Tuesday, 2026-08-25

gouthamrhttps://bugs.launchpad.net/keystone/+bug/2153453 is now public14:34
gouthamrhttps://bugs.launchpad.net/keystone/+bug/2158538 is now public 14:34
xeko/14:35
gouthamrhey xek, ty for working on these ^ watching #openstack-keystone for the patches and prepping the OSSA in parallel14:39
xekgouthamr ack14:39
fungii'm still on hand to review the ossa repo change( for the advisory/advisories15:04
opendevreviewGoutham Pacha Ravi proposed openstack/ossa master: Add OSSA-2026-037  https://review.opendev.org/c/openstack/ossa/+/100232415:22
gouthamrty fungi... xek: ^ please fact check15:22
fungion it15:22
gouthamrgah, noticed something i discussed with xek but didn't add here15:25
gouthamrfixing15:25
fungijust a heads up that the second stable/2025.1 patch is flagged as in merge-conflict with the branch by gerrit15:28
gouthamryes, xek, that would be something folks would ask for ^15:29
fungii was going to bring it up in #openstack-keystone but don't see xek in that channel15:29
xeklooking15:29
opendevreviewGoutham Pacha Ravi proposed openstack/ossa master: Add OSSA-2026-037  https://review.opendev.org/c/openstack/ossa/+/100232415:30
fungixek: https://review.opendev.org/c/openstack/keystone/+/1002308 specifically15:31
fungitime check, we're an hour past the indicated publication time, so if what's in 1002324 is correct enough to secure systems we can always make further clarifications about which bug was which in a future revision16:00
xekplease just add the missing review links16:05
gouthamrah, 16:05
fungiare the bugs not fixed in master without those? one is even wip16:06
gouthamr(yes, i was commenting on gerrit with the same question)16:07
fungifurther work to future-proof the software is out of scope for the advisory, which is the list of fixes users need to deploy right now to address the risks described16:08
xekwhat do you mean by future-proof?16:09
fungixek: let me rephrase: do users need to apply those patches you listed on master *right now* to avoid the described vulnerabilities being exploited by their users?16:10
fungithat is the purpose of a security advisory16:10
xekyes16:11
fungiso why aren't they needed on stable branches?16:11
xekwhich ones? they are all needed on stable branches, the only discussion is that those restrict some operations which weren't restricted before and thus may make something that was permitted before not work16:12
xeklike using application credentials in one scope to create ec2 credentials in another scope16:13
fungiyou listed patches which weren't sent to downstream stakeholders16:14
xekor log in with an ec2 credential meant for swift, have all keystone api access including changing the project_id in the ec2 credential you logged in, changing it's scope, having access to swift objects from another project, etc.16:14
fungiit seems like we're misaligned on how security advisories work and what it is we're publishing16:14
xekyes, because while I was on PTO someone made a decision to remove part of my work16:15
fungithis should have just all been done in public, it's way to complicated to fix in secret, but that's not an argument we have time for right now16:15
gouthamrhttps://review.opendev.org/c/openstack/keystone/+/1002330 is advisory worthy.  If your patches are ready, we can use the public bug workflow to advisory it right after? 16:16
xekI agree, it's way easier, I said this before, but it feels like re-developing these features, the bugs go deep16:16
fungiredeveloping features is definitely out of scope for our advisory process. openstack security advisories are strictly for backportable backward-compatible fixes that can be patched entirely in code with no action required by the operator to secure their systems other than applying the patches16:20
xekso advisories would be only for simple changes? is the complication of the remediation the bar on whether we create a security advisory?16:21
fungithat's a very big part of the determination, yes16:22
fungisome things can't be fixed through code alone in a way that's backportable for existing deployments without the operator needing to take additional steps, and we have a separate publication workflow for those more complicated situations16:24
xekI'm trying to determine what's left if we don't include the ec2 ban from keystone API and https://review.opendev.org/c/openstack/keystone/+/1002330 for the current advisory...16:24
xekyou don't argue against including https://review.opendev.org/c/openstack/keystone/+/1002289 ?16:26
gouthamrno i'm having second thoughts there16:26
gouthamrexplain that to me please, 1002301 makes it so that empty methods are considered "delegated". 16:27
gouthamr(1002301 is in the advisory)16:27
fungiwhat bug does change 1002289 fix? the change doesn't seem to include any bug reference in its commit message16:29
gouthamri think it fixes "ec2credential" - and that's the OSSN we're referring to16:30
