Constructed types with a recursive definition (such as can be found in
PKCS7) could eventually exceed the stack given malicious input with
excessive recursion. Therefore we limit the stack depth.
Samuel Weiser [Fri, 9 Feb 2018 13:11:47 +0000 (14:11 +0100)]
consttime flag changed
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Kurt Roeckx <kurt@roeckx.be> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5170)
Samuel Weiser [Wed, 31 Jan 2018 12:10:55 +0000 (13:10 +0100)]
used ERR set/pop mark
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Kurt Roeckx <kurt@roeckx.be> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5170)
Samuel Weiser [Tue, 5 Dec 2017 14:55:17 +0000 (15:55 +0100)]
Replaced variable-time GCD with consttime inversion to avoid side-channel attacks on RSA key generation
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Kurt Roeckx <kurt@roeckx.be> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5170)
Bernd Edlinger [Wed, 21 Mar 2018 16:27:44 +0000 (17:27 +0100)]
Cleanup the s_time command.
Various code-cleanups.
Use SSL_CTX_set_mode(ctx, SSL_MODE_AUTO_RETRY) insead of handling
SSL_ERROR_WANT_READ everywhere.
Turn off the linger option on connected sockets to avoid failure.
Add BIO_set_conn_mode(conn, BIO_SOCK_NODELAY) to improve thruput.
Continue test even without -cipher option as in 1.0.2.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5698)
Matt Caswell [Tue, 16 Jan 2018 11:26:50 +0000 (11:26 +0000)]
Don't wait for dry at the end of a handshake
For DTLS/SCTP we were waiting for a dry event during the call to
tls_finish_handshake(). This function just tidies up various internal
things, and after it completes the handshake is over. I can find no good
reason for waiting for a dry event here, and nothing in RFC6083 suggests
to me that we should need to. More importantly though it seems to be
wrong. It is perfectly possible for a peer to send app data/alerts/new
handshake while we are still cleaning up our handshake. If this happens
then we will never get the dry event and so we cannot continue.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5085)
Matt Caswell [Tue, 16 Jan 2018 10:48:01 +0000 (10:48 +0000)]
Check for alerts while waiting for a dry event
At a couple of points in a DTLS/SCTP handshake we need to wait for a dry
event before continuing. However if an alert has been sent by the peer
then we will never receive that dry event and an infinite loop results.
This commit changes things so that we attempt to read a message if we
are waiting for a dry event but haven't got one yet. This should never
succeed, but any alerts will be processed.
Fixes #4763
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5085)
Benjamin Kaduk [Fri, 26 Jan 2018 17:16:21 +0000 (11:16 -0600)]
Do not cache sessions with zero sid_ctx_length when SSL_VERIFY_PEER
The sid_ctx is something of a "certificate request context" or a
"session ID context" -- something from the application that gives
extra indication of what sort of thing this session is/was for/from.
Without a sid_ctx, we only know that there is a session that we
issued, but it could have come from a number of things, especially
with an external (shared) session cache. Accordingly, when resuming,
we will hard-error the handshake when presented with a session with
zero-length sid_ctx and SSL_VERIFY_PEER is set -- we simply have no
information about the peer to verify, so the verification must fail.
In order to prevent these future handshake failures, proactively
decline to add the problematic sessions to the session cache.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5175)
Johannes Bauer [Tue, 20 Mar 2018 19:34:41 +0000 (20:34 +0100)]
Make pkeyutl a bit more user-friendly
Give meaningful error messages when the user incorrectly uses pkeyutl;
backport to OpenSSL_1_1_0-stable, cherrypicked from f6add6ac2c42df37d63b36dbef43e701875893d7.
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5699)
Richard Levitte [Mon, 19 Mar 2018 19:33:50 +0000 (20:33 +0100)]
s_client, s_server: do generic SSL configuration first, specialization after
We did the SSL_CONF_cmd() pass last of all things that could affect
the SSL ctx. However, the results of this, for example:
-max_protocol TLSv1.3 -tls1_2
... would mean that the protocol min got set to TLSv1.2 and the
protocol max to TLSv1.3, when they should clearly both be TLSv1.2.
