问题描述
作为标题，我想编译hello_world.c程序并在qemu-system-aarch64上运行它。这是程序：
#include <stdio.h>
int main()
{
printf("Hello World!\n");
}
从 https://releases.linaro.org/components/toolchain/binaries/latest-7/aarch64-elf/（这是裸机目录），我可以看到这些工具链：
folder aarch64-elf -
folder aarch64-linux-gnu -
folder aarch64_be-elf -
folder aarch64_be-linux-gnu -
folder arm-eabi -
folder arm-linux-gnueabi -
folder arm-linux-gnueabihf -
folder armeb-eabi -
folder armeb-linux-gnueabi -
folder armeb-linux-gnueabihf -
folder armv8l-linux-gnueabihf
所以我选择了 aarch64-elf（这是正确的吗？）并将其安装在我的 ubuntu 16.04 机器上，并将 bin 目录添加到路径中。
如果我只执行 aarch64-elf-gcc hello_world.c，我会收到 _exit、_sbrk、_write、_close、_lseek、_read、_fstat、_isatty 函数的未定义引用错误。所以我尝试添加 -spec=aem.ve-specs 并且它没有抱怨（我不确定这是否正确）。
我尝试运行 qemu。
qemu-system-aarch64 -M virt -cpu cortex-a57 -nographic -smp 1 -m 2048 -kernel a.out
它没有给我任何打印。我应该在这里更改什么？
解决方法
你是对的，你可以使用 qemu-system-aarch64 来实现你的目标。 根据您确切想要做什么，您有多种选择：
-
使用 qemu 的 semihosting 模式，以及带有
--specs=rdimon.specs的 gccnewlib,或使用其他半主机库，例如 Arm Trusted Firmware 源代码中提供的库- 此后的示例使用这种方法。 -
提供您自己的
syscalls.c，并使用--specs=nosys.specsld选项，以便您可以在裸机程序中使用 newlib：我建议阅读来自 Francesco 的优秀文章Balau 博客上的 Balducci - 此后的示例使用这种方法。 -
使用更像裸机的方法，如下文所述：它确实使用
sprintf()和pl011机器的qemu-virtUART 来显示结果字符串.
gcc_arm64_ram.ld：
/******************************************************************************
* @file gcc_arm32.ld
* @brief GNU Linker Script for Cortex-M based device
* @version V2.0.0
* @date 21. May 2019
******************************************************************************/
/*
* Copyright (c) 2009-2019 Arm Limited. All rights reserved.
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License,Version 2.0 (the License); you may
* not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
* www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing,software
* distributed under the License is distributed on an AS IS BASIS,WITHOUT
* WARRANTIES OR CONDITIONS OF ANY KIND,either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
MEMORY
{
RAM (rwx) : ORIGIN = __RAM_BASE,LENGTH = __RAM_SIZE
}
/* Linker script to place sections and symbol values. Should be used together
* with other linker script that defines memory regions FLASH and RAM.
* It references following symbols,which must be defined in code:
* Reset_Handler : Entry of reset handler
*
* It defines following symbols,which code can use without definition:
* __exidx_start
* __exidx_end
* __copy_table_start__
* __copy_table_end__
* __zero_table_start__
* __zero_table_end__
* __etext
* __data_start__
* __preinit_array_start
* __preinit_array_end
* __init_array_start
* __init_array_end
* __fini_array_start
* __fini_array_end
* __data_end__
* __bss_start__
* __bss_end__
* __end__
* end
* __HeapLimit
* __StackLimit
* __StackTop
* __stack
*/
ENTRY(Reset_Handler)
SECTIONS
{
.text :
{
KEEP(*(.vectors))
*(.text*)
KEEP(*(.init))
KEEP(*(.fini))
/* .ctors */
*crtbegin.o(.ctors)
*crtbegin?.o(.ctors)
*(EXCLUDE_FILE(*crtend?.o *crtend.o) .ctors)
*(SORT(.ctors.*))
*(.ctors)
/* .dtors */
*crtbegin.o(.dtors)
*crtbegin?.o(.dtors)
*(EXCLUDE_FILE(*crtend?.o *crtend.o) .dtors)
*(SORT(.dtors.*))
*(.dtors)
*(.rodata*)
KEEP(*(.eh_frame*))
} > RAM
/*
* SG veneers:
* All SG veneers are placed in the special output section .gnu.sgstubs. Its start address
* must be set,either with the command line option �--section-start� or in a linker script,* to indicate where to place these veneers in memory.
