Due to the dynamic allocation that was added to rand_pool_add_begin
this function could now return a null pointer where it was previously
guaranteed to succeed. But the return value of this function does
not need to be checked by design.
Move rand_pool_grow from rand_pool_add_begin to rand_pool_bytes_needed.
Make an allocation error persistent to avoid falling back to less secure
or blocking entropy sources.
Fixes: a6a66e4511 ("Make rand_pool buffers more dynamic in their sizing.")
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/9687)
(cherry picked from commit fa3eb248e29ca8031e6a14e8a2c6f3cd58b5450e)
Since commit 7c226dfc43 a chained DRBG does not add additional
data anymore when reseeding from its parent. The reason is that
the size of the additional data exceeded the allowed size when
no derivation function was used.
This commit provides an alternative fix: instead of adding the
entire DRBG's complete state, we just add the DRBG's address
in memory, thereby providing some distinction between the different
DRBG instances.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9802)
When the new OpenSSL CSPRNG was introduced in version 1.1.1,
it was announced in the release notes that it would be fork-safe,
which the old CSPRNG hadn't been.
The fork-safety was implemented using a fork count, which was
incremented by a pthread_atfork handler. Initially, this handler
was enabled by default. Unfortunately, the default behaviour
had to be changed for other reasons in commit b5319bdbd0, so
the new OpenSSL CSPRNG failed to keep its promise.
This commit restores the fork-safety using a different approach.
It replaces the fork count by a fork id, which coincides with
the process id on UNIX-like operating systems and is zero on other
operating systems. It is used to detect when an automatic reseed
after a fork is necessary.
To prevent a future regression, it also adds a test to verify that
the child reseeds after fork.
CVE-2019-1549
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matt Caswell <matt@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/9802)
CLA: trivial
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Shane Lontis <shane.lontis@oracle.com>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/9295)
The rand pool support allocates maximal sized buffers -- this is typically
12288 bytes in size. These pools are allocated in secure memory which is a
scarse resource. They are also allocated per DRBG of which there are up to two
per thread.
This change allocates 64 byte pools and grows them dynamically if required.
64 is chosen to be sufficiently large so that pools do not normally need to
grow.
Reviewed-by: Bernd Edlinger <bernd.edlinger@hotmail.de>
(Merged from https://github.com/openssl/openssl/pull/9428)
(cherry picked from commit a6a66e4511)
The additional data allocates 12K per DRBG instance in the
secure memory, which is not necessary. Also nonces are not
considered secret.
[extended tests]
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/9424)
... to make the intended use more clear and differentiate
it from the data member "adin_pool".
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/7575)
(cherry picked from commit 31f32abb8e)
Reviewed-by: Paul Yang <yang.yang@baishancloud.com>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/7474)
(cherry picked from commit 21311777ad)
This bug was introduced by #7382 which enhanced RAND_add() to
accept large buffer sizes. As a consequence, RAND_add() now fails
for buffer sizes less than 32 bytes (i.e. less than 256 bits).
In addition, rand_drbg_get_entropy() forgets to reset the attached
drbg->pool in the case of an error, which leads to the heap corruption.
The problem occurred with RAND_load_file(), which reads the file in
chunks of 1024 bytes each. If the size of the final chunk is less than
32 bytes, then RAND_add() fails, whence RAND_load_file() fails
silently for buffer sizes n = k * 1024 + r with r = 1,...,31.
This commit fixes the heap corruption only. The other issues will
be addressed in a separate pull request.
Thanks to Gisle Vanem for reporting this issue.
Fixes#7449
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/7455)
(cherry picked from commit 5b4cb385c1)
In pull request #4328 the seeding of the DRBG via RAND_add()/RAND_seed()
was implemented by buffering the data in a random pool where it is
picked up later by the rand_drbg_get_entropy() callback. This buffer
was limited to the size of 4096 bytes.
When a larger input was added via RAND_add() or RAND_seed() to the DRBG,
the reseeding failed, but the error returned by the DRBG was ignored
by the two calling functions, which both don't return an error code.
As a consequence, the data provided by the application was effectively
ignored.
