Merge branch 'merge'
[deliverable/linux.git] / arch / powerpc / kernel / misc_64.S
1 /*
2 * arch/powerpc/kernel/misc64.S
3 *
4 * This file contains miscellaneous low-level functions.
5 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
6 *
7 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
8 * and Paul Mackerras.
9 * Adapted for iSeries by Mike Corrigan (mikejc@us.ibm.com)
10 * PPC64 updates by Dave Engebretsen (engebret@us.ibm.com)
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/config.h>
20 #include <linux/sys.h>
21 #include <asm/unistd.h>
22 #include <asm/errno.h>
23 #include <asm/processor.h>
24 #include <asm/page.h>
25 #include <asm/cache.h>
26 #include <asm/ppc_asm.h>
27 #include <asm/asm-offsets.h>
28 #include <asm/cputable.h>
29 #include <asm/thread_info.h>
30
31 .text
32
33 /*
34 * Returns (address we are running at) - (address we were linked at)
35 * for use before the text and data are mapped to KERNELBASE.
36 */
37
38 _GLOBAL(reloc_offset)
39 mflr r0
40 bl 1f
41 1: mflr r3
42 LOAD_REG_IMMEDIATE(r4,1b)
43 subf r3,r4,r3
44 mtlr r0
45 blr
46
47 /*
48 * add_reloc_offset(x) returns x + reloc_offset().
49 */
50 _GLOBAL(add_reloc_offset)
51 mflr r0
52 bl 1f
53 1: mflr r5
54 LOAD_REG_IMMEDIATE(r4,1b)
55 subf r5,r4,r5
56 add r3,r3,r5
57 mtlr r0
58 blr
59
60 _GLOBAL(get_msr)
61 mfmsr r3
62 blr
63
64 _GLOBAL(get_dar)
65 mfdar r3
66 blr
67
68 _GLOBAL(get_srr0)
69 mfsrr0 r3
70 blr
71
72 _GLOBAL(get_srr1)
73 mfsrr1 r3
74 blr
75
76 _GLOBAL(get_sp)
77 mr r3,r1
78 blr
79
80 #ifdef CONFIG_IRQSTACKS
81 _GLOBAL(call_do_softirq)
82 mflr r0
83 std r0,16(r1)
84 stdu r1,THREAD_SIZE-112(r3)
85 mr r1,r3
86 bl .__do_softirq
87 ld r1,0(r1)
88 ld r0,16(r1)
89 mtlr r0
90 blr
91
92 _GLOBAL(call___do_IRQ)
93 mflr r0
94 std r0,16(r1)
95 stdu r1,THREAD_SIZE-112(r5)
96 mr r1,r5
97 bl .__do_IRQ
98 ld r1,0(r1)
99 ld r0,16(r1)
100 mtlr r0
101 blr
102 #endif /* CONFIG_IRQSTACKS */
103
104 /*
105 * To be called by C code which needs to do some operations with MMU
106 * disabled. Note that interrupts have to be disabled by the caller
107 * prior to calling us. The code called _MUST_ be in the RMO of course
108 * and part of the linear mapping as we don't attempt to translate the
109 * stack pointer at all. The function is called with the stack switched
110 * to this CPU emergency stack
111 *
112 * prototype is void *call_with_mmu_off(void *func, void *data);
113 *
114 * the called function is expected to be of the form
115 *
116 * void *called(void *data);
117 */
118 _GLOBAL(call_with_mmu_off)
119 mflr r0 /* get link, save it on stackframe */
120 std r0,16(r1)
121 mr r1,r5 /* save old stack ptr */
122 ld r1,PACAEMERGSP(r13) /* get emerg. stack */
123 subi r1,r1,STACK_FRAME_OVERHEAD
124 std r0,16(r1) /* save link on emerg. stack */
125 std r5,0(r1) /* save old stack ptr in backchain */
126 ld r3,0(r3) /* get to real function ptr (assume same TOC) */
127 bl 2f /* we need LR to return, continue at label 2 */
128
129 ld r0,16(r1) /* we return here from the call, get LR and */
130 ld r1,0(r1) /* .. old stack ptr */
131 mtspr SPRN_SRR0,r0 /* and get back to virtual mode with these */
132 mfmsr r4
133 ori r4,r4,MSR_IR|MSR_DR
134 mtspr SPRN_SRR1,r4
135 rfid
136
137 2: mtspr SPRN_SRR0,r3 /* coming from above, enter real mode */
138 mr r3,r4 /* get parameter */
139 mfmsr r0
140 ori r0,r0,MSR_IR|MSR_DR
141 xori r0,r0,MSR_IR|MSR_DR
142 mtspr SPRN_SRR1,r0
143 rfid
144
145
146 .section ".toc","aw"
147 PPC64_CACHES:
148 .tc ppc64_caches[TC],ppc64_caches
149 .section ".text"
150
151 /*
152 * Write any modified data cache blocks out to memory
153 * and invalidate the corresponding instruction cache blocks.
