代码拉取完成，页面将自动刷新
同步操作将从 泰晓科技/Linux Lab 强制同步，此操作会覆盖自 Fork 仓库以来所做的任何修改，且无法恢复！！！
确定后同步将在后台操作，完成时将刷新页面，请耐心等待。
Subscribe Wechat：
This project aims to create a Qemu-based Linux development Lab to easier the learning, development and testing of Linux Kernel.
Linux Lab is open source with no warranty – use at your own risk.
Homepage
Repository
Related Projects:
Cloud Lab
Linux 0.11 Lab
CS630 Qemu Lab
Now, Linux Lab becomes an intergrated Linux learning, development and testing environment, it supports:
| Items | Description |
|---|---|
| Boards | Qemu based, 8+ main Architectures, 15+ popular boards |
| Components | Uboot, Linux / Modules, Buildroot, Qemu, Linux v2.6.10 ~ 5.x supported |
| Prebuilt | All of above components has been prebuilt |
| Rootfs | Support include initrd, harddisk, mmc and nfs, Debian availab for ARM |
| Docker | Cross toolchains available in one command, external ones configurable |
| Acess | Access via web browsers, available everywhere via web vnc or web ssh |
| Network | Builtin bridge networking, every board has network (except Raspi3) |
| Boot | Support serial port, curses (ssh friendly) and graphic booting |
| Testing | Support automatic testing via make test target |
| Debugging | debuggable via make debug target |
Continue reading for more features and usage.
About 9 years ago, a tinylinux proposal: Work on Tiny Linux Kernel accepted by embedded linux foundation, therefore I have worked on this project for serveral months.
During the project cycle, several scripts written to verify if the adding tiny features (e.g. gc-sections) breaks the other kernel features on the main cpu architectures.
These scripts uses qemu-system-ARCH as the cpu/board simulator, basic boot+function tests have been done for ftrace+perf, accordingly, defconfigs, rootfs, test scripts have been prepared, at that time, all of them were simply put in a directory, without a design or holistic consideration.
They have slept in my harddisk for several years without any attention, untill one day, docker and novnc came to my world, at first, Linux 0.11 Lab was born, after that, Linux Lab was designed to unify all of the above scripts, defconfigs, rootfs and test scripts.
Docker is required by Linux Lab, please install it at first:
Linux, Mac OSX, Windows 10
older Windows
Docker Toolbox or Virtualbox/Vmware + Linux
Before running Linux Lab, please make sure the following command works without sudo and without any issue:
$ docker run hello-world
In China, to use docker service normally, please must configure one of chinese docker mirror sites, for example:
More docker related issues, such as download slowly, download timeout and download errors, are cleary documented in the 6.1 section of FAQs.
The other issues, please read the official docker docs.
Notes for Windows User:
Please make sure your Windows version support docker: Official Docker Documentation
Linux Lab only tested with 'Git Bash' in Windows, please must use with it
If installed via Docker Toolbox, please enter into the /mnt/sda1 directory of the default system on Virtualbox, otherwise, after poweroff, the data will be lost for the default /root directory is only mounted in DRAM.
$ cd /mnt/sda1
For Linux, please simply choose one directory in ~/Downloads or ~/Documents.
$ cd ~/Documents
For Mac OSX, to compile Linux normally, please create a case sensitive filesystem as the working space at first:
$ hdiutil -type SPARSE create -size 60g -fs "Case-sensitive Journaled HFS+" -volname labspace labspace.dmg
$ hdiutil attach -mountpoint ~/Documents/labspace -no-browse labspace.dmg
$ cd ~/Documents/labspace
Notes: Docker Images, Linux and Buildroot source code require many storage space, please reserve at least 50G for them.
Use Ubuntu system as an example:
Download cloud lab framework, pull images and checkout linux-lab repository:
$ git clone https://gitee.com/tinylab/cloud-lab.git
$ cd cloud-lab/ && tools/docker/choose linux-lab
Launch the lab and login with the user and password printed in the console:
$ tools/docker/run linux-lab
Re-login the lab via web browser:
$ tools/docker/vnc linux-lab
The other login methods:
$ tools/docker/webvnc linux-lab # The same as tools/docker/vnc
$ tools/docker/webssh linux-lab
$ tools/docker/ssh linux-lab
$ tools/docker/bash linux-lab
Summary of login methods:
| Login Method | Description | Default User | Where |
|---|---|---|---|
| webvnc | web desktop | ubuntu | anywhere via internet |
| webssh | web ssh | ubuntu | anywhere via internet |
| bash | docker bash | ubuntu | localhost |
| ssh | normal ssh | ubuntu | localhost |
| vnc | normal vnc | ubuntu | localhost+VNC client |
Since vnc clients differs from operating systems, we use webvnc by default to make sure auto login vnc for all systems.
If really want to use local vnc clients, please use the 'IP' and 'Normal Password' printed by tools/docker/webvnc, for example, vinagre or remmina can be used by Ubuntu users.
$ tools/docker/vnc linux-lab
...
Please login via VNC Client with:
IP: 172.17.0.3
User: 7827c9 (Only for noVNC)
Password: nl7fxd (Normal)
Password: fmkv7w (View)
...
$ sudo apt-get install vinagre
// After connected, input the above 'Normal' Password, the 'View' one is only for students.
