync
220	slbia
221	isync
222	lbz	r3,0(r3)
223	sync
224	mtspr	SPRN_HID4,r6
225	isync
226	slbia
227	isync
228	mtmsrd	r7
229	sync
230	isync
231	blr
232
233	/*
234 * Do an IO access in real mode
235 */
236_GLOBAL(real_writeb)
237	mfmsr	r7
238	ori	r0,r7,MSR_DR
239	xori	r0,r0,MSR_DR
240	sync
241	mtmsrd	r0
242	sync
243	isync
244	mfspr	r6,SPRN_HID4
245	rldicl	r5,r6,32,0
246	ori	r5,r5,0x100
247	rldicl	r5,r5,32,0
248	sync
249	mtspr	SPRN_HID4,r5
250	isync
251	slbia
252	isync
253	stb	r3,0(r4)
254	sync
255	mtspr	SPRN_HID4,r6
256	isync
257	slbia
258	isync
259	mtmsrd	r7
260	sync
261	isync
262	blr
263#endif /* defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) */
264
265#ifdef CONFIG_PPC_PASEMI
266
267_GLOBAL(real_205_readb)
268	mfmsr	r7
269	ori	r0,r7,MSR_DR
270	xori	r0,r0,MSR_DR
271	sync
272	mtmsrd	r0
273	sync
274	isync
275	LBZCIX(R3,R0,R3)
276	isync
277	mtmsrd	r7
278	sync
279	isync
280	blr
281
282_GLOBAL(real_205_writeb)
283	mfmsr	r7
284	ori	r0,r7,MSR_DR
285	xori	r0,r0,MSR_DR
286	sync
287	mtmsrd	r0
288	sync
289	isync
290	STBCIX(R3,R0,R4)
291	isync
292	mtmsrd	r7
293	sync
294	isync
295	blr
296
297#endif /* CONFIG_PPC_PASEMI */
298
299
300#if defined(CONFIG_CPU_FREQ_PMAC64) || defined(CONFIG_CPU_FREQ_MAPLE)
301/*
302 * SCOM access functions for 970 (FX only for now)
303 *
304 * unsigned long scom970_read(unsigned int address);
305 * void scom970_write(unsigned int address, unsigned long value);
306 *
307 * The address passed in is the 24 bits register address. This code
308 * is 970 specific and will not check the status bits, so you should
309 * know what you are doing.
310 */
311_GLOBAL(scom970_read)
312	/* interrupts off */
313	mfmsr	r4
314	ori	r0,r4,MSR_EE
315	xori	r0,r0,MSR_EE
316	mtmsrd	r0,1
317
318	/* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
319	 * (including parity). On current CPUs they must be 0'd,
320	 * and finally or in RW bit
321	 */
322	rlwinm	r3,r3,8,0,15
323	ori	r3,r3,0x8000
324
325	/* do the actual scom read */
326	sync
327	mtspr	SPRN_SCOMC,r3
328	isync
329	mfspr	r3,SPRN_SCOMD
330	isync
331	mfspr	r0,SPRN_SCOMC
332	isync
333
334	/* XXX:	fixup result on some buggy 970's (ouch ! we lost a bit, bah
335	 * that's the best we can do). Not implemented yet as we don't use
336	 * the scom on any of the bogus CPUs yet, but may have to be done
337	 * ultimately
338	 */
339
340	/* restore interrupts */
341	mtmsrd	r4,1
342	blr
343
344
345_GLOBAL(scom970_write)
346	/* interrupts off */
347	mfmsr	r5
348	ori	r0,r5,MSR_EE
349	xori	r0,r0,MSR_EE
350	mtmsrd	r0,1
351
352	/* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
353	 * (including parity). On current CPUs they must be 0'd.
354	 */
355
356	rlwinm	r3,r3,8,0,15
357
358	sync
359	mtspr	SPRN_SCOMD,r4      /* write data */
360	isync
361	mtspr	SPRN_SCOMC,r3      /* write command */
362	isync
363	mfspr	3,SPRN_SCOMC
364	isync
365
366	/* restore interrupts */
367	mtmsrd	r5,1
368	blr
369#endif /* CONFIG_CPU_FREQ_PMAC64 || CONFIG_CPU_FREQ_MAPLE */
370
371/* kexec_wait(phys_cpu)
372 *
373 * wait for the flag to change, indicating this kernel is going away but
374 * the slave code for the next one is at addresses 0 to 100.
375 *
376 * This is used by all slaves, even those that did not find a matching
377 * paca in the secondary startup code.
378 *
379 * Physical (hardware) cpu id should be in r3.