154 *
155 * flush_icache_range(unsigned long start, unsigned long stop)
156 *
157 * flush all bytes from start through stop-1 inclusive
158 */
159
160 _KPROBE(__flush_icache_range)
161
162 /*
163 * Flush the data cache to memory
164 *
165 * Different systems have different cache line sizes
166 * and in some cases i-cache and d-cache line sizes differ from
167 * each other.
168 */
169 ld r10,PPC64_CACHES@toc(r2)
170 lwz r7,DCACHEL1LINESIZE(r10)/* Get cache line size */
171 addi r5,r7,-1
172 andc r6,r3,r5 /* round low to line bdy */
173 subf r8,r6,r4 /* compute length */
174 add r8,r8,r5 /* ensure we get enough */
175 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of cache line size */
176 srw. r8,r8,r9 /* compute line count */
177 beqlr /* nothing to do? */
178 mtctr r8
179 1: dcbst 0,r6
180 add r6,r6,r7
181 bdnz 1b
182 sync
183
184 /* Now invalidate the instruction cache */
185
186 lwz r7,ICACHEL1LINESIZE(r10) /* Get Icache line size */
187 addi r5,r7,-1
188 andc r6,r3,r5 /* round low to line bdy */
189 subf r8,r6,r4 /* compute length */
190 add r8,r8,r5
191 lwz r9,ICACHEL1LOGLINESIZE(r10) /* Get log-2 of Icache line size */
192 srw. r8,r8,r9 /* compute line count */
193 beqlr /* nothing to do? */
194 mtctr r8
195 2: icbi 0,r6
196 add r6,r6,r7
197 bdnz 2b
198 isync
199 blr
200 .previous .text
201 /*
202 * Like above, but only do the D-cache.
203 *
204 * flush_dcache_range(unsigned long start, unsigned long stop)
205 *
206 * flush all bytes from start to stop-1 inclusive
207 */
208 _GLOBAL(flush_dcache_range)
209
210 /*
211 * Flush the data cache to memory
212 *
213 * Different systems have different cache line sizes
214 */
215 ld r10,PPC64_CACHES@toc(r2)
216 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
217 addi r5,r7,-1
218 andc r6,r3,r5 /* round low to line bdy */
219 subf r8,r6,r4 /* compute length */
220 add r8,r8,r5 /* ensure we get enough */
221 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */
222 srw. r8,r8,r9 /* compute line count */
223 beqlr /* nothing to do? */
224 mtctr r8
225 0: dcbst 0,r6
226 add r6,r6,r7
227 bdnz 0b
228 sync
229 blr
230
231 /*
232 * Like above, but works on non-mapped physical addresses.
233 * Use only for non-LPAR setups ! It also assumes real mode
234 * is cacheable. Used for flushing out the DART before using
235 * it as uncacheable memory
236 *
237 * flush_dcache_phys_range(unsigned long start, unsigned long stop)
238 *
239 * flush all bytes from start to stop-1 inclusive
240 */
241 _GLOBAL(flush_dcache_phys_range)
242 ld r10,PPC64_CACHES@toc(r2)
243 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
244 addi r5,r7,-1
245 andc r6,r3,r5 /* round low to line bdy */
246 subf r8,r6,r4 /* compute length */
247 add r8,r8,r5 /* ensure we get enough */
248 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */
249 srw. r8,r8,r9 /* compute line count */
250 beqlr /* nothing to do? */
251 mfmsr r5 /* Disable MMU Data Relocation */
252 ori r0,r5,MSR_DR
253 xori r0,r0,MSR_DR
254 sync
255 mtmsr r0
256 sync
257 isync
258 mtctr r8
259 0: dcbst 0,r6
260 add r6,r6,r7
261 bdnz 0b
262 sync
263 isync
264 mtmsr r5 /* Re-enable MMU Data Relocation */
265 sync
266 isync
267 blr
268
269 _GLOBAL(flush_inval_dcache_range)
270 ld r10,PPC64_CACHES@toc(r2)
271 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
272 addi r5,r7,-1
273 andc r6,r3,r5 /* round low to line bdy */
274 subf r8,r6,r4 /* compute length */
275 add r8,r8,r5 /* ensure we get enough */
276 lwz r9,DCACHEL1LOGLINESIZE(r10)/* Get log-2 of dcache line size */
277 srw. r8,r8,r9 /* compute line count */
278 beqlr /* nothing to do? */
279 sync
280 isync
281 mtctr r8
282 0: dcbf 0,r6
283 add r6,r6,r7
284 bdnz 0b
285 sync
286 isync
287 blr
288
289
290 /*
291 * Flush a particular page from the data cache to RAM.
