This guide helps to configure the hibernation on a default Fedora35 (also worked fine in previous Fedora34) installation by using a swap file. The Fedora35 installation comes with btrfs as default filesystem. Also, it comes with a zram swap device:
$ swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 8G 0B 100
This device reserves a physical memory area in which all the content will be compressed (at its input) and uncompressed (at its output). But, we cannot hibernate with this type of swap space as it is only in memory.
After feedback received from contributors of this guide, seems we can say that this guide will work in both, encrypted and unencrypted setups.
important note before you continue: Some of us are experimenting problems with nvidia proprietary drivers. See below comments for more information on the investigation.
Nowadays we have lots of RAM in our laptops. In my case, a guy with a laptop for programming, i only use swap space for hibernation. When i move from one location to another, i dont want to loose all my open stuff nor consuming battery while i am moving.
I want to preserve the zram swap device of the default configuration.
This are the steps we need to perform in our system:
- The hibernation is triggered by the user.
- The swap file is activated.
- The zram device is deactivated. If there are any memory pages present in zram, they will be moved to the activated (2) swap file.
- Hibernate the laptop.
And the following sequence for resuming:
- Power on the computer.
- Restore system state from the swap file at boot time.
- Activate the zram device.
- Deactivate the swap file. That could cause the zram device to start compressing data.
Seems there are efforts in order to make this much better, even with dynamic sized swap files, so only generating swap files for hibernation with the needed size on each moment, and more user friendly.
This guide will make use of a fixed size swap file, since probably for going further is better to just contribute in the mentioned systemd issue .
Default Fedora35 installations comes with btrfs as default filesystem. Such file system comes with the subvolume feature. Subvolumes are not partitions. They are just a logical separations of a filesystem at a file level. Some kind of operations like snapshots, will try to include the swap file. We need to prevent this by isolating the swap file into its own subvolume:
btrfs subvolume create /swap
The above will ensure the swap file will not be taken into account in other snaphots as they are not recursive in other subvolumes.
Next is to create our swapfile for hibernation. Ensure the specified size is enough for saving the contents of the RAM + the uncompressed contents of the zram swap space. How to determine the size ? that might be a rough estimation per use case. As root:
touch /swap/swapfile
chattr +C /swap/swapfile ## Needed to disable Copy On Write on the file.
fallocate --length 33GiB /swap/swapfile ## Please, calculate your size as mentioned in above comments.
chmod 600 /swap/swapfile
mkswap /swap/swapfile
Now lets prepare the path of the resume operation. We need add to the initramfs
the necessary modules for resuming the system. We need to create a new file at /etc/dracut.conf.d/resume.conf
with the following content:
add_dracutmodules+=" resume "
After that we need to regenerate the initramfs
:
dracut -f
Next, we have to add the resume
and resume_offset
into the GRUB_CMDLINE_LINUX
, so that Grub can instruct the kernel coordinates where the swap file resides, in order to resume the system.
For gathering the resume
param we need the partition UUID in which the swap file is stored:
$ findmnt -no UUID -T /swap/swapfile
dbb0f71f-8fe9-491e-bce7-4e0e3125ecb8
Now lets gather the last needed data, we need the resume_offset
. That is, the physical offset of our swap file in the file system. For doing that, we need to follow this guide.
Now we need to instruct GRUB to initialize the kernel with this coordinates. Edit the /etc/defaut/grub
and grab there the parameters we just calculated:
GRUB_CMDLINE_LINUX="rd.luks.uuid=luks-4369a407-2be1-4f37-9764-ff848a0f2089 resume=UUID=dbb0f71f-8fe9-491e-bce7-4e0e3125ecb8 resume_offset=2459934 rhgb quiet"
Now lets re-configure the grub (UEFI setup assumed):
grub2-mkconfig -o /boot/grub2/grub.cfg
Note: We only want the swap file to be used when the hibernation takes place and in the resume stage. we are not going to configure fstab
entries. To achieve that, one of the options is to use systemd
. We are going to configure 2 systemd services. One for preparing the hibernation and the other one for resuming it:
For enabling the swap file and disabling the zram swap device before hibernation, lets create the file /etc/systemd/system/hibernate-preparation.service
:
[Unit]
Description=Enable swap file and disable zram before hibernate
Before=systemd-hibernate.service
[Service]
User=root
Type=oneshot
ExecStart=/bin/bash -c "/usr/sbin/swapon /swap/swapfile && /usr/sbin/swapoff /dev/zram0"
[Install]
WantedBy=systemd-hibernate.service
The order is important in the above service definition. First of all we enable the swap file, which should have enough space to store the contents of the "in use RAM", plus the contents of the uncompressed zram swap device. Secondly, we disable the zram swap device. At that moment, the kernel will start moving all the memory pages from the zram swap device to the swap file if needed. After all of that, the hibernation will take place only in the swap file. Last step for the above script to have effect is to install this service in systemd
:
# systemctl enable hibernate-preparation.service
Created symlink /etc/systemd/system/systemd-hibernate.service.wants/hibernate-preparation.service → /etc/systemd/system/hibernate-preparation.service.
