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
11 * PPC44x port. Copyright (C) 2011, IBM Corporation
12 * Author: Suzuki Poulose <suzuki@in.ibm.com>
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.
21 #include <linux/sys.h>
22 #include <asm/unistd.h>
23 #include <asm/errno.h>
26 #include <asm/cache.h>
27 #include <asm/cputable.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>
35 #include <asm/ptrace.h>
36 #include <asm/export.h>
41 * We store the saved ksp_limit in the unused part
42 * of the STACK_FRAME_OVERHEAD
44 _GLOBAL(call_do_softirq)
47 lwz r10,THREAD+KSP_LIMIT(r2)
48 addi r11,r3,THREAD_INFO_GAP
49 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
52 stw r11,THREAD+KSP_LIMIT(r2)
57 stw r10,THREAD+KSP_LIMIT(r2)
62 * void call_do_irq(struct pt_regs *regs, struct thread_info *irqtp);
67 lwz r10,THREAD+KSP_LIMIT(r2)
68 addi r11,r4,THREAD_INFO_GAP
69 stwu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r4)
72 stw r11,THREAD+KSP_LIMIT(r2)
77 stw r10,THREAD+KSP_LIMIT(r2)
82 * This returns the high 64 bits of the product of two 64-bit numbers.
94 1: beqlr cr1 /* all done if high part of A is 0 */
108 * reloc_got2 runs through the .got2 section adding an offset
113 lis r7,__got2_start@ha
114 addi r7,r7,__got2_start@l
116 addi r8,r8,__got2_end@l
136 * call_setup_cpu - call the setup_cpu function for this cpu
137 * r3 = data offset, r24 = cpu number
139 * Setup function is called with:
141 * r4 = ptr to CPU spec (relocated)
143 _GLOBAL(call_setup_cpu)
144 addis r4,r3,cur_cpu_spec@ha
145 addi r4,r4,cur_cpu_spec@l
148 lwz r5,CPU_SPEC_SETUP(r4)
155 #if defined(CONFIG_CPU_FREQ_PMAC) && defined(CONFIG_6xx)
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
161 _GLOBAL(low_choose_750fx_pll)
167 /* If switching to PLL1, disable HID0:BTIC */
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 ? */
185 /* Store new HID1 image */
186 CURRENT_THREAD_INFO(r6, r1)
189 addis r6,r6,nap_save_hid1@ha
190 stw r4,nap_save_hid1@l(r6)
192 /* If switching to PLL0, enable HID0:BTIC */
207 _GLOBAL(low_choose_7447a_dfs)
213 /* Calc new HID1 value */
215 insrwi r4,r3,1,9 /* insert parameter into bit 9 */
225 #endif /* CONFIG_CPU_FREQ_PMAC && CONFIG_6xx */
228 * complement mask on the msr then "or" some values on.
229 * _nmask_and_or_msr(nmask, value_to_or)
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 */
243 * Do an IO access in real mode
261 * Do an IO access in real mode
278 #endif /* CONFIG_40x */
282 * Flush instruction cache.
283 * This is a no-op on the 601.
285 #ifndef CONFIG_PPC_8xx
286 _GLOBAL(flush_instruction_cache)
287 #if defined(CONFIG_4xx)
299 #elif CONFIG_FSL_BOOKE
302 ori r3,r3,L1CSR0_CFI|L1CSR0_CLFC
303 /* msync; isync recommended here */
307 END_FTR_SECTION_IFSET(CPU_FTR_UNIFIED_ID_CACHE)
309 ori r3,r3,L1CSR1_ICFI|L1CSR1_ICLFR
313 rlwinm r3,r3,16,16,31
315 beqlr /* for 601, do nothing */
316 /* 603/604 processor - use invalidate-all bit in HID0 */
320 #endif /* CONFIG_4xx */
323 EXPORT_SYMBOL(flush_instruction_cache)
324 #endif /* CONFIG_PPC_8xx */
327 * Write any modified data cache blocks out to memory
328 * and invalidate the corresponding instruction cache blocks.
