]> granicus.if.org Git - openssl/log
openssl
9 years agoServer side version negotiation rewrite
Matt Caswell [Fri, 27 Mar 2015 23:01:51 +0000 (23:01 +0000)]
Server side version negotiation rewrite

This commit changes the way that we do server side protocol version
negotiation. Previously we had a whole set of code that had an "up front"
state machine dedicated to the negotiating the protocol version. This adds
significant complexity to the state machine. Historically the justification
for doing this was the support of SSLv2 which works quite differently to
SSLv3+. However, we have now removed support for SSLv2 so there is little
reason to maintain this complexity.

The one slight difficulty is that, although we no longer support SSLv2, we
do still support an SSLv3+ ClientHello in an SSLv2 backward compatible
ClientHello format. This is generally only used by legacy clients. This
commit adds support within the SSLv3 code for these legacy format
ClientHellos.

Server side version negotiation now works in much the same was as DTLS,
i.e. we introduce the concept of TLS_ANY_VERSION. If s->version is set to
that then when a ClientHello is received it will work out the most
appropriate version to respond with. Also, SSLv23_method and
SSLv23_server_method have been replaced with TLS_method and
TLS_server_method respectively. The old SSLv23* names still exist as
macros pointing at the new name, although they are deprecated.

Subsequent commits will look at client side version negotiation, as well of
removal of the old s23* code.

Reviewed-by: Kurt Roeckx <kurt@openssl.org>
9 years agoRemove unused #ifdef's from header files
Rich Salz [Thu, 14 May 2015 19:36:09 +0000 (15:36 -0400)]
Remove unused #ifdef's from header files

And remove a duplicate comment, probably from a merge hiccup.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoUse #error in openssl/srp.h
Rich Salz [Thu, 14 May 2015 19:21:36 +0000 (15:21 -0400)]
Use #error in openssl/srp.h

Follow the same convention the other OPENSSL_NO_xxx header files
do, and use #error instead of making the header file be a no-op.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoMake up for a missed 'make update' update.
Rich Salz [Fri, 15 May 2015 10:48:14 +0000 (06:48 -0400)]
Make up for a missed 'make update' update.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoUpdate mkdef for moved header file.
Rich Salz [Thu, 14 May 2015 19:57:30 +0000 (15:57 -0400)]
Update mkdef for moved header file.

crypto/cryptlib.h moved to crypto/include/internal; update
the script.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agomake depend
Richard Levitte [Thu, 14 May 2015 15:38:31 +0000 (17:38 +0200)]
make depend

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd -Iinclude to crypto/ compiles
Richard Levitte [Thu, 14 May 2015 15:36:13 +0000 (17:36 +0200)]
Add -Iinclude to crypto/ compiles

The move of headers from crypto/ to crypto/include/internal/ needs
this extra inclusion directory or the build fails.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoIdentify and move common internal libcrypto header files
Richard Levitte [Thu, 14 May 2015 14:56:48 +0000 (16:56 +0200)]
Identify and move common internal libcrypto header files

There are header files in crypto/ that are used by a number of crypto/
submodules.  Move those to crypto/include/internal and adapt the
affected source code and Makefiles.

The header files that got moved are:

crypto/cryptolib.h
crypto/md32_common.h

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdjust unixly mk1mf after introduction of tkey
Richard Levitte [Thu, 14 May 2015 13:55:59 +0000 (15:55 +0200)]
Adjust unixly mk1mf after introduction of tkey

Added depencies on the public variants of some keys in test to Makefile.
Added the newly introduced key files from test/ in the list of files
to copy in util/pl/unix.pl.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agomake depend
Richard Levitte [Thu, 14 May 2015 12:58:15 +0000 (14:58 +0200)]
make depend

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoIdentify and move OpenSSL internal header files
Richard Levitte [Thu, 14 May 2015 12:54:49 +0000 (14:54 +0200)]
Identify and move OpenSSL internal header files

There are header files in crypto/ that are used by the rest of
OpenSSL.  Move those to include/internal and adapt the affected source
code, Makefiles and scripts.

The header files that got moved are:

crypto/constant_time_locl.h
crypto/o_dir.h
crypto/o_str.h

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoMove definition of INTxx_MIN et al to internal header
Richard Levitte [Thu, 14 May 2015 06:44:06 +0000 (08:44 +0200)]
Move definition of INTxx_MIN et al to internal header

Having the INTxx_MIN et al macros defined in a public header is
unnecessary and risky.  Also, it wasn't done for all platforms that
might need it.

