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 * Adapted for iSeries by Mike Corrigan (mikejc@us.ibm.com) 8 * PPC64 updates by Dave Engebretsen (engebret@us.ibm.com) 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public License 12 * as published by the Free Software Foundation; either version 13 * 2 of the License, or (at your option) any later version. 14 * 15 */ 16 17#include <linux/sys.h> 18#include <asm/unistd.h> 19#include <asm/errno.h> 20#include <asm/processor.h> 21#include <asm/page.h> 22#include <asm/cache.h> 23#include <asm/ppc_asm.h> 24#include <asm/asm-offsets.h> 25#include <asm/cputable.h> 26#include <asm/thread_info.h> 27#include <asm/kexec.h> 28#include <asm/ptrace.h> 29 30 .text 31 32_GLOBAL(call_do_softirq) 33 mflr r0 34 std r0,16(r1) 35 stdu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3) 36 mr r1,r3 37 bl .__do_softirq 38 ld r1,0(r1) 39 ld r0,16(r1) 40 mtlr r0 41 blr 42 43_GLOBAL(call_handle_irq) 44 ld r8,0(r6) 45 mflr r0 46 std r0,16(r1) 47 mtctr r8 48 stdu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r5) 49 mr r1,r5 50 bctrl 51 ld r1,0(r1) 52 ld r0,16(r1) 53 mtlr r0 54 blr 55 56 .section ".toc","aw" 57PPC64_CACHES: 58 .tc ppc64_caches[TC],ppc64_caches 59 .section ".text" 60 61/* 62 * Write any modified data cache blocks out to memory 63 * and invalidate the corresponding instruction cache blocks. 64 * 65 * flush_icache_range(unsigned long start, unsigned long stop) 66 * 67 * flush all bytes from start through stop-1 inclusive 68 */ 69 70_KPROBE(__flush_icache_range) 71 72/* 73 * Flush the data cache to memory 74 * 75 * Different systems have different cache line sizes 76 * and in some cases i-cache and d-cache line sizes differ from 77 * each other. 78 */ 79 ld r10,PPC64_CACHES@toc(r2) 80 lwz r7,DCACHEL1LINESIZE(r10)/* Get cache line size */ 81 addi r5,r7,-1 82 andc r6,r3,r5 /* round low to line bdy */ 83 subf r8,r6,r4 /* compute length */ 84 add r8,r8,r5 /* ensure we get enough */ 85 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of cache line size */ 86 srw. r8,r8,r9 /* compute line count */ 87 beqlr /* nothing to do? */ 88 mtctr r8 891: dcbst 0,r6 90 add r6,r6,r7 91 bdnz 1b 92 sync 93 94/* Now invalidate the instruction cache */ 95 96 lwz r7,ICACHEL1LINESIZE(r10) /* Get Icache line size */ 97 addi r5,r7,-1 98 andc r6,r3,r5 /* round low to line bdy */ 99 subf r8,r6,r4 /* compute length */ 100 add r8,r8,r5 101 lwz r9,ICACHEL1LOGLINESIZE(r10) /* Get log-2 of Icache line size */ 102 srw. r8,r8,r9 /* compute line count */ 103 beqlr /* nothing to do? */ 104 mtctr r8 1052: icbi 0,r6 106 add r6,r6,r7 107 bdnz 2b 108 isync 109 blr 110 .previous .text 111/* 112 * Like above, but only do the D-cache. 