387 lines
11 KiB
ArmAsm
387 lines
11 KiB
ArmAsm
/*
|
|
* Copyright (C) 2007-2009 Michal Simek <monstr@monstr.eu>
|
|
* Copyright (C) 2007-2009 PetaLogix
|
|
* Copyright (C) 2006 Atmark Techno, Inc.
|
|
*
|
|
* MMU code derived from arch/ppc/kernel/head_4xx.S:
|
|
* Copyright (c) 1995-1996 Gary Thomas <gdt@linuxppc.org>
|
|
* Initial PowerPC version.
|
|
* Copyright (c) 1996 Cort Dougan <cort@cs.nmt.edu>
|
|
* Rewritten for PReP
|
|
* Copyright (c) 1996 Paul Mackerras <paulus@cs.anu.edu.au>
|
|
* Low-level exception handers, MMU support, and rewrite.
|
|
* Copyright (c) 1997 Dan Malek <dmalek@jlc.net>
|
|
* PowerPC 8xx modifications.
|
|
* Copyright (c) 1998-1999 TiVo, Inc.
|
|
* PowerPC 403GCX modifications.
|
|
* Copyright (c) 1999 Grant Erickson <grant@lcse.umn.edu>
|
|
* PowerPC 403GCX/405GP modifications.
|
|
* Copyright 2000 MontaVista Software Inc.
|
|
* PPC405 modifications
|
|
* PowerPC 403GCX/405GP modifications.
|
|
* Author: MontaVista Software, Inc.
|
|
* frank_rowand@mvista.com or source@mvista.com
|
|
* debbie_chu@mvista.com
|
|
*
|
|
* This file is subject to the terms and conditions of the GNU General Public
|
|
* License. See the file "COPYING" in the main directory of this archive
|
|
* for more details.
|
|
*/
|
|
|
|
#include <linux/init.h>
|
|
#include <linux/linkage.h>
|
|
#include <asm/thread_info.h>
|
|
#include <asm/page.h>
|
|
#include <linux/of_fdt.h> /* for OF_DT_HEADER */
|
|
|
|
#ifdef CONFIG_MMU
|
|
#include <asm/setup.h> /* COMMAND_LINE_SIZE */
|
|
#include <asm/mmu.h>
|
|
#include <asm/processor.h>
|
|
|
|
.section .data
|
|
.global empty_zero_page
|
|
.align 12
|
|
empty_zero_page:
|
|
.space PAGE_SIZE
|
|
.global swapper_pg_dir
|
|
swapper_pg_dir:
|
|
.space PAGE_SIZE
|
|
|
|
#endif /* CONFIG_MMU */
|
|
|
|
.section .rodata
|
|
.align 4
|
|
endian_check:
|
|
.word 1
|
|
|
|
__HEAD
|
|
ENTRY(_start)
|
|
#if CONFIG_KERNEL_BASE_ADDR == 0
|
|
brai TOPHYS(real_start)
|
|
.org 0x100
|
|
real_start:
|
|
#endif
|
|
|
|
mts rmsr, r0
|
|
/* Disable stack protection from bootloader */
|
|
mts rslr, r0
|
|
addi r8, r0, 0xFFFFFFFF
|
|
mts rshr, r8
|
|
/*
|
|
* According to Xilinx, msrclr instruction behaves like 'mfs rX,rpc'
|
|
* if the msrclr instruction is not enabled. We use this to detect
|
|
* if the opcode is available, by issuing msrclr and then testing the result.
|
|
* r8 == 0 - msr instructions are implemented
|
|
* r8 != 0 - msr instructions are not implemented
|
|
*/
|
|
mfs r1, rmsr
|
|
msrclr r8, 0 /* clear nothing - just read msr for test */
|
|
cmpu r8, r8, r1 /* r1 must contain msr reg content */
|
|
|
|
/* r7 may point to an FDT, or there may be one linked in.
|
|
if it's in r7, we've got to save it away ASAP.