$ vinagre 172.17.0.3
If want a newer version, we must back up any local changes at first, for example, save the container:
$ tools/docker/commit linux-lab
And then update everything:
$ tools/docker/update linux-lab
If fails, please try to clean up the containers:
$ tools/docker/rm-all
Or even clean up the whole environments:
$ tools/docker/clean-all
Then rerurn linux lab:
$ tools/docker/rerun linux-lab
Issue the following command to boot the prebuilt kernel and rootfs on the default vexpress-a9 board:
$ make boot
Login as root user without password(password is empty), just input root and press Enter:
Welcome to Linux Lab
linux-lab login: root
# uname -a
Linux linux-lab 5.1.0 #3 SMP Thu May 30 08:44:37 UTC 2019 armv7l GNU/Linux
List builtin boards:
$ make list
[ aarch64/raspi3 ]:
ARCH = arm64
CPU ?= cortex-a53
LINUX ?= v5.1
ROOTDEV ?= /dev/mmcblk0
[ aarch64/virt ]:
ARCH = arm64
CPU ?= cortex-a57
LINUX ?= v5.1
ROOTDEV ?= /dev/vda
[ arm/versatilepb ]:
ARCH = arm
CPU ?= arm926t
LINUX ?= v5.1
ROOTDEV ?= /dev/ram0
[ arm/vexpress-a9 ]:
ARCH = arm
CPU ?= cortex-a9
LINUX ?= v5.1
ROOTDEV ?= /dev/ram0
[ i386/pc ]:
ARCH = x86
CPU ?= i686
LINUX ?= v5.1
ROOTDEV ?= /dev/ram0
[ mipsel/malta ]:
ARCH = mips
CPU ?= mips32r2
LINUX ?= v5.1
ROOTDEV ?= /dev/ram0
[ ppc/g3beige ]:
ARCH = powerpc
CPU ?= generic
LINUX ?= v5.1
ROOTDEV ?= /dev/ram0
[ riscv32/virt ]:
ARCH = riscv
CPU ?= any
LINUX ?= v5.0.13
ROOTDEV ?= /dev/vda
[ riscv64/virt ]:
ARCH = riscv
CPU ?= any
LINUX ?= v5.1
ROOTDEV ?= /dev/vda
[ x86_64/pc ]:
ARCH = x86
CPU ?= x86_64
LINUX ?= v5.1
ROOTDEV ?= /dev/ram0
By default, the default board: vexpress-a9 is used, we can configure, build and boot for a specific board with BOARD, for example:
$ make BOARD=malta
$ make boot
If using board, it only works on-the-fly, the setting will not be saved, this is helpful to run multiple boards at the same and not to disrupt each other:
$ make board=malta boot
This allows to run multi boards in different terminals or background at the same time.
Check the board specific configuration:
$ cat boards/arm/vexpress-a9/Makefile
The 'Plugin' feature is supported by Linux Lab, to allow boards being added and maintained in standalone git repositories. Standalone repository is very important to ensure Linux Lab itself not grow up big and big while more and more boards being added in.
Book examples or the boards with a whole new cpu architecture benefit from such feature a lot, for book examples may use many boards and a new cpu architecture may need require lots of new packages (such as cross toolchains and the architecture specific qemu system tool).
Here maintains the available plugins:
Every board has its own configuration, some can be changed on demand, for example, memory size, linux version, buildroot version, qemu version and the other external devices, such as serial port, network devices and so on.
The configure method is very simple, just edit it by referring to current values (boards/<BOARD>/Makefile), this command open local configuration (boards/<BOARD>/.labconfig) via vim:
$ make local-edit
But please don't make a big change once, we often only need to tune linux version, this command is better for such case:
$ make list-linux
v4.12 v4.5.5 v5.0.10 [v5.1]
$ make local-config LINUX=v5.0.10
$ make list-linux
v4.12 v4.5.5 [v5.0.10] v5.1
If want to upstream your local changes, please use board-edit and board-config, otherwise, local-edit and local-config are preferrable, for they will avoid conflicts while pulling remote updates.
v0.3+ version add target dependency by default, so, if want to compile a kernel, just run:
$ make kernel-build
Or
$ make build kernel
It will do everything required, of course, we still be able to run the targets explicitly.
And futher, with the timestamping support, finished targets will not be run again during the late operations, if still want, just clean the stamp and run it again:
$ make cleanstamp kernel-build
$ make kernel-build
Or
$ make force-kernel-build
To clean all of the stamp files:
$ make cleanstamp kernel
This function also support uboot, root and qemu.
Download board specific package and the kernel, buildroot source code:
$ make source APP="bsp kernel root uboot"
Or
$ make source APP=all
Or
$ make source all
Download one by one:
$ make bsp-source
$ make kernel-source
$ make root-source
$ make uboot-source
Or
$ make source bsp
$ make source kernel
$ make source root
$ make source uboot
Checkout the target version of kernel and builroot:
$ make checkout APP="kernel root"
Checkout them one by one:
$ make kernel-checkout
$ make root-checkout
Or
$ make checkout kernel
$ make checkout root
If checkout not work due to local changes, save changes and run to get a clean environment:
$ make kernel-cleanup
$ make root-cleanup
Or
$ make cleanup kernel
$ make cleanup root
The same to qemu and uboot.
Apply available patches in boards/<BOARD>/bsp/patch/linux and patch/linux/:
$ make kernel-patch
Or
$ make patch kernel
Configure kernel and buildroot with defconfig:
$ make defconfig APP="kernel root"
Configure one by one, by default, use the defconfig in boards/<BOARD>/bsp/:
$ make kernel-defconfig
$ make root-defconfig
Or
$ make defconfig kernel
$ make defconfig root
Configure with specified defconfig:
$ make B=raspi3
$ make kernel-defconfig KCFG=bcmrpi3_defconfig
$ make root-defconfig KCFG=raspberrypi3_64_defconfig
If only defconfig name specified, search boards/ at first, and then the default configs path of buildroot, u-boot and linux-stable respectivly: buildroot/configs, u-boot/configs, linux-stable/arch//configs.
