Loading...
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#include <asm/export.h>
37
38 .text
39
40/*
41 * We store the saved ksp_limit in the unused part
42 * of the STACK_FRAME_OVERHEAD
43 */
44_GLOBAL(call_do_softirq)
45 mflr r0
46 stw r0,4(r1)
47 lwz r10,THREAD+KSP_LIMIT(r2)
48 addi r11,r3,THREAD_INFO_GAP
49 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
50 mr r1,r3
51 stw r10,8(r1)
52 stw r11,THREAD+KSP_LIMIT(r2)
53 bl __do_softirq
54 lwz r10,8(r1)
55 lwz r1,0(r1)
56 lwz r0,4(r1)
57 stw r10,THREAD+KSP_LIMIT(r2)
58 mtlr r0
59 blr
60
61/*
62 * void call_do_irq(struct pt_regs *regs, struct thread_info *irqtp);
63 */
64_GLOBAL(call_do_irq)
65 mflr r0
66 stw r0,4(r1)
67 lwz r10,THREAD+KSP_LIMIT(r2)
68 addi r11,r4,THREAD_INFO_GAP
69 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
70 mr r1,r4
71 stw r10,8(r1)
72 stw r11,THREAD+KSP_LIMIT(r2)
73 bl __do_irq
74 lwz r10,8(r1)
75 lwz r1,0(r1)
76 lwz r0,4(r1)
77 stw r10,THREAD+KSP_LIMIT(r2)
78 mtlr r0
79 blr
80
81/*
82 * This returns the high 64 bits of the product of two 64-bit numbers.
83 */
84_GLOBAL(mulhdu)
85 cmpwi r6,0
86 cmpwi cr1,r3,0
87 mr r10,r4
88 mulhwu r4,r4,r5
89 beq 1f
90 mulhwu r0,r10,r6
91 mullw r7,r10,r5
92 addc r7,r0,r7
93 addze r4,r4
941: beqlr cr1 /* all done if high part of A is 0 */
95 mullw r9,r3,r5
96 mulhwu r10,r3,r5
97 beq 2f
98 mullw r0,r3,r6
99 mulhwu r8,r3,r6
100 addc r7,r0,r7
101 adde r4,r4,r8
102 addze r10,r10
1032: addc r4,r4,r9
104 addze r3,r10
105 blr
106
107/*
108 * reloc_got2 runs through the .got2 section adding an offset
109 * to each entry.
110 */
111_GLOBAL(reloc_got2)
112 mflr r11
113 lis r7,__got2_start@ha
114 addi r7,r7,__got2_start@l
115 lis r8,__got2_end@ha
116 addi r8,r8,__got2_end@l
117 subf r8,r7,r8
118 srwi. r8,r8,2
119 beqlr
120 mtctr r8
121 bl 1f
1221: mflr r0
123 lis r4,1b@ha
124 addi r4,r4,1b@l
125 subf r0,r4,r0
126 add r7,r0,r7
1272: lwz r0,0(r7)
128 add r0,r0,r3
129 stw r0,0(r7)
130 addi r7,r7,4
131 bdnz 2b
132 mtlr r11
133 blr
134
135/*
136 * call_setup_cpu - call the setup_cpu function for this cpu
137 * r3 = data offset, r24 = cpu number
138 *
139 * Setup function is called with:
140 * r3 = data offset
141 * r4 = ptr to CPU spec (relocated)
142 */
143_GLOBAL(call_setup_cpu)
144 addis r4,r3,cur_cpu_spec@ha
145 addi r4,r4,cur_cpu_spec@l
146 lwz r4,0(r4)
147 add r4,r4,r3
148 lwz r5,CPU_SPEC_SETUP(r4)
149 cmpwi 0,r5,0
150 add r5,r5,r3
151 beqlr
152 mtctr r5
153 bctr
154
155#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
156
157/* This gets called by via-pmu.c to switch the PLL selection
158 * on 750fx CPU. This function should really be moved to some
159 * other place (as most of the cpufreq code in via-pmu
160 */
161_GLOBAL(low_choose_750fx_pll)
162 /* Clear MSR:EE */
163 mfmsr r7
164 rlwinm r0,r7,0,17,15
165 mtmsr r0
166
167 /* If switching to PLL1, disable HID0:BTIC */
168 cmplwi cr0,r3,0
169 beq 1f
170 mfspr r5,SPRN_HID0
171 rlwinm r5,r5,0,27,25
172 sync
173 mtspr SPRN_HID0,r5
174 isync
175 sync
176
1771:
178 /* Calc new HID1 value */
179 mfspr r4,SPRN_HID1 /* Build a HID1:PS bit from parameter */
180 rlwinm r5,r3,16,15,15 /* Clear out HID1:PS from value read */
181 rlwinm r4,r4,0,16,14 /* Could have I used rlwimi here ? */
182 or r4,r4,r5
183 mtspr SPRN_HID1,r4
184
185 /* Store new HID1 image */
186 CURRENT_THREAD_INFO(r6, r1)
187 lwz r6,TI_CPU(r6)
188 slwi r6,r6,2
189 addis r6,r6,nap_save_hid1@ha
190 stw r4,nap_save_hid1@l(r6)
191
192 /* If switching to PLL0, enable HID0:BTIC */
193 cmplwi cr0,r3,0
194 bne 1f
195 mfspr r5,SPRN_HID0
196 ori r5,r5,HID0_BTIC
197 sync
198 mtspr SPRN_HID0,r5
199 isync
200 sync
201
2021:
203 /* Return */
204 mtmsr r7
205 blr
206
207_GLOBAL(low_choose_7447a_dfs)
208 /* Clear MSR:EE */
209 mfmsr r7
210 rlwinm r0,r7,0,17,15
211 mtmsr r0
212
213 /* Calc new HID1 value */
214 mfspr r4,SPRN_HID1
215 insrwi r4,r3,1,9 /* insert parameter into bit 9 */
216 sync
217 mtspr SPRN_HID1,r4
218 sync
219 isync
220
221 /* Return */
222 mtmsr r7
223 blr
224
225#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
226
227/*
228 * complement mask on the msr then "or" some values on.