|
|
We ensure r7 points to a valid FDT, just in case the bootloader
|
|
is broken or non-existent */
|
|
beqi r7, no_fdt_arg /* NULL pointer? don't copy */
|
|
/* Does r7 point to a valid FDT? Load HEADER magic number */
|
|
/* Run time Big/Little endian platform */
|
|
/* Save 1 as word and load byte - 0 - BIG, 1 - LITTLE */
|
|
lbui r11, r0, TOPHYS(endian_check)
|
|
beqid r11, big_endian /* DO NOT break delay stop dependency */
|
|
lw r11, r0, r7 /* Big endian load in delay slot */
|
|
lwr r11, r0, r7 /* Little endian load */
|
|
big_endian:
|
|
rsubi r11, r11, OF_DT_HEADER /* Check FDT header */
|
|
beqi r11, _prepare_copy_fdt
|
|
or r7, r0, r0 /* clear R7 when not valid DTB */
|
|
bnei r11, no_fdt_arg /* No - get out of here */
|
|
_prepare_copy_fdt:
|
|
or r11, r0, r0 /* incremment */
|
|
ori r4, r0, TOPHYS(_fdt_start)
|
|
ori r3, r0, (0x8000 - 4)
|
|
_copy_fdt:
|
|
lw r12, r7, r11 /* r12 = r7 + r11 */
|
|
sw r12, r4, r11 /* addr[r4 + r11] = r12 */
|
|
addik r11, r11, 4 /* increment counting */
|
|
bgtid r3, _copy_fdt /* loop for all entries */
|
|
addik r3, r3, -4 /* descrement loop */
|
|
no_fdt_arg:
|
|
|
|
#ifdef CONFIG_MMU
|
|
|
|
#ifndef CONFIG_CMDLINE_BOOL
|
|
/*
|
|
* handling command line
|
|
* copy command line directly to cmd_line placed in data section.
|
|
*/
|
|
beqid r5, skip /* Skip if NULL pointer */
|
|
or r11, r0, r0 /* incremment */
|
|
ori r4, r0, cmd_line /* load address of command line */
|
|
tophys(r4,r4) /* convert to phys address */
|
|
ori r3, r0, COMMAND_LINE_SIZE - 1 /* number of loops */
|
|
_copy_command_line:
|
|
/* r2=r5+r6 - r5 contain pointer to command line */
|
|
lbu r2, r5, r11
|
|
beqid r2, skip /* Skip if no data */
|
|
sb r2, r4, r11 /* addr[r4+r6]= r2 */
|
|
addik r11, r11, 1 /* increment counting */
|
|
bgtid r3, _copy_command_line /* loop for all entries */
|
|
addik r3, r3, -1 /* decrement loop */
|
|
addik r5, r4, 0 /* add new space for command line */
|
|
tovirt(r5,r5)
|
|
skip:
|
|
#endif /* CONFIG_CMDLINE_BOOL */
|
|
|
|
#ifdef NOT_COMPILE
|
|
/* save bram context */
|
|
or r11, r0, r0 /* incremment */
|
|
ori r4, r0, TOPHYS(_bram_load_start) /* save bram context */
|
|
ori r3, r0, (LMB_SIZE - 4)
|
|
_copy_bram:
|
|
lw r7, r0, r11 /* r7 = r0 + r6 */
|
|
sw r7, r4, r11 /* addr[r4 + r6] = r7 */
|
|
addik r11, r11, 4 /* increment counting */
|
|
bgtid r3, _copy_bram /* loop for all entries */
|
|
addik r3, r3, -4 /* descrement loop */
|
|
#endif
|
|
/* We have to turn on the MMU right away. */
|
|
|
|
/*
|
|
* Set up the initial MMU state so we can do the first level of
|
|
* kernel initialization. This maps the first 16 MBytes of memory 1:1
|
|
* virtual to physical.
|
|
*/
|
|
nop
|
|
addik r3, r0, MICROBLAZE_TLB_SIZE -1 /* Invalidate all TLB entries */
|
|
_invalidate:
|
|
mts rtlbx, r3
|
|
mts rtlbhi, r0 /* flush: ensure V is clear */
|
|
mts rtlblo, r0
|
|
bgtid r3, _invalidate /* loop for all entries */
|
|
addik r3, r3, -1
|
|
/* sync */
|
|
|
|
/* Setup the kernel PID */
|
|
mts rpid,r0 /* Load the kernel PID */
|
|
nop
|
|
bri 4
|
|
|
|
/*
|
|
* We should still be executing code at physical address area
|
|
* RAM_BASEADDR at this point. However, kernel code is at
|
|
* a virtual address. So, set up a TLB mapping to cover this once
|
|
* translation is enabled.