$ make kernel-menuconfig
$ make root-menuconfig
Or
$ make menuconfig kernel
$ make menuconfig root
$ make kernel-olddefconfig
$ make root-olddefconfig
$ make uboot-olddefconfig
Or
$ make olddefconfig kernel
$ make olddefconfig root
$ make olddefconfig uboot
Build kernel and buildroot together:
$ make build APP="kernel root"
Build them one by one:
$ make kernel-build # make kernel
$ make root-build # make root
Or
$ make build kernel
$ make build root
Save all of the configs and rootfs/kernel/dtb images:
$ make save APP="kernel root"
$ make saveconfig APP="kernel root"
Save configs and images to boards/<BOARD>/bsp/:
$ make kernel-saveconfig
$ make root-saveconfig
$ make root-save
$ make kernel-save
Or
$ make saveconfig kernel
$ make saveconfig root
$ make save kernel
$ make save root
Boot with serial port (nographic) by default, exit with CTRL+a x, poweroff, reboot or pkill qemu (See poweroff hang):
$ make boot
Boot with graphic (Exit with CTRL+ALT+2 quit):
$ make b=pc boot G=1 LINUX=v5.1
$ make b=versatilepb boot G=1 LINUX=v5.1
$ make b=g3beige boot G=1 LINUX=v5.1
$ make b=malta boot G=1 LINUX=v2.6.36
$ make b=vexpress-a9 boot G=1 LINUX=v4.6.7 // LINUX=v3.18.39 works too
Note: real graphic boot require LCD and keyboard drivers, the above boards work well, with linux v5.1,
raspi3 and malta has tty0 console but without keyboard input.
vexpress-a9 and virt has no LCD support by default, but for the latest qemu, it is able to boot
with G=1 and switch to serial console via the 'View' menu, this can not be used to test LCD and
keyboard drivers. XOPTS specify the eXtra qemu options.
$ make b=vexpress-a9 CONSOLE=ttyAMA0 boot G=1 LINUX=v5.1
$ make b=raspi3 CONSOLE=ttyAMA0 XOPTS="-serial vc -serial vc" boot G=1 LINUX=v5.1
Boot with curses graphic (friendly to ssh login, not work for all boards, exit with ESC+2 quit or ALT+2 quit):
$ make b=pc boot G=2 LINUX=v4.6.7
Boot with PreBuilt Kernel, Dtb and Rootfs:
$ make boot PBK=1 PBD=1 PBR=1
or
$ make boot k=0 d=0 r=0
or
$ make boot kernel=0 dtb=0 root=0
Boot with new kernel, dtb and rootfs if exists:
$ make boot PBK=0 PBD=0 PBR=0
or
$ make boot k=1 d=1 r=1
or
$ make boot kernel=1 dtb=1 root=1
Boot with new kernel and uboot, build them if not exists:
$ make boot BUILD="kernel uboot"
Boot without Uboot (only versatilepb and vexpress-a9 boards tested):
$ make boot U=0
Boot with different rootfs (depends on board, check /dev/ after boot):
$ make boot ROOTDEV=/dev/ram // support by all boards, basic boot method
$ make boot ROOTDEV=/dev/nfs // depends on network driver, only raspi3 not work
$ make boot ROOTDEV=/dev/sda
$ make boot ROOTDEV=/dev/mmcblk0
$ make boot ROOTDEV=/dev/vda // virtio based block device
Boot with extra kernel command line (XKCLI = eXtra Kernel Command LIne):
$ make boot ROOTDEV=/dev/nfs XKCLI="init=/bin/bash"
List supported options:
$ make list ROOTDEV
$ make list BOOTDEV
$ make list CCORI
$ make list NETDEV
$ make list LINUX
$ make list UBOOT
$ make list QEMU
And more xxx-list are also supported with list xxx, for example:
$ make list features
$ make list modules
$ make list gcc
A tool named scripts/config in linux kernel is helpful to get/set the kernel
config options non-interactively, based on it, both of kernel-getconfig
and kernel-setconfig are added to tune the kernel options, with them, we
can simply "enable/disable/setstr/setval/getstate" of a kernel option or many
at the same time:
Get state of a kernel module:
$ make kernel-getconfig m=minix_fs
Getting kernel config: MINIX_FS ...
output/aarch64/linux-v5.1-virt/.config:CONFIG_MINIX_FS=m
Enable a kernel module:
$ make kernel-setconfig m=minix_fs
Setting kernel config: m=minix_fs ...
output/aarch64/linux-v5.1-virt/.config:CONFIG_MINIX_FS=m
Enable new kernel config: minix_fs ...
More control commands of kernel-setconfig including y, n, c, o, s, v:
| Option | Description |
|---|---|
y |
build the modules in kernel or enable anther kernel options. |
c |
build the modules as pluginable modules, just like m. |
o |
build the modules as pluginable modules, just like m. |
n |
disable a kernel option. |
s |
RTC_SYSTOHC_DEVICE="rtc0", set the rtc device to rtc0 |
v |
v=PANIC_TIMEOUT=5, set the kernel panic timeout to 5 secs. |
Operates many options in one command line:
$ make kernel-setconfig m=tun,minix_fs y=ikconfig v=panic_timeout=5 s=DEFAULT_HOSTNAME=linux-lab n=debug_info
$ make kernel-getconfig o=tun,minix,ikconfig,panic_timeout,hostname
Build all internel kernel modules:
$ make modules
$ make modules-install
$ make root-rebuild // not need for nfs boot
$ make boot
List available modules in modules/, boards/<BOARD>/bsp/modules/:
$ make module-list
If m argument specified, list available modules in modules/, boards/<BOARD>/bsp/modules/ and linux-stable/:
$ make module-list m=hello
1 m=hello ; M=$PWD/modules/hello
$ make module-list m=tun,minix
1 c=TUN ; m=tun ; M=drivers/net
2 c=MINIX_FS ; m=minix ; M=fs/minix
Enable one kernel module:
$ make kernel-getconfig m=minix_fs
Getting kernel config: MINIX_FS ...
output/aarch64/linux-v5.1-virt/.config:CONFIG_MINIX_FS=m
$ make kernel-setconfig m=minix_fs
Setting kernel config: m=minix_fs ...
output/aarch64/linux-v5.1-virt/.config:CONFIG_MINIX_FS=m
Enable new kernel config: minix_fs ...