113 * 114 * flush_dcache_range(unsigned long start, unsigned long stop) 115 * 116 * flush all bytes from start to stop-1 inclusive 117 */ 118_GLOBAL(flush_dcache_range) 119 120/* 121 * Flush the data cache to memory 122 * 123 * Different systems have different cache line sizes 124 */ 125 ld r10,PPC64_CACHES@toc(r2) 126 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */ 127 addi r5,r7,-1 128 andc r6,r3,r5 /* round low to line bdy */ 129 subf r8,r6,r4 /* compute length */ 130 add r8,r8,r5 /* ensure we get enough */ 131 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */ 132 srw. r8,r8,r9 /* compute line count */ 133 beqlr /* nothing to do? */ 134 mtctr r8 1350: dcbst 0,r6 136 add r6,r6,r7 137 bdnz 0b 138 sync 139 blr 140 141/* 142 * Like above, but works on non-mapped physical addresses. 143 * Use only for non-LPAR setups ! It also assumes real mode 144 * is cacheable. Used for flushing out the DART before using 145 * it as uncacheable memory 146 * 147 * flush_dcache_phys_range(unsigned long start, unsigned long stop) 148 * 149 * flush all bytes from start to stop-1 inclusive 150 */ 151_GLOBAL(flush_dcache_phys_range) 152 ld r10,PPC64_CACHES@toc(r2) 153 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */ 154 addi r5,r7,-1 155 andc r6,r3,r5 /* round low to line bdy */ 156 subf r8,r6,r4 /* compute length */ 157 add r8,r8,r5 /* ensure we get enough */ 158 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */ 159 srw. r8,r8,r9 /* compute line count */ 160 beqlr /* nothing to do? */ 161 mfmsr r5 /* Disable MMU Data Relocation */ 162 ori r0,r5,MSR_DR 163 xori r0,r0,MSR_DR 164 sync 165 mtmsr r0 166 sync 167 isync 168 mtctr r8 1690: dcbst 0,r6 170 add r6,r6,r7 171 bdnz 0b 172 sync 173 isync 174 mtmsr r5 /* Re-enable MMU Data Relocation */ 175 sync 176 isync 177 blr 178 179_GLOBAL(flush_inval_dcache_range) 180 ld r10,PPC64_CACHES@toc(r2) 181 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */ 182 addi r5,r7,-1 183 andc r6,r3,r5 /* round low to line bdy */ 184 subf r8,r6,r4 /* compute length */ 185 add r8,r8,r5 /* ensure we get enough */ 186 lwz r9,DCACHEL1LOGLINESIZE(r10)/* Get log-2 of dcache line size */ 187 srw. r8,r8,r9 /* compute line count */ 188 beqlr /* nothing to do? */ 189 sync 190 isync 191 mtctr r8 1920: dcbf 0,r6 193 add r6,r6,r7 194 bdnz 0b 195 sync 196 isync 197 blr 198 199 200/* 201 * Flush a particular page from the data cache to RAM. 202 * Note: this is necessary because the instruction cache does *not* 203 * snoop from the data cache. 204 * 205 * void __flush_dcache_icache(void *page) 206 */ 207_GLOBAL(__flush_dcache_icache) 208/* 209 * Flush the data cache to memory 210 * 211 * Different systems have different cache line sizes 212 */ 213 214/* Flush the dcache */ 215 ld r7,PPC64_CACHES@toc(r2) 216 clrrdi r3,r3,PAGE_SHIFT /* Page align */ 217 lwz r4,DCACHEL1LINESPERPAGE(r7) /* Get # dcache lines per page */ 218 lwz r5,DCACHEL1LINESIZE(r7) /* Get dcache line size */ 219 mr r6,r3 220 mtctr r4 2210: dcbst 0,r6 222 add r6,r6,r5 223 bdnz 0b 224 sync 225 226/* Now invalidate the icache */ 227 228 lwz r4,ICACHEL1LINESPERPAGE(r7) /* Get # icache lines per page */ 229 lwz r5,ICACHEL1LINESIZE(r7) /* Get icache line size */ 230 mtctr r4 2311: icbi 0,r3 232 add r3,r3,r5 233 bdnz 1b 234 isync 235 blr 236 237 238#if defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) 239/* 240 * Do an IO access in real mode 241 */ 242_GLOBAL(real_readb) 243 mfmsr r7 244 ori r0,r7,MSR_DR 245 xori r0,r0,MSR_DR 246 sync 247 mtmsrd r0 248 sync 249 isync 250 mfspr r6,SPRN_HID4 251 rldicl r5,r6,32,0 252 ori r5,r5,0x100 253 rldicl r5,r5,32,0 254 sync 255 mtspr SPRN_HID4,r5 256 isync 257 slbia 258 isync 259 lbz r3,0(r3) 260 sync 261 mtspr SPRN_HID4,r6 262 isync 263 slbia 264 isync 265 mtmsrd r7 266 sync 267 isync 268 blr 269 270 /* 271 * Do an IO access in real mode 272 */ 273_GLOBAL(real_writeb) 274 mfmsr r7 275 ori r0,r7,MSR_DR 276 xori r0,r0,MSR_DR 277 sync 278 mtmsrd r0 279 sync 280 isync 281 mfspr r6,SPRN_HID4 282 rldicl r5,r6,32,0 283 ori r5,r5,0x100 284 rldicl r5,r5,32,0 285 sync 286 mtspr SPRN_HID4,r5 287 isync 288 slbia 289 isync 290 stb r3,0(r4) 291 sync 292 mtspr SPRN_HID4,r6 293 isync 294 slbia 295 isync 296 mtmsrd r7 297 sync 298 isync 299 blr 300#endif /* defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) */ 301 302#ifdef CONFIG_PPC_PASEMI 303 304/* No support in all binutils for these yet, so use defines */ 305#define LBZCIX(RT,RA,RB) .long (0x7c0006aa|(RT<<21)|(RA<<16)|(RB << 11)) 306#define STBCIX(RS,RA,RB) .long (0x7c0007aa|(RS<<21)|(RA<<16)|(RB << 11)) 307 308 309_GLOBAL(real_205_readb) 310 mfmsr r7 311 ori r0,r7,MSR_DR 312 xori r0,r0,MSR_DR 313 sync 314 mtmsrd r0 315 sync 316 isync 317 LBZCIX(r3,0,r3) 318 isync 319 mtmsrd r7 320 sync 321 isync 322 blr 323 324_GLOBAL(real_205_writeb) 325 mfmsr r7 326 ori r0,r7,MSR_DR 327 xori r0,r0,MSR_DR 328 sync 329 mtmsrd r0 330 sync 331 isync 332 STBCIX(r3,0,r4) 333 isync 334 mtmsrd r7 335 sync 336 isync 337 blr 338 339#endif /* CONFIG_PPC_PASEMI */ 340 341 342#if defined(CONFIG_CPU_FREQ_PMAC64) || defined(CONFIG_CPU_FREQ_MAPLE) 343/* 344 * SCOM access functions for 970 (FX only for now) 345 * 346 * unsigned long scom970_read(unsigned int address); 347 * void scom970_write(unsigned int address, unsigned long value); 348 * 349 * The address passed in is the 24 bits register address. This code 350 * is 970 specific and will not check the status bits, so you should 351 * know what you are doing. 352 */ 353_GLOBAL(scom970_read) 354 /* interrupts off */ 355 mfmsr r4 356 ori r0,r4,MSR_EE 357 xori r0,r0,MSR_EE 358 mtmsrd r0,1 359 360 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits 361 * (including parity). On current CPUs they must be 0'd, 362 * and finally or in RW bit 363 */ 364 rlwinm r3,r3,8,0,15 365 ori r3,r3,0x8000 366 367 /* do the actual scom read */ 368 sync 369 mtspr SPRN_SCOMC,r3 370 isync 371 mfspr r3,SPRN_SCOMD 372 isync 373 mfspr r0,SPRN_SCOMC 374 isync 375 376 /* XXX: fixup result on some buggy 970's (ouch ! we lost a bit, bah 377 * that's the best we can do). Not implemented yet as we don't use 378 * the scom on any of the bogus CPUs yet, but may have to be done 379 * ultimately 380 */ 381 382 /* restore interrupts */ 383 mtmsrd r4,1 384 blr 385 386 387_GLOBAL(scom970_write) 388 /* interrupts off */ 389 mfmsr r5 390 ori r0,r5,MSR_EE 391 xori r0,r0,MSR_EE 392 mtmsrd r0,1 393 394 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits 395 * (including parity). On current CPUs they must be 0'd. 396 */ 397 398 rlwinm r3,r3,8,0,15 399 400 sync 401 mtspr SPRN_SCOMD,r4 /* write data */ 402 isync 403 mtspr SPRN_SCOMC,r3 /* write command */ 404 isync 405 mfspr 3,SPRN_SCOMC 406 isync 407 408 /* restore interrupts */ 409 mtmsrd r5,1 410 blr 411#endif /* CONFIG_CPU_FREQ_PMAC64 || CONFIG_CPU_FREQ_MAPLE */ 412 413 414/* 415 * Create a kernel thread 416 * kernel_thread(fn, arg, flags) 417 */ 418_GLOBAL(kernel_thread) 419 std r29,-24(r1) 420 std r30,-16(r1) 421 stdu r1,-STACK_FRAME_OVERHEAD(r1) 422 mr r29,r3 423 mr r30,r4 424 ori r3,r5,CLONE_VM /* flags */ 425 oris r3,r3,(CLONE_UNTRACED>>16) 426 li r4,0 /* new sp (unused) */ 427 li r0,__NR_clone 428 sc 429 bns+ 1f /* did system call indicate error? */ 430 neg r3,r3 /* if so, make return code negative */ 4311: cmpdi 0,r3,0 /* parent or child? */ 432 bne 2f /* return if parent */ 433 li r0,0 434 stdu r0,-STACK_FRAME_OVERHEAD(r1) 435 ld r2,8(r29) 436 ld r29,0(r29) 437 mtlr r29 /* fn addr in lr */ 438 mr r3,r30 /* load arg and call fn */ 439 blrl 440 li r0,__NR_exit /* exit after child exits */ 441 li r3,0 442 sc 4432: addi r1,r1,STACK_FRAME_OVERHEAD 444 ld r29,-24(r1) 445 ld r30,-16(r1) 446 blr 447 448/* 449 * disable_kernel_fp() 450 * Disable the FPU. 451 */ 452_GLOBAL(disable_kernel_fp) 453 mfmsr r3 454 rldicl r0,r3,(63-MSR_FP_LG),1 455 rldicl r3,r0,(MSR_FP_LG+1),0 456 mtmsrd r3 /* disable use of fpu now */ 457 isync 458 blr 459 460/* kexec_wait(phys_cpu) 461 * 462 * wait for the flag to change, indicating this kernel is going away but 463 * the slave code for the next one is at addresses 0 to 100. 464 * 465 * This is used by all slaves, even those that did not find a matching 466 * paca in the secondary startup code. 467 * 468 * Physical (hardware) cpu id should be in r3. 469 */ 470_GLOBAL(kexec_wait) 471 bl 1f 4721: mflr r5 473 addi r5,r5,kexec_flag-1b 474 47599: HMT_LOW 476#ifdef CONFIG_KEXEC /* use no memory without kexec */ 477 lwz r4,0(r5) 478 cmpwi 0,r4,0 479 bnea 0x60 480#endif 481 b 99b 482 483/* this can be in text because we won't change it until we are 484 * running in real anyways 485 */ 486kexec_flag: 487 .long 0 488 489 490#ifdef CONFIG_KEXEC 491 492/* kexec_smp_wait(void) 493 * 494 * call with interrupts off 495 * note: this is a terminal routine, it does not save lr 496 * 497 * get phys id from paca 498 * switch to real mode 499 * mark the paca as no longer used 500 * join other cpus in kexec_wait(phys_id) 501 */ 502_GLOBAL(kexec_smp_wait) 503 lhz r3,PACAHWCPUID(r13) 504 bl real_mode 505 506 li r4,KEXEC_STATE_REAL_MODE 507 stb r4,PACAKEXECSTATE(r13) 508 SYNC 509 510 b .kexec_wait 511 512/* 513 * switch to real mode (turn mmu off) 514 * we use the early kernel trick that the hardware ignores bits 515 * 0 and 1 (big endian) of the effective address in real mode 516 * 517 * don't overwrite r3 here, it is live for kexec_wait above. 