292 * Note: this is necessary because the instruction cache does *not*
293 * snoop from the data cache.
294 *
295 * void __flush_dcache_icache(void *page)
296 */
297 _GLOBAL(__flush_dcache_icache)
298 /*
299 * Flush the data cache to memory
300 *
301 * Different systems have different cache line sizes
302 */
303
304 /* Flush the dcache */
305 ld r7,PPC64_CACHES@toc(r2)
306 clrrdi r3,r3,PAGE_SHIFT /* Page align */
307 lwz r4,DCACHEL1LINESPERPAGE(r7) /* Get # dcache lines per page */
308 lwz r5,DCACHEL1LINESIZE(r7) /* Get dcache line size */
309 mr r6,r3
310 mtctr r4
311 0: dcbst 0,r6
312 add r6,r6,r5
313 bdnz 0b
314 sync
315
316 /* Now invalidate the icache */
317
318 lwz r4,ICACHEL1LINESPERPAGE(r7) /* Get # icache lines per page */
319 lwz r5,ICACHEL1LINESIZE(r7) /* Get icache line size */
320 mtctr r4
321 1: icbi 0,r3
322 add r3,r3,r5
323 bdnz 1b
324 isync
325 blr
326
327 /*
328 * I/O string operations
329 *
330 * insb(port, buf, len)
331 * outsb(port, buf, len)
332 * insw(port, buf, len)
333 * outsw(port, buf, len)
334 * insl(port, buf, len)
335 * outsl(port, buf, len)
336 * insw_ns(port, buf, len)
337 * outsw_ns(port, buf, len)
338 * insl_ns(port, buf, len)
339 * outsl_ns(port, buf, len)
340 *
341 * The *_ns versions don't do byte-swapping.
342 */
343 _GLOBAL(_insb)
344 cmpwi 0,r5,0
345 mtctr r5
346 subi r4,r4,1
347 blelr-
348 00: lbz r5,0(r3)
349 eieio
350 stbu r5,1(r4)
351 bdnz 00b
352 twi 0,r5,0
353 isync
354 blr
355
356 _GLOBAL(_outsb)
357 cmpwi 0,r5,0
358 mtctr r5
359 subi r4,r4,1
360 blelr-
361 00: lbzu r5,1(r4)
362 stb r5,0(r3)
363 bdnz 00b
364 sync
365 blr
366
367 _GLOBAL(_insw)
368 cmpwi 0,r5,0
369 mtctr r5
370 subi r4,r4,2
371 blelr-
372 00: lhbrx r5,0,r3
373 eieio
374 sthu r5,2(r4)
375 bdnz 00b
376 twi 0,r5,0
377 isync
378 blr
379
380 _GLOBAL(_outsw)
381 cmpwi 0,r5,0
382 mtctr r5
383 subi r4,r4,2
384 blelr-
385 00: lhzu r5,2(r4)
386 sthbrx r5,0,r3
387 bdnz 00b
388 sync
389 blr
390
391 _GLOBAL(_insl)
392 cmpwi 0,r5,0
393 mtctr r5
394 subi r4,r4,4
395 blelr-
396 00: lwbrx r5,0,r3
397 eieio
398 stwu r5,4(r4)
399 bdnz 00b
400 twi 0,r5,0
401 isync
402 blr
403
404 _GLOBAL(_outsl)
405 cmpwi 0,r5,0
406 mtctr r5
407 subi r4,r4,4
408 blelr-
409 00: lwzu r5,4(r4)
410 stwbrx r5,0,r3
411 bdnz 00b
412 sync
413 blr
414
415 /* _GLOBAL(ide_insw) now in drivers/ide/ide-iops.c */
416 _GLOBAL(_insw_ns)
417 cmpwi 0,r5,0
418 mtctr r5
419 subi r4,r4,2
420 blelr-
421 00: lhz r5,0(r3)
422 eieio
423 sthu r5,2(r4)
424 bdnz 00b
425 twi 0,r5,0
426 isync
427 blr
428
429 /* _GLOBAL(ide_outsw) now in drivers/ide/ide-iops.c */
430 _GLOBAL(_outsw_ns)
431 cmpwi 0,r5,0
432 mtctr r5
433 subi r4,r4,2
434 blelr-
435 00: lhzu r5,2(r4)
436 sth r5,0(r3)
437 bdnz 00b
438 sync
439 blr
440
441 _GLOBAL(_insl_ns)
442 cmpwi 0,r5,0
443 mtctr r5
444 subi r4,r4,4
445 blelr-
446 00: lwz r5,0(r3)
