Linux 如何编译baremetal hello_world.c并在qemu-system-aarch64上运行它？

作为标题，我想编译hello_world.c程序并在qemu-system-aarch64上运行它。节目如下：

#include <stdio.h>
int main()
{
printf("hello world!\n");
}

所以我选择了aarch64 elf（正确吗？）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。 如果我只是执行

aarch64 elf gcc hello\u world.c

我会得到未定义的_exit、_sbrk、_write、_close、_lseek、_read、_fstat、_isatty函数的引用错误。所以我尝试添加-spec=aem.ve-specs，但它没有抱怨（我不确定这是否正确）。 我试着运行qemu

qemu-system-aarch64 -M virt -cpu cortex-a57 -nographic -smp 1 -m 2048 -kernel a.out

---

而且它没有给我任何印子。我应该在这里更改什么？

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，默认情况下，简单编译C源代码不会运行裸机。您使用

printf

，编译器随后会在

stdio.h-标准缓冲输入/输出

库中找到它。按照您现在的编译方式，这是动态链接到您的程序的。也就是说，当调用printf时，程序在执行过程中跳入libc。您可以通过readelf看到这一点

readelf --dynamic a.out | grep NEEDED
0x0000000000000001 (NEEDED)             Shared library: [libc.so.6]

Qemu模拟机器，但不模拟操作系统。现在将动态链接的程序作为内核/OS传递。如果没有libc库，内核如何知道打印？更重要的是，由于基本引导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），它如何知道如何执行任何操作

我假设您不想编写ARM64内核，只想进入ARM64的“裸机”开发。可能只是下载Aarch64 Linux发行版（例如）

您可以创建虚拟驱动器，安装操作系统，并在该虚拟机中进行开发，以获得虚拟“裸机”开发；）比如说什么

---

如果我的假设是错误的，而你确实想进入操作系统开发，那么这个问题太长了，无法在这里回答。但我建议您研究和。

您是对的，您可以使用qemu-system-aarch64来实现您的目标。 根据您的具体需要，您有多种选择：

使用qemu模式，以及gcc

--specs=rdimon.specs

和

newlib、

或使用另一个半宿主库，如Arm Trusted固件源代码中提供的库-下面的示例使用此方法

提供您自己的

syscalls.c

，并使用

--specs=nosys.specs

ld

选项，这样您就可以在纯金属程序中使用newlib：我建议您在博客上阅读Francesco Balducci的优秀文章-下面的示例使用这种方法

使用一种更像纯金属的方法，如下文所述：它确实使用

sprintf（）

和

qemu-virt

机器的

pl011

UART来显示结果字符串

gcc\u arm64\u 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")
}

__RAM_BASE = 0x40000000;
__RAM_SIZE =  0x08000000;
__STACK_SIZE = 0x00100000;
__HEAP_SIZE  =  0x00100000;
INCLUDE gcc_arm64_ram.ld

qemu-virt-aarch64.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")
}

__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 */
    }
}

#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);
}

qemu-virt-aarch64.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 */
    }
}

#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);
}

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)

---

您可以非常轻松地编译aarch64 linux elf二进制文件，然后运行它，但不能使用

qemu-system-aarch64

，而是

qemu-aarch64

。你确定这不是你想做的吗？我的印象是你可能真的想要这个。@谢谢，我想要的是使用系统模式。qemu-system-aarch64。这些天我们正在开发一个SoC。然后你需要一个内核，或者至少一些嵌入式库来工作。这里您只有一个已编译的二进制文件，但没有内核，没有libc。我们不能在qemu上运行裸机程序吗？这是一个重要的问题。我认为这应该是可能的。（几年前我们在qemu上运行linux和rtems，但不是baremtal）是的，你可以。但是如果在C代码中有一个

printf（）

，它应该做什么？它应该将输出写入标准输出，标准输出应该连接到一些终端硬件。这可以是qemu硬件虚拟化为您提供的模拟硬件，也可以是真正的硬件（可能是一些LCD显示器）。与此（虚拟或物理）硬件交互的例程应该在哪里？交叉编译器将C代码（包括

printf

）编译为二进制代码。但是printf例程的实现应该在哪里呢。所以它都是静态链接的（因为它是纯金属工具链，而不是用于编译要在linux上运行的应用程序的linux工具链）。我的理解是，如果我们使用裸金属编译器，基本的工作是由编译器完成的（创建堆栈、堆等）。我们现在正在开发SoC，所以我必须为我们的芯片准备qemu（现在是非常初始的开发阶段），我想用裸金属程序验证qemu模型（如果reg写入，读取是为一个伪外围设备完成的。）哇，这是一个彻底的解释！非常感谢你，我很快就会试试这个。