c0d3 :: j0rg3

Lately, I’ve been doing a lot of work from a laptop running Kali. Engaged in pursuit of a new job, I’m brushing up on some old tools and skills, exploring some bits that have changed.

My primary desktop rig is currently running Arch because I love the fine grain control and the aggressive releases. Over the years, I’ve Gentoo’d and Slacked, Crunchbanged, BSD’d, Solarised, et cet. And I’ve a fondness for all of them, especially the security-minded focus of OpenBSD. But, these days we’re usually on Arch or Kali. Initially, I went with Black Arch on the laptop but I felt the things and ways I was fixing things were too specific to my situation to be good material for posts.

Anyway, I wanted to get Bitmessage running, corresponding to another post I have in drafts. On Kali, it wasn’t going well so I put it on the Arch box and just ran it over the network. A reasonable solution if you’re in my house but also the sort of solution that will keep a hacker up at night.

If you’re lucky, there’s someone maintaining a package for the piece of software that you want to run. However, that’s often not the case.

If I correctly recall, to “fix” the problem with Bitmessage on Kali would’ve required the manual installation an older version of libraries that were already present. Those libraries should, in fact, be all ebony and ivory, living together in harmony. However, I just didn’t love the idea of that solution. I wanted to find an approach that would be useful on a broader scale.

Enter containerization/virtualization!

Wanting the lightest solution, I quickly went to Docker and realized something. I have not before built a Docker container for a GUI application. And Bitmessage’s CLI/daemon mode doesn’t provide the fluid UX that I wanted. Well, the easy way to get a GUI out of a Docker container is to forward DISPLAY as an evironment variable (i.e., docker run -e DISPLAY=$DISPLAY). Splendid!

Except that it doesn’t work on current Kali which is using QT4. There’s a when graphical apps are run as root and though it is fixed in QT5, we are using current Kali. And that means we are, by default, uid 0 and QT4.

I saw a bunch of workarounds that seemed to have spotty (at best) rates of success including seting QT’s graphics system to Native and giving Xorg over to root. They, mostly, seemed to be cargo cult solutions.

What made the most sense to my (generally questionable) mind was to use X forwarding. Since I had already been running Bitmessage over X forwarding from my Arch box, I knew it should work just the same.

To be completely truthful, the first pass I took at this was with Vagrant mostly because it’s SO easy. Bring up your Vagrant Box and then:
vagrant ssh -- -X
Viola!

Having proof of concept, I wanted a Docker container. The reason for this is practical. Vagrant, while completely awesome, has substantially more overhead than Docker by virtualizing the kernel. We don’t want a separate kernel running for each application. Therefore Docker is the better choice for this project.

Also, we want this whole thing to be seemless. We want to run the command bitmessage and it should fire up with minimal awkwardness and hopefully no extra steps. That is we do not want to run the Docker container then SSH into it and execute Bitmessage as individual steps. Even though that’s going to be how we begin.

The Bitmessage wiki accurately describes how to install the software so we’ll focus on the SSH setup. Though when we build the Dockerfile we will need to add SSH to the list from the wiki.

We’re going to want the container to start so that the SSH daemon is ready. Until then we can’t SSH (with X forwarding) into the container. Then we’ll want to use SSH to kick off the Bitmessage application, drawing the graphical interface using our host system’s X11.

We’re going to take advantage of Docker’s -v --volume option which allows us to specify a directory on our host system to be mounted inside our container. Using this feature, we’ll generate our SSH keys on the host and make them automatically available inside the container. We’ll tuck the keys inside the directory that Bitmessage uses for storing its configuration and data. That way Bitmessage’s configuration and stored messages can be persistent between runs — and all of your pieces are kept in a single place.

When we generate the container /etc/ssh/sshd_config is configured to allow root login without password only (i.e., using keys). So here’s how we’ll get this done:
mkdir -p ~/.config/PyBitmessage/keys #Ensure that our data directories exist
cd ~/.config/PyBitmessage/keys
ssh-keygen -b 4096 -P "" -C $"$(whoami)@$(hostname)-$(date -I)" -f docker-bitmessage-keys #Generate our SSH keys
ln -fs docker-bitmessage-keys.pub authorized_keys #for container to see pubkey


Build our container (sources available at Github and Docker) and we’ll make the script to handle Bitmessage to our preferences. #!/bin/bash
# filename: bitmessage
set -euxo pipefail

