User:needle
Meet me in the #gentoo IRC support channel on libera.chat
SMW
This wiki has SMW support. Special:Version#mw-version-ext
- create easy examples for demonstration.
Create simple example on Logging table - available software using SMW.
SHA-512 to Blowfish migration testing
Before reading furhter. Read about bcrypt
rounds
or cost
, take a close look at the generated salt
. Use a professional password when using bcrypt.This is a test, done before migration. If the test succeeds, the target system is considered safe for migration. The steps described here, should work for every other user that is to be migrated. This short description is meant for nodes only with few users.
Take sure your target system supports Blowfish, apparently it is not available everywhere on every linux:
user $
man 3 crypt
Over here it looks like in depicted table, this man 3 crypto
ID Method ──────────────────────────────────────────────────────────────────────────── 1 MD5 2a Blowfish (not in mainline glibc; added in some Linux distributions) 5 SHA-256 (since glibc 2.7) 6 SHA-512 (since glibc 2.7)
Create a test user here larry the target system:
root #
useradd -m -G users,wheel,audio -s /bin/bash larry
Emerge
Review and set USE flags before emerging the package, enable the bcrypt USE flag
root #
emerge --ask sys-apps/shadow
Verify the package sys-apps/shadow has the
bcrypt
USE flag enabled. And the package has been rebuild using that flag.
Generate
Now, configure bcrypt to create really safe password hashes.
If you are expert in fastfood security done is 5 seconds, here a quick cli to generate a hash:
user $
htpasswd -bnBC 15 "" G3n70o_L1nuX-r0ck5?! | tr -d ':\n'
$2y$15$ibqikJGVNIsDx3LcQF0DduUaa0ropb9wG8bbEkEHWIqPtD3T52cQK%
Generated prefix, here $2y$ is NOT interesting. There is NO difference between $2a$ or $2y$ in any sense. If implemented correctly, all created hashes are working the same. No matter which prefix, is generated using Blowfish.
Replace
Now using a text editor edit the /etc/shadow file
root #
vim /etc/shadow
And replace following test user created SHA-512 hash
... larry:$6$W2LZ5IsI$KVrGRLf7YbTPKA.t/4gvwOr4wtHBdvF6DYpSV93ZvkdkNy0qZFu0VMt7Igy7EzW8GIEED8tVdD5vq2/HpMn7b0:16134:0:99999:7::: ...
With this generated bcrypt hash, notice the cost of bcrypt, it is not the real time of rounds depicted below, it is fake here for the examle:
... larry:$2y$15$ibqikJGVNIsDx3LcQF0DduUaa0ropb9wG8bbEkEHWIqPtD3T52cQK:16134:0:99999:7::: ...
Verify
Open a SSH connection to localhost, using that changed test username, here larry:
user $
ssh larry@localhost
If authentication succeeds, then you are ready to plan the migration, on working users.
Clean up
After the testing is finished, remove larry from the system
root #
userdel larry
etckeeper whitelist configuration
Create a .gitignore file before running the etckeeper initialization.
This example below shows how to save explicit files to the etckeeper repository. This is the reverse approach, compared to the default etckeeper configuration, which
The first entry *
ignores ALL files in the /etc directory, following !
negated entries mark the interesting files that will be saved to the repository.
Configuration example for saving these 3 files:
- /etc/crontab
- /etc/inittab
- /etc/resolv.conf
/etc/.gitignore
# ignore everything
*
# now add interesting files
!crontab
!inittab
!resolv.conf
Configuration example for /etc/apache2/ directory:
/etc/.gitignore
# ignore everything
*
# now add interesting files and dirs
!apache2/
!apache2/*
Configuration example for directories containing subdirectcories with interesting files:
- /etc/apache2/httpd.conf
- /etc/apache2/vhosts.d/
- /etc/apache2/modules.d/
/etc/.gitignore
# ignore everything
*
# now add interesting files and dirs
!apache2/
!apache2/httpd.conf
!apache2/modules.d/
!apache2/modules.d/*
!apache2/vhosts.d/
!apache2/vhosts.d/*
Automated rebuild of portage packages
This solution relies on the sys-process/cronie and the usage of anacron USE flag.
