mbox series

[0/5] Rework socket creation to better integrate with systemd

Message ID 20231023022529.15081-1-felix.moessbauer@siemens.com
Headers show
Series Rework socket creation to better integrate with systemd | expand

Message

Felix Moessbauer Oct. 23, 2023, 2:25 a.m. UTC 
This series is a follow up of "core: fail if running under systemd but not sockets".
As discussed on the ML, the implemented approach of failing in case no sockets are
are provided was not sufficient and by that cannot be integrated.

The new series solves all issues which were brought up in the discussion by
following a different strategy. Instead of trying to guess if we are running as
systemd service, we simply track which sockets we manually create and register
cleanup handlers for them. This also includes some cleanup of the existing socket
creation infrastructure (p1-p3).

The patches p4,p5 take care of creating the sockets at the correct path when running
on a systemd bootet system (no matter if as service or via the shell).

Best regards,
Felix Moessbauer
Siemens AG

Felix Moessbauer (5):
  fix: properly cleanup sockets on SIGTERM
  move listener_create to network_utils.c
  remove irrelevant sd_booted values
  create ipc sockets below /run on systemd
  docs: improve swupdate.socket systemd example

 core/Makefile               |   1 +
 core/network_thread.c       |  78 +----------------------
 core/network_utils.c        | 124 ++++++++++++++++++++++++++++++++++++
 core/progress_thread.c      |  21 +-----
 core/swupdate.c             |   8 ++-
 doc/source/swupdate.rst     |  13 ++--
 include/network_interface.h |   1 -
 include/network_utils.h     |  32 ++++++++++
 ipc/network_ipc.c           |  10 ++-
 ipc/progress_ipc.c          |  10 ++-
 10 files changed, 188 insertions(+), 110 deletions(-)
 create mode 100644 core/network_utils.c
 create mode 100644 include/network_utils.h

Comments

Stefano Babic Oct. 23, 2023, 9:12 a.m. UTC | #1 
Hi Felix,

On 23.10.23 04:25, 'Felix Moessbauer' via swupdate wrote:
> This series is a follow up of "core: fail if running under systemd but not sockets".
> As discussed on the ML, the implemented approach of failing in case no sockets are
> are provided was not sufficient and by that cannot be integrated.
> 
> The new series solves all issues which were brought up in the discussion by
> following a different strategy. Instead of trying to guess if we are running as
> systemd service, we simply track which sockets we manually create and register
> cleanup handlers for them. This also includes some cleanup of the existing socket
> creation infrastructure (p1-p3).

Thanks for the series, this makes absolutely sense and implements what 
we discussed before.

> 
> The patches p4,p5 take care of creating the sockets at the correct path when running
> on a systemd bootet system (no matter if as service or via the shell).

First 3 patches are a nice cleanup and I plan to apply soon. Patch 4 
requires some discussion.

Best regards,
Stefano

> 
> Best regards,
> Felix Moessbauer
> Siemens AG
> 
> Felix Moessbauer (5):
>    fix: properly cleanup sockets on SIGTERM
>    move listener_create to network_utils.c
>    remove irrelevant sd_booted values
>    create ipc sockets below /run on systemd
>    docs: improve swupdate.socket systemd example
> 
>   core/Makefile               |   1 +
>   core/network_thread.c       |  78 +----------------------
>   core/network_utils.c        | 124 ++++++++++++++++++++++++++++++++++++
>   core/progress_thread.c      |  21 +-----
>   core/swupdate.c             |   8 ++-
>   doc/source/swupdate.rst     |  13 ++--
>   include/network_interface.h |   1 -
>   include/network_utils.h     |  32 ++++++++++
>   ipc/network_ipc.c           |  10 ++-
>   ipc/progress_ipc.c          |  10 ++-
>   10 files changed, 188 insertions(+), 110 deletions(-)
>   create mode 100644 core/network_utils.c
>   create mode 100644 include/network_utils.h
>