# open Docker container:
# port 8444 available, sharing local directories for SSH and Bitmessage data
# detatched, interactive, pseudo-tty (-dit)
# record container ID in $DID (Docker ID)
DID=$(docker run -p 8444:8444 -v ~/.config/PyBitmessage/:/root/.config/PyBitmessage -v ~/.config/PyBitmessage/keys/:/root/.ssh/ -dit j0rg3/bitmessage-gui bash)

# find IP address of new container, record in $DIP (Docker IP)
DIP=$(docker inspect $DID | grep IPAddress | cut -d '"' -f 4)

# pause for one second to allow container's SSHD to come online
sleep 1

# SSH into container and execute Bitmessage
ssh -oStrictHostKeyChecking=no -oUserKnownHostsFile=/dev/null -oIdentityFile=~/.config/PyBitmessage/keys/docker-bitmessage-keys -X $DIP ./PyBitmessage/src/bitmessagemain.py

# close container if Bitmessage is closed
docker kill $DID


Okay, let’s make it executable: chmod +x bitmessage

Put a link to it where it can be picked up system-wide: ln -fs ~/docker-bitmessage/bitmessage /usr/local/bin/bitmessage

There we have it! We now have a functional Bitmessage inside a Docker container. \o/

In a future post we’ll look at using eCryptfs to further protect our Bitmessage data stores.

  Project files: Github and Docker

Mucking about with a fresh copy of Kali brings to attention that it’s packaged with an Armitage that doesn’t correctly work.

I know what you’re thinking… Good. Type the commands into Msfconsole like a real man, y’uh lazy good-fer-naught! And, in practice, that was my immediate solution. But I can’t resist a good tinker when things are misbehaving.

I was anticipating that the problem would be thoroughly solved when I ixquicked it. That was partially correct. Surprised, however, when apt-get update && apt-get upgrade didn’t fix the issue. More surprised at the age of the issue. Most surprised that I could see lots of evidence that users have been plagued by this issue — but no clear work arounds were quickly found.

Guess what we’re doing today?

Okay. The issue is quite minor but just enough to be heartbreaking to the fledgling pentester trying to get a VM off the ground.

In brief, the owner of Armitage’s Github explains:

The MSF Scans feature in Armitage parses output from Metasploit’s portscan/tcp module and uses these results to build a list of targets it should run various Metasploit auxiliary modules against. A recent-ish update to the Metasploit Framework changed the format of the portscan/tcp module output. A patch to fix this issue just needs to account for the new format of the portscan/tcp module.

That is, a colon makes it into the input for the Msfconsole command to define RHOSTS. I.e.: set RHOSTS 172.16.223.150: - 172.16.223.150

An other kind coder tweaked the regex and submitted the patch and pull request, which was successfully incorporated into the project.

Sadly, things have stalled out there. So if this problem is crippling your rig, let’s fix it!

We just want a fresh copy of the project.
root@kali:~/armitage# git clone https://github.com/rsmudge/armitage

Cloning into ‘armitage’…
remote: Counting objects: 7564, done.
remote: Total 7564 (delta 0), reused 0 (delta 0), pack-reused 7564
Receiving objects: 100% (7564/7564), 47.12 MiB | 2.91 MiB/s, done.
Resolving deltas: 100% (5608/5608), done.

Kali is Debian-based and we’re going to need Apache Ant:
root@kali:~/armitage# apt-get install ant

Then, we’ll build our new fella:
root@kali:~/armitage# cd armitage
root@kali:~/armitage# ./package.sh

Buildfile: /root/test/armitage/build.xml

clean:

BUILD SUCCESSFUL
Total time: 0 seconds
Buildfile: /root/test/armitage/build.xml

init:
[mkdir] Created dir: /root/test/armitage/bin

compile:
[javac] Compiling 111 source files to /root/test/armitage/bin
[javac] depend attribute is not supported by the modern compiler
[javac] Note: /root/test/armitage/src/ui/MultiFrame.java uses or overrides a deprecated API.
[javac] Note: Recompile with -Xlint:deprecation for details.
[javac] Note: Some input files use unchecked or unsafe operations.
[javac] Note: Recompile with -Xlint:unchecked for details.