Build one kernel module (e.g. minix.ko):
$ make module M=fs/minix/
Or
$ make module m=minix
Install and clean the module:
$ make module-install M=fs/minix/
$ make module-clean M=fs/minix/
More flexible usage:
$ make kernel-setconfig m=tun
$ make kernel x=tun.ko M=drivers/net
$ make kernel x=drivers/net/tun.ko
$ make kernel-run drivers/net/tun.ko
Build external kernel modules (the same as internel modules):
$ make module m=hello
Or
$ make kernel x=$PWD/modules/hello/hello.ko
Kernel features are abstracted in `feature/linux/, including their configurations patchset, it can be used to manage both of the out-of-mainline and in-mainline features.
$ make feature-list
[ feature/linux ]:
+ 9pnet
+ core
- debug
- module
+ ftrace
- v2.6.36
* env.g3beige
* env.malta
* env.pc
* env.versatilepb
- v2.6.37
* env.g3beige
+ gcs
- v2.6.36
* env.g3beige
* env.malta
* env.pc
* env.versatilepb
+ kft
- v2.6.36
* env.malta
* env.pc
+ uksm
- v2.6.38
Verified boards and linux versions are recorded there, so, it should work without any issue if the environment not changed.
For example, to enable kernel modules support, simply do:
$ make feature f=module
$ make kernel-olddefconfig
$ make kernel
For kft feature in v2.6.36 for malta board:
$ make BOARD=malta
$ export LINUX=v2.6.36
$ make kernel-checkout
$ make kernel-patch
$ make kernel-defconfig
$ make feature f=kft
$ make kernel-olddefconfig
$ make kernel
$ make boot
Choose one of the tested boards: versatilepb and vexpress-a9.
$ make BOARD=vexpress-a9
Download Uboot:
$ make uboot-source
Checkout the specified version:
$ make uboot-checkout
Patching with necessary changes, BOOTDEV and ROOTDEV available, use flash by default.
$ make uboot-patch
Use tftp, sdcard or flash explicitly, should run make uboot-checkout before a new uboot-patch:
$ make uboot-patch BOOTDEV=tftp
$ make uboot-patch BOOTDEV=sdcard
$ make uboot-patch BOOTDEV=flash
BOOTDEV is used to specify where to store and load the images for uboot, ROOTDEV is used to tell kernel where to load the rootfs.
Configure:
$ make uboot-defconfig
$ make uboot-menuconfig
Building:
$ make uboot
Boot with BOOTDEV and ROOTDEV, use flash by default:
$ make boot U=1
Use tftp, sdcard or flash explicitly:
$ make boot U=1 BOOTDEV=tftp
$ make boot U=1 BOOTDEV=sdcard
$ make boot U=1 BOOTDEV=flash
We can also change ROOTDEV during boot, for example:
$ make boot U=1 BOOTDEV=flash ROOTDEV=/dev/nfs
Clean images if want to update ramdisk, dtb and uImage:
$ make uboot-images-clean
$ make uboot-clean
Save uboot images and configs:
$ make uboot-save
$ make uboot-saveconfig
Builtin qemu may not work with the newest linux kernel, so, we need compile and add external prebuilt qemu, this has been tested on vexpress-a9 and virt board.
At first, build qemu-system-ARCH:
$ make B=vexpress-a9
$ make qemu-download
$ make qemu-checkout
$ make qemu-patch
$ make qemu-defconfig
$ make qemu
$ make qemu-save
qemu-ARCH-static and qemu-system-ARCH can not be compiled together. to build
qemu-ARCH-static, please enable QEMU_US=1 in board specific Makefile and
rebuild it.
If QEMU and QTOOL specified, the one in bsp submodule will be used in advance of one installed in system, but the first used is the one just compiled if exists.
While porting to newer kernel, Linux 5.0 hangs during boot on qemu 2.5, after compiling a newer qemu 2.12.0, no hang exists. please take notice of such issue in the future kernel upgrade.
The pace of Linux mainline is very fast, builtin toolchains can not keep up, to reduce the maintaining pressure, external toolchain feature is added. for example, ARM64/virt, CCVER and CCPATH has been added for it.
List available prebuilt toolchains:
$ make gcc-list
Download, decompress and enable the external toolchain:
$ make gcc
Switch compiler version if exists, for example:
$ make gcc-switch CCORI=internal GCC=4.7
$ make gcc-switch CCORI=linaro
If not external toolchain there, the builtin will be used back.
If no builtin toolchain exists, please must use this external toolchain feature, currently, aarch64, arm, riscv, mipsel, ppc, i386, x86_64 support such feature.
GCC version can be configured in board specific Makefile for Linux, Uboot, Qemu and Root, for example:
GCC[LINUX_v2.6.11.12] = 4.4
With this configuration, GCC will be switched automatically during defconfig and compiling of the specified Linux v2.6.11.12.
To build host tools, host gcc should be configured too(please specify b=i386/pc explicitly):
$ make gcc-list b=i386/pc
$ make gcc-switch CCORI=internal GCC=4.8 b=i386/pc
Builtin rootfs is minimal, is not enough for complex application development, which requires modern Linux distributions.
Such a type of rootfs has been introduced and has been released as docker image, ubuntu 18.04 is added for arm32v7 at first, more later.