So we move those numbers to an internal header file, do the math
ourselves and make sure to account for the integer representations we
know of.

This introduces include/internal, which is unproblematic since we
already use -I$(TOP)/include everywhere.  This directory is different
from crypto/include/internal, as the former is more general internal
headers for all of OpenSSL, while the latter is for libcrypto only.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agouse unit64_t for CPUID and timestamp code
Dr. Stephen Henson [Wed, 13 May 2015 16:47:13 +0000 (17:47 +0100)]
use unit64_t for CPUID and timestamp code

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoAllow use of standard integer types.
Dr. Stephen Henson [Wed, 13 May 2015 16:34:53 +0000 (17:34 +0100)]
Allow use of standard integer types.

Include appropriate headers for standard integer types in e_os2.h

This should use stdint.h, inttypes.h or a workaround for systems which
have neither.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoAdd missing ctype.h
Rich Salz [Wed, 13 May 2015 18:41:53 +0000 (14:41 -0400)]
Add missing ctype.h

Had old patch, forgot to push/patch this to master

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoFix ocsp bugs
Rich Salz [Sat, 2 May 2015 03:36:11 +0000 (23:36 -0400)]
Fix ocsp bugs

Various bugs found by Viktor, Emilia, Matt, etc.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoRT3841: memset() cipher_data when allocated
Rich Salz [Fri, 8 May 2015 16:23:56 +0000 (12:23 -0400)]
RT3841: memset() cipher_data when allocated

If an EVP implementation (such as an engine) fails out early, it's
possible to call EVP_CIPHER_CTX_cleanup() which will call
ctx->cipher->cleanup() before the cipher_data has been initialized
via ctx->cipher->init().  Guarantee it's all-bytes-zero as soon as
it is allocated.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoRT1207: document SSL_COMP_free_compression_methods.
Rich Salz [Fri, 8 May 2015 20:06:50 +0000 (16:06 -0400)]
RT1207: document SSL_COMP_free_compression_methods.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoRemove some outdated #defines.
Rich Salz [Tue, 12 May 2015 18:13:01 +0000 (14:13 -0400)]
Remove some outdated #defines.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoutil/incore update.
Andy Polyakov [Mon, 20 Apr 2015 13:53:12 +0000 (15:53 +0200)]
util/incore update.

Support cross-compile platforms that don't support FINGERPRINT_premain,
e.g. VxWorks.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agobn/Makefile: give MacOS X hand to compiler armv8-mont module.
Andy Polyakov [Mon, 20 Apr 2015 18:48:59 +0000 (20:48 +0200)]
bn/Makefile: give MacOS X hand to compiler armv8-mont module.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agobn/asm/armv8-mont.pl: boost performance.
Andy Polyakov [Mon, 20 Apr 2015 13:13:24 +0000 (15:13 +0200)]
bn/asm/armv8-mont.pl: boost performance.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoConfigurations/10-main.conf: update iOS commentary.
Andy Polyakov [Tue, 21 Apr 2015 07:13:27 +0000 (09:13 +0200)]
Configurations/10-main.conf: update iOS commentary.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoaes/asm/bsaes-armv7.pl: fix compilation with Xcode 6.3.
Andy Polyakov [Mon, 20 Apr 2015 21:18:08 +0000 (23:18 +0200)]
aes/asm/bsaes-armv7.pl: fix compilation with Xcode 6.3.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoaes/asm/aesni-sha256-x86_64.pl: fix Windows compilation failure with old assembler.
Andy Polyakov [Tue, 21 Apr 2015 12:52:17 +0000 (14:52 +0200)]
aes/asm/aesni-sha256-x86_64.pl: fix Windows compilation failure with old assembler.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoCall of memcmp with null pointers in obj_cmp()
Hanno Böck [Mon, 11 May 2015 10:33:37 +0000 (11:33 +0100)]
Call of memcmp with null pointers in obj_cmp()

The function obj_cmp() (file crypto/objects/obj_dat.c) can in some
situations call memcmp() with a null pointer and a zero length.

