depends: Bump OpenSSL to 1.0.1m #5929

pull laanwj wants to merge 1 commits into bitcoin:master from laanwj:2015_03_bump_openssl changing 1 files +2 −2

laanwj commented at 5:24 PM on March 19, 2015: member

Nothing that should affect consensus, but bumping to the most recent OpenSSL version seems prudent for the TLS usage.

depends: Bump OpenSSL to 1.0.1m

Nothing that should affect consensus, but bumping to the most
recent OpenSSL version seems prudent for the TLS usage.

4f96fb8236

laanwj added the label Build system on Mar 19, 2015
gmaxwell commented at 5:27 PM on March 19, 2015: contributor

I'm concerned that the diff between this and the prior version is about 750kloc, it's huge even with whitespace elimination. It's basically unreviewable. We need to let the OpenSSL project know that this is unacceptable in an urgent security update backport release.
laanwj commented at 5:36 PM on March 19, 2015: member

Yes, it's crazy.
luke-jr commented at 3:16 PM on March 21, 2015: member

FWIW, Gentoo's security team believes this patch backports only the security fixes to 1.0.1l: http://sources.gentoo.org/cgi-bin/viewvc.cgi/gentoo-x86/dev-libs/openssl/files/openssl-1.0.1l-CVE-2015-0286.patch?revision=1.1 See also: https://bugs.gentoo.org/show_bug.cgi?id=543552
gmaxwell commented at 9:46 PM on March 21, 2015: contributor

Debian is apparently doing the same thing.
theuni commented at 1:34 PM on March 22, 2015: member

@luke-jr thanks for pointing that out.

Here's a change to take Gentoo's patch rather than 1.0.1m if we decide to go that route. Seems like a sane approach to me, but the patch would need a thorough audit first.

https://github.com/theuni/bitcoin/commit/3b0c532c5985aa3744706108a5b8efb91698db27
theuni commented at 6:56 PM on March 23, 2015: member
As discussed on IRC: It looks like Gentoo only went for a subset of the changes.

I did my own quick+dirty backport based on changes presented in the changelog. I don't claim that it's correct (probably far from it), but it illustrates the differences and the approach we may wish to use instead: https://github.com/theuni/openssl/tree/OpenSSL_1_0_1m-backport

Compared to gentoo's patch: https://gist.github.com/theuni/10a5345be3ef2daf1d70
- Gentoo modified ASN1_item_d2i() in crypto/asn1/tasn_dec.c in a way that differs from upstream. I'm not sure where theirs comes from.
- Gentoo misses CVE-2015-0288: https://github.com/openssl/openssl/commit/51527f1e3564f210e984fe5b654c45d34e4f03d7
- Gentoo misses https://github.com/openssl/openssl/commit/bc2e18a3c818ae7e2d8c996b6648aa4ae8e3ee28. No CVE here, but it was necessary to take before cd56a08d4e1dcae6a0ad8a5b39512fb80ccd1b73 for CVE-2015-0293.
- Gentoo misses https://github.com/openssl/openssl/commit/8ca79fcbf45ea2ed505679be20e1a8a4c3df07cf (Part of CVE-2015-0293. I'm unsure if this is only cosmetic)
- Gentoo misses https://github.com/openssl/openssl/commit/a4517be9e348634ac64f9cf093131e13e8c03e38. Follow-up of CVE-2015-0209. Doesn't look significant.
I'm not at all familiar with the openssl codebase and certainly not qualified to do the backporting, the above is just an at-a-glance review.
laanwj commented at 9:59 AM on March 24, 2015: member

I'm not at all familiar with the openssl codebase and certainly not qualified to do the backporting

Same here. I'm not at all happy to get into the maintenance of OpenSSL forks. It is so easy to get something wrong, as Debian did back with the entropy issue. If we think 1.0.1m is too risky compared to the fixes, I'd rather just forego the version bump (and work harder on getting rid of OpenSSL).
laanwj commented at 12:35 PM on March 30, 2015: member

Closing this, seems not worth the risk.
laanwj closed this on Mar 30, 2015
MarcoFalke locked this on Sep 8, 2021

Contributors

Labels