oweals/openssl.git
20 years agomake update
Richard Levitte [Tue, 25 May 2004 09:41:00 +0000 (09:41 +0000)]
make update

20 years agoMove some COMP functions to be inside the #ifndef OPENSSL_NO_COMP
Richard Levitte [Thu, 20 May 2004 23:47:57 +0000 (23:47 +0000)]
Move some COMP functions to be inside the #ifndef OPENSSL_NO_COMP
wrapping preprocessor directive.  This also removes a duplicate
declaration.

20 years agoSHA-256/-512 test and benchmark.
Andy Polyakov [Thu, 20 May 2004 21:49:38 +0000 (21:49 +0000)]
SHA-256/-512 test and benchmark.

20 years agoWhile size_t-fying let's not forget to update documentation:-)
Andy Polyakov [Thu, 20 May 2004 21:39:50 +0000 (21:39 +0000)]
While size_t-fying let's not forget to update documentation:-)

20 years agoSHA-256/-512 update. A bug fix, SHA-512 tune-up for AMD64, hook for SSE2
Andy Polyakov [Thu, 20 May 2004 21:24:41 +0000 (21:24 +0000)]
SHA-256/-512 update. A bug fix, SHA-512 tune-up for AMD64, hook for SSE2
code, Makefile update.

20 years agoStress collector/padding function.
Andy Polyakov [Thu, 20 May 2004 21:20:19 +0000 (21:20 +0000)]
Stress collector/padding function.

20 years agoFinal API adaptation. Final, "all openssl" performance numbers [not mixture
Andy Polyakov [Thu, 20 May 2004 21:18:09 +0000 (21:18 +0000)]
Final API adaptation. Final, "all openssl" performance numbers [not mixture
of different implementations]. Real-life performance improvement is rated
at 2-3x, not 6x as preliminary announced.

20 years agoDelete unused function from libeay.num, replace with one
Dr. Stephen Henson [Wed, 19 May 2004 17:08:05 +0000 (17:08 +0000)]
Delete unused function from libeay.num, replace with one
that does exist.

20 years agoDelete obsolete and unimplemented function.
Dr. Stephen Henson [Wed, 19 May 2004 17:05:02 +0000 (17:05 +0000)]
Delete obsolete and unimplemented function.

20 years agoAdd SHA256 and SHA512 algorithms to mkdef.pl.
Dr. Stephen Henson [Wed, 19 May 2004 17:03:59 +0000 (17:03 +0000)]
Add SHA256 and SHA512 algorithms to mkdef.pl.

Fix mkdef.pl script to avoid infinite loop when
parsing sha.h.

20 years agoX509_policy_lib_init is declared but not defined, so it raises havoc
Richard Levitte [Wed, 19 May 2004 14:19:51 +0000 (14:19 +0000)]
X509_policy_lib_init is declared but not defined, so it raises havoc
when trying to build a shared library on VMS or Windows...

20 years agoAfter the latest round of header-hacking, regenerate the dependencies in
Geoff Thorpe [Mon, 17 May 2004 19:26:06 +0000 (19:26 +0000)]
After the latest round of header-hacking, regenerate the dependencies in
the Makefiles. NB: this commit is probably going to generate a huge posting
and it is highly uninteresting to read.

20 years agoDeprecate the recursive includes of bn.h from various API headers (asn1.h,
Geoff Thorpe [Mon, 17 May 2004 19:14:22 +0000 (19:14 +0000)]
Deprecate the recursive includes of bn.h from various API headers (asn1.h,
dh.h, dsa.h, ec.h, ecdh.h, ecdsa.h, rsa.h), as the opaque bignum types are
already declared in ossl_typ.h. Add explicit includes for bn.h in those C
files that need access to structure internals or API functions+macros.

20 years agoRemove some unnecessary recursive includes from the internal apps.h header,
Geoff Thorpe [Mon, 17 May 2004 19:05:32 +0000 (19:05 +0000)]
Remove some unnecessary recursive includes from the internal apps.h header,
and include bn.h in those C files that need bignum functionality.

20 years agoBecause of recent reductions in header interdependencies, these files need
Geoff Thorpe [Mon, 17 May 2004 19:01:15 +0000 (19:01 +0000)]
Because of recent reductions in header interdependencies, these files need
to include crypto.h directly.

