// Copyright 2016 The Fuchsia Authors
//
// Use of this source code is governed by a MIT-style
// license that can be found in the LICENSE file or at
// https://opensource.org/licenses/MIT

#include <asm.h>
#include <arch/arm64/mmu.h>
#include <arch/arm64.h>
#include <arch/asm_macros.h>
#include <arch/defines.h>
#include <zircon/tls.h>

#ifndef __has_feature
#define __has_feature(x) 0
#endif

/*
 * Register use:
 *  x0-x3   Arguments
 *  x9-x15  Scratch
 *  x19-x28 Globals
 */
tmp                     .req x9
tmp2                    .req x10
wtmp2                   .req w10

cpuid                   .req x19
page_table0             .req x20
page_table1             .req x21
kernel_vaddr            .req x22
uart_tx                 .req x23
uart                    .req x24

// This code is purely position-independent and generates no relocations
// that need boot-time fixup; gen-kaslr-fixup.sh ensures this (and would
// ignore it if this code were in .text.boot, so don't put it there).
.text
FUNCTION(_start)
    movlit   uart_tx, 0x078af040
    movlit   uart, 0x078af100

    /* Save the Boot info for the primary CPU only */
    mrs     cpuid, mpidr_el1
    ubfx    cpuid, cpuid, #0, #15 /* mask Aff0 and Aff1 fields */
    cbnz    cpuid, .Lno_save_bootinfo
    /* save x0 in zbi_paddr */
    adrp    tmp, zbi_paddr
    str     x0, [tmp, #:lo12:zbi_paddr]
    /* save entry point physical address in kernel_entry_paddr */
    adrp    tmp, kernel_entry_paddr
    adr     tmp2, _start
    str     tmp2, [tmp, #:lo12:kernel_entry_paddr]
    adrp    tmp2, arch_boot_el
    mrs     x2, CurrentEL
    str     x2, [tmp2, #:lo12:arch_boot_el]
.Lno_save_bootinfo:

    bl      arm64_elX_to_el1
    bl      arch_invalidate_cache_all

    /* enable caches so atomics and spinlocks work */
    mrs     tmp, sctlr_el1
    orr     tmp, tmp, #(1<<12) /* Enable icache */
    orr     tmp, tmp, #(1<<2)  /* Enable dcache/ucache */
    msr     sctlr_el1, tmp

    // This can be any arbitrary (page-aligned) address >= KERNEL_ASPACE_BASE.
    // TODO(SEC-31): Choose it randomly.
    adr_global  tmp, kernel_relocated_base
    ldr     kernel_vaddr, [tmp]

    // Load the base of the translation tables.
    adr_global page_table0, tt_trampoline
    adr_global page_table1, arm64_kernel_translation_table

    // Send secondary cpus over to a waiting spot for the primary to finish.
    cbnz    cpuid, .Lmmu_enable_secondary

    // The fixup code appears right after the kernel image (at __data_end in
    // our view).  Note this code overlaps with the kernel's bss!  It
    // expects x0 to contain the actual runtime address of __code_start.
    mov     x0, kernel_vaddr
    bl      __data_end

    /* clear out the kernel's bss using current physical location */
    /* NOTE: Relies on __bss_start and _end being 16 byte aligned */
.Ldo_bss:
    adr_global tmp, __bss_start
    adr_global tmp2, _end
    sub     tmp2, tmp2, tmp
    cbz     tmp2, .Lbss_loop_done
.Lbss_loop:
    sub     tmp2, tmp2, #16
    stp     xzr, xzr, [tmp], #16
    cbnz    tmp2, .Lbss_loop
.Lbss_loop_done:

    /* set up a functional stack pointer */
    adr_global tmp, boot_cpu_kstack_end
    mov     sp, tmp

    /* make sure the boot allocator is given a chance to figure out where
     * we are loaded in physical memory. */
    bl      boot_alloc_init
/*
    mov     tmp, 0x30
    str     tmp, [uart_tx]
    mov     tmp, 0x1
    str     tmp, [uart_tx]
    mov     tmp, 'B'
    str     tmp, [uart]
*/

    /* save the physical address the kernel is loaded at */
    adr_global x0, __code_start
    adr_global x1, kernel_base_phys
    str     x0, [x1]

    /* set up the mmu according to mmu_initial_mappings */

    /* clear out the kernel translation table */
    mov     tmp, #0
.Lclear_top_page_table_loop:
    str     xzr, [page_table1, tmp, lsl #3]
    add     tmp, tmp, #1
    cmp     tmp, #MMU_KERNEL_PAGE_TABLE_ENTRIES_TOP
    bne     .Lclear_top_page_table_loop

