3 * Copyright (C) 1995-1996 Gary Thomas (gdt@linuxppc.org)
5 * Rewritten by Cort Dougan (cort@cs.nmt.edu) for PReP
6 * Copyright (C) 1996 Cort Dougan <cort@cs.nmt.edu>
7 * Adapted for Power Macintosh by Paul Mackerras.
8 * Low-level exception handlers and MMU support
9 * rewritten by Paul Mackerras.
10 * Copyright (C) 1996 Paul Mackerras.
12 * Adapted for 64bit PowerPC by Dave Engebretsen, Peter Bergner, and
13 * Mike Corrigan {engebret|bergner|mikejc}@us.ibm.com
15 * This file contains the entry point for the 64-bit kernel along
16 * with some early initialization code common to all 64-bit powerpc
19 * This program is free software; you can redistribute it and/or
20 * modify it under the terms of the GNU General Public License
21 * as published by the Free Software Foundation; either version
22 * 2 of the License, or (at your option) any later version.
25 #include <linux/threads.h>
29 #include <asm/ppc_asm.h>
30 #include <asm/asm-offsets.h>
32 #include <asm/cputable.h>
33 #include <asm/setup.h>
34 #include <asm/hvcall.h>
35 #include <asm/thread_info.h>
36 #include <asm/firmware.h>
37 #include <asm/page_64.h>
38 #include <asm/irqflags.h>
39 #include <asm/kvm_book3s_asm.h>
40 #include <asm/ptrace.h>
41 #include <asm/hw_irq.h>
43 /* The physical memory is laid out such that the secondary processor
44 * spin code sits at 0x0000...0x00ff. On server, the vectors follow
45 * using the layout described in exceptions-64s.S
49 * Entering into this code we make the following assumptions:
51 * For pSeries or server processors:
52 * 1. The MMU is off & open firmware is running in real mode.
53 * 2. The kernel is entered at __start
54 * -or- For OPAL entry:
55 * 1. The MMU is off, processor in HV mode, primary CPU enters at 0
56 * with device-tree in gpr3. We also get OPAL base in r8 and
57 * entry in r9 for debugging purposes
58 * 2. Secondary processors enter at 0x60 with PIR in gpr3
60 * For Book3E processors:
61 * 1. The MMU is on running in AS0 in a state defined in ePAPR
62 * 2. The kernel is entered at __start
69 /* NOP this out unconditionally */
71 b .__start_initialization_multiplatform
74 /* Catch branch to 0 in real mode */
77 /* Secondary processors spin on this value until it becomes nonzero.
78 * When it does it contains the real address of the descriptor
79 * of the function that the cpu should jump to to continue
82 .globl __secondary_hold_spinloop
83 __secondary_hold_spinloop:
86 /* Secondary processors write this value with their cpu # */
87 /* after they enter the spin loop immediately below. */
88 .globl __secondary_hold_acknowledge
89 __secondary_hold_acknowledge:
92 #ifdef CONFIG_RELOCATABLE
93 /* This flag is set to 1 by a loader if the kernel should run
94 * at the loaded address instead of the linked address. This
95 * is used by kexec-tools to keep the the kdump kernel in the
96 * crash_kernel region. The loader is responsible for
97 * observing the alignment requirement.
99 /* Do not move this variable as kexec-tools knows about it. */
103 .long 0x72756e30 /* "run0" -- relocate to 0 by default */
108 * The following code is used to hold secondary processors
109 * in a spin loop after they have entered the kernel, but
110 * before the bulk of the kernel has been relocated. This code
111 * is relocated to physical address 0x60 before prom_init is run.
112 * All of it must fit below the first exception vector at 0x100.
113 * Use .globl here not _GLOBAL because we want __secondary_hold
114 * to be the actual text address, not a descriptor.
