Steve Holme [Sat, 18 May 2019 16:30:16 +0000 (17:30 +0100)]
http_ntlm_wb: Handle auth for only a single request
Currently when the server responds with 401 on NTLM authenticated
connection (re-used) we consider it to have failed. However this is
legitimate and may happen when for example IIS is set configured to
'authPersistSingleRequest' or when the request goes thru a proxy (with
'via' header).
Implemented by imploying an additional state once a connection is
re-used to indicate that if we receive 401 we need to restart
authentication.
Daniel Stenberg [Tue, 14 May 2019 14:36:15 +0000 (16:36 +0200)]
cleanup: remove FIXME and TODO comments
They serve very little purpose and mostly just add noise. Most of them
have been around for a very long time. I read them all before removing
or rephrasing them.
Steve Holme [Wed, 8 May 2019 10:36:08 +0000 (11:36 +0100)]
url: Remove duplicate clean up of the winbind variables in conn_shutdown()
Given that Curl_disconnect() calls Curl_http_auth_cleanup_ntlm() prior
to calling conn_shutdown() and it in turn performs this, there is no
need to perform the same action in conn_shutdown().
In Curl_multissl_version() it was possible to overflow the passed in
buffer if the generated version string exceeded the size of the buffer.
Fix by inverting the logic, and also make sure to not exceed the local
buffer during the string generation.
Closes #3863 Reported-by: nevv on HackerOne/curl Reviewed-by: Jay Satiro Reviewed-by: Daniel Stenberg
- No longer show verbose output for the conncache closure handle.
The offending commit was added so that the conncache closure handle
would inherit verbose mode from the user's easy handle. (Note there is
no way for the user to set options for the closure handle which is why
that was necessary.) Other debug settings such as the debug function
were not also inherited since we determined that could lead to crashes
if the user's per-handle private data was used on an unexpected handle.
The reporter here says he has a debug function to capture the verbose
output, and does not expect or want any output to stderr; however
because the conncache closure handle does not inherit the debug function
the verbose output for that handle does go to stderr.
There are other plausible scenarios as well such as the user redirects
stderr on their handle, which is also not inherited since it could lead
to crashes when used on an unexpected handle.
Short of allowing the user to set options for the conncache closure
handle I don't think there's much we can safely do except no longer
inherit the verbose setting.
Jay Satiro [Wed, 8 May 2019 07:39:53 +0000 (03:39 -0400)]
http: Ignore HTTP/2 prior knowledge setting for HTTP proxies
- Do not switch to HTTP/2 for an HTTP proxy that is not tunnelling to
the destination host.
We already do something similar for HTTPS proxies by not sending h2. [1]
Prior to this change setting CURL_HTTP_VERSION_2_PRIOR_KNOWLEDGE would
incorrectly use HTTP/2 to talk to the proxy, which is not something we
support (yet?). Also it's debatable whether or not that setting should
apply to HTTP/2 proxies.
OpenSSL: Report -fips in version if OpenSSL is built with FIPS
Older versions of OpenSSL report FIPS availabilty via an OPENSSL_FIPS
define. It uses this define to determine whether to publish -fips at
the end of the version displayed. Applications that utilize the version
reported by OpenSSL will see a mismatch if they compare it to what curl
reports, as curl is not modifying the version in the same way. This
change simply adds a check to see if OPENSSL_FIPS is defined, and will
alter the reported version to match what OpenSSL itself provides. This
only appears to be applicable in versions of OpenSSL <1.1.1
Frank Gevaerts [Fri, 26 Apr 2019 09:23:15 +0000 (11:23 +0200)]
nss: allow fifos and character devices for certificates.
Currently you can do things like --cert <(cat ./cert.crt) with (at least) the
openssl backend, but that doesn't work for nss because is_file rejects fifos.
I don't actually know if this is sufficient, nss might do things internally
(like seeking back) that make this not work, so actual testing is needed.
Daniel Stenberg [Tue, 30 Apr 2019 09:14:38 +0000 (11:14 +0200)]
multi: provide Curl_multiuse_state to update information
As soon as a TLS backend gets ALPN conformation about the specific HTTP
version it can now set the multiplex situation for the "bundle" and
trigger moving potentially queued up transfers to the CONNECT state.
Daniel Stenberg [Tue, 30 Apr 2019 09:12:12 +0000 (11:12 +0200)]
process_pending_handles: mark queued transfers as previously pending
With transfers being queued up, we only move one at a a time back to the
CONNECT state but now we mark moved transfers so that when a moved
transfer is confirmed "successful" (it connected) it will trigger the
move of another pending transfer. Previously, it would otherwise wait
until the transfer was done before doing this. This makes queued up
pending transfers get processed (much) faster.
In case the name pointer isn't set (due to memory pressure most likely)
we need to skip the prefix matching and reject with a badcookie to avoid
a possible NULL pointer dereference.
Closes #3820 #3821 Reported-by: Jonathan Moerman Reviewed-by: Daniel Stenberg <daniel@haxx.se>
Daniel Stenberg [Mon, 29 Apr 2019 06:00:49 +0000 (08:00 +0200)]
CURL_MAX_INPUT_LENGTH: largest acceptable string input size
This limits all accepted input strings passed to libcurl to be less than
CURL_MAX_INPUT_LENGTH (8000000) bytes, for these API calls:
curl_easy_setopt() and curl_url_set().
The 8000000 number is arbitrary picked and is meant to detect mistakes
or abuse, not to limit actual practical use cases. By limiting the
acceptable string lengths we also reduce the risk of integer overflows
all over.
NOTE: This does not apply to `CURLOPT_POSTFIELDS`.
Steve Holme [Sat, 13 Apr 2019 20:47:56 +0000 (21:47 +0100)]
ntlm: Support the NT response in the type-3 when OpenSSL doesn't include MD4
Just like we do for mbed TLS, use our local implementation of MD4 when
OpenSSL doesn't support it. This allows a type-3 message to include the
NT response.
Kerberos was incorrectly indented as a subsection under FTP, which is
incorrect as they are both top level sections. A fix for this was first
attempted in commit fef38a0898322f285401c5ff2f5e7c90dbf3be63 but that
was a few paddles short of being complete.
Daniel Stenberg [Sat, 20 Apr 2019 10:19:47 +0000 (12:19 +0200)]
docs/BUG-BOUNTY: bug bounty time [skip ci]
Introducing the curl bug bounty program on hackerone. We now recommend
filing security issues directly in the hackerone ticket system which
only is readable to curl security team members.
Steve Holme [Thu, 18 Apr 2019 20:54:35 +0000 (21:54 +0100)]
sasl: Don't send authcid as authzid for the PLAIN mechanism as per RFC 4616
RFC 4616 specifies the authzid is optional in the client authentication
message and that the server will derive the authorisation identity
(authzid) from the authentication identity (authcid) when not specified
by the client.
projects / curl / log