329 * This is a no-op on the 601.
331 * flush_icache_range(unsigned long start, unsigned long stop)
333 _GLOBAL(flush_icache_range)
336 blr /* for 601, do nothing */
337 END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
338 rlwinm r3,r3,0,0,31 - L1_CACHE_SHIFT
340 addi r4,r4,L1_CACHE_BYTES - 1
341 srwi. r4,r4,L1_CACHE_SHIFT
346 addi r3,r3,L1_CACHE_BYTES
348 sync /* wait for dcbst's to get to ram */
352 addi r6,r6,L1_CACHE_BYTES
355 /* Flash invalidate on 44x because we are passed kmapped addresses and
356 this doesn't work for userspace pages due to the virtually tagged
360 sync /* additional sync needed on g4 */
363 _ASM_NOKPROBE_SYMBOL(flush_icache_range)
364 EXPORT_SYMBOL(flush_icache_range)
367 * Flush a particular page from the data cache to RAM.
368 * Note: this is necessary because the instruction cache does *not*
369 * snoop from the data cache.
370 * This is a no-op on the 601 which has a unified cache.
372 * void __flush_dcache_icache(void *page)
374 _GLOBAL(__flush_dcache_icache)
378 END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
379 rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */
380 li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */
383 0: dcbst 0,r3 /* Write line to ram */
384 addi r3,r3,L1_CACHE_BYTES
388 /* We don't flush the icache on 44x. Those have a virtual icache
389 * and we don't have access to the virtual address here (it's
390 * not the page vaddr but where it's mapped in user space). The
391 * flushing of the icache on these is handled elsewhere, when
392 * a change in the address space occurs, before returning to
395 BEGIN_MMU_FTR_SECTION
397 END_MMU_FTR_SECTION_IFSET(MMU_FTR_TYPE_44x)
398 #endif /* CONFIG_44x */
401 addi r6,r6,L1_CACHE_BYTES
409 * Flush a particular page from the data cache to RAM, identified
410 * by its physical address. We turn off the MMU so we can just use
411 * the physical address (this may be a highmem page without a kernel
414 * void __flush_dcache_icache_phys(unsigned long physaddr)
416 _GLOBAL(__flush_dcache_icache_phys)
419 blr /* for 601, do nothing */
420 END_FTR_SECTION_IFSET(CPU_FTR_COHERENT_ICACHE)
422 rlwinm r0,r10,0,28,26 /* clear DR */
425 rlwinm r3,r3,0,0,31-PAGE_SHIFT /* Get page base address */
426 li r4,PAGE_SIZE/L1_CACHE_BYTES /* Number of lines in a page */
429 0: dcbst 0,r3 /* Write line to ram */
430 addi r3,r3,L1_CACHE_BYTES
435 addi r6,r6,L1_CACHE_BYTES
438 mtmsr r10 /* restore DR */
441 #endif /* CONFIG_BOOKE */
444 * Copy a whole page. We use the dcbz instruction on the destination
445 * to reduce memory traffic (it eliminates the unnecessary reads of
446 * the destination into cache). This requires that the destination
449 #define COPY_16_BYTES \
465 #if MAX_COPY_PREFETCH > 1
466 li r0,MAX_COPY_PREFETCH
470 addi r11,r11,L1_CACHE_BYTES
472 #else /* MAX_COPY_PREFETCH == 1 */
474 li r11,L1_CACHE_BYTES+4
475 #endif /* MAX_COPY_PREFETCH */
476 li r0,PAGE_SIZE/L1_CACHE_BYTES - MAX_COPY_PREFETCH
484 #if L1_CACHE_BYTES >= 32
486 #if L1_CACHE_BYTES >= 64
489 #if L1_CACHE_BYTES >= 128
499 crnot 4*cr0+eq,4*cr0+eq
500 li r0,MAX_COPY_PREFETCH
503 EXPORT_SYMBOL(copy_page)
506 * Extended precision shifts.