116 .globl __secondary_hold
118 #ifndef CONFIG_PPC_BOOK3E
121 mtmsrd r24 /* RI on */
123 /* Grab our physical cpu number */
126 /* Tell the master cpu we're here */
127 /* Relocation is off & we are located at an address less */
128 /* than 0x100, so only need to grab low order offset. */
129 std r24,__secondary_hold_acknowledge-_stext(0)
132 /* All secondary cpus wait here until told to start. */
133 100: ld r4,__secondary_hold_spinloop-_stext(0)
137 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
138 ld r4,0(r4) /* deref function descriptor */
142 /* Make sure that patched code is visible */
149 /* This value is used to mark exception frames on the stack. */
152 .tc ID_72656773_68657265[TC],0x7265677368657265
156 * On server, we include the exception vectors code here as it
157 * relies on absolute addressing which is only possible within
158 * this compilation unit
160 #ifdef CONFIG_PPC_BOOK3S
161 #include "exceptions-64s.S"
164 _GLOBAL(generic_secondary_thread_init)
167 /* turn on 64-bit mode */
170 /* get a valid TOC pointer, wherever we're mapped at */
173 #ifdef CONFIG_PPC_BOOK3E
174 /* Book3E initialization */
176 bl .book3e_secondary_thread_init
178 b generic_secondary_common_init
181 * On pSeries and most other platforms, secondary processors spin
182 * in the following code.
183 * At entry, r3 = this processor's number (physical cpu id)
185 * On Book3E, r4 = 1 to indicate that the initial TLB entry for
186 * this core already exists (setup via some other mechanism such
187 * as SCOM before entry).
189 _GLOBAL(generic_secondary_smp_init)
193 /* turn on 64-bit mode */
196 /* get a valid TOC pointer, wherever we're mapped at */
199 #ifdef CONFIG_PPC_BOOK3E
200 /* Book3E initialization */
203 bl .book3e_secondary_core_init
206 generic_secondary_common_init:
207 /* Set up a paca value for this processor. Since we have the
208 * physical cpu id in r24, we need to search the pacas to find
209 * which logical id maps to our physical one.
211 LOAD_REG_ADDR(r13, paca) /* Load paca pointer */
212 ld r13,0(r13) /* Get base vaddr of paca array */
214 addi r13,r13,PACA_SIZE /* know r13 if used accidentally */
215 b .kexec_wait /* wait for next kernel if !SMP */
217 LOAD_REG_ADDR(r7, nr_cpu_ids) /* Load nr_cpu_ids address */
218 lwz r7,0(r7) /* also the max paca allocated */
219 li r5,0 /* logical cpu id */
220 1: lhz r6,PACAHWCPUID(r13) /* Load HW procid from paca */
221 cmpw r6,r24 /* Compare to our id */
223 addi r13,r13,PACA_SIZE /* Loop to next PACA on miss */
225 cmpw r5,r7 /* Check if more pacas exist */
228 mr r3,r24 /* not found, copy phys to r3 */
229 b .kexec_wait /* next kernel might do better */
232 #ifdef CONFIG_PPC_BOOK3E
233 addi r12,r13,PACA_EXTLB /* and TLB exc frame in another */
234 mtspr SPRN_SPRG_TLB_EXFRAME,r12
237 /* From now on, r24 is expected to be logical cpuid */
240 /* See if we need to call a cpu state restore handler */
241 LOAD_REG_ADDR(r23, cur_cpu_spec)
243 ld r23,CPU_SPEC_RESTORE(r23)
250 3: LOAD_REG_ADDR(r3, spinning_secondaries) /* Decrement spinning_secondaries */
258 lbz r23,PACAPROCSTART(r13) /* Test if this processor should */
261 beq 4b /* Loop until told to go */
263 sync /* order paca.run and cur_cpu_spec */
264 isync /* In case code patching happened */
266 /* Create a temp kernel stack for use before relocation is on. */
267 ld r1,PACAEMERGSP(r13)
268 subi r1,r1,STACK_FRAME_OVERHEAD
275 * Assumes we're mapped EA == RA if the MMU is on.
277 #ifdef CONFIG_PPC_BOOK3S
280 andi. r0,r3,MSR_IR|MSR_DR
288 b . /* prevent speculative execution */
293 * Here is our main kernel entry point. We support currently 2 kind of entries
294 * depending on the value of r5.