518 */ 519real_mode: /* assume normal blr return */ 5201: li r9,MSR_RI 521 li r10,MSR_DR|MSR_IR 522 mflr r11 /* return address to SRR0 */ 523 mfmsr r12 524 andc r9,r12,r9 525 andc r10,r12,r10 526 527 mtmsrd r9,1 528 mtspr SPRN_SRR1,r10 529 mtspr SPRN_SRR0,r11 530 rfid 531 532 533/* 534 * kexec_sequence(newstack, start, image, control, clear_all()) 535 * 536 * does the grungy work with stack switching and real mode switches 537 * also does simple calls to other code 538 */ 539 540_GLOBAL(kexec_sequence) 541 mflr r0 542 std r0,16(r1) 543 544 /* switch stacks to newstack -- &kexec_stack.stack */ 545 stdu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3) 546 mr r1,r3 547 548 li r0,0 549 std r0,16(r1) 550 551 /* save regs for local vars on new stack. 552 * yes, we won't go back, but ... 553 */ 554 std r31,-8(r1) 555 std r30,-16(r1) 556 std r29,-24(r1) 557 std r28,-32(r1) 558 std r27,-40(r1) 559 std r26,-48(r1) 560 std r25,-56(r1) 561 562 stdu r1,-STACK_FRAME_OVERHEAD-64(r1) 563 564 /* save args into preserved regs */ 565 mr r31,r3 /* newstack (both) */ 566 mr r30,r4 /* start (real) */ 567 mr r29,r5 /* image (virt) */ 568 mr r28,r6 /* control, unused */ 569 mr r27,r7 /* clear_all() fn desc */ 570 mr r26,r8 /* spare */ 571 lhz r25,PACAHWCPUID(r13) /* get our phys cpu from paca */ 572 573 /* disable interrupts, we are overwriting kernel data next */ 574 mfmsr r3 575 rlwinm r3,r3,0,17,15 576 mtmsrd r3,1 577 578 /* copy dest pages, flush whole dest image */ 579 mr r3,r29 580 bl .kexec_copy_flush /* (image) */ 581 582 /* turn off mmu */ 583 bl real_mode 584 585 /* copy 0x100 bytes starting at start to 0 */ 586 li r3,0 587 mr r4,r30 /* start, aka phys mem offset */ 588 li r5,0x100 589 li r6,0 590 bl .copy_and_flush /* (dest, src, copy limit, start offset) */ 5911: /* assume normal blr return */ 592 593 /* release other cpus to the new kernel secondary start at 0x60 */ 594 mflr r5 595 li r6,1 596 stw r6,kexec_flag-1b(5) 597 598 /* clear out hardware hash page table and tlb */ 599 ld r5,0(r27) /* deref function descriptor */ 600 mtctr r5 601 bctrl /* ppc_md.hpte_clear_all(void); */ 602 603/* 604 * kexec image calling is: 605 * the first 0x100 bytes of the entry point are copied to 0 606 * 607 * all slaves branch to slave = 0x60 (absolute) 608 * slave(phys_cpu_id); 609 * 610 * master goes to start = entry point 611 * start(phys_cpu_id, start, 0); 612 * 613 * 614 * a wrapper is needed to call existing kernels, here is an approximate 615 * description of one method: 616 * 617 * v2: (2.6.10) 618 * start will be near the boot_block (maybe 0x100 bytes before it?) 619 * it will have a 0x60, which will b to boot_block, where it will wait 620 * and 0 will store phys into struct boot-block and load r3 from there, 621 * copy kernel 0-0x100 and tell slaves to back down to 0x60 again 622 * 623 * v1: (2.6.9) 624 * boot block will have all cpus scanning device tree to see if they 625 * are the boot cpu ????? 626 * other device tree differences (prop sizes, va vs pa, etc)... 627 */ 628 mr r3,r25 # my phys cpu 629 mr r4,r30 # start, aka phys mem offset 630 mtlr 4 631 li r5,0 632 blr /* image->start(physid, image->start, 0); */ 633#endif /* CONFIG_KEXEC */ 634