508 * Updated to be valid for shift counts from 0 to 63 inclusive.
511 * R3/R4 has 64 bit value
515 * ashrdi3: arithmetic right shift (sign propagation)
516 * lshrdi3: logical right shift
517 * ashldi3: left shift
521 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
522 addi r7,r5,32 # could be xori, or addi with -32
523 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
524 rlwinm r8,r7,0,32 # t3 = (count < 32) ? 32 : 0
525 sraw r7,r3,r7 # t2 = MSW >> (count-32)
526 or r4,r4,r6 # LSW |= t1
527 slw r7,r7,r8 # t2 = (count < 32) ? 0 : t2
528 sraw r3,r3,r5 # MSW = MSW >> count
529 or r4,r4,r7 # LSW |= t2
531 EXPORT_SYMBOL(__ashrdi3)
535 slw r3,r3,r5 # MSW = count > 31 ? 0 : MSW << count
536 addi r7,r5,32 # could be xori, or addi with -32
537 srw r6,r4,r6 # t1 = count > 31 ? 0 : LSW >> (32-count)
538 slw r7,r4,r7 # t2 = count < 32 ? 0 : LSW << (count-32)
539 or r3,r3,r6 # MSW |= t1
540 slw r4,r4,r5 # LSW = LSW << count
541 or r3,r3,r7 # MSW |= t2
543 EXPORT_SYMBOL(__ashldi3)
547 srw r4,r4,r5 # LSW = count > 31 ? 0 : LSW >> count
548 addi r7,r5,32 # could be xori, or addi with -32
549 slw r6,r3,r6 # t1 = count > 31 ? 0 : MSW << (32-count)
550 srw r7,r3,r7 # t2 = count < 32 ? 0 : MSW >> (count-32)
551 or r4,r4,r6 # LSW |= t1
552 srw r3,r3,r5 # MSW = MSW >> count
553 or r4,r4,r7 # LSW |= t2
555 EXPORT_SYMBOL(__lshrdi3)
558 * 64-bit comparison: __cmpdi2(s64 a, s64 b)
559 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
571 EXPORT_SYMBOL(__cmpdi2)
573 * 64-bit comparison: __ucmpdi2(u64 a, u64 b)
574 * Returns 0 if a < b, 1 if a == b, 2 if a > b.
586 EXPORT_SYMBOL(__ucmpdi2)
593 rlwimi r9,r4,24,16,23
594 rlwimi r10,r3,24,16,23
598 EXPORT_SYMBOL(__bswapdi2)
601 _GLOBAL(start_secondary_resume)
603 CURRENT_THREAD_INFO(r1, r1)
604 addi r1,r1,THREAD_SIZE-STACK_FRAME_OVERHEAD
606 stw r3,0(r1) /* Zero the stack frame pointer */
609 #endif /* CONFIG_SMP */
612 * This routine is just here to keep GCC happy - sigh...
619 * Must be relocatable PIC code callable as a C function.
621 .globl relocate_new_kernel
624 /* r4 = reboot_code_buffer */
625 /* r5 = start_address */
627 #ifdef CONFIG_FSL_BOOKE
633 #define ENTRY_MAPPING_KEXEC_SETUP
634 #include "fsl_booke_entry_mapping.S"
635 #undef ENTRY_MAPPING_KEXEC_SETUP
642 #elif defined(CONFIG_44x)
644 /* Save our parameters */
649 #ifdef CONFIG_PPC_47x
650 /* Check for 47x cores */
653 cmplwi cr0,r3,PVR_476FPE@h
655 cmplwi cr0,r3,PVR_476@h
657 cmplwi cr0,r3,PVR_476_ISS@h
659 #endif /* CONFIG_PPC_47x */
662 * Code for setting up 1:1 mapping for PPC440x for KEXEC
664 * We cannot switch off the MMU on PPC44x.
