Check that the default signature type is allowed

TLS < 1.2 has fixed signature algorithms: MD5+SHA1 for RSA and SHA1 for the
others. TLS 1.2 sends a list of supported ciphers, but allows not sending
it in which case SHA1 is used. TLS 1.3 makes sending the list mandatory.

When we didn't receive a list from the client, we always used the
defaults without checking that they are allowed by the configuration.

Reviewed-by: Paul Dale <paul.dale@oracle.com>
GH: #10784
(cherry picked from commit b0031e5dc2c8c99a6c04bc7625aa00d3d20a59a5)

Replace apps/server.pem with certificate with a sha256 signature.

It replaces apps/server.pem that used a sha1 signature with a copy of
test/certs/servercert.pem that is uses sha256.

This caused the dtlstest to start failing. It's testing connection
sbetween a dtls client and server. In particular it was checking that if
we drop a record that the handshake recovers and still completes
successfully. The test iterates a number of times. The first time
through it drops the first record. The second time it drops the second
one, and so on. In order to do this it has a hard-coded value for the
expected number of records it should see in a handshake. That's ok
because we completely control both sides of the handshake and know what
records we expect to see. Small changes in message size would be
tolerated because that is unlikely to have an impact on the number of
records. Larger changes in message size however could increase or
decrease the number of records and hence cause the test to fail.

This particular test uses a mem bio which doesn't have all the CTRLs
that the dgram BIO has. When we are using a dgram BIO we query that BIO
to determine the MTU size. The smaller the MTU the more fragmented
handshakes become. Since the mem BIO doesn't report an MTU we use a
rather small default value and get quite a lot of records in our
handshake. This has the tendency to increase the likelihood of the
number of records changing in the test if the message size changes.

It so happens that the new server certificate is smaller than the old
one. AFAICT this is probably because the DNs for the Subject and Issuer
are significantly shorter than previously. The result is that the number
of records used to transmit the Certificate message is one less than it
was before. This actually has a knock on impact for subsequent messages
and how we fragment them resulting in one less ServerKeyExchange record
too (the actual size of the ServerKeyExchange message hasn't changed,
but where in that message it gets fragmented has). In total the number
of records used in the handshake has decreased by 2 with the new
server.pem file.

Reviewed-by: Paul Dale <paul.dale@oracle.com>
GH: #10784
(cherry picked from commit 5fd72d96a592c3c4ef28ff11c6ef334a856b0cd1)

openssl-config: add example libssl system-defaults

Provide a "simple" example for affecting the systemwide default behavior
of libssl.  The large number of mandatory nested sections makes this
less simple than the main description might suggest.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10937)

(cherry picked from commit 3472082b4b6d73e0803a7c47f03e96ec0a69f77b)

Remove remaining references to crypto/include

Configure creates an empty crypto/include which
gets not cleaned up with make distclean.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10893)

Fix TLS not using aes_cbc_hmac_sha ciphers

AES_ASM was repaced by AESNI_ASM, so use that.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10892)

Fix unwind info in crypto/rc4/asm/rc4-x86_64.pl

Move .cfi_startproc to the right place for RC4.  Add missing
.cfi_startproc and .cfi_endproc to RC4_options.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/10872)

(cherry picked from commit 967ef73013becef2aec3439f8c45204b24121018)

For all assembler scripts where it matters, recognise clang > 9.x

Fixes #10853

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10857)

Update SSL_CTX_sess_set_new_cb(3) docs for refcounts

The existing documentation for the new-session callback was unclear
about the requirements on the callback with respect to reference-handling
of the session object being created.  Be more explicit about the
(non-)requirements on the callback code for "success" (1) and "ignore"
(0) return values.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10848)

(cherry picked from commit 188d4ec82a9b0085ac5841cce3eda95efb94f2b4)

Fix documentation of return value for EVP_Digest{Sign,Verify}Init()

They never returned the negative values that the documentation stated.

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10847)

Avoid leak in error path of asn1_parse2

CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10794)

(cherry picked from commit 6a165fab239ec5b00b3cd68169a63b509207177d)

Configure: use $list_separator_re only for defines and includes

This regexp was used a bit too uncontrolled, which had it split flag
values where it should not have.

Fixes #10792

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10793)

(cherry picked from commit bbe486cf6154df3d3aaedbae6c5b82d4ed31a5f8)

fix a glitch in the documentation of OCSP_sendreq_bio()

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10713)

Fix pkeyutl -verifyrecover

When performing a pkeyutl -verifyrecover operation the input file is not
a hash - it is the signature itself. Therefore don't do the check to make
sure it looks like a hash.