20 years agoI can't verify this directly, but recent changes will probably require that
Geoff Thorpe [Mon, 17 May 2004 18:58:47 +0000 (18:58 +0000)]
I can't verify this directly, but recent changes will probably require that
the cryptodev implementation include bn.h directly (when building with
OPENSSL_NO_DEPRECATED that is).

20 years agoThe inclusion of bn.h from the engine.h API header has been deprecated, so
Geoff Thorpe [Mon, 17 May 2004 18:56:15 +0000 (18:56 +0000)]
The inclusion of bn.h from the engine.h API header has been deprecated, so
the engine implementations need to include bn.h to manipulate bignums.

20 years agoDeprecate quite a few recursive includes from the ssl.h API header and
Geoff Thorpe [Mon, 17 May 2004 18:53:47 +0000 (18:53 +0000)]
Deprecate quite a few recursive includes from the ssl.h API header and
remove some unnecessary includes from the internal header ssl_locl.h. This
then requires adding includes for bn.h in four C files.

20 years agoDeprecate some recursive includes from the store.h API header, and put back
Geoff Thorpe [Mon, 17 May 2004 18:49:06 +0000 (18:49 +0000)]
Deprecate some recursive includes from the store.h API header, and put back
required includes back via the internal header and str_lib.c.

20 years agoReduce dependencies on crypto.h by moving the opaque definition of
Geoff Thorpe [Mon, 17 May 2004 18:39:00 +0000 (18:39 +0000)]
Reduce dependencies on crypto.h by moving the opaque definition of
CRYPTO_EX_DATA and the new/free/dup callback prototypes to ossl_typ.h.

20 years agoMoving opaque definitions to ossl_typ.h lets us reduce header dependencies.
Geoff Thorpe [Mon, 17 May 2004 18:01:28 +0000 (18:01 +0000)]
Moving opaque definitions to ossl_typ.h lets us reduce header dependencies.
Deprecate inclusion of crypto.h from ui.h.

20 years agoMake reservations for FIPS code in HEAD branch, so that the moment FIPS
Andy Polyakov [Mon, 17 May 2004 15:49:13 +0000 (15:49 +0000)]
Make reservations for FIPS code in HEAD branch, so that the moment FIPS
comes in we have required macros in place.

20 years agoMake some more API types opaquely available from ossl_typ.h, meaning the
Geoff Thorpe [Sat, 15 May 2004 18:32:08 +0000 (18:32 +0000)]
Make some more API types opaquely available from ossl_typ.h, meaning the
corresponding headers are only required for API functions or structure
details. This now includes the bignum types and BUF_MEM. Subsequent commits
will remove various dependencies on bn.h and buffer.h and update the
makefile dependencies.

20 years agoThis file implements various functions that have since been redefined as
Geoff Thorpe [Sat, 15 May 2004 18:26:15 +0000 (18:26 +0000)]
This file implements various functions that have since been redefined as
macros. I'm removing this from the NO_DEPRECATED build.

20 years agoFixes so alerts are sent properly in s3_pkt.c
Dr. Stephen Henson [Sat, 15 May 2004 17:55:07 +0000 (17:55 +0000)]
Fixes so alerts are sent properly in s3_pkt.c

PR: 851

20 years agosize_t-fication of message digest APIs. We should size_t-fy more APIs...
Andy Polyakov [Sat, 15 May 2004 11:29:55 +0000 (11:29 +0000)]
size_t-fication of message digest APIs. We should size_t-fy more APIs...

20 years agoReimplement old functions, so older software that link to libcrypto
Richard Levitte [Fri, 14 May 2004 17:56:30 +0000 (17:56 +0000)]
Reimplement old functions, so older software that link to libcrypto
don't crash and burn.

20 years agoSynchronise o_str.c between 0.9.8-dev and 0.9.7-stable.
Richard Levitte [Thu, 13 May 2004 22:39:56 +0000 (22:39 +0000)]
Synchronise o_str.c between 0.9.8-dev and 0.9.7-stable.

20 years agomake update
Richard Levitte [Thu, 13 May 2004 21:38:47 +0000 (21:38 +0000)]
make update

20 years agoLet's make life easier and have the VMS version of the configuration be
Richard Levitte [Thu, 13 May 2004 21:38:26 +0000 (21:38 +0000)]
Let's make life easier and have the VMS version of the configuration be
generated from the Unixly configuration file.

