Now that ossl_bsearch is capable of using a thunking function, lets
create a thunking function to use for the STACK_OF macros.
The problem we're addressing is one that gives rise to ubsan issues.
clang-16 forward have a ubsan test that confirms that the target symbol
that we call through a pointer matches the type of the pointer itself.
for instance
int foo(void *a, void *b)
{
...
}
int (*fooptr)(char *ac, int *bc) = foo;
fooptr(&charval, &intval);
is strictly speaking in C undefined behavior (even though in normal
operation this works as expected). Newer compilers are strict about
this however, as several security frameworks operate with an expectation
that this constraint is met.
See https://github.com/openssl/openssl/issues/22896#issuecomment-1837266357
for details.
So we need to create a thunking function. The sole purpose of this
thunking function is to accept the "real" comparison function for the
STACK_OF macros, along with the two items to compare of the type that
they are passed as from the calling function, and do the convervsion of
both the comparison function and the data pointers to the types that the
real comparison function expects
So we:
1) Modify the DEFINE_STACK_OF macros to create this thunking function
2) Add an OPENSSL_sk_set_cmp_thunks api to set the comparison function
3) modify the requisite places in the stack code to use the thunking
function when available
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Nikola Pajkovsky <nikolap@openssl.org>
MergeDate: Sat Feb 7 18:11:14 2026
(Merged from https://github.com/openssl/openssl/pull/29640)
Add the initial groundwork to allow for the use of a thunking function
with bsearch. Normally our comparison function signature doesn't match
the type of the pointer we call it through, leading to ubsan errors,
this lets those signatures match and gives us a place to do the proper
casting
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Nikola Pajkovsky <nikolap@openssl.org>
MergeDate: Sat Feb 7 18:11:11 2026
(Merged from https://github.com/openssl/openssl/pull/29640)
If we have a comparison function, and the array was sorted,
check to see if we are inserting in the correct location.
if so do not clear is_sorted.
This allows for element locations found with OPENSSL_sk_find_ex
to be used to insert elements in the correct location and preserve
the sorting order without the need to sort the stack again.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/28533)
Matt noticed "It's kind of weird that we are forced to call sort on
a newly created and empty stack. It feels like an empty stack should
have the "sorted" flag by default on creation"
I am incluined to agree. This change ensures tht empty or 1 element
stacks are marked as sorted, as per the existing comment in the
file.
Since this involved changing the various duplication routines to
also ensure that sorted was preserved for such stacks, I also
noticed the duplication code was largely duplicated. I
took the opportunity to deduplicate the duplication code while
making these changes.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Paul Dale <ppzgs1@gmail.com>
(Merged from https://github.com/openssl/openssl/pull/28509)
Most of the callers do not actually check for
the special -1 return condition because they do not
pass NULL to it. It is also extremely improbable that
any code depends on this -1 return value in this condition
so it can be safely changed to 0 return.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/22930)
Since OPENSSL_malloc() and friends report ERR_R_MALLOC_FAILURE, and
at least handle the file name and line number they are called from,
there's no need to report ERR_R_MALLOC_FAILURE where they are called
directly, or when SSLfatal() and RLAYERfatal() is used, the reason
`ERR_R_MALLOC_FAILURE` is changed to `ERR_R_CRYPTO_LIB`.
There were a number of places where `ERR_R_MALLOC_FAILURE` was reported
even though it was a function from a different sub-system that was
called. Those places are changed to report ERR_R_{lib}_LIB, where
{lib} is the name of that sub-system.
Some of them are tricky to get right, as we have a lot of functions
that belong in the ASN1 sub-system, and all the `sk_` calls or from
the CRYPTO sub-system.
Some extra adaptation was necessary where there were custom OPENSSL_malloc()
wrappers, and some bugs are fixed alongside these changes.
Reviewed-by: Tomas Mraz <tomas@openssl.org>
Reviewed-by: Hugo Landau <hlandau@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/19301)
It is possible for the stack of X509_OBJECTs held in an X509_STORE_CTX to
have a custom compare function associated with it. Normally (by default)
this uses X509_NAME_cmp(). The X509_STORE_CTX_get1_issuer() function
assumed that it would always be X509_NAME_cmp().