BUILD SUCCESSFUL
Total time: 2 seconds
Buildfile: /root/test/armitage/build.xml

init:

compile:

jar:
[unzip] Expanding: /root/test/armitage/lib/sleep.jar into /root/test/armitage/bin
[unzip] Expanding: /root/test/armitage/lib/jgraphx.jar into /root/test/armitage/bin
[unzip] Expanding: /root/test/armitage/lib/msgpack-0.6.12-devel.jar into /root/test/armitage/bin
[unzip] Expanding: /root/test/armitage/lib/postgresql-9.1-901.jdbc4.jar into /root/test/armitage/bin
[unzip] Expanding: /root/test/armitage/lib/javassist-3.15.0-GA.jar into /root/test/armitage/bin
[copy] Copying 4 files to /root/test/armitage/bin/scripts-cortana
[jar] Building jar: /root/test/armitage/armitage.jar
[jar] Building jar: /root/test/armitage/cortana.jar

BUILD SUCCESSFUL
Total time: 1 second
armitage/
armitage/readme.txt
armitage/teamserver
armitage/cortana.jar
armitage/armitage.jar
armitage/armitage-logo.png
armitage/armitage
armitage/whatsnew.txt
adding: readme.txt (deflated 55%)
adding: armitage.exe (deflated 49%)
adding: cortana.jar (deflated 5%)
adding: armitage.jar (deflated 5%)
adding: whatsnew.txt (deflated 65%)
armitage/
armitage/readme.txt
armitage/teamserver
armitage/cortana.jar
armitage/armitage.jar
armitage/armitage-logo.png
armitage/armitage
armitage/whatsnew.txt
Archive: ../../armitage.zip
inflating: readme.txt
inflating: armitage.exe
inflating: cortana.jar
inflating: armitage.jar
inflating: whatsnew.txt

And here, best I can guess from messages read, is where a lot of people are running into trouble. We have successfully produced our new working copy of armitage. However, it is in our own local directory and will not be run if we just enter the command: armitage

Let’s review how to figure out what we want to do about that.

First, we want to verify what happens when we run the command armitage.
root@kali:~/armitage# which armitage

/usr/bin/armitage

Good! Let’s check and see what that does!
root@kali:~/armitage# head /usr/bin/armitage

#!/bin/sh

cd /usr/share/armitage/
exec ./armitage “$@”

Almost there! It’s running /usr/share/armitage/armitage with whatever variables we’ve passed in. We’ll check that out.
root@kali:~/armitage# head /usr/share/armitage/armitage

#!/bin/sh
java -XX:+AggressiveHeap -XX:+UseParallelGC -jar armitage.jar $@

We have enough information to assemble a solution.

I trust that the people behind Kali and Armitage will get this corrected so I don’t want to suggest a solution that would replace the armitage command and prevent an updated version from running later. So, let’s just make a temporary replacement?

Hereafter, we can use the command ‘tmparmitage’ (either CLI or ALT-F2) to run our fresh version until things catch up.

And, of course, to save you the time, weary hacker:

Download here:
    armitage_quick_fix.sh

Some time ago I had begun work on my own Pastebin-type project with a few goals. Basically, I wanted to eat all the cakes — and have them too.

The project, honestly, was going swimmingly when derailed by the goings-on of life.

One of the interesting components of the project was, of course, choosing crypto implementations. There are know shortcomings to handling it in JS but that’s still the most convenient for some users. Outside of the browser, server-side, you had all the same questions about which solution was best. Which protocol(s) should be available?

Well, I’ve just learned about a project which I would have loved to have available back then. Project Wycheproof can help you test your crypto solutions against known problems and attacks. Featuring 80 tests probing at 40 known bugs, here’s a snip from the introduction:

Project Wycheproof has tests for the most popular crypto algorithms, including

AES-EAXAES-GCM

AES-GCM

DH

DHIES

DSA

ECDH

ECDSA

ECIES

RSA

The tests detect whether a library is vulnerable to many attacks, including

Invalid curve attacks

Biased nonces in digital signature schemes

Of course, all Bleichenbacher’s attacks

And many more — we have over 80 test cases

Interesting stuff with exciting potential!

Seems I’ve always just a few more things going on than I can comfortably handle. One of those is an innocent little server holding the beginnings of a new project.

If you expose a server to the Internet, very quickly your ports are getting scanned and tested. If you’ve an SSH server, there are going to be attempts to login as ‘root’ which is why it is ubiquitously advised that you disable root login. Also why many advise against allowing passwords at all.

We could talk for days about improvements; it’s usually not difficult to introduce some form of two-factor authentication (2FA) for sensitive points of entry such as SSH. You can install monitoring software like Logwatch which can summarize important points from your logs, such as: who has logged via SSH, how many times root was used, etc.