20 years agoCHANGES to mention improved PowerPC platform support.
Andy Polyakov [Thu, 13 May 2004 13:58:44 +0000 (13:58 +0000)]
CHANGES to mention improved PowerPC platform support.

20 years agoSHA-224/-256/-384/-512 implementation. This is just sheer code commit.
Andy Polyakov [Thu, 13 May 2004 13:48:33 +0000 (13:48 +0000)]
SHA-224/-256/-384/-512 implementation. This is just sheer code commit.
Makefile modifications, make test, etc. will appear later...

20 years agoMake self signing option of 'x509' use random serial numbers too.
Dr. Stephen Henson [Wed, 12 May 2004 18:20:37 +0000 (18:20 +0000)]
Make self signing option of 'x509' use random serial numbers too.

20 years agoSSE2 SHA512_Transform implementation. No, it's not used anywhere yet and
Andy Polyakov [Thu, 6 May 2004 10:41:07 +0000 (10:41 +0000)]
SSE2 SHA512_Transform implementation. No, it's not used anywhere yet and
is subject to change as C implementation is added...

20 years agoSSE2 accelerated bn_mul_add_words. Code is currently disabled till proper
Andy Polyakov [Thu, 6 May 2004 10:36:49 +0000 (10:36 +0000)]
SSE2 accelerated bn_mul_add_words. Code is currently disabled till proper
config and run-time support is added.
PR: 788
Submitted by: <dean@arctic.org>
Reviewed by: <appro>

Obtained from: http://arctic.org/~dean/crypto/rsa.html

20 years agoSupport for IA-32 SSE2 instruction set.
Andy Polyakov [Thu, 6 May 2004 10:31:09 +0000 (10:31 +0000)]
Support for IA-32 SSE2 instruction set.

20 years agoRemove the creation of $(INSTALL_PREFIX)$(OPENSSLDIR)/lib, since we don't
Richard Levitte [Thu, 6 May 2004 09:46:41 +0000 (09:46 +0000)]
Remove the creation of $(INSTALL_PREFIX)$(OPENSSLDIR)/lib, since we don't
use it.

Notified by Frédéric L. W. Meunier <0@pervalidus.tk> in PR 713

20 years agoWhen the pointer 'from' changes, it's stored length needs to change as
Richard Levitte [Thu, 6 May 2004 09:33:22 +0000 (09:33 +0000)]
When the pointer 'from' changes, it's stored length needs to change as
well.

Notified by Frank Kardel <kardel@acm.org> in PR 879.

20 years agoFix realloc usage in ec_curve.c
Geoff Thorpe [Tue, 4 May 2004 20:08:55 +0000 (20:08 +0000)]
Fix realloc usage in ec_curve.c

Submitted by: Nils Larsch
Reviewed by: Geoff Thorpe

20 years ago- update from current 0.9.6-stable CHANGES file
Bodo Möller [Tue, 4 May 2004 01:15:48 +0000 (01:15 +0000)]
- update from current 0.9.6-stable CHANGES file
- update from current 0.9.7-stable CHANGES file:

  Now here we have "CHANGES between 0.9.7e and 0.9.8", and I hope
  that all patches mentioned for 0.9.7d and 0.9.7e actually are
  in the CVS HEAD, i.e. what is to become 0.9.8.

  I have rewritten the 'openssl ca -create_serial' entry (0.9.8)
  so that it explains the earlier change that is now listed (0.9.7e).

  The ENGINE_set_default typo bug entry has been moved from 0.9.8
  to 0.9.7b, which is where it belongs.

20 years agoThe new BN_CTX code makes this sort of abuse unnecessary.
Geoff Thorpe [Wed, 28 Apr 2004 18:34:39 +0000 (18:34 +0000)]
The new BN_CTX code makes this sort of abuse unnecessary.

20 years agoOops! Typo in ./config...
Andy Polyakov [Tue, 27 Apr 2004 22:17:25 +0000 (22:17 +0000)]
Oops! Typo in ./config...

20 years agoImproved PowerPC support. Proper ./config support for ppc targets,
Andy Polyakov [Tue, 27 Apr 2004 22:05:50 +0000 (22:05 +0000)]
Improved PowerPC support. Proper ./config support for ppc targets,
especially for AIX. But most important BIGNUM assembler implementation
submitted by IBM.

