Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Date: Sat, 26 Aug 2023 00:23:59 +0200
From: Solar Designer <>
Cc: Vegard Nossum <>, Jiri Kosina <>,
	Donald Buczek <>,
	Greg KH <>
Subject: linux-distros list policy and Linux kernel, again


This is sort of continuation to a thread I started in here last year, at
the time focusing on allowing "embargoes" on publicly fixed Linux kernel
issues, which we ended up agreeing to:

BTW, this recent thread on "Quality standards for embargoed code" is
relevant to that previous topic:

(It's harder/worse to test code changes that are not yet public, which
is something we also recently discussed in context of and granted an
exception to the curl project.)

This time, there are at least two more points of contention to discuss.

Thank you Donald Buczek for bringing this to oss-security (and beating
me to it), with your specific needs explained:

Here I'd like to start a new thread just on oss-security (no cross-post,
but some people CC'ed) and with more context given.

For some years until recently, the Linux kernel's
Documentation/admin-guide/security-bugs.rst directed people to also
report "sensitive bugs, such as those that might lead to privilege
escalations" to linux-distros.  On one hand, that was great -
cooperation between Linux kernel upstream and Linux distros.  On the
other hand, the specific wording did not fully match linux-distros
policy, yet it bypassed a visit to the wiki page on linux-distros with
the instructions and policy on it by directly giving out the posting
address.  This resulted in people reporting issues to linux-distros
without having actually read and accepted the policy.

Last year, Vegard Nossum suggested edits (and then revised them this
year) to try and make things much clearer for reporters:

As I recall, when Vegard offered this edited documentation (the second
URL above) to people who reported something to linux-distros but were
confused, they said that version of the documentation was indeed much
clearer.  However, the changes didn't get merged, which I speculate is
in part a result of them changing several things at once and thus making
it harder for all reviewers to agree on them.  Besides, I am not sure
this was the right direction: maybe rather than duplicate and explain
linux-distros policy aspects in Linux kernel documentation, it was
better to refer to linux-distros instructions and omit the linux-distros
posting address from the kernel documentation.

Then there was the recent lengthy linux-distros and s@k.o thread on the
StackRot (CVE-2023-3269) vulnerability, publicly disclosed here:

In that private thread, two linux-distros policy aspects came up as
being inconsistent with s@k.o preferences:

1. For s@k.o, public disclosure is typically in up to 7 days since fix
is ready, but for linux-distros the deadline is 14 days max since
linux-distros is notified even if the fix is not ready.  Apparently,
this one makes Greg KH particularly unhappy about linux-distros.

2. _If_ the reporter shares PoCs/exploits with linux-distros, then per
current policy those should also be made public (within up to 7 days
more after the vulnerability is publicly disclosed as such).  Linus
Torvalds said that this must be up to the reporter only, and we should
not be imposing such policy.  In a sense, this is already up to the
reporter - if they disagree, they just shouldn't post PoCs/exploits to
linux-distros - can instead post an offer to share with individual
distros if they want to, or not share at all.  However, this is
problematic in practice because not everyone reads the rules before
posting and sometimes people change their mind during the embargo time
(or are required to "change their mind" by their employer, etc.)

I had already recognized both of these as problems - e.g., the latter
one also came up in Mathias Krause's disclosure of another Linux kernel
vulnerability in 2022:

At the time, I thought that maybe increasing the maximum delay for
PoC/exploit publication from 7 to 30 days would be appropriate, but we
did not formally do it.  I effectively granted a similar exception to
Ruihan Li for StackRot this time, even though the full exploit wasn't
even shared with linux-distros and Ruihan was willing to publish the
PoCs that he did share on time.  I also informed Ruihan of this policy
aspect early enough, so we dodged that bullet that time, but at the same
time made Linus aware of this general issue, triggering his criticism.
(That's fine, it's good to have criticism.)

