In NPN and ALPN, the protocol is renegotiated upon resumption. Test that
resumption picks up changes to the extension.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Move custom server and client options from the test dictionary to an
"extra" section of each server/client. Rename test expectations to say
"Expected".
This is a big but straightforward change. Primarily, this allows us to
specify multiple server and client contexts without redefining the
custom options for each of them. For example, instead of
"ServerNPNProtocols", "Server2NPNProtocols", "ResumeServerNPNProtocols",
we now have, "NPNProtocols".
This simplifies writing resumption and SNI tests. The first application
will be resumption tests for NPN and ALPN.
Regrouping the options also makes it clearer which options apply to the
server, which apply to the client, which configure the test, and which
are test expectations.
Reviewed-by: Richard Levitte <levitte@openssl.org>
These were guarded by $disabled{tests}. However, 'tests' is disabled
if we configure 'no-stdio', which means that we don't detect the lack
of OPENSSL_NO_STDIO guards in our public header files. So we move the
generation and build of test/buildtest_*.c to be unconditional.
Reviewed-by: Rich Salz <rsalz@openssl.org>
SSL_set_rbio() and SSL_set_wbio() are new functions in 1.1.0 and really
should be called SSL_set0_rbio() and SSL_set0_wbio(). The old
implementation was not consistent with what "set0" means though as there
were special cases around what happens if the rbio and wbio are the same.
We were only ever taking one reference on the BIO, and checking everywhere
whether the rbio and wbio are the same so as not to double free.
A better approach is to rename the functions to SSL_set0_rbio() and
SSL_set0_wbio(). If an existing BIO is present it is *always* freed
regardless of whether the rbio and wbio are the same or not. It is
therefore the callers responsibility to ensure that a reference is taken
for *each* usage, i.e. one for the rbio and one for the wbio.
The legacy function SSL_set_bio() takes both the rbio and wbio in one go
and sets them both. We can wrap up the old behaviour in the implementation
of that function, i.e. previously if the rbio and wbio are the same in the
call to this function then the caller only needed to ensure one reference
was passed. This behaviour is retained by internally upping the ref count.
This commit was inspired by BoringSSL commit f715c423224.
RT#4572
Reviewed-by: Rich Salz <rsalz@openssl.org>
This adds some simple SSL BIO tests that check for pushing and popping of
BIOs into the chain. These tests would have caught the bugs fixed in the
previous three commits, if combined with a crypto-mdebug build.
Reviewed-by: Rich Salz <rsalz@openssl.org>
The SSL_set_bio() function has some complicated ownership rules. This adds a
test to make sure it all works as expected.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Don't emit duplicate server/client sections when they are
identical. Instead, just point to the same section.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Add tests for resuming with a different client version.
This happens in reality when clients persist sessions on disk through
upgrades.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Systematically test every server-side version downgrade or upgrade.
Client version upgrade or downgrade could be tested analogously but will
be done in a later change.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Fix some indentation at the same time
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1292)
Add some more tests for sessions following on from the previous commit
to ensure the callbacks are called when appropriate.
Reviewed-by: Richard Levitte <levitte@openssl.org>
In TLS during ClientAuth if the CA is not recognised you should get an
UnknownCA alert. In SSLv3 this does not exist and you should get a
BadCertificate alert.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
The Client Auth tests were not correctly setting the Protocol, so that this
aspect had no effect. It was testing the same thing lots of times for
TLSv1.2 every time.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
In light of potential UKS (unknown key share) attacks on some
applications, primarily browsers, despite RFC761, name checks are
by default applied with DANE-EE(3) TLSA records. Applications for
which UKS is not a problem can optionally disable DANE-EE(3) name
checks via the new SSL_CTX_dane_set_flags() and friends.
Reviewed-by: Rich Salz <rsalz@openssl.org>
subject alternate names.
Add nameConstraints tests incluing DNS, IP and email tests both in
subject alt name extension and subject name.
Reviewed-by: Richard Levitte <levitte@openssl.org>
If application uses any of Windows-specific interfaces, make it
application developer's respondibility to include <windows.h>.
Rationale is that <windows.h> is quite "toxic" and is sensitive
to inclusion order (most notably in relation to <winsock2.h>).
It's only natural to give complete control to the application developer.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
Now that INCLUDE considers both the source and build trees, no need
for the rel2abs perl fragment hacks any more.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1264)
Reviewed-by: Andy Polyakov <appro@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1264)
Observe that the old tests were partly ill-defined:
setting sn_server1 but not sn_server2 in ssltest_old.c does not enable
the SNI callback.