Submitted by: Peter Waltenberg <pwalten@au1.ibm.com>
Reviewed by: appro

20 years agoMake ASN1 code work again...
Dr. Stephen Henson [Tue, 27 Apr 2004 18:33:40 +0000 (18:33 +0000)]
Make ASN1 code work again...

20 years agoWith the new dynamic BN_CTX implementation, there should be no need for
Geoff Thorpe [Tue, 27 Apr 2004 13:24:51 +0000 (13:24 +0000)]
With the new dynamic BN_CTX implementation, there should be no need for
additional contexts.

20 years agoThe problem of rsa key-generation getting stuck in a loop for (pointlessly)
Geoff Thorpe [Mon, 26 Apr 2004 15:38:44 +0000 (15:38 +0000)]
The problem of rsa key-generation getting stuck in a loop for (pointlessly)
small key sizes seems to result from the code continually regenerating the
same prime value once the range is small enough. From my tests, this change
fixes the problem by setting an escape velocity of 3 repeats for the second
of the two primes.

PR: 874

20 years agoAllow RSA key-generation to specify an arbitrary public exponent. Jelte
Geoff Thorpe [Mon, 26 Apr 2004 15:31:35 +0000 (15:31 +0000)]
Allow RSA key-generation to specify an arbitrary public exponent. Jelte
proposed the change and submitted the patch, I jiggled it slightly and
adjusted the other parts of openssl that were affected.

PR: 867
Submitted by: Jelte Jansen
Reviewed by: Geoff Thorpe

20 years agoMore ASN1 reformat/tidy.
Dr. Stephen Henson [Sun, 25 Apr 2004 12:46:39 +0000 (12:46 +0000)]
More ASN1 reformat/tidy.

20 years agoReformat/tidy some of the ASN1 code.
Dr. Stephen Henson [Sat, 24 Apr 2004 17:02:48 +0000 (17:02 +0000)]
Reformat/tidy some of the ASN1 code.

20 years agoFix leak.
Dr. Stephen Henson [Thu, 22 Apr 2004 12:37:16 +0000 (12:37 +0000)]
Fix leak.

PR:870

20 years agoAs far as I can tell, the bugfix this comment refers to was committed to
Geoff Thorpe [Wed, 21 Apr 2004 15:12:20 +0000 (15:12 +0000)]
As far as I can tell, the bugfix this comment refers to was committed to
0.9.7-stable as well as HEAD (and doesn't apply to the 0.9.6-engine
variant).

20 years agoExtend the index parameter checking from sk_value to sk_set(). Also tidy up
Geoff Thorpe [Wed, 21 Apr 2004 15:08:56 +0000 (15:08 +0000)]
Extend the index parameter checking from sk_value to sk_set(). Also tidy up
some similar code elsewhere.

Thanks to Francesco Petruzzi for bringing this to my attention.

20 years agoNew option to 'x509' -next_serial. This outputs the certificate
Dr. Stephen Henson [Wed, 21 Apr 2004 12:46:20 +0000 (12:46 +0000)]
New option to 'x509' -next_serial. This outputs the certificate
serial number plus 1 to the output file. Its purpose is to allow
serial number files to be initialized when random serial numbers
are used.

20 years agoUse X509_get_serialNumber() instead of accessing internals in x509.c
Dr. Stephen Henson [Wed, 21 Apr 2004 12:43:21 +0000 (12:43 +0000)]
Use X509_get_serialNumber() instead of accessing internals in x509.c

20 years agoReduce chances of issuer and serial number duplication by use of random
Dr. Stephen Henson [Tue, 20 Apr 2004 12:05:26 +0000 (12:05 +0000)]
Reduce chances of issuer and serial number duplication by use of random
initial serial numbers.

PR: 842

20 years agoWhooaaaaa, the BN_CTX_DEBUG macro really produces output these
Richard Levitte [Tue, 20 Apr 2004 11:53:33 +0000 (11:53 +0000)]
Whooaaaaa, the BN_CTX_DEBUG macro really produces output these
days...  A little too much for my tests, currently...

20 years agoPrint the debug thingies on stderr instead of stdout. If for nothing
Richard Levitte [Tue, 20 Apr 2004 10:57:07 +0000 (10:57 +0000)]
Print the debug thingies on stderr instead of stdout.  If for nothing
else then at least so bc doesn't have problems parsing the output from
bntest :-).

