[pkg-cryptsetup-devel] Bug#768314: Bug#768314: cryptsetup: Passphrase prompt rolls by without stopping (fwd)

[Tomas Pospisek]

Hello systemd maintainers & Laurent,

bug #768314 [0] has been reassigned to the release-notes. It's 
about a user not being able to enter his cryptsetup 
password. A solution seems to be to install plymouth.

It seems, that's a known problem, as noted in your titanpad [1]. However I 
could not find a respective issue in the BTS entries for systemd [2] - is 
there one that tracks this problem?

Do you have any idea how this problem should be resolved for jessie?

The original owner of the bug report suggests raising plymouth to a 
"Recommends" dependency.

I suggested that systemd could recognize that there are mounted crypted 
partitions and suggest to the user to install plymouth at postinst time.

Another workaround would be to just document the problem in the release 
notes and hope users won't run into it.

Do you see or prefer any other approach?

Second question: in the titanpad entry you write "The plan with plymouth 
0.9.0-9 is to not require any modification to the kernel cmdline and 
enable the I/O multiplexing functionality by default when the pkg is 
installed."

As far as I can see from the plymouth changelog [3] no solution was 
implemented for this problem [4] in 0.9.0-9 and there hasn't been any 
activity in that bug report since Nov 14th. What's the current goal/aim 
wrt that bug?

Thanks,
*t

[0] https://bugs.debian.org/768314
[1] https://debian.titanpad.com/23?
[2] http://bugs.debian.org/systemd
[3] http://metadata.ftp-master.debian.org/changelogs//main/p/plymouth/plymouth_0.9.0-9_changelog
[4] https://bugs.debian.org/768329

@bugs.debian.org ---------- Forwarded message ----------
Date: Tue, 30 Dec 2014 23:11:58 +0100
From: Jonas Meurer <jonas at freesources.org>
To: Tomas Pospisek <tpo at sourcepole.ch>, 768314 at bugs.debian.org
Cc: Kjetil Kjernsmo <kjetil at kjernsmo.net>
Subject: Re: [pkg-cryptsetup-devel] Bug#768314: cryptsetup: Passphrase prompt
     rolls by without stopping

Hi Tomas,

thanks for taking care of the bugreport.

Am 30.12.2014 um 19:27 schrieb Tomas Pospisek:
> Hello Jonas & Kjetil,
>
> (context: I'm reading through release-notes bug reports).
>
> I'm not sure I understand what you are expecting as a result by
> cloning/reassigning this to the release notes -
>
> Let me try to understand the problem:
>
> * if there's an encrypted partition, then systemd, who aparently would be
>   responsible to do so will not prompt for the password, if plymouth is
>   not installed.
>
> Is my understanding of the problem correct?

Yes. Actually, it is even more complicated, but your understanding is
correct:

Systemd includes its own dm-crypt/cryptsetup device unlocking functions.
With systemd as init system, it processes all dm-crypt encrypted devices
that shall be unlocked during the boot process and *after* initramfs.
I don't know systemd, but from the bugreports I learned that it
apparently doesn't implement a proper mechanism to prompt for user input
itself. Instead it relies on plymouth doing that task. As a result,
systemd without plymouth doesn't wait for user input at unlocking
dm-crypt devices but instead continues to print boot logging output to
the console.

> So I think the right thing to do would be, that during the upgrade the
> systemd postinstallation should check whether there are some mounted
> partitions that are crypted and then recommend to install plymouth. Do
> you concur?

I would even go futher and say that systemd should recommend plymouth in
any case. Still, if it's only recommended and not a hard dependency, the
discovered behaviour should be documented in the release notes in my eyes.

> Otherwise, should the release-notes recommend to install plymouth to the
> user if s/he has crypted partitions that should get mounted during boot?

Yes, that's what needs to be done at least.

> Ideally IMHO the release notes should also explain the problem in
> sufficient technical detail to allow the user to take his own steps to
> further understand the problem and to choose an alternative solution if
> he deems so.
>
> Optimally you could suggest a wording?

Unfortunately I've not enough knowledge about systemd to propose a
wording. But feel free to use anything I wrote in the bugreport for a draft.

Cheers,
  jonas