]>
Commit | Line | Data |
---|---|---|
9994a338 PM |
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 | * | |
3d1229d6 ME |
8 | * kexec bits: |
9 | * Copyright (C) 2002-2003 Eric Biederman <ebiederm@xmission.com> | |
10 | * GameCube/ppc32 port Copyright (C) 2004 Albert Herranz | |
11 | * | |
9994a338 PM |
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 | ||
9994a338 PM |
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> | |
3d1229d6 ME |
30 | #include <asm/processor.h> |
31 | #include <asm/kexec.h> | |
9994a338 PM |
32 | |
33 | .text | |
34 | ||
f2783c15 PM |
35 | /* |
36 | * This returns the high 64 bits of the product of two 64-bit numbers. | |
37 | */ | |
38 | _GLOBAL(mulhdu) | |
39 | cmpwi r6,0 | |
40 | cmpwi cr1,r3,0 | |
41 | mr r10,r4 | |
42 | mulhwu r4,r4,r5 | |
43 | beq 1f | |
44 | mulhwu r0,r10,r6 | |
45 | mullw r7,r10,r5 | |
46 | addc r7,r0,r7 | |
47 | addze r4,r4 | |
48 | 1: beqlr cr1 /* all done if high part of A is 0 */ | |
49 | mr r10,r3 | |
50 | mullw r9,r3,r5 | |
51 | mulhwu r3,r3,r5 | |
52 | beq 2f | |
53 | mullw r0,r10,r6 | |
54 | mulhwu r8,r10,r6 | |
55 | addc r7,r0,r7 | |
56 | adde r4,r4,r8 | |
57 | addze r3,r3 | |
58 | 2: addc r4,r4,r9 | |
59 | addze r3,r3 | |
60 | blr | |
61 | ||
9994a338 PM |
62 | /* |
63 | * sub_reloc_offset(x) returns x - reloc_offset(). | |
64 | */ | |
65 | _GLOBAL(sub_reloc_offset) | |
66 | mflr r0 | |
67 | bl 1f | |
68 | 1: mflr r5 | |
69 | lis r4,1b@ha | |
70 | addi r4,r4,1b@l | |
71 | subf r5,r4,r5 | |
72 | subf r3,r5,r3 | |
73 | mtlr r0 | |
74 | blr | |
75 | ||
76 | /* | |
77 | * reloc_got2 runs through the .got2 section adding an offset | |
78 | * to each entry. | |
79 | */ | |
80 | _GLOBAL(reloc_got2) | |
81 | mflr r11 | |
82 | lis r7,__got2_start@ha | |
83 | addi r7,r7,__got2_start@l | |
84 | lis r8,__got2_end@ha | |
85 | addi r8,r8,__got2_end@l | |
86 | subf r8,r7,r8 | |
87 | srwi. r8,r8,2 | |
88 | beqlr | |
89 | mtctr r8 | |
90 | bl 1f | |
91 | 1: mflr r0 | |
92 | lis r4,1b@ha | |
93 | addi r4,r4,1b@l | |
94 | subf r0,r4,r0 | |
95 | add r7,r0,r7 | |
96 | 2: lwz r0,0(r7) | |
97 | add r0,r0,r3 | |
98 | stw r0,0(r7) | |
99 | addi r7,r7,4 | |
100 | bdnz 2b | |
101 | mtlr r11 | |
102 | blr | |
103 | ||
9994a338 PM |
104 | /* |
105 | * call_setup_cpu - call the setup_cpu function for this cpu | |
106 | * r3 = data offset, r24 = cpu number | |
107 | * | |
108 | * Setup function is called with: | |
109 | * r3 = data offset | |
110 | * r4 = ptr to CPU spec (relocated) | |
111 | */ | |
112 | _GLOBAL(call_setup_cpu) | |
113 | addis r4,r3,cur_cpu_spec@ha | |
114 | addi r4,r4,cur_cpu_spec@l | |
115 | lwz r4,0(r4) | |
116 | add r4,r4,r3 | |
117 | lwz r5,CPU_SPEC_SETUP(r4) | |
b26f100d | 118 | cmpwi 0,r5,0 |
9994a338 PM |
119 | add r5,r5,r3 |
120 | beqlr | |
121 | mtctr r5 | |
122 | bctr | |
123 | ||
124 | #if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx) | |
125 | ||
126 | /* This gets called by via-pmu.c to switch the PLL selection | |
127 | * on 750fx CPU. This function should really be moved to some | |
128 | * other place (as most of the cpufreq code in via-pmu | |
129 | */ | |
130 | _GLOBAL(low_choose_750fx_pll) | |
131 | /* Clear MSR:EE */ | |
132 | mfmsr r7 | |
133 | rlwinm r0,r7,0,17,15 | |
134 | mtmsr r0 | |
135 | ||
136 | /* If switching to PLL1, disable HID0:BTIC */ | |
137 | cmplwi cr0,r3,0 | |
138 | beq 1f | |
139 | mfspr r5,SPRN_HID0 | |
140 | rlwinm r5,r5,0,27,25 | |
141 | sync | |
142 | mtspr SPRN_HID0,r5 | |
143 | isync | |
144 | sync | |
145 | ||
146 | 1: | |
147 | /* Calc new HID1 value */ | |
