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)
* Add SSL_VERIFY_POST_HANDSHAKE to postpone client authentication until after
the initial handshake.
* Update SSL_VERIFY_CLIENT_ONCE now only sends out one CertRequest regardless
of when the certificate authentication takes place; either initial handshake,
re-negotiation, or post-handshake authentication.
Add 'RequestPostHandshake' and 'RequirePostHandshake' SSL_CONF options that
add the SSL_VERIFY_POST_HANDSHAKE to the 'Request' and 'Require' options
Add support to s_client:
* Enabled automatically when cert is configured
* Can be forced enabled via -force_pha
Add support to s_server:
* Use 'c' to invoke PHA in s_server
* Remove some dead code
Update documentation
Update unit tests:
* Illegal use of PHA extension
* TLSv1.3 certificate tests
DTLS and TLS behave ever-so-slightly differently. So, when DTLS1.3 is
implemented, it's PHA support state machine may need to be different.
Add a TODO and a #error
Update handshake context to deal with PHA.
The handshake context for TLSv1.3 post-handshake auth is up through the
ClientFinish message, plus the CertificateRequest message. Subsequent
Certificate, CertificateVerify, and Finish messages are based on this
handshake context (not the Certificate message per se, but it's included
after the hash). KeyUpdate, NewSessionTicket, and prior Certificate
Request messages are not included in post-handshake authentication.
After the ClientFinished message is processed, save off the digest state
for future post-handshake authentication. When post-handshake auth occurs,
copy over the saved handshake context into the "main" handshake digest.
This effectively discards the any KeyUpdate or NewSessionTicket messages
and any prior post-handshake authentication.
This, of course, assumes that the ID-22 did not mean to include any
previous post-handshake authentication into the new handshake transcript.
This is implied by section 4.4.1 that lists messages only up to the
first ClientFinished.
Reviewed-by: Ben Kaduk <kaduk@mit.edu> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4964)
Richard Levitte [Sun, 28 Jan 2018 08:49:46 +0000 (09:49 +0100)]
Make sure that apps/openssl prefixes its output with '# ' during tests
The reason to do this is that some output might start with an 'ok',
which TAP catches and takes for TAP output. The TAP compatible way is
to make all output it shouldn't catch look like comments.
We do this by setting the environment variable HARNESS_OSSL_PREFIX
during tests. When that is set, apps/openssl uses BIO_f_linebuffer
and sets its prefix to the content of that environment variable.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5224)
Richard Levitte [Wed, 31 Jan 2018 13:15:52 +0000 (14:15 +0100)]
Apps: divide the modules in direct command modules, support library and init
Most modules are direct implementations of openssl application
sub-commands, but some constitute a support library, which can be used
by more than one program (and is, incidently, by test/uitest).
For practical purposes, we place the support library modules in a
private, static library.
Finally, there are some modules that don't have direct references in
the rest of the apps code, but are still crucial. See them as some
kind of extra crt0 or similar for your platform.
Inspiration from David von Oheimb
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5222)
Richard Levitte [Wed, 31 Jan 2018 10:13:10 +0000 (11:13 +0100)]
apps: Don't include progs.h in apps.h
Everything in apps includes apps.h, because that one declares apps
internal library routines. However, progs.h doesn't declare library
routines, but rather the main commands and their options, and there's
no reason why the library modules should include it.
So, remove the inclusion of progs.h from apps.h and add that inclusion
in all command source files.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5222)
Benjamin Kaduk [Fri, 26 Jan 2018 15:32:40 +0000 (09:32 -0600)]
Restore clearing of init_lock after free
The behavior of resetting the init_lock value to NULL after
freeing it during OPENSSL_cleanup() was added as part of the
global lock commits that were just reverted, but there is desire
to retain this behavior for clarity.
It is unclear that the library would actually remain usable in
any form after OPENSSL_cleanup(), since the required re-initialization
occurs under a CRYPTO_ONCE check that cannot be reset at cleanup time.
That said, a NULL dereference is probably more friendly behavior
in these treacherous waters than using freed memory would be.
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/5089)
But there were some intervening commits that made neither revert apply
cleanly, so instead do it all as one shot.
The crypto global locks were an attempt to cope with the awkward
POSIX semantics for pthread_atfork(); its documentation (the "RATIONALE"
section) indicates that the expected usage is to have the prefork handler
lock all "global" locks, and the parent and child handlers release those
locks, to ensure that forking happens with a consistent (lock) state.
However, the set of functions available in the child process is limited
to async-signal-safe functions, and pthread_mutex_unlock() is not on
the list of async-signal-safe functions! The only synchronization
primitives that are async-signal-safe are the semaphore primitives,
which are not really appropriate for general-purpose usage.
