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>
26 #include <linux/init.h>
30 #include <asm/ppc_asm.h>
31 #include <asm/asm-offsets.h>
33 #include <asm/cputable.h>
34 #include <asm/setup.h>
35 #include <asm/hvcall.h>
36 #include <asm/thread_info.h>
37 #include <asm/firmware.h>
38 #include <asm/page_64.h>
39 #include <asm/irqflags.h>
40 #include <asm/kvm_book3s_asm.h>
41 #include <asm/ptrace.h>
42 #include <asm/hw_irq.h>
44 /* The physical memory is laid out such that the secondary processor
45 * spin code sits at 0x0000...0x00ff. On server, the vectors follow
46 * using the layout described in exceptions-64s.S
50 * Entering into this code we make the following assumptions:
52 * For pSeries or server processors:
53 * 1. The MMU is off & open firmware is running in real mode.
54 * 2. The kernel is entered at __start
55 * -or- For OPAL entry:
56 * 1. The MMU is off, processor in HV mode, primary CPU enters at 0
57 * with device-tree in gpr3. We also get OPAL base in r8 and
58 * entry in r9 for debugging purposes
59 * 2. Secondary processors enter at 0x60 with PIR in gpr3
61 * For Book3E processors:
62 * 1. The MMU is on running in AS0 in a state defined in ePAPR
63 * 2. The kernel is entered at __start
70 /* NOP this out unconditionally */
73 b __start_initialization_multiplatform
76 /* Catch branch to 0 in real mode */
79 /* Secondary processors spin on this value until it becomes non-zero.
80 * When non-zero, it contains the real address of the function the cpu
84 .globl __secondary_hold_spinloop
85 __secondary_hold_spinloop:
88 /* Secondary processors write this value with their cpu # */
89 /* after they enter the spin loop immediately below. */
90 .globl __secondary_hold_acknowledge
91 __secondary_hold_acknowledge:
94 #ifdef CONFIG_RELOCATABLE
95 /* This flag is set to 1 by a loader if the kernel should run
96 * at the loaded address instead of the linked address. This
97 * is used by kexec-tools to keep the the kdump kernel in the
98 * crash_kernel region. The loader is responsible for
99 * observing the alignment requirement.
101 /* Do not move this variable as kexec-tools knows about it. */
105 .long 0x72756e30 /* "run0" -- relocate to 0 by default */
110 * The following code is used to hold secondary processors
111 * in a spin loop after they have entered the kernel, but
112 * before the bulk of the kernel has been relocated. This code
113 * is relocated to physical address 0x60 before prom_init is run.
114 * All of it must fit below the first exception vector at 0x100.
115 * Use .globl here not _GLOBAL because we want __secondary_hold
116 * to be the actual text address, not a descriptor.
118 .globl __secondary_hold
121 #ifndef CONFIG_PPC_BOOK3E
124 mtmsrd r24 /* RI on */
126 /* Grab our physical cpu number */
128 /* stash r4 for book3e */
131 /* Tell the master cpu we're here */
132 /* Relocation is off & we are located at an address less */
133 /* than 0x100, so only need to grab low order offset. */
134 std r24,__secondary_hold_acknowledge-_stext(0)
138 #ifdef CONFIG_PPC_BOOK3E
141 /* All secondary cpus wait here until told to start. */
142 100: ld r12,__secondary_hold_spinloop-_stext(r26)
146 #if defined(CONFIG_SMP) || defined(CONFIG_KEXEC)
147 #ifdef CONFIG_PPC_BOOK3E
153 * it may be the case that other platforms have r4 right to
154 * begin with, this gives us some safety in case it is not
156 #ifdef CONFIG_PPC_BOOK3E
161 /* Make sure that patched code is visible */
168 /* This value is used to mark exception frames on the stack. */
171 .tc ID_72656773_68657265[TC],0x7265677368657265
175 * On server, we include the exception vectors code here as it
176 * relies on absolute addressing which is only possible within
177 * this compilation unit
179 #ifdef CONFIG_PPC_BOOK3S
180 #include "exceptions-64s.S"
183 _GLOBAL(generic_secondary_thread_init)
186 /* turn on 64-bit mode */
189 /* get a valid TOC pointer, wherever we're mapped at */
193 #ifdef CONFIG_PPC_BOOK3E
194 /* Book3E initialization */
196 bl book3e_secondary_thread_init
198 b generic_secondary_common_init
201 * On pSeries and most other platforms, secondary processors spin
202 * in the following code.