148 | mfspr r4,SPRN_HID1 /* Build a HID1:PS bit from parameter */ | |
149 | rlwinm r5,r3,16,15,15 /* Clear out HID1:PS from value read */ | |
150 | rlwinm r4,r4,0,16,14 /* Could have I used rlwimi here ? */ | |
151 | or r4,r4,r5 | |
152 | mtspr SPRN_HID1,r4 | |
153 | ||
154 | /* Store new HID1 image */ | |
155 | rlwinm r6,r1,0,0,18 | |
156 | lwz r6,TI_CPU(r6) | |
157 | slwi r6,r6,2 | |
158 | addis r6,r6,nap_save_hid1@ha | |
159 | stw r4,nap_save_hid1@l(r6) | |
160 | ||
161 | /* If switching to PLL0, enable HID0:BTIC */ | |
162 | cmplwi cr0,r3,0 | |
163 | bne 1f | |
164 | mfspr r5,SPRN_HID0 | |
165 | ori r5,r5,HID0_BTIC | |
166 | sync | |
167 | mtspr SPRN_HID0,r5 | |
168 | isync | |
169 | sync | |
170 | ||
171 | 1: | |
172 | /* Return */ | |
173 | mtmsr r7 | |
174 | blr | |
175 | ||
176 | _GLOBAL(low_choose_7447a_dfs) | |
177 | /* Clear MSR:EE */ | |
178 | mfmsr r7 | |
179 | rlwinm r0,r7,0,17,15 | |
180 | mtmsr r0 | |
181 | ||
182 | /* Calc new HID1 value */ | |
183 | mfspr r4,SPRN_HID1 | |
184 | insrwi r4,r3,1,9 /* insert parameter into bit 9 */ | |
185 | sync | |
186 | mtspr SPRN_HID1,r4 | |
187 | sync | |
188 | isync | |
189 | ||
190 | /* Return */ | |
191 | mtmsr r7 | |
192 | blr | |
193 | ||
194 | #endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */ | |
195 | ||
196 | /* | |
197 | * complement mask on the msr then "or" some values on. | |
198 | * _nmask_and_or_msr(nmask, value_to_or) | |
199 | */ | |
200 | _GLOBAL(_nmask_and_or_msr) | |
201 | mfmsr r0 /* Get current msr */ | |
202 | andc r0,r0,r3 /* And off the bits set in r3 (first parm) */ | |
203 | or r0,r0,r4 /* Or on the bits in r4 (second parm) */ | |
204 | SYNC /* Some chip revs have problems here... */ | |
205 | mtmsr r0 /* Update machine state */ | |
206 | isync | |
207 | blr /* Done */ | |
208 | ||
9dae8afd BH |
209 | #ifdef CONFIG_40x |
210 | ||
211 | /* | |
212 | * Do an IO access in real mode | |
213 | */ | |
214 | _GLOBAL(real_readb) | |
215 | mfmsr r7 | |
216 | ori r0,r7,MSR_DR | |
217 | xori r0,r0,MSR_DR | |
218 | sync | |
219 | mtmsr r0 | |
220 | sync | |
221 | isync | |
222 | lbz r3,0(r3) | |
223 | sync | |
224 | mtmsr r7 | |
225 | sync | |
226 | isync | |
227 | blr | |
228 | ||
229 | /* | |
230 | * Do an IO access in real mode | |
231 | */ | |
232 | _GLOBAL(real_writeb) | |
233 | mfmsr r7 | |
234 | ori r0,r7,MSR_DR | |
235 | xori r0,r0,MSR_DR | |
236 | sync | |
237 | mtmsr r0 | |
238 | sync | |
239 | isync | |
240 | stb r3,0(r4) | |
241 | sync | |
242 | mtmsr r7 | |
243 | sync | |
244 | isync | |
245 | blr | |
246 | ||
247 | #endif /* CONFIG_40x */ | |
9994a338 PM |
248 | |
249 | /* | |
250 | * Flush MMU TLB | |
251 | */ | |
252 | _GLOBAL(_tlbia) | |
253 | #if defined(CONFIG_40x) | |
254 | sync /* Flush to memory before changing mapping */ | |
255 | tlbia | |
256 | isync /* Flush shadow TLB */ | |
257 | #elif defined(CONFIG_44x) | |
258 | li r3,0 | |
259 | sync | |
260 | ||
261 | /* Load high watermark */ | |
262 | lis r4,tlb_44x_hwater@ha | |
263 | lwz r5,tlb_44x_hwater@l(r4) | |
264 | ||
265 | 1: tlbwe r3,r3,PPC44x_TLB_PAGEID | |
266 | addi r3,r3,1 | |
267 | cmpw 0,r3,r5 | |
268 | ble 1b | |
269 | ||
270 | isync | |
271 | #elif defined(CONFIG_FSL_BOOKE) | |
272 | /* Invalidate all entries in TLB0 */ | |
273 | li r3, 0x04 | |
274 | tlbivax 0,3 | |
275 | /* Invalidate all entries in TLB1 */ | |
276 | li r3, 0x0c | |
277 | tlbivax 0,3 | |
278 | /* Invalidate all entries in TLB2 */ | |
279 | li r3, 0x14 | |
280 | tlbivax 0,3 | |
281 | /* Invalidate all entries in TLB3 */ | |
282 | li r3, 0x1c | |
