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>
30 #ifdef CONFIG_IRQSTACKS
31 _GLOBAL(call_do_softirq)
34 stdu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
42 _GLOBAL(call_handle_irq)
47 stdu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r5)
54 #endif /* CONFIG_IRQSTACKS */
58 .tc ppc64_caches[TC],ppc64_caches
62 * Write any modified data cache blocks out to memory
63 * and invalidate the corresponding instruction cache blocks.
65 * flush_icache_range(unsigned long start, unsigned long stop)
67 * flush all bytes from start through stop-1 inclusive
70 _KPROBE(__flush_icache_range)
73 * Flush the data cache to memory
75 * Different systems have different cache line sizes
76 * and in some cases i-cache and d-cache line sizes differ from
79 ld r10,PPC64_CACHES@toc(r2)
80 lwz r7,DCACHEL1LINESIZE(r10)/* Get cache line size */
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? */
94 /* Now invalidate the instruction cache */
96 lwz r7,ICACHEL1LINESIZE(r10) /* Get Icache line size */
98 andc r6,r3,r5 /* round low to line bdy */
99 subf r8,r6,r4 /* compute length */
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? */
112 * Like above, but only do the D-cache.
114 * flush_dcache_range(unsigned long start, unsigned long stop)
116 * flush all bytes from start to stop-1 inclusive
118 _GLOBAL(flush_dcache_range)
121 * Flush the data cache to memory
123 * Different systems have different cache line sizes
125 ld r10,PPC64_CACHES@toc(r2)
126 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
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? */
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
147 * flush_dcache_phys_range(unsigned long start, unsigned long stop)
149 * flush all bytes from start to stop-1 inclusive
151 _GLOBAL(flush_dcache_phys_range)
152 ld r10,PPC64_CACHES@toc(r2)
153 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
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 */
174 mtmsr r5 /* Re-enable MMU Data Relocation */
179 _GLOBAL(flush_inval_dcache_range)
180 ld r10,PPC64_CACHES@toc(r2)
181 lwz r7,DCACHEL1LINESIZE(r10) /* Get dcache line size */
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? */
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.
205 * void __flush_dcache_icache(void *page)
207 _GLOBAL(__flush_dcache_icache)
209 * Flush the data cache to memory
211 * Different systems have different cache line sizes
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 */
226 /* Now invalidate the icache */
228 lwz r4,ICACHEL1LINESPERPAGE(r7) /* Get # icache lines per page */
229 lwz r5,ICACHEL1LINESIZE(r7) /* Get icache line size */
238 #if defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE)
240 * Do an IO access in real mode
271 * Do an IO access in real mode
300 #endif /* defined(CONFIG_PPC_PMAC) || defined(CONFIG_PPC_MAPLE) */
302 #ifdef CONFIG_PPC_PASEMI
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))
309 _GLOBAL(real_205_readb)
324 _GLOBAL(real_205_writeb)
339 #endif /* CONFIG_PPC_PASEMI */
342 #ifdef CONFIG_CPU_FREQ_PMAC64
344 * SCOM access functions for 970 (FX only for now)
346 * unsigned long scom970_read(unsigned int address);
347 * void scom970_write(unsigned int address, unsigned long value);
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.
353 _GLOBAL(scom970_read)
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
367 /* do the actual scom read */
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
382 /* restore interrupts */
387 _GLOBAL(scom970_write)
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.
401 mtspr SPRN_SCOMD,r4 /* write data */
403 mtspr SPRN_SCOMC,r3 /* write command */
408 /* restore interrupts */
411 #endif /* CONFIG_CPU_FREQ_PMAC64 */
415 * Create a kernel thread
416 * kernel_thread(fn, arg, flags)
418 _GLOBAL(kernel_thread)
421 stdu r1,-STACK_FRAME_OVERHEAD(r1)
424 ori r3,r5,CLONE_VM /* flags */
425 oris r3,r3,(CLONE_UNTRACED>>16)
426 li r4,0 /* new sp (unused) */
429 cmpdi 0,r3,0 /* parent or child? */
430 bne 1f /* return if parent */
432 stdu r0,-STACK_FRAME_OVERHEAD(r1)
435 mtlr r29 /* fn addr in lr */
436 mr r3,r30 /* load arg and call fn */
438 li r0,__NR_exit /* exit after child exits */
441 1: addi r1,r1,STACK_FRAME_OVERHEAD
447 * disable_kernel_fp()
450 _GLOBAL(disable_kernel_fp)
452 rldicl r0,r3,(63-MSR_FP_LG),1
453 rldicl r3,r0,(MSR_FP_LG+1),0
454 mtmsrd r3 /* disable use of fpu now */
458 #ifdef CONFIG_ALTIVEC
460 #if 0 /* this has no callers for now */
462 * disable_kernel_altivec()
465 _GLOBAL(disable_kernel_altivec)
467 rldicl r0,r3,(63-MSR_VEC_LG),1
468 rldicl r3,r0,(MSR_VEC_LG+1),0
469 mtmsrd r3 /* disable use of VMX now */
475 * giveup_altivec(tsk)
476 * Disable VMX for the task given as the argument,
477 * and save the vector registers in its thread_struct.