This is invalid behaviour. When compiling openssl with undefined
behaviour sanitizer (add -fsanitize=undefined to compile flags) this
can be seen. One example that triggers this behaviour is the pkcs7
command (but there are others, e.g. I've seen it with the timestamp
function):
apps/openssl pkcs7 -in test/testp7.pem

What happens is that obj_cmp takes objects of the type ASN1_OBJECT and
passes their ->data pointer to memcmp. Zero-sized ASN1_OBJECT
structures can have a null pointer as data.

RT#3816

Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd CHANGES entry for Kerberos removal
Matt Caswell [Tue, 12 May 2015 12:12:16 +0000 (13:12 +0100)]
Add CHANGES entry for Kerberos removal

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoRemove remaining Kerberos references
Matt Caswell [Tue, 12 May 2015 11:14:13 +0000 (12:14 +0100)]
Remove remaining Kerberos references

Following on from the removal of libcrypto and libssl support for Kerberos
this commit removes all remaining references to Kerberos.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoRemove Kerberos support from libcrypto
Matt Caswell [Tue, 12 May 2015 10:49:21 +0000 (11:49 +0100)]
Remove Kerberos support from libcrypto

Remove libcrypto support for Kerberos following on from the previous commit
which removed it from libssl.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoRemove Kerberos support from libssl
Matt Caswell [Tue, 12 May 2015 09:27:53 +0000 (10:27 +0100)]
Remove Kerberos support from libssl

Remove RFC2712 Kerberos support from libssl. This code and the associated
standard is no longer considered fit-for-purpose.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoRemove Kerberos support from apps
Matt Caswell [Tue, 12 May 2015 09:35:51 +0000 (10:35 +0100)]
Remove Kerberos support from apps

Remove Kerberos related options from the apps to prepare for the
subsequent commits which will remove libcrypto and libssl support for
Kerberos.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoDon't allow a CCS when expecting a CertificateVerify
Matt Caswell [Wed, 6 May 2015 20:31:16 +0000 (21:31 +0100)]
Don't allow a CCS when expecting a CertificateVerify

Currently we set change_cipher_spec_ok to 1 before calling
ssl3_get_cert_verify(). This is because this message is optional and if it
is not sent then the next thing we would expect to get is the CCS. However,
although it is optional, we do actually know whether we should be receiving
one in advance. If we have received a client cert then we should expect
a CertificateVerify message. By the time we get to this point we will
already have bombed out if we didn't get a Certificate when we should have
done, so it is safe just to check whether |peer| is NULL or not. If it is
we won't get a CertificateVerify, otherwise we will. Therefore we should
change the logic so that we only attempt to get the CertificateVerify if
we are expecting one, and not allow a CCS in this scenario.

Whilst this is good practice for TLS it is even more important for DTLS.
In DTLS messages can be lost. Therefore we may be in a situation where a
CertificateVerify message does not arrive even though one was sent. In that
case the next message the server will receive will be the CCS. This could
also happen if messages get re-ordered in-flight. In DTLS if
|change_cipher_spec_ok| is not set and a CCS is received it is ignored.
However if |change_cipher_spec_ok| *is* set then a CCS arrival will
immediately move the server into the next epoch. Any messages arriving for
the previous epoch will be ignored. This means that, in this scenario, the
handshake can never complete. The client will attempt to retransmit
missing messages, but the server will ignore them because they are the wrong
epoch. The server meanwhile will still be waiting for the CertificateVerify
which is never going to arrive.

RT#2958

Reviewed-by: Emilia Käsper <emilia@openssl.org>
9 years agoFix the heap corruption in libeay32!OBJ_add_object.
Gunnar Kudrjavets [Tue, 12 May 2015 22:44:45 +0000 (15:44 -0700)]
Fix the heap corruption in libeay32!OBJ_add_object.

Original 'sizeof(ADDED_OBJ)' was replaced with 'sizeof(*ao)'. However,
they return different sizes. Therefore as the result heap gets corrupted
and at some point later debug version of malloc() detects the corruption.

On x86 we can observe that as follows:

sizeof(*ao) == 4
sizeof(*ao[0]) == sizeof(ADDED_OBJ) == 8

Issue reproduces with either enabling CRT debug heap or Application
Verifier's full-page heap.

