2 * arch/powerpc/kernel/misc64.S
4 * This file contains miscellaneous low-level functions.
5 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
7 * Largely rewritten by Cort Dougan (cort@cs.nmt.edu)
9 * Adapted for iSeries by Mike Corrigan (mikejc@us.ibm.com)
10 * PPC64 updates by Dave Engebretsen (engebret@us.ibm.com)
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/config.h>
20 #include <linux/sys.h>
21 #include <asm/unistd.h>
22 #include <asm/errno.h>
23 #include <asm/processor.h>
25 #include <asm/cache.h>
26 #include <asm/ppc_asm.h>
27 #include <asm/asm-offsets.h>
28 #include <asm/cputable.h>
29 #include <asm/thread_info.h>
34 * Returns (address we are running at) - (address we were linked at)
35 * for use before the text and data are mapped to KERNELBASE.
48 * add_reloc_offset(x) returns x + reloc_offset().
50 _GLOBAL(add_reloc_offset)
80 #ifdef CONFIG_IRQSTACKS
81 _GLOBAL(call_do_softirq)
84 stdu r1,THREAD_SIZE-112(r3)
92 _GLOBAL(call_handle_IRQ_event)
95 stdu r1,THREAD_SIZE-112(r6)
102 #endif /* CONFIG_IRQSTACKS */
105 * To be called by C code which needs to do some operations with MMU
106 * disabled. Note that interrupts have to be disabled by the caller
107 * prior to calling us. The code called _MUST_ be in the RMO of course
108 * and part of the linear mapping as we don't attempt to translate the
109 * stack pointer at all. The function is called with the stack switched
110 * to this CPU emergency stack
112 * prototype is void *call_with_mmu_off(void *func, void *data);
114 * the called function is expected to be of the form
116 * void *called(void *data);
118 _GLOBAL(call_with_mmu_off)
119 mflr r0 /* get link, save it on stackframe */
121 mr r1,r5 /* save old stack ptr */
122 ld r1,PACAEMERGSP(r13) /* get emerg. stack */
123 subi r1,r1,STACK_FRAME_OVERHEAD
124 std r0,16(r1) /* save link on emerg. stack */
125 std r5,0(r1) /* save old stack ptr in backchain */
126 ld r3,0(r3) /* get to real function ptr (assume same TOC) */
127 bl 2f /* we need LR to return, continue at label 2 */
129 ld r0,16(r1) /* we return here from the call, get LR and */
130 ld r1,0(r1) /* .. old stack ptr */
131 mtspr SPRN_SRR0,r0 /* and get back to virtual mode with these */
133 ori r4,r4,MSR_IR|MSR_DR
137 2: mtspr SPRN_SRR0,r3 /* coming from above, enter real mode */
138 mr r3,r4 /* get parameter */
140 ori r0,r0,MSR_IR|MSR_DR
141 xori r0,r0,MSR_IR|MSR_DR
148 .tc ppc64_caches[TC],ppc64_caches
152 * Write any modified data cache blocks out to memory
153 * and invalidate the corresponding instruction cache blocks.
155 * flush_icache_range(unsigned long start, unsigned long stop)
157 * flush all bytes from start through stop-1 inclusive
160 _KPROBE(__flush_icache_range)
163 * Flush the data cache to memory
165 * Different systems have different cache line sizes
166 * and in some cases i-cache and d-cache line sizes differ from
169 ld r10,PPC64_CACHES@toc(r2)
170 lwz r7,DCACHEL1LINESIZE(r10)/* Get cache line size */
172 andc r6,r3,r5 /* round low to line bdy */
173 subf r8,r6,r4 /* compute length */
174 add r8,r8,r5 /* ensure we get enough */
175 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of cache line size */
176 srw. r8,r8,r9 /* compute line count */
177 beqlr /* nothing to do? */
184 /* Now invalidate the instruction cache */
186 lwz r7,ICACHEL1LINESIZE(r10) /* Get Icache line size */
188 andc r6,r3,r5 /* round low to line bdy */
189 subf r8,r6,r4 /* compute length */
191 lwz r9,ICACHEL1LOGLINESIZE(r10) /* Get log-2 of Icache line size */
192 srw. r8,r8,r9 /* compute line count */
193 beqlr /* nothing to do? */
202 * Like above, but only do the D-cache.