Run it via docker directly:
$ docker run -it tinylab/arm32v7-ubuntu
Extract it out and run in Linux Lab:
ARM32/vexpress-a9 (user: root, password: root):
$ tools/root/docker/extract.sh tinylab/arm32v7-ubuntu arm
$ make boot B=vexpress-a9 U=0 V=1 MEM=1024M ROOTDEV=/dev/nfs ROOTFS=$PWD/prebuilt/fullroot/tmp/tinylab-arm32v7-ubuntu
ARM64/raspi3 (user: root, password: root):
$ tools/root/docker/extract.sh tinylab/arm64v8-ubuntu arm
$ make boot B=raspi3 V=1 ROOTDEV=/dev/mmcblk0 ROOTFS=$PWD/prebuilt/fullroot/tmp/tinylab-arm64v8-ubuntu
More rootfs from docker can be found:
$ docker search arm64 | egrep "ubuntu|debian"
arm64v8/ubuntu Ubuntu is a Debian-based Linux operating system 25
arm64v8/debian Debian is a Linux distribution that's composed 20
Compile the kernel with debugging options:
$ make feature f=debug
$ make kernel-olddefconfig
$ make kernel
Compile with one thread:
$ make kernel JOBS=1
And then debug it directly:
$ make debug
It will open a new terminal, load the scripts from .kernel_gdbinit, run gdb automatically.
It equals to:
$ make debug linux
or
$ make boot DEBUG=linux
to automate debug testing:
$ make test-debug linux
or
$ make test DEBUG=linux
find out the code line of a kernel panic address:
$ make kernel-calltrace func+offset/length
to debug uboot with .uboot_gdbinit:
$ make debug uboot
or
$ make boot DEBUG=uboot
to automate uboot debug testing:
$ make test-debug uboot
or
$ make test DEBUG=uboot
Use aarch64/virt as the demo board here.
$ make BOARD=virt
Prepare for testing, install necessary files/scripts in system/:
$ make rootdir
$ make root-install
$ make root-rebuild
Simply boot and poweroff (See poweroff hang):
$ make test
Don't poweroff after testing:
$ make test TEST_FINISH=echo
Run guest test case:
$ make test TEST_CASE=/tools/ftrace/trace.sh
Run guest test cases (COMMAND_LINE_SIZE must be big enough, e.g. 4096, see cmdline_size feature below):
$ make test TEST_BEGIN=date TEST_END=date TEST_CASE='ls /root,echo hello world'
Reboot the guest system for several times:
$ make test TEST_REBOOT=2
NOTE: reboot may 1) hang, 2) continue; 3) timeout killed, TEST_TIMEOUT=30; 4) timeout continue, TIMEOUT_CONTINUE=1
Test a feature of a specified linux version on a specified board(cmdline_size feature is for increase COMMAND_LINE_SIZE to 4096):
$ make test f=kft LINUX=v2.6.36 b=malta TEST_PREPARE=board-init,kernel-cleanup
NOTE: board-init and kernel-cleanup make sure test run automatically, but kernel-cleanup is not safe, please save your code before use it!!
Test a kernel module:
$ make test m=hello
Test multiple kernel modules:
$ make test m=exception,hello
Test modules with specified ROOTDEV, nfs boot is used by default, but some boards may not support network:
$ make test m=hello,exception TEST_RD=/dev/ram0
Run test cases while testing kernel modules (test cases run between insmod and rmmod):
$ make test m=exception TEST_BEGIN=date TEST_END=date TEST_CASE='ls /root,echo hello world' TEST_PREPARE=board-init,kernel-cleanup f=cmdline_size
Run test cases while testing internal kernel modules:
$ make test m=lkdtm TEST_BEGIN='mount -t debugfs debugfs /mnt' TEST_CASE='echo EXCEPTION ">" /mnt/provoke-crash/DIRECT'
Run test cases while testing internal kernel modules, pass kernel arguments:
$ make test m=lkdtm lkdtm_args='cpoint_name=DIRECT cpoint_type=EXCEPTION'
Run test without feature-init (save time if not necessary, FI=FEATURE_INIT):
$ make test m=lkdtm lkdtm_args='cpoint_name=DIRECT cpoint_type=EXCEPTION' FI=0
Or
$ make raw-test m=lkdtm lkdtm_args='cpoint_name=DIRECT cpoint_type=EXCEPTION'
Run test with module and the module's necessary dependencies (check with make kernel-menuconfig):
$ make test m=lkdtm y=runtime_testing_menu,debug_fs lkdtm_args='cpoint_name=DIRECT cpoint_type=EXCEPTION'
Run test without feature-init, boot-init, boot-finish and no TEST_PREPARE:
$ make boot-test m=lkdtm lkdtm_args='cpoint_name=DIRECT cpoint_type=EXCEPTION'
Test a kernel module and make some targets before testing:
$ make test m=exception TEST=kernel-checkout,kernel-patch,kernel-defconfig
Test everything in one command (from download to poweroff, see poweroff hang):
$ make test TEST=kernel,root TEST_PREPARE=board-init,kernel-cleanup,root-cleanup
Test everything in one command (with uboot while support, e.g. vexpress-a9):
$ make test TEST=kernel,root,uboot TEST_PREPARE=board-init,kernel-cleanup,root-cleanup,uboot-cleanup
Test kernel hang during boot, allow to specify a timeout, timeout must happen while system hang:
$ make test TEST_TIMEOUT=30s
Test kernel debug:
$ make test DEBUG=1
To transfer files between Qemu Board and Host, three methods are supported by default:
Simply put the files with a relative path in system/, install and rebuild the rootfs:
$ cd system/
$ mkdir system/root/
$ touch system/root/new_file
$ make root-install
$ make root-rebuild
$ make boot G=1
Boot the board with ROOTDEV=/dev/nfs:
$ make boot ROOTDEV=/dev/nfs
Host:
$ make env-dump VAR=ROOTDIR
ROOTDIR="/labs/linux-lab/boards/<BOARD>/bsp/root/<BUILDROOT_VERSION>/rootfs"
Using tftp server of host from the Qemu board with the tftp command.
Host:
$ ifconfig br0
inet addr:172.17.0.3 Bcast:172.17.255.255 Mask:255.255.0.0
$ cd tftpboot/
$ ls tftpboot
kft.patch kft.log
Qemu Board:
$ ls
kft_data.log
$ tftp -g -r kft.patch 172.17.0.3
$ tftp -p -r kft.log -l kft_data.log 172.17.0.3
Note: while put file from Qemu board to host, must create an empty file in host firstly. Buggy?
