From: "Paul Barker" <paul@pbarker.dev>
To: "Yoann Congal" <yoann.congal@smile.fr>,
"Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at
Cisco)" <dkelaiya@cisco.com>,
<openembedded-core@lists.openembedded.org>
Subject: Re: [OE-core][wrynose][PATCH v2 2/5] curl: fix CVE-2026-11564
Date: Tue, 17 Nov 2026 09:50:24 +0000 [thread overview]
Message-ID: <DMXE5AL1MFRL.SJ70FR1E8XH2@pbarker.dev> (raw)
In-Reply-To: <DMWR7AJXCNJ1.105TD3KM0QL29@smile.fr>
Hi Darsh,
I'm following up on what Yoann has already said and adding some comments
of my own here.
On Mon Nov 16, 2026 at 3:51 PM GMT, Yoann Congal wrote:
> On Fri Oct 23, 2026 at 11:33 AM CEST, Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco) wrote:
>> From: Darsh Kelaiya <dkelaiya@cisco.com>
>>
>> This patch applies the upstream fix as referenced in [2],
>> using the commit shown in [1].
>>
>> Retain the curl 8.19.0 option-completion API and existing opt-in
>> automatic client certificate behavior. Reset native CA options per
>> transfer without adding the newer initial-origin restriction.
>>
>> [1] https://github.com/curl/curl/commit/d69bfad3fa3daf5e72331f6870667607828d5891
>> [2] https://curl.se/docs/CVE-2026-11564.html
>>
>> Signed-off-by: Darsh Kelaiya <dkelaiya@cisco.com>
>> ---
>> Changes in v2:
>> - Clarify that Schannel selection remains opt-in and disabled by
>> default, preserving the existing CURLSSLOPT_AUTO_CLIENT_CERT flag.
>> - Document the per-transfer native-CA reset and omitted newer
>> initial-origin restriction. Schannel redirect behavior has not been
>> runtime tested.
>> ---
>> .../curl/curl/CVE-2026-11564.patch | 139 ++++++++++++++++++
>> meta/recipes-support/curl/curl_8.19.0.bb | 1 +
>> 2 files changed, 140 insertions(+)
>> create mode 100644 meta/recipes-support/curl/curl/CVE-2026-11564.patch
>>
>> diff --git a/meta/recipes-support/curl/curl/CVE-2026-11564.patch b/meta/recipes-support/curl/curl/CVE-2026-11564.patch
>> new file mode 100644
>> index 0000000000..86f0181630
>> --- /dev/null
>> +++ b/meta/recipes-support/curl/curl/CVE-2026-11564.patch
>> @@ -0,0 +1,139 @@
>> +From d69bfad3fa3daf5e72331f6870667607828d5891 Mon Sep 17 00:00:00 2001
>> +From: Stefan Eissing <stefan@eissing.org>
>> +Date: Mon, 8 Jun 2026 10:11:30 +0200
>> +Subject: [PATCH] ssl native_ca_store: always reinit
>> +
>> +Add bit `native_ca_store_opt` to keep the setting of
>> +CURLOPT_(PROXY_)SSL_OPTIONS and use that to calculate every easy
>> +transfer if a native CA store shall be used or not.
>> +
>> +This avoids `native_ca_store` getting stuck on TRUE after being set
>> +once.
>> +
>> +Closes #21902
>> +
>> +CVE: CVE-2026-11564
>> +Upstream-Status: Backport [https://github.com/curl/curl/commit/d69bfad3fa3daf5e72331f6870667607828d5891]
>> +
>> +Backport Changes:
>> +- Applied the option-completion helper to lib/vtls/vtls.c because curl
>> + 8.19.0 does not have lib/vtls/vtls_config.c.
The vtls_config.c file was added by the commit you link to below. It's
commit message is:
When a transfer goes against another origin than the initial one, do not
add the following to the ssl configuration: client cert, client key, srp
user/pass, pinned key.
That seems like a fix worth having. The diff is large, but hopefully
most of it is moving code around rather than fully re-writing it.
If that commit is backported as well, does the upstream patch for this
CVE apply cleanly?
>> +- Kept auto_client_cert derived from CURLSSLOPT_AUTO_CLIENT_CERT, as
>> + in the old set_ssl_options(). This preserves the existing Schannel
>> + opt-in behavior; automatic selection remains disabled by default.
>> +- Omitted the initial-origin restriction from the newer origin-aware
>> + configuration, introduced by [1].
>> + Curl 8.19.0 has no equivalent origin state. This backport resets
>> + native CA options per transfer without adding that restriction.
How has this been tested? Have you confirmed that this still resolves
the vulnerability?
