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)
7 * Adapted for iSeries by Mike Corrigan (mikejc@us.ibm.com)
8 * PPC64 updates by Dave Engebretsen (engebret@us.ibm.com)
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.
17 #include <linux/sys.h>
18 #include <asm/unistd.h>
19 #include <asm/errno.h>
20 #include <asm/processor.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>
42 #ifdef CONFIG_IRQSTACKS
43 _GLOBAL(call_do_softirq)
46 stdu r1,THREAD_SIZE-112(r3)
54 _GLOBAL(call_handle_irq)
59 stdu r1,THREAD_SIZE-112(r5)
66 #endif /* CONFIG_IRQSTACKS */
70 .tc ppc64_caches[TC],ppc64_caches
74 * Write any modified data cache blocks out to memory
75 * and invalidate the corresponding instruction cache blocks.
77 * flush_icache_range(unsigned long start, unsigned long stop)
79 * flush all bytes from start through stop-1 inclusive
82 _KPROBE(__flush_icache_range)
85 * Flush the data cache to memory
87 * Different systems have different cache line sizes
88 * and in some cases i-cache and d-cache line sizes differ from
91 ld r10,PPC64_CACHES@toc(r2)
92 lwz r7,DCACHEL1LINESIZE(r10)/* Get cache line size */
94 andc r6,r3,r5 /* round low to line bdy */
95 subf r8,r6,r4 /* compute length */
96 add r8,r8,r5 /* ensure we get enough */
97 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of cache line size */
98 srw. r8,r8,r9 /* compute line count */
99 beqlr /* nothing to do? */
106 /* Now invalidate the instruction cache */
108 lwz r7,ICACHEL1LINESIZE(r10) /* Get Icache line size */
110 andc r6,r3,r5 /* round low to line bdy */
111 subf r8,r6,r4 /* compute length */
113 lwz r9,ICACHEL1LOGLINESIZE(r10) /* Get log-2 of Icache line size */
114 srw. r8,r8,r9 /* compute line count */
115 beqlr /* nothing to do? */
124 * Like above, but only do the D-cache.
126 * flush_dcache_range(unsigned long start, unsigned long stop)
128 * flush all bytes from start to stop-1 inclusive
130 _GLOBAL(flush_dcache_range)
133 * Flush the data cache to memory
135 * Different systems have different cache line sizes
137 ld r10,PPC64_CACHES@toc(r2)
138 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
140 andc r6,r3,r5 /* round low to line bdy */
141 subf r8,r6,r4 /* compute length */
142 add r8,r8,r5 /* ensure we get enough */
143 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */
144 srw. r8,r8,r9 /* compute line count */
145 beqlr /* nothing to do? */
154 * Like above, but works on non-mapped physical addresses.
155 * Use only for non-LPAR setups ! It also assumes real mode
156 * is cacheable. Used for flushing out the DART before using
157 * it as uncacheable memory
159 * flush_dcache_phys_range(unsigned long start, unsigned long stop)
161 * flush all bytes from start to stop-1 inclusive
163 _GLOBAL(flush_dcache_phys_range)
164 ld r10,PPC64_CACHES@toc(r2)
165 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
167 andc r6,r3,r5 /* round low to line bdy */
168 subf r8,r6,r4 /* compute length */
169 add r8,r8,r5 /* ensure we get enough */
170 lwz r9,DCACHEL1LOGLINESIZE(r10) /* Get log-2 of dcache line size */
171 srw. r8,r8,r9 /* compute line count */
172 beqlr /* nothing to do? */
173 mfmsr r5 /* Disable MMU Data Relocation */
186 mtmsr r5 /* Re-enable MMU Data Relocation */
191 _GLOBAL(flush_inval_dcache_range)
192 ld r10,PPC64_CACHES@toc(r2)
193 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
195 andc r6,r3,r5 /* round low to line bdy */
196 subf r8,r6,r4 /* compute length */
197 add r8,r8,r5 /* ensure we get enough */
198 lwz r9,DCACHEL1LOGLINESIZE(r10)/* Get log-2 of dcache line size */
199 srw. r8,r8,r9 /* compute line count */
200 beqlr /* nothing to do? */
213 * Flush a particular page from the data cache to RAM.
214 * Note: this is necessary because the instruction cache does *not*
215 * snoop from the data cache.
217 * void __flush_dcache_icache(void *page)
219 _GLOBAL(__flush_dcache_icache)
221 * Flush the data cache to memory
223 * Different systems have different cache line sizes
226 /* Flush the dcache */
227 ld r7,PPC64_CACHES@toc(r2)
228 clrrdi r3,r3,PAGE_SHIFT /* Page align */
229 lwz r4,DCACHEL1LINESPERPAGE(r7) /* Get # dcache lines per page */
230 lwz r5,DCACHEL1LINESIZE(r7) /* Get dcache line size */
238 /* Now invalidate the icache */
240 lwz r4,ICACHEL1LINESPERPAGE(r7) /* Get # icache lines per page */
241 lwz r5,ICACHEL1LINESIZE(r7) /* Get icache line size */
250 * identify_cpu and calls setup_cpu
251 * In: r3 = base of the cpu_specs array
252 * r4 = address of cur_cpu_spec
253 * r5 = relocation offset
255 _GLOBAL(identify_cpu)
258 lwz r8,CPU_SPEC_PVR_MASK(r3)
260 lwz r9,CPU_SPEC_PVR_VALUE(r3)
263 addi r3,r3,CPU_SPEC_ENTRY_SIZE
268 ld r4,CPU_SPEC_SETUP(r3)
275 /* Calling convention for cpu setup is r3=offset, r4=cur_cpu_spec */
281 * do_cpu_ftr_fixups - goes through the list of CPU feature fixups
282 * and writes nop's over sections of code that don't apply for this cpu.