To enable 9p virtio for a new board, please refer to qemu 9p setup. qemu must be compiled with --enable-virtfs, and kernel must enable the necessary options.
Reconfigure the kernel with:
CONFIG_NET_9P=y
CONFIG_NET_9P_VIRTIO=y
CONFIG_NET_9P_DEBUG=y (Optional)
CONFIG_9P_FS=y
CONFIG_9P_FS_POSIX_ACL=y
CONFIG_PCI=y
CONFIG_VIRTIO_PCI=y
CONFIG_PCI_HOST_GENERIC=y (only needed for the QEMU Arm 'virt' board)
If using -virtfs or -device virtio-9p-pci option for qemu, must enable the above PCI related options, otherwise will not work:
9pnet_virtio: no channels available for device hostshare
mount: mounting hostshare on /hostshare failed: No such file or directory
-device virtio-9p-device requires less kernel options.
To enable the above options, please simply type:
$ make feature f=9pnet
$ make kernel-olddefconfig
Docker host:
$ modprobe 9pnet_virtio
$ lsmod | grep 9p
9pnet_virtio 17519 0
9pnet 72068 1 9pnet_virtio
Host:
$ make BOARD=virt
$ make root-install # Install mount/umount scripts, ref: system/etc/init.d/S50sharing
$ make root-rebuild
$ touch hostshare/test # Create a file in host
$ make boot U=0 ROOTDEV=/dev/ram0 PBR=1 SHARE=1
$ make boot SHARE=1 SHARE_DIR=modules # for external modules development
$ make boot SHARE=1 SHARE_DIR=output/aarch64/linux-v5.1-virt/ # for internal modules learning
$ make boot SHARE=1 SHARE_DIR=examples # for c/assembly learning
Qemu Board:
$ ls /hostshare/ # Access the file in guest
test
$ touch /hostshare/guest-test # Create a file in guest
Verified boards with Linux v5.1:
| boards | Status |
|---|---|
| aarch64/virt | virtio-9p-device (virtio-9p-pci breaks nfsroot) |
| arm/vexpress-a9 | only work with virtio-9p-device and without uboot booting |
| arm/versatilepb | only work with virtio-9p-pci |
| x86_64/pc | only work with virtio-9p-pci |
| i386/pc | only work with virtio-9p-pci |
| riscv64/virt | work with virtio-9p-pci and virtio-9p-dev |
| riscv32/virt | work with virtio-9p-pci and virtio-9p-dev |
Linux Lab has added many assembly examples in examples/assembly:
$ cd examples/assembly
$ ls
aarch64 arm mips64el mipsel powerpc powerpc64 README.md x86 x86_64
$ make -s -C aarch64/
Hello, ARM64!
Linux Lab allows to access Makefile goals easily via xxx-run, for example:
$ make kernel-run help
$ make kernel-run menuconfig
$ make root-run help
$ make root-run busybox-menuconfig
$ make uboot-run help
$ make uboot-run menuconfig
-run goals allows to run sub-make goals of kernel, root and uboot directly without entering into their own building directory.
This introduces how to add a new board for Linux Lab.
list the boards, use arm as an example:
$ qemu-system-arm -M ?
Use vexpress-a9 as an example:
$ mkdir boards/arm/vexpress-a9/
Use versatilepb as an example:
$ cp boards/arm/versatilebp/Makefile boards/arm/vexpress-a9/Makefile
Comment everything, add minimal ones and then others.
Please refer to doc/qemu/qemu-doc.html or the online one http://qemu.weilnetz.de/qemu-doc.html.
We need to prepare the configs for linux, buildroot and even uboot.
Buildroot has provided many examples about buildroot and kernel configuration:
buildroot: buildroot/configs/qemu_ARCH_BOARD_defconfig
kernel: buildroot/board/qemu/ARCH-BOARD/linux-VERSION.config
Uboot has also provided many default configs:
uboot: u-boot/configs/vexpress_ca9x4_defconfig
Kernel itself also:
kernel: linux-stable/arch/arm/configs/vexpress_defconfig
Linux Lab itself also provide many working configs too, the -clone target is a
good helper to utilize existing configs:
$ make list kernel
v4.12 v5.0.10 v5.1
$ make kernel-clone LINUX=v5.1 LINUX_NEW=v5.4
$ make kernel-menuconfig
$ make kernel-saveconfig
$ make list root
2016.05 2019.02.2
$ make root-clone BUILDROOT=2019.02.2 BUILDROOT_NEW=2019.11
$ make root-menuconfig
$ make root-saveconfig
Edit the configs and Makefile untill they match our requirements.
$ make kernel-menuconfig
$ make root-menuconfig
$ make board-edit
The configuration must be put in boards/<BOARD>/ and named with necessary
version info, use raspi3 as an example:
$ make kernel-saveconfig
$ make root-saveconfig
$ ls boards/aarch64/raspi3/bsp/configs/
buildroot_2019.02.2_defconfig linux_v5.1_defconfig
2019.02.2 is the buildroot version, v5.1 is the kernel version, both of these
variables should be configured in boards/<BOARD>/Makefile.
Please use tag instead of branch, use kernel as an example:
$ cd linux-stable
$ git tag
...
v5.0
...
v5.1
..
v5.1.1
v5.1.5
...
If want v5.1 kernel, just put a line "LINUX = v5.1" in boards/<BOARD>/Makefile.