However, the state consistency problem that the global locks were
attempting to solve is not actually a serious problem, particularly for
OpenSSL. That is, we can consider four cases of forking application
that might use OpenSSL:
(1) Single-threaded, does not call into OpenSSL in the child (e.g.,
the child calls exec() immediately)
For this class of process, no locking is needed at all, since there is
only ever a single thread of execution and the only reentrancy is due to
signal handlers (which are themselves limited to async-signal-safe
operation and should not be doing much work at all).
(2) Single-threaded, calls into OpenSSL after fork()
The application must ensure that it does not fork() with an unexpected
lock held (that is, one that would get unlocked in the parent but
accidentally remain locked in the child and cause deadlock). Since
OpenSSL does not expose any of its internal locks to the application
and the application is single-threaded, the OpenSSL internal locks
will be unlocked for the fork(), and the state will be consistent.
(OpenSSL will need to reseed its PRNG in the child, but that is
an orthogonal issue.) If the application makes use of locks from
libcrypto, proper handling for those locks is the responsibility of
the application, as for any other locking primitive that is available
for application programming.
(3) Multi-threaded, does not call into OpenSSL after fork()
As for (1), the OpenSSL state is only relevant in the parent, so
no particular fork()-related handling is needed. The internal locks
are relevant, but there is no interaction with the child to consider.
(4) Multi-threaded, calls into OpenSSL after fork()
This is the case where the pthread_atfork() hooks to ensure that all
global locks are in a known state across fork() would come into play,
per the above discussion. However, these "calls into OpenSSL after
fork()" are still subject to the restriction to async-signal-safe
functions. Since OpenSSL uses all sorts of locking and libc functions
that are not on the list of safe functions (e.g., malloc()), this
case is not currently usable and is unlikely to ever be usable,
independently of the locking situation. So, there is no need to
go through contortions to attempt to support this case in the one small
area of locking interaction with fork().
In light of the above analysis (thanks @davidben and @achernya), go
back to the simpler implementation that does not need to distinguish
"library-global" locks or to have complicated atfork handling for locks.
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/5089)
crypto/rand/rand_lib.c: fix undefined reference to `clock_gettime'
Some older glibc versions require the `-lrt` linker option for
resolving the reference to `clock_gettime'. Since it is not desired
to add new library dependencies in version 1.1.1, the call to
clock_gettime() is replaced by a call to gettimeofday() for the
moment. It will be added back in version 1.2.
Signed-off-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com> Reviewed-by: Richard Levitte <levitte@openssl.org> Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/5199)
Configure: Fix configdata.pm shorthand for --dump, should be -d
INSTALL: Mention 'aria' algorithm for no-<alg>
Signed-off-by: Peter Meerwald-Stadler <pmeerw@pmeerw.net> Reviewed-by: Paul Dale <paul.dale@oracle.com> Reviewed-by: Richard Levitte <levitte@openssl.org> Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/5215)
Matt Caswell [Mon, 29 Jan 2018 14:19:52 +0000 (14:19 +0000)]
Move decisions about whether to accept reneg into the state machine
If a server receives an unexpected ClientHello then we may or may not
accept it. Make sure all such decisions are made in the state machine
and not in the record layer. This also removes a disparity between the
TLS and the DTLS code. The TLS code was making this decision in the
record layer, while the DTLS code was making it later.
Finally it also solves a problem where a warning alert was being sent
during tls_setup_handshake() and the function was returning a failure
return code. This is problematic because it can be called from a
transition function - which we only allow fatal errors to occur in.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5190)
Richard Levitte [Mon, 29 Jan 2018 05:19:17 +0000 (06:19 +0100)]
Make configdata.pm runnable and move all display of information there
The "make variable" information displayed by Configure was selective
and incomplete, and possibly undesirable (too verbose).
Instead, we make configdata.pm and have the user run it to get the
information they desire, and also make it possible to have it perform
a reconfiguration.
Possibilities so far:
perl configdata.pm --dump Displays everything (i.e. the
combined output from
--command-line, --environment,
--make-variables and
--build-parameters.
perl configdata.pm --command-line Displays the config command
line.
perl configdata.pm --envirnoment Displays the recorded
environment variables.
perl configdata.pm --make-variables Displays the configured "make
variables".
perl configdata.pm --build-parameters Displays the build file and
the template files to create
it.
perl configdata.pm --reconfigure Re-runs the configuration with
the recorded environment
variables.
