Issue metadata
Sign in to add a comment
|
CVE-2017-7526 gcrypt RSA side-channel |
||||||||||||||||||||
Issue descriptionUserAgent: Mozilla/5.0 (X11; Linux x86_64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/60.0.3112.66 Safari/537.36 Steps to reproduce the problem: gcrypt version 1.6.3 is in Chrome OS. gcrypt 1.7.8 fixes CVE-2017-7529. What is the expected behavior? What went wrong? Need newer gcrypt Did this work before? N/A Chrome version: 60.0.3112.66 Channel: beta OS Version: Flash Version:
,
Jul 17 2017
1.7.8 is in gentoo, so we can probably just upgrade to address this.
,
Jul 17 2017
,
Jul 17 2017
Seems potentially feasible to run the side-channel attack from NaCl. (Whether any part of Chrome OS actually does the relevant operations using gcrypt rather than some other library is another question.)
,
Jul 17 2017
we disable libgcrypt in most packages. a cursory check suggests crda (for verification?) & samba (for kerberos auth) are the only ones currently using it.
,
Jul 18 2017
I've taken a quick look: libgcrypt-1.7.8 depends on >=libgpg-error-1.13. The latter fails to build (even the most recent 1.27-r1 from gentoo) due to its inability to deal with our x86_64-cros-linux-gnu CHOST. Furthermore, the libgcrypt configure script checks for the libgpg-error version installed in the system, but fails to do so correctly for cross compilation (i.e. invokes the SDK's /usr/bin/gpg-error-config). Finally, there seems to be some issue with the libgcrypt build not selecting the correct asm implementations for crypto algorithms, leading to link errors. After hacking around all these, I got a successful x86_64 build at least. Here's the hacked up version of the ebuilds for reference: https://chromium-review.googlesource.com/575134 There's definitely work to be done to fix up the packages to build properly. I'm going to throw this over the wall to the chromad team who owns samba on Chrome OS. Folks, can you sort out the upgrade soonish?
,
Jul 18 2017
,
Jul 18 2017
Sure, I'll take a look after M61 branch.
,
Jul 18 2017
,
Jul 24 2017
,
Jul 24 2017
Re - libgpg-error-1.27-r1 not compiling due to cros CHOST:
I've modified Mattias' hack to make it compile on amd64, x86 and arm. mkheader.c has a hardcoded map for CHOST where we could add cros, but that seems ugly since having everything in the ebuild is easier to read. I'm not sure what the right solution is here.
Re - libgcrypt using /usr/bin/gpg-error-config instead of /build/${BOARD}/usr/bin/gpg-error-config:
I've added
export GPG_ERROR_CONFIG="${EROOT}/usr/bin/gpg-error-config"
and verified that does the right thing.
Re - libgcrypt not selecting the correct asm implementations:
I've played around with the ebuild and made it compile on all 3 platforms without disabling asm. Not sure what exactly is going on, though. I believe mpi/config.links picks the asm based on platform.
,
Jul 26 2017
Hey Mike, mind to take a look at the CL? Is this an acceptable solution? https://chromium-review.googlesource.com/575134
,
Jul 27 2017
,
Dec 21 2017
@ljusten, any plans to submit the CL mentioned in #C12?
,
Jan 30 2018
ljusten@, are you still intending to submit the CL to fix this?
,
Jan 30 2018
Mike, could you take a look at the CL, please? Looks like I've linked the wrong one in #12. https://chromium-review.googlesource.com/c/chromiumos/overlays/portage-stable/+/594728
,
Mar 7 2018
,
Apr 19 2018
,
May 30 2018
,
Jul 25
,
Sep 5
,
Oct 17
,
Nov 26
,
Nov 26
,
Dec 5
,
Jan 16
(6 days ago)
vapier@ can you please comment on #18? |
|||||||||||||||||||||
►
Sign in to add a comment |
|||||||||||||||||||||
Comment 1 by keescook@chromium.org
, Jul 17 2017Labels: -Restrict-View-SecurityTeam Security_Impact-Stable Security_Severity-Low