Samba Security Process

From SambaWiki
Revision as of 08:00, 17 June 2023 by Abartlet (talk | contribs) (Add directory listings to 'Some specific things not to report to Samba')
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)

Below is the process the Samba Team follows to prepare a security release and co-ordinate the disclosure of a security vulnerability.

Reporting Security Defects in Samba

Please report all security issues or defects in Samba or concerning our infrastructure to and never on IRC, public mailing lists or in Bugzilla!

For clarity, a Samba Team member will create the private Bugzilla issue mentioned below.

Some specific things not to report to Samba

A public security process for Samba

The following process is followed by the Samba Team Member working on the security issue.

We publish it as a service to the Free Software and broader software development community as a well-worked, practical process for responsible disclosure and to our aid our Samba community in knowing how we handle serious security issues in our code.

Security issues are usually reported via

1.) Someone should feel responsible and make sure that it's really a
    security defect. If yes, we need one person who is responsible for
    this specific issue beyond this point. This person coordinates the
    further actions and might delegate tasks, of course.

2.) Create a bug report with limited access to "Samba core developers"
    only!  This is done by selecting 'Show advanced fields' and at the
    very bottom:

   'Only users in all of the selected groups can view this bug:

    [x] Samba Core Developers

2a.) Change the summary so that it starts with "[EMBARGOED][SECURITY]"

    For example:

    "[EMBARGOED][SECURITY] dsdb: time wrap at end of unix time

2b.) Add all information including reproducer etc. to the bug report.

3.) Write an initial advisory and attach it to the bug report. A
    template is available here:

3a.) Confirm with the initial reporter that they are willing to be
     publicly thanked and any affiliation they wish to have noted.

4.) Do a CVSSv3.1 calculation ([2]) and add it to the bug report and the

4a.) Stop and ask for advise on if the CVSSv3.1
     score is less than 5.0.  This is a heavy-weight process with
     costs to many parties.

4b.) Even for a CVSSv3.1 of more than 5.0 we do not do a security
     release where the impact is:

     * Only a Denial of Service (not a use after free) and the
       automated internal process restart will allow other clients to
       reconnect eventually.

       For clarity, the shared LDAP or KDC process in the AD DC will
       restart, so a simple NULL de-reference crash there would not
       trigger a release.  However a similar issue in the main smbd
       listener preventing new connections being established may
       require a release.

       This is similar to, but an extension of, our existing rule that
       crashing the smbd belonging to the connecting user is not a

     * Mitigated by the selection of strong passwords.

       We will not do a security release simply because weak passwords
       are permitted to be selected (eg our password policy tools fail
       to exclude all weak passwords) or a weak password can be brute
       forced online (eg avoiding a bad password lockout).

     These issues are real, important and should be backported to
     maintained releases, but do not justify the costs of coordinated
     disclosure or a seperate security release.

     However, we do:
      - ask for a CVE (step 5.)
      - place the CVE on the bug (step 5b.)
      - leave or make the bug public (step 14. and 14a.)
      - mark the commits with the CVE (step 6a.)

4c.) In general we do not take security release steps at all where the
     issue is:

     * Samba's historical acceptence of now-weak cryptography

       Unless of course a new novel attack is found, rather than just
       the slow march of time.

     * Reliant on a timing attack

       For example newer Samba versions use a constant-time memcmp()
       from GnuTLS.  The failure of this to be constant-time or the
       lack of use of constant-time methods in a particular place will
       just be fixed normally.

     * User enumeration / username discovery

       It is assumed that in our deployed environments usernames are
       available to an attacker.  In the AD DC Kerberos will readily
       confirm that a user does or does not exist.

     * Relient on offline brute force of a weak password

       The NTLM challenge response and Kerberos traffic (without FAST,
       which is rare) is well known to expose a user's password to a
       brute force attack.

     However do ask for advise on if you feel
     concerned that any particular issue may be an exception.

5.) Ask for a CVE (usually Red Hat Product Security
    <> does help out on this one).

To:      Red Hat Product Security <>
Subject: Another Samba CVE (bug XXXX)

Red Hat Product Security,

Can I please get a CVE for what will be Samba Bugzilla:

The Samba team has evaluated this issue and determined that it will
require a CVE per our process.

Details are confidential for now, but will be clearly marked with that
bug (so you can connect the threads once vendors are notified or the
issue is made public).


    If the bug is already public, remove
     "are confidential for now, but "
    and feel free to describe the issue in cleartext.

    DO NOT attach the advisory from step 3.  This avoids two issues:
     - If Red Hat gets the advisory, they need to get GDPR permission for
       the reporter name and we don't want to block on the back and forth
       for that (see 3a).
     - It keeps all vendors on the same basis.

     See [1] for GPG key etc if the circumstances require it (if for
     example you do feel the need to disclose details or seek detailed

5b.) If for any reason Red Hat is unable to assist with obtaining a
     CVE number in time, the backup process is to fill in this form (explain the situation in the

5c.) After receiving the CVE number, please add it as an
     alias (this is a seperate bugzilla field) on the bug report
     and to the summary (title), eg:
     CVE-2038-12345 [EMBARGOED][SECURITY]...