447 eieio
448 stwu r5,4(r4)
449 bdnz 00b
450 twi 0,r5,0
451 isync
452 blr
453
454 _GLOBAL(_outsl_ns)
455 cmpwi 0,r5,0
456 mtctr r5
457 subi r4,r4,4
458 blelr-
459 00: lwzu r5,4(r4)
460 stw r5,0(r3)
461 bdnz 00b
462 sync
463 blr
464
465 /*
466 * identify_cpu and calls setup_cpu
467 * In: r3 = base of the cpu_specs array
468 * r4 = address of cur_cpu_spec
469 * r5 = relocation offset
470 */
471 _GLOBAL(identify_cpu)
472 mfpvr r7
473 1:
474 lwz r8,CPU_SPEC_PVR_MASK(r3)
475 and r8,r8,r7
476 lwz r9,CPU_SPEC_PVR_VALUE(r3)
477 cmplw 0,r9,r8
478 beq 1f
479 addi r3,r3,CPU_SPEC_ENTRY_SIZE
480 b 1b
481 1:
482 sub r0,r3,r5
483 std r0,0(r4)
484 ld r4,CPU_SPEC_SETUP(r3)
485 cmpdi 0,r4,0
486 add r4,r4,r5
487 beqlr
488 ld r4,0(r4)
489 add r4,r4,r5
490 mtctr r4
491 /* Calling convention for cpu setup is r3=offset, r4=cur_cpu_spec */
492 mr r4,r3
493 mr r3,r5
494 bctr
495
496 /*
497 * do_cpu_ftr_fixups - goes through the list of CPU feature fixups
498 * and writes nop's over sections of code that don't apply for this cpu.
499 * r3 = data offset (not changed)
500 */
501 _GLOBAL(do_cpu_ftr_fixups)
502 /* Get CPU 0 features */
503 LOAD_REG_IMMEDIATE(r6,cur_cpu_spec)
504 sub r6,r6,r3
505 ld r4,0(r6)
506 sub r4,r4,r3
507 ld r4,CPU_SPEC_FEATURES(r4)
508 /* Get the fixup table */
509 LOAD_REG_IMMEDIATE(r6,__start___ftr_fixup)
510 sub r6,r6,r3
511 LOAD_REG_IMMEDIATE(r7,__stop___ftr_fixup)
512 sub r7,r7,r3
513 /* Do the fixup */
514 1: cmpld r6,r7
515 bgelr
516 addi r6,r6,32
517 ld r8,-32(r6) /* mask */
518 and r8,r8,r4
519 ld r9,-24(r6) /* value */
520 cmpld r8,r9
521 beq 1b
522 ld r8,-16(r6) /* section begin */
523 ld r9,-8(r6) /* section end */
524 subf. r9,r8,r9
525 beq 1b
526 /* write nops over the section of code */
527 /* todo: if large section, add a branch at the start of it */
528 srwi r9,r9,2
529 mtctr r9
530 sub r8,r8,r3
531 lis r0,0x60000000@h /* nop */
532 3: stw r0,0(r8)
533 andi. r10,r4,CPU_FTR_SPLIT_ID_CACHE@l
534 beq 2f
535 dcbst 0,r8 /* suboptimal, but simpler */
536 sync
537 icbi 0,r8
538 2: addi r8,r8,4
539 bdnz 3b
540 sync /* additional sync needed on g4 */
541 isync
542 b 1b
543
544 #if defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE)
545 /*
546 * Do an IO access in real mode
547 */
548 _GLOBAL(real_readb)
549 mfmsr r7
550 ori r0,r7,MSR_DR
551 xori r0,r0,MSR_DR
552 sync
553 mtmsrd r0
554 sync
555 isync
556 mfspr r6,SPRN_HID4
557 rldicl r5,r6,32,0
558 ori r5,r5,0x100
559 rldicl r5,r5,32,0
560 sync
561 mtspr SPRN_HID4,r5
562 isync
563 slbia
564 isync
565 lbz r3,0(r3)
566 sync
567 mtspr SPRN_HID4,r6
568 isync
569 slbia
570 isync
571 mtmsrd r7
572 sync
573 isync
574 blr
575
576 /*
577 * Do an IO access in real mode
578 */
579 _GLOBAL(real_writeb)
580 mfmsr r7
581 ori r0,r7,MSR_DR
582 xori r0,r0,MSR_DR
583 sync
584 mtmsrd r0
585 sync
586 isync