Overall, even though these issues were mostly avoided for StackRot in
particular, bringing them up in this thread ended up being the last
straw for the Linux kernel security folks to say they don't want to deal
with linux-distros anymore, resulting in this edit to the documentation:

This has a spiteful commit message and a calmer almost constructive
actual change.  Knowing that the status quo was problematic and not
having a better solution ready (other than Vegard's elaborate edits), I
actually approved of this edit as well.  It says:

The kernel security team strongly recommends that reporters of potential
security issues NEVER contact the "linux-distros" mailing list until
AFTER discussing it with the kernel security team.  Do not Cc: both
lists at once.  You may contact the linux-distros mailing list after a
fix has been agreed on and you fully understand the requirements that
doing so will impose on you and the kernel community.

The different lists have different goals and the linux-distros rules do
not contribute to actually fixing any potential security problems.

(I take issue with the second paragraph, as I explained in the message
Donald forwarded in here earlier today.)

So the kernel security team can, after having arrived at a fix, choose
to direct the reporter to also contact linux-distros.  I don't know
whether this would actually be happening or not.  Maybe some friendly
dialogue (which I hope this message can be a part of) and agreeing on
things could make that more likely.

I think the two linux-distros policy aspects above do not preclude such
cooperation.  However, I am willing to discuss possible policy changes
(or perhaps Linux kernel exceptions).

I recognize that 14 days might not always be enough to get a fix ready,
especially not for many of the CPU microarchitectural issues being
handled since mid-2017 and affecting many proprietary OSes as well (who
may be used to much longer disclosure timelines and would object to
Linux's earlier fix and disclosure).  I am glad that almost none of such
issues were brought to (linux-)distros, and never when it was still more
than 14 days until public disclosure.  We had a lot of luck there.
Linux kernel security team has its own mailing list (encrypted, so more
secure than s@k.o) for handling of those, which is great:

I mean, this is "great" within the constraints of the rest of the
industry.  Of course, I'd very much like disclosure timelines for that
kind of issues to also become much shorter.  This is just not the case.

In terms of (linux-)distros list policy, what can we do here?  Accept up
to 7 days since fix is ready and thus accept arbitrarily long embargoes
and more likely have issues "requiring" such embargoes brought to the
list?  BTW, for CPU microarchitectural issues, that would probably need
to be for the full distros list, not limited to Linux, and from what I
know disclosure timelines for such issues may be 3 to 12+ months.

As to publishing PoCs/exploits, this is already mitigated by the Linux
kernel documentation edit making it less likely (but far from
impossible) that people would send stuff to linux-distros without being
aware of the policy.  We could further mitigate this issue by allowing
up to 30 days (but perhaps suggesting at most 7 days?)  As I recall, how
much time to allow here was first discussed in context of an Exim issue
that was expected to be so easy to independently come up with an exploit
for that even 7 days seemed excessive, so 24 hours was agreed upon for
that occasion:

However, issues vary, and maybe 1 to 30 days on a case-by-case basis is

Another option is to completely remove this requirement (only for Linux
kernel issues or for all), but there are reasons to keep it:

1. To reduce the value of information staying in linux-distros members'
mailboxes/computers to potential attackers.

The issues with highly valuable information staying in there are that
such people/resources become more lucrative attack targets, and that
compromises/leaks might not be detected, in which case the situation is
worse than with full publication (only bad actors benefit).

2. To let the wider community benefit from those PoCs/exploits.

There are many uses for PoCs/exploits other than testing by upstream
developers or attacking systems in the wild.  Other uses include:
testing of (possibly different or backported) fixes by other distros
(not all of which are on linux-distros for a variety of reasons) and by
advanced users, reuses by other developers and security researchers to
better understand the problem and as templates for further PoC/exploit
development (such as in determining and demonstrating exploitability of
further related bugs), testing and improvement of (non-)effectiveness of
general security hardening mitigations and of security monitoring tools.

I'd appreciate any well-reasoned votes and constructive suggestions.
Maybe there are good ideas that didn't cross my mind yet.



Powered by blists - more mailing lists

Please check out the Open Source Software Security Wiki, which is counterpart to this mailing list.

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.