666 * 1) Invalidate all the mappings except the one we are running from.
667 * 2) Create a tmp mapping for our code in the other address space(TS) and
668 * jump to it. Invalidate the entry we started in.
669 * 3) Create a 1:1 mapping for 0-2GiB in chunks of 256M in original TS.
670 * 4) Jump to the 1:1 mapping in original TS.
671 * 5) Invalidate the tmp mapping.
673 * - Based on the kexec support code for FSL BookE
678 * Load the PID with kernel PID (0).
679 * Also load our MSR_IS and TID to MMUCR for TLB search.
686 oris r3,r3,PPC44x_MMUCR_STS@h
692 * Invalidate all the TLB entries except the current entry
693 * where we are running from
695 bl 0f /* Find our address */
696 0: mflr r5 /* Make it accessible */
697 tlbsx r23,0,r5 /* Find entry we are in */
698 li r4,0 /* Start at TLB entry 0 */
699 li r3,0 /* Set PAGEID inval value */
700 1: cmpw r23,r4 /* Is this our entry? */
701 beq skip /* If so, skip the inval */
702 tlbwe r3,r4,PPC44x_TLB_PAGEID /* If not, inval the entry */
704 addi r4,r4,1 /* Increment */
705 cmpwi r4,64 /* Are we done? */
706 bne 1b /* If not, repeat */
709 /* Create a temp mapping and jump to it */
710 andi. r6, r23, 1 /* Find the index to use */
711 addi r24, r6, 1 /* r24 will contain 1 or 2 */
713 mfmsr r9 /* get the MSR */
714 rlwinm r5, r9, 27, 31, 31 /* Extract the MSR[IS] */
715 xori r7, r5, 1 /* Use the other address space */
717 /* Read the current mapping entries */
718 tlbre r3, r23, PPC44x_TLB_PAGEID
719 tlbre r4, r23, PPC44x_TLB_XLAT
720 tlbre r5, r23, PPC44x_TLB_ATTRIB
722 /* Save our current XLAT entry */
725 /* Extract the TLB PageSize */
726 li r10, 1 /* r10 will hold PageSize */
727 rlwinm r11, r3, 0, 24, 27 /* bits 24-27 */
729 /* XXX: As of now we use 256M, 4K pages */
730 cmpwi r11, PPC44x_TLB_256M
732 rotlwi r10, r10, 28 /* r10 = 256M */
735 cmpwi r11, PPC44x_TLB_4K
737 rotlwi r10, r10, 12 /* r10 = 4K */
740 rotlwi r10, r10, 10 /* r10 = 1K */
744 * Write out the tmp 1:1 mapping for this code in other address space
745 * Fixup EPN = RPN , TS=other address space
747 insrwi r3, r7, 1, 23 /* Bit 23 is TS for PAGEID field */
749 /* Write out the tmp mapping entries */
750 tlbwe r3, r24, PPC44x_TLB_PAGEID
751 tlbwe r4, r24, PPC44x_TLB_XLAT
752 tlbwe r5, r24, PPC44x_TLB_ATTRIB
754 subi r11, r10, 1 /* PageOffset Mask = PageSize - 1 */
755 not r10, r11 /* Mask for PageNum */
757 /* Switch to other address space in MSR */
758 insrwi r9, r7, 1, 26 /* Set MSR[IS] = r7 */
762 addi r8, r8, (2f-1b) /* Find the target offset */
764 /* Jump to the tmp mapping */
770 /* Invalidate the entry we were executing from */
772 tlbwe r3, r23, PPC44x_TLB_PAGEID
774 /* attribute fields. rwx for SUPERVISOR mode */