What does that anacron USE flag do, verify using the euse tool:
user $
euse -i anacron
[- ] anacron (sys-process/cronie): Install the periodic anacron command scheduler.
it is a added feature or function to cronie. The anacron USE flag re-schedules missed cron jobs for machines that are not 24/7 online, like f.e. laptops, workstations. Apart from that it is working like an usual cron scheduler. This feature does not rely on the separate anacron package.
Install cronie:
root #
emerge --ask sys-process/cronie
Schedule daily rebuild by adding following file in the /etc/cron.daily/ directory:
user:Sam suggested sanity checks are missing the pre-upgrade and post-emerge routines are not handled by this script. Read Portage_log and elogv for final solution.
/etc/cron.daily/portage
#!/bin/sh
#
# Sync portage using eix-sync
# -U Do not touch the database, do not show differences
# -T Do not measure time
/usr/bin/eix-sync -U -T
if [ $? -eq 0 ]; then
logger "eix-sync has finished."
else
logger "eix-sync has exited with error code: $?"
fi
# Now update the database
/usr/bin/eix-update
if [ $? -eq 0 ]; then
logger "eix-update has finished."
else
logger "eix-update has exited with error code: $?"
fi
# Emerge world packages. For skipping bugged ebuild,
# add "EMERGE_DEFAULT_OPTS= --keep-going"* to make.conf file
emerge -uDN @world
if [ $? -eq 0 ]; then
logger "emerge --world has finished."
else
logger "emerge --world has exited with error code: $?"
fi
# And keep everything working
emerge @preserved-rebuild
if [ $? -eq 0 ]; then
logger "emerge @preserved-rebuild has finished."
else
logger "emerge @preserved-rebuild has exited with error code: $?"
fi
# Write a message to syslog portage rebuild has finished now.
logger "daily cron portage update has finished with exit status: $?"
Make the /etc/cron.daily/portage file executable by adding the +x flag:
root #
chmod +x /etc/cron.daily/portage
This will schedule run the eix-sync and ebuild jobs, at ~03:00 AM. And if the job has been missed because host was turned off, the job gets scheduled after a the host has been turned on again.
Now it would be nice to see, what has been rebuild and how it worked out without using any complex commands. Add a bash script to the ~/bin directory of the root user. The script runs 2 qlop commands showing the results from beginning of the day. qlop is part of app-portage/portage-utils ebuild.
/root/bin/emergelog.sh
#!/bin/sh
qlop -H -s -d '1 day ago'
qlop -H -m -u -d '1 day ago'
Make the /root/bin/emergelog.sh file executable by adding the +x flag:
root #
chmod +x ~/bin/emergelog.sh
Because I am even to tazy to run that command manually, I add following lines to the /root/.profile file, this calls the upper /root/bin/emergelog.sh, each time the root user authenticates to this host. Additionally this lists the last 8 lines of the /var/log/emerge.log file:
/root/.profile
echo "Last emerged packages:"
sh ~/bin/emergelog.sh
echo ""
echo "Last emerge.log entries:"
tail -n 8 /var/log/emerge.log
echo ""
And this is the prompt how it looks like after successful authentication:
root #
Last emerged packages:
2020-04-21T02:30:13 *** gentoo
2020-04-21T03:22:44 >>> dev-util/re2c
2020-04-21T03:32:45 >>> net-misc/whois
2020-04-21T08:20:39 >>> dev-libs/libpcre2
Last emerge.log entries:
1587455666: *** Finished. Cleaning up...
1587455669: *** exiting successfully.
1587455669: *** terminating.
1587455675: Started emerge on: Apr 21, 2020 09:54:34
1587455675: *** emerge --keep-going @preserved-rebuild
1587455687: *** Finished. Cleaning up...
1587455690: *** exiting successfully.
1587455691: *** terminating.