20 years agomake update
Geoff Thorpe [Mon, 19 Apr 2004 18:33:41 +0000 (18:33 +0000)]
make update

20 years ago"make update" noticed a new function.
Geoff Thorpe [Mon, 19 Apr 2004 18:32:19 +0000 (18:32 +0000)]
"make update" noticed a new function.

20 years agoMore updates for the header cleanups (and apologies, again, for not having
Geoff Thorpe [Mon, 19 Apr 2004 18:30:41 +0000 (18:30 +0000)]
More updates for the header cleanups (and apologies, again, for not having
consolidated these prior to committing).

20 years agoWhen generating dependencies in the makefiles, generate the reduced
Geoff Thorpe [Mon, 19 Apr 2004 18:19:24 +0000 (18:19 +0000)]
When generating dependencies in the makefiles, generate the reduced
dependencies of the OPENSSL_NO_DEPRECATED mode. This prevents dependencies
being reproduced for "deprecated" header behaviour when a developer doesn't
define the symbol (with the subsequent CVS wars that can ensue).

20 years agoheader cleanup in apps/
Geoff Thorpe [Mon, 19 Apr 2004 18:13:07 +0000 (18:13 +0000)]
header cleanup in apps/

20 years ago(oops) Apologies all, that last header-cleanup commit was from the wrong
Geoff Thorpe [Mon, 19 Apr 2004 18:09:28 +0000 (18:09 +0000)]
(oops) Apologies all, that last header-cleanup commit was from the wrong
tree. This further reduces header interdependencies, and makes some
associated cleanups.

20 years agoReduce header interdependencies, initially in engine.h (the rest of the
Geoff Thorpe [Mon, 19 Apr 2004 17:46:04 +0000 (17:46 +0000)]
Reduce header interdependencies, initially in engine.h (the rest of the
changes are the fallout). As this could break source code that doesn't
directly include headers for interfaces it uses, changes to recursive
includes are covered by the OPENSSL_NO_DEPRECATED symbol. It's better to
define this when building and using openssl, and then adapt code where
necessary - this is how to stay current. However the mechanism exists for
the lethargic.

20 years agoClear error if unique_subject lookup fails.
Dr. Stephen Henson [Thu, 15 Apr 2004 00:32:19 +0000 (00:32 +0000)]
Clear error if unique_subject lookup fails.

20 years agoAdd some root CAs.
Dr. Stephen Henson [Tue, 13 Apr 2004 17:47:37 +0000 (17:47 +0000)]
Add some root CAs.

20 years agoAvoid undefined results when the parameter is out of range.
Geoff Thorpe [Fri, 2 Apr 2004 06:25:11 +0000 (06:25 +0000)]
Avoid undefined results when the parameter is out of range.

20 years agoDon't use C++ reserved word.
Dr. Stephen Henson [Thu, 1 Apr 2004 22:23:46 +0000 (22:23 +0000)]
Don't use C++ reserved word.

20 years agoOops forgot CHANGES entry.
Dr. Stephen Henson [Wed, 31 Mar 2004 12:55:33 +0000 (12:55 +0000)]
Oops forgot CHANGES entry.

20 years agoNew function X509_POLICY_NODE_print()
Dr. Stephen Henson [Wed, 31 Mar 2004 12:17:24 +0000 (12:17 +0000)]
New function X509_POLICY_NODE_print()

20 years agoAdd symbol hacks for some long names.
Richard Levitte [Mon, 29 Mar 2004 08:13:49 +0000 (08:13 +0000)]
Add symbol hacks for some long names.
make update

20 years agoThis is essentially Intel 32-bit compiler tune-up. To start with all
Andy Polyakov [Sun, 28 Mar 2004 21:27:47 +0000 (21:27 +0000)]
This is essentially Intel 32-bit compiler tune-up. To start with all
available compiler versions generated bogus machine code trying to
compile new crypto/des/cfb_enc.c. Secondly, 8th version defines
__GNUC__ macro, but fails to compile *some* inline assembler correctly.
Note that all versions of icc implement MSC-like _lrot[rl] intrinsic,
which is used now instead of offensive asm. Finally, unnecessary linker
dependencies are eliminated. Most notably dependency from libirc.a
caused trouble at application start-up, if libcrypto.so is linked with
-Bsymbolic (which it is).