This commit fixes the problem by a more efficient implementation which
does not copy the data in memory and by raising the buffer the size limit
to INT32_MAX (2 gigabytes). This is less than the NIST limit of 2^35 bits
but it was chosen intentionally to avoid platform dependent problems
like integer sizes and/or signed/unsigned conversion.
Additionally, the DRBG is now less permissive on errors: In addition to
pushing a message to the openssl error stack, it enters the error state,
which forces a reinstantiation on next call.
Thanks go to Dr. Falko Strenzke for reporting this issue to the
openssl-security mailing list. After internal discussion the issue
has been categorized as not being security relevant, because the DRBG
reseeds automatically and is fully functional even without additional
randomness provided by the application.
Fixes#7381
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/7382)
(cherry picked from commit 3064b55134)
Fixes#7022
In pull request #6432 a change was made to keep the handles to the
random devices opened in order to avoid reseeding problems for
applications in chroot environments.
As a consequence, the handles of the random devices were leaked at exit
if the random generator was not used by the application. This happened,
because the call to RAND_set_rand_method(NULL) in rand_cleanup_int()
triggered a call to the call_once function do_rand_init, which opened
the random devices via rand_pool_init().
Thanks to GitHub user @bwelling for reporting this issue.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/7023)
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Tim Hudson <tjh@openssl.org>
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/6778)
This allows operation inside a chroot environment without having the
random device present.
A new call, RAND_keep_random_devices_open(), has been introduced that can
be used to control file descriptor use by the random seed sources. Some
seed sources maintain open file descriptors by default, which allows
such sources to operate in a chroot(2) jail without the associated device
nodes being available.
Reviewed-by: Matt Caswell <matt@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/6432)
rand_pool_bytes_needed() was constructed in such a way that the
smallest acceptable entropy factor was 1 entropy bits per 8 bits of
data. At the same time, we have a DRBG_MINMAX_FACTOR that allows
weaker source, as small as 1 bit of entropy per 128 bits of data.
The conclusion is that rand_pool_bytes_needed() needs to change to
support weaker entropy sources. We therefore change the input of
entropy per byte to be an entropy factor instead. This entropy factor
expresses how many bits of data it takes (on average) to get 1 bit of
entropy.
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/6150)
Fixes#5849
In pull request #5503 a fallback was added which adds a random nonce of
security_strength/2 bits if no nonce callback is provided. This change raised
the entropy requirements form 256 to 384 bit, which can cause problems on some
platforms (e.g. VMS, see issue #5849).
The requirements for the nonce are given in section 8.6.7 of NIST SP 800-90Ar1:
A nonce may be required in the construction of a seed during instantiation
in order to provide a security cushion to block certain attacks.
The nonce shall be either:
a) A value with at least (security_strength/2) bits of entropy, or
b) A value that is expected to repeat no more often than a
(security_strength/2)-bit random string would be expected to repeat.
Each nonce shall be unique to the cryptographic module in which instantiation
is performed, but need not be secret. When used, the nonce shall be considered
to be a critical security parameter.
This commit implements a nonce of type b) in order to lower the entropy
requirements during instantiation back to 256 bits.
The formulation "shall be unique to the cryptographic module" above implies
that the nonce needs to be unique among (with high probability) among all
DRBG instances in "space" and "time". We try to achieve this goal by creating a
nonce of the following form
nonce = app-specific-data || high-resolution-utc-timestamp || counter
Where || denotes concatenation. The application specific data can be something
like the process or group id of the application. A utc timestamp is used because
it increases monotonically, provided the system time is synchronized. This approach
may not be perfect yet for a FIPS evaluation, but it should be good enough for the
moment.
This commit also harmonizes the implementation of the get_nonce() and the
get_additional_data() callbacks and moves the platform specific parts from
rand_lib.c into rand_unix.c, rand_win.c, and rand_vms.c.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5920)
When these two functions returned zero, it could mean:
1. that an error occured. In their case, the error is an overflow of
the pool, i.e. the correct response from the caller would be to
stop trying to fill the pool.
2. that there isn't enought entropy acquired yet, i.e. the correct
response from the caller would be to try and add more entropy to
the pool.