DenyHosts works primarily with SSH and asks very little from you in way of configuration, especially if you’re using a package manager to install a version that is configured for the distribution on which you’re working. If you’re installing from source you may need to find where are your SSH logs (e.g., /var/log/secure, /var/log/auth.log). It’s extremely easy to set up DenyHosts to synchronize so that you’re automatically blocking widely-known offenders whether or not they’re after your server.

In contrast, Fail2ban is going to take more work to get set up. However, it is extremely configurable and works with any log file you point it toward which means that it can watch anything (e.g., FTP, web traffic, mail traffic). You define your own jails which means you can ban problematic IP addresses according to preference. Ban bad HTTP attempts from HTTP only or stick their noses in the virtual corner and don’t accept any traffic from them until they’ve served their time-out by completely disallowing their traffic. You can even use Fail2ban to scan its own logs, so repeating offenders can be locked out for longer.

Today we’re going to assume that you’ve a new server that shouldn’t be seeing any traffic except from you and any others involved in the project. In that case, you probably want to block traffic pretty aggressively. If you’ve physical access to the server (or the ability to work with staff at the datacenter) then it’s better to err in the direction of accidentally blocking good guys than trying to be overly fault-tolerant.

The server we’re working on today is a Debian Wheezy system. It has become a common misconception that Ubuntu and Debian are, intents and purposes, interchangeable. They’re similar in many respects and Ubuntu is great preparation for using Debian but they are not the same. The differences, I think, won’t matter for this exercise but I am unsure because this was written using Wheezy.

Several minutes after bringing my new server online, I started seeing noise in the logs. I was still getting set up and really didn’t want to stop and take protective measures but there’s no point in securing a server after its been compromised. The default Fail2ban configuration was too forgiving for my use. It was scanning for 10 minutes and banning for 10 minutes. Since only a few people should be accessing this server, there’s no reason for anyone to be trying a different password every 15 minutes (for hours).

I found a ‘close-enough’ script and modified it. Here, we’ll deal with a simplified version.

First, lets create a name for these ne’er-do-wells in iptables:
iptables -N bad_traffic

For this one, we’ll use Perl. We’ll look at our Apache log files to find people sniffing ‘round and we’ll block their traffic. Specifically, we’re going to check Apache’s ‘error.log’ for the phrases ‘File does not exist’ and ‘client denied by server configuration’ and block people causing those errors. This would be excessive for servers intended to serve the general populace. For a personal project, it works just fine as a ‘DO NOT DISTURB’ sign.

That gives us some great, utterly unforgiving, blockage. Looking at the IP addresses attempting to pry, I noticed that most of them were on at least one of the popular block-lists.

So let’s make use of some of those block-lists! I found a program intended to apply those lists locally but, of course, it didn’t work for me. Here’s a similar program; this one will use ipset for managing the block-list though only minor changes would be needed to use iptables as above:

#!/bin/bash
IP_TMP=ip.tmp
IP_BLACKLIST_TMP=ip-blacklist.tmp

IP_BLACKLIST=ip-blacklist.conf

WIZ_LISTS="chinese nigerian russian lacnic exploited-servers"

BLACKLISTS=(
"http://danger.rulez.sk/projects/bruteforceblocker/blist.php" # BruteForceBlocker IP List
"http://rules.emergingthreats.net/blockrules/compromised-ips.txt" # Emerging Threats - Compromised IPs
"http://www.spamhaus.org/drop/drop.txt" # Spamhaus Don't Route Or Peer List (DROP)
"http://www.spamhaus.org/drop/edrop.txt" # Spamhaus Don't Route Or Peer List (DROP) Extended
"http://cinsscore.com/list/ci-badguys.txt" # C.I. Army Malicious IP List
"http://www.openbl.org/lists/base.txt" # OpenBL.org 90 day List
"http://www.autoshun.org/files/shunlist.csv" # Autoshun Shun List
"http://lists.blocklist.de/lists/all.txt" # blocklist.de attackers
)

for address in "${BLACKLISTS[@]}"
do
echo -e "\nFetching $address\n"
curl "$address" >> $IP_TMP
done

for list in $WIZ_LISTS
do
wget "http://www.wizcrafts.net/$list-iptables-blocklist.html" -O - >> $IP_TMP
done

wget 'http://wget-mirrors.uceprotect.net/rbldnsd-all/dnsbl-3.uceprotect.net.gz' -O - | gunzip | tee -a $IP_TMP