380 */
381_GLOBAL(kexec_wait)
382	bl	1f
3831:	mflr	r5
384	addi	r5,r5,kexec_flag-1b
385
38699:	HMT_LOW
387#ifdef CONFIG_KEXEC_CORE	/* use no memory without kexec */
388	lwz	r4,0(r5)
389	cmpwi	0,r4,0
390	beq	99b
391#ifdef CONFIG_PPC_BOOK3S_64
392	li	r10,0x60
393	mfmsr	r11
394	clrrdi	r11,r11,1	/* Clear MSR_LE */
395	mtsrr0	r10
396	mtsrr1	r11
397	rfid
398#else
399	/* Create TLB entry in book3e_secondary_core_init */
400	li	r4,0
401	ba	0x60
402#endif
403#endif
404
405/* this can be in text because we won't change it until we are
406 * running in real anyways
407 */
408kexec_flag:
409	.long	0
410
411
412#ifdef CONFIG_KEXEC_CORE
413#ifdef CONFIG_PPC_BOOK3E
414/*
415 * BOOK3E has no real MMU mode, so we have to setup the initial TLB
416 * for a core to identity map v:0 to p:0.  This current implementation
417 * assumes that 1G is enough for kexec.
418 */
419kexec_create_tlb:
420	/*
421	 * Invalidate all non-IPROT TLB entries to avoid any TLB conflict.
422	 * IPROT TLB entries should be >= PAGE_OFFSET and thus not conflict.
423	 */
424	PPC_TLBILX_ALL(0,R0)
425	sync
426	isync
427
428	mfspr	r10,SPRN_TLB1CFG
429	andi.	r10,r10,TLBnCFG_N_ENTRY	/* Extract # entries */
430	subi	r10,r10,1	/* Last entry: no conflict with kernel text */
431	lis	r9,MAS0_TLBSEL(1)@h
432	rlwimi	r9,r10,16,4,15		/* Setup MAS0 = TLBSEL | ESEL(r9) */
433
434/* Set up a temp identity mapping v:0 to p:0 and return to it. */
435#if defined(CONFIG_SMP) || defined(CONFIG_PPC_E500MC)
436#define M_IF_NEEDED	MAS2_M
437#else
438#define M_IF_NEEDED	0
439#endif
440	mtspr	SPRN_MAS0,r9
441
442	lis	r9,(MAS1_VALID|MAS1_IPROT)@h
443	ori	r9,r9,(MAS1_TSIZE(BOOK3E_PAGESZ_1GB))@l
444	mtspr	SPRN_MAS1,r9
445
446	LOAD_REG_IMMEDIATE(r9, 0x0 | M_IF_NEEDED)
447	mtspr	SPRN_MAS2,r9
448
449	LOAD_REG_IMMEDIATE(r9, 0x0 | MAS3_SR | MAS3_SW | MAS3_SX)
450	mtspr	SPRN_MAS3,r9
451	li	r9,0
452	mtspr	SPRN_MAS7,r9
453
454	tlbwe
455	isync
456	blr
457#endif
458
459/* kexec_smp_wait(void)
460 *
461 * call with interrupts off
462 * note: this is a terminal routine, it does not save lr
463 *
464 * get phys id from paca
465 * switch to real mode
466 * mark the paca as no longer used
467 * join other cpus in kexec_wait(phys_id)
468 */
469_GLOBAL(kexec_smp_wait)
470	lhz	r3,PACAHWCPUID(r13)
471	bl	real_mode
472
473	li	r4,KEXEC_STATE_REAL_MODE
474	stb	r4,PACAKEXECSTATE(r13)
475	SYNC
476
477	b	kexec_wait
478
479/*
480 * switch to real mode (turn mmu off)
481 * we use the early kernel trick that the hardware ignores bits
482 * 0 and 1 (big endian) of the effective address in real mode
483 *
484 * don't overwrite r3 here, it is live for kexec_wait above.
485 */
486real_mode:	/* assume normal blr return */
487#ifdef CONFIG_PPC_BOOK3E
488	/* Create an identity mapping. */
489	b	kexec_create_tlb
490#else
4911:	li	r9,MSR_RI
492	li	r10,MSR_DR|MSR_IR
493	mflr	r11		/* return address to SRR0 */
494	mfmsr	r12
495	andc	r9,r12,r9
496	andc	r10,r12,r10
497
498	mtmsrd	r9,1
499	mtspr	SPRN_SRR1,r10
500	mtspr	SPRN_SRR0,r11
501	rfid
502#endif
503
504/*
505 * kexec_sequence(newstack, start, image, control, clear_all(),
506	          copy_with_mmu_off)
507 *
508 * does the grungy work with stack switching and real mode switches
509 * also does simple calls to other code
510 */
511
512_GLOBAL(kexec_sequence)
513	mflr	r0
514	std	r0,16(r1)
515
516	/* switch stacks to newstack -- &kexec_stack.stack */
517	stdu	r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
518	mr	r1,r3
519
520	li	r0,0
521	std	r0,16(r1)
522
523BEGIN_FTR_SECTION
524	/*
525	 * This is the best time to turn AMR/IAMR off.