*/
/*
.gnu.sgstubs :
{
. = ALIGN(32);
} > RAM
*/
.ARM.extab :
{
*(.ARM.extab* .gnu.linkonce.armextab.*)
} > RAM
__exidx_start = .;
.ARM.exidx :
{
*(.ARM.exidx* .gnu.linkonce.armexidx.*)
} > RAM
__exidx_end = .;
.copy.table :
{
. = ALIGN(16);
__copy_table_start__ = .;
LONG (__etext)
LONG (__data_start__)
LONG (__data_end__ - __data_start__)
/* Add each additional data section here */
/*
LONG (__etext2)
LONG (__data2_start__)
LONG (__data2_end__ - __data2_start__)
*/
__copy_table_end__ = .;
} > RAM
.zero.table :
{
. = ALIGN(16);
__zero_table_start__ = .;
/* Add each additional bss section here */
/*
LONG (__bss2_start__)
LONG (__bss2_end__ - __bss2_start__)
*/
__zero_table_end__ = .;
} > RAM
/**
* Location counter can end up 2byte aligned with narrow Thumb code but
* __etext is assumed by startup code to be the LMA of a section in RAM
* which must be 4byte aligned
*/
__etext = ALIGN(16);
.data : AT (__etext)
{
__data_start__ = .;
*(vtable)
*(.data)
*(.data.*)
. = ALIGN(16);
/* preinit data */
PROVIDE_HIDDEN (__preinit_array_start = .);
KEEP(*(.preinit_array))
PROVIDE_HIDDEN (__preinit_array_end = .);
. = ALIGN(16);
/* init data */
PROVIDE_HIDDEN (__init_array_start = .);
KEEP(*(SORT(.init_array.*)))
KEEP(*(.init_array))
PROVIDE_HIDDEN (__init_array_end = .);
. = ALIGN(16);
/* finit data */
PROVIDE_HIDDEN (__fini_array_start = .);
KEEP(*(SORT(.fini_array.*)))
KEEP(*(.fini_array))
PROVIDE_HIDDEN (__fini_array_end = .);
KEEP(*(.jcr*))
. = ALIGN(16);
/* All data end */
__data_end__ = .;
} > RAM
/*
* Secondary data section,optional
*
* Remember to add each additional data section
* to the .copy.table above to asure proper
* initialization during startup.
*/
/*
__etext2 = ALIGN(16);
.data2 : AT (__etext2)
{
. = ALIGN(16);
__data2_start__ = .;
*(.data2)
*(.data2.*)
. = ALIGN(16);
__data2_end__ = .;
} > RAM2
*/
.bss :
{
. = ALIGN(16);
__bss_start__ = .;
*(.bss)
*(.bss.*)
*(COMMON)
. = ALIGN(16);
__bss_end__ = .;
} > RAM AT > RAM
/*
* Secondary bss section,optional
*
* Remember to add each additional bss section
* to the .zero.table above to asure proper
* initialization during startup.
*/
/*
.bss2 :
{
. = ALIGN(16);
__bss2_start__ = .;
*(.bss2)
*(.bss2.*)
. = ALIGN(16);
__bss2_end__ = .;
} > RAM2 AT > RAM2
*/
.heap (COPY) :
{
. = ALIGN(16);
__end__ = .;
PROVIDE(end = .);
. = . + __HEAP_SIZE;
. = ALIGN(16);
__HeapLimit = .;
} > RAM
.stack (ORIGIN(RAM) + LENGTH(RAM) - __STACK_SIZE) (COPY) :
{
. = ALIGN(16);
__StackLimit = .;
. = . + __STACK_SIZE;
. = ALIGN(16);
__StackTop = .;
} > RAM
PROVIDE(__stack = __StackTop);
/* Check if data + heap + stack exceeds RAM limit */
ASSERT(__StackLimit >= __HeapLimit,"region RAM overflowed with stack")
}
qemu-virt-aarch64.ld：
__RAM_BASE = 0x40000000;
__RAM_SIZE = 0x08000000;
__STACK_SIZE = 0x00100000;
__HEAP_SIZE = 0x00100000;
INCLUDE gcc_arm64_ram.ld
startup.s：
.title startup64.s
.arch armv8-a
.text
.section .text.startup,"ax"
.globl Reset_Handler
Reset_Handler:
ldr x0,=__StackTop
mov sp,x0
bl main
wait: wfe
b wait
.end
pl011.c：
#include <stdint.h>
static volatile unsigned int * const UART0DR = ( unsigned int * ) ( uintptr_t * ) 0x9000000;
int putchar(int c)
{
*UART0DR = c; /* Transmit char */
return c;
