libssh2 1.11.1 release?

[vszakats]

There has been a fair amount of mostly fixes and tidy-ups since 1.11.0.

196 commits at the moment: libssh2-1.11.0...f0e23da

With niceties such as OpenSSL 3 no-deprecated compatibility, REUSE
compatibility, disablable MD5, AES-GCM-encrypted private keys, CMake
improvements, *BSD CI tests, a few build regression fixes and more. Almost
all of these backed by new CI tests (up to 84 from 61). We also lose the
Makefile.mk and NMakefile build methods. And fix CVE-2023-48795.

It'd be real nice to include the [email protected] patch
in this release. [next time]

And still pending #1243, which is green now and ready for review. [DONE]

After this release it came up to do an ABI break next. This would allow
to address this long-term issue: #1220 and those listed in TODO. In case
we agree to commit to this.

/cc @libssh2/dev-team

[bagder]

I'm 👍 for a release. We mostly need to get the RELEASE-NOTES done, then I think I can do all the mechanics necessary to ship.

[vszakats]

Release notes updates here:
https://github.com/libssh2/libssh2/blob/52c382f7aae2a8bff4ce3b4b83d18dde30ea11e2/RELEASE-NOTES

It's a fresh script + manual work. I did my best, but please let me know
if somebody or something is missing or might need improvement.

Helper script here:
https://gist.github.com/vszakats/70dbe907de571c2beaa5e793d4f868da

PR for further updates:
#1279

[vszakats]

If nobody has anything else pending, we're ready for release.

[pm-cfs]

Would be great to get the Terrapin fix in there for CVE-2023-48795 - given the timeframe between releases, addressing this major SSH issue now seems timely.

[vszakats]

We merged the fix, it will be in the release.

[ivoanjo]

+1 would be really cool to see a release out with #1058 :)

[willco007]

@MichaelBuckley is investigating a possible regression, please hold off until we give an all clear.

[willco007]

OK, fix has been landed. All clear.

[vszakats]

@willco007 & @MichaelBuckley: Thanks!

Any chance somebody would be kind to peer-review my #1297 and #1301?
These are also security related and it'd be nice to include them in the release,
and that could really use such review and approval to avoid any mistake there.

[MichaelBuckley]

I'll take a look in a little bit here.

[vszakats]

@MichaelBuckley Thank you for you reviews! Both patches are merged now.

[edward-shen]

Hey all! Just wondering on the status of this. Are there any remaining blockers left? Or are we just waiting on a cut to be made?

[vszakats]

There is #1303 that sprung from the review of #1301; adding futher low-level fail checks. Nothing else.

[vszakats]

Merged now.

[vszakats]

Comments and suggestions are appreciated for deprecation
plans for obsolete algos: #1307

[tio-checo]

Hi guys,
I am wondering if there is any ETA for 1.11.1 release. In particular, I am interested in fix for CVE-2023-48795.
Thank you,
Jan

[pm-cfs]

Starting to get comments from users using ChaCha as many servers rollout their Terrapin updates.
Any way we can get a release out in the next 5-7 days?

Lots of other good stuff in here that would be nice to have too.
Thanks for all the hard work; we do appreciate it!

[miertschink]

I am still confused if the use of libssh2 makes my setup really vulnerable to CVE-2023-48795

afaik libssh2 does not support ChaCha20-Poly1305 and also not any MAC algorithm suffixed [email protected]

but... here is part of the FAQ from terrapin-attack.com:
In more technical terms, if your SSH implementations supports (and is configured to offer) the [email protected] encryption algorithm, or any encryption algorithm suffixed -cbc in combination with any MAC algorithm suffixed [email protected], you are vulnerable to Terrapin.

So even if there is no release (that enables support of strict key exchange) I am still not vulnerable to this attack as libssh2 does not support this combination.

Is that correct?

[tio-checo]

Actually, libssh2.org lists [email protected] & [email protected] as supported MAC algorithms.

[miertschink]

ah, I see
I am using PHP with libssh2 and looked it up here: https://www.php.net/manual/en/function.ssh2-connect.php
there is this parameter "methods" where a client_to_server or server_to_client configuration array can be passed
The "Supported Values" there doesn't mention these.
However there is a hint that says
"* - Supported Values are dependent on methods supported by underlying library. See » libssh2 documentation for additional information."

Ok, basically I am vulnerable.
Thank you!

[YamsThePotato]

Hello! Is it still planned to release a 1.11.1 release? I'm specifically interested in getting the fix for CVE-2023-48795.

[sburnwal]

I am interested in this too CVE-2023-48795. I understand it would be released but would like to know a rough estimate of the timeline to release it.

[vszakats]

After processing the backlog and merging more improvements and fixes,
the repo is (again) fit for a release 1.11.1.

[ #1315 is a larger PR pending addition. It may take some more while
to merge it, and perhaps better fitting 1.12.0. ] [merged]

/cc @bagder

[Tejaswikandula]

Does the release 1.11.1 including all commits from latest master. I'm interested to know if #1314 would be part of 1.11.1 release

[willco007]

Yes it would.

[dmondrik]

Any update on a date when we might see the release tag created?

[Jan-E]

I am admiring all the work that @vszakats is doing the last time for LIbSSH2, but really would like to see a release as well. The Terrapin fix was already merged in December. Is anything blocking the 1.11.1 release?

[vszakats]

To make the wait more useful for everyone, we appreciate testing and feedback
on current master (as a "release candidate") while a release is cut.

Also for somebody interested in mbedTLS, we have an unfixed regression with
mbedTLS 3.6.0: 2e4c5ec

libssh2/src/mbedtls.c

Lines 1295 to 1315 in 3b23e03

	#if MBEDTLS_VERSION_NUMBER >= 0x03060000
	/* FIXME: implement this functionality via a public API */
	(void)session;
	(void)filename;
	(void)pwd;
	data = NULL;
	data_len = 0;
	#else
	if(mbedtls_pk_load_file(filename, &data, &data_len))
	goto cleanup;
	mbedtls_pk_init(&pkey);
	if(_libssh2_mbedtls_parse_eckey(ctx, &pkey, session,
	data, data_len, pwd) == 0)
	goto cleanup;
	_libssh2_mbedtls_parse_openssh_key(ctx, session, data, data_len, pwd);
	cleanup:
	#endif

I was thinking to make it a compile-time failure outside of CI, to avoid
runtime surprises. That's the only pending thing I have. But a proper fix of
course would be much better. → #1393

[vszakats]

We also have this unfixed regression with wolfSSL, since 1.11.0:
#1020, #1299