Conceptually, this is a squashed version of:
Revert "Address feedback"
This reverts commit 75551e07bd.
and
Revert "Add CRYPTO_thread_glock_new"
This reverts commit ed6b2c7938.
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)
Breaks djgpp, masks a common kernel function name.
Thanks to Gisle Vanem for pointing this out.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3776)
Various initialization functions modify this table, which can cause heap
corruption in the absence of external synchronization.
Some stats are modified from OPENSSL_LH_retrieve, where callers aren't
expecting to have to take out an exclusive lock. Switch to using atomic
operations for those stats.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3525)
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>
We are using strcmp() as the cmp_func, where in the EDK2 environment
strcmp actually ends up being the external AsciiStrCmp() function —
an EFI library function defined with the Microsoft ABI.
This means that we can't just assign function pointers to it, since
in GCC-hosted builds the ABI of any function *not* explicitly marked
EFIAPI is the native SysV ABI.
Arguably this stupidity ought to be resolved on the UEFI side, but in
the general case that would mean that we need to provide ABI-compatible
wrappers for *all* the "standard" functions, just in case they're used
like this.
And in fact we already have a workaround here for DEC C. So instead of
playing games with casting function pointers, it's nicer just to use a
simple function to wrap the strcmp() call. That cleans up the DEC C
workaround, *and* it works around the UEFI bogosity at the same time.
Signed-off-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Kurt Roeckx <kurt@openssl.org>
Make LHASH_OF use static inline functions.
Add new lh_get_down_load and lh_set_down_load functions and their
typesafe inline equivalents.
Make lh_error a function instead of a macro.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Only two macros CRYPTO_MDEBUG and CRYPTO_MDEBUG_ABORT to control this.
If CRYPTO_MDEBUG is not set, #ifdef out the whole debug machinery.
(Thanks to Jakob Bohm for the suggestion!)
Make the "change wrapper functions" be the only paradigm.
Wrote documentation!
Format the 'set func' functions so their paramlists are legible.
Format some multi-line comments.
Remove ability to get/set the "memory debug" functions at runtme.
Remove MemCheck_* and CRYPTO_malloc_debug_init macros.
Add CRYPTO_mem_debug(int flag) function.
Add test/memleaktest.
Rename CRYPTO_malloc_init to OPENSSL_malloc_init; remove needless calls.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Applications wishing to include their own stacks now just need to include
DEFINE_STACK_OF(foo)
in a header file.
Reviewed-by: Richard Levitte <levitte@openssl.org>
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>
Miscellaneous unchecked malloc fixes. Also fixed some mem leaks on error
paths as I spotted them along the way.
Reviewed-by: Tim Hudson <tjh@openssl.org>
The mkstack.pl script now generates the entire safestack.h file.
It generates output that follows the coding style.
Also, removed all instances of the obsolete IMPLEMENT_STACK_OF
macro.
Reviewed-by: Andy Polyakov <appro@openssl.org>
knock-on work than expected - they've been extracted into a patch
series that can be completed elsewhere, or in a different branch,
before merging back to HEAD.
example) are declared with some extra linkage information. This
generates a warning when using the function name as a value to a
regular function pointer with the "correct" definition of the
function. Therefore, use a macro to cast the appropriate function on
VMS.
callbacks, and their prototypes were consistent as they were. These casts
need reversing.
Also, I personally find line breaks during parameter lists (ie a line
ending in a comma) easier to read at a glance than line breaks at the end
of a function call and before a dereference on the return value (ie a line
ending in a closed-bracket followed by a line starting with "->").
"doall" functions to using type-safe wrappers. As and where required, this
can be replaced by redeclaring the underlying callbacks to use the
underlying "void"-based prototypes (eg. if performance suffers from an
extra level of function invocation).
LHASH code, this evil was uncovered. The cast was obscuring the fact that
the function was prototyped to take 2 parameters when in fact it is being
used as a callback that should take only one. Anyway, the function itself
ignores the second parameter (thankfully). A proper cure is on the way but
for now this corrects the inconsistency.
Win32 but it is getting there...
Update mkdef.pl to handle ASN1_ANY and fix headers.
Stop various VC++ warnings.
Include some fixes from "Peter 'Luna' Runestig"
<peter@runestig.com>
Remove external declaration for des_set_weak_key_flag:
it doesn't exist.
functions need to be constified, and therefore meant a number of easy
changes a little everywhere.
Now, if someone could explain to me why OBJ_dup() cheats...
DECLARE/IMPLEMENT macros now exist to create type (and prototype) safe
wrapper functions that avoid the use of function pointer casting yet retain
type-safety for type-specific callbacks. However, most of the usage within
OpenSSL itself doesn't really require the extra function because the hash
and compare callbacks are internal functions declared only for use by the
hash table. So this change catches all those cases and reimplements the
functions using the base-level LHASH prototypes and does per-variable
casting inside those functions to convert to the appropriate item type.
The exception so far is in ssl_lib.c where the hash and compare callbacks
are not static - they're exposed in ssl.h so their prototypes should not be
changed. In this last case, the IMPLEMENT_LHASH_*** macros have been left
intact.
casts) used in the lhash code are about as horrible and evil as they can
be. For starters, the callback prototypes contain empty parameter lists.
Yuck.
This first change defines clearer prototypes - including "typedef"'d
function pointer types to use as "hash" and "compare" callbacks, as well as
the callbacks passed to the lh_doall and lh_doall_arg iteration functions.
Now at least more explicit (and clear) casting is required in all of the
dependant code - and that should be included in this commit.
The next step will be to hunt down and obliterate some of the function
pointer casting being used when it's not necessary - a particularly evil
variant exists in the implementation of lh_doall.
This is mostly a work around for the old VC++ problem
that it treats func() as func(void).
Various prototypes had been added to 'compare' function
pointers that triggered this. This could be fixed by removing
the prototype, adding function pointer casts to every call or
changing the passed function to use the expected arguments.
I mostly did the latter.
The mkdef.pl script was modified to remove the typesafe
functions which no longer exist.
Oh and some functions called OPENSSL_freeLibrary() were
changed back to FreeLibrary(), wonder how that happened :-)
After some messing around this seems to work but needs
a few more tests. Working out the syntax for sk_set_cmp_func()
(cast it to a function that itself returns a function pointer)
was painful :-(
Needs some testing to see what other compilers think of this
syntax.
Also needs similar stuff for ASN1_SET_OF etc etc.
like Malloc, Realloc and especially Free conflict with already existing names
on some operating systems or other packages. That is reason enough to change
the names of the OpenSSL memory allocation macros to something that has a
better chance of being unique, like prepending them with OPENSSL_.
This change includes all the name changes needed throughout all C files.