Fixes #9658

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9731)

(cherry picked from commit 5ffc33244cd4d66e47dfa66ce89cb38d0f3074cc)

Revert "Move random-related defines from e_os.h to rand_unix.c"

This reverts commit 7b18d1a53f932391bbc599a4717d6f98a597849c, which moved the
DEVRANDOM and DEVRANDOM_EGD defines into rand_unix.c. That change introduced
the regression that the compiler complains about missing declarations in
apps/version.c when OpenSSL is configured using `--with-rand-seed=devrandom`
(resp. `--with-rand-seed=egd`):

apps/version.c:173:42: error: 'DEVRANDOM' undeclared
             static const char *dev[] = { DEVRANDOM, NULL };
                                          ^~~~~~~~~

Fixes #10759

Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10764)

Don't store an HMAC key for longer than we need

The HMAC_CTX structure stores the original key in case the ctx is reused
without changing the key.

However, HMAC_Init_ex() checks its parameters such that the only code path
where the stored key is ever used is in the case where HMAC_Init_ex is
called with a NULL key and an explicit md is provided which is the same as
the md that was provided previously. But in that case we can actually reuse
the pre-digested key that we calculated last time, so we can refactor the
code not to use the stored key at all.

With that refactor done it is no longer necessary to store the key in the
ctx at all. This means that long running ctx's will not keep the key in
memory for any longer than required. Note though that the digested key
*is* still kept in memory for the duration of the life of the ctx.

Fixes #10743

Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10763)

Fix incorrect return code on ECDSA key verification

ECDSA_do_verify() is a function that verifies a ECDSA signature given a hash and a public EC key. The function is supposed to return 1 on valid signature, 0 on invalid signature and -1 on error. Previously, we returned 0 if the key did not have a verify_sig method. This is actually an error case and not an invalid signature. Consequently, this patch updates the return code to -1.

Fixes #8766

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10693)

(cherry picked from commit 26583f6aa8dc28e3598e61db66e54e2fdf8b195f)

Fix disabled ecdsa in apps/speed

This came from f3fdfbf78c6b. run = 1 should be done in pkey_print_message
as well, otherwise other tests printed with pkey_print_message won't run.

Change-Id: I0ba0b05256ad6509ada4735b26d10f8a73fd89ec

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/10710)

(cherry picked from commit 6e49b514067a2b6a30d064d2ae1fdfd8050c184b)

Fix side channel in ecp_nistz256-armv8.pl

This change addresses a potential side-channel vulnerability in
the internals of nistz256 low level operations for armv8.

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit f5a659b6dfcc735a62c712dcca64d116d2289b97)

Fix side channel in the ecp_nistz256.c reference implementation

This is only used if configured with
./config -DECP_NISTZ256_REFERENCE_IMPLEMENTATION

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit 7d4716648e8348dea862e198b9395478fae01907)

Improve side channel fix in ecp_nistz256-x86_64.pl

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit e9fe87950db2e6169029b2ecf3ed09d64265bc9c)

Fix side channel in ecp_nistz256-armv4.pl

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit 3d139746ca72f0906c036d0a4a3e176c7b61ed1b)

Fix side channel in ecp_nistz256-x86.pl

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit 0de3399b691f025153c8001045d5eeb0909dfd7a)

Avoid leaking intermediate states in point doubling special case.

Cherry picked from
https://github.com/google/boringssl/commit/12d9ed670da3edd64ce8175cfe0e091982989c18

Reviewed-by: Nicola Tuveri <nic.tuv@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit 2baea7c7e0896658b74956cac6084dd7e82e8c1b)

Fix potential SCA vulnerability in some EC_METHODs

This commit addresses a potential side-channel vulnerability in the
internals of some elliptic curve low level operations.
The side-channel leakage appears to be tiny, so the severity of this
issue is rather low.

The issue was reported by David Schrammel and Samuel Weiser.

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/9239)

(cherry picked from commit 3cb914c463ed1c9e32cfb773d816139a61b6ad5f)

Run make update

The New Year has caused various files to appear out of date to "make
update". This causes Travis to fail. Therefore we update those files.

