Funda of Linux

Posted on January 2, 2022May 9, 2026Back End, Virtual Machine

All about Linux here

WSL openClawd

Official Tutorial for WSL setup {Link}.

Set up Ubuntu in WSL with

wsl --install Ubuntu-24.04 --location D:\WSL\OpenClawd\ --name OpenClawd

Follow the general installation {this} with

curl -fsSL https://openclaw.ai/install.sh | bash

After disclaimer and

QuickStart

, I choose the

Qwen

which suggest 2k free queries per day. It will jump to Qwen website for registration and API authorization.
For channel (chat usage), minimal usage is localhost only, but i pick Slack because I have access to it most of time.

ChatBot (Channel)

Inspired by eagles, I decided the name

boobook

, an owl – wisdom but overlooked predator.
In slack, {this} page explains types of tokens. Choose

bot token

then it lead to {this} page that shows bots are part of an app following {this} page.
After naming it and choose a workspace, we go to

OAuth&Permission

Bot Token Scope

section. I gaurantee the following scopes:

chat:write

user:read

im:history

assistant:write

im.read

im.write

for sending messages, identifying users, and read chat history, and direct messages.
Find this graph tutorial {Link} suggesting turn on Socket Mode and Event Subscription.
Install the app to a workspace and copy the bot-token starting with

xobx-

, also apply for App token back to the basic info page of the slack app, here i only grant the most basic

connections:write

scope.
In the

App Home->Message Tab

, tick

Allow users to send message...

box.
Skip skills for now.

The first message in slack will trigger pairing request, and use the following command to check and approve it.

openclaw pairing list slack

openclaw pairing approve slack <YOUR_USER_ID>

Notice that gateway is automatic installed allowing WebUI local dashboard.

Model Providers

The 20k tokens runs out in less than 1 hour. I applied for Gemini pro and ready to deploy. Then I found this blog {Link} arguing offload to different models, start with cheap primary model then supplement with expensive ones. A example config is {here}.

Trying to upgrade configs to overcome limited token problem. For now, just call

openclaw onboard

to add

Gemini-pro-2.5-flash-lite

as the primary model to make it work.
Gemini now only allow 20 requests per day.
Trying

Amazon-Claude-Haiku

, $5 per 1M input /5M output tokens, too expensive
Z.ai charge $9 per 3 months.
{Bailian-AliCloud} granted 3-mon trial but no clear API config tutorial yet. Explore later.

Find {this-CN} graphic tutorial, which mentione

openclaw gateway restart

to restart the openclaw services

FInd {this-CN} post with graphic instruction to get free Nvidia API from this link: https://build.nvidia.com/
WIth the following config

"models": {
    "mode": "merge",
    "providers": {
      "nvidia": {
        "baseUrl": "https://integrate.api.nvidia.com/v1",
        "apiKey": "nvapi- 前面复制的Key，也可以用环境变量来配置",
        "api": "openai-completions",
        "models": [
          {
            "id": "minimaxai/minimax-m2.1",
            "name": "Nvidia Minimax m2",
            "reasoning": false,
            "input": [
              "text"
            ],
            "cost": {
              "input": 0,
              "output": 0,
              "cacheRead": 0,
              "cacheWrite": 0
            },
            "contextWindow": 32768,
            "maxTokens": 8192
          }
        ]
      }
    }
  },
  "agents": {
    "defaults": {
      "model": {
        "primary": "nvidia/minimaxai/minimax-m2.1"
      },
      "models": {
        "nvidia/minimaxai/minimax-m2.1": {
          "alias": "Nvidia Minimax m2"
        }
      },
      "maxConcurrent": 4,
      "subagents": {
        "maxConcurrent": 8
      }
    }
  },

Kimi 2.5k is free for limited time on OpenCode now {off_doc}. This is also invalid since it quickly exceed the rate limit somehow.
{This} blog teach how to deploy local LLM with details. Many of it is done by OpenClaw already, just need to redirect.
Currently mainly use {vector-engine} tutorial to set up cheap running APIs. As long as the

providers

model

ID

, and

auth

are correct, the config will run.

first small trial to automate Lethe

Testing feedback function.

WSL Kali

As for 250426, simply use

wsl --install kali-linux --location D:\WSL\kali_arp\ --name kali_arp

to set up virtual machine in customized directory and nickname

Following this video: {bilibili}.
Use

namp ip_range

to scan devices.

