error strings and a hash table storing per-thread error state) go via an
ERR_FNS function table. The first time an ERR operation occurs, the
implementation that will be used (from then on) is set to the internal
"defaults" implementation if it has not already been set. The actual LHASH
tables are only accessed by this implementation.
This is primarily for modules that can be loaded at run-time and bound into
an application (or a shared-library version of OpenSSL). If the module has
its own statically-linked copy of OpenSSL code - this mechanism allows it
to *not* create and use ERR information in its own linked "ERR" code, but
instead to use and interact with the state stored in the loader
(application or shared library). The loader calls ERR_get_implementation()
and the return value is what the module should use when calling its own
copy of ERR_set_implementation().
dependant code has to directly increment the "references" value of each
such structure using the corresponding lock. Apart from code duplication,
this provided no "REF_CHECK/REF_PRINT" checking and violated
encapsulation.
setting stack (actually, array) values in ex_data. So only increment the
global counters if the underlying CRYPTO_get_ex_new_index() call succeeds.
This change doesn't make "ex_data" right (see the comment at the head of
ex_data.c to know why), but at least makes the source code marginally less
frustrating.
setting stack (actually, array) values in ex_data. So only increment the
global counters if the underlying CRYPTO_get_ex_new_index() call succeeds.
This change doesn't make "ex_data" right (see the comment at the head of
ex_data.c to know why), but at least makes the source code marginally less
frustrating.
His comments are:
1) Changes all references for `True64' to be `Tru64', which is the correct
spelling for the OS name.
2) Makes `alpha-cc' be the same as `alpha164-cc', and adds an `alphaold-cc'
entry that is the same as the previous `alpha-cc'. The reason is that most
people these days are using the newer compiler, so it should be the default.
3) Adds a bit of commentary to Configure, regarding the name changes of
the OS over the years, so it's not so confusing to people that haven't been
with the OS for a while.
4) Adds an `alpha-cc-rpath' target (which is *not* selected automatically
by Configure under any circumstance) that builds an RPATH into the
shared libraries. This is explained in the comment in Configure. It's
very very useful for people that want it, and people that don't want it
just shouldn't choose that target.
5) Adds the `-pthread' flag as the best way to get POSIX thread support
from the newer compiler.
6) Updates the Makefile targets, so that when the `alpha164-cc', `alpha-cc',
or `alpha-cc-rpath' target is what Configure is set to use, it uses a Makefile
target that includes the `-msym' option when building the shared library.
This is a performance enhancement.
7) Updates `config' so that if it detects you're running version 4 or 5
of the OS, it automatically selects `alpha-cc', but uses `alphaold-cc'
for versions 1-3 of the OS.
8) Updates the comment in opensslv.h, fixing both the OS name typo and
adding a reference to IRIX 6.x, since the shared library semantics are
virtually identical there.
DES's keyschedules.
I know these two should be separate, and I'll back out the DES changes if they
are deemed to be an error.
Note that there is a memory leak lurking in SSL somewhere in this version.
HP-UX in common in ./config). Note that for the moment of this writing
none of 64-bit platforms pass bntest. I'm committing this anyway as it's
too frustrating to patch snapshots over and over while 0.9.6 is known to
work.
Split private key PEM and normal PEM handling. Private key
handling needs to link in stuff like PKCS#8.
Relocate the ASN1 *_dup() functions, to the relevant ASN1
modules using new macro IMPLEMENT_ASN1_DUP_FUNCTION. Previously
these were all in crypto/x509/x_all.c along with every ASN1
BIO/fp function which linked in *every* ASN1 function if
a single dup was used.
Move the authority key id ASN1 structure to a separate file.
This is used in the X509 routines and its previous location
linked in all the v3 extension code.
Also move ASN1_tag2bit to avoid linking in a_bytes.c which
is now largely obsolete.
So far under Linux stripped binary with single PEM_read_X509
is now 238K compared to 380K before these changes.
reduce linker bloat. For example the
single line:
PEM_read_X509()
results in a binary of around 400K in Linux!
