Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.
RHEL Engineering is moving the tracking of its product development work on RHEL 6 through RHEL 9 to Red Hat Jira (issues.redhat.com). If you're a Red Hat customer, please continue to file support cases via the Red Hat customer portal. If you're not, please head to the "RHEL project" in Red Hat Jira and file new tickets here. Individual Bugzilla bugs in the statuses "NEW", "ASSIGNED", and "POST" are being migrated throughout September 2023. Bugs of Red Hat partners with an assigned Engineering Partner Manager (EPM) are migrated in late September as per pre-agreed dates. Bugs against components "kernel", "kernel-rt", and "kpatch" are only migrated if still in "NEW" or "ASSIGNED". If you cannot log in to RH Jira, please consult article #7032570. That failing, please send an e-mail to the RH Jira admins at rh-issues@redhat.com to troubleshoot your issue as a user management inquiry. The email creates a ServiceNow ticket with Red Hat. Individual Bugzilla bugs that are migrated will be moved to status "CLOSED", resolution "MIGRATED", and set with "MigratedToJIRA" in "Keywords". The link to the successor Jira issue will be found under "Links", have a little "two-footprint" icon next to it, and direct you to the "RHEL project" in Red Hat Jira (issue links are of type "https://issues.redhat.com/browse/RHEL-XXXX", where "X" is a digit). This same link will be available in a blue banner at the top of the page informing you that that bug has been migrated.

Bug 1792173

Summary: Ordering cycle introduced by tangd.socket during boot
Product: Red Hat Enterprise Linux 8 Reporter: Renaud Métrich <rmetrich>
Component: tangAssignee: Sergio Correia <scorreia>
Status: CLOSED DUPLICATE QA Contact: BaseOS QE Security Team <qe-baseos-security>
Severity: high Docs Contact:
Priority: high    
Version: 8.1CC: dapospis, scorreia
Target Milestone: rcFlags: pm-rhel: mirror+
Target Release: 8.0   
Hardware: All   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2020-02-13 11:12:46 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Attachments:
Description Flags
Proposed updated tangd units none

Description Renaud Métrich 2020-01-17 08:23:49 UTC 
Description of problem:

When enabling tangd.socket on a RHEL8 system, an ordering cycle is created (and deleted) at boot, causing various issues around "sockets.target" to happen, usually "tangd.socket" gets not started:

-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
[...] sockets.target: Found ordering cycle on tangd.socket/start
[...] sockets.target: Found dependency on tangd-keygen.service/start
[...] sockets.target: Found dependency on basic.target/start
[...] sockets.target: Found dependency on sockets.target/start
[...] sockets.target: Job tangd.socket/start deleted to break ordering cycle starting with sockets.target/start
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

The reason behind this is "tangd.socket" requires "tangd-keygen.service" which itself must start after "basic.target" (it's the default dependencies for services), but this cannot happen because "sockets.target" is supposed to be reached before "basic.target".

The exact same unit files on RHEL7 do not produce this ordering cycle, even though they should in theory.

There are hence 2 issues there:

(Issue 1). something must have changed in systemd-239 compared to systemd-219 (RHEL7), causing "sockets.target" to pull "tangd.socket" even though "tangd.socket" is installed in "multi-user.target"

(Issue 2). the tangd.socket unit must be fixed, it is quite non-sense to have a socket depend on service units (a socket is just "idle" until used)


Version-Release number of selected component (if applicable):

tang-7-1.el8.x86_64, also applies to Upstream tang probably (at least applies to tang-7-3 which is "latest community" on Arch Linux ... running on my raspberrypi :-))


How reproducible:

This is 100% reproducible with RHEL 8.1 latest.


Steps to Reproduce:
1. Install tang

  # yum -y install tang

2. Enable the socket

  # systemctl enable tangd.socket

3. Reboot

Actual results:

Ordering cycle broken for sockets.target


Expected results:

No ordering cycle


Additional info:

(Issue 1) related to systemd will be investigated in another BZ.

(Issue 2) proposed fix in comment to follow.
Comment 1 Renaud Métrich 2020-01-17 08:35:08 UTC 
Currently, we have "tangd.socket" require/after "tangd-keygen.service" and "tangd-update.service". Additionally the socket is installed in "multi-user.target":

/usr/lib/systemd/system/tangd.socket:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
[Unit]
Description=Tang Server socket
Requires=tangd-keygen.service
Requires=tangd-update.service
Requires=tangd-update.path
After=tangd-keygen.service
After=tangd-update.service

[Socket]
ListenStream=80
Accept=true

[Install]
WantedBy=multi-user.target
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

It looks weird to me that a *socket* depends on *services* or *paths*.
Due to having a socket, we have systemd automatically add a "After" dependency on "sockets.target" (this is Issue 1 to be investigate), which creates the ordering cycle.
For some reason, the same dependency was happening on RHEL7 but didn't create an ordering cycle.

To fix all this, I propose the following changes:

1. Implement a "regular" socket, installed in "sockets.target" with no dependencies

/usr/lib/systemd/system/tangd.socket:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
[Unit]
Description=Tang Server socket

[Socket]
ListenStream=80
Accept=true

[Install]
WantedBy=sockets.target
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

2. Move the dependencies in "tangd@.service" unit which is where dependencies have to be handled

/usr/lib/systemd/system/tangd@.service:
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------
[Unit]
Description=Tang Server
Requires=tangd-keygen.service
After=tangd-keygen.service

[Service]
StandardInput=socket
StandardOutput=socket
StandardError=journal
ExecStart=/usr/libexec/tangd /var/cache/tang
User=tang
-------- 8< ---------------- 8< ---------------- 8< ---------------- 8< --------

Note here that I removed the dependency on "tangd-update.path" and "tangd-update.service" since I consider that they are not needed at all:
- tangd-update.path monitors changes in Tang keys, this has nothing to do with "tangd@.service" instances
- tangd-update.service updates the cache with new keys, this has nothing to do with "tangd@.service" instances either

Please correct me if I'm wrong with the logic.

With this in place, the ordering cycle just disappear, "tangd.socket" becomes a regular socket, and, when it triggers an instance of "tangd@.service", systemd will check that there is indeed Tang keys to use.

Note for packaging: the postinstall RPM scriptlet should re-enable the socket upon package upgrade since it moved from "multi-user.target" to "sockets.target".
Comment 2 Renaud Métrich 2020-01-17 08:38:59 UTC 
Created attachment 1652978 [details]
Proposed updated tangd units
Comment 3 Renaud Métrich 2020-01-17 10:28:29 UTC 
Issue 1 is tracked by BZ #1792219