Because of this ambiguity, the returned zero turns out to be useless.
This change makes the returned value more consistent. 1 means the
addition of new entropy was successful, 0 means it wasn't. To know if
the pool has been filled enough, the caller will have to call some
other function, such as rand_pool_entropy_available().
Fixes#5846
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/5876)
If a nonce is required and the get_nonce callback is NULL, request 50%
more entropy following NIST SP800-90Ar1 section 9.1.
Reviewed-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
GH: #5503
This avoids lock contention.
Reviewed-by: Tim Hudson <tjh@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/5547)
There is a requirements of having access to a live entropy source
which we can't do with the default callbacks. If you need prediction
resistance you need to set up your own callbacks that follow the
requirements of NIST SP 800-90C.
Reviewed-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
GH: #5402
Fixes#4403
This commit moves the internal header file "internal/rand.h" to
<openssl/rand_drbg.h>, making the RAND_DRBG API public.
The RAND_POOL API remains private, its function prototypes were
moved to "internal/rand_int.h" and converted to lowercase.
Documentation for the new API is work in progress on GitHub #5461.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5462)
We currently don't support the algorithm from NIST SP 800-90C
10.1.2 to use a weaker DRBG as source
Reviewed-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
GH: #5506
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
GH: #5400
In PR #5295 it was decided that the locking api should remain private
and used only inside libcrypto. However, the locking functions were added
back to `libcrypto.num` by `mkdef.pl`, because the function prototypes
were still listed in `internal/rand.h`. (This header contains functions
which are internal, but shared between libcrypto and libssl.)
This commit moves the prototypes to `rand_lcl.h` and changes the names
to lowercase, following the convention therein. It also corrects an
outdated documenting comment.
Reviewed-by: Richard Levitte <levitte@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5375)
This commit adds three new accessors to the internal DRBG lock
int RAND_DRBG_lock(RAND_DRBG *drbg)
int RAND_DRBG_unlock(RAND_DRBG *drbg)
int RAND_DRBG_enable_locking(RAND_DRBG *drbg)
The three shared DRBGs are intended to be used concurrently, so they
have locking enabled by default. It is the callers responsibility to
guard access to the shared DRBGs by calls to RAND_DRBG_lock() and
RAND_DRBG_unlock().
All other DRBG instances don't have locking enabled by default, because
they are intendended to be used by a single thread. If it is desired,
locking can be enabled by using RAND_DRBG_enable_locking().
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/5294)
Remove the timer and TSC additional input code and instead provide a single
routine that attempts to use the "best" timer/counter available on the
system. It attempts to use TSC, then various OS dependent resources and
finally several tries to obtain the date. If any of these timer/counters
is successful, the rest are skipped.
No randomness is credited for this.
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
(Merged from https://github.com/openssl/openssl/pull/5231)
The functions RAND_bytes() and RAND_priv_bytes() are now both based
on a common implementation using RAND_DRBG_bytes() (if the default
OpenSSL rand method is active). This not only simplifies the code
but also has the advantage that additional input from a high precision
timer is added on every generate call if the timer is available.
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/5251)
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)
Some older glibc versions require the `-lrt` linker option for
resolving the reference to `clock_gettime'. Since it is not desired
to add new library dependencies in version 1.1.1, the call to
clock_gettime() is replaced by a call to gettimeofday() for the
moment. It will be added back in version 1.2.
Signed-off-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Richard Levitte <levitte@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
(Merged from https://github.com/openssl/openssl/pull/5199)
A third shared DRBG is added, the so called master DRBG. Its sole purpose
is to reseed the two other shared DRBGs, the public and the private DRBG.
The randomness for the master DRBG is either pulled from the os entropy
sources, or added by the application using the RAND_add() call.
The master DRBG reseeds itself automatically after a given number of generate
requests, but can also be reseeded using RAND_seed() or RAND_add().
A reseeding of the master DRBG is automatically propagated to the public
and private DRBG. This construction fixes the problem, that up to now
the randomness provided by RAND_add() was added only to the public and
not to the private DRBG.