6.) Write patches and tests and add them to the bug report.  Mark them
    WIP in the description until they are 'final'.
    (Even 'final' patches may change after being subject to review of

6a.) ALL commits in the patch should have a CVE prefix and a BUG tag eg:

  CVE-2038-12345 dsdb: time wrap at end of unix time

  At time_t rollover all AD security is lost due to integer wrap on
  32-bit systems


  Signed-off-by: Time Lord <>

6b.) Ask for early review of your WIP and final master patches ASAP

6c.) Prepare backports for all affected and supported versions

     (1 file per version, even if identical).

     Files should be named $CVE-$DESCRIPTION-$MAJOR-$VERSION.patch


6d.) Run *private* autobuild (see doc/autobuild.txt) or *private*
     Gitlab CI for each CVE and each maintained branch.  (This allows
     the release manager to defer some patches if problems appear
     later in the process).

6e.) Mark the patch that has passed with the ci-passed flag in the
     bugzilla attachment details.

6f.) Once each CVE patch and branch has passed, ask for review ASAP.

6g.) Get one explicit review flag in Bugzilla (yours, as a team member,
     is implicit and CI passed is now a distinct flag).

7.) Once all patches and the advisory are available and reviewed, ask
    the release manager for a release date.

7a.) Between now and the release, the release manager or their
     delegate will run a *private* autobuild of the exact combination
     of patches the be released.

7b.) If multiple issues are being released in a batch, create a
     tracking bug that is blocked by the CVEs to be addressed in this

8.) Finish the advisory and attach it to the bug report as
    advisory-$CVE.txt.  Ask for review.  (This must be after getting
    the release date so the correct version numbers can be included).

9.) ~10 days before the release date, confirm every patch and the CVE
    text still has the correct reviews (due to additional feedback for

9a.) add samba-vendor to the CC list in the bug report,

9b.) open the bug report by checking "Core developers, and Vendors
     shipping Samba as part of their products" and removing "Samba
     Core Developers".

9c.) Add the planned release date to the bug or the tracking bug if
     that has been created.  This will be the first useful e-mail seen
     by our vendors.

10.) 7 days before the release date, send an e-mail to, and indicating there will be a security
     release and the broad component it impacts.  This is to allow
     large installations time to prepare for security patching.

     For example:

Cc:	 Upstream Samba Technical Mailing list <>,
Subject: Heads-up: Security Releases ahead!

this is a heads-up that there will be Samba security updates on
Tuesday, XXXXX YYth. Please make sure that your Samba
servers will be updated soon after the release!

Impacted components:
 - AD DC (CVSS 7.5, High)
 - client (CVSS 5.9, Medium)
 - file server / classic DC (CVSS 6.8 Medium)

10a.) Note that this represents the final GO/NO-GO point.  In order to
      facilitate creation of tarballs etc, after this date the patches
      and CVE text *must not change* or else the release will need to
      be publicly rescheduled, patches dropped or other appropriate
      action taken at the absolute discretion of the release manager.

11.) The release manager or their delegate will prepare the tarballs
     in the *stable* branches:
    - git fetch
    - git rebase
    - git merge -ff-only the VERSION bump commit from the corresponding test
    - apply the security patchset
    - WHATSNEW.txt: add the release notes
    - add an empty announce.samba-$VERSION.quotation.txt file
    - GNUPGHOME=$PATH_TO_THE_RELEASE_KEY script/ samba-stable create

    On the release day:
    Upload the files and publish the announcements as described in steps
    3.) and 4.) in team-info/doc/releases/howto_releases.txt 

12.) Additional steps for the release manager:
     - upload the signed security patch(es) to the ftp server:
     - push the advisory to the samba-web git repository:
     - update samba-web/history/security.html

     Merge the tags to the test branches:
     - git fetch
     - git rebase
     - git merge $TAG
     - VERSION: bump version
     - push without autobuild:
       LC_SAMBA_AUTOBUILD_PUSH=1 git push-v4-X-test-skip-autobuild

13.) After shipping the releases, the responsible person must make
     sure that the patches find their way into the master branch and
     remove the [EMBARGOED] tag from the bug report.

14.) Mark as 'private' any sensitive comments or attachments

14a.) Then un-select the "Core developers, and Vendors shipping Samba
      as part of their products" restriction.

15.) Address any minor improvements that were suggested after the
     patches were frozen and incorporate those into commits in master.

16.) Close out bug report when patches have been pushed to *all* relevant

17.) Now that any regressions etc will have been broadly noticed,
     remove samba-vendor from the CC list of the now public bug, ask
     that they subscribe individually if still interested.


Template security advisory

The template security advisory is here: File:Template security advisory.txt

In Video

There is a Video of a presentation by Jeremy Allison about this security process in practice.