However, if we see the SSL_CONF_cmd() switches as generic and those
internal to s_client and s_server as specialisations, we get something
that makes a little more sense.
Reviewed-by: Tim Hudson <tjh@openssl.org> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5679)
Richard Levitte [Mon, 19 Mar 2018 08:08:06 +0000 (09:08 +0100)]
Enhance ssltestlib's create_ssl_ctx_pair to take min and max proto version
Have all test programs using that function specify those versions.
Additionally, have the remaining test programs that use SSL_CTX_new
directly specify at least the maximum protocol version.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5662)
Richard Levitte [Wed, 14 Mar 2018 16:31:20 +0000 (17:31 +0100)]
Stop test/shlibloadtest.c from failing in a regression test
When doing a regression test, it's obvious that the version
test/shlibloadtest is built for will not be the same as the library
version. So we change the test to check for assumed compatibility.
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5620)
Matt Caswell [Fri, 23 Feb 2018 19:48:11 +0000 (19:48 +0000)]
Allow multiple entries without a Subject even if unique_subject == yes
It is quite likely for there to be multiple certificates with empty
subjects, which are still distinct because of subjectAltName. Therefore
we allow multiple certificates with an empty Subject even if
unique_subject is set to yes.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5627)
Matt Caswell [Fri, 23 Feb 2018 18:28:47 +0000 (18:28 +0000)]
Report a readable error on a duplicate cert in ca app
Commit 87e8feca (16 years ago!) introduced a bug where if we are
attempting to insert a cert with a duplicate subject name, and
duplicate subject names are not allowed (which is the default),
then we get an unhelpful error message back (error number 2). Prior
to that commit we got a helpful error message which displayed details
of the conflicting entry in the database.
That commit was itself attempting to fix a bug with the noemailDN option
where we were setting the subject field in the database too early
(before extensions had made any amendments to it).
This PR moves the check for a conflicting Subject name until after all
changes to the Subject have been made by extensions etc.
This also, co-incidentally fixes the ca crashing bug described in issue
5109.
Fixes #5109
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5627)
Andy Polyakov [Sat, 10 Mar 2018 15:19:33 +0000 (16:19 +0100)]
Configurations/unix-Makefile.tmpl: overhaul assembler make rules.
So far assembly modules were built as .pl->.S->.s followed by .s->.o.
This posed a problem in build_all_generated rule if it was executed
on another computer. So we change rule sequence to .pl->.S and then
.S->.s->.o.
Richard Levitte [Thu, 8 Mar 2018 11:01:28 +0000 (12:01 +0100)]
Configure: catch the build tree configdata.pm
There are things depending on configdata.pm. However, it's perfectly
possible that there is one in the source directory from a previous
build, and that might disrupt an out of source build. To avoid this
conflict, make sure never to use the source tree configdata.pm in that
case, i.e. make the hard assumption that it's a generated file in the
build tree, which it is.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5546)
Richard Levitte [Wed, 7 Mar 2018 14:58:04 +0000 (15:58 +0100)]
Configure: don't mangle the directory again when checking DEPEND inclusion
When generating the correct inclusion directory for DEPEND, we mangled
it to be relative to the build or the source directory. However, the
value we handle already come with a correct directory, so we only need
to use it as is.
Fixes #5543
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5546)
Andy Polyakov [Tue, 6 Mar 2018 22:33:06 +0000 (23:33 +0100)]
Configurations/10-main.conf: add -fno-common back to darwin-ppc-cc.
-fno-common was removed for all Darwin targets in 0c8734198d4282f6997965a03cd2e0ceaf207549 with rationale "it's either
'ranlib -c' or '-fno-common'." However, it's still absolutely required
in 32-bit darwin-ppc-cc. And when trying things out I didn't quite
see why it was formulated as one-or-another choice, as 'ranlib -c'
shouldn't [and doesn't] have problems with object modules without
commons. [Well, to be frank, I didn't manage to reproduce the problem
the modification was meaning to resolve either...]