    /* void arm64_boot_map(pte_t* kernel_table0, vaddr_t vaddr, paddr_t paddr, size_t len, pte_t flags); */

    mov x0, KERNEL_ASPACE_BASE
    bl debug_print_int

    mov x0, ARCH_PHYSMAP_SIZE
    bl debug_print_int


    /* map a large run of physical memory at the base of the kernel's address space */
    mov     x0, page_table1
    mov     x1, KERNEL_ASPACE_BASE
    mov     x2, 0
    mov     x3, ARCH_PHYSMAP_SIZE
    movlit  x4, MMU_PTE_KERNEL_DATA_FLAGS
    bl      arm64_boot_map

    mov x0, kernel_vaddr
    bl debug_print_int

    adr_global x0, __code_start
    bl debug_print_int

    adr_global x2, __code_start
    adr_global x3, _end
    sub     x0, x3, x2
    bl debug_print_int

    /* map the kernel to a fixed address */
    /* note: mapping the kernel here with full rwx, this will get locked down later in vm initialization; */
    mov     x0, page_table1
    mov     x1, kernel_vaddr
    adr_global x2, __code_start
    adr_global x3, _end
    sub     x3, x3, x2
    mov     x4, MMU_PTE_KERNEL_RWX_FLAGS
    bl      arm64_boot_map

    /* Prepare tt_trampoline page table.
     * this will identity map the 1GB page holding the physical address of this code.
     * Used to temporarily help us get switched to the upper virtual address. */

    /* Zero tt_trampoline translation tables */
    mov     tmp, #0
.Lclear_tt_trampoline:
    str     xzr, [page_table0, tmp, lsl#3]
    add     tmp, tmp, #1
    cmp     tmp, #MMU_PAGE_TABLE_ENTRIES_IDENT
    blt     .Lclear_tt_trampoline

    /* Setup mapping at phys -> phys */
    adr     tmp, .Lmmu_on_pc
    lsr     tmp, tmp, #MMU_IDENT_TOP_SHIFT    /* tmp = paddr index */
    movlit  tmp2, MMU_PTE_IDENT_FLAGS
    add     tmp2, tmp2, tmp, lsl #MMU_IDENT_TOP_SHIFT  /* tmp2 = pt entry */

    str     tmp2, [page_table0, tmp, lsl #3]  /* tt_trampoline[paddr index] = pt entry */

    /* mark page tables as set up, so secondary cpus can fall through */
    adr_global tmp, page_tables_not_ready
    str     wzr, [tmp]
    b       .Lpage_tables_ready

.Lmmu_enable_secondary:
    adr_global tmp, page_tables_not_ready
    /* trap any secondary cpus until the primary has set up the page tables */
.Lpage_tables_not_ready:
    ldr     wtmp2, [tmp]
    cbnz    wtmp2, .Lpage_tables_not_ready
.Lpage_tables_ready:

    /* set up the mmu */

    /* Invalidate TLB */
    tlbi    vmalle1is
    isb
    dsb     sy

    /* Initialize Memory Attribute Indirection Register */
    movlit  tmp, MMU_MAIR_VAL
    msr     mair_el1, tmp

    /* Initialize TCR_EL1 */
    /* set cacheable attributes on translation walk */
    /* (SMP extensions) non-shareable, inner write-back write-allocate */
    movlit  tmp, MMU_TCR_FLAGS_IDENT
    msr     tcr_el1, tmp

    // When mapping kernelspace and userspace with different
    // ASIDs, check if tcr_el1 is not getting updated (ie; set to 0)
    // In this case, move the ASID to TTBR1 else, this will result in
    // an invalid mapping
    mrs     tmp, tcr_el1
    cbnz    tmp, .Lttbr1_el1_asid

    movlit  tmp, MMU_TCR_FLAGS_IDENT1
    msr     tcr_el1, tmp

.Lttbr1_el1_asid:
    isb

    /* Write ttbr with phys addr of the translation table */
    msr     ttbr0_el1, page_table0
    msr     ttbr1_el1, page_table1
    isb

    /* Read SCTLR */
    mrs     tmp, sctlr_el1

//hit this    b   .

    /* Turn on the MMU */
    orr     tmp, tmp, #0x1

    /* Write back SCTLR */
    msr     sctlr_el1, tmp
.Lmmu_on_pc:
    isb


//did get here    b   .