203 * At entry, r3 = this processor's number (physical cpu id)
205 * On Book3E, r4 = 1 to indicate that the initial TLB entry for
206 * this core already exists (setup via some other mechanism such
207 * as SCOM before entry).
209 _GLOBAL(generic_secondary_smp_init)
214 /* turn on 64-bit mode */
217 /* get a valid TOC pointer, wherever we're mapped at */
221 #ifdef CONFIG_PPC_BOOK3E
222 /* Book3E initialization */
225 bl book3e_secondary_core_init
228 generic_secondary_common_init:
229 /* Set up a paca value for this processor. Since we have the
230 * physical cpu id in r24, we need to search the pacas to find
231 * which logical id maps to our physical one.
233 LOAD_REG_ADDR(r13, paca) /* Load paca pointer */
234 ld r13,0(r13) /* Get base vaddr of paca array */
236 addi r13,r13,PACA_SIZE /* know r13 if used accidentally */
237 b kexec_wait /* wait for next kernel if !SMP */
239 LOAD_REG_ADDR(r7, nr_cpu_ids) /* Load nr_cpu_ids address */
240 lwz r7,0(r7) /* also the max paca allocated */
241 li r5,0 /* logical cpu id */
242 1: lhz r6,PACAHWCPUID(r13) /* Load HW procid from paca */
243 cmpw r6,r24 /* Compare to our id */
245 addi r13,r13,PACA_SIZE /* Loop to next PACA on miss */
247 cmpw r5,r7 /* Check if more pacas exist */
250 mr r3,r24 /* not found, copy phys to r3 */
251 b kexec_wait /* next kernel might do better */
254 #ifdef CONFIG_PPC_BOOK3E
255 addi r12,r13,PACA_EXTLB /* and TLB exc frame in another */
256 mtspr SPRN_SPRG_TLB_EXFRAME,r12
259 /* From now on, r24 is expected to be logical cpuid */
262 /* See if we need to call a cpu state restore handler */
263 LOAD_REG_ADDR(r23, cur_cpu_spec)
265 ld r12,CPU_SPEC_RESTORE(r23)
268 #if !defined(_CALL_ELF) || _CALL_ELF != 2
274 3: LOAD_REG_ADDR(r3, spinning_secondaries) /* Decrement spinning_secondaries */
282 lbz r23,PACAPROCSTART(r13) /* Test if this processor should */
285 beq 4b /* Loop until told to go */
287 sync /* order paca.run and cur_cpu_spec */
288 isync /* In case code patching happened */
290 /* Create a temp kernel stack for use before relocation is on. */
291 ld r1,PACAEMERGSP(r13)
292 subi r1,r1,STACK_FRAME_OVERHEAD
299 * Assumes we're mapped EA == RA if the MMU is on.
301 #ifdef CONFIG_PPC_BOOK3S
304 andi. r0,r3,MSR_IR|MSR_DR
312 b . /* prevent speculative execution */
317 * Here is our main kernel entry point. We support currently 2 kind of entries
318 * depending on the value of r5.
320 * r5 != NULL -> OF entry, we go to prom_init, "legacy" parameter content
323 * r5 == NULL -> kexec style entry. r3 is a physical pointer to the
324 * DT block, r4 is a physical pointer to the kernel itself
327 __start_initialization_multiplatform:
328 /* Make sure we are running in 64 bits mode */
331 /* Get TOC pointer (current runtime address) */
334 /* find out where we are now */
336 0: mflr r26 /* r26 = runtime addr here */
337 addis r26,r26,(_stext - 0b)@ha
338 addi r26,r26,(_stext - 0b)@l /* current runtime base addr */
341 * Are we booted from a PROM Of-type client-interface ?