This is a optional step and could be useful on always on systems, for the update routine to be sane. If the update routine has changed the daemons libriaries, that particular service would need a unattended restart. To accomplish ths use following tool:
root #
emerge --ask app-admin/needrestart
Default configuration needs to be adjusted to the own system. Do not rely on defaults. app-admin/needstart needs further configuration. The list of all configuration files:
user $
tree /etc/needrestart
This solution works 99% of the time. If packages fail to build, this needs to be resolved manually by doing a world rebuild, and inspecting what has gone wrong.
laptop_mode laptop roaming howto
Additional software
These packages are all needed to get it to run:
- openrc — a dependency-based init system for Unix-like systems that maintains compatibility with the system-provided init system - with enabled USE flag netifrc
- wpa_supplicant — a Wi-Fi supplicant
- sys-apps/ifplugd - Brings up/down ethernet ports automatically with cable detection
- dhcpcd — a popular DHCP client capable of handling both IPv4 and IPv6 configuration.
- app-laptop/laptop-mode-tools - Linux kernel laptop_mode user-space utilities
Configuration
OpenRC
Managing daemon status and interfaces reflecting the current powerlevel AC or running on battery. This can be accomplished by using sys-apps/openrc. OpenRC configuration and management is more complex compared to the app-laptop/laptop-mode-tools configuration approach, but also much more flexible. A simplified openrc configuration is needed. Dynamic services are handled by laptop-mode-tools. Overview of running daemons handled by openrc runlevel default, note laptop_mode daemon is started here:
user $
rc-status default
Runlevel: default lm_sensors [ started ] sysklogd [ started ] sensord [ started ] alsasound [ started ] acpid [ started ] cupsd [ started ] cronie [ started ] chronyd [ started ] laptop_mode [ started ] local [ started ] sshd [ started ]
Following daemons need to be managed by laptop-tools:
user $
rc-status default
net.eth0 [ started ] net.wlan0 [ started ] sshd [ started ] cupsd [ started ]
Remove the daemons from the openrc default startup level
root #
rc-update del net.eth0 default
root #
rc-update del net.wlan0 default
root #
rc-update del net.sshd default
root #
rc-update del net.cupsd default
Verify the default startup of openrc:
user $
rc-status default
Runlevel: default lm_sensors [ started ] sysklogd [ started ] sensord [ started ] alsasound [ started ] acpid [ started ] cronie [ started ] chronyd [ started ] laptop_mode [ started ] local [ started ]
Laptop-mode-tools
The laptop_mode tools dynamic configuration relies on default 2 ACPI levels:
- laptop is running on AC power
- laptop is running on battery
laptop-mode tools has 2 according ACPI states named batt and lm-ac:
- batt
- lm-ac
- nolm-ac
The 3-rd state nolm-ac (laptop-mode tools daemon NOT running) is not used.
Get an overview of the laptop-mode directory:
user $
tree -L 1 /etc/laptop-mode
/etc/laptop-mode ├── batt-start ├── batt-stop ├── conf.d ├── laptop-mode.conf ├── lm-ac-start ├── lm-ac-stop ├── lm-profiler.conf ├── modules ├── nolm-ac-start └── nolm-ac-stop
Each of the 3 predefined states batt lm-ac and nolm-ac have a -start and -stop suffix in the directory structure. There is also a conf.d directory for services configuration that would be handled by laptop-mode and a modules directory for modules to be used explicitelly.
The goal is reached when the laptop automatically determines which daemons need to be started and which need to be stopped depending on the ACPI battery level.
There are 2 states in which the laptop is working:
- laptop is docked, ac connected, wired access, printing available, ssh daemon running
- laptop is not docked, battery, wireless access, no priting available, no ssh daemon runni
Adjust the previosly removed daemons to laptop-mode. Change to the battery level. wlan is the only one service needed while running on battery.
Change to the directory:
root #
cd /etc/laptop-mode/batt-start/
Create a symlink to daemons to be run while on battery:
root #
ln -s /etc/init.d/net.wlan0 .
Change to the directory /etc/laptop-mode/batt-stop/:
root #
cd /etc/laptop-mode/batt-stop/
Create a symlink to deamons to be stopped while on battery:
root #
ln -s /etc/init.d/cupsd .
root #
ln -s /etc/init.d/net.eth0 .
root #
ln -s /etc/init.d/sshd .