229 * _nmask_and_or_msr(nmask, value_to_or)
230 */
231_GLOBAL(_nmask_and_or_msr)
232 mfmsr r0 /* Get current msr */
233 andc r0,r0,r3 /* And off the bits set in r3 (first parm) */
234 or r0,r0,r4 /* Or on the bits in r4 (second parm) */
235 SYNC /* Some chip revs have problems here... */
236 mtmsr r0 /* Update machine state */
237 isync
238 blr /* Done */
239
240#ifdef CONFIG_40x
241
242/*
243 * Do an IO access in real mode
244 */
245_GLOBAL(real_readb)
246 mfmsr r7
247 rlwinm r0,r7,0,~MSR_DR
248 sync
249 mtmsr r0
250 sync
251 isync
252 lbz r3,0(r3)
253 sync
254 mtmsr r7
255 sync
256 isync
257 blr
258
259 /*
260 * Do an IO access in real mode
261 */
262_GLOBAL(real_writeb)
263 mfmsr r7
264 rlwinm r0,r7,0,~MSR_DR
265 sync
266 mtmsr r0
267 sync
268 isync
269 stb r3,0(r4)
270 sync
271 mtmsr r7
272 sync
273 isync
274 blr
275
276#endif /* CONFIG_40x */
277
278
279/*
280 * Flush instruction cache.
281 * This is a no-op on the 601.
282 */
283#ifndef CONFIG_PPC_8xx
284_GLOBAL(flush_instruction_cache)
285#if defined(CONFIG_4xx)
286#ifdef CONFIG_403GCX
287 li r3, 512
288 mtctr r3
289 lis r4, KERNELBASE@h
2901: iccci 0, r4
291 addi r4, r4, 16
292 bdnz 1b
293#else
294 lis r3, KERNELBASE@h
295 iccci 0,r3
296#endif
297#elif defined(CONFIG_FSL_BOOKE)
298BEGIN_FTR_SECTION
299 mfspr r3,SPRN_L1CSR0
300 ori r3,r3,L1CSR0_CFI|L1CSR0_CLFC
301 /* msync; isync recommended here */
302 mtspr SPRN_L1CSR0,r3
303 isync
304 blr
305END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
306 mfspr r3,SPRN_L1CSR1
307 ori r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
308 mtspr SPRN_L1CSR1,r3
309#else
310 mfspr r3,SPRN_PVR
311 rlwinm r3,r3,16,16,31
312 cmpwi 0,r3,1
313 beqlr /* for 601, do nothing */
314 /* 603/604 processor - use invalidate-all bit in HID0 */
315 mfspr r3,SPRN_HID0
316 ori r3,r3,HID0_ICFI
317 mtspr SPRN_HID0,r3
318#endif /* CONFIG_4xx */
319 isync
320 blr
321EXPORT_SYMBOL(flush_instruction_cache)
322#endif /* CONFIG_PPC_8xx */
323
324/*
325 * Write any modified data cache blocks out to memory
326 * and invalidate the corresponding instruction cache blocks.
327 * This is a no-op on the 601.
328 *
329 * flush_icache_range(unsigned long start, unsigned long stop)
330 */
331_GLOBAL(flush_icache_range)
332BEGIN_FTR_SECTION
333 PURGE_PREFETCHED_INS
334 blr /* for 601, do nothing */
335END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
336 rlwinm r3,r3,0,0,31 - L1_CACHE_SHIFT
337 subf r4,r3,r4
338 addi r4,r4,L1_CACHE_BYTES - 1
339 srwi. r4,r4,L1_CACHE_SHIFT
340 beqlr
341 mtctr r4
342 mr r6,r3
3431: dcbst 0,r3
344 addi r3,r3,L1_CACHE_BYTES
345 bdnz 1b
346 sync /* wait for dcbst's to get to ram */
347#ifndef CONFIG_44x
348 mtctr r4
3492: icbi 0,r6
350 addi r6,r6,L1_CACHE_BYTES
351 bdnz 2b
352#else
353 /* Flash invalidate on 44x because we are passed kmapped addresses and
354 this doesn't work for userspace pages due to the virtually tagged
355 icache. Sigh. */
356 iccci 0, r0
357#endif
358 sync /* additional sync needed on g4 */
359 isync
360 blr
361_ASM_NOKPROBE_SYMBOL(flush_icache_range)
362EXPORT_SYMBOL(flush_icache_range)
363
364/*
365 * Flush a particular page from the data cache to RAM.
366 * Note: this is necessary because the instruction cache does *not*
367 * snoop from the data cache.
368 * This is a no-op on the 601 which has a unified cache.
369 *
370 * void __flush_dcache_icache(void *page)
371 */
372_GLOBAL(__flush_dcache_icache)
373BEGIN_FTR_SECTION
374 PURGE_PREFETCHED_INS
375 blr
376END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
377 rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */
378 li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */
379 mtctr r4
380 mr r6,r3
3810: dcbst 0,r3 /* Write line to ram */
382 addi r3,r3,L1_CACHE_BYTES
383 bdnz 0b
384 sync
385#ifdef CONFIG_44x
386 /* We don't flush the icache on 44x. Those have a virtual icache
387 * and we don't have access to the virtual address here (it's
388 * not the page vaddr but where it's mapped in user space). The
389 * flushing of the icache on these is handled elsewhere, when
390 * a change in the address space occurs, before returning to
391 * user space
392 */
393BEGIN_MMU_FTR_SECTION
394 blr
395END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
396#endif /* CONFIG_44x */
397 mtctr r4
3981: icbi 0,r6
399 addi r6,r6,L1_CACHE_BYTES
400 bdnz 1b
401 sync
402 isync
403 blr
404
405#ifndef CONFIG_BOOKE
406/*
407 * Flush a particular page from the data cache to RAM, identified
408 * by its physical address. We turn off the MMU so we can just use
409 * the physical address (this may be a highmem page without a kernel
410 * mapping).
411 *
412 * void __flush_dcache_icache_phys(unsigned long physaddr)
413 */
414_GLOBAL(__flush_dcache_icache_phys)
415BEGIN_FTR_SECTION
416 PURGE_PREFETCHED_INS
417 blr /* for 601, do nothing */
418END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
419 mfmsr r10
420 rlwinm r0,r10,0,28,26 /* clear DR */
421 mtmsr r0
422 isync
423 rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */
424 li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */
425 mtctr r4
426 mr r6,r3
4270: dcbst 0,r3 /* Write line to ram */
428 addi r3,r3,L1_CACHE_BYTES
429 bdnz 0b
430 sync
431 mtctr r4
4321: icbi 0,r6
433 addi r6,r6,L1_CACHE_BYTES
434 bdnz 1b
435 sync
436 mtmsr r10 /* restore DR */
437 isync
438 blr
439#endif /* CONFIG_BOOKE */
440
441/*
442 * Copy a whole page. We use the dcbz instruction on the destination
443 * to reduce memory traffic (it eliminates the unnecessary reads of
444 * the destination into cache). This requires that the destination
445 * is cacheable.
