Univention Bugzilla – Bug 41693
linux: Multiple security issues (3.3)
Last modified: 2016-10-27 18:21:56 CEST
Upstream Debian (Jessie) package version 3.16.7-ckt25-2+deb8u2 fixes the following issues:
CVE ID : CVE-2015-7515 CVE-2016-0821 CVE-2016-1237 CVE-2016-1583
CVE-2016-2117 CVE-2016-2143 CVE-2016-2184 CVE-2016-2185
CVE-2016-2186 CVE-2016-2187 CVE-2016-3070 CVE-2016-3134
CVE-2016-3136 CVE-2016-3137 CVE-2016-3138 CVE-2016-3140
CVE-2016-3156 CVE-2016-3157 CVE-2016-3672 CVE-2016-3951
CVE-2016-3955 CVE-2016-3961 CVE-2016-4470 CVE-2016-4482
CVE-2016-4485 CVE-2016-4486 CVE-2016-4565 CVE-2016-4569
CVE-2016-4578 CVE-2016-4580 CVE-2016-4581 CVE-2016-4805
CVE-2016-4913 CVE-2016-4997 CVE-2016-4998 CVE-2016-5243
Several vulnerabilities have been discovered in the Linux kernel that
may lead to a privilege escalation, denial of service or information
CVE-2015-7515, CVE-2016-2184, CVE-2016-2185, CVE-2016-2186,
CVE-2016-2187, CVE-2016-3136, CVE-2016-3137, CVE-2016-3138,
Ralf Spenneberg of OpenSource Security reported that various USB
drivers do not sufficiently validate USB descriptors. This
allowed a physically present user with a specially designed USB
device to cause a denial of service (crash).
Solar Designer noted that the list 'poisoning' feature, intended
to mitigate the effects of bugs in list manipulation in the
kernel, used poison values within the range of virtual addresses
that can be allocated by user processes.
David Sinquin discovered that nfsd does not check permissions when
setting ACLs, allowing users to grant themselves permissions to a
file by setting the ACL.
Jann Horn of Google Project Zero reported that the eCryptfs
filesystem could be used together with the proc filesystem to
cause a kernel stack overflow. If the ecryptfs-utils package is
installed, local users could exploit this, via the
mount.ecryptfs_private program, for denial of service (crash) or
possibly for privilege escalation.
Justin Yackoski of Cryptonite discovered that the Atheros L2
ethernet driver incorrectly enables scatter/gather I/O. A remote
attacker could take advantage of this flaw to obtain potentially
sensitive information from kernel memory.
Marcin Koscielnicki discovered that the fork implementation in the
Linux kernel on s390 platforms mishandles the case of four
page-table levels, which allows local users to cause a denial of
service (system crash).
Jan Stancek of Red Hat discovered a local denial of service
vulnerability in AIO handling.
The Google Project Zero team found that the netfilter subsystem does
not sufficiently validate filter table entries. A user with the
CAP_NET_ADMIN capability could use this for denial of service
(crash) or possibly for privilege escalation. Debian disables
unprivileged user namespaces by default, if locally enabled with the
kernel.unprivileged_userns_clone sysctl, this allows privilege
Solar Designer discovered that the IPv4 implementation in the Linux
kernel did not perform the destruction of inet device objects
properly. An attacker in a guest OS could use this to cause a denial
of service (networking outage) in the host OS.
CVE-2016-3157 / XSA-171
Andy Lutomirski discovered that the x86_64 (amd64) task switching
implementation did not correctly update the I/O permission level
when running as a Xen paravirtual (PV) guest. In some
configurations this would allow local users to cause a denial of
service (crash) or to escalate their privileges within the guest.
Hector Marco and Ismael Ripoll noted that it was possible to disable
Address Space Layout Randomisation (ASLR) for x86_32 (i386) programs
by removing the stack resource limit. This made it easier for local
users to exploit security flaws in programs that have the setuid or
setgid flag set.
It was discovered that the cdc_ncm driver would free memory
prematurely if certain errors occurred during its initialisation.
This allowed a physically present user with a specially designed
USB device to cause a denial of service (crash) or possibly to
escalate their privileges.