345 b __boot_from_prom /* yes -> prom */
347 /* Save parameters */
350 #ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
351 /* Save OPAL entry */
356 #ifdef CONFIG_PPC_BOOK3E
357 bl start_initialization_book3e
360 /* Setup some critical 970 SPRs before switching MMU off */
363 cmpwi r0,0x39 /* 970 */
365 cmpwi r0,0x3c /* 970FX */
367 cmpwi r0,0x44 /* 970MP */
369 cmpwi r0,0x45 /* 970GX */
371 1: bl __cpu_preinit_ppc970
374 /* Switch off MMU if not already off */
377 #endif /* CONFIG_PPC_BOOK3E */
380 #ifdef CONFIG_PPC_OF_BOOT_TRAMPOLINE
381 /* Save parameters */
389 * Align the stack to 16-byte boundary
390 * Depending on the size and layout of the ELF sections in the initial
391 * boot binary, the stack pointer may be unaligned on PowerMac
395 #ifdef CONFIG_RELOCATABLE
396 /* Relocate code for where we are now */
401 /* Restore parameters */
408 /* Do all of the interaction with OF client interface */
411 #endif /* #CONFIG_PPC_OF_BOOT_TRAMPOLINE */
413 /* We never return. We also hit that trap if trying to boot
414 * from OF while CONFIG_PPC_OF_BOOT_TRAMPOLINE isn't selected */
418 #ifdef CONFIG_RELOCATABLE
419 /* process relocations for the final address of the kernel */
420 lis r25,PAGE_OFFSET@highest /* compute virtual base of kernel */
422 lwz r7,__run_at_load-_stext(r26)
423 cmplwi cr0,r7,1 /* flagged to stay where we are ? */
431 * We need to run with _stext at physical address PHYSICAL_START.
432 * This will leave some code in the first 256B of
433 * real memory, which are reserved for software use.
435 * Note: This process overwrites the OF exception vectors.
437 li r3,0 /* target addr */
438 #ifdef CONFIG_PPC_BOOK3E
439 tovirt(r3,r3) /* on booke, we already run at PAGE_OFFSET */
441 mr. r4,r26 /* In some cases the loader may */
442 beq 9f /* have already put us at zero */
443 li r6,0x100 /* Start offset, the first 0x100 */
444 /* bytes were copied earlier. */
445 #ifdef CONFIG_PPC_BOOK3E
446 tovirt(r6,r6) /* on booke, we already run at PAGE_OFFSET */
449 #ifdef CONFIG_RELOCATABLE
451 * Check if the kernel has to be running as relocatable kernel based on the
452 * variable __run_at_load, if it is set the kernel is treated as relocatable
453 * kernel, otherwise it will be moved to PHYSICAL_START
455 lwz r7,__run_at_load-_stext(r26)
459 /* just copy interrupts */
460 LOAD_REG_IMMEDIATE(r5, __end_interrupts - _stext)
464 lis r5,(copy_to_here - _stext)@ha
465 addi r5,r5,(copy_to_here - _stext)@l /* # bytes of memory to copy */
467 bl copy_and_flush /* copy the first n bytes */
468 /* this includes the code being */
470 addis r8,r3,(4f - _stext)@ha /* Jump to the copy of this code */
471 addi r12,r8,(4f - _stext)@l /* that we just made */
476 p_end: .llong _end - _stext
478 4: /* Now copy the rest of the kernel up to _end */
479 addis r5,r26,(p_end - _stext)@ha
480 ld r5,(p_end - _stext)@l(r5) /* get _end */
481 5: bl copy_and_flush /* copy the rest */
483 9: b start_here_multiplatform
486 * Copy routine used to copy the kernel to start at physical address 0
487 * and flush and invalidate the caches as needed.
488 * r3 = dest addr, r4 = source addr, r5 = copy limit, r6 = start offset
489 * on exit, r3, r4, r5 are unchanged, r6 is updated to be >= r5.