This first step separates some of the PEM functions and
avoids linking in some PKCS#7 and PKCS#12 code.
does not contain more bytes than the RSA modulus 'n' - it does not check
that the input is strictly *less* than 'n'. Whether this should be the
case or not is open to debate - however, due to security problems with
returning miscalculated CRT results, the 'rsa_mod_exp' implementation in
rsa_eay.c now performs a public-key exponentiation to verify the CRT result
and in the event of an error will instead recalculate and return a non-CRT
(more expensive) mod_exp calculation. As the mod_exp of 'I' is equivalent
to the mod_exp of 'I mod n', and the verify result is automatically between
0 and n-1 inclusive, the verify only matches the input if 'I' was less than
'n', otherwise even a correct CRT calculation is only congruent to 'I' (ie.
they differ by a multiple of 'n'). Rather than rejecting correct
calculations and doing redundant and slower ones instead, this changes the
equality check in the verification code to a congruence check.
SSL according to RFC 2712. His comment is:
This is a patch to openssl-SNAP-20010702 to support Kerberized SSL
authentication. I'm expecting to have the full kssl-0.5 kit up on
sourceforge by the end of the week. The full kit includes patches
for mod-ssl, apache, and a few text clients. The sourceforge URL
is http://sourceforge.net/projects/kssl/ .
Thanks to a note from Simon Wilkinson I've replaced my KRB5 AP_REQ
message with a real KerberosWrapper struct. I think this is fully
RFC 2712 compliant now, including support for the optional
authenticator field. I also added openssl-style ASN.1 macros for
a few Kerberos structs; see crypto/krb5/ if you're interested.
Add new extension functions which work with NCONF.
Tidy up extension config routines and remove redundant code.
Fix NCONF_get_number().
Todo: more testing of apps to see they still work...
application do that.
NOTE: there's no requirement for other UI_METHODs to avoid this kind
of loop. For example, a GUI UI_METHOD would probably check the
lengths of the answers from within instead of being constantly
redisplayed for everything that is wrong.
Implement UI controls. Current controls are the possibility to output
the OpenSSL error stack on the same channel from within UI_process()
and to check if the same user interface can be redone without being
rebuilt (this is often more a question of philosophy than
technicalities).
applications to use EVP. Add missing calls to HMAC_cleanup() and
don't assume HMAC_CTX can be copied using memcpy().
Note: this is almost identical to the patch submitted to openssl-dev
by Verdon Walker <VWalker@novell.com> except some redundant
EVP_add_digest_()/EVP_cleanup() calls were removed and some changes
made to avoid compiler warnings.
with arbitrary arguments instead of just a string.
- Change the key loaders to take a UI_METHOD instead of a callback
function pointer. NOTE: this breaks binary compatibility with
earlier versions of OpenSSL [engine].
- Addapt the nCipher code for these new conditions and add a card
insertion callback.
prompting, application-defined prompts, the possibility to use
defaults (for example default passwords from somewhere else) and
interrupts/cancelations.
In the new crypto/ui/, this was changed into tty (which is usually
/dev/tty), i.e. the FILE * used for reading passwords from the user.
However stdio buffering for read/write streams is not without pitfalls
(passwords would be echoed on some systems).
To avoid problems, split tty into tty_in and tty_out (which are
opened separately).
(for new functions...). One might still want to be able to pass down
a user-data pointer to be used by the UI. However, ex_data doesn't
quite cut it, since that means the appropriate index to it might need
to be shared between parts that aren't really related in that sense,
and would require the currently hidden (static) index holders to be
uncovered. Not a good thing. Therefore, add the possibility to add a
user-data pointer to a UI.
details (performance numbers and accompanying discussions:-). Note that
the code is not engaged in ./Configure yet. I'll add it later this week
along with updates for .spec file.
Submitted by:
Reviewed by:
PR:
passwords that were given to the key loading functions were completely
ignored, at least in the ncipher code, and then we made the assumption
that the callback wanted a prompt as user argument.