Signed-off-by: Dr. Matthias St. Pierre <Matthias.St.Pierre@ncp-e.com>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
(Merged from https://github.com/openssl/openssl/pull/4402)
The drbg's lock must be held across calls to RAND_DRBG_generate()
to prevent simultaneous modification of internal state.
This was observed in practice with simultaneous SSL_new() calls attempting
to seed the (separate) per-SSL RAND_DRBG instances from the global
rand_drbg instance; this eventually led to simultaneous calls to
ctr_BCC_update() attempting to increment drbg->bltmp_pos for their
respective partial final block, violating the invariant that bltmp_pos < 16.
The AES operations performed in ctr_BCC_blocks() makes the race window
quite easy to trigger. A value of bltmp_pos greater than 16 induces
catastrophic failure in ctr_BCC_final(), with subtraction overflowing
and leading to an attempt to memset() to zero a very large range,
which eventually reaches an unmapped page and segfaults.
Provide the needed locking in get_entropy_from_parent(), as well as
fixing a similar issue in RAND_priv_bytes(). There is also an
unlocked call to RAND_DRBG_generate() in ssl_randbytes(), but the
requisite serialization is already guaranteed by the requirements on
the application's usage of SSL objects, and no further locking is
needed for correct behavior. In that case, leave a comment noting
the apparent discrepancy and the reason for its safety (at present).
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4328)
Reseeding is handled very differently by the classic RAND_METHOD API
and the new RAND_DRBG api. These differences led to some problems when
the new RAND_DRBG was made the default OpenSSL RNG. In particular,
RAND_add() did not work as expected anymore. These issues are discussed
on the thread '[openssl-dev] Plea for a new public OpenSSL RNG API'
and in Pull Request #4328. This commit fixes the mentioned issues,
introducing the following changes:
- Replace the fixed size RAND_BYTES_BUFFER by a new RAND_POOL API which
facilitates collecting entropy by the get_entropy() callback.
- Don't use RAND_poll()/RAND_add() for collecting entropy from the
get_entropy() callback anymore. Instead, replace RAND_poll() by
RAND_POOL_acquire_entropy().
- Add a new function rand_drbg_restart() which tries to get the DRBG
in an instantiated state by all means, regardless of the current
state (uninstantiated, error, ...) the DRBG is in. If the caller
provides entropy or additional input, it will be used for reseeding.
- Restore the original documented behaviour of RAND_add() and RAND_poll()
(namely to reseed the DRBG immediately) by a new implementation based
on rand_drbg_restart().
- Add automatic error recovery from temporary failures of the entropy
source to RAND_DRBG_generate() using the rand_drbg_restart() function.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Kurt Roeckx <kurt@roeckx.be>
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Ben Kaduk <kaduk@mit.edu>
(Merged from https://github.com/openssl/openssl/pull/4328)
The one creating the DRBG should instantiate it, it's there that we
know which parameters we should use to instantiate it.
This splits the rand init in two parts to avoid a deadlock
because when the global drbg is created it wands to call
rand_add on the global rand method.
Reviewed-by: Rich Salz <rsalz@openssl.org>
Reviewed-by: Paul Dale <paul.dale@oracle.com>
GH: #4268
The DRBG callbacks 'get_entropy()' and 'cleanup_entropy()' are designed
in such a way that the randomness buffer does not have to be allocated
by the calling function. It receives the address of a dynamically
allocated buffer from get_entropy() and returns this address to
cleanup_entropy(), where it is freed. If these two calls are properly
paired, the address can be stored in a stack local variable of the
calling function, so there is no need for having a 'randomness' member
(and a 'filled' member) in 'RAND_DRBG'.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4266)
With the introduction of RAND_poll_ex(), the `RAND_add()` calls were
replaced by meaningless cb(...). This commit changes the 'cb(...)'
calls back to 'rand_add(...)' calls by changing the signature as follows:
-int RAND_poll_ex(RAND_poll_fn cb, void *arg);
+int RAND_poll_ex(RAND_poll_cb rand_add, void *arg);
Changed the function typedef name to 'RAND_poll_cb' to emphasize the fact
that the function type represents a callback function.
Reviewed-by: Paul Dale <paul.dale@oracle.com>
Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/4266)