    // Map our current physical PC to the virtual PC and jump there.
    // PC = next_PC - __code_start + kernel_vaddr
    adr     tmp, .Lmmu_on_vaddr
    adr     tmp2, __code_start
    sub     tmp, tmp, tmp2
    add     tmp, tmp, kernel_vaddr

// got here    b   .

    br      tmp

.Lmmu_on_vaddr:
// didn't    b   .

    /* Disable trampoline page-table in ttbr0 */
    movlit  tmp, MMU_TCR_FLAGS_KERNEL
    msr     tcr_el1, tmp
    isb

// didn't    b   .

    /* Invalidate TLB */
    tlbi    vmalle1
    isb

    cbnz    cpuid, .Lsecondary_boot

// didn't get here    b   .

    // set up the boot stack for real
    adr_global tmp, boot_cpu_kstack_end
    mov     sp, tmp

    // Set the thread pointer early so compiler-generated references
    // to the stack-guard and unsafe-sp slots work.  This is not a
    // real 'struct thread' yet, just a pointer to (past, actually)
    // the two slots used by the ABI known to the compiler.  This avoids
    // having to compile-time disable safe-stack and stack-protector
    // code generation features for all the C code in the bootstrap
    // path, which (unlike on x86, e.g.) is enough to get annoying.
    adr_global tmp, boot_cpu_fake_thread_pointer_location
    msr     tpidr_el1, tmp

    // set the per cpu pointer for cpu 0
    adr_global x18, arm64_percpu_array

// didn't hit this    b   .

    // Choose a good (ideally random) stack-guard value as early as possible.
    bl      choose_stack_guard
    mrs     tmp, tpidr_el1
    str     x0, [tmp, #ZX_TLS_STACK_GUARD_OFFSET]
    // Don't leak the value to other code.
    mov     x0, xzr

// didn't hit this    b   .

    bl  lk_main
    b   .

.Lsecondary_boot:
    bl      arm64_get_secondary_sp
    cbz     x0, .Lunsupported_cpu_trap
    mov     sp, x0
    msr     tpidr_el1, x1

    bl      arm64_secondary_entry

.Lunsupported_cpu_trap:
    wfe
    b       .Lunsupported_cpu_trap
END_FUNCTION(_start)

.ltorg

// These are logically .bss (uninitialized data).  But they're set before
// clearing the .bss, so put them in .data so they don't get zeroed.
.data
    .balign 64
DATA(arch_boot_el)
    .quad 0xdeadbeef00ff00ff
END_DATA(arch_boot_el)
DATA(zbi_paddr)
    .quad -1
END_DATA(zbi_paddr)
DATA(kernel_entry_paddr)
    .quad -1
END_DATA(kernel_entry_paddr)

DATA(page_tables_not_ready)
    .long       1
END_DATA(page_tables_not_ready)

    .balign 8
LOCAL_DATA(boot_cpu_fake_arch_thread)
    .quad 0xdeadbeef1ee2d00d // stack_guard
#if __has_feature(safe_stack)
    .quad boot_cpu_unsafe_kstack_end
#else
    .quad 0
#endif
LOCAL_DATA(boot_cpu_fake_thread_pointer_location)
END_DATA(boot_cpu_fake_arch_thread)

.bss
LOCAL_DATA(boot_cpu_kstack)
    .skip ARCH_DEFAULT_STACK_SIZE
    .balign 16
LOCAL_DATA(boot_cpu_kstack_end)
END_DATA(boot_cpu_kstack)

#if __has_feature(safe_stack)
LOCAL_DATA(boot_cpu_unsafe_kstack)
    .skip ARCH_DEFAULT_STACK_SIZE
    .balign 16
LOCAL_DATA(boot_cpu_unsafe_kstack_end)
END_DATA(boot_cpu_unsafe_kstack)
#endif

.section .bss.prebss.translation_table, "aw", @nobits
.align 3 + MMU_PAGE_TABLE_ENTRIES_IDENT_SHIFT
DATA(tt_trampoline)
    .skip 8 * MMU_PAGE_TABLE_ENTRIES_IDENT
END_DATA(tt_trampoline)

// This symbol is used by image.S
.global IMAGE_ELF_ENTRY
IMAGE_ELF_ENTRY = _start

// This symbol is used by gdb python to know the base of the kernel module
.global KERNEL_BASE_ADDRESS
KERNEL_BASE_ADDRESS = KERNEL_BASE