204 * flush_dcache_range(unsigned long start, unsigned long stop)
206 * flush all bytes from start to stop-1 inclusive
208 _GLOBAL(flush_dcache_range)
211 * Flush the data cache to memory
213 * Different systems have different cache line sizes
215 ld r10,PPC64_CACHES@toc(r2)
216 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
218 andc r6,r3,r5 /* round low to line bdy */
219 subf r8,r6,r4 /* compute length */
220 add r8,r8,r5 /* ensure we get enough */
221 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */
222 srw. r8,r8,r9 /* compute line count */
223 beqlr /* nothing to do? */
232 * Like above, but works on non-mapped physical addresses.
233 * Use only for non-LPAR setups ! It also assumes real mode
234 * is cacheable. Used for flushing out the DART before using
235 * it as uncacheable memory
237 * flush_dcache_phys_range(unsigned long start, unsigned long stop)
239 * flush all bytes from start to stop-1 inclusive
241 _GLOBAL(flush_dcache_phys_range)
242 ld r10,PPC64_CACHES@toc(r2)
243 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
245 andc r6,r3,r5 /* round low to line bdy */
246 subf r8,r6,r4 /* compute length */
247 add r8,r8,r5 /* ensure we get enough */
248 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */
249 srw. r8,r8,r9 /* compute line count */
250 beqlr /* nothing to do? */
251 mfmsr r5 /* Disable MMU Data Relocation */
264 mtmsr r5 /* Re-enable MMU Data Relocation */
269 _GLOBAL(flush_inval_dcache_range)
270 ld r10,PPC64_CACHES@toc(r2)
271 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
273 andc r6,r3,r5 /* round low to line bdy */
274 subf r8,r6,r4 /* compute length */
275 add r8,r8,r5 /* ensure we get enough */
276 lwz r9,DCACHEL1LOGLINESIZE(r10)/* Get log-2 of dcache line size */
277 srw. r8,r8,r9 /* compute line count */
278 beqlr /* nothing to do? */
291 * Flush a particular page from the data cache to RAM.
292 * Note: this is necessary because the instruction cache does *not*
293 * snoop from the data cache.
295 * void __flush_dcache_icache(void *page)
297 _GLOBAL(__flush_dcache_icache)
299 * Flush the data cache to memory
301 * Different systems have different cache line sizes
304 /* Flush the dcache */
305 ld r7,PPC64_CACHES@toc(r2)
306 clrrdi r3,r3,PAGE_SHIFT /* Page align */
307 lwz r4,DCACHEL1LINESPERPAGE(r7) /* Get # dcache lines per page */
308 lwz r5,DCACHEL1LINESIZE(r7) /* Get dcache line size */
316 /* Now invalidate the icache */
318 lwz r4,ICACHEL1LINESPERPAGE(r7) /* Get # icache lines per page */
319 lwz r5,ICACHEL1LINESIZE(r7) /* Get icache line size */
328 * I/O string operations
330 * insb(port, buf, len)
331 * outsb(port, buf, len)
332 * insw(port, buf, len)
333 * outsw(port, buf, len)
334 * insl(port, buf, len)
335 * outsl(port, buf, len)
336 * insw_ns(port, buf, len)
337 * outsw_ns(port, buf, len)
338 * insl_ns(port, buf, len)
339 * outsl_ns(port, buf, len)
341 * The *_ns versions don't do byte-swapping.
415 /* _GLOBAL(ide_insw) now in drivers/ide/ide-iops.c */
429 /* _GLOBAL(ide_outsw) now in drivers/ide/ide-iops.c */
466 * identify_cpu and calls setup_cpu
467 * In: r3 = base of the cpu_specs array
468 * r4 = address of cur_cpu_spec
469 * r5 = relocation offset
471 _GLOBAL(identify_cpu)
474 lwz r8,CPU_SPEC_PVR_MASK(r3)
476 lwz r9,CPU_SPEC_PVR_VALUE(r3)
479 addi r3,r3,CPU_SPEC_ENTRY_SIZE
484 ld r4,CPU_SPEC_SETUP(r3)
489 /* Calling convention for cpu setup is r3=offset, r4=cur_cpu_spec */
495 * do_cpu_ftr_fixups - goes through the list of CPU feature fixups
496 * and writes nop's over sections of code that don't apply for this cpu.