--verbose can be used to have --reconfigure be a bit more verbose.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5185)
Kurt Roeckx [Fri, 17 Nov 2017 14:00:35 +0000 (15:00 +0100)]
Add RAND_DRBG_bytes
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/4752)
nickthetait [Sun, 28 Jan 2018 19:15:23 +0000 (20:15 +0100)]
Create troubleshooting subsection in INSTALL file
Fixes: #5130 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/5178)
Richard Levitte [Sat, 27 Jan 2018 13:56:06 +0000 (14:56 +0100)]
Treat C++ flags more like C flags, and only if C++ compiler specified
C++ flags got the same config target value as C flags, but then
nothing else happened while C flags get all kinds of stuff added to
them (especially when --strict-warnings is used).
Now, C++ flags get the exact same treatment as C flags. However, this
only happens when a C++ compiler is specified, to avoid confusing
messages about added C++ flags.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5181)
Steve Linsell [Sun, 28 Jan 2018 11:01:04 +0000 (12:01 +0100)]
Update copyright year in mkerr.pl
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/5166)
Richard Levitte [Sat, 27 Jan 2018 12:01:44 +0000 (13:01 +0100)]
We need Unixly defaults for config targets that don't inherit a BASE
Ideally, each config target should inherit a base to get their
platform specific defaults. Unfortunately, that is currently not the
case, so we duplicate the Unixly defaults from the BASE_unix template
into the DEFAULT template.
Reviewed-by: Tim Hudson <tjh@openssl.org> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5177)
AR (GNU compatible)
ARFLAGS (GNU Compatible)
AS (GNU Compatible)
ASFLAGS (GNU Compatible)
CC (GNU Compatible)
CFLAGS (GNU Compatible)
CXX (GNU Compatible)
CXXFLAGS (GNU Compatible)
CPP (GNU Compatible)
CPPFLAGS (GNU Compatible)
CPPDEFINES List of CPP macro definitions. Alternative for -D
CPPINCLUDES List of CPP inclusion directories. Alternative for -I
HASHBANGPERL Perl invocation to be inserted after '#!' in public
perl scripts.
LDFLAGS (GNU Compatible)
LDLIBS (GNU Compatible)
RANLIB Program to generate library archive index
RC Program to manipulate Windows resources
RCFLAGS Flags for $(RC)
RM (GNU Compatible)
Setting one of these overrides the corresponding data from our config
targets. However, flags given directly on the configuration command
line are additional, and are therefore added to the flags coming from
one of the variables above or the config target.
Fixes #2420
Reviewed-by: Tim Hudson <tjh@openssl.org> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5177)
Richard Levitte [Tue, 23 Jan 2018 12:54:55 +0000 (13:54 +0100)]
Processing GNU-style "make variables" - separate CPP flags from C flags
C preprocessor flags get separated from C flags, which has the
advantage that we don't get loads of macro definitions and inclusion
directory specs when linking shared libraries, DSOs and programs.
This is a step to add support for "make variables" when configuring.
Reviewed-by: Tim Hudson <tjh@openssl.org> Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5177)
Benjamin Kaduk [Fri, 26 Jan 2018 15:21:08 +0000 (09:21 -0600)]
Fix ssl-trace with TLS 1.3 draft-23 PSS sigalgs
The latest TLS 1.3 draft split the RSA-PSS signature schemes into
two versions that indicate the OID of the RSA key being used.
This forced us to rename the preprocessor defines for the sigalg
values, and the ssl-trace code was not adopted to match, since
it was not enabled int the default build.
Belatedly update the ssl_sigalg_tbl in the trace code to match.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5174)
Benjamin Kaduk [Wed, 24 Jan 2018 20:45:08 +0000 (14:45 -0600)]
Add TLSProxy tests for signature_algorithms_cert
We don't need to send this extension in normal operation since
we are our own X.509 library, but add some test cases that force
the extension to be sent and exercise our code to process the extension.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5068)
Benjamin Kaduk [Thu, 11 Jan 2018 17:47:12 +0000 (11:47 -0600)]
Add support for the TLS 1.3 signature_algorithms_cert extension
The new extension is like signature_algorithms, but only for the
signature *on* the certificate we will present to the peer (the
old signature_algorithms extension is still used for signatures that
we *generate*, i.e., those over TLS data structures).
We do not need to generate this extension, since we are the same
implementation as our X.509 stack and can handle the same types
of signatures, but we need to be prepared to receive it, and use the received
information when selecting what certificate to present.
There is a lot of interplay between signature_algorithms_cert and
signature_algorithms, since both affect what certificate we can
use, and thus the resulting signature algorithm used for TLS messages.
So, apply signature_algorithms_cert (if present) as a filter on what
certificates we can consider when choosing a certificate+sigalg
pair.
