From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from gate001.proxmox.com (gate001.proxmox.com [45.144.208.40]) by lore.proxmox.com (Postfix) with ESMTPS id F13F51FF0DE for ; Thu, 19 Nov 2026 14:37:55 +0100 (CET) Received: from gate001.proxmox.com (localhost.localdomain [127.0.0.1]) by gate001.proxmox.com (Proxmox) with ESMTP id B99F321697; Thu, 19 Nov 2026 14:37:55 +0100 (CET) Mime-Version: 1.0 Content-Transfer-Encoding: quoted-printable Content-Type: text/plain; charset=UTF-8 Date: Thu, 19 Nov 2026 14:37:46 +0100 Message-Id: To: "Thomas Lamprecht" , "Mark Atwood" , Subject: Re: [PATCH cluster v2 1/1] datacenter config: add ticket-signature-hash option X-Mailer: aerc 0.20.0 References: <20261113030646.3533-1-mark@wolfssl.com> <20261113030646.3533-2-mark@wolfssl.com> <2a835c03-2787-48f8-a8e4-60bd05e35107@proxmox.com> In-Reply-To: From: "Shannon Sterz" X-Bm-Milter-Handled: 55990f41-d878-4baa-be0a-ee34c49e34d2 X-Bm-Transport-Timestamp: 1795095466545 X-SPAM-LEVEL: Spam detection results: 0 AWL 0.541 Adjusted score from AWL reputation of From: address DMARC_MISSING 0.1 Missing DMARC policy KAM_DMARC_STATUS 0.01 Test Rule for DKIM or SPF Failure with Strict Alignment (newer systems) RCVD_IN_DNSWL_MED -2.3 Sender listed at https://www.dnswl.org/, medium trust SPF_HELO_NONE 0.001 SPF: HELO does not publish an SPF Record SPF_PASS -0.001 SPF: sender matches SPF record Message-ID-Hash: O27OJ4S3UB4EXINK3O5JKMCXKBTQWOZJ X-Message-ID-Hash: O27OJ4S3UB4EXINK3O5JKMCXKBTQWOZJ X-MailFrom: s.sterz@proxmox.com X-Mailman-Rule-Misses: dmarc-mitigation; no-senders; approved; loop; banned-address; emergency; member-moderation; nonmember-moderation; administrivia; implicit-dest; max-recipients; max-size; news-moderation; no-subject; digests; suspicious-header X-Mailman-Version: 3.3.10 Precedence: list List-Id: Proxmox VE development discussion List-Help: List-Owner: List-Post: List-Subscribe: List-Unsubscribe: On Thu Nov 19, 2026 at 1:38 PM CET, Thomas Lamprecht wrote: > Thanks for your input, more than appreciated. Of course! > Am 18.11.26 um 10:40 schrieb Shannon Sterz: >> Using the cluster version guard helper [2] would come with one downside: >> It als errors when not all nodes are online or the version information >> is unreadable. In that case we'd probably want to fall back to SHA-1, >> which for FIPS compliant nodes would mean that they can't log in if a >> single node goes down. While logging in with TFA is already an issue in >> an unquorate state, this would arguably be worse for FIPS users. > > I mean, FIPS compliance might require quorum to allow login in the > first place? as only with quorum you can ensure that the user config > and its access control lists etc. are up-to-date and the user is allowed > to login. We argued convenience-over-security for the common case, but > for security compliance the latter often should be weighted higher I > think. So this might be actually a feature for FIPS? > > But in any case rather orthogonal to this here. Yes, I suppose requiring quorum in FIPS mode may be reasonable and should be considered. However, I think I may have been unclear. My argument did not relate to whether a FIPS cluster is quorate or not. Consider the following: In FIPS mode, if we use the current implementation of `assert_min_cluster_version()`, no new tickets across the cluster could be created if a *single* node goes down or doesn't provide version information. This applies even if the cluster is quorate (and, therefore, ACLs etc. are current and correct). This means for FIPS users during ordinary maintenance operations such as a single node reboot, nobody would be able to do a regular login via the GUI or API. An attacker could easily use this during an attack to make patching vulnerabilities via the GUI or API harder for admins. If I am a high risk user and I run a large FIPS hardened cluster, an attacker could lock me out of the GUI and API with a DDoS attack on a single node. So I'd argue using that check directly makes the situation quote a bit worse for FIPS users. >> So we'd need to keep using SHA-256 at least until a new node joins with >> a version that's too old. We could do that with a non-dying variant of >> the helper that reads the stale info from pmxcfs as usually version >> numbers only go up. However, nodes with an unknown version would remain >> an issue. Alternatively we may want to add a fallback to SHA-256 if >> signing with SHA-1 fails or allow an override similar to this approach. >> >> The nice thing about this option is that it only depends on the state of >> the local node. Even in an unquorate state the cluster config stays >> readable as pmxcfs switches to read-only mode. > > hmm, using an option like this certainly makes it a bit simpler to > reason about when to do what, so can be fine by me. We can switch the > default here then e.g. in PVE 10 (we require all nodes to be on latest > 9.x before upgrade anyway) or if we're more cautious then 11. And there > might be other algos in the future, so OK to have this config as > infrastructure for that. Yes, moving toward SHA-256 as the default with PVE 10 seems reasonable to me. >> [1]: https://lore.proxmox.com/all/DMKS80347XID.3KPXPXVUHQ3ZG@proxmox.com= / >> [2]: https://git.proxmox.com/?p=3Dpve-cluster.git;a=3Dcommitdiff;h=3D168= ac114df