283 | tlbivax 0,3 | |
284 | msync | |
285 | #ifdef CONFIG_SMP | |
286 | tlbsync | |
287 | #endif /* CONFIG_SMP */ | |
288 | #else /* !(CONFIG_40x || CONFIG_44x || CONFIG_FSL_BOOKE) */ | |
289 | #if defined(CONFIG_SMP) | |
290 | rlwinm r8,r1,0,0,18 | |
291 | lwz r8,TI_CPU(r8) | |
292 | oris r8,r8,10 | |
293 | mfmsr r10 | |
294 | SYNC | |
295 | rlwinm r0,r10,0,17,15 /* clear bit 16 (MSR_EE) */ | |
296 | rlwinm r0,r0,0,28,26 /* clear DR */ | |
297 | mtmsr r0 | |
298 | SYNC_601 | |
299 | isync | |
300 | lis r9,mmu_hash_lock@h | |
301 | ori r9,r9,mmu_hash_lock@l | |
302 | tophys(r9,r9) | |
303 | 10: lwarx r7,0,r9 | |
304 | cmpwi 0,r7,0 | |
305 | bne- 10b | |
306 | stwcx. r8,0,r9 | |
307 | bne- 10b | |
308 | sync | |
309 | tlbia | |
310 | sync | |
311 | TLBSYNC | |
312 | li r0,0 | |
313 | stw r0,0(r9) /* clear mmu_hash_lock */ | |
314 | mtmsr r10 | |
315 | SYNC_601 | |
316 | isync | |
317 | #else /* CONFIG_SMP */ | |
318 | sync | |
319 | tlbia | |
320 | sync | |
321 | #endif /* CONFIG_SMP */ | |
322 | #endif /* ! defined(CONFIG_40x) */ | |
323 | blr | |
324 | ||
325 | /* | |
326 | * Flush MMU TLB for a particular address | |
327 | */ | |
328 | _GLOBAL(_tlbie) | |
329 | #if defined(CONFIG_40x) | |
e701d269 BH |
330 | /* We run the search with interrupts disabled because we have to change |
331 | * the PID and I don't want to preempt when that happens. | |
332 | */ | |
333 | mfmsr r5 | |
334 | mfspr r6,SPRN_PID | |
335 | wrteei 0 | |
336 | mtspr SPRN_PID,r4 | |
9994a338 | 337 | tlbsx. r3, 0, r3 |
e701d269 BH |
338 | mtspr SPRN_PID,r6 |
339 | wrtee r5 | |
9994a338 PM |
340 | bne 10f |
341 | sync | |
342 | /* There are only 64 TLB entries, so r3 < 64, which means bit 25 is clear. | |
343 | * Since 25 is the V bit in the TLB_TAG, loading this value will invalidate | |
344 | * the TLB entry. */ | |
345 | tlbwe r3, r3, TLB_TAG | |
346 | isync | |
347 | 10: | |
e701d269 | 348 | |
9994a338 | 349 | #elif defined(CONFIG_44x) |
e701d269 BH |
350 | mfspr r5,SPRN_MMUCR |
351 | rlwimi r5,r4,0,24,31 /* Set TID */ | |
9994a338 | 352 | |
aa1cf632 DG |
353 | /* We have to run the search with interrupts disabled, even critical |
354 | * and debug interrupts (in fact the only critical exceptions we have | |
355 | * are debug and machine check). Otherwise an interrupt which causes | |
356 | * a TLB miss can clobber the MMUCR between the mtspr and the tlbsx. */ | |
e701d269 | 357 | mfmsr r4 |
aa1cf632 DG |
358 | lis r6,(MSR_EE|MSR_CE|MSR_ME|MSR_DE)@ha |
359 | addi r6,r6,(MSR_EE|MSR_CE|MSR_ME|MSR_DE)@l | |
e701d269 | 360 | andc r6,r4,r6 |
aa1cf632 | 361 | mtmsr r6 |
e701d269 | 362 | mtspr SPRN_MMUCR,r5 |
9994a338 | 363 | tlbsx. r3, 0, r3 |
e701d269 | 364 | mtmsr r4 |
9994a338 PM |
365 | bne 10f |
366 | sync | |
367 | /* There are only 64 TLB entries, so r3 < 64, | |
368 | * which means bit 22, is clear. Since 22 is | |
369 | * the V bit in the TLB_PAGEID, loading this | |
370 | * value will invalidate the TLB entry. | |
371 | */ | |
372 | tlbwe r3, r3, PPC44x_TLB_PAGEID | |
373 | isync | |
374 | 10: | |
375 | #elif defined(CONFIG_FSL_BOOKE) | |
376 | rlwinm r4, r3, 0, 0, 19 | |
377 | ori r5, r4, 0x08 /* TLBSEL = 1 */ | |
378 | ori r6, r4, 0x10 /* TLBSEL = 2 */ | |
379 | ori r7, r4, 0x18 /* TLBSEL = 3 */ | |
380 | tlbivax 0, r4 | |
381 | tlbivax 0, r5 | |
382 | tlbivax 0, r6 | |
383 | tlbivax 0, r7 | |
384 | msync | |
385 | #if defined(CONFIG_SMP) | |
386 | tlbsync | |
387 | #endif /* CONFIG_SMP */ | |
388 | #else /* !(CONFIG_40x || CONFIG_44x || CONFIG_FSL_BOOKE) */ | |
389 | #if defined(CONFIG_SMP) | |