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10739)

Update tls13_enc.c

Fix double + in hkdflabel declaration (FIXES #10675)
CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10700)

(cherry picked from commit 2de5a5fbdd14f514e962cccfe90482c37786c183)

Temporarily disable external pyca tests

The pyca-cryptography external test has been failing for a long time.
It looks like upstream needs to make some changes to adapt to 1.1.1.

Backported from #10689

[extended tests]

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10722)

Update the krb5 submodule

Bring us up to date with upstream's 1.17.1 release.  Among other
things, it includes commit c2497d46b4bad473e164943d67b58cd1ae261c3a
which fixes several issues that affect running the test suite under
Travis CI.  Hopefully those will work transitively for us as well.

[extended tests]

Reviewed-by: Matt Caswell <matt@openssl.org>
(cherry picked from commit 3e73f558af600ea068bb2132988c31ddb444e13e)

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10690)

Fix a race condition in the speed command

The timer alarm sets run = 0, while the benchmark
does run = 1 in the initialization code.  That is
a race condition, if the timer goes off too early
the benchmark runs forever.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10680)

(cherry picked from commit f3fdfbf78c6bfc97abf9c70b03859a28ebf6b66d)

Add some missing cfi frame info in rc4-md5-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10679)

(cherry picked from commit b2a00f62209add348deb8283c588ddbd572dc216)

Add some missing cfi frame info in poly1305-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10678)

(cherry picked from commit 048fa13e5ef4ccd730561f79a6c91f38365994d1)

Add some missing cfi frame info in aesni-gcm-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10677)

(cherry picked from commit 275a048ffc1585a731e39d7e8e3f53766e8f48d7)

Add some missing cfi frame info in x25519-x86_64.pl

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10676)

(cherry picked from commit 9d079f2744b9b624c6fe75f95fc0f766ef88ffcf)

Fix aesni_cbc_sha256_enc_avx2 backtrace info

We store a secondary frame pointer info for the debugger
in the red zone.  This fixes a crash in the unwinder when
this function is interrupted.

Additionally the missing cfi function annotation is added
to aesni_cbc_sha256_enc_shaext.

[extended tests]

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10674)

(cherry picked from commit 665de4d48aef2507022a7d74f5c7f6e339d5e6bc)

Add some missing cfi frame info in ecp_nistz256-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10672)

(cherry picked from commit eff5076a78502d1ac04669e44127d4bd7c0a9ce7)

Fix evp_extra_test with no-dh

The new DH test in evp_extra_test.c broke the no-dh build so we add some
guards to fix it.

Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/10644)

(cherry picked from commit 501fcfb8cfc1aa114ffde437039c2dc2827554ae)

Add some missing cfi frame info in aesni-sha and sha-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10655)

(cherry picked from commit b0d3442efc10b635863b915c2d014345f6e5a219)

Add some missing cfi frame info in keccak1600-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10654)

(cherry picked from commit 95bbe6eff7aadc681e282ec957379b49d6f80ca8)

Add some missing cfi frame info in aesni-x86_64.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10653)

(cherry picked from commit a5fe7825b970a6c937118a4f707f9ad367413794)

Add some missing cfi frame info in rsaz-x86_64

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10652)

(cherry picked from commit 013c2e8d1a272df444f47b8b54de1d51bc499887)

Add some missing cfi frame info in x86_64-mont5.pl

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10651)

(cherry picked from commit 0190c52ab8b4cdf5fe577b3d924576167c892a15)

Add some missing cfi frame info in camellia-x86_64.pl

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10642)

(cherry picked from commit 6b913be708f98b1d971586d38e608218ee6de6fa)

Fix regression on x509 keyform argument

In OpenSSL pre 1.1.0, 'openssl x509 -keyform engine' was possible
and supported.  In 1.1.0, type of keyform argument is OPT_FMT_PEMDER
which doesn't support engine. This changes type of keyform argument
to OPT_FMT_PDE which means PEM, DER or engine and updates the manpage
including keyform and CAkeyform.

This restores the pre 1.1.0 behavior.

This issue is very similar than https://github.com/openssl/openssl/issues/4366

CLA: trivial

Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10609)

(cherry picked from commit 0ab6fc79a9a63370be1a615729dc2a6ed0d6c89b)

Configuration: compute openssl_other_defines and related info later

The computation of macros and configdata.pm related data from %disabled
was done much too early, leaving later disablings without real support.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10565)

Disable devcryptoeng on newer OpenBSD versions

It's reported that /dev/crypto support has been dropped in OpenBSD 5.7.