446 */
447#define COPY_16_BYTES \
448 lwz r6,4(r4); \
449 lwz r7,8(r4); \
450 lwz r8,12(r4); \
451 lwzu r9,16(r4); \
452 stw r6,4(r3); \
453 stw r7,8(r3); \
454 stw r8,12(r3); \
455 stwu r9,16(r3)
456
457_GLOBAL(copy_page)
458 addi r3,r3,-4
459 addi r4,r4,-4
460
461 li r5,4
462
463#if MAX_COPY_PREFETCH > 1
464 li r0,MAX_COPY_PREFETCH
465 li r11,4
466 mtctr r0
46711: dcbt r11,r4
468 addi r11,r11,L1_CACHE_BYTES
469 bdnz 11b
470#else /* MAX_COPY_PREFETCH == 1 */
471 dcbt r5,r4
472 li r11,L1_CACHE_BYTES+4
473#endif /* MAX_COPY_PREFETCH */
474 li r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
475 crclr 4*cr0+eq
4762:
477 mtctr r0
4781:
479 dcbt r11,r4
480 dcbz r5,r3
481 COPY_16_BYTES
482#if L1_CACHE_BYTES >= 32
483 COPY_16_BYTES
484#if L1_CACHE_BYTES >= 64
485 COPY_16_BYTES
486 COPY_16_BYTES
487#if L1_CACHE_BYTES >= 128
488 COPY_16_BYTES
489 COPY_16_BYTES
490 COPY_16_BYTES
491 COPY_16_BYTES
492#endif
493#endif
494#endif
495 bdnz 1b
496 beqlr
497 crnot 4*cr0+eq,4*cr0+eq
498 li r0,MAX_COPY_PREFETCH
499 li r11,4
500 b 2b
501EXPORT_SYMBOL(copy_page)
502
503/*
504 * Extended precision shifts.
505 *
506 * Updated to be valid for shift counts from 0 to 63 inclusive.
507 * -- Gabriel
508 *
509 * R3/R4 has 64 bit value
510 * R5 has shift count
511 * result in R3/R4
512 *
513 * ashrdi3: arithmetic right shift (sign propagation)
514 * lshrdi3: logical right shift
515 * ashldi3: left shift
516 */
517_GLOBAL(__ashrdi3)
518 subfic r6,r5,32
519 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
520 addi r7,r5,32 # could be xori, or addi with -32
521 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
522 rlwinm r8,r7,0,32 # t3 = (count < 32) ? 32 : 0
523 sraw r7,r3,r7 # t2 = MSW >> (count-32)
524 or r4,r4,r6 # LSW |= t1
525 slw r7,r7,r8 # t2 = (count < 32) ? 0 : t2
526 sraw r3,r3,r5 # MSW = MSW >> count
527 or r4,r4,r7 # LSW |= t2
528 blr
529EXPORT_SYMBOL(__ashrdi3)
530
531_GLOBAL(__ashldi3)
532 subfic r6,r5,32
533 slw r3,r3,r5 # MSW = count > 31 ? 0 : MSW << count
534 addi r7,r5,32 # could be xori, or addi with -32
535 srw r6,r4,r6 # t1 = count > 31 ? 0 : LSW >> (32-count)
536 slw r7,r4,r7 # t2 = count < 32 ? 0 : LSW << (count-32)
537 or r3,r3,r6 # MSW |= t1
538 slw r4,r4,r5 # LSW = LSW << count
539 or r3,r3,r7 # MSW |= t2
540 blr
541EXPORT_SYMBOL(__ashldi3)
542
543_GLOBAL(__lshrdi3)
544 subfic r6,r5,32
545 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
546 addi r7,r5,32 # could be xori, or addi with -32
547 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
548 srw r7,r3,r7 # t2 = count < 32 ? 0 : MSW >> (count-32)
549 or r4,r4,r6 # LSW |= t1
550 srw r3,r3,r5 # MSW = MSW >> count
551 or r4,r4,r7 # LSW |= t2
552 blr
553EXPORT_SYMBOL(__lshrdi3)
554
555/*
556 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
557 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
558 */
559_GLOBAL(__cmpdi2)
560 cmpw r3,r5
561 li r3,1
562 bne 1f
563 cmplw r4,r6
564 beqlr
5651: li r3,0
566 bltlr
567 li r3,2
568 blr
569EXPORT_SYMBOL(__cmpdi2)
570/*
571 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
572 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
573 */
574_GLOBAL(__ucmpdi2)
575 cmplw r3,r5
576 li r3,1
577 bne 1f
578 cmplw r4,r6
579 beqlr
5801: li r3,0
581 bltlr
582 li r3,2
583 blr
584EXPORT_SYMBOL(__ucmpdi2)
585
586_GLOBAL(__bswapdi2)
587 rotlwi r9,r4,8
588 rotlwi r10,r3,8
589 rlwimi r9,r4,24,0,7
590 rlwimi r10,r3,24,0,7
591 rlwimi r9,r4,24,16,23
592 rlwimi r10,r3,24,16,23
593 mr r3,r9
594 mr r4,r10
595 blr
596EXPORT_SYMBOL(__bswapdi2)
597
598#ifdef CONFIG_SMP
599_GLOBAL(start_secondary_resume)
600 /* Reset stack */
601 CURRENT_THREAD_INFO(r1, r1)
602 addi r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
603 li r3,0
604 stw r3,0(r1) /* Zero the stack frame pointer */
605 bl start_secondary
606 b .
607#endif /* CONFIG_SMP */
608
609/*
610 * This routine is just here to keep GCC happy - sigh...
611 */
612_GLOBAL(__main)
613 blr
614
615#ifdef CONFIG_KEXEC_CORE
616 /*
617 * Must be relocatable PIC code callable as a C function.
