This commit fixes a bug in sub_cipherlists() when using sockets to test whether a server supports a set of ciphers with SSLv2. Code that is supposed to extract the list of SSLv2 ciphers to test for is incorrectly reading from the list of non-SSLv2 ciphers.
As noted in https://github.com/drwetter/testssl.sh/pull/1963#issuecomment-954935175, there are two places in which debugme1() is called in which the command to be run prints something to stdout, which is then redirected to a file. Now that debugme1() includes a redirect from stdout to stderr, these two lines do not work as expected. The output appears in the terminal and an empty file is created.
This commit fixes the issue by not using the debugme1() function in these two places.
This commit changes a few lines of code that checks OpenSSL version numbers to recognize 3.0.1 and 3.1.0 in addtion to 3.0.0.
According to https://www.openssl.org/blog/blog/2018/11/28/version, API/ABI compatibility is guaranteed for the same MAJOR version number (starting with version 3), so it should be sufficient to just check that $OSSL_VER_MAJOR is 3.
This commit adds PrioritzeChaCha detection, as requested in #1733. If the server (1) enforces a cipher order for a given protocol, (2) supports at least one non-ChaCha cipher for that protocol, and (3) supports at least one ChaCha cipher that would normally be less preferred than the ChaCha cipher, then a test is performed to see if the server selects the ChaCha cipher if it is listed first in the ClientHello.
While OpenSSL 1.0.2-chacha supports more than 128 cipher suites, other versions of OpenSSL tend to support far fewer than that.
Calling check_tls12_pref() adds to the run time of cipher_pref_check(), but is only needed if OpenSSL would otherwise send a ClientHello with more than 128 cipher suites. So, this commit adds another check and only calls check_tls12_pref() if the server has the size limit bug and OpenSSL supports more than 127 ciphers.
The code added by #2024 creates a cipher list with two consecutive colons. While this doesn't seem to be a problem, the commit removes the extra colon.
Some servers are configured to prioritize ChaCha ciphers if those ciphers are preferred by the client, even if the server is generally configured to use the server's cipher preferences rather than the client's. As a result of this, if a ChaCha cipher appears in the ClientHello before a non-ChaCha cipher, the server may select the ChaCha cipher even if the server is configured to prefer the non-ChaCha cipher.
In a few cases, e.g., cloudflare.com for TLS 1.2, this affects the ordering of the ciphers presented by cipher_pref_check(). This commit fixes the problem by having cipher_pref_check() (and check_tls12_pref()) always place any ChaCha ciphers at the end of the cipher list in the ClientHello. This ensures that cipher_pref_check() presents the ciphers in the server's preference order.
This commit fixes an error in s_client_options() when the function is called with an empty "-cipher" list and $OPENSSL supports "@SECLEVEL". This happens, for example, when ciphers_by_strength() is called for TLS 1.3.
The call to `openssl s_client` will fail is the cipher option is provided with an empty list or a list that just contains "@SECLEVEL=0". So, "@SECLEVEL=0" should only be added if the "$ciphers" list is non-empty. If "$ciphers" remains empty, then the "-cipher" option will not be added to the command line.
As noted in #2016 the detection of curves in find_openssl_binary()
copuld be improved.
This commit removes another connect call. Also it corrects the port statement
for LibreSSL-like pre-checks as port 0 generally seems to me better suited.
Inline documentation was added.
It worked so far with openssl 1.0.2, 1.1.1 and 3.0 and LibreSSL 3.4.
This needs to be backported to 3.0.
In order to be in line with 3.0 this adds a statement to avoid double port
statements for WSL users in a precheck condition
$OPENSSL s_client -groups $curve -connect 127.0.0.1:0:8443
This commit changes that to $OPENSSL s_client -groups $curve -connect 127.0.0.1:8443
which could cause a delay for WSL users and needs further investigation.
Not clear why port 8443 was chosen.
OpenSSL shows certificate serial numbers >35 with a LF (0A). Testssl.sh
just output that which makes JSON invalid and displays the LF in the terminal
too.