Fixes #10552

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10565)

Fix unwind info for some trivial functions

While stack unwinding works with gdb here, the
function _Unwind_Backtrace gives up when something outside
.cfi_startproc/.cfi_endproc is found in the call stack, like
OPENSSL_cleanse, OPENSSL_atomic_add, OPENSSL_rdtsc, CRYPTO_memcmp
and other trivial functions which don't save anything in the stack.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/10635)

(cherry picked from commit 8913378a552e470c66277c47b19699f20b84aa3b)

apps/speed.c: Fix eddsa sign and verify output with -multi option

Fixes #10261
CLA: trivial

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10285)

(cherry picked from commit 1ac7e15375be39c8f03171c02658cf703f58217a)

Update the HISTORY entry for RSA_get0_pss_params()

Make a note of when this function was first introduced

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10568)

Backport the RSA_get0_pss_params() function from master

This is a missing accessor in order to obtain PSS parameters from an
RSA key, which should also be available in 1.1.1.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10568)

Test that EVP_PKEY_set1_DH() correctly identifies the DH type

Provide a test to check tat when we assign a DH object we know whether
we are dealing with PKCS#3 or X9.42 DH keys.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10593)

(cherry picked from commit e295de1d8433ed07092845cb6c56aa424ff35c6d)

Ensure EVP_PKEY_set1_DH detects X9.42 keys

OpenSSL supports both PKCS#3 and X9.42 DH keys. By default we use PKCS#3
keys. The function `EVP_PKEY_set1_DH` was assuming that the supplied DH
key was a PKCS#3 key. It should detect what type of key it is and assign
the correct type as appropriate.

Fixes #10592

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10593)

(cherry picked from commit 32c869ffaba67822602ea9fec611272ff8e8db58)

Return 1 when openssl req -addext kv is duplicated

CLA: trivial

Fixes #10273

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10578)

(cherry picked from commit 1aeec3dbc2d62f902698b1eba9ed31cbd436f9dc)

rand_lib.c: fix null pointer dereferences after RAND_get_rand_method() failure

RAND_get_rand_method() can return a NULL method pointer in the case of a
malloc failure, so don't dereference it without a check.

Reported-by: Zu-Ming Jiang (detected by FIFUZZ)
Fixes #10480

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10490)

util/mkerr.pl: revert accidental change of header guards

This change was backported accidentally from master in commit fbbfd128c9aa.

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10623)

Parse large GOST ClientKeyExchange messages

Large GOST ClientKeyExchange messages are sent by VipNet CSP, one of
Russian certified products implementing GOST TLS, when a server
certificate contains 512-bit keys.

This behaviour was present in 1.0.2 branch and needs to be restored.
Backport of #10376

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10376)

Check return value after loading config file

CLA: trivial

Reviewed-by: Paul Yang <kaishen.yy@antfin.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10607)

(cherry picked from commit dd0139f416257ec5632414ed3ad8c61d07ba07ec)

Enrich arm64 tests in Travis matrix

1, Remove simple test just with --strict-warnings enabled.
2, Share the three common envs with amd64.
3, Add matrix item running test in bionic(default xenial) for arm64.
4, Enable MSan test on arm64 for extended test.
5, Enable UBSan test on arm64 for extended test.

Change-Id: Ic1f2c5e39ee6fbafed6ede74a925301121463520

Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10519)

(cherry picked from commit 77fedcdc8cf5446354a1e7ff9123d8ada5b79b06)

Fix docs for CRYPTO_secure_allocated

Fixes #9300

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10523)

(cherry picked from commit 20c09f00346aec3a20ceacc9c9d44c5f050e13dd)

Fix some typos

Reported-by: misspell-fixer <https://github.com/vlajos/misspell-fixer>
CLA: trivial

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10544)

(cherry picked from commit 79c44b4e3044aee9dc9618850d4f1ce067757b4b)

Difference between EVP_CipherInit and EVP_CipherInit_ex

Fixes #10455

Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10550)

Add a CHANGES entry for CVE-2019-1551

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10575)

Add a test case for rsaz_512_sqr overflow handling

[extended tests]

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10575)

Improve the overflow handling in rsaz_512_sqr

We have always a carry in %rcx or %rbx in range 0..2
from the previous stage, that is added to the result
of the 64-bit square, but the low nibble of any square
can only be 0, 1, 4, 9.

