Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Fri, 11 Aug 2017 00:01:25 +0200
From: Solar Designer <>
Cc: Brad Spengler <>,
	Vladis Dronov <>
Subject: Re: [CVE-2017-7533] kernel: inotify: a race between inotify_handle_event() and sys_rename()


Thank you for asking the right questions, Brad.  Thank you for providing
additional detail, Vladis.  I'll add some more:

On Mon, Aug 07, 2017 at 09:55:36AM -0400, Vladis Dronov wrote:
> As for the timeline as we understand it, we were notified about the flaw, we've discovered the flaw
> being actually already fixed in the upstream, we've notified linux-distros@ and in a week we've made
> this public announce.

For linux-distros, cases like this are a gray area.  On one hand, we
state that the list is for non-public security issues only, and if
anything already public is wrongly sent to that list then it should be
posted to oss-security right away.  On the other hand, sometimes it
happens that (at least) the reporter considers the issue semi-public
rather than fully public - typically when public information about the
issue is incomplete, relatively obscure, or/and is in a relatively
obscure place.  Then it's sometimes tough to decide between insisting on
immediate posting to oss-security or leaving distros some time to
prepare updates before the issue turns from semi-public to fully public
(if this is even a valid distinction, on which opinions vary).
Sometimes the issue being semi-public results in a shorter
(semi-)embargo period than what would normally be used for a similar
issue that is considered fully private.

In this specific case, the issue itself was never sent to the
linux-distros list itself.  Instead, Red Hat posted the following on
July 26:

| we've been informed of a local privilege escalation issue in the Linux
| kernel. CVE-2017-7533 has been assigned to this issue. The unembargo
| date is set to August 3rd (Thursday next week) 15:00 UTC.
| For detailed information please email Vladis Dronov [1]
|   [1]

This kind of indirect disclosure is unusual, but it happens from time to
time.  I did e-mail Vladis for the detail, and learned from the helpful
reply that the issue was actually semi-public.  There was no discussion
(that I am aware of) on whether this specific issue being semi-public
required bringing it to oss-security earlier than initially proposed.
Even if there were, it'd only affect the last one week out of two months.

We could insist that semi-public issues being brought to (linux-)distros
be posted to oss-security right away, but this would result in distros
who learn of such issues from elsewhere choosing not to inform their
fellow distros, or at least not via the (linux-)distros list, as the
distros would not want to make things worse for themselves and for most
of their users (except for some advanced users, who could apply
workarounds) by making issues fully public before updates are ready.

My reconstruction of the timeline is as follows:

2017-05-31 postings to linux-fsdevel by Leilei Lin of Alibaba Group
2017-07-06 private Bug created in Red Hat Bugzilla (now public, edited)
2017-07-07 upstream fix
2017-07-26 notification to linux-distros
2017-08-03 posting to oss-security, full-disclosure, Bugtraq

Indeed, I am unhappy about a two-month timeline like this and about
semi-public issues in general (the worst state an issue can be in),
but I think the last week was actually the least problematic: multiple
distros were finally aware and could work on fixes, still before the
issue got widespread attention.


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.