grep -o '^[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}[/][0-9]\{1,3\}' $IP_TMP | tee -a $IP_BLACKLIST_TMP
grep -o '^[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}\.[0-9]\{1,3\}[^/]' $IP_TMP | tee -a $IP_BLACKLIST_TMP

sed -i 's/\t//g' $IP_BLACKLIST_TMP
sort -u $IP_BLACKLIST_TMP | tee $IP_BLACKLIST

rm $IP_TMP
rm $IP_BLACKLIST_TMP

wc -l $IP_BLACKLIST

if hash ipset 2>/dev/null
then
ipset flush bloxlist
while IFS= read -r ip
do
ipset add bloxlist $ip
done < $IP_BLACKLIST
else
echo -e '\nipset not found\n'
echo -e "\nYour bloxlist file is: $IP_BLACKLIST\n"
fi


Download here:
    bad_traffic.pl
    bloxlist.sh

Last we were discussing the structure and design of your own CLI-centric blog platform, we had some crude methods of starting and resuming posts before publishing.

Today, let’s explore a little more into setting up a bloging-friendly environment because we need to either make the experience of blogging easy or we’ll grow tired of the hassle and lose interest.

We can reasonably anticipate that we won’t want to beleaguered with repetitious typing of HTML bits. If we’re going to apply paragraph tags, hyperlinks, codeblocks, etc. with any frequency, that task is best to be simplified. Using Vim as our preferred editor, we will use Tim Pope’s brilliant plug-ins ‘surround’ and ‘repeat’, combined with abbreviations to take away the tedium.

The plug-ins just need dropped into your Vim plugin directory (~/.vim/plugin/). The directory may not exist if you don’t have any plug-ins yet. That’s no problem, though. Let’s grab the plugins:

Expand the archives into the appropriate directories:

Ta-da! Your Vim is now configured to quickly wrap (surround) in any variety of markup. When working on a blog, you might use <p> tags a lot by putting your cursor amid the paragraph and typing yss<p>. The plug-in will wrap it with opening and closing paragraph tags. Move to your next paragraph and then press . to repeat.

That out of the way, let’s take advantage of Vim’s abbreviations for some customization. In our .vimrc file, we can define a few characters that Vim will expand according to their definition. For example, you might use:
ab <gclb> <code class="prettyprint lang-bsh linenums:1">
Then, any time you type <gclb> and bress <enter>, you’ll get:
<code class="prettyprint lang-bsh linenums:1">

The next time that we take a look at blogitecture, we will focus on making the posts convenient to manage from our CLI.

Recently, I ordered a Yubikey and, in the comments section of the order, I promised to write about the product. At the time, I assumed that there was going to be something about which to write: (at least a few) steps of setting up and configuration or a registration process. They’ve made the task of writing about it difficult, by making the process of using it so easy.

Plug it in. The light turns solid green and you push the button when you need to enter the key. That’s the whole thing!

Physically, the device has a hole for a keychain or it can slip easily into your wallet. It draws power from the USB port on the computer, so there’s none stored in the device, meaning it should be completely unfazed if you accidentally get it wet.

Let’s take a look at the device.

> lsusb | grep Yubico


Bus 005 Device 004: ID 1050:0010 Yubico.com Yubikey

> lsusb -v -s5:4

Bus 005 Device 004: ID 1050:0010 Yubico.com Yubikey
Couldn't open device, some information will be missing
Device Descriptor:
  bLength                18
  bDescriptorType         1
  bcdUSB               2.00
  bDeviceClass            0 (Defined at Interface level)
  bDeviceSubClass         0 
  bDeviceProtocol         0 
  bMaxPacketSize0         8
  idVendor           0x1050 Yubico.com
  idProduct          0x0010 Yubikey
  bcdDevice            2.41
  iManufacturer           1 
  iProduct                2 
  iSerial                 0 
  bNumConfigurations      1
  Configuration Descriptor:
    bLength                 9
    bDescriptorType         2
    wTotalLength           34
    bNumInterfaces          1
    bConfigurationValue     1
    iConfiguration          0 
    bmAttributes         0x80
      (Bus Powered)
    MaxPower               30mA
    Interface Descriptor:
      bLength                 9
      bDescriptorType         4
      bInterfaceNumber        0
      bAlternateSetting       0
      bNumEndpoints           1
      bInterfaceClass         3 Human Interface Device
      bInterfaceSubClass      1 Boot Interface Subclass
      bInterfaceProtocol      1 Keyboard
      iInterface              0 
        HID Device Descriptor:
          bLength                 9
          bDescriptorType        33
          bcdHID               1.11
          bCountryCode            0 Not supported
          bNumDescriptors         1
          bDescriptorType        34 Report
          wDescriptorLength      71
         Report Descriptors: 
           ** UNAVAILABLE **
      Endpoint Descriptor:
        bLength                 7
        bDescriptorType         5
        bEndpointAddress     0x81  EP 1 IN
        bmAttributes            3
          Transfer Type            Interrupt
          Synch Type               None
          Usage Type               Data
        wMaxPacketSize     0x0008  1x 8 bytes
        bInterval              10