Therefore one "adcq $0, %rdx" can be removed.
Likewise in the ADX code we can remove one
"adcx %rbp, $out" since %rbp is always 0, and carry is
also zero, therefore that is a no-op.

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10575)

Fix an overflow bug in rsaz_512_sqr

There is an overflow bug in the x64_64 Montgomery squaring procedure used in
exponentiation with 512-bit moduli. No EC algorithms are affected. Analysis
suggests that attacks against 2-prime RSA1024, 3-prime RSA1536, and DSA1024 as a
result of this defect would be very difficult to perform and are not believed
likely. Attacks against DH512 are considered just feasible. However, for an
attack the target would have to re-use the DH512 private key, which is not
recommended anyway. Also applications directly using the low level API
BN_mod_exp may be affected if they use BN_FLG_CONSTTIME.

CVE-2019-1551

Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/10575)

Fix exit issue in travisci

Ungraceful 'exit' probably causes unexpeced error on background activity.
So replace 'exit' with recommended 'travis_terminate'. Also see
https://travis-ci.community/t/exit-0-cannot-exit-successfully-on-arm/5731/4

Change-Id: I382bd93a3e15ecdf305bab23fc4adefbf0348ffb

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10561)

(cherry picked from commit 6df44cf65fbc7e150965149d7e681ac3e22d11d8)

Remove spurious space from file.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10562)

(cherry picked from commit 7ba46774b7f1b2e83d44323a8831b615a49f9f37)

Update NOTES.ANDROID for newer NDK versions + small fixes.

Fixes #8941

Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10478)

(cherry picked from commit d3a27c5ee45a29edd9c0d60ad5929f67996f89fd)

Add a test for NULL chunks in encrypt/decrypt

Issue #8675 describes a problem where calling EVP_DecryptUpdate() with an
empty chunk causes the result to be different compared to if you do not
use an empty chunk. This adds a test for that case.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9057)

EVP_*Update: ensure that input NULL with length 0 isn't passed

Even with custom ciphers, the combination in == NULL && inl == 0
should not be passed down to the backend cipher function.  The reason
is that these are the values passed by EVP_*Final, and some of the
backend cipher functions do check for these to see if a "final" call
is made.

An exception is made for CCM mode which has special handling for the case
where inl == 0: this may mean the total plaintext or ciphertext length is 0.

This is based on an original commit by Richard Levitte.

Fixes #8675

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9057)

apps/ocsp.c: sock_timeout -> socket_timeout

It appears that 'sock_timeout' is defined at least with DJGPP, so we
rename our symbol and hope the new name isn't taken.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10515)

(cherry picked from commit e9b95e42fbae668cb605287fa462a0d5f58b9caf)

Fix L<EVP_MD_CTX_set_pkey_ctx> links

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10420)

i2b_PVK(): Use Encrypt, not Decrypt

We used EVP_EncryptInit_ex() to initialise, but EVP_DecryptUpdate()
and EVP_DecryptFinal_ex() to actually perform encryption.  This worked
long ago, when the Encrypt and Decrypt variants were the same, but
doesn't now (actually haven't for a very long time).

This shows how seldom PVK is actually used.

Fixes #9338

Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10521)

doc/man7/proxy-certificates.pod: New guide for proxy certificates

This replaces doc/HOWTO/proxy_certificates.txt

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10507)

(cherry picked from commit 2a03823606b2d7e73e5dc890c3202e186511151f)

UI_UTIL_wrap_read_pem_callback(): when |cb| is NULL, use PEM_def_callback

Fixes #10444

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10447)

(cherry picked from commit 72a5412b4858cc7c5627a121f78685a2a4065521)

main: avoid a NULL dereference on initialisation.

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10495)

(cherry picked from commit acc7b9fb5c162c2ca522e5e1e09d1efbde8dc6a0)

Fix a race condition in SNI handling

As was done for ciphers, supported groups, and EC point formats in
https://github.com/openssl/openssl/pull/9162, only write the negotiated
SNI hostname value to the session object when not resuming, even for
TLS 1.3 resumptions.  Otherwise, when using a stateful session cache
(as is done by default when 0-RTT data is enabled), we can have multiple
SSLs active using the same in-memory session object, which leads to
double-frees and similar race conditions in the SNI handler prior
to this commit.

