i am trying to set up Passbolt on Ubuntu 22.04 LTS
During the step of checking the entropy I noticed that it is stuck at 256.
I tried to install rng-tools and haveged to increase it but it is still at the value of 256.
I am running Ubuntu 22.04.1 LTS / 5.15.0-48-generic on a VM.
I have read that in a recent update the entropy value was capped at 256 but I can’t figure out if this in any way would hinder Passbolt from working properly.
I would be glad if anyone could help me out
Are you using Ubuntu Desktop or Ubuntu Server ?
It’s not a bug or an issue, this has been picked up by thousands of unix users. Everyone running kernel v-5.15.0-47 and up would have the entropy hardcoded at 256.
I am not too sure how passbolt would react to low entropy as i always aim for 3000+ preferably 4096
There was a large commit done for 5.15.X, you can find it here kernel/git/stable/linux.git - Linux kernel stable tree
If you would like to change the hard coded value, it is located
Jump to line 551 and you will see it is set to 8, which would be 256.
8*32 = 256
POOL_BITS = BLAKE2S_HASH_SIZE * 8,
POOL_READY_BITS = POOL_BITS, /* When crng_init->CRNG_READY */
POOL_EARLY_BITS = POOL_READY_BITS / 2 /* When crng_init->CRNG_EARLY */
if you would like to have lets say 4096, you would need to change the hardcoded value from 8 to 128.
128*32 = 4096
Entropy matters a lot, here is another reference particular to passbolt. Passbolt Help | Why should I install haveged on virtual environments?
I am using Ubuntu Server.
Thanks for your detailed clarification.
Related kernel commit explanation for reference: [PATCH 5.15 038/145] random: use computational hash for entropy extraction - Greg Kroah-Hartman
@remy maybe use of haveged needs to be revisited in docs.
As I am reading through the patch comments, it appears that the structuring of the random number generation has changed and running out of entropy (or a security concern of an attacker zeroing it out) is no longer a risk?
Thanks guys for the information, very good to know, and yes, the entropy help page should be updated.
Could you perhaps try debian 11 as i have actively installed and used it without any entropy issues.
That is exactly what i was thinking. I am not sure why this move was initiated and why it was thought to be a “good idea”.
The commit needs to be revisited and fixed.
Maybe I’m reading it wrong, but it seems better to me if the rng process is started with a controllable encryption algorithm rather than vulnerable noise sourcing which can be compromised, leading to no randomness. In this way entropy is always secure and cannot be exhausted.