}
void putchar_uart0( int c )
{
*UART0DR = c; /* Transmit char */
}
void putc_uart0( int c )
{
*UART0DR = c; /* Transmit char */
}
void print_uart0( const char * s )
{
while( *s != '\0' ) /* Loop until end of string */
{
*UART0DR = ( unsigned int ) ( *s ); /* Transmit char */
s++; /* Next char */
}
}
void puts_uart0( const char * s )
{
while( *s != '\0' ) /* Loop until end of string */
{
*UART0DR = ( unsigned int ) ( *s ); /* Transmit char */
if (*s == '\n') {
*UART0DR = ( unsigned int ) ( '\r' );
}
s++; /* Next char */
}
}
pl011.h：
#pragma once
#ifdef __cplusplus
extern "C" {
#endif
void putchar_uart0( int c );
void print_uart0( const char * s );
void putc_uart0( int c );
void puts_uart0( const char * s );
#ifdef __cplusplus
}
#endif
qemu-virt-aarch64.c：
#include <stdio.h>
#include <sys/types.h>
#include <sys/stat.h>
#include <unistd.h>
#include <errno.h>
#include "pl011.h"
// angel/semihosting interface
#define SYS_WRITE0 0x04
static uint64_t semihosting_call(uint32_t operation,uint64_t parameter)
{
__asm("HLT #0xF000");
}
// syscall stubs
int _close (int fd)
{
errno = EBADF;
return -1;
}
int _isatty (int fd)
{
return 1;
}
int _fstat (int fd,struct stat * st)
{
errno = EBADF;
return -1;
}
off_t _lseek (int fd,off_t ptr,int dir)
{
errno = EBADF;
return (off_t) -1;
}
int _read (int fd,void *ptr,size_t len)
{
errno = EBADF;
return -1;
}
int _write (int fd,const char *ptr,size_t len)
{
for (size_t i = 0; i < len; i++) {
putchar_uart0(ptr[i]);
}
return len;
}
void main()
{
char buffer[BUFSIZ];
uint64_t regCurrentEL;
__asm volatile ("mrs %0,CurrentEL" : "=r" (regCurrentEL));
// UART0
sprintf(buffer,"Hello EL%d World!\n",(regCurrentEL >> 2) & 0b11);
puts_uart0(buffer);
// angel/semihosting interface
sprintf(buffer,"Hello semi-hosted EL%d World!\n",(regCurrentEL >> 2) & 0b11);
semihosting_call(SYS_WRITE0,(uint64_t) (uintptr_t) buffer);
// newlib - custom syscalls.c,with _write() using UART0
printf("Hello EL%d World! (syscalls version)\n",(regCurrentEL >> 2) & 0b11);
}
请注意，省略了负责初始化 .bss 部分的代码。
编译：
/opt/arm/9/gcc-arm-9.2-2019.12-x86_64-aarch64-none-elf/bin/aarch64-none-elf-gcc -I. -O0 -ggdb -mtune=cortex-a53 -nostartfiles -ffreestanding --specs=nosys.specs -L. -Wl,-T,qemu-virt-aarch64.ld -o virt.elf startup.s pl011.c qemu-virt-aarch64.c
运行：
/opt/qemu-5.2.0/bin/qemu-system-aarch64 -semihosting -m 128M -nographic -monitor none -serial stdio -machine virt,gic-version=2,secure=on,virtualization=on -cpu cortex-a53 -kernel virt.elf
Hello EL3 World!
Hello semi-hosted EL3 World!
Hello EL3 World! (syscalls version)
请注意，我只有 x86/amd64 开发经验。但我认为你最初的一些假设是有缺陷的。
首先，简单地编译 C 源代码默认情况下不会运行裸机。您使用 printf，您的编译器随后会在 stdio.h — standard buffered input/output 库中找到它。你现在编译的方式，这是动态链接到你的程序。也就是说，您的程序在调用 printf 时会在执行期间跳转到 libc。你可以用 readelf 看到这个。
readelf --dynamic a.out | grep NEEDED
0x0000000000000001 (NEEDED) Shared library: [libc.so.6]
Qemu 模拟机器，但它不模拟操作系统。您现在将动态链接的程序作为内核/操作系统传递。如果没有 libc 库，您的内核如何知道要打印？更重要的是，它如何知道如何做任何事情，因为基本的启动步骤不是由您的程序完成的（设置和初始化 RAM、设置设备树等）。
我假设您不想编写 ARM64 内核，而只想进入 ARM64“裸机”开发。也许只是下载 Aarch64 Linux 发行版（例如 Debian）？
您可以创建一个虚拟驱动器，安装操作系统，然后在该虚拟机中进行开发，以获得虚拟的“裸机”开发；) 例如 this 之类的东西？
如果我的假设是错误的，而您确实想要从事操作系统开发，那么这个问题在这里回答太长了。但我建议查看 tutorials 和 documentation。