There’s not a great deal to be seen here. As it tells you right on Yubico’s site, the device presents as a keyboard and it “types” out its key when you press the button, adding another long and complex password to combine with the long and complex password that you’re already using.

Keep in mind that this device is unable to protect you from keyloggers, some of which are hardware-based. It’s critically important that you are very, very careful about where you’re sticking your Yubikey. Even Yubico cannot protect us from ourselves.

In this writer’s opinion, it is vitally important that we take reasonable measures now to help insure anonymity, lest we create a situation where privacy no longer exists, and the simple want of, becomes suspicious.

Here’s how to configure your browser to automatically use a search engine that respects your privacy.

Chrome:

1. Click Settings.
2. Click “Set pages” in the “On startup” section.
3. Enter https://ixquick.com/eng/ in the “Add a new page” text field.
4. Click OK.
5. Click “Manage search engines…”
6. At the bottom of the “Search Engines” dialog, click in the “Add a new search engine” field.
7. Enter
   ixquick
ixquick.com
https://ixquick.com/do/search?lui=english&language=english&cat=web&query=%s
8. Click “Make Default”.
9. Click “Done”.

Firefox:

1. Click the Tools Menu.
2. Click Options.
3. Click the General tab.
4. In “When Firefox Starts” dropdown, select “Show my home page”.
5. Enter https://ixquick.com/eng/ in the “Home Page” text field.
6. Click one of the English options here.
7. Check box for “Start using it right away.”
8. Click “Add”.

Opera:

1. Click “Manage Search Engines
2. Click “Add”
3. Enter
   Name: ixquick
   Keyword: x
   Address: https://ixquick.com/do/search?lui=english&language=english&cat=web&query=%s
4. Check “Use as default search engine”
5. Click “OK”

Internet Explorer:

  _     ___  _ __        ___   _ _____ ___ 
 | |   / _ \| |\ \      / / | | |_   _|__ \
 | |  | | | | | \ \ /\ / /| | | | | |   / /
 | |__| |_| | |__\ V  V / | |_| | | |  |_| 
 |_____\___/|_____\_/\_/   \___/  |_|  (_) 

(This is not a good strategy for privacy.)

Congratulations!

You are now one step closer to not having every motion on the Internet recorded.

This is a relatively small measure, though. You can improve your resistance to prying eyes (e.g., browser fingerprinting) by using the Torbrowser Bundle, or even better, Tails, and routing your web usage through Tor, i2p, or FreeNet.

If you would like more on subjects like anonymyzing, privacy and security then drop me a line via email or Bitmessage me: BM-2D9tDkYEJSTnEkGDKf7xYA5rUj2ihETxVR

Working on remote servers, some tools are practically ubiquitous — while others are harder to come by. Even if you’ve the authority to install your preferred tools on every server you visit, it’s not always something you want to do. If you’ve hopped on to a friend’s server just to troubleshoot a problem, there is little reason to install tools that your friend is not in the habit of using. Some servers, for security reasons, are very tightly locked down to include only a core set of tools, to complicate the job of any prying intruders. Or perhaps it is a machine that you normally use through a graphical interface but on this occasion you need to work from the CLI.

These are very compelling reasons to get comfortable, at the very least, with tools like Vim, mail, grep and sed. Eventually, you’re likely to encounter a situation where only the classic tools are available. If you aren’t competent with those tools, you’ll end up facing the obstacle of how to get files from the server to your local environment where you can work and, subsequently, how to get the files back when you’re done. In a secured environment, this may not be possible without violating protocols.

Let’s take a look at how we can build a makeshift system monitor using some common tools. This particular configuration is for a server running PHP, MySQL and has the tools Htop and mytop installed. These can easily be replaced with top and a small script to SHOW FULL PROCESSLIST, if needed. The point here is illustrative, to provide a template to be modified according to each specific environment.

(Note: I generally prefer tmux to Gnu Screen but screen is the tool more likely to be already installed, so we’ll use it for this example.)