618 */
619 .globl relocate_new_kernel
620relocate_new_kernel:
621 /* r3 = page_list */
622 /* r4 = reboot_code_buffer */
623 /* r5 = start_address */
624
625#ifdef CONFIG_FSL_BOOKE
626
627 mr r29, r3
628 mr r30, r4
629 mr r31, r5
630
631#define ENTRY_MAPPING_KEXEC_SETUP
632#include "fsl_booke_entry_mapping.S"
633#undef ENTRY_MAPPING_KEXEC_SETUP
634
635 mr r3, r29
636 mr r4, r30
637 mr r5, r31
638
639 li r0, 0
640#elif defined(CONFIG_44x)
641
642 /* Save our parameters */
643 mr r29, r3
644 mr r30, r4
645 mr r31, r5
646
647#ifdef CONFIG_PPC_47x
648 /* Check for 47x cores */
649 mfspr r3,SPRN_PVR
650 srwi r3,r3,16
651 cmplwi cr0,r3,PVR_476FPE@h
652 beq setup_map_47x
653 cmplwi cr0,r3,PVR_476@h
654 beq setup_map_47x
655 cmplwi cr0,r3,PVR_476_ISS@h
656 beq setup_map_47x
657#endif /* CONFIG_PPC_47x */
658
659/*
660 * Code for setting up 1:1 mapping for PPC440x for KEXEC
661 *
662 * We cannot switch off the MMU on PPC44x.
663 * So we:
664 * 1) Invalidate all the mappings except the one we are running from.
665 * 2) Create a tmp mapping for our code in the other address space(TS) and
666 * jump to it. Invalidate the entry we started in.
667 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
668 * 4) Jump to the 1:1 mapping in original TS.
669 * 5) Invalidate the tmp mapping.
670 *
671 * - Based on the kexec support code for FSL BookE
672 *
673 */
674
675 /*
676 * Load the PID with kernel PID (0).
677 * Also load our MSR_IS and TID to MMUCR for TLB search.
678 */
679 li r3, 0
680 mtspr SPRN_PID, r3
681 mfmsr r4
682 andi. r4,r4,MSR_IS@l
683 beq wmmucr
684 oris r3,r3,PPC44x_MMUCR_STS@h
685wmmucr:
686 mtspr SPRN_MMUCR,r3
687 sync
688
689 /*
690 * Invalidate all the TLB entries except the current entry
691 * where we are running from
692 */
693 bl 0f /* Find our address */
6940: mflr r5 /* Make it accessible */
695 tlbsx r23,0,r5 /* Find entry we are in */
696 li r4,0 /* Start at TLB entry 0 */
697 li r3,0 /* Set PAGEID inval value */
6981: cmpw r23,r4 /* Is this our entry? */
699 beq skip /* If so, skip the inval */
700 tlbwe r3,r4,PPC44x_TLB_PAGEID /* If not, inval the entry */
701skip:
702 addi r4,r4,1 /* Increment */
703 cmpwi r4,64 /* Are we done? */
704 bne 1b /* If not, repeat */
705 isync
706
707 /* Create a temp mapping and jump to it */
708 andi. r6, r23, 1 /* Find the index to use */
709 addi r24, r6, 1 /* r24 will contain 1 or 2 */
710
711 mfmsr r9 /* get the MSR */
712 rlwinm r5, r9, 27, 31, 31 /* Extract the MSR[IS] */
713 xori r7, r5, 1 /* Use the other address space */
714
715 /* Read the current mapping entries */
716 tlbre r3, r23, PPC44x_TLB_PAGEID
717 tlbre r4, r23, PPC44x_TLB_XLAT
718 tlbre r5, r23, PPC44x_TLB_ATTRIB
719
720 /* Save our current XLAT entry */
721 mr r25, r4
722
723 /* Extract the TLB PageSize */
724 li r10, 1 /* r10 will hold PageSize */
725 rlwinm r11, r3, 0, 24, 27 /* bits 24-27 */
726
727 /* XXX: As of now we use 256M, 4K pages */
728 cmpwi r11, PPC44x_TLB_256M
729 bne tlb_4k
730 rotlwi r10, r10, 28 /* r10 = 256M */
731 b write_out
732tlb_4k:
733 cmpwi r11, PPC44x_TLB_4K
734 bne default
735 rotlwi r10, r10, 12 /* r10 = 4K */
736 b write_out
737default:
738 rotlwi r10, r10, 10 /* r10 = 1K */
739
740write_out:
741 /*
742 * Write out the tmp 1:1 mapping for this code in other address space
743 * Fixup EPN = RPN , TS=other address space
744 */
745 insrwi r3, r7, 1, 23 /* Bit 23 is TS for PAGEID field */
746
747 /* Write out the tmp mapping entries */
748 tlbwe r3, r24, PPC44x_TLB_PAGEID
749 tlbwe r4, r24, PPC44x_TLB_XLAT
750 tlbwe r5, r24, PPC44x_TLB_ATTRIB
751
752 subi r11, r10, 1 /* PageOffset Mask = PageSize - 1 */
753 not r10, r11 /* Mask for PageNum */
754
755 /* Switch to other address space in MSR */
756 insrwi r9, r7, 1, 26 /* Set MSR[IS] = r7 */
757
758 bl 1f
7591: mflr r8
760 addi r8, r8, (2f-1b) /* Find the target offset */
761
762 /* Jump to the tmp mapping */
763 mtspr SPRN_SRR0, r8
764 mtspr SPRN_SRR1, r9
765 rfi
766
7672:
768 /* Invalidate the entry we were executing from */
769 li r3, 0
770 tlbwe r3, r23, PPC44x_TLB_PAGEID
771
772 /* attribute fields. rwx for SUPERVISOR mode */
773 li r5, 0
774 ori r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
775
776 /* Create 1:1 mapping in 256M pages */
777 xori r7, r7, 1 /* Revert back to Original TS */
778
779 li r8, 0 /* PageNumber */
780 li r6, 3 /* TLB Index, start at 3 */
781
782next_tlb:
783 rotlwi r3, r8, 28 /* Create EPN (bits 0-3) */
784 mr r4, r3 /* RPN = EPN */
785 ori r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
786 insrwi r3, r7, 1, 23 /* Set TS from r7 */
787
788 tlbwe r3, r6, PPC44x_TLB_PAGEID /* PageID field : EPN, V, SIZE */
789 tlbwe r4, r6, PPC44x_TLB_XLAT /* Address translation : RPN */
790 tlbwe r5, r6, PPC44x_TLB_ATTRIB /* Attributes */
791
792 addi r8, r8, 1 /* Increment PN */
793 addi r6, r6, 1 /* Increment TLB Index */
794 cmpwi r8, 8 /* Are we done ? */
795 bne next_tlb
796 isync
797
798 /* Jump to the new mapping 1:1 */
799 li r9,0
800 insrwi r9, r7, 1, 26 /* Set MSR[IS] = r7 */
801
802 bl 1f
8031: mflr r8
804 and r8, r8, r11 /* Get our offset within page */
805 addi r8, r8, (2f-1b)
806
807 and r5, r25, r10 /* Get our target PageNum */
808 or r8, r8, r5 /* Target jump address */
809
810 mtspr SPRN_SRR0, r8
811 mtspr SPRN_SRR1, r9
812 rfi
8132:
814 /* Invalidate the tmp entry we used */
815 li r3, 0
816 tlbwe r3, r24, PPC44x_TLB_PAGEID
817 sync
818 b ppc44x_map_done
819
820#ifdef CONFIG_PPC_47x
821
822 /* 1:1 mapping for 47x */
823
824setup_map_47x:
825
826 /*
827 * Load the kernel pid (0) to PID and also to MMUCR[TID].