497 * r3 = data offset (not changed)
499 _GLOBAL(do_cpu_ftr_fixups)
500 /* Get CPU 0 features */
501 LOADADDR(r6,cur_cpu_spec)
505 ld r4,CPU_SPEC_FEATURES(r4)
506 /* Get the fixup table */
507 LOADADDR(r6,__start___ftr_fixup)
509 LOADADDR(r7,__stop___ftr_fixup)
515 ld r8,-32(r6) /* mask */
517 ld r9,-24(r6) /* value */
520 ld r8,-16(r6) /* section begin */
521 ld r9,-8(r6) /* section end */
524 /* write nops over the section of code */
525 /* todo: if large section, add a branch at the start of it */
529 lis r0,0x60000000@h /* nop */
531 andi. r10,r4,CPU_FTR_SPLIT_ID_CACHE@l
533 dcbst 0,r8 /* suboptimal, but simpler */
538 sync /* additional sync needed on g4 */
542 #if defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE)
544 * Do an IO access in real mode
575 * Do an IO access in real mode
604 #endif /* defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) */
607 * SCOM access functions for 970 (FX only for now)
609 * unsigned long scom970_read(unsigned int address);
610 * void scom970_write(unsigned int address, unsigned long value);
612 * The address passed in is the 24 bits register address. This code
613 * is 970 specific and will not check the status bits, so you should
614 * know what you are doing.
616 _GLOBAL(scom970_read)
623 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
624 * (including parity). On current CPUs they must be 0'd,
625 * and finally or in RW bit
630 /* do the actual scom read */
639 /* XXX: fixup result on some buggy 970's (ouch ! we lost a bit, bah
640 * that's the best we can do). Not implemented yet as we don't use
641 * the scom on any of the bogus CPUs yet, but may have to be done
645 /* restore interrupts */
650 _GLOBAL(scom970_write)
657 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
658 * (including parity). On current CPUs they must be 0'd.
664 mtspr SPRN_SCOMD,r4 /* write data */
666 mtspr SPRN_SCOMC,r3 /* write command */
671 /* restore interrupts */
677 * Create a kernel thread
678 * kernel_thread(fn, arg, flags)
680 _GLOBAL(kernel_thread)
683 stdu r1,-STACK_FRAME_OVERHEAD(r1)
686 ori r3,r5,CLONE_VM /* flags */
687 oris r3,r3,(CLONE_UNTRACED>>16)
688 li r4,0 /* new sp (unused) */
691 cmpdi 0,r3,0 /* parent or child? */
692 bne 1f /* return if parent */
694 stdu r0,-STACK_FRAME_OVERHEAD(r1)
697 mtlr r29 /* fn addr in lr */
698 mr r3,r30 /* load arg and call fn */
700 li r0,__NR_exit /* exit after child exits */
703 1: addi r1,r1,STACK_FRAME_OVERHEAD
709 * disable_kernel_fp()
712 _GLOBAL(disable_kernel_fp)
714 rldicl r0,r3,(63-MSR_FP_LG),1
715 rldicl r3,r0,(MSR_FP_LG+1),0
716 mtmsrd r3 /* disable use of fpu now */
720 #ifdef CONFIG_ALTIVEC
722 #if 0 /* this has no callers for now */
724 * disable_kernel_altivec()
727 _GLOBAL(disable_kernel_altivec)
729 rldicl r0,r3,(63-MSR_VEC_LG),1
730 rldicl r3,r0,(MSR_VEC_LG+1),0
731 mtmsrd r3 /* disable use of VMX now */
737 * giveup_altivec(tsk)
738 * Disable VMX for the task given as the argument,
739 * and save the vector registers in its thread_struct.
740 * Enables the VMX for use in the kernel on return.