All that is now changed, and the application author is forced to give
a callback function of type pem_callback_cb and possibly an argument
for it, just as for all other functions that want to generate password
prompting.
NOTE: this change creates binary and source incompatibilities with
previous versions of OpenSSL [engine]. It's worth it this time, to
get it right (or at least better and with a chance that it'll work).
Only use trust settings if either trust or reject settings
are present, otherwise use compatibility mode. This stops
root CAs being rejected if they have alias of keyid set.
For some unknown reason fopen("con", "w") is the
only way to make this work. Using "r+" and "w+"
causes the fopen call to fail and the fallback
(using stdin) doesn't work because writing to stdin
fails.
like des_read_password and friends (backward compatibility functions
using this new API are provided). The purpose is to remove prompting
functions from the DES code section as well as provide for prompting
through dialog boxes in a window system and the like.
For those, unless the environment variables RANDFILE or HOME are
defined (the default case!), RAND_file_name() will return NULL.
This change adds a default HOME for those platforms.
To add a default HOME for any platform, just define DEFAULT_HOME in
the proper place, wrapped in appropriate #ifdef..#endif, in e_os.h.
few statements equivalent to "ENGINE_add(ENGINE_openssl())" etc. The inner
call to ENGINE_openssl() (as with other functions like it) orphans a
structural reference count. Second, the ENGINE_cleanup() function also
needs to clean up the functional reference counts held internally as the
list of "defaults" (ie. as used when RSA_new() requires an appropriate
ENGINE reference). So ENGINE_clear_defaults() was created and is called
from within ENGINE_cleanup(). Third, some of the existing code was
logically broken in its treatment of reference counts and locking (my
fault), so the necessary bits have been restructured and tidied up.
To test this stuff, compiling with ENGINE_REF_COUNT_DEBUG will cause every
reference count change (both structural and functional) to log a message to
'stderr'. Using with "openssl engine" for example shows this in action
quite well as the 'engine' sub-command cleans up after itself properly.
Also replaced some spaces with tabs.
* "ex_data" - a CRYPTO_EX_DATA structure in the ENGINE structure itself
that allows an ENGINE to store its own information there rather than in
global variables. It follows the declarations and implementations used
in RSA code, for better or worse. However there's a problem when storing
state with ENGINEs because, unlike related structure types in OpenSSL,
there is no ENGINE-vs-ENGINE_METHOD separation. Because of what ENGINE
is, it has method pointers as its structure elements ... which leads
to;
* ENGINE_FLAGS_BY_ID_COPY - if an ENGINE should not be used just as a
reference to an "implementation" (eg. to get to a hardware device), but
should also be able to maintain state, then this flag can be set by the
ENGINE implementation. The result is that any call to ENGINE_by_id()
will not result in the existing ENGINE being returned (with its
structural reference count incremented) but instead a new copy of the
ENGINE will be returned that can maintain its own state independantly of
any other copies returned in the past or future. Eg. key-generation
might involve a series of ENGINE-specific control commands to set
algorithms, sizes, module-keys, ids, ACLs, etc. A final command could
generate the key. An ENGINE doing this would *have* to declare
ENGINE_FLAGS_BY_ID_COPY so that the state of that process can be
maintained "per-handle" and unaffected by other code having a reference
to the same ENGINE structure.
This change adds some basic control commands to the existing ENGINEs
(except the software 'openssl' engine). All these engines currently load
shared-libraries for hardware APIs, so they've all been given "SO_PATH"
commands that will configure the chosen ENGINE to load its shared library
from the given path. Eg. by calling;
ENGINE_ctrl_cmd_string(e, "SO_PATH", <path>, 0).
The nCipher 'chil' ENGINE has also had "FORK_CHECK" and "THREAD_LOCKING"
commands added so these settings could be handled via application-level
configuration rather than in application source code.
Changes to "openssl engine" to test and examine these control commands will
be made shortly. It will also provide the necessary tips to application
programs wanting to support these dynamic control commands.