828 * Also set the MSR IS->MMUCR STS
829 */
830 li r3, 0
831 mtspr SPRN_PID, r3 /* Set PID */
832 mfmsr r4 /* Get MSR */
833 andi. r4, r4, MSR_IS@l /* TS=1? */
834 beq 1f /* If not, leave STS=0 */
835 oris r3, r3, PPC47x_MMUCR_STS@h /* Set STS=1 */
8361: mtspr SPRN_MMUCR, r3 /* Put MMUCR */
837 sync
838
839 /* Find the entry we are running from */
840 bl 2f
8412: mflr r23
842 tlbsx r23, 0, r23
843 tlbre r24, r23, 0 /* TLB Word 0 */
844 tlbre r25, r23, 1 /* TLB Word 1 */
845 tlbre r26, r23, 2 /* TLB Word 2 */
846
847
848 /*
849 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
850 * of 4k page size in all 4 ways (0-3 in r3).
851 * This would invalidate the entire UTLB including the one we are
852 * running from. However the shadow TLB entries would help us
853 * to continue the execution, until we flush them (rfi/isync).
854 */
855 addis r3, 0, 0x8000 /* specify the way */
856 addi r4, 0, 0 /* TLB Word0 = (EPN=0, VALID = 0) */
857 addi r5, 0, 0
858 b clear_utlb_entry
859
860 /* Align the loop to speed things up. from head_44x.S */
861 .align 6
862
863clear_utlb_entry:
864
865 tlbwe r4, r3, 0
866 tlbwe r5, r3, 1
867 tlbwe r5, r3, 2
868 addis r3, r3, 0x2000 /* Increment the way */
869 cmpwi r3, 0
870 bne clear_utlb_entry
871 addis r3, 0, 0x8000
872 addis r4, r4, 0x100 /* Increment the EPN */
873 cmpwi r4, 0
874 bne clear_utlb_entry
875
876 /* Create the entries in the other address space */
877 mfmsr r5
878 rlwinm r7, r5, 27, 31, 31 /* Get the TS (Bit 26) from MSR */
879 xori r7, r7, 1 /* r7 = !TS */
880
881 insrwi r24, r7, 1, 21 /* Change the TS in the saved TLB word 0 */
882
883 /*
884 * write out the TLB entries for the tmp mapping
885 * Use way '0' so that we could easily invalidate it later.
886 */
887 lis r3, 0x8000 /* Way '0' */
888
889 tlbwe r24, r3, 0
890 tlbwe r25, r3, 1
891 tlbwe r26, r3, 2
892
893 /* Update the msr to the new TS */
894 insrwi r5, r7, 1, 26
895
896 bl 1f
8971: mflr r6
898 addi r6, r6, (2f-1b)
899
900 mtspr SPRN_SRR0, r6
901 mtspr SPRN_SRR1, r5
902 rfi
903
904 /*
905 * Now we are in the tmp address space.
906 * Create a 1:1 mapping for 0-2GiB in the original TS.
907 */
9082:
909 li r3, 0
910 li r4, 0 /* TLB Word 0 */
911 li r5, 0 /* TLB Word 1 */
912 li r6, 0
913 ori r6, r6, PPC47x_TLB2_S_RWX /* TLB word 2 */
914
915 li r8, 0 /* PageIndex */
916
917 xori r7, r7, 1 /* revert back to original TS */
918
919write_utlb:
920 rotlwi r5, r8, 28 /* RPN = PageIndex * 256M */
921 /* ERPN = 0 as we don't use memory above 2G */
922
923 mr r4, r5 /* EPN = RPN */
924 ori r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
925 insrwi r4, r7, 1, 21 /* Insert the TS to Word 0 */
926
927 tlbwe r4, r3, 0 /* Write out the entries */
928 tlbwe r5, r3, 1
929 tlbwe r6, r3, 2
930 addi r8, r8, 1
931 cmpwi r8, 8 /* Have we completed ? */
932 bne write_utlb
933
934 /* make sure we complete the TLB write up */
935 isync
936
937 /*
938 * Prepare to jump to the 1:1 mapping.
939 * 1) Extract page size of the tmp mapping
940 * DSIZ = TLB_Word0[22:27]
941 * 2) Calculate the physical address of the address
942 * to jump to.
