Hi list,
Am Mittwoch, dem 24.05.2023 um 09:59 +0100 schrieb Michael Tremer:
Hello Jonatan,
Thanks for this *huge* patchset.
What I do not quite understand from your email is the motivation for this. Starting the SSH daemon a little bit sooner is not justifying such massive change.
Starting the SSH-Daemon earlier was not the only reason. The main reason was
a) to debug problems with the red interface via ssh and via console b) Wait infinitely for a dhcp lease. We currently only wait 30 seconds. If we do not get a lease, red stays down forever.
I consider this a highly risky patchset as we are touching code that is already incredibly fragile and the smallest changes might introduce more problems.
I know that this is a risky patch set. But not touching these code does not improve it. It gets worse each day. Never touch a running system turns against us here. So both change and no change have a risk here.
My further plan for this patch set is to split it up even further. My builder is currently running with all the patches you already agreed to. I will send this patch set in the next days. When this is merged, I will go on to review the other patches.
Greetings Jonatan
But I have to say that I like how you split this into small patches that can be reviewed easier.
On 23 May 2023, at 18:23, Jonatan Schlag jonatan.schlag@ipfire.org wrote:
Hi,
this is the first try for a patch series which makes the startup of local and uplink networks independent. This resolves in the end #11502. It should further allow us to keep waiting for a DHCP lease, because at least our local networks are all up when we get to the red network. This might have some small quirks, even when my testing so far shows no problems. I still think it is now better to share result, then to wait any longer. As this touches quite important parts of our network stack, I might be a could idea to merge this in two stages. After the patch, titled "network startup: Clean up duplicated Code ", might be a good breaking point. But in the end, this is up to you. Please keep in mind, I only tested the full patch set:
Remove ipsec interface creation from network startup Remove Start/Stop links for client175 Use bash as shebang in network initscripts network initscripts: check if the zone in the current config exists network initscripts: Remove code for old zone scheme network scripts: remove check for AUTOCONNECT network startup: Reload routing informations for every interface network startup: Always cleanup before red gets started network startup: check for correct action at start network startup: Refactor how cmd args are processed network startup: Clean up duplicated Code network script: add extra scripts for action that depend on a network network startup: Add scripts for local and uplink network startup: Start local and uplink network independent network startup: Only work with configured zones
config/rootfiles/common/aarch64/initscripts | 31 +++-- config/rootfiles/common/riscv64/initscripts | 31 +++-- config/rootfiles/common/x86_64/initscripts | 31 +++-- lfs/initscripts | 36 +++--- src/initscripts/networking/any | 31 ++++- src/initscripts/networking/functions.network | 14 +- src/initscripts/networking/red | 5 +- .../networking/red.up/99-pakfire-update | 2 +- src/initscripts/system/depends-on-network | 40 ++++++ src/initscripts/system/network | 121 ++++++++-------
10 files changed, 218 insertions(+), 124 deletions(-) create mode 100644 src/initscripts/system/depends-on-network
Looking forward to your feedback
Greetings Jonatan