Fix this, and also explicitly test both flavours of SNI mismatch (ignore
/ fatal alert). Tests still pass.
Reviewed-by: Rich Salz <rsalz@openssl.org>
The previous commit fixed a problem where fragmented alerts would cause an
infinite loop. This commit adds a test for these fragmented alerts.
Reviewed-by: Andy Polyakov <appro@openssl.org>
The previous fix contained a mistake, where any absolute path in
%directories would be cleared away instead of just being left alone.
Reviewed-by: Rich Salz <rsalz@openssl.org>
We recalculate the location of the directories we keep track of.
However, we did so after having moved to the new directory already, so
the data we did the calculations from were possibly not quite correct.
This change moves the calculations to happen before moving to the new
directory.
This issue is sporadic, and possibly dependent on the platform.
Reviewed-by: Matt Caswell <matt@openssl.org>
This extends 'req' to take more than one DN component, and to take
them as full DN components and not just CN values. All other commands
are changed to pass "CN = $cn" instead of just a CN value.
This adds 'genpc', which differs from the other 'gen*' commands by not
calling 'req', and expect the result from 'req' to come through stdin.
Finally, test/certs/setup.sh gets the commands needed to generate a
few proxy certificates.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Stephen Henson <steve@openssl.org>
Reorder arguments to follow convention.
Also allow r/s to be NULL in DSA_SIG_get0, similarly to ECDSA_SIG_get0.
This complements GH1193 which adds non-const setters.
Reviewed-by: Rich Salz <rsalz@openssl.org>
On some platforms we can't startup the TLSProxy due to environmental
problems (e.g. network set up on the build machine). These aren't OpenSSL
problems so we shouldn't treat them as test failures. Just visibly
indicate that we are skipping the test.
We only skip the first time we attempt to start up the proxy. If that works
then everything else should do...if not we should probably investigate and
so report as a failure.
This also removes test_networking...there is a danger that this turns into
a test of user's environmental set up rather than OpenSSL.
Reviewed-by: Richard Levitte <levitte@openssl.org>
VMS C doesn't allow symbols longer than 31 characters. We do the
automatic shortening with the library files, but not otherwise (to
make sure to work the VMS C magic).
For consistency, I shortened other similar symbols in the same manner.
Reviewed-by: Rich Salz <rsalz@openssl.org>
This commit adds some session API tests, and in particular tests the
modified behaviour of SSL_set_session() introduced in the last commit. To
do this I have factored out some common code from the asynciotest into a
new ssltestlib.c file. I've also renamed getsettest to sslapitest as this
more closely matches what it now is!
Reviewed-by: Rich Salz <rsalz@openssl.org>
The SNI tests introduced a redundant "server2" section into every test
configuration. Copy this automatically from "server" unless testing SNI,
to reduce noise in the generated confs.
Also remove duplicate SSL_TEST_CTX_create (merge conflict error).
Reviewed-by: Rich Salz <rsalz@openssl.org>
Some Linux platforms have a suitably recent kernel to support AFALG, but
apparently you still can't actually create an afalg socket. This extends
the afalg_chk_platform() function to additionally check whether we can
create an AFALG socket. We also amend the afalgtest to not report a
failure to load the engine as a test failure. A failure to load is almost
certainly due to platform environmental issues, and not an OpenSSL problem.
RT 4434
Reviewed-by: Andy Polyakov <appro@openssl.org>
- Only send SNI in SNI tests. This allows us to test handshakes without
the SNI extension as well.
- Move all handshake-specific machinery to handshake_helper.c
- Use enum types to represent the enum everywhere
(Resorting to plain ints can end in sign mismatch when the enum is
represented by an unsigned type.)
Reviewed-by: Rich Salz <rsalz@openssl.org>
We already test in EC_POINT_oct2point that points are on the curve. To
be on the safe side, move this check to
EC_POINT_set_affine_coordinates_* so as to also check point coordinates
received through some other method.
We do not check projective coordinates, though, as
- it's unlikely that applications would be receiving this primarily
internal representation from untrusted sources, and
- it's possible that the projective setters are used in a setting where
performance matters.