390 | rlwinm r8,r1,0,0,18 | |
391 | lwz r8,TI_CPU(r8) | |
392 | oris r8,r8,11 | |
393 | mfmsr r10 | |
394 | SYNC | |
395 | rlwinm r0,r10,0,17,15 /* clear bit 16 (MSR_EE) */ | |
396 | rlwinm r0,r0,0,28,26 /* clear DR */ | |
397 | mtmsr r0 | |
398 | SYNC_601 | |
399 | isync | |
400 | lis r9,mmu_hash_lock@h | |
401 | ori r9,r9,mmu_hash_lock@l | |
402 | tophys(r9,r9) | |
403 | 10: lwarx r7,0,r9 | |
404 | cmpwi 0,r7,0 | |
405 | bne- 10b | |
406 | stwcx. r8,0,r9 | |
407 | bne- 10b | |
408 | eieio | |
409 | tlbie r3 | |
410 | sync | |
411 | TLBSYNC | |
412 | li r0,0 | |
413 | stw r0,0(r9) /* clear mmu_hash_lock */ | |
414 | mtmsr r10 | |
415 | SYNC_601 | |
416 | isync | |
417 | #else /* CONFIG_SMP */ | |
418 | tlbie r3 | |
419 | sync | |
420 | #endif /* CONFIG_SMP */ | |
421 | #endif /* ! CONFIG_40x */ | |
422 | blr | |
423 | ||
424 | /* | |
425 | * Flush instruction cache. | |
426 | * This is a no-op on the 601. | |
427 | */ | |
428 | _GLOBAL(flush_instruction_cache) | |
429 | #if defined(CONFIG_8xx) | |
430 | isync | |
431 | lis r5, IDC_INVALL@h | |
432 | mtspr SPRN_IC_CST, r5 | |
433 | #elif defined(CONFIG_4xx) | |
434 | #ifdef CONFIG_403GCX | |
435 | li r3, 512 | |
436 | mtctr r3 | |
437 | lis r4, KERNELBASE@h | |
438 | 1: iccci 0, r4 | |
439 | addi r4, r4, 16 | |
440 | bdnz 1b | |
441 | #else | |
442 | lis r3, KERNELBASE@h | |
443 | iccci 0,r3 | |
444 | #endif | |
445 | #elif CONFIG_FSL_BOOKE | |
446 | BEGIN_FTR_SECTION | |
447 | mfspr r3,SPRN_L1CSR0 | |
448 | ori r3,r3,L1CSR0_CFI|L1CSR0_CLFC | |
449 | /* msync; isync recommended here */ | |
450 | mtspr SPRN_L1CSR0,r3 | |
451 | isync | |
452 | blr | |
4508dc21 | 453 | END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE) |
9994a338 PM |
454 | mfspr r3,SPRN_L1CSR1 |
455 | ori r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR | |
456 | mtspr SPRN_L1CSR1,r3 | |
457 | #else | |
458 | mfspr r3,SPRN_PVR | |
459 | rlwinm r3,r3,16,16,31 | |
460 | cmpwi 0,r3,1 | |
461 | beqlr /* for 601, do nothing */ | |
462 | /* 603/604 processor - use invalidate-all bit in HID0 */ | |
463 | mfspr r3,SPRN_HID0 | |
464 | ori r3,r3,HID0_ICFI | |
465 | mtspr SPRN_HID0,r3 | |
466 | #endif /* CONFIG_8xx/4xx */ | |
467 | isync | |
468 | blr | |
469 | ||
470 | /* | |
471 | * Write any modified data cache blocks out to memory | |
472 | * and invalidate the corresponding instruction cache blocks. | |
473 | * This is a no-op on the 601. | |
474 | * | |
475 | * flush_icache_range(unsigned long start, unsigned long stop) | |
476 | */ | |
26ef5c09 | 477 | _GLOBAL(__flush_icache_range) |
9994a338 PM |
478 | BEGIN_FTR_SECTION |
479 | blr /* for 601, do nothing */ | |
4508dc21 | 480 | END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE) |
7dffb720 | 481 | li r5,L1_CACHE_BYTES-1 |
9994a338 PM |
482 | andc r3,r3,r5 |
483 | subf r4,r3,r4 | |
484 | add r4,r4,r5 | |
7dffb720 | 485 | srwi. r4,r4,L1_CACHE_SHIFT |
9994a338 PM |
486 | beqlr |
487 | mtctr r4 | |
488 | mr r6,r3 | |
489 | 1: dcbst 0,r3 | |
7dffb720 | 490 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
491 | bdnz 1b |
492 | sync /* wait for dcbst's to get to ram */ | |
493 | mtctr r4 | |
494 | 2: icbi 0,r6 | |
7dffb720 | 495 | addi r6,r6,L1_CACHE_BYTES |
9994a338 PM |
496 | bdnz 2b |
497 | sync /* additional sync needed on g4 */ | |
498 | isync | |
499 | blr | |
500 | /* | |
501 | * Write any modified data cache blocks out to memory. | |
502 | * Does not invalidate the corresponding cache lines (especially for | |
503 | * any corresponding instruction cache). | |
504 | * | |