587 mfspr r6,SPRN_HID4
588 rldicl r5,r6,32,0
589 ori r5,r5,0x100
590 rldicl r5,r5,32,0
591 sync
592 mtspr SPRN_HID4,r5
593 isync
594 slbia
595 isync
596 stb r3,0(r4)
597 sync
598 mtspr SPRN_HID4,r6
599 isync
600 slbia
601 isync
602 mtmsrd r7
603 sync
604 isync
605 blr
606 #endif /* defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) */
607
608 /*
609 * SCOM access functions for 970 (FX only for now)
610 *
611 * unsigned long scom970_read(unsigned int address);
612 * void scom970_write(unsigned int address, unsigned long value);
613 *
614 * The address passed in is the 24 bits register address. This code
615 * is 970 specific and will not check the status bits, so you should
616 * know what you are doing.
617 */
618 _GLOBAL(scom970_read)
619 /* interrupts off */
620 mfmsr r4
621 ori r0,r4,MSR_EE
622 xori r0,r0,MSR_EE
623 mtmsrd r0,1
624
625 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
626 * (including parity). On current CPUs they must be 0'd,
627 * and finally or in RW bit
628 */
629 rlwinm r3,r3,8,0,15
630 ori r3,r3,0x8000
631
632 /* do the actual scom read */
633 sync
634 mtspr SPRN_SCOMC,r3
635 isync
636 mfspr r3,SPRN_SCOMD
637 isync
638 mfspr r0,SPRN_SCOMC
639 isync
640
641 /* XXX: fixup result on some buggy 970's (ouch ! we lost a bit, bah
642 * that's the best we can do). Not implemented yet as we don't use
643 * the scom on any of the bogus CPUs yet, but may have to be done
644 * ultimately
645 */
646
647 /* restore interrupts */
648 mtmsrd r4,1
649 blr
650
651
652 _GLOBAL(scom970_write)
653 /* interrupts off */
654 mfmsr r5
655 ori r0,r5,MSR_EE
656 xori r0,r0,MSR_EE
657 mtmsrd r0,1
658
659 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
660 * (including parity). On current CPUs they must be 0'd.
661 */
662
663 rlwinm r3,r3,8,0,15
664
665 sync
666 mtspr SPRN_SCOMD,r4 /* write data */
667 isync
668 mtspr SPRN_SCOMC,r3 /* write command */
669 isync
670 mfspr 3,SPRN_SCOMC
671 isync
672
673 /* restore interrupts */
674 mtmsrd r5,1
675 blr
676
677
678 /*
679 * Create a kernel thread
680 * kernel_thread(fn, arg, flags)
681 */
682 _GLOBAL(kernel_thread)
683 std r29,-24(r1)
684 std r30,-16(r1)
685 stdu r1,-STACK_FRAME_OVERHEAD(r1)
686 mr r29,r3
687 mr r30,r4
688 ori r3,r5,CLONE_VM /* flags */
689 oris r3,r3,(CLONE_UNTRACED>>16)
690 li r4,0 /* new sp (unused) */
691 li r0,__NR_clone
692 sc
693 cmpdi 0,r3,0 /* parent or child? */
694 bne 1f /* return if parent */
695 li r0,0
696 stdu r0,-STACK_FRAME_OVERHEAD(r1)
697 ld r2,8(r29)
698 ld r29,0(r29)
699 mtlr r29 /* fn addr in lr */
700 mr r3,r30 /* load arg and call fn */
701 blrl
702 li r0,__NR_exit /* exit after child exits */
703 li r3,0
704 sc
705 1: addi r1,r1,STACK_FRAME_OVERHEAD
706 ld r29,-24(r1)
707 ld r30,-16(r1)
708 blr
709
710 /*
711 * disable_kernel_fp()
712 * Disable the FPU.