Basic debugging data from the moment the corruption is first detected:

0:000:x86> |
.  0    id: 283c        create  name: openssl.exe
0:000:x86> kcn
 #
00 MSVCR120D!_heap_alloc_dbg_impl
01 MSVCR120D!_nh_malloc_dbg_impl
02 MSVCR120D!_nh_malloc_dbg
03 MSVCR120D!malloc
04 LIBEAY32!default_malloc_ex
05 LIBEAY32!CRYPTO_malloc
06 LIBEAY32!lh_insert
07 LIBEAY32!OBJ_add_object
08 LIBEAY32!OBJ_create
09 openssl!add_oid_section
0a openssl!req_main
0b openssl!do_cmd
0c openssl!main
0d openssl!__tmainCRTStartup
0e openssl!mainCRTStartup
0f KERNEL32!BaseThreadInitThunk
10 ntdll_77d60000!__RtlUserThreadStart
11 ntdll_77d60000!_RtlUserThreadStart

Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoFix cipherlist order.
Dr. Stephen Henson [Tue, 12 May 2015 18:49:06 +0000 (19:49 +0100)]
Fix cipherlist order.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoMake COMP_CTX and COMP_METHOD opaque
Rich Salz [Fri, 8 May 2015 16:05:36 +0000 (12:05 -0400)]
Make COMP_CTX and COMP_METHOD opaque

Since COMP_METHOD is now defined in comp_lcl.h, it is no
longer possible to create new TLS compression methods without
using the OpenSSL source.  Only ZLIB is supported by default.
Also, since the types are opaque, #ifdef guards to use "char *"
instead of the real type aren't necessary.

The changes are actually minor.  Adding missing copyright to some
files makes the diff misleadingly big.

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoAdditional X509_ALGOR documentation
Dr. Stephen Henson [Tue, 12 May 2015 01:05:17 +0000 (02:05 +0100)]
Additional X509_ALGOR documentation

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoUse p==NULL not !p (in if statements, mainly)
Rich Salz [Wed, 6 May 2015 17:43:59 +0000 (13:43 -0400)]
Use p==NULL not !p (in if statements, mainly)

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoAdd missing NULL check in X509V3_parse_list()
Kurt Cancemi [Thu, 7 May 2015 20:12:33 +0000 (16:12 -0400)]
Add missing NULL check in X509V3_parse_list()

Matt's note: I added a call to X509V3err to Kurt's original patch.

RT#3840

Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd missing terminating NULL to speed_options table.
Kurt Cancemi [Thu, 7 May 2015 22:19:20 +0000 (18:19 -0400)]
Add missing terminating NULL to speed_options table.

This would cause memory corruption in opt_init() because it relies on the
terminating NULL.

RT#3842

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoFix for memcpy() and strcmp() being undefined.
Bjoern D. Rasmussen [Sun, 10 May 2015 05:51:29 +0000 (07:51 +0200)]
Fix for memcpy() and strcmp() being undefined.

clang says: "s_cb.c:958:9: error: implicitly declaring library function
'memcpy'"

Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoCheck sk_SSL_CIPHER_new_null return value
Matt Caswell [Sat, 9 May 2015 05:51:25 +0000 (13:51 +0800)]
Check sk_SSL_CIPHER_new_null return value

If sk_SSL_CIPHER_new_null() returns NULL then ssl_bytes_to_cipher_list()
should also return NULL.

Based on an original patch by mrpre <mrpre@163.com>.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd SSL_use_certificate_chain_file function
Dr. Stephen Henson [Thu, 7 May 2015 13:59:08 +0000 (14:59 +0100)]
Add SSL_use_certificate_chain_file function

Add SSL_use_certiicate_chain file functions: this is works the same
way as SSL_CTX_use_certificate_chain_file but for an SSL structure.

Update SSL_CONF code to use the new function.
Update docs.
Update ordinals.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoFix typo in valid_star
Viktor Dukhovni [Thu, 7 May 2015 17:43:36 +0000 (13:43 -0400)]
Fix typo in valid_star

Reviewed-by: Rich Salz <rsalz@akamai.com>
9 years agoUse "==0" instead of "!strcmp" etc
Rich Salz [Wed, 6 May 2015 18:56:14 +0000 (14:56 -0400)]
Use "==0" instead of "!strcmp" etc

For the various string-compare routines (strcmp, strcasecmp, str.*cmp)
use "strcmp()==0" instead of "!strcmp()"

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoDigest cached records if not sending a certificate.
Dr. Stephen Henson [Wed, 6 May 2015 23:04:48 +0000 (00:04 +0100)]
Digest cached records if not sending a certificate.

If server requests a certificate, but the client doesn't send one, cache
digested records. This is an optimisation and ensures the correct finished
mac is used when extended master secret is used with client authentication.