505 | * clean_dcache_range(unsigned long start, unsigned long stop) | |
506 | */ | |
507 | _GLOBAL(clean_dcache_range) | |
7dffb720 | 508 | li r5,L1_CACHE_BYTES-1 |
9994a338 PM |
509 | andc r3,r3,r5 |
510 | subf r4,r3,r4 | |
511 | add r4,r4,r5 | |
7dffb720 | 512 | srwi. r4,r4,L1_CACHE_SHIFT |
9994a338 PM |
513 | beqlr |
514 | mtctr r4 | |
515 | ||
516 | 1: dcbst 0,r3 | |
7dffb720 | 517 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
518 | bdnz 1b |
519 | sync /* wait for dcbst's to get to ram */ | |
520 | blr | |
521 | ||
522 | /* | |
523 | * Write any modified data cache blocks out to memory and invalidate them. | |
524 | * Does not invalidate the corresponding instruction cache blocks. | |
525 | * | |
526 | * flush_dcache_range(unsigned long start, unsigned long stop) | |
527 | */ | |
528 | _GLOBAL(flush_dcache_range) | |
7dffb720 | 529 | li r5,L1_CACHE_BYTES-1 |
9994a338 PM |
530 | andc r3,r3,r5 |
531 | subf r4,r3,r4 | |
532 | add r4,r4,r5 | |
7dffb720 | 533 | srwi. r4,r4,L1_CACHE_SHIFT |
9994a338 PM |
534 | beqlr |
535 | mtctr r4 | |
536 | ||
537 | 1: dcbf 0,r3 | |
7dffb720 | 538 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
539 | bdnz 1b |
540 | sync /* wait for dcbst's to get to ram */ | |
541 | blr | |
542 | ||
543 | /* | |
544 | * Like above, but invalidate the D-cache. This is used by the 8xx | |
545 | * to invalidate the cache so the PPC core doesn't get stale data | |
546 | * from the CPM (no cache snooping here :-). | |
547 | * | |
548 | * invalidate_dcache_range(unsigned long start, unsigned long stop) | |
549 | */ | |
550 | _GLOBAL(invalidate_dcache_range) | |
7dffb720 | 551 | li r5,L1_CACHE_BYTES-1 |
9994a338 PM |
552 | andc r3,r3,r5 |
553 | subf r4,r3,r4 | |
554 | add r4,r4,r5 | |
7dffb720 | 555 | srwi. r4,r4,L1_CACHE_SHIFT |
9994a338 PM |
556 | beqlr |
557 | mtctr r4 | |
558 | ||
559 | 1: dcbi 0,r3 | |
7dffb720 | 560 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
561 | bdnz 1b |
562 | sync /* wait for dcbi's to get to ram */ | |
563 | blr | |
564 | ||
9994a338 PM |
565 | /* |
566 | * Flush a particular page from the data cache to RAM. | |
567 | * Note: this is necessary because the instruction cache does *not* | |
568 | * snoop from the data cache. | |
569 | * This is a no-op on the 601 which has a unified cache. | |
570 | * | |
571 | * void __flush_dcache_icache(void *page) | |
572 | */ | |
573 | _GLOBAL(__flush_dcache_icache) | |
574 | BEGIN_FTR_SECTION | |
4508dc21 DG |
575 | blr |
576 | END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE) | |
9994a338 | 577 | rlwinm r3,r3,0,0,19 /* Get page base address */ |
7dffb720 | 578 | li r4,4096/L1_CACHE_BYTES /* Number of lines in a page */ |
9994a338 PM |
579 | mtctr r4 |
580 | mr r6,r3 | |
581 | 0: dcbst 0,r3 /* Write line to ram */ | |
7dffb720 | 582 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
583 | bdnz 0b |
584 | sync | |
b98ac05d BH |
585 | #ifndef CONFIG_44x |
586 | /* We don't flush the icache on 44x. Those have a virtual icache | |
587 | * and we don't have access to the virtual address here (it's | |
588 | * not the page vaddr but where it's mapped in user space). The | |
589 | * flushing of the icache on these is handled elsewhere, when | |
590 | * a change in the address space occurs, before returning to | |
591 | * user space | |
592 | */ | |
9994a338 PM |
593 | mtctr r4 |
594 | 1: icbi 0,r6 | |
7dffb720 | 595 | addi r6,r6,L1_CACHE_BYTES |
9994a338 PM |
596 | bdnz 1b |
597 | sync | |
598 | isync | |
b98ac05d | 599 | #endif /* CONFIG_44x */ |
9994a338 PM |
600 | blr |
601 | ||
602 | /* | |