gouthamri can prep the OSSN in parallel and mark it WIP16:30
fungireminder, we're now 1.5 hours past when we told downstream stakeholders it would be okay to discuss this in public, and we don't have an advisory published about anything yet16:31
gouthamryes, i vote on retaining the existing advisory as is. I'll make the small correction on "independently reported". 16:31
opendevreviewGoutham Pacha Ravi proposed openstack/ossa master: Add OSSA-2026-037  https://review.opendev.org/c/openstack/ossa/+/100232416:33
fungilgtm16:34
gouthamrwe're racing in comments16:43
gouthamrif i'm missing something, flag it please16:43
gouthamri asked if you wanted the "ec2credential" part any more obvious than I already stated.. but, i'd not.. i'm hoping the OSSN for that can go out in a couple days 16:44
xekok, so if we go without https://review.opendev.org/c/openstack/keystone/+/1002289, we also go without https://review.opendev.org/c/openstack/keystone/+/1002082 because it doesn't work without it16:45
gouthamrxek: both of those are precisely the OSSN content?16:46
xekbut still, without the 3rd patch, ec2 api can still be used to create long-lived credentials16:47
xekin another project16:47
gouthamrits not me/VMT that are holding the line on that - our opinion is to fix that up publicly (as you're doing) and get reviews and supply an OSSN to operators16:47
gouthamryes16:47
gouthamrand that's a known gap - we've indicated that to operators 16:48
xekright, I'm only saying the OSSA shouldn't say that it fixes that16:48
xekok, let's go with this, I see your comment16:49
fungidoesn't the notes section of the advisory cover it?16:49
gouthamrit does imo16:50
fungiright, at least that seemed to be the intent of including that note16:51
fungiif it doesn't accurately capture it, we can update via errata later16:51
fungii've approved the advisory now that it has xek's +116:52
gouthamrty! working on the emails now16:52
xekI +116:53
xekwe should have really drafted this together on the bug beforehand16:54
fungii recommend sending those out asap while zuul does its thing to get the site updates in parallel16:54
fungixek: yes, normally we draft the prose for at least the impact description, affected versions, et cetera in comments on the bug when it's being done under private embargo16:55
fungito give the project maintainers an opportunity to catch misstatements before publication day16:55
gouthamromergod, rendering sucks16:59
opendevreviewGoutham Pacha Ravi proposed openstack/ossa master: Add OSSA-2026-037  https://review.opendev.org/c/openstack/ossa/+/100232417:00
gouthamrmy bad on that one, a teeny yaml mistake17:00
fungik17:00
gouthamri've the emails prepped if you can push this back to the gate :(17:00
fungireapproved17:00
gouthamr++ ty17:01
fungithe e-mails can go out now even if the site isn't updated yet, since they don't refer to one another17:01
gouthamr++emails sent17:06
fungiaccepted the one for openstack-announce17:12
gouthamrthanks fungi 17:12
gouthamrhttps://bugs.launchpad.net/keystone/+bug/2159643 is now public17:19
opendevreviewMerged openstack/ossa master: Add OSSA-2026-037  https://review.opendev.org/c/openstack/ossa/+/100232418:11
xekfungi, gouthamr, on the complexity thing - to date, we were trying to merge security issues, to release those in larger batches, since it was easier to process and test them together, avoiding merge conflicts, but we could also decide the other way - splitting the issues, like I did with https://bugs.launchpad.net/keystone/+bug/2159643 just now, and even this issue can be split further, since parts of it were discovered independently in19:04
xekhttps://bugs.launchpad.net/keystone/+bug/215893119:04
fungiyes, part of the challenge too is that people kept putting comments in some private bugs about other also private bugs, and we were stuck being able to make any of them public until all of them were public19:05
gouthamrokay, 2158931 is private.. this channel is publicly logged. Are you okay with me opening it up? 19:05
xekyes, I already poked you out of band, it contains the same findings19:07
xekand it was already referenced in the opened bug19:07
gouthamrack19:07
gouthamrhttps://bugs.launchpad.net/keystone/+bug/2158931 is now public19:07
gouthamrfungi xek: does it make sense to have a new LP for turning off EC2 credentials?19:27
gouthamrit will make it easier to lump all the related work there, and, i can reference all of this in the OSSN to tie the story together19:28
xekyes, maybe not just ec2, also application credentials and maybe others19:28
gouthamryou mean token rescoping with the others? ec2 has a whole host of problems to warrant its own?19:29
fungias long as it's public, sounds like a fine idea to me, however is needed to best organize what's being done19:30
fungii think it's up to the keystone maintainers how they want to track that work, whether bug or blueprint or spec or just in the change commit messages19:30
gouthamryeah, public - folks are already aware now19:30
