【物联网设备端开发】ESP开发工具：QEMU的使用方法

概要

本文提供了一些运行QEMU的ESP特定说明。有关QEMU的一般使用问题，请参阅官方文档：https://www.qemu.org/documentation/.

编译 QEMU

准备工作

在此之前，请查看有关构建先决条件的QEMU文档。如果你在Linux主机上构建QEMU，你可以参考 这篇文章.
请确保您的系统上安装了libgcrypt ( Ubuntu 中的libgcrypt-devel, Arch 中的 libgcrypt, macOS Homebrew 中的libgcrypt ).

配置

要生成 ninja 构建文件，我们需要先配置项目，可以使用以下命令：

./configure --target-list=xtensa-softmmu \
    --enable-gcrypt \
    --enable-slirp \
    --enable-debug --enable-sanitizers \
    --enable-sdl \
    --disable-strip --disable-user \
    --disable-capstone --disable-vnc \
    --disable-gtk

为了减少第一次编译时间，可以添加的 --disable 选项。可启用或禁用的选项的完整列表可以通过 ./configure --help 命令获取。

If you need a graphical interface for the virtual machine, make sure to provide --enable-sdl or --enable-gtk or both.
如果您需要虚拟机的图形界面，请添加 --enable-sdl 或 --enable-gtk配置，或两者都添加。

编译

成功配置项目后，可以使用 ninja 来构建它：

ninja -C build

编译可能需要几分钟，具体取决于之前启用或禁用的组件。编译完成后，会创建build/qemu-system-xtensa 可执行文件。

编译ESP-IDF程序进行仿真

QEMU for ESP32 target is ready, it already includes the first stage bootloader, located on the ROM on the real chip, which is mainly responsible for initializing the peripherals, like the UART and, more importantly, the SPI Flash. The latter must contain the second stage bootloader and the program to run.

Thus, in this section, we are going to create a flash image that combines the (second stage) bootloader, the partition table and the application to run. This can be done using esptool.py merge_bin command, supported in esptool.py 3.1 or later. Let’s suppose that the ESP-IDf project has just been compiled successfully, the following commands will create that flash image:

cd build
esptool.py --chip esp32 merge_bin --fill-flash-size 4MB -o flash_image.bin @flash_args

Here, flash_args is a file generated by ESP-IDF build system in the build directory, it contains the list of names of binary files and corresponding flash addresses. merge_bin command takes this list and creates the binary image of the whole flash. --fill-flash-size 4MB argument specifies the total flash size.

ESP32 target in QEMU supports flash of size 2, 4, 8 and 16MB, creating an image with any other size will result in an error.

Notes

• For “Secure Boot” feature in ESP-IDF, we recommend separate command to flash bootloader and hence flash_args file do not have corresponding entry. However, you may modify flash_args file to add entry for bootloader.bin as per below:

  0x1000 bootloader/bootloader.bin
  

• It is also possible to use esptool.py to “flash” the application into QEMU, but QEMU needs to be started with the right strapping mode. See Bootstrapping Mode section below for more info.

Run QEMU

Without attaching GDB

If you don’t need to debug the guest application, you can execute QEMU without attaching GDB to it:

build/qemu-system-xtensa -nographic \
    -machine esp32 \
    -drive file=flash_image.bin,if=mtd,format=raw

Where flash_image.bin is the SPI flash image generated previously.

With GDB server, waiting for a connection

If you need to debug the guest application, you can execute QEMU with -s -S options. This tells QEMU not to start the CPU after initializing the virtual machine. It will wait for a connection from a GDB client:

build/qemu-system-xtensa -nographic -s -S \
    -machine esp32 \
    -drive file=flash_image.bin,if=mtd,format=raw

Where flash_image.bin is the SPI flash image generated previously.

Then, to connect the GDB client, use the following command:

xtensa-esp32-elf-gdb build/app-name.elf \
    -ex "target remote :1234" \
    -ex "monitor system_reset" \
    -ex "tb app_main" -ex "c"

The last line sets a breakpoint in app_main function of the guest application and starts the virtual CPU with c. If you need to put a breakpoint in any other functions or if you don’t need to start the CPU directly, please adapt this last line.