By implementing OPENSSL_sk_find_all() function we can avoid explicitly
using X509_NAME_cmp() in X509_STORE_CTX_get1_issuer().
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/14728)
We put almost everything in these internal static libraries:
libcommon Block building code that can be used by all
our implementations, legacy and non-legacy
alike.
libimplementations All non-legacy algorithm implementations and
only them. All the code that ends up here is
agnostic to the definitions of FIPS_MODE.
liblegacy All legacy implementations.
libnonfips Support code for the algorithm implementations.
Built with FIPS_MODE undefined. Any code that
checks that FIPS_MODE isn't defined must end
up in this library.
libfips Support code for the algorithm implementations.
Built with FIPS_MODE defined. Any code that
checks that FIPS_MODE is defined must end up
in this library.
The FIPS provider module is built from providers/fips/*.c and linked
with libimplementations, libcommon and libfips.
The Legacy provider module is built from providers/legacy/*.c and
linked with liblegacy, libcommon and libcrypto.
If module building is disabled, the object files from liblegacy and
libcommon are added to libcrypto and the Legacy provider becomes a
built-in provider.
The Default provider module is built-in, so it ends up being linked
with libimplementations, libcommon and libnonfips. For libcrypto in
form of static library, the object files from those other libraries
are simply being added to libcrypto.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/10088)
OBJ_bsearch_ and OBJ_bsearch_ex_ are generic functions that don't
really belong with the OBJ API, but should rather be generic utility
functions. The ending underscore indicates that they are considered
internal, even though they are declared publicly.
Since crypto/stack/stack.c uses OBJ_bsearch_ex_, the stack API ends up
depending on the OBJ API, which is unnecessary, and carries along
other dependencies.
Therefor, a generic internal function is created, ossl_bsearch().
This removes the unecessary dependencies.
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/8899)
This reverts commit 8839324450.
Removing these checks changes the behaviour of the API which is not
appropriate for a minor release. This also fixes a failure in the
fuzz tests when building with no-comp.
Reviewed-by: Tim Hudson <tjh@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6895)
Checks are left in OPENSSL_sk_shift, OPENSSL_sk_pop and OPENSSL_sk_num.
This is because these are used as "opportunistic" readers, pulling
whatever datai, if any, set by somebody else. All calls that add data
don't check for stack being NULL, because caller should have checked
if stack was actually created.
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/6860)
This allows the caller to guarantee that there is sufficient space for a
number of insertions without reallocation.
The expansion ratio when reallocating the array is reduced to 1.5 rather than 2.
Change bounds testing to use a single size rather than both INT_MAX and
SIZE_MAX. This simplifies some of the tests.
Switch the stack pointers to data from char * to void *
Reviewed-by: Andy Polyakov <appro@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4386)
If OPENSSL_sk_insert() calls OPENSSL_realloc() and it fails, it was leaking
the originally allocated memory.
Reviewed-by: Rich Salz <rsalz@openssl.org>
We were casting num_alloc to size_t in lots of places, or just using it in
a context where size_t makes more sense - so convert it. This simplifies
the code a bit.
Also tweak the style in stack.c a bit following on from the previous
commit
Reviewed-by: Rich Salz <rsalz@openssl.org>
Rename sk_xxx to OPENSSL_sk_xxx and _STACK to OPENSSL_STACK
Rename lh_xxx API to OPENSSL_LH_xxx and LHASH_NODE to OPENSSL_LH_NODE
Make lhash stuff opaque.
Use typedefs for function pointers; makes the code simpler.
Remove CHECKED_xxx macros.
Add documentation; remove old X509-oriented doc.
Add API-compat names for entire old API
Reviewed-by: Dr. Stephen Henson <steve@openssl.org>
The Unix build was the last to retain the classic build scheme. The
new unified scheme has matured enough, even though some details may
need polishing.
Reviewed-by: Rich Salz <rsalz@openssl.org>
openssl-machine
Neil Horman
Bob Beck
Eugene Syromiatnikov
Frederik Wedel-Heinen
Richard Levitte
Tomas Mraz
Matt Caswell
Pauli
Dr. David von Oheimb
Andy Polyakov
Rich Salz
Richard Levitte
Paul Yang
FdaSilvaYY
Pauli
Guido Vranken
Dr. Stephen Henson