296 * r5 != NULL -> OF entry, we go to prom_init, "legacy" parameter content
299 * r5 == NULL -> kexec style entry. r3 is a physical pointer to the
300 * DT block, r4 is a physical pointer to the kernel itself
303 _GLOBAL(__start_initialization_multiplatform)
304 /* Make sure we are running in 64 bits mode */
307 /* Get TOC pointer (current runtime address) */
310 /* find out where we are now */
312 0: mflr r26 /* r26 = runtime addr here */
313 addis r26,r26,(_stext - 0b)@ha
314 addi r26,r26,(_stext - 0b)@l /* current runtime base addr */
317 * Are we booted from a PROM Of-type client-interface ?
321 b .__boot_from_prom /* yes -> prom */
323 /* Save parameters */
326 #ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
327 /* Save OPAL entry */
332 #ifdef CONFIG_PPC_BOOK3E
333 bl .start_initialization_book3e
334 b .__after_prom_start
336 /* Setup some critical 970 SPRs before switching MMU off */
339 cmpwi r0,0x39 /* 970 */
341 cmpwi r0,0x3c /* 970FX */
343 cmpwi r0,0x44 /* 970MP */
345 cmpwi r0,0x45 /* 970GX */
347 1: bl .__cpu_preinit_ppc970
350 /* Switch off MMU if not already off */
352 b .__after_prom_start
353 #endif /* CONFIG_PPC_BOOK3E */
355 _INIT_STATIC(__boot_from_prom)
356 #ifdef CONFIG_PPC_OF_BOOT_TRAMPOLINE
357 /* Save parameters */
365 * Align the stack to 16-byte boundary
366 * Depending on the size and layout of the ELF sections in the initial
367 * boot binary, the stack pointer may be unaligned on PowerMac
371 #ifdef CONFIG_RELOCATABLE
372 /* Relocate code for where we are now */
377 /* Restore parameters */
384 /* Do all of the interaction with OF client interface */
387 #endif /* #CONFIG_PPC_OF_BOOT_TRAMPOLINE */
389 /* We never return. We also hit that trap if trying to boot
390 * from OF while CONFIG_PPC_OF_BOOT_TRAMPOLINE isn't selected */
393 _STATIC(__after_prom_start)
394 #ifdef CONFIG_RELOCATABLE
395 /* process relocations for the final address of the kernel */
396 lis r25,PAGE_OFFSET@highest /* compute virtual base of kernel */
398 lwz r7,__run_at_load-_stext(r26)
399 cmplwi cr0,r7,1 /* flagged to stay where we are ? */
407 * We need to run with _stext at physical address PHYSICAL_START.
408 * This will leave some code in the first 256B of
409 * real memory, which are reserved for software use.
411 * Note: This process overwrites the OF exception vectors.
413 li r3,0 /* target addr */
414 #ifdef CONFIG_PPC_BOOK3E
415 tovirt(r3,r3) /* on booke, we already run at PAGE_OFFSET */
417 mr. r4,r26 /* In some cases the loader may */
418 beq 9f /* have already put us at zero */
419 li r6,0x100 /* Start offset, the first 0x100 */
420 /* bytes were copied earlier. */
421 #ifdef CONFIG_PPC_BOOK3E
422 tovirt(r6,r6) /* on booke, we already run at PAGE_OFFSET */
425 #ifdef CONFIG_CRASH_DUMP
427 * Check if the kernel has to be running as relocatable kernel based on the
428 * variable __run_at_load, if it is set the kernel is treated as relocatable
429 * kernel, otherwise it will be moved to PHYSICAL_START
431 lwz r7,__run_at_load-_stext(r26)
435 li r5,__end_interrupts - _stext /* just copy interrupts */
439 lis r5,(copy_to_here - _stext)@ha
440 addi r5,r5,(copy_to_here - _stext)@l /* # bytes of memory to copy */
442 bl .copy_and_flush /* copy the first n bytes */
443 /* this includes the code being */
445 addis r8,r3,(4f - _stext)@ha /* Jump to the copy of this code */
446 addi r8,r8,(4f - _stext)@l /* that we just made */
450 p_end: .llong _end - _stext
452 4: /* Now copy the rest of the kernel up to _end */
453 addis r5,r26,(p_end - _stext)@ha
454 ld r5,(p_end - _stext)@l(r5) /* get _end */
455 5: bl .copy_and_flush /* copy the rest */
457 9: b .start_here_multiplatform
460 * Copy routine used to copy the kernel to start at physical address 0
461 * and flush and invalidate the caches as needed.