GDB and toolchain compatibility

If you are using esp2021r1 or an earlier toolchain release, GDB may report the following error when connecting to QEMU:

Remote 'g' packet reply is too long (expected 420 bytes, got 628 bytes):

In that case, you need to set the following environment variable beforehand:

export QEMU_XTENSA_CORE_REGS_ONLY=1

When this environment variable is set, QEMU will only send the values of non-privileged registers to GDB.

If you are using later releases of Xtensa toolchain, i.e. esp-2021r2 and later, GDB will work out-of-the-box, without the need to set this environment variable.

Hardware crypto support

Starting from IDF 4.1, the following hardware crypto features are enabled by default: AES, SHA, RSA.

All of them are implemented in QEMU for ESP32 target. However, please note that the SHA emulation currently doesn’t support concurrent operations with different SHA types.

Ethernet support

Support for Opencores Ethernet MAC in ESP-IDF is added in https://github.com/espressif/esp-idf/commit/31dac92e5f0daac98190fd603df213a0a25a3807.

• When running protocols examples, enable CONFIG_EXAMPLE_CONNECT_ETHERNET and CONFIG_EXAMPLE_USE_OPENETH.
• When running a custom app, enable CONFIG_ETH_USE_OPENETH and initialize the Ethernet driver as it is done in examples/common_components/protocol_examples_common/connect.c (look for esp_eth_mac_new_openeth).

When starting QEMU, use open_eth network device.

User mode networking

For example, to start networking in user mode (TCP/UDP only, emulated device is behind NAT), add the following option to the QEMU command line:

-nic user,model=open_eth

Some ESP projects (specifically running TCP listeners) might need port forwarding to be setup,

-nic user,model=open_eth,id=lo0,hostfwd=tcp:127.0.0.1:PORT_HOST-:PORT_GUEST

(e.g. asio-echo-server sets up a server on 2222 by default, so hostfwd=tcp:127.0.0.1:2222-:2222 enables to nc localhost 2222 from host machine)

Bridge/tap networking

• Tap networking

Specifying bootstrapping mode

To specify the desired strapping mode, it is necessary to add the following argument when running QEMU:

-global driver=esp32.gpio,property=strap_mode,value=0x0f

This sets the value of GPIO_STRAP register.

• Use 0x12 for flash boot mode (default)
• Use 0x0f for UART-only download mode (since the SDIO part is not implemented)

Specifying eFuse storage

Add extra arguments to the command line:

-drive file=qemu_efuse.bin,if=none,format=raw,id=efuse
-global driver=nvram.esp32.efuse,property=drive,value=efuse

The first argument creates a block device backed by qemu_efuse.bin file, with identifier efuse. The second line configures nvram.esp32.efuse device to use this block device for storage.

The file must be created before starting QEMU:

dd if=/dev/zero bs=1 count=124 of=/tmp/qemu_efuse.bin

124 bytes is the total size of ESP32 eFuse blocks.

Note

Specifying eFuse storage is mandatory to test out any platform security features like “Secure Boot” or “Flash Encryption”.

Emulating ESP32 ECO3

For the application to detect the emulated chip as ESP32 ECO3, the following virtual efuses must be set:

• CHIP_VER_REV1
• CHIP_VER_REV2

Here is the corresponding efuse file (in hexadecimal, produced using xxd -p):

000000000000000000000000008000000000000000001000000000000000
000000000000000000000000000000000000000000000000000000000000
000000000000000000000000000000000000000000000000000000000000
000000000000000000000000000000000000000000000000000000000000
00000000

To convert this (efuse.hex) back to binary, run xxd -r -p efuse.hex qemu_efuse.bin.

Alternatively, these bits can be set using espefuse:

espefuse.py --port=socket://localhost:5555 burn_efuse CHIP_VER_REV1
espefuse.py --port=socket://localhost:5555 burn_efuse CHIP_VER_REV2

Disabling the watchdogs

By default, Timer Group watchdog timers are emulated, and TG0 WDT is enabled at reset. It is sometimes useful to disable these watchdog timers. This can be done by adding the following to the command line:

-global driver=timer.esp32.timg,property=wdt_disable,value=true

This disables the emulation of TG watchdog timers. Even if the application configures them, they will not fire.

