Bug#75250: ssh: syslogin_perform_logout: logout() returned an error

2025-07-14 Thread Paul Slootman
On Sat 12 Jul 2025, Chris Hofstaedtler wrote:
> On Sat, Jul 12, 2025 at 03:17:15PM +0100, Richard Lewis wrote:
> > The draft release-notes for trixie do have some information on this,
> > https://www.debian.org/releases/trixie/release-notes/issues.en.html#things-to-be-aware-of-while-upgrading-to-releasename
> > 
> > but it we dont mention who(1) at all - should we say it continues to work?
> > (im not sure if it matters enough?)
> 
> It seems wrong to me to list all things that just continue to work.

I agree.
I mentioned that who(1) still worked, to indicate that apparently utmp
is deprecated (I did miss the documentation on that). So sshd should not
report an error if it fails to update utmp .


Paul



Bug#75250: ssh: syslogin_perform_logout: logout() returned an error

2025-07-12 Thread Chris Hofstaedtler
On Sat, Jul 12, 2025 at 03:17:15PM +0100, Richard Lewis wrote:
> The draft release-notes for trixie do have some information on this,
> https://www.debian.org/releases/trixie/release-notes/issues.en.html#things-to-be-aware-of-while-upgrading-to-releasename
> 
> but it we dont mention who(1) at all - should we say it continues to work?
> (im not sure if it matters enough?)

It seems wrong to me to list all things that just continue to work.

> https://manpages.debian.org/testing/coreutils/who.1.en.html stills refer
> to /var/run/utmp - not sure if that is just a documentation bug of if
> who is broken too ('who' doesnt return anything in a chroot for me)?

In a chroot, who should (most of the time) not show anything, 
because there was never a login.

Chris



Bug#75250: ssh: syslogin_perform_logout: logout() returned an error

2025-07-12 Thread Richard Lewis
The following message is a courtesy copy of an article
that has been posted to gmane.linux.debian.devel.ssh as well.

Paul Slootman  writes:

> It seems that on a fresh trixie install, nothing creates /run/utmp
> anymore.  Is that a bug or is utmp deprecated? who and who -b still seem
> to work, so not sure how who finds its information, and if this means
> utmp is not required anymore.

The draft release-notes for trixie do have some information on this,
https://www.debian.org/releases/trixie/release-notes/issues.en.html#things-to-be-aware-of-while-upgrading-to-releasename

but it we dont mention who(1) at all - should we say it continues to work?
(im not sure if it matters enough?)

https://manpages.debian.org/testing/coreutils/who.1.en.html stills refer
to /var/run/utmp - not sure if that is just a documentation bug of if
who is broken too ('who' doesnt return anything in a chroot for me)?



Bug#75250: ssh: syslogin_perform_logout: logout() returned an error

2025-07-12 Thread Chris Hofstaedtler
On Sat, Jul 12, 2025 at 03:26:57PM +0200, Paul Slootman wrote:
> It seems that on a fresh trixie install, nothing creates /run/utmp
> anymore.  Is that a bug or is utmp deprecated?

utmp is deprecated, indeed.

Chris



Bug#75250: ssh: syslogin_perform_logout: logout() returned an error

2025-07-12 Thread Paul Slootman
It seems that on a fresh trixie install, nothing creates /run/utmp
anymore.  Is that a bug or is utmp deprecated? who and who -b still seem
to work, so not sure how who finds its information, and if this means
utmp is not required anymore.

According to Googel Gemini (sorry...):

Yes, it is by design that /run/utmp is no longer provided in Debian
Trixie.

This change is primarily due to:

Year 2038 Problem (Y2K38): The traditional utmp file format uses 32-bit
timestamps, which are susceptible to the Year 2038 problem. After
January 19, 2038, these timestamps will overflow, leading to incorrect
date and time readings. Upstream developers for the utmp format have
been unwilling to change it.


Modern systemd Session Management: systemd (which Debian uses as its
init system) has its own, more robust, and modern session management
capabilities. Information about active sessions, login times, and other
user accounting data is now primarily managed by systemd itself, often
exposed via logind and its API.

So maybe sshd should ignore any error connected to utmp, instead of
throwing an error.


Thanks,
Paul



Bug#75250: ssh: syslogin_perform_logout: logout() returned an error

2025-06-17 Thread Frank Engler
Package: openssh-server
Version: 1:10.0p1-5
Followup-For: Bug #75250
X-Debbugs-Cc: [email protected]
Control: tags -1 patch

Dear Maintainer,

this message is the result of a missing (/var)/run/utmp. OpenSSH tries
to write, but this file does not exists in a default installation. To
resolve this issue we can create /run/utmp.
diff -urN a/openssh-10.0p1/debian/openssh-server.tmpfiles 
b/openssh-10.0p1/debian/openssh-server.tmpfiles
--- a/openssh-10.0p1/debian/openssh-server.tmpfiles 2025-05-09 
14:40:49.0 +0200
+++ b/openssh-10.0p1/debian/openssh-server.tmpfiles 2025-06-17 
01:01:14.3 +0200
@@ -1 +1,2 @@
 x /tmp/sshauth.*
+f /run/utmp 0664 root utmp