943 */
944 rlwinm r10, r24, 0, 22, 27
945
946 cmpwi r10, PPC47x_TLB0_4K
947 bne 0f
948 li r10, 0x1000 /* r10 = 4k */
949 bl 1f
950
9510:
952 /* Defaults to 256M */
953 lis r10, 0x1000
954
955 bl 1f
9561: mflr r4
957 addi r4, r4, (2f-1b) /* virtual address of 2f */
958
959 subi r11, r10, 1 /* offsetmask = Pagesize - 1 */
960 not r10, r11 /* Pagemask = ~(offsetmask) */
961
962 and r5, r25, r10 /* Physical page */
963 and r6, r4, r11 /* offset within the current page */
964
965 or r5, r5, r6 /* Physical address for 2f */
966
967 /* Switch the TS in MSR to the original one */
968 mfmsr r8
969 insrwi r8, r7, 1, 26
970
971 mtspr SPRN_SRR1, r8
972 mtspr SPRN_SRR0, r5
973 rfi
974
9752:
976 /* Invalidate the tmp mapping */
977 lis r3, 0x8000 /* Way '0' */
978
979 clrrwi r24, r24, 12 /* Clear the valid bit */
980 tlbwe r24, r3, 0
981 tlbwe r25, r3, 1
982 tlbwe r26, r3, 2
983
984 /* Make sure we complete the TLB write and flush the shadow TLB */
985 isync
986
987#endif
988
989ppc44x_map_done:
990
991
992 /* Restore the parameters */
993 mr r3, r29
994 mr r4, r30
995 mr r5, r31
996
997 li r0, 0
998#else
999 li r0, 0
1000
1001 /*
1002 * Set Machine Status Register to a known status,
1003 * switch the MMU off and jump to 1: in a single step.
1004 */
1005
1006 mr r8, r0
1007 ori r8, r8, MSR_RI|MSR_ME
1008 mtspr SPRN_SRR1, r8
1009 addi r8, r4, 1f - relocate_new_kernel
1010 mtspr SPRN_SRR0, r8
1011 sync
1012 rfi
1013
10141:
1015#endif
1016 /* from this point address translation is turned off */
1017 /* and interrupts are disabled */
1018
1019 /* set a new stack at the bottom of our page... */
1020 /* (not really needed now) */
1021 addi r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1022 stw r0, 0(r1)
1023
1024 /* Do the copies */
1025 li r6, 0 /* checksum */
1026 mr r0, r3
1027 b 1f
1028
10290: /* top, read another word for the indirection page */
1030 lwzu r0, 4(r3)
1031
10321:
1033 /* is it a destination page? (r8) */
1034 rlwinm. r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1035 beq 2f
1036
1037 rlwinm r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1038 b 0b
1039
10402: /* is it an indirection page? (r3) */
1041 rlwinm. r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1042 beq 2f
1043
1044 rlwinm r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1045 subi r3, r3, 4
1046 b 0b
1047
10482: /* are we done? */
1049 rlwinm. r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1050 beq 2f
1051 b 3f
1052
10532: /* is it a source page? (r9) */
1054 rlwinm. r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1055 beq 0b
1056
1057 rlwinm r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1058
1059 li r7, PAGE_SIZE / 4
1060 mtctr r7
1061 subi r9, r9, 4
1062 subi r8, r8, 4
10639:
1064 lwzu r0, 4(r9) /* do the copy */
1065 xor r6, r6, r0
1066 stwu r0, 4(r8)
1067 dcbst 0, r8
1068 sync
1069 icbi 0, r8
1070 bdnz 9b
1071
1072 addi r9, r9, 4
1073 addi r8, r8, 4
1074 b 0b
1075
10763:
1077
1078 /* To be certain of avoiding problems with self-modifying code
1079 * execute a serializing instruction here.
1080 */
1081 isync
1082 sync
1083
1084 mfspr r3, SPRN_PIR /* current core we are running on */
1085 mr r4, r5 /* load physical address of chunk called */
1086
1087 /* jump to the entry point, usually the setup routine */
1088 mtlr r5
1089 blrl
1090
10911: b 1b
1092
1093relocate_new_kernel_end:
1094
1095 .globl relocate_new_kernel_size
1096relocate_new_kernel_size:
1097 .long relocate_new_kernel_end - relocate_new_kernel
1098#endif
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 *
9 */
10
11#include <linux/sys.h>
12#include <asm/unistd.h>
13#include <asm/errno.h>
14#include <asm/reg.h>
15#include <asm/page.h>
16#include <asm/cache.h>
17#include <asm/cputable.h>
18#include <asm/mmu.h>
19#include <asm/ppc_asm.h>
20#include <asm/thread_info.h>
21#include <asm/asm-offsets.h>
22#include <asm/processor.h>
23#include <asm/bug.h>
24#include <asm/ptrace.h>
25#include <asm/export.h>
26#include <asm/feature-fixups.h>
27
28 .text
29
30/*
31 * We store the saved ksp_limit in the unused part
32 * of the STACK_FRAME_OVERHEAD
33 */
34_GLOBAL(call_do_softirq)
35 mflr r0
36 stw r0,4(r1)
37 lwz r10,THREAD+KSP_LIMIT(r2)
38 stw r3, THREAD+KSP_LIMIT(r2)
39 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
40 mr r1,r3
41 stw r10,8(r1)
42 bl __do_softirq
43 lwz r10,8(r1)
44 lwz r1,0(r1)
45 lwz r0,4(r1)
46 stw r10,THREAD+KSP_LIMIT(r2)
47 mtlr r0
48 blr
49
50/*
51 * void call_do_irq(struct pt_regs *regs, void *sp);
52 */
53_GLOBAL(call_do_irq)
54 mflr r0
55 stw r0,4(r1)
56 lwz r10,THREAD+KSP_LIMIT(r2)
57 stw r4, THREAD+KSP_LIMIT(r2)
58 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
59 mr r1,r4
60 stw r10,8(r1)
61 bl __do_irq
62 lwz r10,8(r1)
63 lwz r1,0(r1)
64 lwz r0,4(r1)
65 stw r10,THREAD+KSP_LIMIT(r2)
66 mtlr r0
67 blr
68
69/*
70 * This returns the high 64 bits of the product of two 64-bit numbers.
71 */
72_GLOBAL(mulhdu)
73 cmpwi r6,0
74 cmpwi cr1,r3,0
75 mr r10,r4
76 mulhwu r4,r4,r5
77 beq 1f
78 mulhwu r0,r10,r6
79 mullw r7,r10,r5
80 addc r7,r0,r7
81 addze r4,r4
821: beqlr cr1 /* all done if high part of A is 0 */
83 mullw r9,r3,r5
84 mulhwu r10,r3,r5
85 beq 2f
86 mullw r0,r3,r6
87 mulhwu r8,r3,r6
88 addc r7,r0,r7
89 adde r4,r4,r8
90 addze r10,r10
912: addc r4,r4,r9
92 addze r3,r10
93 blr
94
95/*
96 * reloc_got2 runs through the .got2 section adding an offset
97 * to each entry.