The RTC watchdog timer is not emulated yet, so it doesn’t need to be disabled.

Using esptool.py and espefuse.py to interact with QEMU

1. Start QEMU:

   build/qemu-system-xtensa -nographic \
       -machine esp32 \
       -drive file=flash_image.bin,if=mtd,format=raw \
       -global driver=esp32.gpio,property=strap_mode,value=0x0f \
       -drive file=qemu_efuse.bin,if=none,format=raw,id=efuse \
       -global driver=nvram.esp32.efuse,property=drive,value=efuse \
       -serial tcp::5555,server,nowait
   

   The final line redirects the emulated UART to TCP port 5555 (QEMU acts as a server).

   Type q and press Enter at any time to quit.

2. Run esptool.py:

   esptool.py -p socket://localhost:5555 flash_id
   

   Flashing with idf.py also works:

   export ESPPORT=socket://localhost:5555
   idf.py flash
   

3. Or, run espefuse.py:

   espefuse.py --port socket://localhost:5555 --do-not-confirm burn_custom_mac 00:11:22:33:44:55
   

Note: esptool can not reset the emulated chip using the RTS signal, because the state of RTS is not transmitted over TCP to QEMU. To reset the emulated chip, run system_reset command in QEMU console (started at step 1).

Specifying ROM ELF file

If -kernel and -bios arguments are not given, ESP32 (rev. 3) ROM code will be loaded. This ROM code binary is included in the repository. To specify the ROM code ELF file to load, pass the filename with a -bios <filename> argument.

Using flash encryption

Self-encryption workflow

1. In the IDF application, enable CONFIG_SECURE_FLASH_ENC_ENABLED through menuconfig, and build it
2. Build the flash image as per the instructions from the Compiling the ESP-IDF program to emulate section.
3. Create qemu_efuse.bin as highlighted in the Specifying eFuse storage section.
4. Execute qemu-system-xtensa using the following command:

   build/qemu-system-xtensa -nographic -machine esp32 \
       -drive file=/path/to/qemu_efuse.bin,if=none,format=raw,id=efuse   \
       -global driver=nvram.esp32.efuse,property=drive,value=efuse       \
       -drive file=/path/to/flash_image.bin,if=mtd,format=raw            \
       -global driver=timer.esp32.timg,property=wdt_disable,value=true
   

Adding PSRAM

QEMU “memory size” option can be used to enable PSRAM emulation. By default, no PSRAM is added to the machine. You can add 2MB or 4MB PSRAM using -m 2M or -m 4M command line options, respectively.

Note that PSRAM MMU is not emulated yet, so things like bank switching (himem in IDF) do not work.

Using SD cards

QEMU emulates SD/MMC host controller used in ESP32. To add an SD card to the system, create an image and pass it to QEMU.

1. Create a raw image file, for example, 64 MB:

   $ dd if=/dev/zero bs=$((1024*1024)) count=64 of=sd_image.bin
   

2. Add the following argument when running QEMU:

   -drive file=sd_image.bin,if=sd,format=raw
   

If you need to create a large SD card image, it is recommended to use sparse cqow2 images instead of raw ones. Consult QEMU manual about qemu-img tool for details.

Only one SD card is supported at a time. You can use either slot 0 or slot 1 of the SD/MMC controller in the application code.

Enabling graphical user interface (GUI)

The ESP32 QEMU implementation implements a virtual RGB panel, absent on the real hardware, that can be used to show graphical interface. It is associated to a virtual frame buffer that can be used to populate the pixels to show. It is also possible to use the target internal RAM as a frame buffer.

To enable the graphical interface, while keeping the serial output in the console, use the following command line:

build/qemu-system-xtensa \
    -machine esp32 \
    -drive file=flash_image.bin,if=mtd,format=raw
    -display sdl \
    -serial stdio

If gtk backend was enabled when compiling QEMU, it is possible to replace -display sdl with -display gtk