Reviewed-by: Rich Salz <rsalz@openssl.org>
When session tickets are used, it's possible that SNI might swtich the
SSL_CTX on an SSL. Normally, this is not a problem, because the
initial_ctx/session_ctx are used for all session ticket/id processes.
However, when the SNI callback occurs, it's possible that the callback
may update the options in the SSL from the SSL_CTX, and this could
cause SSL_OP_NO_TICKET to be set. If this occurs, then two bad things
can happen:
1. The session ticket TLSEXT may not be written when the ticket expected
flag is set. The state machine transistions to writing the ticket, and
the client responds with an error as its not expecting a ticket.
2. When creating the session ticket, if the ticket key cb returns 0
the crypto/hmac contexts are not initialized, and the code crashes when
trying to encrypt the session ticket.
To fix 1, if the ticket TLSEXT is not written out, clear the expected
ticket flag.
To fix 2, consider a return of 0 from the ticket key cb a recoverable
error, and write a 0 length ticket and continue. The client-side code
can explicitly handle this case.
Fix these two cases, and add unit test code to validate ticket behavior.
Reviewed-by: Emilia Käsper <emilia@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/1098)
Various fixes to get the following to compile:
./config no-asm -ansi -D_DEFAULT_SOURCE
RT4479
RT4480
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
The previous commit changed how we handle out-of-context empty records.
This commit adds some tests for the various scenarios. There are three
tests:
1: Check that if we inject an out-of-context empty record then we fail
2: Check that if we inject an in-context empty record then we succeed
3: Check that if we inject too many in-context empty records then we fail.
Reviewed-by: Andy Polyakov <appro@openssl.org>
The previous commit fixed a bug with BN_mod_word() which would have been
caught if we had a test for it. This commit adds one.
Reviewed-by: Andy Polyakov <appro@openssl.org>
If we're going to redirect STDERR and STDOUT, it's better to do it by
the book. This fix is a straight copy of examples in the perlfunc
manual.
Reviewed-by: Matt Caswell <matt@openssl.org>
The flags RSA_FLAG_NO_CONSTTIME, DSA_FLAG_NO_EXP_CONSTTIME and
DH_FLAG_NO_EXP_CONSTTIME which previously provided the ability to switch
off the constant time implementation for RSA, DSA and DH have been made
no-ops and deprecated.
Reviewed-by: Richard Levitte <levitte@openssl.org>
In rare cases, the shell we run test programs in may have complaints.
Shut those up unless testing verbosely.
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
Since one generates files that the other depends on, there's no
real reason to keep them separate. Since they were both different
aspects of 'openssl req', the merge ends up in 25-test_req.t.
This also makes cleanup easier.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Generate small test programs to check that external programs can be
built with our stuff at a very basic level.
For now, they check that each of our header files can be included
individually without compile failure.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Port client auth tests to the new framework, add coverage. The old tests
were only testing success, and only for some protocol versions; the new
tests add all protocol versions and various failure modes.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Add copyright to missing assembler files.
Add copyrights to missing test/* files.
Add copyrights
Various source and misc files.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Trying to use normal perl conditions to conditionally 'use' a perl
module didn't quite work. Using the 'if' module to do so does work.
Reviewed-by: Andy Polyakov <appro@openssl.org>
As it turns out default glob's behaviour for quoted argument varies
from version to version, making it impossible to Configure or run
tests in some cases. The reason for quoting globs was to accommodate
source path with spaces in its name, which was treated by default glob
as multiple paths. File::Glob::glob on the other hand doesn't consider
spaces as delimiters and therefore works with unquoted patterns.
[Unfortunaltely File::Glob::glob, being too csh-ly, doesn't work
on VMS, hence the "pinning" is conditional.]
Reviewed-by: Richard Levitte <levitte@openssl.org>
If using the msys console then msys attempts to "fix" command line
arguments to convert them from Unix style to Windows style. One of the
things it does is to look for arguments seperated by colons. This it
assumes is a list of file paths, so it replaces the colon with a semi-colon.
This was causing one of our tests to fail when calling the "req" command
line app. We were attempting to create a new DSA key and passing the
argument "dsa:../apps/dsa1024.pem". This is exactly what we intended but
Msys mangles it to "dsa;../apps/dsa1024.pem" and the command fails.
There doesn't seem to be a way to suppress Msys name mangling. Fortunately
we can work around this issue by generating the DSA key in a separate step
by calling "gendsa".
RT#4255
Reviewed-by: Richard Levitte <levitte@openssl.org>