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoMake sig_app, sigx_app static
Rich Salz [Wed, 6 May 2015 20:03:07 +0000 (16:03 -0400)]
Make sig_app, sigx_app static

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoMake -CAserial a type 's' option
Richard Levitte [Wed, 6 May 2015 16:50:57 +0000 (18:50 +0200)]
Make -CAserial a type 's' option

The file name given to -CAserial might not exist yet.  The
-CAcreateserial option decides if this is ok or not.

Previous to this change, -CAserial was a type '<' option, and in that
case, the existence of the file given as argument is tested quite
early, and is a failure if it doesn't.  With the type 's' option, the
argument is just a string that the application can do whatever it
wants with.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd a -CAserial argument for signing the user cert request
Richard Levitte [Wed, 6 May 2015 16:48:48 +0000 (18:48 +0200)]
Add a -CAserial argument for signing the user cert request

This is just to make sure that option is tested on a Unix build.  This
option is already present in ms/testss.bat, so it's an easy steal.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoSSL_CONF table reorganisation.
Dr. Stephen Henson [Sun, 26 Apr 2015 12:56:40 +0000 (13:56 +0100)]
SSL_CONF table reorganisation.

Add command line switch entries to table and return SSL_CONF_TYPE_NONE for
them in SSL_CONF_cmd_value_type.

Update docs.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoInitialize potentially uninitialized local variables
Gunnar Kudrjavets [Wed, 6 May 2015 09:16:55 +0000 (10:16 +0100)]
Initialize potentially uninitialized local variables

Compiling OpenSSL code with MSVC and /W4 results in a number of warnings.
One category of warnings is particularly interesting - C4701 (potentially
uninitialized local variable 'name' used). This warning pretty much means
that there's a code path which results in uninitialized variables being used
or returned. Depending on compiler, its options, OS, values in registers
and/or stack, the results can be nondeterministic. Cases like this are very
hard to debug so it's rational to fix these issues.

This patch contains a set of trivial fixes for all the C4701 warnings (just
initializing variables to 0 or NULL or appropriate error code) to make sure
that deterministic values will be returned from all the execution paths.

RT#3835

Signed-off-by: Matt Caswell <matt@openssl.org>
Matt's note: All of these appear to be bogus warnings, i.e. there isn't
actually a code path where an unitialised variable could be used - its just
that the compiler hasn't been able to figure that out from the logic. So
this commit is just about silencing spurious warnings.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoFix s_server version specific methods
Matt Caswell [Wed, 6 May 2015 10:17:07 +0000 (11:17 +0100)]
Fix s_server version specific methods

A copy&paste error as a result of the big apps cleanup broke the version
specific methods in s_server.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agomemset, memcpy, sizeof consistency fixes
Rich Salz [Mon, 4 May 2015 22:00:15 +0000 (18:00 -0400)]
memset, memcpy, sizeof consistency fixes

Just as with the OPENSSL_malloc calls, consistently use sizeof(*ptr)
for memset and memcpy.  Remove needless casts for those functions.
For memset, replace alternative forms of zero with 0.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoZLIB compression deserves a better comment
Richard Levitte [Tue, 5 May 2015 22:47:42 +0000 (00:47 +0200)]
ZLIB compression deserves a better comment

What could be better than to refer to the RFC that defines it?

Reviewed-by: Stephen Henson <steve@openssl.org>
9 years agoFix windows build
Matt Caswell [Tue, 5 May 2015 21:09:01 +0000 (22:09 +0100)]
Fix windows build

Fix error in WIN32_rename() introduced by commit b4faea50c35.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agomake update
Richard Levitte [Tue, 5 May 2015 19:49:51 +0000 (21:49 +0200)]
make update

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoRemove the last traces of the fake RLE compression
Richard Levitte [Tue, 5 May 2015 19:48:43 +0000 (21:48 +0200)]
Remove the last traces of the fake RLE compression

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoAdd more error state transitions (DTLS)
Matt Caswell [Thu, 30 Apr 2015 10:50:08 +0000 (11:50 +0100)]
Add more error state transitions (DTLS)

Ensure all fatal errors transition into the new error state for DTLS.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd more error state transitions (client)
Matt Caswell [Thu, 30 Apr 2015 10:32:35 +0000 (11:32 +0100)]
Add more error state transitions (client)