603 | * Flush a particular page from the data cache to RAM, identified | |
604 | * by its physical address. We turn off the MMU so we can just use | |
605 | * the physical address (this may be a highmem page without a kernel | |
606 | * mapping). | |
607 | * | |
608 | * void __flush_dcache_icache_phys(unsigned long physaddr) | |
609 | */ | |
610 | _GLOBAL(__flush_dcache_icache_phys) | |
611 | BEGIN_FTR_SECTION | |
612 | blr /* for 601, do nothing */ | |
4508dc21 | 613 | END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE) |
9994a338 PM |
614 | mfmsr r10 |
615 | rlwinm r0,r10,0,28,26 /* clear DR */ | |
616 | mtmsr r0 | |
617 | isync | |
618 | rlwinm r3,r3,0,0,19 /* Get page base address */ | |
7dffb720 | 619 | li r4,4096/L1_CACHE_BYTES /* Number of lines in a page */ |
9994a338 PM |
620 | mtctr r4 |
621 | mr r6,r3 | |
622 | 0: dcbst 0,r3 /* Write line to ram */ | |
7dffb720 | 623 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
624 | bdnz 0b |
625 | sync | |
626 | mtctr r4 | |
627 | 1: icbi 0,r6 | |
7dffb720 | 628 | addi r6,r6,L1_CACHE_BYTES |
9994a338 PM |
629 | bdnz 1b |
630 | sync | |
631 | mtmsr r10 /* restore DR */ | |
632 | isync | |
633 | blr | |
634 | ||
635 | /* | |
636 | * Clear pages using the dcbz instruction, which doesn't cause any | |
637 | * memory traffic (except to write out any cache lines which get | |
638 | * displaced). This only works on cacheable memory. | |
639 | * | |
640 | * void clear_pages(void *page, int order) ; | |
641 | */ | |
642 | _GLOBAL(clear_pages) | |
7dffb720 | 643 | li r0,4096/L1_CACHE_BYTES |
9994a338 PM |
644 | slw r0,r0,r4 |
645 | mtctr r0 | |
646 | #ifdef CONFIG_8xx | |
647 | li r4, 0 | |
648 | 1: stw r4, 0(r3) | |
649 | stw r4, 4(r3) | |
650 | stw r4, 8(r3) | |
651 | stw r4, 12(r3) | |
652 | #else | |
653 | 1: dcbz 0,r3 | |
654 | #endif | |
7dffb720 | 655 | addi r3,r3,L1_CACHE_BYTES |
9994a338 PM |
656 | bdnz 1b |
657 | blr | |
658 | ||
659 | /* | |
660 | * Copy a whole page. We use the dcbz instruction on the destination | |
661 | * to reduce memory traffic (it eliminates the unnecessary reads of | |
662 | * the destination into cache). This requires that the destination | |
663 | * is cacheable. | |
664 | */ | |
665 | #define COPY_16_BYTES \ | |
666 | lwz r6,4(r4); \ | |
667 | lwz r7,8(r4); \ | |
668 | lwz r8,12(r4); \ | |
669 | lwzu r9,16(r4); \ | |
670 | stw r6,4(r3); \ | |
671 | stw r7,8(r3); \ | |
672 | stw r8,12(r3); \ | |
673 | stwu r9,16(r3) | |
674 | ||
675 | _GLOBAL(copy_page) | |
676 | addi r3,r3,-4 | |
677 | addi r4,r4,-4 | |
678 | ||
679 | #ifdef CONFIG_8xx | |
680 | /* don't use prefetch on 8xx */ | |
7dffb720 | 681 | li r0,4096/L1_CACHE_BYTES |
9994a338 PM |
682 | mtctr r0 |
683 | 1: COPY_16_BYTES | |
684 | bdnz 1b | |
685 | blr | |
686 | ||
687 | #else /* not 8xx, we can prefetch */ | |
688 | li r5,4 | |
689 | ||
690 | #if MAX_COPY_PREFETCH > 1 | |
691 | li r0,MAX_COPY_PREFETCH | |
692 | li r11,4 | |
693 | mtctr r0 | |
694 | 11: dcbt r11,r4 | |
7dffb720 | 695 | addi r11,r11,L1_CACHE_BYTES |
9994a338 PM |
696 | bdnz 11b |
697 | #else /* MAX_COPY_PREFETCH == 1 */ | |
698 | dcbt r5,r4 | |
7dffb720 | 699 | li r11,L1_CACHE_BYTES+4 |
9994a338 | 700 | #endif /* MAX_COPY_PREFETCH */ |
7dffb720 | 701 | li r0,4096/L1_CACHE_BYTES - MAX_COPY_PREFETCH |
9994a338 PM |
702 | crclr 4*cr0+eq |
703 | 2: | |
704 | mtctr r0 | |
705 | 1: | |
706 | dcbt r11,r4 | |
707 | dcbz r5,r3 | |
708 | COPY_16_BYTES | |
7dffb720 | 709 | #if L1_CACHE_BYTES >= 32 |
9994a338 | 710 | COPY_16_BYTES |
7dffb720 | 711 | #if L1_CACHE_BYTES >= 64 |
9994a338 PM |