Fortunately, since draft-ietf-tls-tls13-22, there is no requirement
that the SNI hostname be preserved across TLS 1.3 resumption, and thus
not a need to continually update the session object with the "current"
value (to be used when producing session tickets, so that the subsequent
resumption can be checked against the current value).  So we can just
relax the logic and only write to the session object for initial handshakes.
This still leaves us in a somewhat inconsistent state, since if the SNI value
does change across handshakes, the session object will continue to record
the initial handshake's value, even if that bears no relation to the
current handshake.  The current SSL_get_servername() implementation
prefers the value from the session if s->hit, but a more complete fix
for that and related issues is underway in
https://github.com/openssl/openssl/pull/10018; there is no need to wait
for the complete fix for SNI name handling in order to close the
race condition and avoid runtime crashes.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10441)

(cherry picked from commit 2a5385511051d33be8d2b20d7669d8b1862fe510)

EVP p_lib: Add NULL check to EVP_PKEY_missing_parameters.

Check for NULL and return error if so.
This can possibly be called from apps/ca.c with a NULL argument.

Reviewed-by: Paul Yang <kaishen.yy@antfin.com>
(Merged from https://github.com/openssl/openssl/pull/10474)

(cherry picked from commit ab5c77b4766e0992751d86560193ca42b49cf316)

Engine: Add NULL check.

Add NULL check for return from pkey_asn1_meths.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10473)

(cherry picked from commit 9bada854de16bcc1a9dc199b4b352b19ab6897fc)

ECDSA: don't clear free memory after verify.

Verifications are public, there is no need to clear the used storage before
freeing it.

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10475)

(cherry picked from commit cff7d199e0dc51ae939de5fb7702aab2a9ef30fc)

Fix sha512_block_data_order_avx2 backtrace info

We store a secondary frame pointer info for the debugger
in the red zone.

Fixes #8853

[extended tests]

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9624)

(cherry picked from commit 9ce91035bcf7d74fe15c94650f3bc1f89b7c0f07)

Allow specifying the tag after AAD in CCM mode (2)

In addition to 67c81ec3 which introduced this behavior in CCM mode
docs but only implemented it for AES-CCM.

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10331)

(cherry picked from commit f7382fbbd846dd3bdea6b8c03b6af22faf0ab94f)

Conflicts:
test/recipes/30-test_evp_data/evpciph.txt

s390x assembly pack: fix bn_mul_comba4

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10454)

(cherry picked from commit 97a986f78289fef71bf8778dc4763458e983750c)

Add missing EVP_PKEY_METHOD accessors for digestsign and digestverify

Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10388)

(cherry picked from commit 2555285fa5e4248ad4a5a0bc14ae4606443856c2)

travis.yml: add arch s390x target

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10446)

(cherry picked from commit 985412f8c14853b9936852bc7ef4d9438db27b88)

testutil/init.c rename to testutil/testutil_init.c

Avoid conflicts with some linkers.

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10439)

(cherry picked from commit e74b5dcf16dfd7c91d9f9a7e69c447f00d778e17)

Conflicts:
test/build.info

chacha_enc.c: fix for EBCDIC platforms

Signed-off-by: Joerg Schmidbauer <jschmidb@de.ibm.com>
Reviewed-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10417)

(cherry picked from commit c31950b964a2f3f7b9e6ad98076954178ee1e77d)

Extend docs for EC_POINT conversion functions

Add more explicit documentation about the relation between
EC_POINT_point2oct(), EC_POINT_point2hex(), EC_POINT_point2bn() and
their reverse.

In particular highlight that EC_POINT_point2oct() and
EC_POINT_oct2point() conform to, respectively, Sec. 2.3.3 and Sec. 2.3.4
of the SECG SEC 1 standard (which is the normative reference for the
already mentioned RFC 5480), highlighting with a note how this affect
the encoding/decoding of the point at infinity (which in contrast with
any other valid generic point of a curve is assigned an exceptional
fixed octet string encoding, i.e., 0x00).

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10329)

(cherry picked from commit 3cc26f2eba8a8c16ac559e68c05c094d7ea6bd8b)

Add self-test for EC_POINT_hex2point

Adds tests for each curve to ensure that encodings obtained through
EC_POINT_hex2point() can be fed to EC_POINT_point2hex() yielding a point
identical to the one from which the encoding is generated.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10329)

(cherry picked from commit 35ed029b5a488924890fda2487c87f664361a33b)

Fix EC_POINT_bn2point() for BN_zero()

EC_POINT_bn2point() rejected BIGNUMs with a zero value.

