blog.crox.net

Let's say you want some output to be echoed to the console but also to be passed to a pipe. Specifically, in my case I wanted a message to be displayed to the user and also to be sent to syslog. You could of course just print it twice, but in some cases it's not possible, or it makes it more complicated than it should be.

There is a simple solution with tee:

echo "a test" | tee >(logger)

tcpdump and wireshark are the tools that usually come to mind when you have to capture network traffic. But in some situations where you have to record a large amount of data and you want to avoid losing packets, tcpdump has some limitations. When I was hit myself by the tcpdump packet loss problem, I quickly found out that I was not alone and that a number of people had already researched the topic and/or provided alternatives.*

In particular, I found two different tools to perform the task: Corey Satten's gulp (http://corey.elsewhere.org/gulp/) and lindump from HP Labs (http://tesla.hpl.hp.com/opensource/)

I also found two interesting papers about capturing high volumes of traffic: http://www.usenix.org/events/fast09/tech/full_papers/anderson/anderson_html/ and http://docs.di.fc.ul.pt/jspui/bitstream/10455/3299/1/thesis-nhenriqu.pdf (the second quotes the first one among others, and also contains useful info to optimally spread the load among different cores)

After some tests I quickly became a happy gulp user, and thanks to the software being open source I was able to add features to it that I missed from the latest tcpdump versions:

-n - allows to change the default filename template
-t - allows to add a timestamp to the filename
-G - rotate pcap file every n seconds
-F - allows to skip the check for an ethernet interface
-Z - allows to specify a command to post-process each capture file

I've sent a patch to Corey Satten, who intends to setup a repository to hold the various contributions he gets for gulp. In the meanwhile, you can find my changes in the attached file (02-gulp-ntGFZ.patch.gz). For your convenience and for completeness, I also provide here the patch from Guy Harris that fixes issues on 64 bit systems (see http://seclists.org/wireshark/2009/Oct/105, apply that one first).

Updates:
2012-08-22: new version of my patch to fix the issue reported by SgtMalicious
2017-02-03: long-standing bug fixed

Downloads:
01-gulp-amd64.patch.gz: fix issues with 64-bit systems
02-gulp-ntGFZ.patch.gz: additional functionality as described above
gulp-1.58-crox.tgz: source with both patches applied

* other people have reported a performance drop with libpcap version 1.0 compared to previous builds, see http://thread.gmane.org/gmane.network.tcpdump.devel/4629 or http://seclists.org/tcpdump/2010/q3/index.html#11

1. create and enable a "regular" swap partition (fdisk / mkswap / swapon)

2. install ecryptfs-utils and run ecryptfs-setup-swap

sudo apt-get install ecryptfs-utils
sudo ecryptfs-setup-swap



                

When performing tests you may sometimes want to send specially crafted icmp packets. hping is a handy tool for that.

However, the default behavior is to refuse to send "unsupported" Type/Code combinations. eg

hping3 -c 1 --icmp -C 33 -K 0 192.168.70.1
HPING 192.168.70.1 (wlan0 192.168.70.10): icmp mode set, 28 headers + 0 data bytes
[send_icmp] Unsupported icmp type!

Fortunately, there is an (undocumented) --force-icmp option that you can add to bypass the check:

hping3 -c 1 --icmp --force-icmp -C 33 -K 0 192.168.70.1
HPING 192.168.70.1 (wlan0 192.168.70.10): icmp mode set, 28 headers + 0 data bytes

--- 192.168.70.1 hping statistic ---
1 packets transmitted, 0 packets received, 100% packet loss
round-trip min/avg/max = 0.0/0.0/0.0 ms

In Ubuntu Karmic (and possibly Jaunty ?), when you install libsane, it adds a file /lib/udev/rules.d/40-libsane.rules which contains rules that match on supported scanners and set the environment variable "libsane_matched" to "yes".

This in turn triggers the following in /lib/udev/rules.d/70-acl.rules:

# USB scanners
ENV{libsane_matched}=="yes", ENV{ACL_MANAGE}="1"

<snip>

# apply ACL for all locally logged in users
LABEL="acl_apply", ENV{ACL_MANAGE}=="?*", TEST=="/var/run/ConsoleKit/database", \
  RUN+="udev-acl --action=$env{ACTION} --device=$env{DEVNAME}"

In the end, the result is that an ACL is created for the device, which allows locally logged in users to use it (read/write permission). eg for my scanner:

~# lsusb
Bus 001 Device 005: ID 04b8:011c Seiko Epson Corp. Perfection 3200
~# ls -l /dev/bus/usb/001/005
crw-rw-r--+ 1 root root 189, 4 2009-12-28 00:11 /dev/bus/usb/001/005
~# getfacl /dev/bus/usb/001/005
getfacl: Removing leading '/' from absolute path names
# file: dev/bus/usb/001/005
# owner: root
# group: root
user::rw-
user:crox:rw-
group::rw-
mask::rw-
other::r--
~#

However, I also wanted to allow access to the scanner from other workstations through saned. In older Ubuntu versions, you could just add saned (or whatever user the service runs as) to the scanner group. This no longer works since the device belongs to root:root, and ACLs are added for specific users. The solution that works for me is to create a file /etc/udev/rules.d/99-sane-group.rules with the following contents:

# change group to scanner for sane devices
ENV{libsane_matched}=="yes", GROUP="scanner"

Then you just need to run

sudo udevadm trigger

and the group of the device magically changes to scanner.

Of course you could also add a similar rule specifically for a certain device instead, in my case this would work too:

ATTRS{idVendor}=="1d6b", ATTRS{idProduct}=="0002", GROUP="scanner"

To check that it worked, run the following:

sudo su -s /bin/bash -c 'scanimage -L' saned