712 | COPY_16_BYTES |
713 | COPY_16_BYTES | |
7dffb720 | 714 | #if L1_CACHE_BYTES >= 128 |
9994a338 PM |
715 | COPY_16_BYTES |
716 | COPY_16_BYTES | |
717 | COPY_16_BYTES | |
718 | COPY_16_BYTES | |
719 | #endif | |
720 | #endif | |
721 | #endif | |
722 | bdnz 1b | |
723 | beqlr | |
724 | crnot 4*cr0+eq,4*cr0+eq | |
725 | li r0,MAX_COPY_PREFETCH | |
726 | li r11,4 | |
727 | b 2b | |
728 | #endif /* CONFIG_8xx */ | |
729 | ||
730 | /* | |
731 | * void atomic_clear_mask(atomic_t mask, atomic_t *addr) | |
732 | * void atomic_set_mask(atomic_t mask, atomic_t *addr); | |
733 | */ | |
734 | _GLOBAL(atomic_clear_mask) | |
735 | 10: lwarx r5,0,r4 | |
736 | andc r5,r5,r3 | |
737 | PPC405_ERR77(0,r4) | |
738 | stwcx. r5,0,r4 | |
739 | bne- 10b | |
740 | blr | |
741 | _GLOBAL(atomic_set_mask) | |
742 | 10: lwarx r5,0,r4 | |
743 | or r5,r5,r3 | |
744 | PPC405_ERR77(0,r4) | |
745 | stwcx. r5,0,r4 | |
746 | bne- 10b | |
747 | blr | |
748 | ||
9994a338 PM |
749 | /* |
750 | * Extended precision shifts. | |
751 | * | |
752 | * Updated to be valid for shift counts from 0 to 63 inclusive. | |
753 | * -- Gabriel | |
754 | * | |
755 | * R3/R4 has 64 bit value | |
756 | * R5 has shift count | |
757 | * result in R3/R4 | |
758 | * | |
759 | * ashrdi3: arithmetic right shift (sign propagation) | |
760 | * lshrdi3: logical right shift | |
761 | * ashldi3: left shift | |
762 | */ | |
763 | _GLOBAL(__ashrdi3) | |
764 | subfic r6,r5,32 | |
765 | srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count | |
766 | addi r7,r5,32 # could be xori, or addi with -32 | |
767 | slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count) | |
768 | rlwinm r8,r7,0,32 # t3 = (count < 32) ? 32 : 0 | |
769 | sraw r7,r3,r7 # t2 = MSW >> (count-32) | |
770 | or r4,r4,r6 # LSW |= t1 | |
771 | slw r7,r7,r8 # t2 = (count < 32) ? 0 : t2 | |
772 | sraw r3,r3,r5 # MSW = MSW >> count | |
773 | or r4,r4,r7 # LSW |= t2 | |
774 | blr | |
775 | ||
776 | _GLOBAL(__ashldi3) | |
777 | subfic r6,r5,32 | |
778 | slw r3,r3,r5 # MSW = count > 31 ? 0 : MSW << count | |
779 | addi r7,r5,32 # could be xori, or addi with -32 | |
780 | srw r6,r4,r6 # t1 = count > 31 ? 0 : LSW >> (32-count) | |
781 | slw r7,r4,r7 # t2 = count < 32 ? 0 : LSW << (count-32) | |
782 | or r3,r3,r6 # MSW |= t1 | |
783 | slw r4,r4,r5 # LSW = LSW << count | |
784 | or r3,r3,r7 # MSW |= t2 | |
785 | blr | |
786 | ||
787 | _GLOBAL(__lshrdi3) | |
788 | subfic r6,r5,32 | |
789 | srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count | |
790 | addi r7,r5,32 # could be xori, or addi with -32 | |
791 | slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count) | |
792 | srw r7,r3,r7 # t2 = count < 32 ? 0 : MSW >> (count-32) | |
793 | or r4,r4,r6 # LSW |= t1 | |
794 | srw r3,r3,r5 # MSW = MSW >> count | |
795 | or r4,r4,r7 # LSW |= t2 | |
796 | blr | |
797 | ||
798 | _GLOBAL(abs) | |
799 | srawi r4,r3,31 | |
800 | xor r3,r3,r4 | |
801 | sub r3,r3,r4 | |
802 | blr | |
803 | ||
9994a338 PM |
804 | /* |
805 | * Create a kernel thread | |
806 | * kernel_thread(fn, arg, flags) | |
807 | */ | |
808 | _GLOBAL(kernel_thread) | |
809 | stwu r1,-16(r1) | |
810 | stw r30,8(r1) | |
811 | stw r31,12(r1) | |
812 | mr r30,r3 /* function */ | |
813 | mr r31,r4 /* argument */ | |
814 | ori r3,r5,CLONE_VM /* flags */ | |
815 | oris r3,r3,CLONE_UNTRACED>>16 | |
816 | li r4,0 /* new sp (unused) */ | |
817 | li r0,__NR_clone | |
818 | sc | |
819 | cmpwi 0,r3,0 /* parent or child? */ | |
820 | bne 1f /* return if parent */ | |
821 | li r0,0 /* make top-level stack frame */ | |