|
|
*/
|
|
|
|
addik r3,r0, CONFIG_KERNEL_START /* Load the kernel virtual address */
|
|
tophys(r4,r3) /* Load the kernel physical address */
|
|
|
|
/* start to do TLB calculation */
|
|
addik r12, r0, _end
|
|
rsub r12, r3, r12
|
|
addik r12, r12, CONFIG_LOWMEM_SIZE >> PTE_SHIFT /* that's the pad */
|
|
|
|
or r9, r0, r0 /* TLB0 = 0 */
|
|
or r10, r0, r0 /* TLB1 = 0 */
|
|
|
|
addik r11, r12, -0x1000000
|
|
bgei r11, GT16 /* size is greater than 16MB */
|
|
addik r11, r12, -0x0800000
|
|
bgei r11, GT8 /* size is greater than 8MB */
|
|
addik r11, r12, -0x0400000
|
|
bgei r11, GT4 /* size is greater than 4MB */
|
|
/* size is less than 4MB */
|
|
addik r11, r12, -0x0200000
|
|
bgei r11, GT2 /* size is greater than 2MB */
|
|
addik r9, r0, 0x0100000 /* TLB0 must be 1MB */
|
|
addik r11, r12, -0x0100000
|
|
bgei r11, GT1 /* size is greater than 1MB */
|
|
/* TLB1 is 0 which is setup above */
|
|
bri tlb_end
|
|
GT4: /* r11 contains the rest - will be either 1 or 4 */
|
|
ori r9, r0, 0x400000 /* TLB0 is 4MB */
|
|
bri TLB1
|
|
GT16: /* TLB0 is 16MB */
|
|
addik r9, r0, 0x1000000 /* means TLB0 is 16MB */
|
|
TLB1:
|
|
/* must be used r2 because of subtract if failed */
|
|
addik r2, r11, -0x0400000
|
|
bgei r2, GT20 /* size is greater than 16MB */
|
|
/* size is >16MB and <20MB */
|
|
addik r11, r11, -0x0100000
|
|
bgei r11, GT17 /* size is greater than 17MB */
|
|
/* kernel is >16MB and < 17MB */
|
|
GT1:
|
|
addik r10, r0, 0x0100000 /* means TLB1 is 1MB */
|
|
bri tlb_end
|
|
GT2: /* TLB0 is 0 and TLB1 will be 4MB */
|
|
GT17: /* TLB1 is 4MB - kernel size <20MB */
|
|
addik r10, r0, 0x0400000 /* means TLB1 is 4MB */
|
|
bri tlb_end
|
|
GT8: /* TLB0 is still zero that's why I can use only TLB1 */
|
|
GT20: /* TLB1 is 16MB - kernel size >20MB */
|
|
addik r10, r0, 0x1000000 /* means TLB1 is 16MB */
|
|
tlb_end:
|
|
|
|
/*
|
|
* Configure and load two entries into TLB slots 0 and 1.
|
|
* In case we are pinning TLBs, these are reserved in by the
|
|
* other TLB functions. If not reserving, then it doesn't
|
|
* matter where they are loaded.
|
|
*/
|
|
andi r4,r4,0xfffffc00 /* Mask off the real page number */
|
|
ori r4,r4,(TLB_WR | TLB_EX) /* Set the write and execute bits */
|
|
|
|
/*
|
|
* TLB0 is always used - check if is not zero (r9 stores TLB0 value)
|
|
* if is use TLB1 value and clear it (r10 stores TLB1 value)
|
|
*/
|
|
bnei r9, tlb0_not_zero
|
|
add r9, r10, r0
|
|
add r10, r0, r0
|
|
tlb0_not_zero:
|
|
|
|
/* look at the code below */
|
|
ori r30, r0, 0x200
|
|
andi r29, r9, 0x100000
|
|
bneid r29, 1f
|
|
addik r30, r30, 0x80
|
|
andi r29, r9, 0x400000
|
|
bneid r29, 1f
|
|
addik r30, r30, 0x80
|
|
andi r29, r9, 0x1000000
|
|
bneid r29, 1f
|
|
addik r30, r30, 0x80
|
|
1:
|
|
andi r3,r3,0xfffffc00 /* Mask off the effective page number */
|
|
ori r3,r3,(TLB_VALID)
|
|
or r3, r3, r30
|
|
|
|
/* Load tlb_skip size value which is index to first unused TLB entry */
|
|
lwi r11, r0, TOPHYS(tlb_skip)
|
|
mts rtlbx,r11 /* TLB slow 0 */
|
|
|
|
mts rtlblo,r4 /* Load the data portion of the entry */
|
|
