2 * This file contains miscellaneous low-level functions.
3 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
5 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
9 * Copyright (C) 2002-2003 Eric Biederman <ebiederm@xmission.com>
10 * GameCube/ppc32 port Copyright (C) 2004 Albert Herranz
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.
19 #include <linux/sys.h>
20 #include <asm/unistd.h>
21 #include <asm/errno.h>
24 #include <asm/cache.h>
25 #include <asm/cputable.h>
27 #include <asm/ppc_asm.h>
28 #include <asm/thread_info.h>
29 #include <asm/asm-offsets.h>
30 #include <asm/processor.h>
31 #include <asm/kexec.h>
36 #ifdef CONFIG_IRQSTACKS
37 _GLOBAL(call_do_softirq)
40 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
48 _GLOBAL(call_handle_irq)
52 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r5)
59 #endif /* CONFIG_IRQSTACKS */
62 * This returns the high 64 bits of the product of two 64-bit numbers.
74 1: beqlr cr1 /* all done if high part of A is 0 */
89 * sub_reloc_offset(x) returns x - reloc_offset().
91 _GLOBAL(sub_reloc_offset)
103 * reloc_got2 runs through the .got2 section adding an offset
108 lis r7,__got2_start@ha
109 addi r7,r7,__got2_start@l
111 addi r8,r8,__got2_end@l
131 * call_setup_cpu - call the setup_cpu function for this cpu
132 * r3 = data offset, r24 = cpu number
134 * Setup function is called with:
136 * r4 = ptr to CPU spec (relocated)
138 _GLOBAL(call_setup_cpu)
139 addis r4,r3,cur_cpu_spec@ha
140 addi r4,r4,cur_cpu_spec@l
143 lwz r5,CPU_SPEC_SETUP(r4)
150 #if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
152 /* This gets called by via-pmu.c to switch the PLL selection
153 * on 750fx CPU. This function should really be moved to some
154 * other place (as most of the cpufreq code in via-pmu
156 _GLOBAL(low_choose_750fx_pll)
162 /* If switching to PLL1, disable HID0:BTIC */
173 /* Calc new HID1 value */
174 mfspr r4,SPRN_HID1 /* Build a HID1:PS bit from parameter */
175 rlwinm r5,r3,16,15,15 /* Clear out HID1:PS from value read */
176 rlwinm r4,r4,0,16,14 /* Could have I used rlwimi here ? */
180 /* Store new HID1 image */
181 rlwinm r6,r1,0,0,(31-THREAD_SHIFT)
184 addis r6,r6,nap_save_hid1@ha
185 stw r4,nap_save_hid1@l(r6)
187 /* If switching to PLL0, enable HID0:BTIC */
202 _GLOBAL(low_choose_7447a_dfs)
208 /* Calc new HID1 value */
210 insrwi r4,r3,1,9 /* insert parameter into bit 9 */
220 #endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
223 * complement mask on the msr then "or" some values on.
224 * _nmask_and_or_msr(nmask, value_to_or)
226 _GLOBAL(_nmask_and_or_msr)
227 mfmsr r0 /* Get current msr */
228 andc r0,r0,r3 /* And off the bits set in r3 (first parm) */
229 or r0,r0,r4 /* Or on the bits in r4 (second parm) */
230 SYNC /* Some chip revs have problems here... */
231 mtmsr r0 /* Update machine state */
238 * Do an IO access in real mode
256 * Do an IO access in real mode
273 #endif /* CONFIG_40x */
277 * Flush instruction cache.
278 * This is a no-op on the 601.
280 _GLOBAL(flush_instruction_cache)
281 #if defined(CONFIG_8xx)
284 mtspr SPRN_IC_CST, r5
285 #elif defined(CONFIG_4xx)
297 #elif CONFIG_FSL_BOOKE
300 ori r3,r3,L1CSR0_CFI|L1CSR0_CLFC
301 /* msync; isync recommended here */
305 END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
307 ori r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
311 rlwinm r3,r3,16,16,31
313 beqlr /* for 601, do nothing */
314 /* 603/604 processor - use invalidate-all bit in HID0 */
318 #endif /* CONFIG_8xx/4xx */
323 * Write any modified data cache blocks out to memory
324 * and invalidate the corresponding instruction cache blocks.
325 * This is a no-op on the 601.