713 */
714 _GLOBAL(disable_kernel_fp)
715 mfmsr r3
716 rldicl r0,r3,(63-MSR_FP_LG),1
717 rldicl r3,r0,(MSR_FP_LG+1),0
718 mtmsrd r3 /* disable use of fpu now */
719 isync
720 blr
721
722 #ifdef CONFIG_ALTIVEC
723
724 #if 0 /* this has no callers for now */
725 /*
726 * disable_kernel_altivec()
727 * Disable the VMX.
728 */
729 _GLOBAL(disable_kernel_altivec)
730 mfmsr r3
731 rldicl r0,r3,(63-MSR_VEC_LG),1
732 rldicl r3,r0,(MSR_VEC_LG+1),0
733 mtmsrd r3 /* disable use of VMX now */
734 isync
735 blr
736 #endif /* 0 */
737
738 /*
739 * giveup_altivec(tsk)
740 * Disable VMX for the task given as the argument,
741 * and save the vector registers in its thread_struct.
742 * Enables the VMX for use in the kernel on return.
743 */
744 _GLOBAL(giveup_altivec)
745 mfmsr r5
746 oris r5,r5,MSR_VEC@h
747 mtmsrd r5 /* enable use of VMX now */
748 isync
749 cmpdi 0,r3,0
750 beqlr- /* if no previous owner, done */
751 addi r3,r3,THREAD /* want THREAD of task */
752 ld r5,PT_REGS(r3)
753 cmpdi 0,r5,0
754 SAVE_32VRS(0,r4,r3)
755 mfvscr vr0
756 li r4,THREAD_VSCR
757 stvx vr0,r4,r3
758 beq 1f
759 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
760 lis r3,MSR_VEC@h
761 andc r4,r4,r3 /* disable FP for previous task */
762 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
763 1:
764 #ifndef CONFIG_SMP
765 li r5,0
766 ld r4,last_task_used_altivec@got(r2)
767 std r5,0(r4)
768 #endif /* CONFIG_SMP */
769 blr
770
771 #endif /* CONFIG_ALTIVEC */
772
773 _GLOBAL(execve)
774 li r0,__NR_execve
775 sc
776 bnslr
777 neg r3,r3
778 blr
779
780 /* kexec_wait(phys_cpu)
781 *
782 * wait for the flag to change, indicating this kernel is going away but
783 * the slave code for the next one is at addresses 0 to 100.
784 *
785 * This is used by all slaves.
786 *
787 * Physical (hardware) cpu id should be in r3.
788 */
789 _GLOBAL(kexec_wait)
790 bl 1f
791 1: mflr r5
792 addi r5,r5,kexec_flag-1b
793
794 99: HMT_LOW
795 #ifdef CONFIG_KEXEC /* use no memory without kexec */
796 lwz r4,0(r5)
797 cmpwi 0,r4,0
798 bnea 0x60
799 #endif
800 b 99b
801
802 /* this can be in text because we won't change it until we are
803 * running in real anyways
804 */
805 kexec_flag:
806 .long 0
807
808
809 #ifdef CONFIG_KEXEC
810
811 /* kexec_smp_wait(void)
812 *
813 * call with interrupts off
814 * note: this is a terminal routine, it does not save lr
815 *
816 * get phys id from paca
817 * set paca id to -1 to say we got here
818 * switch to real mode
819 * join other cpus in kexec_wait(phys_id)
820 */
821 _GLOBAL(kexec_smp_wait)
822 lhz r3,PACAHWCPUID(r13)
823 li r4,-1
824 sth r4,PACAHWCPUID(r13) /* let others know we left */
825 bl real_mode
826 b .kexec_wait
827
828 /*
829 * switch to real mode (turn mmu off)
830 * we use the early kernel trick that the hardware ignores bits
831 * 0 and 1 (big endian) of the effective address in real mode
832 *
833 * don't overwrite r3 here, it is live for kexec_wait above.