98 */
99_GLOBAL(reloc_got2)
100 mflr r11
101 lis r7,__got2_start@ha
102 addi r7,r7,__got2_start@l
103 lis r8,__got2_end@ha
104 addi r8,r8,__got2_end@l
105 subf r8,r7,r8
106 srwi. r8,r8,2
107 beqlr
108 mtctr r8
109 bl 1f
1101: mflr r0
111 lis r4,1b@ha
112 addi r4,r4,1b@l
113 subf r0,r4,r0
114 add r7,r0,r7
1152: lwz r0,0(r7)
116 add r0,r0,r3
117 stw r0,0(r7)
118 addi r7,r7,4
119 bdnz 2b
120 mtlr r11
121 blr
122
123/*
124 * call_setup_cpu - call the setup_cpu function for this cpu
125 * r3 = data offset, r24 = cpu number
126 *
127 * Setup function is called with:
128 * r3 = data offset
129 * r4 = ptr to CPU spec (relocated)
130 */
131_GLOBAL(call_setup_cpu)
132 addis r4,r3,cur_cpu_spec@ha
133 addi r4,r4,cur_cpu_spec@l
134 lwz r4,0(r4)
135 add r4,r4,r3
136 lwz r5,CPU_SPEC_SETUP(r4)
137 cmpwi 0,r5,0
138 add r5,r5,r3
139 beqlr
140 mtctr r5
141 bctr
142
143#if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_PPC_BOOK3S_32)
144
145/* This gets called by via-pmu.c to switch the PLL selection
146 * on 750fx CPU. This function should really be moved to some
147 * other place (as most of the cpufreq code in via-pmu
148 */
149_GLOBAL(low_choose_750fx_pll)
150 /* Clear MSR:EE */
151 mfmsr r7
152 rlwinm r0,r7,0,17,15
153 mtmsr r0
154
155 /* If switching to PLL1, disable HID0:BTIC */
156 cmplwi cr0,r3,0
157 beq 1f
158 mfspr r5,SPRN_HID0
159 rlwinm r5,r5,0,27,25
160 sync
161 mtspr SPRN_HID0,r5
162 isync
163 sync
164
1651:
166 /* Calc new HID1 value */
167 mfspr r4,SPRN_HID1 /* Build a HID1:PS bit from parameter */
168 rlwinm r5,r3,16,15,15 /* Clear out HID1:PS from value read */
169 rlwinm r4,r4,0,16,14 /* Could have I used rlwimi here ? */
170 or r4,r4,r5
171 mtspr SPRN_HID1,r4
172
173#ifdef CONFIG_SMP
174 /* Store new HID1 image */
175 lwz r6,TASK_CPU(r2)
176 slwi r6,r6,2
177#else
178 li r6, 0
179#endif
180 addis r6,r6,nap_save_hid1@ha
181 stw r4,nap_save_hid1@l(r6)
182
183 /* If switching to PLL0, enable HID0:BTIC */
184 cmplwi cr0,r3,0
185 bne 1f
186 mfspr r5,SPRN_HID0
187 ori r5,r5,HID0_BTIC
188 sync
189 mtspr SPRN_HID0,r5
190 isync
191 sync
192
1931:
194 /* Return */
195 mtmsr r7
196 blr
197
198_GLOBAL(low_choose_7447a_dfs)
199 /* Clear MSR:EE */
200 mfmsr r7
201 rlwinm r0,r7,0,17,15
202 mtmsr r0
203
204 /* Calc new HID1 value */
205 mfspr r4,SPRN_HID1
206 insrwi r4,r3,1,9 /* insert parameter into bit 9 */
207 sync
208 mtspr SPRN_HID1,r4
209 sync
210 isync
211
212 /* Return */
213 mtmsr r7
214 blr
215
216#endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_PPC_BOOK3S_32 */
217
218/*
219 * complement mask on the msr then "or" some values on.
220 * _nmask_and_or_msr(nmask, value_to_or)
221 */
222_GLOBAL(_nmask_and_or_msr)
223 mfmsr r0 /* Get current msr */
224 andc r0,r0,r3 /* And off the bits set in r3 (first parm) */
225 or r0,r0,r4 /* Or on the bits in r4 (second parm) */
226 SYNC /* Some chip revs have problems here... */
227 mtmsr r0 /* Update machine state */
228 isync
229 blr /* Done */
230
231#ifdef CONFIG_40x
232
233/*
234 * Do an IO access in real mode
235 */
236_GLOBAL(real_readb)
237 mfmsr r7
238 rlwinm r0,r7,0,~MSR_DR
239 sync
240 mtmsr r0
241 sync
242 isync
243 lbz r3,0(r3)
244 sync
245 mtmsr r7
246 sync
247 isync
248 blr
249_ASM_NOKPROBE_SYMBOL(real_readb)
250
251 /*
252 * Do an IO access in real mode
253 */
254_GLOBAL(real_writeb)
255 mfmsr r7
256 rlwinm r0,r7,0,~MSR_DR
257 sync
258 mtmsr r0
259 sync
260 isync
261 stb r3,0(r4)
262 sync
263 mtmsr r7
264 sync
265 isync
266 blr
267_ASM_NOKPROBE_SYMBOL(real_writeb)
268
269#endif /* CONFIG_40x */
270
271
272/*
273 * Flush instruction cache.
274 * This is a no-op on the 601.
275 */
276#ifndef CONFIG_PPC_8xx
277_GLOBAL(flush_instruction_cache)
278#if defined(CONFIG_4xx)
279 lis r3, KERNELBASE@h
280 iccci 0,r3
281#elif defined(CONFIG_FSL_BOOKE)
282#ifdef CONFIG_E200
283 mfspr r3,SPRN_L1CSR0
284 ori r3,r3,L1CSR0_CFI|L1CSR0_CLFC
285 /* msync; isync recommended here */
286 mtspr SPRN_L1CSR0,r3
287 isync
288 blr
289#endif
290 mfspr r3,SPRN_L1CSR1
291 ori r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
292 mtspr SPRN_L1CSR1,r3
293#elif defined(CONFIG_PPC_BOOK3S_601)
294 blr /* for 601, do nothing */
295#else
296 /* 603/604 processor - use invalidate-all bit in HID0 */
297 mfspr r3,SPRN_HID0
298 ori r3,r3,HID0_ICFI
299 mtspr SPRN_HID0,r3
300#endif /* CONFIG_4xx */
301 isync
302 blr
303EXPORT_SYMBOL(flush_instruction_cache)
304#endif /* CONFIG_PPC_8xx */
305
306/*
307 * Copy a whole page. We use the dcbz instruction on the destination
308 * to reduce memory traffic (it eliminates the unnecessary reads of
309 * the destination into cache). This requires that the destination
310 * is cacheable.