462 * r3 = dest addr, r4 = source addr, r5 = copy limit, r6 = start offset
463 * on exit, r3, r4, r5 are unchanged, r6 is updated to be >= r5.
465 * Note: this routine *only* clobbers r0, r6 and lr
467 _GLOBAL(copy_and_flush)
470 4: li r0,8 /* Use the smallest common */
471 /* denominator cache line */
472 /* size. This results in */
473 /* extra cache line flushes */
474 /* but operation is correct. */
475 /* Can't get cache line size */
476 /* from NACA as it is being */
479 mtctr r0 /* put # words/line in ctr */
480 3: addi r6,r6,8 /* copy a cache line */
484 dcbst r6,r3 /* write it to memory */
486 icbi r6,r3 /* flush the icache line */
498 #ifdef CONFIG_PPC_PMAC
500 * On PowerMac, secondary processors starts from the reset vector, which
501 * is temporarily turned into a call to one of the functions below.
506 .globl __secondary_start_pmac_0
507 __secondary_start_pmac_0:
508 /* NB the entries for cpus 0, 1, 2 must each occupy 8 bytes. */
518 _GLOBAL(pmac_secondary_start)
519 /* turn on 64-bit mode */
524 rldimi r3,r0,40,23 /* clear bit 23 (rm_ci) */
531 /* get TOC pointer (real address) */
534 /* Copy some CPU settings from CPU 0 */
535 bl .__restore_cpu_ppc970
537 /* pSeries do that early though I don't think we really need it */
540 mtmsrd r3 /* RI on */
542 /* Set up a paca value for this processor. */
543 LOAD_REG_ADDR(r4,paca) /* Load paca pointer */
544 ld r4,0(r4) /* Get base vaddr of paca array */
545 mulli r13,r24,PACA_SIZE /* Calculate vaddr of right paca */
546 add r13,r13,r4 /* for this processor. */
547 SET_PACA(r13) /* Save vaddr of paca in an SPRG*/
549 /* Mark interrupts soft and hard disabled (they might be enabled
550 * in the PACA when doing hotplug)
553 stb r0,PACASOFTIRQEN(r13)
554 li r0,PACA_IRQ_HARD_DIS
555 stb r0,PACAIRQHAPPENED(r13)
557 /* Create a temp kernel stack for use before relocation is on. */
558 ld r1,PACAEMERGSP(r13)
559 subi r1,r1,STACK_FRAME_OVERHEAD
563 #endif /* CONFIG_PPC_PMAC */
566 * This function is called after the master CPU has released the
567 * secondary processors. The execution environment is relocation off.
568 * The paca for this processor has the following fields initialized at
570 * 1. Processor number
571 * 2. Segment table pointer (virtual address)
572 * On entry the following are set:
573 * r1 = stack pointer (real addr of temp stack)
574 * r24 = cpu# (in Linux terms)
575 * r13 = paca virtual address
576 * SPRG_PACA = paca virtual address
581 .globl __secondary_start
583 /* Set thread priority to MEDIUM */
586 /* Initialize the kernel stack */
587 LOAD_REG_ADDR(r3, current_set)
588 sldi r28,r24,3 /* get current_set[cpu#] */
590 addi r14,r14,THREAD_SIZE-STACK_FRAME_OVERHEAD
591 std r14,PACAKSAVE(r13)
593 /* Do early setup for that CPU (stab, slb, hash table pointer) */
594 bl .early_setup_secondary
597 * setup the new stack pointer, but *don't* use this until
602 /* Clear backchain so we get nice backtraces */
606 /* Mark interrupts soft and hard disabled (they might be enabled
607 * in the PACA when doing hotplug)
609 stb r7,PACASOFTIRQEN(r13)
610 li r0,PACA_IRQ_HARD_DIS
611 stb r0,PACAIRQHAPPENED(r13)
613 /* enable MMU and jump to start_secondary */
614 LOAD_REG_ADDR(r3, .start_secondary_prolog)
615 LOAD_REG_IMMEDIATE(r4, MSR_KERNEL)
620 b . /* prevent speculative execution */
623 * Running with relocation on at this point. All we want to do is
624 * zero the stack back-chain pointer and get the TOC virtual address
625 * before going into C code.