Handy workflows

Download plain text scripts from git dir

Click the raw button for a random scripts in that folder with format like:
```
https://raw.githubusercontent.com/
```
*
Save that to a
```
prefix.txt
```
file, and save list of scripts names into
```
 files.txt
```

Use the code to iterate all lks.txt, append prefix then use wget to download.

for ln in $(cat files.txt); do printf "%s%s\n" "$(cat prefix.txt)" "$ln"| xargs wget; done;

command info search

info 

# press  to exit Documentation

Linux内存管理与扩容

内存管理

全局占用查看

# 全局占用情况
df -hl

## 注解
df 是报告磁盘空间使用情况的
-h = --human-readable 以常见格式输出大小
-l = --local 只显示本地文件大小

【df命令详解】

逐层查看内存

# 进入目标路径
cd [path]

# 查看当前路径下内存分布情况
du -h -x --max-depth=1

删除日志文件

运行过程中产生的日志，本次删除的是

var/spool/abrt</em></strong>

中的

ccpp-2020-11-28-14:52:27-3689</strong></em>

日志文件，大小4.7G。

按照解释，/spool目录下多为mails, news和运行日志，而abrt是Automatic bug report tool的缩写，系统错误日志可以在core中找到。

删除操作如下（也可见后文）

# 删除文件
rm -f %file_name

# 删除文件夹
# rm -rf %directory_name





linux内存管理知识



各英文缩写的解释如下：



Abbrev. Name Define
LVM Logic Volume Manager 逻辑内存管理器
PV Physic Volume 物理内存
VG Volume Group 根据物理内存分成的内存组
LV Logic Volume 逻辑内存



硬盘挂载流程




可以直接分区后就格式化，也可以借助LVM。
LVM的优势就是多了一层Virtual Group加工，让后期的内存分配更加灵活（而非Logic Group直接与Physical Volume 关联）。



借助LVM的内存管理机制：



Physical Volume将被格式化为基本单元，划分到不同的Virtural Group中，实际划分Logic Group时再统一从Virtual Group中调用内存。




参考：




流程图解析Linux 分区和LVM挂载，带流程图解析

表格形式列出Linux内存管理涉及的各种指令Linux磁盘分区与LVM详解




实战：linux硬盘扩容



Linux公社多图细致解析:https://www.linuxidc.com/Linux/2019-04/158346.htm
涉及流程基本一致，连接物理盘，采用LVM识别常见PV，转化为VG，分配到LV
格式化后将LV分配给现有对应路径下。
使用虚拟机的朋友注意：shared entities 无法expand，离线设置后方可上线。
可以使用lsblk展示当前分区情况



(base) [arc@centos7 ~]$ lsblk
NAME                    MAJ:MIN RM SIZE RO TYPE MOUNTPOINT
sda                       8:0    0  20G  0 disk 
├─sda1                    8:1    0   1G  0 part /boot
└─sda2                    8:2    0  19G  0 part 
  ├─centos_centos7-root 253:0    0  17G  0 lvm  /
  └─centos_centos7-swap 253:1    0   2G  0 lvm  [SWAP]

## 解释
AME : 这是块设备名。

MAJ:MIN : 本栏显示主要和次要设备号。

RM : 本栏显示设备是否可移动设备。
     设备sdb和sr0的RM值等于1，这说明他们是可移动设备。

SIZE : 本栏列出设备的容量大小信息。

RO : 该项表明设备是否为只读。
     设备的RO值为0，表明他们不是只读的。

TYPE :本栏显示块设备是否是磁盘或磁盘上的一个分区。
       sda和sdb是磁盘，而sr0是只读存储（rom）。

MOUNTPOINT : 本栏指出设备挂载的挂载点。

引自：https://blog.csdn.net/qianggezhishen/article/details/55517714



File Management



删除文件/文件夹

Abbrev.	Name	Define
LVM	Logic Volume Manager	逻辑内存管理器
PV	Physic Volume	物理内存
VG	Volume Group	根据物理内存分成的内存组
LV	Logic Volume	逻辑内存

# 删除文件
rm -f %file_name

# 删除文件夹
# rm -rf %directory_name

Packages

查找安装列表

yum list installed | grep 关键字

命令匹配过滤
可支持通配符

卸载

yum remove [package name]

仅支持删除单个包，删除时需要输入全称
但会自动卸载关联包

批量卸载

参考链接

rpm -qa | grep [包名] |xargs rpm -e --nodeps

解释：

```
rpm -qa
```
用 rpm查找所有包
```
rpm -e
```
卸载
```
--nodeps
```
仅卸载当前，不考虑依赖
```
xargs
```
Extended Arguments
形成pipeline传递前一步返回参数(stdin)，加工后（分批etc）传递下一级
常用于批量执行命令，类似于传话筒
扩展阅读：cnblogs的详细说明

查找文件

find

命令简介

find 路径 -name 名称
find / -name 'genom*'
find /usr -name '*.pdf'

关于：为何要加单引号

批量移动/重命名文件

mv file_a file_b file_c -t file_e

复制就是cp
-t应该是 to的意思，指定结果的
如果是到某个目录下，可以用'/'结尾

Ref: CSDN-Copy and Move

count files

ls -l .csv| wc -l

zip and unzip

unzip *.zip -d ./

link

usually refers to symlink: symbol link or soft link

ln -s %target_file_path %link_path

delete the link by

unlink  %link_path

soft v.s. hard

Soft link

is like shortcut in windows. It will be invalid if target file is removed. It is a door to the key.