311 */
312#define COPY_16_BYTES \
313 lwz r6,4(r4); \
314 lwz r7,8(r4); \
315 lwz r8,12(r4); \
316 lwzu r9,16(r4); \
317 stw r6,4(r3); \
318 stw r7,8(r3); \
319 stw r8,12(r3); \
320 stwu r9,16(r3)
321
322_GLOBAL(copy_page)
323 rlwinm r5, r3, 0, L1_CACHE_BYTES - 1
324 addi r3,r3,-4
325
3260: twnei r5, 0 /* WARN if r3 is not cache aligned */
327 EMIT_BUG_ENTRY 0b,__FILE__,__LINE__, BUGFLAG_WARNING
328
329 addi r4,r4,-4
330
331 li r5,4
332
333#if MAX_COPY_PREFETCH > 1
334 li r0,MAX_COPY_PREFETCH
335 li r11,4
336 mtctr r0
33711: dcbt r11,r4
338 addi r11,r11,L1_CACHE_BYTES
339 bdnz 11b
340#else /* MAX_COPY_PREFETCH == 1 */
341 dcbt r5,r4
342 li r11,L1_CACHE_BYTES+4
343#endif /* MAX_COPY_PREFETCH */
344 li r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
345 crclr 4*cr0+eq
3462:
347 mtctr r0
3481:
349 dcbt r11,r4
350 dcbz r5,r3
351 COPY_16_BYTES
352#if L1_CACHE_BYTES >= 32
353 COPY_16_BYTES
354#if L1_CACHE_BYTES >= 64
355 COPY_16_BYTES
356 COPY_16_BYTES
357#if L1_CACHE_BYTES >= 128
358 COPY_16_BYTES
359 COPY_16_BYTES
360 COPY_16_BYTES
361 COPY_16_BYTES
362#endif
363#endif
364#endif
365 bdnz 1b
366 beqlr
367 crnot 4*cr0+eq,4*cr0+eq
368 li r0,MAX_COPY_PREFETCH
369 li r11,4
370 b 2b
371EXPORT_SYMBOL(copy_page)
372
373/*
374 * Extended precision shifts.
375 *
376 * Updated to be valid for shift counts from 0 to 63 inclusive.
377 * -- Gabriel
378 *
379 * R3/R4 has 64 bit value
380 * R5 has shift count
381 * result in R3/R4
382 *
383 * ashrdi3: arithmetic right shift (sign propagation)
384 * lshrdi3: logical right shift
385 * ashldi3: left shift
386 */
387_GLOBAL(__ashrdi3)
388 subfic r6,r5,32
389 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
390 addi r7,r5,32 # could be xori, or addi with -32
391 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
392 rlwinm r8,r7,0,32 # t3 = (count < 32) ? 32 : 0
393 sraw r7,r3,r7 # t2 = MSW >> (count-32)
394 or r4,r4,r6 # LSW |= t1
395 slw r7,r7,r8 # t2 = (count < 32) ? 0 : t2
396 sraw r3,r3,r5 # MSW = MSW >> count
397 or r4,r4,r7 # LSW |= t2
398 blr
399EXPORT_SYMBOL(__ashrdi3)
400
401_GLOBAL(__ashldi3)
402 subfic r6,r5,32
403 slw r3,r3,r5 # MSW = count > 31 ? 0 : MSW << count
404 addi r7,r5,32 # could be xori, or addi with -32
405 srw r6,r4,r6 # t1 = count > 31 ? 0 : LSW >> (32-count)
406 slw r7,r4,r7 # t2 = count < 32 ? 0 : LSW << (count-32)
407 or r3,r3,r6 # MSW |= t1
408 slw r4,r4,r5 # LSW = LSW << count
409 or r3,r3,r7 # MSW |= t2
410 blr
411EXPORT_SYMBOL(__ashldi3)
412
413_GLOBAL(__lshrdi3)
414 subfic r6,r5,32
415 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
416 addi r7,r5,32 # could be xori, or addi with -32
417 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
418 srw r7,r3,r7 # t2 = count < 32 ? 0 : MSW >> (count-32)
419 or r4,r4,r6 # LSW |= t1
420 srw r3,r3,r5 # MSW = MSW >> count
421 or r4,r4,r7 # LSW |= t2
422 blr
423EXPORT_SYMBOL(__lshrdi3)
424
425/*
426 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
427 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
428 */
429_GLOBAL(__cmpdi2)
430 cmpw r3,r5
431 li r3,1
432 bne 1f
433 cmplw r4,r6
434 beqlr
4351: li r3,0
436 bltlr
437 li r3,2
438 blr
439EXPORT_SYMBOL(__cmpdi2)
440/*
441 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
442 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
443 */
444_GLOBAL(__ucmpdi2)
445 cmplw r3,r5
446 li r3,1
447 bne 1f
448 cmplw r4,r6
449 beqlr
4501: li r3,0
451 bltlr
452 li r3,2
453 blr
454EXPORT_SYMBOL(__ucmpdi2)
455
456_GLOBAL(__bswapdi2)
457 rotlwi r9,r4,8
458 rotlwi r10,r3,8
459 rlwimi r9,r4,24,0,7
460 rlwimi r10,r3,24,0,7
461 rlwimi r9,r4,24,16,23
462 rlwimi r10,r3,24,16,23
463 mr r3,r9
464 mr r4,r10
465 blr
466EXPORT_SYMBOL(__bswapdi2)
467
468#ifdef CONFIG_SMP
469_GLOBAL(start_secondary_resume)
470 /* Reset stack */
471 rlwinm r1, r1, 0, 0, 31 - THREAD_SHIFT
472 addi r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
473 li r3,0
474 stw r3,0(r1) /* Zero the stack frame pointer */
475 bl start_secondary
476 b .
477#endif /* CONFIG_SMP */
478
479/*
480 * This routine is just here to keep GCC happy - sigh...
481 */
482_GLOBAL(__main)
483 blr