327 * flush_icache_range(unsigned long start, unsigned long stop)
329 _KPROBE(__flush_icache_range)
331 blr /* for 601, do nothing */
332 END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
333 li r5,L1_CACHE_BYTES-1
337 srwi. r4,r4,L1_CACHE_SHIFT
342 addi r3,r3,L1_CACHE_BYTES
344 sync /* wait for dcbst's to get to ram */
347 addi r6,r6,L1_CACHE_BYTES
349 sync /* additional sync needed on g4 */
353 * Write any modified data cache blocks out to memory.
354 * Does not invalidate the corresponding cache lines (especially for
355 * any corresponding instruction cache).
357 * clean_dcache_range(unsigned long start, unsigned long stop)
359 _GLOBAL(clean_dcache_range)
360 li r5,L1_CACHE_BYTES-1
364 srwi. r4,r4,L1_CACHE_SHIFT
369 addi r3,r3,L1_CACHE_BYTES
371 sync /* wait for dcbst's to get to ram */
375 * Write any modified data cache blocks out to memory and invalidate them.
376 * Does not invalidate the corresponding instruction cache blocks.
378 * flush_dcache_range(unsigned long start, unsigned long stop)
380 _GLOBAL(flush_dcache_range)
381 li r5,L1_CACHE_BYTES-1
385 srwi. r4,r4,L1_CACHE_SHIFT
390 addi r3,r3,L1_CACHE_BYTES
392 sync /* wait for dcbst's to get to ram */
396 * Like above, but invalidate the D-cache. This is used by the 8xx
397 * to invalidate the cache so the PPC core doesn't get stale data
398 * from the CPM (no cache snooping here :-).
400 * invalidate_dcache_range(unsigned long start, unsigned long stop)
402 _GLOBAL(invalidate_dcache_range)
403 li r5,L1_CACHE_BYTES-1
407 srwi. r4,r4,L1_CACHE_SHIFT
412 addi r3,r3,L1_CACHE_BYTES
414 sync /* wait for dcbi's to get to ram */
418 * Flush a particular page from the data cache to RAM.
419 * Note: this is necessary because the instruction cache does *not*
420 * snoop from the data cache.
421 * This is a no-op on the 601 which has a unified cache.
423 * void __flush_dcache_icache(void *page)
425 _GLOBAL(__flush_dcache_icache)
428 END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
429 rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */
430 li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */
433 0: dcbst 0,r3 /* Write line to ram */
434 addi r3,r3,L1_CACHE_BYTES
438 /* We don't flush the icache on 44x. Those have a virtual icache
439 * and we don't have access to the virtual address here (it's
440 * not the page vaddr but where it's mapped in user space). The
441 * flushing of the icache on these is handled elsewhere, when
442 * a change in the address space occurs, before returning to
447 addi r6,r6,L1_CACHE_BYTES
451 #endif /* CONFIG_44x */
455 * Flush a particular page from the data cache to RAM, identified
456 * by its physical address. We turn off the MMU so we can just use
457 * the physical address (this may be a highmem page without a kernel
460 * void __flush_dcache_icache_phys(unsigned long physaddr)
462 _GLOBAL(__flush_dcache_icache_phys)
464 blr /* for 601, do nothing */
465 END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
467 rlwinm r0,r10,0,28,26 /* clear DR */
470 rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */
471 li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */
474 0: dcbst 0,r3 /* Write line to ram */
475 addi r3,r3,L1_CACHE_BYTES
480 addi r6,r6,L1_CACHE_BYTES
483 mtmsr r10 /* restore DR */
488 * Clear pages using the dcbz instruction, which doesn't cause any
489 * memory traffic (except to write out any cache lines which get
490 * displaced). This only works on cacheable memory.
492 * void clear_pages(void *page, int order) ;
495 li r0,PAGE_SIZE/L1_CACHE_BYTES
507 addi r3,r3,L1_CACHE_BYTES
512 * Copy a whole page. We use the dcbz instruction on the destination
513 * to reduce memory traffic (it eliminates the unnecessary reads of
514 * the destination into cache). This requires that the destination
517 #define COPY_16_BYTES \
532 /* don't use prefetch on 8xx */
533 li r0,4096/L1_CACHE_BYTES
539 #else /* not 8xx, we can prefetch */
542 #if MAX_COPY_PREFETCH > 1
543 li r0,MAX_COPY_PREFETCH
547 addi r11,r11,L1_CACHE_BYTES
549 #else /* MAX_COPY_PREFETCH == 1 */
551 li r11,L1_CACHE_BYTES+4
552 #endif /* MAX_COPY_PREFETCH */
553 li r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
561 #if L1_CACHE_BYTES >= 32
563 #if L1_CACHE_BYTES >= 64
566 #if L1_CACHE_BYTES >= 128
576 crnot 4*cr0+eq,4*cr0+eq
577 li r0,MAX_COPY_PREFETCH
580 #endif /* CONFIG_8xx */
583 * void atomic_clear_mask(atomic_t mask, atomic_t *addr)
584 * void atomic_set_mask(atomic_t mask, atomic_t *addr);
586 _GLOBAL(atomic_clear_mask)
593 _GLOBAL(atomic_set_mask)
602 * Extended precision shifts.