776 ori r5, r5, (PPC44x_TLB_SW | PPC44x_TLB_SR | PPC44x_TLB_SX | PPC44x_TLB_G)
778 /* Create 1:1 mapping in 256M pages */
779 xori r7, r7, 1 /* Revert back to Original TS */
781 li r8, 0 /* PageNumber */
782 li r6, 3 /* TLB Index, start at 3 */
785 rotlwi r3, r8, 28 /* Create EPN (bits 0-3) */
786 mr r4, r3 /* RPN = EPN */
787 ori r3, r3, (PPC44x_TLB_VALID | PPC44x_TLB_256M) /* SIZE = 256M, Valid */
788 insrwi r3, r7, 1, 23 /* Set TS from r7 */
790 tlbwe r3, r6, PPC44x_TLB_PAGEID /* PageID field : EPN, V, SIZE */
791 tlbwe r4, r6, PPC44x_TLB_XLAT /* Address translation : RPN */
792 tlbwe r5, r6, PPC44x_TLB_ATTRIB /* Attributes */
794 addi r8, r8, 1 /* Increment PN */
795 addi r6, r6, 1 /* Increment TLB Index */
796 cmpwi r8, 8 /* Are we done ? */
800 /* Jump to the new mapping 1:1 */
802 insrwi r9, r7, 1, 26 /* Set MSR[IS] = r7 */
806 and r8, r8, r11 /* Get our offset within page */
809 and r5, r25, r10 /* Get our target PageNum */
810 or r8, r8, r5 /* Target jump address */
816 /* Invalidate the tmp entry we used */
818 tlbwe r3, r24, PPC44x_TLB_PAGEID
822 #ifdef CONFIG_PPC_47x
824 /* 1:1 mapping for 47x */
829 * Load the kernel pid (0) to PID and also to MMUCR[TID].
830 * Also set the MSR IS->MMUCR STS
833 mtspr SPRN_PID, r3 /* Set PID */
834 mfmsr r4 /* Get MSR */
835 andi. r4, r4, MSR_IS@l /* TS=1? */
836 beq 1f /* If not, leave STS=0 */
837 oris r3, r3, PPC47x_MMUCR_STS@h /* Set STS=1 */
838 1: mtspr SPRN_MMUCR, r3 /* Put MMUCR */
841 /* Find the entry we are running from */
845 tlbre r24, r23, 0 /* TLB Word 0 */
846 tlbre r25, r23, 1 /* TLB Word 1 */
847 tlbre r26, r23, 2 /* TLB Word 2 */
851 * Invalidates all the tlb entries by writing to 256 RPNs(r4)
852 * of 4k page size in all 4 ways (0-3 in r3).
853 * This would invalidate the entire UTLB including the one we are
854 * running from. However the shadow TLB entries would help us
855 * to continue the execution, until we flush them (rfi/isync).
857 addis r3, 0, 0x8000 /* specify the way */
858 addi r4, 0, 0 /* TLB Word0 = (EPN=0, VALID = 0) */
862 /* Align the loop to speed things up. from head_44x.S */
870 addis r3, r3, 0x2000 /* Increment the way */
874 addis r4, r4, 0x100 /* Increment the EPN */
878 /* Create the entries in the other address space */
880 rlwinm r7, r5, 27, 31, 31 /* Get the TS (Bit 26) from MSR */
881 xori r7, r7, 1 /* r7 = !TS */
883 insrwi r24, r7, 1, 21 /* Change the TS in the saved TLB word 0 */
886 * write out the TLB entries for the tmp mapping
887 * Use way '0' so that we could easily invalidate it later.
889 lis r3, 0x8000 /* Way '0' */
895 /* Update the msr to the new TS */
907 * Now we are in the tmp address space.
908 * Create a 1:1 mapping for 0-2GiB in the original TS.