822 | stwu r0,-16(r1) | |
823 | mtlr r30 /* fn addr in lr */ | |
824 | mr r3,r31 /* load arg and call fn */ | |
825 | PPC440EP_ERR42 | |
826 | blrl | |
827 | li r0,__NR_exit /* exit if function returns */ | |
828 | li r3,0 | |
829 | sc | |
830 | 1: lwz r30,8(r1) | |
831 | lwz r31,12(r1) | |
832 | addi r1,r1,16 | |
833 | blr | |
834 | ||
9994a338 PM |
835 | /* |
836 | * This routine is just here to keep GCC happy - sigh... | |
837 | */ | |
838 | _GLOBAL(__main) | |
839 | blr | |
3d1229d6 ME |
840 | |
841 | #ifdef CONFIG_KEXEC | |
842 | /* | |
843 | * Must be relocatable PIC code callable as a C function. | |
844 | */ | |
845 | .globl relocate_new_kernel | |
846 | relocate_new_kernel: | |
847 | /* r3 = page_list */ | |
848 | /* r4 = reboot_code_buffer */ | |
849 | /* r5 = start_address */ | |
850 | ||
851 | li r0, 0 | |
852 | ||
853 | /* | |
854 | * Set Machine Status Register to a known status, | |
855 | * switch the MMU off and jump to 1: in a single step. | |
856 | */ | |
857 | ||
858 | mr r8, r0 | |
859 | ori r8, r8, MSR_RI|MSR_ME | |
860 | mtspr SPRN_SRR1, r8 | |
861 | addi r8, r4, 1f - relocate_new_kernel | |
862 | mtspr SPRN_SRR0, r8 | |
863 | sync | |
864 | rfi | |
865 | ||
866 | 1: | |
867 | /* from this point address translation is turned off */ | |
868 | /* and interrupts are disabled */ | |
869 | ||
870 | /* set a new stack at the bottom of our page... */ | |
871 | /* (not really needed now) */ | |
872 | addi r1, r4, KEXEC_CONTROL_CODE_SIZE - 8 /* for LR Save+Back Chain */ | |
873 | stw r0, 0(r1) | |
874 | ||
875 | /* Do the copies */ | |
876 | li r6, 0 /* checksum */ | |
877 | mr r0, r3 | |
878 | b 1f | |
879 | ||
880 | 0: /* top, read another word for the indirection page */ | |
881 | lwzu r0, 4(r3) | |
882 | ||
883 | 1: | |
884 | /* is it a destination page? (r8) */ | |
885 | rlwinm. r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */ | |
886 | beq 2f | |
887 | ||
888 | rlwinm r8, r0, 0, 0, 19 /* clear kexec flags, page align */ | |
889 | b 0b | |
890 | ||
891 | 2: /* is it an indirection page? (r3) */ | |
892 | rlwinm. r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */ | |
893 | beq 2f | |
894 | ||
895 | rlwinm r3, r0, 0, 0, 19 /* clear kexec flags, page align */ | |
896 | subi r3, r3, 4 | |
897 | b 0b | |
898 | ||
899 | 2: /* are we done? */ | |
900 | rlwinm. r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */ | |
901 | beq 2f | |
902 | b 3f | |
903 | ||
904 | 2: /* is it a source page? (r9) */ | |
905 | rlwinm. r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */ | |
906 | beq 0b | |
907 | ||
908 | rlwinm r9, r0, 0, 0, 19 /* clear kexec flags, page align */ | |
909 | ||
910 | li r7, PAGE_SIZE / 4 | |
911 | mtctr r7 | |
912 | subi r9, r9, 4 | |
913 | subi r8, r8, 4 | |
914 | 9: | |
915 | lwzu r0, 4(r9) /* do the copy */ | |
916 | xor r6, r6, r0 | |
917 | stwu r0, 4(r8) | |
918 | dcbst 0, r8 | |
919 | sync | |
920 | icbi 0, r8 | |
921 | bdnz 9b | |
922 | ||
923 | addi r9, r9, 4 | |
924 | addi r8, r8, 4 | |
925 | b 0b | |
926 | ||
927 | 3: | |
928 | ||
929 | /* To be certain of avoiding problems with self-modifying code | |
930 | * execute a serializing instruction here. | |
931 | */ | |
932 | isync | |
933 | sync | |
934 | ||
935 | /* jump to the entry point, usually the setup routine */ | |
936 | mtlr r5 | |
937 | blrl | |
938 | ||
939 | 1: b 1b | |
940 | ||
941 | relocate_new_kernel_end: | |
942 | ||
943 | .globl relocate_new_kernel_size | |
944 | relocate_new_kernel_size: | |
945 | .long relocate_new_kernel_end - relocate_new_kernel | |
946 | #endif |