d2i_X509.pod: clarify usage of the 'pp' function parameter
The 'pp' function parameters of d2i_TYPE() and i2d_TYPE() are referenced
in the DESCRIPTION section as 'in' resp. 'out'. This commit renames the
references to 'ppin' resp. 'ppout' and adds an explaining sentence.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5365)
Alex Gaynor [Sat, 3 Mar 2018 16:37:07 +0000 (11:37 -0500)]
Fixed a typo in a man page
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5498)
(cherry picked from commit c03dc6427cb8d15ddce735b5e5beef606bdea51d)
Ivan Filenko [Sun, 25 Feb 2018 13:49:27 +0000 (16:49 +0300)]
Fix typo in ASN1_STRING_length doc
CLA: trivial
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/5458)
David Benjamin [Wed, 21 Feb 2018 23:50:21 +0000 (18:50 -0500)]
Always use adr with __thumb2__.
Thumb2 addresses are a bit a mess, depending on whether a label is
interpreted as a function pointer value (for use with BX and BLX) or as
a program counter value (for use with PC-relative addressing). Clang's
integrated assembler mis-assembles this code. See
https://crbug.com/124610#c54 for details.
Instead, use the ADR pseudo-instruction which has clear semantics and
should be supported by every assembler that handles the OpenSSL Thumb2
code. (In other files, the ADR vs SUB conditionals are based on
__thumb2__ already. For some reason, this one is based on __APPLE__, I'm
guessing to deal with an older version of clang assembler.)
It's unclear to me which of clang or binutils is "correct" or if this is
even a well-defined notion beyond "whatever binutils does". But I will
note that https://github.com/openssl/openssl/pull/4669 suggests binutils
has also changed behavior around this before.
Reviewed-by: Andy Polyakov <appro@openssl.org> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5431)
bio_b64.c: prevent base64 filter BIO from decoding out-of-bound data
Fixes #5405, #1381
The base64 filter BIO reads its input in chunks of B64_BLOCK_SIZE bytes.
When processing input in PEM format it can happen in rare cases that
- the trailing PEM marker crosses the boundary of a chunk, and
- the beginning of the following chunk contains valid base64 encoded data.
This happened in issue #5405, where the PEM marker was split into
"-----END CER" and "TIFICATE-----" at the end of the first chunk.
The decoding of the first chunk terminated correctly at the '-' character,
which is treated as an EOF marker, and b64_read() returned. However,
when called the second time, b64_read() read the next chunk and interpreted
the string "TIFICATE" as valid base64 encoded data, adding 6 extra bytes
'4c 81 48 08 04 c4'.
This patch restores the assignment of the error code to 'ctx->cont', which
was deleted accidentally in commit 5562cfaca4f3 and which prevents b64_read()
from reading additional data on subsequent calls.
This issue was observed and reported by Annie Yousar.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5422)
Andy Polyakov [Tue, 20 Feb 2018 11:43:35 +0000 (12:43 +0100)]
test/recipes/80-test_pkcs12.t: handle lack of Win32::API.
So far check for availability of Win32::API served as implicit check
for $^O being MSWin32. Reportedly it's not safe assumption, and check
for MSWin32 has to be explicit.
Reviewed-by: Matt Caswell <matt@openssl.org> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5416)
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5367)
Pavel Kopyl [Fri, 3 Nov 2017 19:18:35 +0000 (22:18 +0300)]
do_body: fix heap-use-after-free.
The memory pointed to by the 'push' is freed by the
X509_NAME_ENTRY_free() in do_body(). The second time
it is referenced to (indirectly) in certify_cert:X509_REQ_free().
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4698)
X509v3_add_ext: free 'sk' if the memory pointed to by it
was malloc-ed inside this function.
X509V3_EXT_add_nconf_sk: return an error if X509v3_add_ext() fails.
This prevents use of a freed memory in do_body:sk_X509_EXTENSION_num().
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4698)
Viktor Dukhovni [Wed, 14 Feb 2018 03:43:15 +0000 (22:43 -0500)]
Avoid fragile aliasing of SHA224/384 update/final
This is purported to save a few cycles, but makes the code less
obvious and more brittle, and in fact breaks on platforms where for
ABI continuity reasons there is a SHA2 implementation in libc, and
so EVP needs to call those to avoid conflicts.