Or clone a kernel config from the old one or the official defconfig:
$ make kernel-clone LINUX_NEW=v5.3 LINUX=v5.1
Or
$ make B=i386/pc
$ pushd linux-stable && git checkout v5.4 && popd
$ make kernel-clone LINUX_NEW=v5.4 KCFG=i386_defconfig
If no tag existed, a virtual tag name with the real commmit number can be configured as following:
LINUX = v2.6.11.12
LINUX[LINUX_v2.6.11.12] = 8e63197f
Linux version specific ROOTFS are also supported:
ROOTFS[LINUX_v2.6.12.6] ?= $(BSP_ROOT)/$(BUILDROOT)/rootfs32.cpio.gz
Use kernel as an example:
$ make kernel-defconfig
$ make kernel-menuconfig
$ make kernel
$ make boot
The same to rootfs, uboot and even qemu.
$ make root-save
$ make kernel-save
$ make uboot-save
$ make root-saveconfig
$ make kernel-saveconfig
$ make uboot-saveconfig
At last, upload the images, defconfigs, patchset to board specific bsp submodule repository.
Firstly, get the remote bsp repository address as following:
$ git remote show origin
* remote origin
Fetch URL: https://gitee.com/tinylab/qemu-aarch64-raspi3/
Push URL: https://gitee.com/tinylab/qemu-aarch64-raspi3/
HEAD branch: master
Remote branch:
master tracked
Local branch configured for 'git pull':
master merges with remote master
Local ref configured for 'git push':
master pushes to master (local out of date)
Then, fork this repository from gitee.com, upload your changes, and send your pull request.
To optimize docker images download speed, please edit DOCKER_OPTS in /etc/default/docker via referring to tools/docker/install.
We assume the docker network is 10.66.0.0/16, if not, we'd better change it as following:
$ sudo vim /etc/default/docker
DOCKER_OPTS="$DOCKER_OPTS --bip=10.66.0.10/16"
$ sudo vim /lib/systemd/system/docker.service
ExecStart=/usr/bin/dockerd -H fd:// --bip=10.66.0.10/16
Please restart docker service and lab container to make this change works:
$ sudo service docker restart
$ tools/docker/rerun linux-lab
If lab network still not work, please try another private network address and eventually to avoid conflicts with LAN address.
The full function of Linux Lab depends on the full docker environment managed by Cloud Lab, so, please really never try and therefore please don't complain about why there are lots of packages missing failures and even the other weird issues.
Linux Lab is designed to use pre-installed environment with the docker technology and save our life by avoiding the packages installation issues in different systems, so, Linux Lab would never support local host using even in the future.
To use the tools under tools without sudo, please make sure add your account to the docker group and reboot your system to take effect:
$ sudo usermod -aG docker $USER
$ newgrp docker
If ping not work, please check one by one:
DNS issue
if ping 8.8.8.8 work, please check /etc/resolv.conf and make sure it is the same as your host configuration.
IP issue
if ping not work, please refer to network conflict issue and change the ip range of docker containers.
This means must configure one of the following docker mirror sites:
Potential methods of configuration in Ubuntu, depends on docker and ubuntu versions:
/etc/default/docker:
echo "DOCKER_OPTS=\"\$DOCKER_OPTS --registry-mirror=<your accelerate address>\""
/lib/systemd/system/docker.service:
ExecStart=/usr/bin/dockerd -H fd:// --bip=10.66.0.10/16 --registry-mirror=<your accelerate address>
/etc/docker/daemon.json:
{
"registry-mirrors": ["<your accelerate address>"]
}
Please restart docker service after change the accelerate address:
$ sudo service docker restart
For the other Linux systems, Windows and MacOS System, please refer to Aliyun Mirror Speedup Document.
If want to restore the installed softwares and related configurations, please save the container manually:
$ tools/docker/commit linux-lab
After host system (include virtual machine) shutdown or reboot, you can restart the lab via the "Linux Lab" icon on the desktop, or just like before, issue this command:
$ tools/docker/run linux-lab
Current implementation doesn't support the direct 'docker start' command, please learn it.
If the above methods still not restart the lab, please refer to the methods mentioned in the 6.3.9 section.
If resume from a suspended host system, the lab will restore automatically, no need to do anything to restart it, just use one of the 4 login methods mentioned in the 2.4 section, for example, start a web browser to connect it:
$ tools/docker/vnc linux-lab
kvm only supports both of qemu-system-i386 and qemu-system-x86_64 currently, and it also requires the cpu and bios support, otherwise, you may get this error log:
modprobe: ERROR: could not insert 'kvm_intel': Operation not supported
Check cpu virtualization support, if nothing output, then, cpu not support virtualization:
$ cat /proc/cpuinfo | egrep --color=always "vmx|svm"
If cpu supports, we also need to make sure it is enabled in bios features, simply reboot your computer, press 'Delete' to enter bios, please make sure the 'Intel virtualization technology' feature is 'enabled'.
Both of the poweroff and reboot commands not work on these boards currently (LINUX=v5.1):
System will directly hang there while running poweroff or reboot, to exit qemu, please pressing CTRL+a x or using pkill qemu.
To test such boards automatically, please make sure setting TEST_TIMEOUT, e.g. make test TEST_TIMEOUT=50.
Welcome to fix up them.
| Where | How |
|---|---|
| Serial Port Console | CTRL+a x |
| Curses based Graphic |
ESC+2 quit Or ALT+2 quit
|
| X based Graphic | CTRL+ALT+2 quit |
| Generic Methods |
poweroff, reboot, kill, pkill
|
That's because the docker image is not updated, just rerun the lab (please must not use tools/docker/restart here for it not using the new docker image):
$ tools/docker/pull linux-lab
$ tools/docker/rerun linux-lab
Or
$ tools/docker/update linux-lab
With tools/docker/update, every docker images and source code will be updated, it is preferred.
If nfs or tftpboot not work, please run modprobe nfsd in host side and restart the net services via /configs/tools/restart-net-servers.sh and please
make sure not use tools/docker/trun.
CTRL+w is used in both of browser and vim, to switch from one window to another, please use CTRL+Left or CTRL+Right key instead, Linux Lab has remapped CTRL+Right to CTRL+w and CTRL+Left to CTRL+p.