This commit fixes that (#2010) by adding filters so that the
serialnumber is not a multiline string.
Also this commit introduces a new function: a size check of the cert serial.
Below 8 bytes the CAB Forum's lower limit is hit which says the *entropy*
from a CSPRNG should be at least 64 bits. It is assumed that below 8 bytes
length this requirement isn't possible to meet (needs to be clarified with
Shannon, 8 bytes seems to low to me).
The high threshold is according to RFC 5280, Section-4.1.2.2 .
See also #2013.
The output has changed, so that on the terminal the serial has one line,
SHA1 and SHA256 each one line. The new json key is "cert_serialNumberLen".
In order to save time and to spare WSL users lame DNS lookups
I removed a couple of connect calls from 14 to 4:
$OPENSSL s_client -ssl2 -connect invalid.
-->
OPENSSL s_client -ssl2
NXCONNECT to localhost IP and port 0 is now the default when WSL
is detected. Not sure whether this is working under all circumstances,
so this needs some testing.
This commit provides a global variable to the RFC 6761 use of "invalid."
which WSL clients doen't seem to handle very well, see #1738, #1812.
With this commit it e.g. is possible to use
NXCONNECT=localhost:0 ./testssl.sh <TARGET>
to save some time.
This commit will be amended later.
Newer configuration files from openssl may include statements
which aren't compatible with our supplied old openssl version.
This commit adds an autodetection of such a file and uses a
openssl.cnf provided by this project then.
A longer while back the section ~ "Testing standard ciphers" was
renamed to "Testing cipher categories". However the internal help
didn't reflect that.
This fixes that, including an addtion to the documentation.
Note: the help still lists "-s --std, --standard" as a cmd line
switch.
* the ignore ~/.digrc option from dig is now parsed from the builtin help
* there was a potential DNS call which is now avoided
* for +noidnout check however there's a call to invalid. added
* the OPENSSL_CONF="" in check_resolver_bins() was moved a few lines
higher to avoid other errors in the terminal
Tested on (EOL) Ubuntu 14.04 which only has dig in an older version
See also #1950
This commit fixes#1961 in the 3.1dev branch by leaving NODEIP set to the server's IP address rather than changing it to the DNS name in the case of STARTTLS XMPP.
In order to address the problem of $OPENSSL s_client not working with STARTTLS XMPP if an IP address is provided to -connect, the -xmpphost option is used to provide the DNS name.
The fact that debugme1() redirects to stderr and the calls to this functions
redo that is deliberately as in the future we might want to use debugme1
without redirection.
... to address #1956 and other places. Similar to #1957,
only for the 3.1dev rolling release branch.
Also it changes debugme1() back? to output debug
statements only when $DEBUG >= 1. Per default here
also stderr is used.
get_server_certificate() includes a few calls to tls_sockets() in which the response will be TLS 1.3 and in which the response will be useless if it cannot be decrypted (since the goal is to obtain the server's certificate). So, these calls to tls_sockets() should specify "all+" rather than "all".
This commit changes run_server_defaults() so that the test for certificate compression is not run in --ssl-native mode. This fixes an issue that was caught by 21_baseline_starttls.t.
This commit adds a check for whether the server supports certificate compression (RFC 8879). If it does, then the list of supprted compression methods is output in the server's preference order.
If the order of the cmdline is '-U --ids-friendly' then we need to make sure we catch --ids-friendly. Normally we do not,
see #1717. The following statement makes sure. In the do-while + case-esac loop the check for --ids-friendly will be
executed again, but it does not hurt
Newer dig versions have an option to ignore $HOME/.digrc, older don't.
This commit adds a patch checking for the availability of such an option and
uses it by default. See #1894 .
If this option doesn't exist then still dig is used and can still lead to
wrong output. Unfortunately Debian-based distros are not very
good at this. Debian 10, Ubuntu 18.04 still use dig 9.11, whereas
Opensuse 15.2 has 9.16. Debian 11 and Ubuntu 20.04 use that too.