A sufficiently good optimizer could simply generate the same entry
points for:
foo(...) { ... }
and
bar(...) { return foo(...); }
but, even without that, the different is negligible, with the
"winner" varying from run to run (openssl speed -evp sha384):
Richard Levitte [Fri, 16 Jun 2017 01:46:41 +0000 (03:46 +0200)]
Add the target 'build_all_generated'
This new target is used to build all generated files and only that.
This can be used to prepare everything that requires things like perl
for a system that lacks perl and then move everything to that system
and do the rest of the build there.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3695)
Matt Caswell [Thu, 8 Feb 2018 14:48:51 +0000 (14:48 +0000)]
Don't calculate the Finished MAC twice
In <= TLSv1.2 a Finished message always comes immediately after a CCS
except in the case of NPN where there is an additional message between
the CCS and Finished. Historically we always calculated the Finished MAC
when we processed the CCS. However to deal with NPN we also calculated it
when we receive the Finished message. Really this should only have been
done if we hand negotiated NPN.
This simplifies the code to only calculate the MAC when we receive the
Finished. In 1.1.1 we need to do it this way anyway because there is no
CCS (except in middlebox compat mode) in TLSv1.3.
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5286)
Richard Levitte [Thu, 8 Feb 2018 11:31:05 +0000 (12:31 +0100)]
util/mkdef.pl: use better array in search of 'DEPRECATEDIN_'
%disabled_algorithms isn't necessarily initialised with the "algos"
'DEPRECATEDIN_1_1_0' etc. However, we know that @known_algorithms has
them all, so use that to find them instead.
Fixes #5157
(where this was reported)
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5282)
Richard Levitte [Wed, 24 Jan 2018 13:17:39 +0000 (14:17 +0100)]
util/mkdef.pl: Trust configdata.pm
This script kept its own database of disablable algorithms, which is a
maintenance problem, as it's not always perfectly in sync with what
Configure does. However, we do have all the data in configdata.pm,
produced by Configure, so let's use that instead.
Also, make sure to parse the *err.h header files, as they contain
function declarations that might not be present elsewhere.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5157)
Rich Salz [Tue, 6 Feb 2018 20:27:03 +0000 (15:27 -0500)]
Remove unused file
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5267)
(cherry picked from commit a3a5d1b73db46274e2cdedaad42fda5ce5cfb2da)
David Benjamin [Wed, 31 Jan 2018 19:47:41 +0000 (14:47 -0500)]
Fix timing leak in BN_from_montgomery_word.
BN_from_montgomery_word doesn't have a constant memory access pattern.
Replace the pointer trick with a constant-time select. There is, of
course, still the bn_correct_top leak pervasive in BIGNUM itself.
See also https://boringssl-review.googlesource.com/22904 from BoringSSL.
Reviewed-by: Andy Polyakov <appro@openssl.org> Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/5228)
David Benjamin [Tue, 23 Jan 2018 18:57:10 +0000 (13:57 -0500)]
Don't leak the exponent bit width in BN_mod_exp_mont_consttime.
The exponent here is one of d, dmp1, or dmq1 for RSA. This value and its
bit length are both secret. The only public upper bound is the bit width
of the corresponding modulus (RSA n, p, and q, respectively).
Although BN_num_bits is constant-time (sort of; see bn_correct_top notes
in preceding patch), this does not fix the root problem, which is that
the windows are based on the minimal bit width, not the upper bound. We
could use BN_num_bits(m), but BN_mod_exp_mont_consttime is public API
and may be called with larger exponents. Instead, use all top*BN_BITS2
bits in the BIGNUM. This is still sensitive to the long-standing
bn_correct_top leak, but we need to fix that regardless.
This may cause us to do a handful of extra multiplications for RSA keys
which are just above a whole number of words, but that is not a standard
RSA key size.
Reviewed-by: Paul Dale <paul.dale@oracle.com> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5154)
David Benjamin [Tue, 23 Jan 2018 18:46:53 +0000 (13:46 -0500)]
Make BN_num_bits_word constant-time.
(This patch was written by Andy Polyakov. I only wrote the commit
message. Mistakes in the analysis are my fault.)