912 li r4, 0 /* TLB Word 0 */
913 li r5, 0 /* TLB Word 1 */
915 ori r6, r6, PPC47x_TLB2_S_RWX /* TLB word 2 */
917 li r8, 0 /* PageIndex */
919 xori r7, r7, 1 /* revert back to original TS */
922 rotlwi r5, r8, 28 /* RPN = PageIndex * 256M */
923 /* ERPN = 0 as we don't use memory above 2G */
925 mr r4, r5 /* EPN = RPN */
926 ori r4, r4, (PPC47x_TLB0_VALID | PPC47x_TLB0_256M)
927 insrwi r4, r7, 1, 21 /* Insert the TS to Word 0 */
929 tlbwe r4, r3, 0 /* Write out the entries */
933 cmpwi r8, 8 /* Have we completed ? */
936 /* make sure we complete the TLB write up */
940 * Prepare to jump to the 1:1 mapping.
941 * 1) Extract page size of the tmp mapping
942 * DSIZ = TLB_Word0[22:27]
943 * 2) Calculate the physical address of the address
946 rlwinm r10, r24, 0, 22, 27
948 cmpwi r10, PPC47x_TLB0_4K
950 li r10, 0x1000 /* r10 = 4k */
954 /* Defaults to 256M */
959 addi r4, r4, (2f-1b) /* virtual address of 2f */
961 subi r11, r10, 1 /* offsetmask = Pagesize - 1 */
962 not r10, r11 /* Pagemask = ~(offsetmask) */
964 and r5, r25, r10 /* Physical page */
965 and r6, r4, r11 /* offset within the current page */
967 or r5, r5, r6 /* Physical address for 2f */
969 /* Switch the TS in MSR to the original one */
978 /* Invalidate the tmp mapping */
979 lis r3, 0x8000 /* Way '0' */
981 clrrwi r24, r24, 12 /* Clear the valid bit */
986 /* Make sure we complete the TLB write and flush the shadow TLB */
994 /* Restore the parameters */
1004 * Set Machine Status Register to a known status,
1005 * switch the MMU off and jump to 1: in a single step.
1009 ori r8, r8, MSR_RI|MSR_ME
1011 addi r8, r4, 1f - relocate_new_kernel
1018 /* from this point address translation is turned off */
1019 /* and interrupts are disabled */
1021 /* set a new stack at the bottom of our page... */
1022 /* (not really needed now) */
1023 addi r1, r4, KEXEC_CONTROL_PAGE_SIZE - 8 /* for LR Save+Back Chain */
1027 li r6, 0 /* checksum */
1031 0: /* top, read another word for the indirection page */
1035 /* is it a destination page? (r8) */
1036 rlwinm. r7, r0, 0, 31, 31 /* IND_DESTINATION (1<<0) */
1039 rlwinm r8, r0, 0, 0, 19 /* clear kexec flags, page align */
1042 2: /* is it an indirection page? (r3) */
1043 rlwinm. r7, r0, 0, 30, 30 /* IND_INDIRECTION (1<<1) */
1046 rlwinm r3, r0, 0, 0, 19 /* clear kexec flags, page align */
1050 2: /* are we done? */
1051 rlwinm. r7, r0, 0, 29, 29 /* IND_DONE (1<<2) */
1055 2: /* is it a source page? (r9) */
1056 rlwinm. r7, r0, 0, 28, 28 /* IND_SOURCE (1<<3) */
1059 rlwinm r9, r0, 0, 0, 19 /* clear kexec flags, page align */
1061 li r7, PAGE_SIZE / 4
1066 lwzu r0, 4(r9) /* do the copy */
1080 /* To be certain of avoiding problems with self-modifying code
1081 * execute a serializing instruction here.
1086 mfspr r3, SPRN_PIR /* current core we are running on */
1087 mr r4, r5 /* load physical address of chunk called */
1089 /* jump to the entry point, usually the setup routine */
1095 relocate_new_kernel_end:
1097 .globl relocate_new_kernel_size
1098 relocate_new_kernel_size:
1099 .long relocate_new_kernel_end - relocate_new_kernel