This change adds some new functionality to the ENGINE code and API to
make it possible for ENGINEs to describe and implement their own control
commands that can be interrogated and used by calling applications at
run-time. The source code includes numerous comments explaining how it all
works and some of the finer details. But basically, an ENGINE will normally
declare an array of ENGINE_CMD_DEFN entries in its ENGINE - and the various
new ENGINE_CTRL_*** command types take care of iterating through this list
of definitions, converting command numbers to names, command names to
numbers, getting descriptions, getting input flags, etc. These
administrative commands are handled directly in the base ENGINE code rather
than in each ENGINE's ctrl() handler, unless they specify the
ENGINE_FLAGS_MANUAL_CMD_CTRL flag (ie. if they're doing something clever or
dynamic with the command definitions).
There is also a new function, ENGINE_cmd_is_executable(), that will
determine if an ENGINE control command is of an "executable" type that
can be used in another new function, ENGINE_ctrl_cmd_string(). If not, the
control command is not supposed to be exposed out to user/config level
access - eg. it could involve the exchange of binary data, returning
results to calling code, etc etc. If the command is executable then
ENGINE_ctrl_cmd_string() can be called using a name/arg string pair. The
control command's input flags will be used to determine necessary
conversions before the control command is called, and commands of this
form will always return zero or one (failure or success, respectively).
This is set up so that arbitrary applications can support control commands
in a consistent way so that tweaking particular ENGINE behaviour is
specific to the ENGINE and the host environment, and independant of the
application or OpenSSL.
Some code demonstrating this stuff in action will applied shortly to the
various ENGINE implementations, as well as "openssl engine" support for
executing arbitrary control commands before and/or after initialising
various ENGINEs.
The existing ENGINEs (including the default 'openssl' software engine) were
static, declared inside the source file for each engine implementation. The
reason this was not going boom was that all the ENGINEs had reference
counts that never hit zero (once linked into the internal list, each would
always have at least 1 lasting structural reference).
To fix this so it will stay standing when an "unload" function is added to
match ENGINE_load_builtin_engines(), the "constructor" functions for each
ENGINE implementation have been changed to dynamically allocate and
construct their own ENGINEs using API functions. The other benefit of this
is that no ENGINE implementation has to include the internal "engine_int.h"
header file any more.
Previously RAND_get_rand_method was returning a non-const pointer, but it
should be const. As with all other such cases, METHOD pointers are stored and
returned as "const". The only methods one should be able to alter are methods
"local" to the relevant code, in which case a non-const handle to the methods
should already exist.
This change has been forced by the constifying of the ENGINE code (before
which RAND_METHOD was the only method pointer in an ENGINE structure that was
not constant).
ENGINE handler functions should take the ENGINE structure as a parameter -
this is because ENGINE structures can be copied, and like other
structure/method setups in OpenSSL, it should be possible for init(),
finish(), ctrl(), etc to adjust state inside the ENGINE structures rather
than globally. This commit includes the dependant changes in the ENGINE
implementations.
Previous changes permanently removed the commented-out old code for where
it was possible to create and use an ENGINE statically, and this code gets
rid of the ENGINE_FLAGS_MALLOCED flag that supported the distinction with
dynamically allocated ENGINEs. It also moves the area for ENGINE_FLAGS_***
values from engine_int.h to engine.h - because it should be possible to
declare ENGINEs just from declarations in exported headers.
* Constify the get/set functions, and add some that functions were missing.
* Add a new 'ENGINE_cpy()' function that will produce a new ENGINE based
copied from an original (except for the references, ie. the new copy will
be like an ENGINE returned from 'ENGINE_new()' - a structural reference).
* Removed the "null parameter" checking in the get/set functions - it is
legitimate to set NULL values as a way of *changing* an ENGINE (ie.
removing a handler that previously existed). Also, passing a NULL pointer
for an ENGINE is obviously wrong for these functions, so don't bother
checking for it. The result is a number of error codes and strings could
be removed.