491 * Note: this routine *only* clobbers r0, r6 and lr
493 _GLOBAL(copy_and_flush)
496 4: li r0,8 /* Use the smallest common */
497 /* denominator cache line */
498 /* size. This results in */
499 /* extra cache line flushes */
500 /* but operation is correct. */
501 /* Can't get cache line size */
502 /* from NACA as it is being */
505 mtctr r0 /* put # words/line in ctr */
506 3: addi r6,r6,8 /* copy a cache line */
510 dcbst r6,r3 /* write it to memory */
512 icbi r6,r3 /* flush the icache line */
525 #ifdef CONFIG_PPC_PMAC
527 * On PowerMac, secondary processors starts from the reset vector, which
528 * is temporarily turned into a call to one of the functions below.
533 .globl __secondary_start_pmac_0
534 __secondary_start_pmac_0:
535 /* NB the entries for cpus 0, 1, 2 must each occupy 8 bytes. */
545 _GLOBAL(pmac_secondary_start)
546 /* turn on 64-bit mode */
551 rldimi r3,r0,40,23 /* clear bit 23 (rm_ci) */
558 /* get TOC pointer (real address) */
562 /* Copy some CPU settings from CPU 0 */
563 bl __restore_cpu_ppc970
565 /* pSeries do that early though I don't think we really need it */
568 mtmsrd r3 /* RI on */
570 /* Set up a paca value for this processor. */
571 LOAD_REG_ADDR(r4,paca) /* Load paca pointer */
572 ld r4,0(r4) /* Get base vaddr of paca array */
573 mulli r13,r24,PACA_SIZE /* Calculate vaddr of right paca */
574 add r13,r13,r4 /* for this processor. */
575 SET_PACA(r13) /* Save vaddr of paca in an SPRG*/
577 /* Mark interrupts soft and hard disabled (they might be enabled
578 * in the PACA when doing hotplug)
581 stb r0,PACASOFTIRQEN(r13)
582 li r0,PACA_IRQ_HARD_DIS
583 stb r0,PACAIRQHAPPENED(r13)
585 /* Create a temp kernel stack for use before relocation is on. */
586 ld r1,PACAEMERGSP(r13)
587 subi r1,r1,STACK_FRAME_OVERHEAD
591 #endif /* CONFIG_PPC_PMAC */
594 * This function is called after the master CPU has released the
595 * secondary processors. The execution environment is relocation off.
596 * The paca for this processor has the following fields initialized at
598 * 1. Processor number
599 * 2. Segment table pointer (virtual address)
600 * On entry the following are set:
601 * r1 = stack pointer (real addr of temp stack)
602 * r24 = cpu# (in Linux terms)
603 * r13 = paca virtual address
604 * SPRG_PACA = paca virtual address
609 .globl __secondary_start
611 /* Set thread priority to MEDIUM */
614 /* Initialize the kernel stack */
615 LOAD_REG_ADDR(r3, current_set)
616 sldi r28,r24,3 /* get current_set[cpu#] */
618 addi r14,r14,THREAD_SIZE-STACK_FRAME_OVERHEAD
619 std r14,PACAKSAVE(r13)
621 /* Do early setup for that CPU (stab, slb, hash table pointer) */
622 bl early_setup_secondary
625 * setup the new stack pointer, but *don't* use this until
630 /* Clear backchain so we get nice backtraces */
634 /* Mark interrupts soft and hard disabled (they might be enabled
635 * in the PACA when doing hotplug)
637 stb r7,PACASOFTIRQEN(r13)
638 li r0,PACA_IRQ_HARD_DIS
639 stb r0,PACAIRQHAPPENED(r13)
641 /* enable MMU and jump to start_secondary */
642 LOAD_REG_ADDR(r3, start_secondary_prolog)
643 LOAD_REG_IMMEDIATE(r4, MSR_KERNEL)
648 b . /* prevent speculative execution */
651 * Running with relocation on at this point. All we want to do is
652 * zero the stack back-chain pointer and get the TOC virtual address
653 * before going into C code.
655 start_secondary_prolog:
658 std r3,0(r1) /* Zero the stack frame pointer */
662 * Reset stack pointer and call start_secondary
663 * to continue with online operation when woken up
664 * from cede in cpu offline.