This behavior indirectly caused failures when converting a point
at infinity through EC_POINT_point2hex() and then back to a point with
EC_POINT_hex2point().

With this change such BIGNUMs are treated like any other and exported to
an octet buffer filled with zero.
It is then EC_POINT_oct2point() (either the default implementation or
the custom one in group->meth->oct2point) to determine if such encoding
maps to a valid point (generally the point at infinity is encoded as
0x00).

Fixes #10258

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10329)

(cherry picked from commit d47c10875656790d146f62ac3c437db54c58dbf7)

Add more tests for apps/req

https://github.com/openssl/openssl/issues/10224#issuecomment-546593113
highlighted that existing testing infrastructure is not covering common
usage patterns of the `req` app.

This commit explicitly adds request generations thorugh the CLI using
RSA, DSA and ECDSA (P-256) keys.

(cherry picked from commit b2a7310af0dd190712bae2e462a7708483dd4628)

Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10369)

SSL: Document SSL_add_{file,dir}_cert_subjects_to_stack()

This also removes the incorrect documentation comments by those
functions.

Reviewed-by: Tomas Mraz <tmraz@fedoraproject.org>
(Merged from https://github.com/openssl/openssl/pull/10403)

Fix misspelled resumption_label for CHARSET_EBCDIC

The resumption_label variable when CHARSET_EBCDIC was enabled, was misspelled.
Instead of evaluating to 'res binder' as expected, it evaluated to 'red binder'.

CLA: trivial

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10396)

(cherry picked from commit 6ed12cec7216c3e81b58f5cafa41775e456feaee)

Workaround for Windows-based GOST implementations

Many Windows-based GOST TLS implementations are unable to extend the
list of supported SignatureAlgorithms because of lack of the necessary
callback in Windows. So for TLS 1.2 it makes sense to imply the support
of GOST algorithms in case when the GOST ciphersuites are present.

This is a backport of #10377 to 1.1.1 branch

Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/10378)

Fix --strict-warnings build

Appease -Wstring-plus-int.

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/9608)

(cherry picked from commit e0249827b3fa81ff6c59fb14ef85d38361dd5e31)

Fix a -Warray-bounds gcc warning in OPENSSL_DIR_read

'__builtin_strncpy' offset [275, 4095] from the object at
'direntry' is out of the bounds of referenced subobject 'd_name'
with type 'char[256]' at offset 19

Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10343)

(cherry picked from commit db5cf86535b305378308c58c52596994e1ece1e6)

conf_def.c: Avoid calling strlen() in a loop

Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/10361)

(cherry picked from commit d1c1fb2d41a627293483d832aaffcb6eca9075f9)

Configure: Make --strict-warnings meaningful with MSVC cl

We also add this to our x86_64 builds on appveyor

(cherry picked from commit b4a7b4ec4acc712b1f22a83966ac986b510f25d8)

Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10365)

BIO_s_connect: add an error state and use it

If no connection could be made, addr_iter will eventually end up being
NULL, and if the user didn't check the returned error value, the
BIO_CONN_S_CONNECT code will be performed again and will crash.

So instead, we add a state BIO_CONN_S_CONNECT_ERROR that we enter into
when we run out of addresses to try.  That state will just simply say
"error" back, until the user does something better with the BIO, such
as free it or reset it.

Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10333)

Don't leak memory in the event of a failure in i2v_GENERAL_NAMES

i2v_GENERAL_NAMES call i2v_GENERAL_NAME repeatedly as required. Each
time i2v_GENERAL_NAME gets called it allocates adds data to the passed in
stack and then returns a pointer to the stack, or NULL on failure. If
the passed in stack is itself NULL then it allocates one.

i2v_GENERAL_NAMES was not correctly handling the case where a NULL gets
returned from i2v_GENERAL_NAME. If a stack had already been allocated then
it just leaked it.

Reviewed-by: Dmitry Belyavskiy <beldmit@gmail.com>
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10300)

(cherry picked from commit 45b244620a74248b46ebe1c85e86437b9641447a)

s390x assembly pack: enable clang build

clang imposes some restrictions on the assembler code that
gcc does not.

Signed-off-by: Patrick Steuer <patrick.steuer@de.ibm.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10330)

(cherry picked from commit 6f93f06135cbbd36c3fe98d63717e8303a5d559b)

Conflicts:
crypto/perlasm/s390x.pm (non-existant)
crypto/s390xcpuid.pl (code to be changed non-existant)