742 _GLOBAL(giveup_altivec)
745 mtmsrd r5 /* enable use of VMX now */
748 beqlr- /* if no previous owner, done */
749 addi r3,r3,THREAD /* want THREAD of task */
757 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
759 andc r4,r4,r3 /* disable FP for previous task */
760 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
764 ld r4,last_task_used_altivec@got(r2)
766 #endif /* CONFIG_SMP */
769 #endif /* CONFIG_ALTIVEC */
771 _GLOBAL(__setup_cpu_power3)
781 /* kexec_wait(phys_cpu)
783 * wait for the flag to change, indicating this kernel is going away but
784 * the slave code for the next one is at addresses 0 to 100.
786 * This is used by all slaves.
788 * Physical (hardware) cpu id should be in r3.
793 addi r5,r5,kexec_flag-1b
796 #ifdef CONFIG_KEXEC /* use no memory without kexec */
803 /* this can be in text because we won't change it until we are
804 * running in real anyways
812 /* kexec_smp_wait(void)
814 * call with interrupts off
815 * note: this is a terminal routine, it does not save lr
817 * get phys id from paca
818 * set paca id to -1 to say we got here
819 * switch to real mode
820 * join other cpus in kexec_wait(phys_id)
822 _GLOBAL(kexec_smp_wait)
823 lhz r3,PACAHWCPUID(r13)
825 sth r4,PACAHWCPUID(r13) /* let others know we left */
830 * switch to real mode (turn mmu off)
831 * we use the early kernel trick that the hardware ignores bits
832 * 0 and 1 (big endian) of the effective address in real mode
834 * don't overwrite r3 here, it is live for kexec_wait above.
836 real_mode: /* assume normal blr return */
839 mflr r11 /* return address to SRR0 */
851 * kexec_sequence(newstack, start, image, control, clear_all())
853 * does the grungy work with stack switching and real mode switches
854 * also does simple calls to other code
857 _GLOBAL(kexec_sequence)
861 /* switch stacks to newstack -- &kexec_stack.stack */
862 stdu r1,THREAD_SIZE-112(r3)
868 /* save regs for local vars on new stack.
869 * yes, we won't go back, but ...
881 /* save args into preserved regs */
882 mr r31,r3 /* newstack (both) */
883 mr r30,r4 /* start (real) */
884 mr r29,r5 /* image (virt) */
885 mr r28,r6 /* control, unused */
886 mr r27,r7 /* clear_all() fn desc */
887 mr r26,r8 /* spare */
888 lhz r25,PACAHWCPUID(r13) /* get our phys cpu from paca */
890 /* disable interrupts, we are overwriting kernel data next */
895 /* copy dest pages, flush whole dest image */
897 bl .kexec_copy_flush /* (image) */
902 /* clear out hardware hash page table and tlb */
903 ld r5,0(r27) /* deref function descriptor */
905 bctrl /* ppc_md.hash_clear_all(void); */
908 * kexec image calling is:
909 * the first 0x100 bytes of the entry point are copied to 0
911 * all slaves branch to slave = 0x60 (absolute)
912 * slave(phys_cpu_id);
914 * master goes to start = entry point
915 * start(phys_cpu_id, start, 0);
918 * a wrapper is needed to call existing kernels, here is an approximate
919 * description of one method:
922 * start will be near the boot_block (maybe 0x100 bytes before it?)
923 * it will have a 0x60, which will b to boot_block, where it will wait
924 * and 0 will store phys into struct boot-block and load r3 from there,
925 * copy kernel 0-0x100 and tell slaves to back down to 0x60 again
928 * boot block will have all cpus scanning device tree to see if they
929 * are the boot cpu ?????
930 * other device tree differences (prop sizes, va vs pa, etc)...
933 /* copy 0x100 bytes starting at start to 0 */
938 bl .copy_and_flush /* (dest, src, copy limit, start offset) */
939 1: /* assume normal blr return */
941 /* release other cpus to the new kernel secondary start at 0x60 */
944 stw r6,kexec_flag-1b(5)
945 mr r3,r25 # my phys cpu
946 mr r4,r30 # start, aka phys mem offset
949 blr /* image->start(physid, image->start, 0); */
950 #endif /* CONFIG_KEXEC */