666 _GLOBAL(start_secondary_resume)
667 ld r1,PACAKSAVE(r13) /* Reload kernel stack pointer */
669 std r3,0(r1) /* Zero the stack frame pointer */
675 * This subroutine clobbers r11 and r12
678 mfmsr r11 /* grab the current MSR */
679 #ifdef CONFIG_PPC_BOOK3E
680 oris r11,r11,0x8000 /* CM bit set, we'll set ICM later */
682 #else /* CONFIG_PPC_BOOK3E */
683 li r12,(MSR_64BIT | MSR_ISF)@highest
692 * This puts the TOC pointer into r2, offset by 0x8000 (as expected
693 * by the toolchain). It computes the correct value for wherever we
694 * are running at the moment, using position-independent code.
696 * Note: The compiler constructs pointers using offsets from the
697 * TOC in -mcmodel=medium mode. After we relocate to 0 but before
698 * the MMU is on we need our TOC to be a virtual address otherwise
699 * these pointers will be real addresses which may get stored and
700 * accessed later with the MMU on. We use tovirt() at the call
701 * sites to handle this.
703 _GLOBAL(relative_toc)
707 ld r2,(p_toc - 0b)(r11)
713 p_toc: .llong __toc_start + 0x8000 - 0b
716 * This is where the main kernel code starts.
718 start_here_multiplatform:
723 /* Clear out the BSS. It may have been done in prom_init,
724 * already but that's irrelevant since prom_init will soon
725 * be detached from the kernel completely. Besides, we need
726 * to clear it now for kexec-style entry.
728 LOAD_REG_ADDR(r11,__bss_stop)
729 LOAD_REG_ADDR(r8,__bss_start)
730 sub r11,r11,r8 /* bss size */
731 addi r11,r11,7 /* round up to an even double word */
732 srdi. r11,r11,3 /* shift right by 3 */
736 mtctr r11 /* zero this many doublewords */
741 #ifdef CONFIG_PPC_EARLY_DEBUG_OPAL
742 /* Setup OPAL entry */
743 LOAD_REG_ADDR(r11, opal)
748 #ifndef CONFIG_PPC_BOOK3E
751 mtmsrd r6 /* RI on */
754 #ifdef CONFIG_RELOCATABLE
755 /* Save the physical address we're running at in kernstart_addr */
756 LOAD_REG_ADDR(r4, kernstart_addr)
761 /* The following gets the stack set up with the regs */
762 /* pointing to the real addr of the kernel stack. This is */
763 /* all done to support the C function call below which sets */
764 /* up the htab. This is done because we have relocated the */
765 /* kernel but are still running in real mode. */
767 LOAD_REG_ADDR(r3,init_thread_union)
769 /* set up a stack pointer */
770 addi r1,r3,THREAD_SIZE
772 stdu r0,-STACK_FRAME_OVERHEAD(r1)
774 /* Do very early kernel initializations, including initial hash table,
775 * stab and slb setup before we turn on relocation. */
777 /* Restore parameters passed from prom_init/kexec */
779 bl early_setup /* also sets r13 and SPRG_PACA */
781 LOAD_REG_ADDR(r3, start_here_common)
786 b . /* prevent speculative execution */
788 /* This is where all platforms converge execution */
791 /* relocation is on at this point */
792 std r1,PACAKSAVE(r13)
794 /* Load the TOC (virtual address) */
797 /* Do more system initializations in virtual mode */
800 /* Mark interrupts soft and hard disabled (they might be enabled
801 * in the PACA when doing hotplug)
804 stb r0,PACASOFTIRQEN(r13)
805 li r0,PACA_IRQ_HARD_DIS
806 stb r0,PACAIRQHAPPENED(r13)
808 /* Generic kernel entry */
815 * We put a few things here that have to be page-aligned.
816 * This stuff goes at the beginning of the bss, which is page-aligned.
822 .globl empty_zero_page
826 .globl swapper_pg_dir
828 .space PGD_TABLE_SIZE