20 years agoEnhance EVP code to generate random symmetric keys of the
Dr. Stephen Henson [Sun, 28 Mar 2004 17:38:00 +0000 (17:38 +0000)]
Enhance EVP code to generate random symmetric keys of the
appropriate form, for example correct DES parity.

Update S/MIME code and EVP_SealInit to use new functions.

PR: 700

20 years agoMake {i2v,v2i}_ASN1_BIT_STRING global.
Dr. Stephen Henson [Sun, 28 Mar 2004 12:40:11 +0000 (12:40 +0000)]
Make {i2v,v2i}_ASN1_BIT_STRING global.

make update

20 years agoRemove obsolete files.
Dr. Stephen Henson [Sun, 28 Mar 2004 12:29:05 +0000 (12:29 +0000)]
Remove obsolete files.

20 years agoAllow CRLs to be passed into X509_STORE_CTX. This is useful when the
Dr. Stephen Henson [Sat, 27 Mar 2004 22:49:28 +0000 (22:49 +0000)]
Allow CRLs to be passed into X509_STORE_CTX. This is useful when the
verified structure can contain its own CRLs (such as PKCS#7 signedData).

Tidy up some of the verify code.

20 years agoExtend OID config module format.
Dr. Stephen Henson [Sat, 27 Mar 2004 13:30:14 +0000 (13:30 +0000)]
Extend OID config module format.

20 years agoFree up BIO properly when using streaming S/MIME sign.
Dr. Stephen Henson [Fri, 26 Mar 2004 00:24:38 +0000 (00:24 +0000)]
Free up BIO properly when using streaming S/MIME sign.

20 years agoRemove BN_CTX debug from debug-steve
Dr. Stephen Henson [Thu, 25 Mar 2004 23:32:06 +0000 (23:32 +0000)]
Remove BN_CTX debug from debug-steve

20 years agoSSL_COMP_get_compression_method is a typo (a missing 's' at the end of
Richard Levitte [Thu, 25 Mar 2004 21:32:30 +0000 (21:32 +0000)]
SSL_COMP_get_compression_method is a typo (a missing 's' at the end of
the symbol name).

20 years agoMove the definition of Win32_rename(), since the macro rename gets undefined
Richard Levitte [Thu, 25 Mar 2004 20:09:00 +0000 (20:09 +0000)]
Move the definition of Win32_rename(), since the macro rename gets undefined
in the middle of the code on Windows, and that disrupts operations in functions
later that use rename()...
PR: 853

20 years agoWrap code starting with a definition.
Richard Levitte [Thu, 25 Mar 2004 20:01:01 +0000 (20:01 +0000)]
Wrap code starting with a definition.
PR: 854

20 years agoChange spaces to symbols in names.
Richard Levitte [Thu, 25 Mar 2004 19:52:34 +0000 (19:52 +0000)]
Change spaces to symbols in names.
PR: 856

20 years agoMake prototypes for some callback pointers.
Richard Levitte [Thu, 25 Mar 2004 16:21:42 +0000 (16:21 +0000)]
Make prototypes for some callback pointers.

20 years agoA couple more cases where RAND_add() gets an integer instead of a
Richard Levitte [Thu, 25 Mar 2004 16:04:02 +0000 (16:04 +0000)]
A couple more cases where RAND_add() gets an integer instead of a
doule as last argument.

20 years agoRAND_add() wants a double as it's last argument.
Richard Levitte [Thu, 25 Mar 2004 15:52:43 +0000 (15:52 +0000)]
RAND_add() wants a double as it's last argument.

20 years agoFix loads of warnings in policy code.
Dr. Stephen Henson [Thu, 25 Mar 2004 13:45:58 +0000 (13:45 +0000)]
Fix loads of warnings in policy code.

I'll remember to try to compile this with warnings enabled next time :-)

20 years agoFix ASN1 warnings.
Dr. Stephen Henson [Thu, 25 Mar 2004 13:37:02 +0000 (13:37 +0000)]
Fix ASN1 warnings.

20 years agoAdjust various bignum functions to use BN_CTX for variables instead of
Geoff Thorpe [Thu, 25 Mar 2004 04:32:24 +0000 (04:32 +0000)]
Adjust various bignum functions to use BN_CTX for variables instead of
locally initialising their own.