Ensure all fatal errors transition into the new error state on the client
side.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd more error state transitions
Matt Caswell [Thu, 30 Apr 2015 10:11:04 +0000 (11:11 +0100)]
Add more error state transitions

Ensure all fatal errors transition into the new error state on the server
side.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoAdd Error state
Matt Caswell [Thu, 23 Apr 2015 19:01:33 +0000 (20:01 +0100)]
Add Error state

Reusing an SSL object when it has encountered a fatal error can
have bad consequences. This is a bug in application code not libssl
but libssl should be more forgiving and not crash.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoUpdate multi-thread FAQ
Rich Salz [Mon, 4 May 2015 15:54:26 +0000 (11:54 -0400)]
Update multi-thread FAQ

Reviewed-by: Matt Caswell <matt@openssl.org>
9 years agoRemove apps cache for gethostbyname
Rich Salz [Mon, 4 May 2015 22:05:26 +0000 (18:05 -0400)]
Remove apps cache for gethostbyname

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agomake update
Matt Caswell [Mon, 4 May 2015 22:28:41 +0000 (23:28 +0100)]
make update

Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
9 years agossl/kssl.c: include missing header to complete SSL structure's defn.
mancha security [Mon, 4 May 2015 06:36:42 +0000 (06:36 +0000)]
ssl/kssl.c: include missing header to complete SSL structure's defn.

Signed-off-by: mancha security <mancha1@zoho.com>
Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
9 years agossl/ssl_asn1.c: Fix typo introduced via cc5b6a03a320f1
mancha security [Sun, 3 May 2015 02:45:45 +0000 (02:45 +0000)]
ssl/ssl_asn1.c: Fix typo introduced via cc5b6a03a320f1

Signed-off-by: mancha security <mancha1@zoho.com>
Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
9 years agoRemove libcrypto to libssl dependency
Matt Caswell [Mon, 4 May 2015 22:15:46 +0000 (23:15 +0100)]
Remove libcrypto to libssl dependency

Remove dependency on ssl_locl.h from v3_scts.c, and incidentally fix a build problem with
kerberos (the dependency meant v3_scts.c was trying to include krb5.h, but without having been
passed the relevanant -I flags to the compiler)

Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
9 years agoRemove the fake RLE compression method.
Rich Salz [Sat, 2 May 2015 15:19:06 +0000 (11:19 -0400)]
Remove the fake RLE compression method.

RLE is a no-op only for testing.  Remove it.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoGH271: Warning on </dev/null to CA.pl
Rich Salz [Sun, 3 May 2015 12:45:27 +0000 (08:45 -0400)]
GH271: Warning on </dev/null to CA.pl

If CA.pl is reading from /dev/null, then "chop $FILE" gives a warning.
Sigh.  Have to add "if $FILE".  This just silences a build warning.
Thanks to GitHub user andrejs-igumenovs for help with this.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoUse safer sizeof variant in malloc
Rich Salz [Sat, 2 May 2015 03:10:31 +0000 (23:10 -0400)]
Use safer sizeof variant in malloc

For a local variable:
        TYPE *p;
Allocations like this are "risky":
        p = OPENSSL_malloc(sizeof(TYPE));
if the type of p changes, and the malloc call isn't updated, you
could get memory corruption.  Instead do this:
        p = OPENSSL_malloc(sizeof(*p));
Also fixed a few memset() calls that I noticed while doing this.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoRT2943: Check sizes if -iv and -K arguments
Richard Levitte [Mon, 4 May 2015 15:34:40 +0000 (17:34 +0200)]
RT2943: Check sizes if -iv and -K arguments

RT2943 only complains about the incorrect check of -K argument size,
we might as well do the same thing with the -iv argument.

Before this, we only checked that the given argument wouldn't give a
bitstring larger than EVP_MAX_KEY_LENGTH.  we can be more precise and
check against the size of the actual cipher used.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoHave -K actually take an argument, and correct help text
Richard Levitte [Mon, 4 May 2015 15:33:34 +0000 (17:33 +0200)]
Have -K actually take an argument, and correct help text

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoFix cut/paste error
Rich Salz [Mon, 4 May 2015 14:34:51 +0000 (10:34 -0400)]
Fix cut/paste error

Was memset with wrong sizeof.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoRT1369: don't do "helpful" access check.
Rich Salz [Mon, 4 May 2015 11:52:27 +0000 (07:52 -0400)]
RT1369: don't do "helpful" access check.