Ignat Korchagin reported that the usbip subsystem did not check
the length of data received for a USB buffer. This allowed denial
of service (crash) or privilege escalation on a system configured
as a usbip client, by the usbip server or by an attacker able to
impersonate it over the network. A system configured as a usbip
server might be similarly vulnerable to physically present users.
CVE-2016-3961 / XSA-174
Vitaly Kuznetsov of Red Hat discovered that Linux allowed the use of
hugetlbfs on x86 (i386 and amd64) systems even when running as a Xen
paravirtualised (PV) guest, although Xen does not support huge
pages. This allowed users with access to /dev/hugepages to cause a
denial of service (crash) in the guest.
David Howells of Red Hat discovered that a local user can trigger a
flaw in the Linux kernel's handling of key lookups in the keychain
subsystem, leading to a denial of service (crash) or possibly to
CVE-2016-4482, CVE-2016-4485, CVE-2016-4486, CVE-2016-4569,
CVE-2016-4578, CVE-2016-4580, CVE-2016-5243, CVE-2016-5244
Kangjie Lu reported that the USB devio, llc, rtnetlink, ALSA
timer, x25, tipc, and rds facilities leaked information from the
Jann Horn of Google Project Zero reported that various components
in the InfiniBand stack implemented unusual semantics for the
write() operation. On a system with InfiniBand drivers loaded,
local users could use this for denial of service or privilege
Tycho Andersen discovered that in some situations the Linux kernel
did not handle propagated mounts correctly. A local user can take
advantage of this flaw to cause a denial of service (system crash).
Baozeng Ding discovered a use-after-free in the generic PPP layer in
the Linux kernel. A local user can take advantage of this flaw to
cause a denial of service (system crash), or potentially escalate
Al Viro found that the ISO9660 filesystem implementation did not
correctly count the length of certain invalid name entries.
Reading a directory containing such name entries would leak
information from kernel memory. Users permitted to mount disks or
disk images could use this to obtain sensitive information.
CVE-2016-4997 / CVE-2016-4998
Jesse Hertz and Tim Newsham discovered that missing input sanitising
in Netfilter socket handling may result in denial of service. Debian
disables unprivileged user namespaces by default, if locally enabled
with the kernel.unprivileged_userns_clone sysctl, this also allows
More CVEs deb8u3: <https://anonscm.debian.org/cgit/kernel/linux.git/log/?h=wheezy-backports>
* ALSA: compress: fix an integer overflow check (CVE-2014-9904)
* [amd64] misc: mic: Fix for double fetch security bug in VOP driver (CVE-2016-5728)
* [powerpc*] tm: Always reclaim in start_thread() for exec() class syscalls (CVE-2016-5828)
* HID: hiddev: validate num_values for HIDIOCGUSAGES, HIDIOCSUSAGES commands (CVE-2016-5829)
* [s390*] sclp_ctl: fix potential information leak with /dev/sclp (CVE-2016-6130)
repo_admin.py -U -p linux -d wheezy-backports -r 3.3 -s errata3.3-0
# linux 3.16.7-ckt25-2+deb8u3~bpo70+1
repo_admin.py -e -r 3.3 -s errata3.3-0 -p linux /var/build/temp/linux
# r16617 | linux-3.16.7-ckt25-2+deb8u3~bpo70+1
r71174 | Bug #41693 kernel: Update to 3.16.7-ckt25-2+deb8u3~bpo70+1
r71177 | Bug #41693 kernel: Update to 3.16.7-ckt25-2+deb8u3~bpo70+1 YANK
QA: uname -a # 3.16.0-ucs197-686-pae
QA: uname -a # 3.16.0-ucs197-amd64
I guess we should pick up the forthcoming yet unreleased tcp-make-challenge-acks-less-predictable.patch from https://anonscm.debian.org/cgit/kernel/linux.git/commit/?h=jessie-security&id=1bd5c3370523e5846019361b33a97c754db76f8d . While it may not be that critical, and there is a workaround via sysctl.conf, it enjoys a certain publicity currently and customers will inquire.