283 * r3 = data offset (not changed)
285 _GLOBAL(do_cpu_ftr_fixups)
286 /* Get CPU 0 features */
287 LOAD_REG_IMMEDIATE(r6,cur_cpu_spec)
291 ld r4,CPU_SPEC_FEATURES(r4)
292 /* Get the fixup table */
293 LOAD_REG_IMMEDIATE(r6,__start___ftr_fixup)
295 LOAD_REG_IMMEDIATE(r7,__stop___ftr_fixup)
301 ld r8,-32(r6) /* mask */
303 ld r9,-24(r6) /* value */
306 ld r8,-16(r6) /* section begin */
307 ld r9,-8(r6) /* section end */
310 /* write nops over the section of code */
311 /* todo: if large section, add a branch at the start of it */
315 lis r0,0x60000000@h /* nop */
317 andi. r10,r4,CPU_FTR_SPLIT_ID_CACHE@l
319 dcbst 0,r8 /* suboptimal, but simpler */
324 sync /* additional sync needed on g4 */
329 * do_fw_ftr_fixups - goes through the list of firmware feature fixups
330 * and writes nop's over sections of code that don't apply for this firmware.
331 * r3 = data offset (not changed)
333 _GLOBAL(do_fw_ftr_fixups)
334 /* Get firmware features */
335 LOAD_REG_IMMEDIATE(r6,powerpc_firmware_features)
338 /* Get the fixup table */
339 LOAD_REG_IMMEDIATE(r6,__start___fw_ftr_fixup)
341 LOAD_REG_IMMEDIATE(r7,__stop___fw_ftr_fixup)
347 ld r8,-32(r6) /* mask */
349 ld r9,-24(r6) /* value */
352 ld r8,-16(r6) /* section begin */
353 ld r9,-8(r6) /* section end */
356 /* write nops over the section of code */
357 /* todo: if large section, add a branch at the start of it */
361 lis r0,0x60000000@h /* nop */
364 dcbst 0,r8 /* suboptimal, but simpler */
367 END_FTR_SECTION_IFSET(CPU_FTR_SPLIT_ID_CACHE)
370 sync /* additional sync needed on g4 */
374 #if defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE)
376 * Do an IO access in real mode
407 * Do an IO access in real mode
436 #endif /* defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) */
438 #ifdef CONFIG_CPU_FREQ_PMAC64
440 * SCOM access functions for 970 (FX only for now)
442 * unsigned long scom970_read(unsigned int address);
443 * void scom970_write(unsigned int address, unsigned long value);
445 * The address passed in is the 24 bits register address. This code
446 * is 970 specific and will not check the status bits, so you should
447 * know what you are doing.
449 _GLOBAL(scom970_read)
456 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
457 * (including parity). On current CPUs they must be 0'd,
458 * and finally or in RW bit
463 /* do the actual scom read */
472 /* XXX: fixup result on some buggy 970's (ouch ! we lost a bit, bah
473 * that's the best we can do). Not implemented yet as we don't use
474 * the scom on any of the bogus CPUs yet, but may have to be done
478 /* restore interrupts */
483 _GLOBAL(scom970_write)
490 /* rotate 24 bits SCOM address 8 bits left and mask out it's low 8 bits
491 * (including parity). On current CPUs they must be 0'd.
497 mtspr SPRN_SCOMD,r4 /* write data */
499 mtspr SPRN_SCOMC,r3 /* write command */
504 /* restore interrupts */
507 #endif /* CONFIG_CPU_FREQ_PMAC64 */
511 * Create a kernel thread
512 * kernel_thread(fn, arg, flags)
514 _GLOBAL(kernel_thread)
517 stdu r1,-STACK_FRAME_OVERHEAD(r1)
520 ori r3,r5,CLONE_VM /* flags */
521 oris r3,r3,(CLONE_UNTRACED>>16)
522 li r4,0 /* new sp (unused) */
525 cmpdi 0,r3,0 /* parent or child? */
526 bne 1f /* return if parent */
528 stdu r0,-STACK_FRAME_OVERHEAD(r1)
531 mtlr r29 /* fn addr in lr */
532 mr r3,r30 /* load arg and call fn */
534 li r0,__NR_exit /* exit after child exits */
537 1: addi r1,r1,STACK_FRAME_OVERHEAD
543 * disable_kernel_fp()
546 _GLOBAL(disable_kernel_fp)
548 rldicl r0,r3,(63-MSR_FP_LG),1
549 rldicl r3,r0,(MSR_FP_LG+1),0
550 mtmsrd r3 /* disable use of fpu now */
554 #ifdef CONFIG_ALTIVEC
556 #if 0 /* this has no callers for now */
558 * disable_kernel_altivec()
561 _GLOBAL(disable_kernel_altivec)
563 rldicl r0,r3,(63-MSR_VEC_LG),1
564 rldicl r3,r0,(MSR_VEC_LG+1),0
565 mtmsrd r3 /* disable use of VMX now */
571 * giveup_altivec(tsk)
572 * Disable VMX for the task given as the argument,
573 * and save the vector registers in its thread_struct.