478 * Enables the VMX for use in the kernel on return.
480 _GLOBAL(giveup_altivec)
483 mtmsrd r5 /* enable use of VMX now */
486 beqlr- /* if no previous owner, done */
487 addi r3,r3,THREAD /* want THREAD of task */
495 ld r4,_MSR-STACK_FRAME_OVERHEAD(r5)
497 andc r4,r4,r3 /* disable FP for previous task */
498 std r4,_MSR-STACK_FRAME_OVERHEAD(r5)
502 ld r4,last_task_used_altivec@got(r2)
504 #endif /* CONFIG_SMP */
507 #endif /* CONFIG_ALTIVEC */
509 /* kexec_wait(phys_cpu)
511 * wait for the flag to change, indicating this kernel is going away but
512 * the slave code for the next one is at addresses 0 to 100.
514 * This is used by all slaves.
516 * Physical (hardware) cpu id should be in r3.
521 addi r5,r5,kexec_flag-1b
524 #ifdef CONFIG_KEXEC /* use no memory without kexec */
531 /* this can be in text because we won't change it until we are
532 * running in real anyways
540 /* kexec_smp_wait(void)
542 * call with interrupts off
543 * note: this is a terminal routine, it does not save lr
545 * get phys id from paca
546 * set paca id to -1 to say we got here
547 * switch to real mode
548 * join other cpus in kexec_wait(phys_id)
550 _GLOBAL(kexec_smp_wait)
551 lhz r3,PACAHWCPUID(r13)
553 sth r4,PACAHWCPUID(r13) /* let others know we left */
558 * switch to real mode (turn mmu off)
559 * we use the early kernel trick that the hardware ignores bits
560 * 0 and 1 (big endian) of the effective address in real mode
562 * don't overwrite r3 here, it is live for kexec_wait above.
564 real_mode: /* assume normal blr return */
567 mflr r11 /* return address to SRR0 */
579 * kexec_sequence(newstack, start, image, control, clear_all())
581 * does the grungy work with stack switching and real mode switches
582 * also does simple calls to other code
585 _GLOBAL(kexec_sequence)
589 /* switch stacks to newstack -- &kexec_stack.stack */
590 stdu r1,THREAD_SIZE-STACK_FRAME_OVERHEAD(r3)
596 /* save regs for local vars on new stack.
597 * yes, we won't go back, but ...
607 stdu r1,-STACK_FRAME_OVERHEAD-64(r1)
609 /* save args into preserved regs */
610 mr r31,r3 /* newstack (both) */
611 mr r30,r4 /* start (real) */
612 mr r29,r5 /* image (virt) */
613 mr r28,r6 /* control, unused */
614 mr r27,r7 /* clear_all() fn desc */
615 mr r26,r8 /* spare */
616 lhz r25,PACAHWCPUID(r13) /* get our phys cpu from paca */
618 /* disable interrupts, we are overwriting kernel data next */
623 /* copy dest pages, flush whole dest image */
625 bl .kexec_copy_flush /* (image) */
630 /* copy 0x100 bytes starting at start to 0 */
632 mr r4,r30 /* start, aka phys mem offset */
635 bl .copy_and_flush /* (dest, src, copy limit, start offset) */
636 1: /* assume normal blr return */
638 /* release other cpus to the new kernel secondary start at 0x60 */
641 stw r6,kexec_flag-1b(5)
643 /* clear out hardware hash page table and tlb */
644 ld r5,0(r27) /* deref function descriptor */
646 bctrl /* ppc_md.hpte_clear_all(void); */
649 * kexec image calling is:
650 * the first 0x100 bytes of the entry point are copied to 0
652 * all slaves branch to slave = 0x60 (absolute)
653 * slave(phys_cpu_id);
655 * master goes to start = entry point
656 * start(phys_cpu_id, start, 0);
659 * a wrapper is needed to call existing kernels, here is an approximate
660 * description of one method:
663 * start will be near the boot_block (maybe 0x100 bytes before it?)
664 * it will have a 0x60, which will b to boot_block, where it will wait
665 * and 0 will store phys into struct boot-block and load r3 from there,
666 * copy kernel 0-0x100 and tell slaves to back down to 0x60 again
669 * boot block will have all cpus scanning device tree to see if they
670 * are the boot cpu ?????
671 * other device tree differences (prop sizes, va vs pa, etc)...
673 mr r3,r25 # my phys cpu
674 mr r4,r30 # start, aka phys mem offset
677 blr /* image->start(physid, image->start, 0); */
678 #endif /* CONFIG_KEXEC */