New issues, currently only fixed in Debian sid (currently at Kernel 4.7):
* Race condition in the audit_log_single_execve_arg function in kernel/auditsc.c in the Linux kernel through 4.7 allows local users to bypass intended character-set restrictions or disrupt system-call auditing by changing a certain string, aka a "double fetch" vulnerability. (CVE-2016-6136)
* Race condition in the ioctl_send_fib function in drivers/scsi/aacraid/commctrl.c in the Linux kernel through 4.7 allows local users to cause a denial of service (out-of-bounds access or system crash) by changing a certain size value, aka a "double fetch" vulnerability. (CVE-2016-6480)
* Linux tcp_xmit_retransmit_queue use after free (CVE-2016-6828)
(In reply to Arvid Requate from comment #3)
> I guess we should pick up the forthcoming yet unreleased
> tcp-make-challenge-acks-less-predictable.patch from
> security&id=1bd5c3370523e5846019361b33a97c754db76f8d . While it may not be
> that critical, and there is a workaround via sysctl.conf, it enjoys a
> certain publicity currently and customers will inquire.
This is CVE-2016-5696
CP 1bd5c3370523e5846019361b33a97c754db76f8d from 3.16
(In reply to Arvid Requate from comment #4)
> * Race condition in the audit_log_single_execve_arg function in
> kernel/auditsc.c in the Linux kernel through 4.7 allows local users to
> bypass intended character-set restrictions or disrupt system-call auditing
> by changing a certain string, aka a "double fetch" vulnerability.
CP e06f0a239348a6a8b357554c0dc4720df11638b2 from 3.16
> * Race condition in the ioctl_send_fib function in
> drivers/scsi/aacraid/commctrl.c in the Linux kernel through 4.7 allows local
> users to cause a denial of service (out-of-bounds access or system crash) by
> changing a certain size value, aka a "double fetch" vulnerability.
CP b713215e8e54aa8852ae307bf4a07bbb2db66bf0 from 3.16
> * Linux tcp_xmit_retransmit_queue use after free (CVE-2016-6828)
CP 121b60af11646210937eba34dfadb30f04f96fb5 from 3.16
CVE-2016-7118; regression in 3.16.36-1.
not needed: 5a092f0ef9f759d382d11770a980b34e375a71b1 from 3.16
r16684 | Bug #41693: linux-3.16.x for UCS-3.3
r72256 | Bug #41693, Bug #42099: More 3.16.7-ckt25-2+deb8u3~bpo70+1
r72306 | Bug #41693, Bug #42099: More 3.16.7-ckt25-2+deb8u3~bpo70+1 YAML
Just a question: Jessie security updated to 3.16.36-1+deb8u1, what is our strategy, is it easier to update to or to continue backporting in the future?
(In reply to Arvid Requate from comment #6)
> Just a question: Jessie security updated to 3.16.36-1+deb8u1, what is our
> strategy, is it easier to update to or to continue backporting in the future?
The "Jessie" linux package does not compile on wheezy; see first entry of <http://metadata.ftp-master.debian.org/changelogs//main/l/linux/linux_3.16.7-ckt25-2+deb8u3~bpo70+1_changelog>
You either have to make the Jessie kernel compile on Wheezy (or even Squeeze which UCS-3 is based on), or need to pick the patches yourself from Jessie missing in Wheezy-Backport.
I've choose to lessen my pain by cherry-picking the missing patches.
Ok, the patches are part of the linux source package 3.16.38 packaged
for Bug 42353:
The patch for CVE-2016-5696:
jessie-security git commit 1bd5c3370523e5846019361b33a97c754db76f8d
linux-stable git commit c6287499662db0b0caee72d6453e445a2c6162af
part of branch linux-3.16.y
The patch for CVE-2016-6136:
jessie-security git commit e06f0a239348a6a8b357554c0dc4720df11638b2
linux-stable git commit c8a3667e957a498865ec82c3d19160f1a330daef
part of branch linux-3.16.y
The patch for CVE-2016-6480:
jessie-security git commit b713215e8e54aa8852ae307bf4a07bbb2db66bf0
now part of 3.16.38-0.1+deb8u2~bpo70+1-errata3.3-0 applied as debian/patches:
The patch for CVE-2016-6828:
jessie-security git commit 121b60af11646210937eba34dfadb30f04f96fb5
now part of 3.16.38-0.1+deb8u2~bpo70+1-errata3.3-0 applied as debian/patches:
*** Bug 42353 has been marked as a duplicate of this bug. ***