Some judgement needs to be exercised when making changes to upstream
security patches as part of the backporting process. You need to explain
and justify the decisions, not just rely on the stable maintainer to
figure it out.
A simple change doesn't need much explanation, but one that may affect
the security properties does. "Simple" in this case doesn't mean small.
>> +
>> +[1] https://github.com/curl/curl/commit/d3391229b0a0eb04f1ffe85f1483a890c5e00009
>> +
>> +(cherry picked from commit d69bfad3fa3daf5e72331f6870667607828d5891)
>> +Signed-off-by: Darsh Kelaiya <dkelaiya@cisco.com>
>> +---
>> + lib/doh.c | 3 ++-
>> + lib/setopt.c | 20 ++------------------
>> + lib/vtls/vtls.c | 21 +++++++++++++++++++++
>> + 3 files changed, 25 insertions(+), 19 deletions(-)
>> +
>> +diff --git a/lib/vtls/vtls.c b/lib/vtls/vtls.c
>> +index 6dcc67d..35cb4f0 100644
>> +--- a/lib/vtls/vtls.c
>> ++++ b/lib/vtls/vtls.c
>> +@@ -306,6 +306,22 @@ static void free_primary_ssl_config(struct ssl_primary_config *sslc)
>> + #endif
>> + }
>> +
>> ++static void ssl_easy_config_compl_options(struct ssl_config_data *sslc)
>> ++{
>> ++ uint8_t options = sslc->primary.ssl_options;
>> ++ /* If set via CURLOPT_(PROXY_)SSL_OPTIONS, we definitely use it.
>> ++ * If not, we switch it on for supported backends if no custom
>> ++ * ca settings exist. */
>> ++ sslc->native_ca_store = !!(options & CURLSSLOPT_NATIVE_CA);
>> ++ sslc->enable_beast = !!(options & CURLSSLOPT_ALLOW_BEAST);
>> ++ sslc->no_partialchain = !!(options & CURLSSLOPT_NO_PARTIALCHAIN);
>> ++ sslc->no_revoke = !!(options & CURLSSLOPT_NO_REVOKE);
>> ++ sslc->revoke_best_effort = !!(options & CURLSSLOPT_REVOKE_BEST_EFFORT);
>> ++ sslc->earlydata = !!(options & CURLSSLOPT_EARLYDATA);
>> ++
>> ++ sslc->auto_client_cert = !!(options & CURLSSLOPT_AUTO_CLIENT_CERT);
>
> I'm sorry, I'm still not conviced that this change from upstream patch
> is safe:
> -+ sslc->auto_client_cert = Curl_peer_equal(origin, initial_origin) &&
> -+ !!(options & CURLSSLOPT_AUTO_CLIENT_CERT);
> ++ sslc->auto_client_cert = !!(options & CURLSSLOPT_AUTO_CLIENT_CERT);
>
> The fact that the function is not available is not reason enough to just
> omit the call. If the function is simple, maybe inline it?
>
>> ++}
>> ++
>> + CURLcode Curl_ssl_easy_config_complete(struct Curl_easy *data)
>> + {
>> + struct ssl_config_data *sslc = &data->set.ssl;
Best regards,
--
Paul Barker
next prev parent reply other threads:[~2026-11-17 9:50 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2026-10-23 9:33 [OE-core][wrynose][PATCH v2 0/5] curl: backport security fixes Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco)
2026-10-23 9:33 ` [OE-core][wrynose][PATCH v2 1/5] curl: fix CVE-2026-8926 Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco)
2026-11-16 14:36 ` Yoann Congal
2026-10-23 9:33 ` [OE-core][wrynose][PATCH v2 2/5] curl: fix CVE-2026-11564 Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco)
2026-11-16 15:51 ` Yoann Congal
2026-11-17 9:50 ` Paul Barker [this message]
2026-10-23 9:33 ` [OE-core][wrynose][PATCH v2 3/5] curl: fix CVE-2026-80255 Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco)
2026-11-17 9:54 ` Yoann Congal
2026-10-23 9:33 ` [OE-core][wrynose][PATCH v2 4/5] curl: fix CVE-2026-82208 Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco)
2026-10-23 9:33 ` [OE-core][wrynose][PATCH v2 5/5] curl: fix CVE-2026-80231 Darsh Kelaiya -X (dkelaiya - E INFOCHIPS PRIVATE LIMITED at Cisco)
2026-11-17 10:31 ` Yoann Congal
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=DMXE5AL1MFRL.SJ70FR1E8XH2@pbarker.dev \
--to=paul@pbarker.dev \
--cc=dkelaiya@cisco.com \
--cc=openembedded-core@lists.openembedded.org \
--cc=yoann.congal@smile.fr \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.