574 * Enables the VMX for use in the kernel on return.
576 _GLOBAL(giveup_altivec)
579 mtmsrd r5 /* enable use of VMX now */
582 beqlr- /* if no previous owner, done */
583 addi r3,r3,THREAD /* want THREAD of task */
591 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
593 andc r4,r4,r3 /* disable FP for previous task */
594 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
598 ld r4,last_task_used_altivec@got(r2)
600 #endif /* CONFIG_SMP */
603 #endif /* CONFIG_ALTIVEC */
605 _GLOBAL(kernel_execve)
612 /* kexec_wait(phys_cpu)
614 * wait for the flag to change, indicating this kernel is going away but
615 * the slave code for the next one is at addresses 0 to 100.
617 * This is used by all slaves.
619 * Physical (hardware) cpu id should be in r3.
624 addi r5,r5,kexec_flag-1b
627 #ifdef CONFIG_KEXEC /* use no memory without kexec */
634 /* this can be in text because we won't change it until we are
635 * running in real anyways
643 /* kexec_smp_wait(void)
645 * call with interrupts off
646 * note: this is a terminal routine, it does not save lr
648 * get phys id from paca
649 * set paca id to -1 to say we got here
650 * switch to real mode
651 * join other cpus in kexec_wait(phys_id)
653 _GLOBAL(kexec_smp_wait)
654 lhz r3,PACAHWCPUID(r13)
656 sth r4,PACAHWCPUID(r13) /* let others know we left */
661 * switch to real mode (turn mmu off)
662 * we use the early kernel trick that the hardware ignores bits
663 * 0 and 1 (big endian) of the effective address in real mode
665 * don't overwrite r3 here, it is live for kexec_wait above.
667 real_mode: /* assume normal blr return */
670 mflr r11 /* return address to SRR0 */
682 * kexec_sequence(newstack, start, image, control, clear_all())
684 * does the grungy work with stack switching and real mode switches
685 * also does simple calls to other code
688 _GLOBAL(kexec_sequence)
692 /* switch stacks to newstack -- &kexec_stack.stack */
693 stdu r1,THREAD_SIZE-112(r3)
699 /* save regs for local vars on new stack.
700 * yes, we won't go back, but ...
712 /* save args into preserved regs */
713 mr r31,r3 /* newstack (both) */
714 mr r30,r4 /* start (real) */
715 mr r29,r5 /* image (virt) */
716 mr r28,r6 /* control, unused */
717 mr r27,r7 /* clear_all() fn desc */
718 mr r26,r8 /* spare */
719 lhz r25,PACAHWCPUID(r13) /* get our phys cpu from paca */
721 /* disable interrupts, we are overwriting kernel data next */
726 /* copy dest pages, flush whole dest image */
728 bl .kexec_copy_flush /* (image) */
733 /* clear out hardware hash page table and tlb */
734 ld r5,0(r27) /* deref function descriptor */
736 bctrl /* ppc_md.hpte_clear_all(void); */
739 * kexec image calling is:
740 * the first 0x100 bytes of the entry point are copied to 0
742 * all slaves branch to slave = 0x60 (absolute)
743 * slave(phys_cpu_id);
745 * master goes to start = entry point
746 * start(phys_cpu_id, start, 0);
749 * a wrapper is needed to call existing kernels, here is an approximate
750 * description of one method:
753 * start will be near the boot_block (maybe 0x100 bytes before it?)
754 * it will have a 0x60, which will b to boot_block, where it will wait
755 * and 0 will store phys into struct boot-block and load r3 from there,
756 * copy kernel 0-0x100 and tell slaves to back down to 0x60 again
759 * boot block will have all cpus scanning device tree to see if they
760 * are the boot cpu ?????
761 * other device tree differences (prop sizes, va vs pa, etc)...
764 /* copy 0x100 bytes starting at start to 0 */
769 bl .copy_and_flush /* (dest, src, copy limit, start offset) */
770 1: /* assume normal blr return */
772 /* release other cpus to the new kernel secondary start at 0x60 */
775 stw r6,kexec_flag-1b(5)
776 mr r3,r25 # my phys cpu
777 mr r4,r30 # start, aka phys mem offset
780 blr /* image->start(physid, image->start, 0); */
781 #endif /* CONFIG_KEXEC */