834 */
835 real_mode: /* assume normal blr return */
836 1: li r9,MSR_RI
837 li r10,MSR_DR|MSR_IR
838 mflr r11 /* return address to SRR0 */
839 mfmsr r12
840 andc r9,r12,r9
841 andc r10,r12,r10
842
843 mtmsrd r9,1
844 mtspr SPRN_SRR1,r10
845 mtspr SPRN_SRR0,r11
846 rfid
847
848
849 /*
850 * kexec_sequence(newstack, start, image, control, clear_all())
851 *
852 * does the grungy work with stack switching and real mode switches
853 * also does simple calls to other code
854 */
855
856 _GLOBAL(kexec_sequence)
857 mflr r0
858 std r0,16(r1)
859
860 /* switch stacks to newstack -- &kexec_stack.stack */
861 stdu r1,THREAD_SIZE-112(r3)
862 mr r1,r3
863
864 li r0,0
865 std r0,16(r1)
866
867 /* save regs for local vars on new stack.
868 * yes, we won't go back, but ...
869 */
870 std r31,-8(r1)
871 std r30,-16(r1)
872 std r29,-24(r1)
873 std r28,-32(r1)
874 std r27,-40(r1)
875 std r26,-48(r1)
876 std r25,-56(r1)
877
878 stdu r1,-112-64(r1)
879
880 /* save args into preserved regs */
881 mr r31,r3 /* newstack (both) */
882 mr r30,r4 /* start (real) */
883 mr r29,r5 /* image (virt) */
884 mr r28,r6 /* control, unused */
885 mr r27,r7 /* clear_all() fn desc */
886 mr r26,r8 /* spare */
887 lhz r25,PACAHWCPUID(r13) /* get our phys cpu from paca */
888
889 /* disable interrupts, we are overwriting kernel data next */
890 mfmsr r3
891 rlwinm r3,r3,0,17,15
892 mtmsrd r3,1
893
894 /* copy dest pages, flush whole dest image */
895 mr r3,r29
896 bl .kexec_copy_flush /* (image) */
897
898 /* turn off mmu */
899 bl real_mode
900
901 /* clear out hardware hash page table and tlb */
902 ld r5,0(r27) /* deref function descriptor */
903 mtctr r5
904 bctrl /* ppc_md.hash_clear_all(void); */
905
906 /*
907 * kexec image calling is:
908 * the first 0x100 bytes of the entry point are copied to 0
909 *
910 * all slaves branch to slave = 0x60 (absolute)
911 * slave(phys_cpu_id);
912 *
913 * master goes to start = entry point
914 * start(phys_cpu_id, start, 0);
915 *
916 *
917 * a wrapper is needed to call existing kernels, here is an approximate
918 * description of one method:
919 *
920 * v2: (2.6.10)
921 * start will be near the boot_block (maybe 0x100 bytes before it?)
922 * it will have a 0x60, which will b to boot_block, where it will wait
923 * and 0 will store phys into struct boot-block and load r3 from there,
924 * copy kernel 0-0x100 and tell slaves to back down to 0x60 again
925 *
926 * v1: (2.6.9)
927 * boot block will have all cpus scanning device tree to see if they
928 * are the boot cpu ?????
929 * other device tree differences (prop sizes, va vs pa, etc)...
930 */
931
932 /* copy 0x100 bytes starting at start to 0 */
933 li r3,0
934 mr r4,r30
935 li r5,0x100
936 li r6,0
937 bl .copy_and_flush /* (dest, src, copy limit, start offset) */
938 1: /* assume normal blr return */
939
940 /* release other cpus to the new kernel secondary start at 0x60 */
941 mflr r5
942 li r6,1
943 stw r6,kexec_flag-1b(5)
944 mr r3,r25 # my phys cpu
945 mr r4,r30 # start, aka phys mem offset
946 mtlr 4
947 li r5,0
948 blr /* image->start(physid, image->start, 0); */
949 #endif /* CONFIG_KEXEC */
This page took 0.056171 seconds and 5 git commands to generate.