fungie.g. with the help of something lightweight like a consistent gerrit change hashtag19:31
gouthamri agree.. but, i wanted to tie up loose ends here. I informed MITRE about the publication and the gap we've left19:31
fungithanks!19:31
gouthamrwhatever you do to "ec2credential", don't do it silently19:31
gouthamrxek: gtema, keystone-cores.. keep us in the loop, so we'll do the OSSN as promised in the advisory and let operators close the holes we've currently pointed at19:32
xekoh, I already proposed it for appcreds https://review.opendev.org/c/openstack/keystone/+/98716019:32
xekfor the advisory, the policy based approach of disabling ec2 should work19:34
xekhttps://bugs.launchpad.net/keystone/+bug/2153453/comments/4419:35
xekthe issue is, I see people in the comments asking how to keep using ec2 credentials, while fixing the vulnerability19:36
fungiseems like the answer to that is pretty simple: you can't19:39
xekhttps://review.opendev.org/c/openstack/keystone/+/1002330 partially disables the problematic APIs, making it safe AFAIK19:46
fungiah, that. asking in which comments?19:50
xekhttps://bugs.launchpad.net/keystone/+bug/2153453/comments/49 - the second point at the end19:52
gouthamri think you responded that "EC2 tokens (for S3 supoprt in Swift)" is unaffected19:52
gouthamrthe same tokens shouldn't be used to generate new auth in keystone, and that's the gap  to be addressed and recommended to be enforced via an OSSN19:53
xekyes, none of the changes block that use case, the discussion was only around whether there is some potential operator that uses outside this usecase19:55
xekone such use case was ec2 tokens rotation, replacing the blobs with PATCH, that's why it's carved out in the above change19:55
gouthamrgood stuff19:55
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0109: EC2-derived tokens retain full privileges  https://review.opendev.org/c/openstack/security-doc/+/100239220:58
xekgouthamrfungidmendiza ^20:59
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0109: EC2-derived tokens retain full privileges  https://review.opendev.org/c/openstack/security-doc/+/100239221:25
gouthamrthanks for working on it xek 21:26
gouthamrhttps://bugs.launchpad.net/keystone/+bug/2153447 is now public21:26
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0109: EC2-derived tokens retain full privileges  https://review.opendev.org/c/openstack/security-doc/+/100239221:26
gouthamrhttps://bugs.launchpad.net/keystone/+bug/2154645 is now public21:30
opendevreviewGoutham Pacha Ravi proposed openstack/ossa master: OSSA-2026-037 Errata 1  https://review.opendev.org/c/openstack/ossa/+/100240322:11
gouthamrfungi: JayF: still around to look at this ^?22:14
fungiyep, reviewing now22:20
fungigouthamr: quick note, you added the errata history section but not the errata section22:22
gouthamr!22:22
fungiotherwise lgtm22:22
opendevreviewGoutham Pacha Ravi proposed openstack/ossa master: OSSA-2026-037 Errata 1  https://review.opendev.org/c/openstack/ossa/+/100240322:23
gouthamri _think_ Red Hat has managed to assign duplicate CVEs.. i'll confirm with them with less-tired eyes22:24
gouthamror maybe xek knows what the confusion is there.. we'll get it resolved either case22:24
fungithat does sometimes happen, but also sometimes what you have is a cve for the upstream bug and then a cve for the bug's impact downstream in the distro22:25
fungithough the latter is unusual and only ideally when the downstream impact is not as described upstream for downstream-specific reasons22:25
fungie.g. additional patches for downstream-only behaviors due to other alterations they're carrying in their fork22:26
gouthamrah possible22:27
gouthamrty for flagging that possibility :) i'd've not dug down that road22:27
gouthamrzuul is quite busy, i'll send the emails and watch this like a hawk22:32
* gouthamr sent22:32
fungiaccepted to openstack-announce22:38
JayFLate +2 on that22:39
opendevreviewMerged openstack/ossa master: OSSA-2026-037 Errata 1  https://review.opendev.org/c/openstack/ossa/+/100240322:45
gouthamr\o/ 22:49
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0110: Self-service password change does not revoke generators  https://review.opendev.org/c/openstack/security-doc/+/100241023:23
*** mrunge_ is now known as mrunge23:24
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0110: Self-service password change does not revoke generators  https://review.opendev.org/c/openstack/security-doc/+/100241023:31
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0110: Self-service password change does not revoke generators  https://review.opendev.org/c/openstack/security-doc/+/100241023:38
opendevreviewGrzegorz Grasza proposed openstack/security-doc master: OSSN-0110: Self-service password change does not revoke generators  https://review.opendev.org/c/openstack/security-doc/+/100241023:41

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!