NB: I've removed the "BN_clear_free()" loops for the exit-paths in some of
these functions, and that may be a major part of the performance
improvements we're seeing. The "free" part can be removed because we're
using BN_CTX. The "clear" part OTOH can be removed because BN_CTX
destruction automatically performs this task, so performing it inside
functions that may be called repeatedly is wasteful. This is currently safe
within openssl due to the fact that BN_CTX objects are never created for
longer than a single high-level operation. However, that is only because
there's currently no mechanism in openssl for thread-local storage. Beyond
that, this might be an issue for applications using the bignum API directly
and caching their own BN_CTX objects. The solution is to introduce a flag
to BN_CTX_start() that allows its variables to be automatically sanitised
on release during BN_CTX_end(). This way any higher-level function (and
perhaps the application) can specify this flag in its own
BN_CTX_start()/BN_CTX_end() pair, and this will cause inner-loop functions
specifying the flag to be ignored so that sanitisation is handled only once
back out at the higher level. I will be implementing this in the near
future.

20 years agoReplace the BN_CTX implementation with my current work. I'm leaving the
Geoff Thorpe [Thu, 25 Mar 2004 04:16:14 +0000 (04:16 +0000)]
Replace the BN_CTX implementation with my current work. I'm leaving the
little TODO list in there as well as the debugging code (only enabled if
BN_CTX_DEBUG is defined).

I'd appreciate as much review and testing as can be spared for this. I'll
commit some changes to other parts of the bignum code shortly to make
better use of this implementation (no more fixed size limitations). Note
also that under identical optimisations, I'm seeing a noticable speed
increase over openssl-0.9.7 - so any feedback to confirm/deny this on other
systems would also be most welcome.

20 years agoAdds warnings about two curves and fixes the "seed" value for two other
Geoff Thorpe [Thu, 25 Mar 2004 03:03:52 +0000 (03:03 +0000)]
Adds warnings about two curves and fixes the "seed" value for two other
curves.

Submitted by: Nils Larsch

20 years ago... and this should likewise fix up those RSA implementations that weren't
Geoff Thorpe [Thu, 25 Mar 2004 02:55:17 +0000 (02:55 +0000)]
... and this should likewise fix up those RSA implementations that weren't
already built and tested.

20 years agoBy adding a BN_CTX parameter to the 'rsa_mod_exp' callback, private key
Geoff Thorpe [Thu, 25 Mar 2004 02:52:04 +0000 (02:52 +0000)]
By adding a BN_CTX parameter to the 'rsa_mod_exp' callback, private key
operations no longer require two distinct BN_CTX structures. This may put
more "strain" on the current BN_CTX implementation (which has a fixed limit
to the number of variables it will hold), but so far this limit is not
triggered by any of the tests pass and I will be changing BN_CTX in the
near future to avoid this problem anyway.

This also changes the default RSA implementation code to use the BN_CTX in
favour of initialising some of its variables locally in each function.

20 years agoDamn, I was a bit hasty with my fix and hadn't spotted the linker
Geoff Thorpe [Thu, 25 Mar 2004 02:41:35 +0000 (02:41 +0000)]
Damn, I was a bit hasty with my fix and hadn't spotted the linker
dependency from asn1.

20 years agoRemove some warnings.
Geoff Thorpe [Thu, 25 Mar 2004 02:24:38 +0000 (02:24 +0000)]
Remove some warnings.

20 years agoProtect against gcc's "warning: cast does not match function type".
Geoff Thorpe [Thu, 25 Mar 2004 02:19:42 +0000 (02:19 +0000)]
Protect against gcc's "warning: cast does not match function type".

20 years agoDon't define fd for platforms that do not use it, as some may not declare fileno...
Richard Levitte [Wed, 24 Mar 2004 10:55:48 +0000 (10:55 +0000)]
Don't define fd for platforms that do not use it, as some may not declare fileno() properly

20 years agoCorrect constness problems.
Richard Levitte [Wed, 24 Mar 2004 10:50:42 +0000 (10:50 +0000)]
Correct constness problems.

20 years agoMake it easier to buld test applications...
Richard Levitte [Wed, 24 Mar 2004 10:50:25 +0000 (10:50 +0000)]
Make it easier to buld test applications...

20 years agoOnly build the PKCS#7 test applications if "pkcs7" is present in
Richard Levitte [Wed, 24 Mar 2004 10:48:50 +0000 (10:48 +0000)]
Only build the PKCS#7 test applications if "pkcs7" is present in
SDIRS.