526	 * key 0 is used in radix for supervisor<->user
527	 * protection, but on hash key 0 is reserved
528	 * ideally we want to enter with a clean state.
529	 * NOTE, we rely on r0 being 0 from above.
530	 */
531	mtspr	SPRN_IAMR,r0
532BEGIN_FTR_SECTION_NESTED(42)
533	mtspr	SPRN_AMOR,r0
534END_FTR_SECTION_NESTED_IFSET(CPU_FTR_HVMODE, 42)
535END_FTR_SECTION_IFSET(CPU_FTR_ARCH_300)
536
537	/* save regs for local vars on new stack.
538	 * yes, we won't go back, but ...
539	 */
540	std	r31,-8(r1)
541	std	r30,-16(r1)
542	std	r29,-24(r1)
543	std	r28,-32(r1)
544	std	r27,-40(r1)
545	std	r26,-48(r1)
546	std	r25,-56(r1)
547
548	stdu	r1,-STACK_FRAME_OVERHEAD-64(r1)
549
550	/* save args into preserved regs */
551	mr	r31,r3			/* newstack (both) */
552	mr	r30,r4			/* start (real) */
553	mr	r29,r5			/* image (virt) */
554	mr	r28,r6			/* control, unused */
555	mr	r27,r7			/* clear_all() fn desc */
556	mr	r26,r8			/* copy_with_mmu_off */
557	lhz	r25,PACAHWCPUID(r13)	/* get our phys cpu from paca */
558
559	/* disable interrupts, we are overwriting kernel data next */
560#ifdef CONFIG_PPC_BOOK3E
561	wrteei	0
562#else
563	mfmsr	r3
564	rlwinm	r3,r3,0,17,15
565	mtmsrd	r3,1
566#endif
567
568	/* We need to turn the MMU off unless we are in hash mode
569	 * under a hypervisor
570	 */
571	cmpdi	r26,0
572	beq	1f
573	bl	real_mode
5741:
575	/* copy dest pages, flush whole dest image */
576	mr	r3,r29
577	bl	kexec_copy_flush	/* (image) */
578
579	/* turn off mmu now if not done earlier */
580	cmpdi	r26,0
581	bne	1f
582	bl	real_mode
583
584	/* copy  0x100 bytes starting at start to 0 */
5851:	li	r3,0
586	mr	r4,r30		/* start, aka phys mem offset */
587	li	r5,0x100
588	li	r6,0
589	bl	copy_and_flush	/* (dest, src, copy limit, start offset) */
5901:	/* assume normal blr return */
591
592	/* release other cpus to the new kernel secondary start at 0x60 */
593	mflr	r5
594	li	r6,1
595	stw	r6,kexec_flag-1b(5)
596
597	cmpdi	r27,0
598	beq	1f
599
600	/* clear out hardware hash page table and tlb */
601#ifdef PPC64_ELF_ABI_v1
602	ld	r12,0(r27)		/* deref function descriptor */
603#else
604	mr	r12,r27
605#endif
606	mtctr	r12
607	bctrl				/* mmu_hash_ops.hpte_clear_all(void); */
608
609/*
610 *   kexec image calling is:
611 *      the first 0x100 bytes of the entry point are copied to 0
612 *
613 *      all slaves branch to slave = 0x60 (absolute)
614 *              slave(phys_cpu_id);
615 *
616 *      master goes to start = entry point
617 *              start(phys_cpu_id, start, 0);
618 *
619 *
620 *   a wrapper is needed to call existing kernels, here is an approximate
621 *   description of one method:
622 *
623 * v2: (2.6.10)
624 *   start will be near the boot_block (maybe 0x100 bytes before it?)
625 *   it will have a 0x60, which will b to boot_block, where it will wait
626 *   and 0 will store phys into struct boot-block and load r3 from there,
627 *   copy kernel 0-0x100 and tell slaves to back down to 0x60 again
628 *
629 * v1: (2.6.9)
630 *    boot block will have all cpus scanning device tree to see if they
631 *    are the boot cpu ?????
632 *    other device tree differences (prop sizes, va vs pa, etc)...
633 */
6341:	mr	r3,r25	# my phys cpu
635	mr	r4,r30	# start, aka phys mem offset
636	mtlr	4
637	li	r5,0
638	blr	/* image->start(physid, image->start, 0); */
639#endif /* CONFIG_KEXEC_CORE */
640