Long keypress not work in novnc client currently, so, long Delete not work, please use alt+delete or alt+backspace instead, more tips:
| Function | Vim | Bash |
|---|---|---|
| begin/end | ^/$ |
Ctrl + a/e |
| forward/backward | w/b |
Ctrl + Home/end |
| cut one word backword | db |
Alt + Delete/backspace |
| cut one word forward | dw |
Alt + d |
| cut all to begin | d^ |
Ctrl + u |
| cut all to end | d$ |
Ctrl + k |
| paste all cutted | p |
Ctrl + y |
In order to switch English/Chinese input method, please use CTRL+s shortcuts, it is used instead of CTRL+space to avoid conflicts with local system.
There are tow methods to tune the screen size, one is auto scaling by noVNC, another is pre-setting during launching.
The first one is setting noVNC before connecting.
* Press the left sidebar of noVNC web page
* Disconnect
* Enable 'Auto Scaling Mode' via 'Settings -> Scaling Mode: -> Local Scaling -> Apply'
* Connect
The second one is setting SCREEN_SIZE while running Linux Lab.
The screen size of lab is captured by xrandr, if not work, please check and set your own, for example:
Get available screen size values:
$ xrandr --current
Screen 0: minimum 1 x 1, current 1916 x 891, maximum 16384 x 16384
Virtual1 connected primary 1916x891+0+0 (normal left inverted right x axis y axis) 0mm x 0mm
1916x891 60.00*+
2560x1600 59.99
1920x1440 60.00
1856x1392 60.00
1792x1344 60.00
1920x1200 59.88
1600x1200 60.00
1680x1050 59.95
1400x1050 59.98
1280x1024 60.02
1440x900 59.89
1280x960 60.00
1360x768 60.02
1280x800 59.81
1152x864 75.00
1280x768 59.87
1024x768 60.00
800x600 60.32
640x480 59.94
Before running rm command, please save all of your data, for example, save the container:
$ tools/docker/commit linux-lab
Choose one and configure it:
$ cd /path/to/cloud-lab
$ tools/docker/rm-all
$ SCREEN_SIZE=800x600 tools/docker/run linux-lab
If want the default one, please remove the manual setting at first:
$ cd /path/to/cloud-lab
$ rm configs/linux-lab/docker/.screen_size
$ tools/docker/rm-all
$ tools/docker/run linux-lab
Open the left sidebar, press the 'Fullscreen' button.
Enable recording
Open the left sidebar, press the 'Settings' button, config 'File/Title/Author/Category/Tags/Description' and enable the 'Record Screen' option.
Start recording
Press the 'Connect' button.
Stop recording
Press the 'Disconnect' button.
Replay recorded video
Press the 'Play' button.
Share it
Videos are stored in 'cloud-lab/recordings', share it with help from showdesk.io.
The VNC connection may hang for some unknown reasons and therefore Linux Lab may not response sometimes, to restore it, please press the flush button of web browser or re-connect after explicitly disconnect.
If VNC login return "Disconnect timeout", wait a while and press the left 'Connect' button again, otherwise, check as following:
At first, check the containers' status (Up: Ok, Exit: Bad):
$ docker ps -a
CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES
19a61ba075b5 tinylab/linux-lab "/tools/lab/run" 4 days ago Up 4 days 22/tcp, 5900/tcp linux-lab-21575
75dae89984c9 tinylab/cloud-ubuntu-web "/startup.sh" 8 days ago Up 8 days ....443/tcp cloud-ubuntu-web
If the status is 'Exit', that means container may be shutdown or may never up, run it again to resume for the shutdown case:
$ tools/docker/run linux-lab
Otherwise, check the running logs:
$ tools/docker/logs linux-lab
If normal, that means the login account and password may have been invalid for some exceptions, please regenerte new account and password with the coming steps:
Note: The clean command will remove some containers and data, please do necessary backup before run it, for example, save the container:
$ tools/docker/commit linux-lab
VNC login fails while using mismatched password, to fix up such issue, please clean up all and rerun it:
$ tools/docker/clean linux-lab
$ tools/docker/rerun linux-lab
If the above command not work, please try this one (It will clean more data, please do necessary backup)
$ tools/docker/clean-all
$ tools/docker/rerun linux-lab
Users report many snap issues, please use apt-get instead:
This means the rootfs.ext2 image may be broken, please remove it and try make boot again, for example:
$ rm boards/aarch64/raspi3/bsp/root/2019.02.2/rootfs.ext2
$ make boot
make boot command can create this image automatically.
This means using a newer gcc than the one linux kernel version supported, the solution is switching to an older gcc version via make gcc-switch, use i386/pc board as an example:
$ make gcc-list
$ make gcc-switch CCORI=internal GCC=4.4
This may happen at make boot while the repository is cloned with root user, please simply update the owner of cloud-lab/ directory:
$ cd /path/to/cloud-lab
$ sudo chown $USER:$USER -R ./
$ tools/docker/rerun linux-lab
To make a consistent working environment, Linux Lab only support using as general user: 'ubuntu'.
This means MAC OSX not use Case sensitive filesystem, create one using hdiutil or Disk Utility yourself:
$ hdiutil create -type SPARSE -size 60g -fs "Case-sensitive Journaled HFS+" -volname labspace labspace.dmg
$ hdiutil attach -mountpoint ~/Documents/labspace -no-browse labspace.dmg
$ cd ~/Documents/labspace
Our contact wechat is tinylab, welcome to join our user & developer discussion group.
Contact us and Sponsor via wechat:
此处可能存在不合适展示的内容，页面不予展示。您可通过相关编辑功能自查并修改。
如您确认内容无涉及 不当用语 / 纯广告导流 / 暴力 / 低俗色情 / 侵权 / 盗版 / 虚假 / 无价值内容或违法国家有关法律法规的内容，可点击提交进行申诉，我们将尽快为您处理。