Don't do access check on destination directory; it breaks when euid/egid
is different from real uid/gid.

Reviewed-by: Richard Levitte <levitte@openssl.org>
Signed-off-by: Rich Salz <rsalz@akamai.com>
9 years agoReturn an error in ASN1_TYPE_unpack_sequence if argument is NULL
Dr. Stephen Henson [Sun, 3 May 2015 23:59:48 +0000 (00:59 +0100)]
Return an error in ASN1_TYPE_unpack_sequence if argument is NULL

Thanks to Brian Carpenter for reporting this issue.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agomore OSSL_NELEM cases
Dr. Stephen Henson [Sun, 3 May 2015 16:32:05 +0000 (17:32 +0100)]
more OSSL_NELEM cases

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agossl/s3_srvr.c: Fix typo introduced via 69f682374868b.
mancha security [Sat, 2 May 2015 22:34:46 +0000 (22:34 +0000)]
ssl/s3_srvr.c: Fix typo introduced via 69f682374868b.

Incorrect name used for SSL_AD_INTERNAL_ERROR.

Signed-off-by: mancha security <mancha1@zoho.com>
Signed-off-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoFixes some typos in doc/ssl/
Alok Menghrajani [Mon, 13 Apr 2015 16:48:06 +0000 (09:48 -0700)]
Fixes some typos in doc/ssl/

This is the last of Alok's PR260
Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoFixes some typos in doc/apps/
Alok Menghrajani [Mon, 13 Apr 2015 16:29:52 +0000 (09:29 -0700)]
Fixes some typos in doc/apps/

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoRT3802: Fixes typos in doc/crypto/
Alok Menghrajani [Mon, 13 Apr 2015 18:05:13 +0000 (11:05 -0700)]
RT3802: Fixes typos in doc/crypto/

Signed-off-by: Rich Salz <rsalz@akamai.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agofix various typo's
Rich Salz [Sat, 2 May 2015 22:42:29 +0000 (18:42 -0400)]
fix various typo's

 https://github.com/openssl/openssl/pull/176 (CHANGES)
 https://rt.openssl.org/Ticket/Display.html?id=3545 (objects.txt)
 https://rt.openssl.org/Ticket/Display.html?id=3796 (verify.pod)

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoAdd OSSL_NELEM macro.
Dr. Stephen Henson [Sat, 2 May 2015 17:30:00 +0000 (18:30 +0100)]
Add OSSL_NELEM macro.

Add OSSL_NELEM macro to e_os.h to determine the number of elements in an
array.

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoRemove obsolete make variables
Richard Levitte [Sat, 2 May 2015 12:42:21 +0000 (14:42 +0200)]
Remove obsolete make variables

ONEDIRS, EDIRS and WDIRS aren't used anywhere.  Most probably remains
from a build system of the past, it's time they get put to rest.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoClean all .o files.
Ben Laurie [Fri, 1 May 2015 16:48:59 +0000 (17:48 +0100)]
Clean all .o files.

Reviewed-by: Rich Salz
9 years agomake X509_VERIFY_PARAM opaque
Dr. Stephen Henson [Sat, 2 May 2015 12:21:34 +0000 (13:21 +0100)]
make X509_VERIFY_PARAM opaque

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoRemove shlib/
Richard Levitte [Sat, 2 May 2015 05:47:35 +0000 (07:47 +0200)]
Remove shlib/

Cleanup of files that haven't been used, touched, or I suspect, even
noticed for a long time.

Reviewed-by: Rich Salz <rsalz@openssl.org>
9 years agoRT3820: Don't call GetDesktopWindow()
Gilles Khouzam [Sat, 2 May 2015 02:20:42 +0000 (22:20 -0400)]
RT3820: Don't call GetDesktopWindow()

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoRT3776: Wrong size for malloc
Rich Salz [Sat, 2 May 2015 02:06:07 +0000 (22:06 -0400)]
RT3776: Wrong size for malloc

Use sizeof *foo parameter, to avoid these errors.

Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
9 years agoFix uninitialized variable.
Hanno Böck [Sat, 2 May 2015 02:27:20 +0000 (22:27 -0400)]
Fix uninitialized variable.

Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Viktor Dukhovni <viktor@openssl.org>
9 years agoRemove outdated RC4 files
Rich Salz [Sat, 2 May 2015 11:23:58 +0000 (07:23 -0400)]
Remove outdated RC4 files

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agofree NULL cleanup -- coda
Rich Salz [Fri, 1 May 2015 18:37:16 +0000 (14:37 -0400)]
free NULL cleanup -- coda

After the finale, the "real" final part. :)  Do a recursive grep with
"-B1 -w [a-zA-Z0-9_]*_free" to see if any of the preceeding lines are
an "if NULL" check that can be removed.

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoRemove goto inside an if(0) block
Rich Salz [Fri, 1 May 2015 18:29:48 +0000 (14:29 -0400)]
Remove goto inside an if(0) block

There were a dozen-plus instances of this construct:
   if (0) { label: ..... }

Reviewed-by: Tim Hudson <tjh@openssl.org>
9 years agoFix build on MacOS.
Ben Laurie [Sun, 19 Apr 2015 13:10:54 +0000 (14:10 +0100)]
Fix build on MacOS.

Reviewed-by: Andy Polyakov
9 years agoUse BN_ULONG format.
Ben Laurie [Sun, 19 Apr 2015 13:34:05 +0000 (14:34 +0100)]
Use BN_ULONG format.

Reviewed-by: Andy Polyakov
9 years agou_len may be unused.
Ben Laurie [Sun, 19 Apr 2015 13:30:56 +0000 (14:30 +0100)]
u_len may be unused.

Reviewed-by: Andy Polyakov
9 years agofree NULL cleanup 11
Rich Salz [Fri, 1 May 2015 14:15:18 +0000 (10:15 -0400)]
free NULL cleanup 11

Don't check for NULL before calling free functions. This gets:
        ERR_STATE_free
        ENGINE_free
        DSO_free
        CMAC_CTX_free
        COMP_CTX_free
        CONF_free
        NCONF_free NCONF_free_data _CONF_free_data
        A sk_free use within OBJ_sigid_free
        TS_TST_INFO_free (rest of TS_ API was okay)
        Doc update for UI_free (all uses were fine)
        X509V3_conf_free
        X509V3_section_free
        X509V3_string_free

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agofree null cleanup finale
Rich Salz [Fri, 1 May 2015 14:02:07 +0000 (10:02 -0400)]
free null cleanup finale

Don't check for NULL before calling OPENSSL_free

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoFix some typo's, silence warnings.
Rich Salz [Fri, 1 May 2015 11:11:17 +0000 (07:11 -0400)]
Fix some typo's, silence warnings.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoRewrite CA.pl.in
Rich Salz [Fri, 1 May 2015 01:44:40 +0000 (21:44 -0400)]
Rewrite CA.pl.in

Reformat CA.pl.in to follow coding style.
Also add "use strict" and "use warnings"
Also modify it to exit properly and report only when succeeded.
And some perl tweaks via Richard.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agofree NULL cleanup 7
Rich Salz [Fri, 1 May 2015 01:37:06 +0000 (21:37 -0400)]
free NULL cleanup 7

This gets BN_.*free:
    BN_BLINDING_free BN_CTX_free BN_FLG_FREE BN_GENCB_free
    BN_MONT_CTX_free BN_RECP_CTX_free BN_clear_free BN_free BUF_MEM_free

Also fix a call to DSA_SIG_free to ccgost engine and remove some #ifdef'd
dead code in engines/e_ubsec.

Reviewed-by: Richard Levitte <levitte@openssl.org>
9 years agoFix buffer overrun in RSA signing
Matt Caswell [Wed, 29 Apr 2015 12:22:18 +0000 (13:22 +0100)]
Fix buffer overrun in RSA signing

The problem occurs in EVP_PKEY_sign() when using RSA with X931 padding.
It is only triggered if the RSA key size is smaller than the digest length.
So with SHA512 you can trigger the overflow with anything less than an RSA
512 bit key. I managed to trigger a 62 byte overflow when using a 16 bit RSA
key. This wasn't sufficient to cause a crash, although your mileage may
vary.

In practice RSA keys of this length are never used and X931 padding is very
rare. Even if someone did use an excessively short RSA key, the chances of
them combining that with a longer digest and X931 padding is very
small. For these reasons I do not believe there is a security implication to
this. Thanks to Kevin Wojtysiak (Int3 Solutions) and Paramjot Oberoi (Int3
Solutions) for reporting this issue.

Reviewed-by: Andy Polyakov <appro@openssl.org>