We’re going to make a set of windows, by a configuration file, to help us keep tabs on what is happening in this system. In so doing, we’ll be using the well-known tools less and watch. More specifically, less +F which tells less to “scroll forward”. Other words, less will continue to read the file making sure any new lines are added to the display. You can exit this mode with CTRL+c, search the file (/), quit(q) or get back into scroll-forward mode with another uppercase F.

Using watch, we’ll include the “-d” flag which tells watch we want to highlight any changes (differences).

We will create a configuration file for screen by typing:

> vim monitor.screenrc

In the file, paste the following:

# Screen setup for system monitoring
# screen -c monitor.screenrc
hardstatus alwayslastline
hardstatus string '%{= kG}[ %{G}%H %{g}][%= %{=kw}%?%-Lw%?%{r}(%{W}%n*%f%t%?(%u)%?%{r})%{w}%?%+Lw%?%?%= %{g}][%{B}%Y-%m-%d %{W}%c %{g}]'

screen -t htop 0 htop
screen -t mem 1 watch -d "free -t -m"
screen -t mpstat 2 watch -d "mpstat -A"
screen -t iostat 3 watch -d "iostat"
screen -t w 4 watch -d "w"
screen -t messages 5 less +F /var/log/messages
screen -t warn 6 less +F /var/log/warn
screen -t database 7 less +F /srv/www/log/db_error
screen -t mytop 8 mytop
screen -t php 9 less +F /srv/www/log/php_error


(Note: -t sets the title, then the window number, followed by the command running in that window)

Save the file (:wq) or, if you’d prefer, you can grab a copy by right-clicking and saving this file.

Then we will execute screen using this configuration, as noted in the comment:

> screen -c monitor.screenrc

Then you can switch between windows using CTRL+a, n (next) or CTRL+a, p (previous).

I use this technique on my own computers, running in a TTY different from the one used by X. If the graphical interface should get flaky, I can simply switch to that TTY (e.g., CTRL+ALT+F5) to see what things are going on — and take corrective actions, if needed.

There may be times when you find your Git repository burdened with scads of untracked files left aside while twiddling, testing bug patches, or what-have-youse.

For the especially scatter-brained among us, these things can go unchecked until a day when the useful bits of a git status scroll off the screen due to utterly unimportant stuff. Well, hopefully unimportant.

But we’d better not just cleave away everything that we haven’t checked in. You wonder:
What if there’s something important in one of those files?

You are so right!

Let’s fix this!

Firstly, we want a solution that’s reproducible. Only want to invent this wheel once, right?

Let’s begin with the play-by-play:

Git, we want a list of what isn’t tracked: git ls-files -o --exclude-standard -z

We’ll back these files up in our home directory (~), using CPIO but we don’t want a poorly-named directory or finding anything will become its own obstacle. So we’ll take use the current date (date +%Y-%m-%d), directory (pwd) and branch we’re using (git branch) and we’ll twist all of it into a meaningful, but appropriate, directory name using sed. git ls-files -o --exclude-standard -z | cpio -pmdu ~/untracked-git-backup-`date +%Y-%m-%d`.`pwd | sed 's,^\(.*/\)\?\([^/]*\),\2,'`.`git branch | grep "*" | sed "s/* //"`/

Ahhhh… Much better. Is there anything left out? Perhaps. What if we decide that moving these files away was a mistake? The kind of mistake that breaks something. If we realize right away, it’s easily-enough undone. But what if we break something and don’t notice for a week or two? It’d probably be best if we had an automated script to put things back the way they were. Let’s do that.

Simple enough. We’ll just take the opposite commands and echo them into a script to be used in case of emergency.

Create the restore script (restore.sh), to excuse faulty memory: echo "(cd ~/untracked-git-backup-`date +%Y-%m-%d`.`pwd | sed 's,^\(.*/\)\?\([^/]*\),\2,'`.`git branch | grep "*" | sed "s/* //"`/; find . -type f \( ! -iname 'restore.sh' \) | cpio -pdm `pwd`)" > ~/untracked-git-backup-`date +%Y-%m-%d`.`pwd | sed 's,^\(.*/\)\?\([^/]*\),\2,'`.`git branch | grep "*" | sed "s/* //"`/restore.sh

Make the restore script executable: chmod u+x ~/untracked-git-backup-`date +%Y-%m-%d`.`pwd | sed 's,^\(.*/\)\?\([^/]*\),\2,'`.`git branch | grep "*" | sed "s/* //"`/restore.sh