We need to disable the swap file when the system just resumed. Lets create the /etc/systemd/system/hibernate-resume.service
file:
[Unit]
Description=Disable swap after resuming from hibernation
After=hibernate.target
[Service]
User=root
Type=oneshot
ExecStart=/usr/sbin/swapoff /swap/swapfile
[Install]
WantedBy=hibernate.target
Then enable it by:
# systemctl enable hibernate-resume.service
Created symlink /etc/systemd/system/hibernate.target.wants/hibernate-resume.service → /etc/systemd/system/hibernate-resume.service.
In order to make the suspend-then-hibernate
sequence to work, please take a look at this great comment and follow the instructions there.
In order to avoid false positives regarding to swap space, due to the zram device existence, we need to disable some checks:
mkdir -p /etc/systemd/system/systemd-logind.service.d/
cat <<-EOF | sudo tee /etc/systemd/system/systemd-logind.service.d/override.conf
[Service]
Environment=SYSTEMD_BYPASS_HIBERNATION_MEMORY_CHECK=1
EOF
mkdir -p /etc/systemd/system/systemd-hibernate.service.d/
cat <<-EOF | sudo tee /etc/systemd/system/systemd-hibernate.service.d/override.conf
[Service]
Environment=SYSTEMD_BYPASS_HIBERNATION_MEMORY_CHECK=1
EOF
Now you must reboot your computer, in order the steps until here take effect.
We also need to allow systemd-sleep system to read the swap file in SELinux . One option for allowing this is to make it fail. After that, use the audit2allow
to do the white listing. Lets go step by step:
-
Try to hibernate:
# systemctl hibernate
This should fail right now. Probably returning you to the display manager login. Also, logging details at
/var/log/audit/audit.log
. -
We can check the event happened with
audit2allow
inspecting the log, it will ouput something similar to this entry among others:# audit2allow -w -a type=AVC msg=audit(1630262756.460:2098): avc: denied { search } for pid=26180 comm="systemd-sleep" name="swap" dev="dm-0" ino=256 scontext=system_u:system_r:systemd_sleep_t:s0 tcontext=system_u:object_r:unlabeled_t:s0 tclass=dir permissive=0 Was caused by: Missing type enforcement (TE) allow rule. You can use audit2allow to generate a loadable module to allow this access.3.
-
To see what rule we must allow, just type:
# audit2allow -b #============= systemd_sleep_t ============== allow systemd_sleep_t unlabeled_t:dir search;
The above rule should be the only one in the output. If not, we could be white listing other elements by accident.
-
Instruct
SELinux
to allow further attemps by executing the following commands:# cd /tmp # audit2allow -b -M systemd_sleep ******************** IMPORTANT *********************** To make this policy package active, execute: semodule -i systemd_sleep.pp # semodule -i systemd_sleep.pp
We should be ready to hibernate now. Lets just try it by:
# systemctl hibernate
After resuming, the swap file should be deactivated, so continuing with the default zram swap device setup:
$ swapon
NAME TYPE SIZE USED PRIO
/dev/zram0 partition 8G 0B 100
As a bonus, you can enable gnome environment hibernate buttons by installing this extension.
Please check that all your devices state are properly restored after resuming.
If you are having unexpected problems, inspectioning the journal will be of help to see errors in systemd scripts:
journalctl -f
Hello guys. It was a pleasure to read all your comments so far. Many thanks for all your contributions ! I fixed the guide to reflect all your findings. I hope its better now for new readers. Allow me to reply to all of you on this comment and summarise the changes:
Update the guide to point out that works for Fedora34/35.
As many of you experimented problems with the
filefrag
method for calculating the swap offset, i just changed that section to point to the archwiki section, which seems is the most reliable way. It is still a mistery for me how is possible that my setup is working fine, after using thefilefrag
method. Lets see if i find some time to properly understand what happened.@miXwui , thanks a lot for all your contributions. Great detailed explanations/investigations ! Not only i learnt a lot, now im also happily running [B] + [C] . I added a link to your comment in the guide, so new readers can benefit from it as they read.
@manujchandra thanks for noticing the problem with the nvidia drivers. I was able to test this on a new desktop i bought. On my case its partially working: It correctly resumes, but seems GNOME is having rendering problems. Maybe the state of the graphic card is not correctly resumed for some reason. I need to investigate a bit more this, but cannot guarantee at what point in time. Until we find a solution, i just wrote an advice at the beginning of the guide warning this. The way i installed the nvidia drivers was through the rpm fusion non free repo. Heres the guide i followed, just in case someone else wants to continue investigations.
@w4tsn many thanks for your contributions ! Just modified the guide to:
Fixed the
grub2-mkconfig
command. Now/boot/grub2/grub.cfg
is used as destination for the output.Now the guide only allows the systemd security exception in selinux.
Seems this guide works perfectly fine with most setups, encrypted/unencrypted. Just clarified that at intro.
Regarding to the post in FedoraMagazine , of course ! go for it. Lets just put a link here whenever its done, so the information is connected.
@ixiotidi thanks to @w4tsn findings (see point above), now we modified the guide to only allow the needed systemd rule in selinux. Please follow that part again and check if that solves your problem.