Hard link

more like mirroring, it points to the inode (fundamental database). It is another key to the door.

SHARCNET Skills

command
```
diff
```
can distinguish the files, useful in version check
Resource manager
```
top
```
, press q> to quit
Append
```
alias lm="load modulus -i"
```
in
```
~/.bashrc
```
to set custom command.

alias

temporary alias, can run
```
alias shortcode='full code'
```

permanent alias can do

touch ~/.bash_aliases # this requires the ~/.bashrc has inherent support lines.
nano ~/.bash_aliases # edit the file
# add the alias you need
source ~/.bash_aliases # run the scripts

## file download
batch download from  dropbox
`bash
wget {dropbox Link} -O download.zip

text pipeline

裁剪

裁剪命令

| tr "\n" "\0" | xargs -d"\0"

tr = truncating删减
此处匹配的是逐字符串匹配 '@anaconda'

批量匹配于修改

对于字符串匹配和批量修改，采用sed命令

sed -i 's/OldString/NewString/g' filename

注意该命令多用于文本文档内修改，
尝试用于xargs修改grep返回命令，失败。

permissions

# individual permission setting
chmod u+w $filename # add write permission to user
chmod o+x $filename # add excute permission to public

# absolute setting mode
chmod 137 $filename 
# set permission as r--rw-rwx
# where r=read, w=write, x=excute
# each 3 are for user, group. public respectively

vim

vim /etc/%YOUR_FILE% to view the file
Press a to insert content
Press Shift+; to activate command mode.
Enter qto quit, or wq to save and quit, or q! to force quit.

visual mode

for interactive response: {Link}

```
v
```
for phrase mode, can also use
```
w
```
to start of line,
```
V
```
for line mode, good for copy
```
Ctrl
```
+
```
b
```
for block mode, good for indent check

Add users

Both

useradd

and

adduser

can do this. But former is in command form thus requires more options. The latter is preferred usually to avoid compatibility problems.

password change

passwd
# Input current pwd
# Input new pwd twice
# Done!

bash

IP reading pipeline
use alias to give a allocate a nickname for a value, the value is wrapped in " ".
Note that a '\' placed before “$" escapes the "$".
grep is a filter, like 'findstr' command in windows shell
awk is a batch operation software do the same operation line by line.
More about GNU awk see this: https://blog.csdn.net/jasonchen_gbd/article/details/54986434
the dollar mark '$' is a regular expression like sign pointing to the instance at the fourth position.
Inspired by https://www.youtube.com/watch?v=PPQ8m8xQAs8, thanks for sharing.

alias myip="ifconfig | grep broadcast | awk '{print \$2}'"

bash scripts with nano

which bash
/usr/bin/bash

cd YOUR_DIR

nano test.sh

# ^x is ctrl+x, M is alt
# write something like this
echo "May I know your name?"
read name
if [ $name ]; then
    echo "Hello, #name !"
else
    echo "I can't hear you!"
fi
# notice two space besides "$name" are vital since bash is space-sensitive
# "fi" in the last row required for loop and judge

Visual Matlab on server

Forwarding Matlab GUI to local
Download: Xming and Putty (Graphic Tutorial)
In Putty Session setup, Connection > SSH > X11. Tick Enable X11 forwarding
Then set

localhost:0.0

Here we assume Xming is listening

localhost:0.0

(Note: In Win OpenSSH, equivalent option is to run

$env:DISPLAY = "localhost:0.0"

in powershell)

ssh with x11-forwarding

ssh -X username@hostname

validate forwarding with

xclock

DOS output in UTF-8

Ref:{StackOverflow}

chcp # check current language
chcp 65001 # switch to UTF-8
chcp [mmm] # change back to your original language

Linux Jobs Pipeline Design

Use
```
&&
```
connect the commands to force all line must succeed.
```
;
```
only count on start or not.
```
||
```
indicate the tasks are alternative. {Link-2daygeek}
Use
```
pstree -p <pid>
```
to check hierachy of process. More multi-thread command see {Link}
Use
```
&
```
at the end of the line to run jobs background.
```
jobs -l
```
to list the pid and job#. {StackXchange}
The
```
wait <PID>
```
and
```
sleep <time>
```
can be links between commands.
A demo is here {Link-Baelung}
The
```
wait
```
only works if the waited process is a child process of current process...