mts rtlbhi,r3 /* Load the tag portion of the entry */
|
|
|
|
/* Increase tlb_skip size */
|
|
addik r11, r11, 1
|
|
swi r11, r0, TOPHYS(tlb_skip)
|
|
|
|
/* TLB1 can be zeroes that's why we not setup it */
|
|
beqi r10, jump_over2
|
|
|
|
/* look at the code below */
|
|
ori r30, r0, 0x200
|
|
andi r29, r10, 0x100000
|
|
bneid r29, 1f
|
|
addik r30, r30, 0x80
|
|
andi r29, r10, 0x400000
|
|
bneid r29, 1f
|
|
addik r30, r30, 0x80
|
|
andi r29, r10, 0x1000000
|
|
bneid r29, 1f
|
|
addik r30, r30, 0x80
|
|
1:
|
|
addk r4, r4, r9 /* previous addr + TLB0 size */
|
|
addk r3, r3, r9
|
|
|
|
andi r3,r3,0xfffffc00 /* Mask off the effective page number */
|
|
ori r3,r3,(TLB_VALID)
|
|
or r3, r3, r30
|
|
|
|
lwi r11, r0, TOPHYS(tlb_skip)
|
|
mts rtlbx, r11 /* r11 is used from TLB0 */
|
|
|
|
mts rtlblo,r4 /* Load the data portion of the entry */
|
|
mts rtlbhi,r3 /* Load the tag portion of the entry */
|
|
|
|
/* Increase tlb_skip size */
|
|
addik r11, r11, 1
|
|
swi r11, r0, TOPHYS(tlb_skip)
|
|
|
|
jump_over2:
|
|
/*
|
|
* Load a TLB entry for LMB, since we need access to
|
|
* the exception vectors, using a 4k real==virtual mapping.
|
|
*/
|
|
/* Use temporary TLB_ID for LMB - clear this temporary mapping later */
|
|
ori r11, r0, MICROBLAZE_LMB_TLB_ID
|
|
mts rtlbx,r11
|
|
|
|
ori r4,r0,(TLB_WR | TLB_EX)
|
|
ori r3,r0,(TLB_VALID | TLB_PAGESZ(PAGESZ_4K))
|
|
|
|
mts rtlblo,r4 /* Load the data portion of the entry */
|
|
mts rtlbhi,r3 /* Load the tag portion of the entry */
|
|
|
|
/*
|
|
* We now have the lower 16 Meg of RAM mapped into TLB entries, and the
|
|
* caches ready to work.
|
|
*/
|
|
turn_on_mmu:
|
|
ori r15,r0,start_here
|
|
ori r4,r0,MSR_KERNEL_VMS
|
|
mts rmsr,r4
|
|
nop
|
|
rted r15,0 /* enables MMU */
|
|
nop
|
|
|
|
start_here:
|
|
#endif /* CONFIG_MMU */
|
|
|
|
/* Initialize small data anchors */
|
|
addik r13, r0, _KERNEL_SDA_BASE_
|
|
addik r2, r0, _KERNEL_SDA2_BASE_
|
|
|
|
/* Initialize stack pointer */
|
|
addik r1, r0, init_thread_union + THREAD_SIZE - 4
|
|
|
|
/* Initialize r31 with current task address */
|
|
addik r31, r0, init_task
|
|
|
|
addik r11, r0, machine_early_init
|
|
brald r15, r11
|
|
nop
|
|
|
|
#ifndef CONFIG_MMU
|
|
addik r15, r0, machine_halt
|
|
braid start_kernel
|
|
nop
|
|
#else
|
|
/*
|
|
* Initialize the MMU.
|
|
*/
|
|
bralid r15, mmu_init
|
|
nop
|
|
|
|
/* Go back to running unmapped so we can load up new values
|
|
* and change to using our exception vectors.
|
|
* On the MicroBlaze, all we invalidate the used TLB entries to clear
|
|
* the old 16M byte TLB mappings.
|
|
*/
|
|
ori r15,r0,TOPHYS(kernel_load_context)
|
|
ori r4,r0,MSR_KERNEL
|
|
mts rmsr,r4
|
|
nop
|
|
bri 4
|
|
rted r15,0
|
|
nop
|
|
|
|
/* Load up the kernel context */
|
|
kernel_load_context:
|
|
ori r5, r0, MICROBLAZE_LMB_TLB_ID
|
|
mts rtlbx,r5
|
|
nop
|
|
mts rtlbhi,r0
|
|
nop
|
|
addi r15, r0, machine_halt
|
|
ori r17, r0, start_kernel
|
|
ori r4, r0, MSR_KERNEL_VMS
|
|
mts rmsr, r4
|
|
nop
|
|
rted r17, 0 /* enable MMU and jump to start_kernel */
|
|
nop
|
|
#endif /* CONFIG_MMU */
|