Lastly, the magic, compressed into one line that will stop if any command does not report success: a='untracked-git-backup-'`date +%Y-%m-%d`.`pwd | sed 's,^\(.*/\)\?\([^/]*\),\2,'`.`git branch | grep "*" | sed "s/* //"`; git ls-files -o --exclude-standard -z | cpio -pmdu ~/$a/ && git clean -d -f && echo "(cd ~/$a/; find . -type f \( ! -iname 'restore.sh' \) | cpio -pdm `pwd`)" > ~/$a/restore.sh && chmod +x ~/$a/restore.sh; unset a

Documentation for this one seems a bit hard to come by but it is one of the things I love about Zsh.

I’ve seen many .bashrc files that have things like:
alias www='cd /var/www'
alias music='cd /home/j0rg3/music'


And that’s a perfectly sensible way to make life a little easier, especially if the paths are very long.

In Zsh, however, we can use the hash command and the shortcut we get from it works fully as the path. Other words, using the version above, if we want to edit ‘index.html’ in the ‘www’ directory, we would have to issue the shortcut to get there and then edit the file, in two steps:
> www
> vim index.html

The improved version in .zshrc would look like:
hash www=/var/www
hash -d www=/var/www


Then, at any time, you can use tilde (~) and your shortcut in place of path.
> vim ~www/index.html

Even better, it integrates with Zsh’s robust completions so you can, for example, type cd ~www/ and then use the tab key to cycle through subdirectories and files.

On this system, I’m using something like this:
(.zshrc)
hash posts=/home/j0rg3/weblog/posts
hash -d posts=/home/j0rg3/weblog/posts

Then we can make a function to create a new post, to paste into .zshrc. Since we want to be able to edit and save, without partial posts becoming visible, while we are working, we’ll use an extra .tmp extension at the end:
post() { vim ~posts/`date +%Y-%m`/`date +%Y%m%d`.$1.txt.tmp }

But, surely there is going to be a point when we need to save a post and finish it later. For now, let’s assume that only a single post will be in limbo at any time. We definitely don’t want to have to remember the exact name of the post — and we don’t want to have hunt it down every time.

We can make those things easier like this:
alias resume="vim `find ~posts/ -name '*.txt.tmp'`"

It’s generally a great idea to have Vim keep backups. Once in awhile, they can really save your bacon.

The other side of that coin, though, is that they can get left behind here and there, eventually causing aggravation.

Here’s a snippet to find and eliminate those files from the current directory down:

This uses find from the current directory down (./) to execute an rm statement on all files with an extension ending in tilde (~)

Thanks for visiting my little spot on the web. This is a Blosxom ‘blog which, for those who don’t know, is a CGI written in Perl using the file-system (rather than a database).

To the CLI-addicted, this is an awesome little product. Accepting, of course, that you’re going to get under the hood if you’re going to make it the product you want. After some modules and hacking, I’m pleased with the result.

My posts are just text files, meaning I start a new one like: vim ~posts/`date +%Y%m%d`.brief.subject.txt

Note: the back-ticks (`) tell the system that you want to execute the command between ticks, and dynamically insert its output into the command. In this case, the command date with these parameters:

1. (+) we’re going to specify a format
2. (%Y) four-digit year
3. (%m) two-digit month
4. (%d) two-digit day

That means the command above will use Vim to edit a text file named ‘20130507.brief.subject.txt’ in the directory I have assigned to the hash of ‘posts’. (using hash this way is a function of Zsh that I’ll cover in another post)

Toss in a line that identifies tags and, since Perl is the beast of Regex, we pick up the tags and make them links, meta-tags, etc.

Things here are likely to change a lot at first, while I twiddle with CSS and hack away at making a Blosxom that perfectly fits my tastes — so don’t be too alarmed if you visit and things look a tad wonky. It just means that I’m tinkering.

Once the saw-horses have been tucked away, I’m going to take the various notes I’ve made during my years in IT and write them out, in a very simple breakdown, aimed at sharing these with people who know little about how to negotiate the command line. The assumption here is that you have an interest in *nix/BSD. If you’ve that and the CLI is not a major part of your computing experience, it probably will be at some point. If you’re working on systems remotely, graphical interfaces often just impede you.

Once you’ve started working on remote machines, the rest is inevitable. You can either remember how to do everything two ways, through a graphical interface and CLI — or just start using the CLI for everything.

So let’s take a little journey through the kinds of things that make me love the CLI.