Start and Stop daemons handled by the battery status after configuration:
user $
tree -L 1 /etc/laptop-mode/batt-st*
tree -L 1 /etc/laptop-mode/batt-st* /etc/laptop-mode/batt-start └── net.wlan0 -> /etc/init.d/net.wlan0 /etc/laptop-mode/batt-stop ├── cupsd -> /etc/init.d/cupsd ├── net.eth0 -> /etc/init.d/net.eth0 └── sshd -> /etc/init.d/sshd
Start and Stop daemons handled by the battery status:
user $
tree -L 1 /etc/laptop-mode/lm-ac-st*
/etc/laptop-mode/lm-ac-start ├── cupsd -> /etc/init.d/cupsd ├── net.eth0 -> /etc/init.d/net.eth0 └── sshd -> /etc/init.d/sshd /etc/laptop-mode/lm-ac-stop └── net.wlan0 -> /etc/init.d/net.wlan0
Verification
Docked laptop and service status:
user $
rc-status default
Runlevel: default lm_sensors [ started ] sysklogd [ started ] sensord [ started ] alsasound [ started ] mpd [ started ] acpid [ started ] cronie [ started ] chronyd [ started ] laptop_mode [ started ] local [ started ] Dynamic Runlevel: hotplugged Dynamic Runlevel: needed/wanted net.eth0 [ started ] cupsd [ started ] Dynamic Runlevel: manual sshd [ started ]
Undocked status:
user $
rc-status default
Runlevel: default lm_sensors [ started ] sysklogd [ started ] sensord [ started ] alsasound [ started ] mpd [ started ] acpid [ started ] cronie [ started ] chronyd [ started ] laptop_mode [ started ] local [ started ] Dynamic Runlevel: hotplugged Dynamic Runlevel: needed/wanted net.wlan0 [ started ]
This is a ready, easy to use, running configuration.
Enable IPv6 privacy extensions (RFC4941)
IPv6 privacy extensions are disabled by default on GNU/linux, they lead to problems if users are not aware of this. To enable privacy extensions on gentoo permanently add following lines and reboot the system:
/etc/sysctl.conf
Enabling IPv6 privacy extensions...
# Enabling IPv6 privacy extensions for specified interfaces.
# here eth0 and wlan0
# net.ipv6.conf.eth0.use_tempaddr = 2
# net.ipv6.conf.wlan0.use_tempaddr = 2
net.ipv6.conf.all.use_tempaddr = 2
net.ipv6.conf.default.use_tempaddr = 2
# Setting q shorter timeout for a temporary IPv6 prefix
# default setting is one day
net.ipv6.conf.eth0.temp_prefered_lft = 14400
net.ipv6.conf.wlan0.temp_prefered_lft = 14400
The setting net.ipv6.conf.all.use_tempaddr is used to propagate its value to all interfaces currently attached to the system. This setting might not work reliably for all interfaces. At least not on my own tested gentoo installations up to kernel 4.14.
There are two old bugs in the Linux kernel bugtracker for this issue:
Wiki templates for templates
Command sequence for old gentoo boxes to update after a long time
This is a basic sequence of commands for updating older gentoo boxes.
Divide and conquer: Update the toolchain first, then resolve the blocks manually afterwards.
Sync portage tree:
root #
eix-sync
Update the portage application:
root #
emerge --oneshot portage
Sync portage again:
root #
eix-sync
Emerge GCC first:
root #
emerge --oneshot gcc
Show availabe GCC compilers:
root #
eselect gcc list
Set the latest available GCC compiler in the list:
root #
eselect gcc set <input>
Check if the desired GCC has been set, apply portage postinstall hint:
root #
eselect gcc list
Emerge latest glibc
root #
emerge --oneshot glibc
Check the latest gentoo related toolchain changes on the wiki, bugs, etc: Project:Toolchain
Emerge latest linux kernel:
root #
emerge gentoo-sources
Show available kernel sources:
root #
eselect kernel list
Set the latest linux kernel version:
root #
eselect kernel set <input>
Emerge latest binutils:
root #
emerge binutils
Show current available binutils:
root #
eselect binutils list
Set the latest binutils version:
root #
eselect binutils set <input>
Verify binutils setting:
root #
eselect binutils list
Emerge latest python
root #
emerge --oneshot python
Emerge latest perl:
root #
emerge --oneshot perl
Emerge latest iproute2
root #
emerge --oneshot iproute2
Update the system with following command, resolve dependency errors:
root #
emerge -vauDN system
Update the system with following command, resolve dependency errors:
root #
emerge -vauDN world
Now it is done.