As part of this addition, we also remove the fallback code that let
keys of type EVP_PKEY_RSA be used to generate RSA-PSS signatures -- the
new rsa_pss_pss_* and rsa_pss_rsae_* signature schemes have pulled
the key type into what is covered by the signature algorithm, so
we should not apply this sort of compatibility workaround.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5068)
Benjamin Kaduk [Thu, 18 Jan 2018 05:21:19 +0000 (23:21 -0600)]
Update documentation for SSL_set1_sigalgs()
These functions can now take both "sig+hash" strings and
algorithm-specific identifiers like "rsa_pss_pss_sha256" that
indicate a particular entry from the TLS signature algorithm
registry.
Also clarify that only the "_list" form allows for the new-style names
(the non-"list" interfaces take sig and hasn NIDs, which cannot
access all of the new-style schemes).
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5068)
Benjamin Kaduk [Wed, 17 Jan 2018 17:55:29 +0000 (11:55 -0600)]
Propagate TLS 1.3 sigalgs through tls1_set_sigalgs()
Our historical SSL{,_CTX}_set_sigalgs() APIs take an array of
NID pairs (hash and signature), and our parser for manually
specifying unified sigalgs (that do not necessarily correspond
to an actual signature+hash pair) was transiting via (the implementation
of) this historical API. The TLS 1.3 draft-23 has introduced
signature schemes that have identical signature type and hash type,
differing only in the (RSA) public key OID, which prevents
the rsa_pss_pss_* schemes from being properly identified and
sent on the wire.
To fix the issue, parse sigalg strings directly into SIGALG_LOOKUP
objects, and pass around an array of uint16 wire protocol values
instead of NID pairs. The old interface is retained for API
compatibility but will become less and less useful with time.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5068)
Benjamin Kaduk [Thu, 11 Jan 2018 19:39:30 +0000 (13:39 -0600)]
Add TLS 1.3 draft-23 PSS signature algorithms
We now have a split in the signature algorithms codepoint space for
whether the certificate's key is for rsaEncryption or a PSS-specific
key, which should let us get rid of some special-casing that we
previously needed to try to coax rsaEncryption keys into performing PSS.
(This will be done in a subsequent commit.)
Send the new PSS-with-PSS-specific key first in our list, so that
we prefer the new technology to the old one.
We need to update the expected certificate type in one test,
since the "RSA-PSS+SHA256" form now corresponds to a public key
of type rsaEncryption, so we should expect the server certificate
type to be just "RSA". If we want to get a server certificate
type of "RSA-PSS", we need to use a new signature algorithm
that cannot be represented as signature+hash, so add a test for that
as well.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5068)
Christian Heimes [Sun, 21 Jan 2018 09:37:59 +0000 (10:37 +0100)]
Fix signature of min/max proto getter
The getters for min and max proto version wrongly passed NULL instead of
0 as third argument to SSL_ctrl() and SSL_CTX_ctrl(). The third argument
is not used, but the error results in a compiler warning:
warning: passing argument 3 of ‘SSL_CTX_ctrl’ makes integer from pointer without a cast [-Wint-conversion]
int v = SSL_CTX_get_max_proto_version(self->ctx);
See https://github.com/openssl/openssl/pull/4364
Signed-off-by: Christian Heimes <christian@python.org> Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Kurt Roeckx <kurt@roeckx.be> Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/5128)
David Cooper [Fri, 18 Aug 2017 13:27:19 +0000 (09:27 -0400)]
Add -rsigopt option to ocsp command
Add a -rsigopt option to the ocsp command that allows signature parameters to be provided for the signing of OCSP responses. The parameters that may be provided to -rsigopt are the same as may be provided to -sigopt in the ca, req, and x509 commands.
This PR also defines a OCSP_basic_sign_ctx() function, which functions in the same way as OCSP_basic_sign(), except that it accepts a EVP_MD_CTX rather than a key and digest. The OCSP_basic_sign_ctx() function is used to implement the -rsigopt option in the ocsp command.
Reviewed-by: Rich Salz <rsalz@openssl.org> Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4190)
Richard Levitte [Tue, 23 Jan 2018 18:13:48 +0000 (19:13 +0100)]
Configure: ensure that a DEPEND generates the correct inclusion directory
We incorrectly assumed that explicit dependencies meant that the
source directory would be added for inclusion. However, if the
dependent file is generated, it's stored in the build directory, and
that should be used for inclusion rather than the source directory.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5153)
Richard Levitte [Mon, 22 Jan 2018 18:03:37 +0000 (19:03 +0100)]
Have EVP_PKEY_asn1_find_str() work more like EVP_PKEY_asn1_find()
EVP_PKEY_asn1_find_str() would search through standard asn1 methods
first, then those added by the application, which EVP_PKEY_asn1_find()
worked the other way around. Also, EVP_PKEY_asn1_find_str() didn't
handle aliases.
This change brings EVP_PKEY_asn1_find_str() closer to EVP_PKEY_asn1_find().
Fixes #5086
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/5137)