BN_num_bits, by way of BN_num_bits_word, currently leaks the
most-significant word of its argument via branching and memory access
pattern.
BN_num_bits is called on RSA prime factors in various places. These have
public bit lengths, but all bits beyond the high bit are secret. This
fully resolves those cases.
There are a few places where BN_num_bits is called on an input where the
bit length is also secret. This does *not* fully resolve those cases as
we still only look at the top word. Today, that is guaranteed to be
non-zero, but only because of the long-standing bn_correct_top timing
leak. Once that is fixed, a constant-time BN_num_bits on such inputs
must count bits on each word.
Instead, those cases should not call BN_num_bits at all. In particular,
BN_mod_exp_mont_consttime uses the exponent bit width to pick windows,
but it should be using the maximum bit width. The next patch will fix
this.
Thanks to Dinghao Wu, Danfeng Zhang, Shuai Wang, Pei Wang, and Xiao Liu
for reporting this issue.
Reviewed-by: Paul Dale <paul.dale@oracle.com> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5154)
added macro to create version number
use the macro to build OPENSSL_VERSION_AT_LEAST(maj,min,fix) so that
customers of libssl (such as ruby-openssl) do not need to be so aware of
openssl version numbers.
includes updates to ssl(7) and OPENSSL_VERSION_NUMBER(3) man page
Reviewed-by: Richard Levitte <levitte@openssl.org> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5212)
(cherry picked from commit 3c5a61dd0f9d9a9eac098419bcaf47d1c296ca81)
Matt Caswell [Wed, 10 May 2017 20:46:14 +0000 (16:46 -0400)]
Add the SSL_OP_NO_RENEGOTIATION option to 1.1.0
This is based on a heavily modified version of commit db0f35dda by Todd
Short from the master branch.
We are adding this because it used to be possible to disable reneg using
the flag SSL3_FLAGS_NO_RENEGOTIATE_CIPHERS in 1.0.2. This is no longer
possible because of the opacity work.
A point to note about this is that if an application built against new
1.1.0 headers (that know about the new option SSL_OP_NO_RENEGOTIATION
option) is run using an older version of 1.1.0 (that doesn't know about
the option) then the option will be accepted but nothing will happen, i.e.
renegotiation will not be prevented. There's probably not much we can do
about that.
Fixes #4739
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4901)
Matt Caswell [Mon, 29 Jan 2018 14:55:44 +0000 (14:55 +0000)]
Make sure we check an incoming reneg ClientHello in DTLS
In TLS we have a check to make sure an incoming reneg ClientHello is
acceptable. The equivalent check is missing in the DTLS code. This means
that if a client does not signal the ability to handle secure reneg in the
initial handshake, then a subsequent reneg handshake should be rejected by
the server. In the DTLS case the reneg was being allowed if the the 2nd
ClientHello had a renegotiation_info extension. This is incorrect.
While incorrect, this does not represent a security issue because if
the renegotiation_info extension is present in the second ClientHello it
also has to be *correct*. Therefore this will only work if both the client
and server believe they are renegotiating, and both know the previous
Finished result. This is not the case in an insecure rengotiation attack.
I have also tidied up the check in the TLS code and given a better check
for determining whether we are renegotiating or not.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5191)
Benjamin Kaduk [Wed, 24 Jan 2018 19:54:46 +0000 (13:54 -0600)]
Fix strict-warnings build on FreeBSD
The cryptodev engine is only available for OpenBSD and FreeBSD,
but for the OS version-specific checks the OpenBSD macro is not
defined on FreeBSD. When building with -Werror and -Wundef (enabled
by strict-warnings), the FreeBSD build fails:
crypto/engine/eng_cryptodev.c:47:7: error: 'OpenBSD' is not defined,
evaluates to 0
[-Werror,-Wundef]
\# if (OpenBSD >= 200112) || ((__FreeBSD_version >= 470101 &&
\# __FreeBSD_versi...
^
The reverse case would be true on OpenBSD (__FreeBSD_version would
not be defined), but since the boolean will short-circuit and this
code is only executed on OpenBSD and FreeBSD, and the line is
already pretty long, leave that out for now.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5160)
projects / openssl / log