604 * Updated to be valid for shift counts from 0 to 63 inclusive.
607 * R3/R4 has 64 bit value
611 * ashrdi3: arithmetic right shift (sign propagation)
612 * lshrdi3: logical right shift
613 * ashldi3: left shift
617 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
618 addi r7,r5,32 # could be xori, or addi with -32
619 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
620 rlwinm r8,r7,0,32 # t3 = (count < 32) ? 32 : 0
621 sraw r7,r3,r7 # t2 = MSW >> (count-32)
622 or r4,r4,r6 # LSW |= t1
623 slw r7,r7,r8 # t2 = (count < 32) ? 0 : t2
624 sraw r3,r3,r5 # MSW = MSW >> count
625 or r4,r4,r7 # LSW |= t2
630 slw r3,r3,r5 # MSW = count > 31 ? 0 : MSW << count
631 addi r7,r5,32 # could be xori, or addi with -32
632 srw r6,r4,r6 # t1 = count > 31 ? 0 : LSW >> (32-count)
633 slw r7,r4,r7 # t2 = count < 32 ? 0 : LSW << (count-32)
634 or r3,r3,r6 # MSW |= t1
635 slw r4,r4,r5 # LSW = LSW << count
636 or r3,r3,r7 # MSW |= t2
641 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
642 addi r7,r5,32 # could be xori, or addi with -32
643 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
644 srw r7,r3,r7 # t2 = count < 32 ? 0 : MSW >> (count-32)
645 or r4,r4,r6 # LSW |= t1
646 srw r3,r3,r5 # MSW = MSW >> count
647 or r4,r4,r7 # LSW |= t2
651 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
652 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
672 * Create a kernel thread
673 * kernel_thread(fn, arg, flags)
675 _GLOBAL(kernel_thread)
679 mr r30,r3 /* function */
680 mr r31,r4 /* argument */
681 ori r3,r5,CLONE_VM /* flags */
682 oris r3,r3,CLONE_UNTRACED>>16
683 li r4,0 /* new sp (unused) */
686 bns+ 1f /* did system call indicate error? */
687 neg r3,r3 /* if so, make return code negative */
688 1: cmpwi 0,r3,0 /* parent or child? */
689 bne 2f /* return if parent */
690 li r0,0 /* make top-level stack frame */
692 mtlr r30 /* fn addr in lr */
693 mr r3,r31 /* load arg and call fn */
696 li r0,__NR_exit /* exit if function returns */
705 * This routine is just here to keep GCC happy - sigh...
712 * Must be relocatable PIC code callable as a C function.
714 .globl relocate_new_kernel
717 /* r4 = reboot_code_buffer */
718 /* r5 = start_address */
723 * Set Machine Status Register to a known status,
724 * switch the MMU off and jump to 1: in a single step.
728 ori r8, r8, MSR_RI|MSR_ME
730 addi r8, r4, 1f - relocate_new_kernel
736 /* from this point address translation is turned off */
737 /* and interrupts are disabled */
739 /* set a new stack at the bottom of our page... */
740 /* (not really needed now) */
741 addi r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
745 li r6, 0 /* checksum */
749 0: /* top, read another word for the indirection page */
753 /* is it a destination page? (r8) */
754 rlwinm. r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
757 rlwinm r8, r0, 0, 0, 19 /* clear kexec flags, page align */
760 2: /* is it an indirection page? (r3) */
761 rlwinm. r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
764 rlwinm r3, r0, 0, 0, 19 /* clear kexec flags, page align */
768 2: /* are we done? */
769 rlwinm. r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
773 2: /* is it a source page? (r9) */
774 rlwinm. r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
777 rlwinm r9, r0, 0, 0, 19 /* clear kexec flags, page align */
784 lwzu r0, 4(r9) /* do the copy */
798 /* To be certain of avoiding problems with self-modifying code
799 * execute a serializing instruction here.
804 /* jump to the entry point, usually the setup routine */
810 relocate_new_kernel_end:
812 .globl relocate_new_kernel_size
813 relocate_new_kernel_size:
814 .long relocate_new_kernel_end - relocate_new_kernel