627 _GLOBAL(start_secondary_prolog)
630 std r3,0(r1) /* Zero the stack frame pointer */
634 * Reset stack pointer and call start_secondary
635 * to continue with online operation when woken up
636 * from cede in cpu offline.
638 _GLOBAL(start_secondary_resume)
639 ld r1,PACAKSAVE(r13) /* Reload kernel stack pointer */
641 std r3,0(r1) /* Zero the stack frame pointer */
647 * This subroutine clobbers r11 and r12
649 _GLOBAL(enable_64b_mode)
650 mfmsr r11 /* grab the current MSR */
651 #ifdef CONFIG_PPC_BOOK3E
652 oris r11,r11,0x8000 /* CM bit set, we'll set ICM later */
654 #else /* CONFIG_PPC_BOOK3E */
655 li r12,(MSR_64BIT | MSR_ISF)@highest
664 * This puts the TOC pointer into r2, offset by 0x8000 (as expected
665 * by the toolchain). It computes the correct value for wherever we
666 * are running at the moment, using position-independent code.
668 _GLOBAL(relative_toc)
672 ld r2,(p_toc - 0b)(r11)
677 p_toc: .llong __toc_start + 0x8000 - 0b
680 * This is where the main kernel code starts.
682 _INIT_STATIC(start_here_multiplatform)
683 /* set up the TOC (real address) */
686 /* Clear out the BSS. It may have been done in prom_init,
687 * already but that's irrelevant since prom_init will soon
688 * be detached from the kernel completely. Besides, we need
689 * to clear it now for kexec-style entry.
691 LOAD_REG_ADDR(r11,__bss_stop)
692 LOAD_REG_ADDR(r8,__bss_start)
693 sub r11,r11,r8 /* bss size */
694 addi r11,r11,7 /* round up to an even double word */
695 srdi. r11,r11,3 /* shift right by 3 */
699 mtctr r11 /* zero this many doublewords */
704 #ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
705 /* Setup OPAL entry */
710 #ifndef CONFIG_PPC_BOOK3E
713 mtmsrd r6 /* RI on */
716 #ifdef CONFIG_RELOCATABLE
717 /* Save the physical address we're running at in kernstart_addr */
718 LOAD_REG_ADDR(r4, kernstart_addr)
723 /* The following gets the stack set up with the regs */
724 /* pointing to the real addr of the kernel stack. This is */
725 /* all done to support the C function call below which sets */
726 /* up the htab. This is done because we have relocated the */
727 /* kernel but are still running in real mode. */
729 LOAD_REG_ADDR(r3,init_thread_union)
731 /* set up a stack pointer */
732 addi r1,r3,THREAD_SIZE
734 stdu r0,-STACK_FRAME_OVERHEAD(r1)
736 /* Do very early kernel initializations, including initial hash table,
737 * stab and slb setup before we turn on relocation. */
739 /* Restore parameters passed from prom_init/kexec */
741 bl .early_setup /* also sets r13 and SPRG_PACA */
743 LOAD_REG_ADDR(r3, .start_here_common)
748 b . /* prevent speculative execution */
750 /* This is where all platforms converge execution */
751 _INIT_GLOBAL(start_here_common)
752 /* relocation is on at this point */
753 std r1,PACAKSAVE(r13)
755 /* Load the TOC (virtual address) */
758 /* Do more system initializations in virtual mode */
761 /* Mark interrupts soft and hard disabled (they might be enabled
762 * in the PACA when doing hotplug)
765 stb r0,PACASOFTIRQEN(r13)
766 li r0,PACA_IRQ_HARD_DIS
767 stb r0,PACAIRQHAPPENED(r13)
769 /* Generic kernel entry */
776 * We put a few things here that have to be page-aligned.
777 * This stuff goes at the beginning of the bss, which is page-aligned.
783 .globl empty_zero_page
787 .globl swapper_pg_dir
789 .space PGD_TABLE_SIZE