| Age | Commit message (Collapse) | Author |
|---|
|
|
|
The resource was not needed as it was never kept between external
crypto calls.
|
|
|
|
enif_open_resource_type() must be called even during repeated loading
of same library. Otherwise the resource type will be deallocated
when the old module instance is purged.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
and remove corresponding specific functions for srp and ecdh
but leave dh_ functions for backward compatibility.
|
|
|
|
and remove curve from the returned tuple.
|
|
|
|
|
|
as well as mpint's for backward compatibility.
|
|
for rsa, dss, and new ecdsa.
No mpint's accepted.
|
|
|
|
Conflicts:
lib/crypto/src/crypto.erl
|
|
Adjust API to better fit in with similar funtions in crypto
|
|
algorithms
add algorithms/0 function that returns a list off compiled in crypto algorithms
and make tests suites with SHA226, SHA256, SHA384 and SHA512 conditional based
on that
|
|
|
|
As a preparation for the new mod_exp_prime and in the quest to reduce
to use of the old mpint format.
|
|
A couple of #defines were missed in my previous patch.
|
|
add ripemd160 message digest support to the crypto app,
includes some test cases.
|
|
|
|
An attempt to enable the use of openssl 0.9.7
|
|
if static linking of openssl is used.
|
|
|
|
|
|
Facts:
crypto nif-lib registers callback functions that openssl uses
for memory management and thread synchronization. The callback
functions can only be set once, openssl does not allow changing the
callback functions.
Problem:
If openssl is dynamicly linked to crypto, you might get s scenario
where the crypto lib is unloaded while leaving openssl loaded
with its old pointers to the unloaded crypto code intact.
If crypto is then reloaded (by init:restart() for example), the crypto
nif-lib might get relocated at a different address. crypto calls
openssl which in turn calls the old invalid callback functions...kaboom.
Solution:
Break apart the callback functions into a separate dynamic lib that
crypto loads with dlopen. When crypto is unloaded the callback lib is
left in place to be reused if/when crypto is loaded again.
|
|
|
|
* ia/ssl/tls1.1and1.2: (46 commits)
ssl: Clean up of code thanks to dialyzer
ssl: Test suite adjustments
ssl & public_key: Prepare for release
ssl: Use crypto:strong_rand_bytes if possible
ssl & public_key: Add use of more "sha-rsa oids"
ssl: Fix inet header option to behave as in inet
ssl: TLS 1.2: fix hash and signature handling
ssl: TLS 1.2: fix Certificate Request list of Accepted Signatur/Hash combinations
ssl: Add Signature Algorithms hello extension from TLS 1.2
ssl: Fix rizzo tests to run as intended
ssl: TLS-1.1 and TLS-1.2 support should not be default until R16
ssl: Signture type bug
ssl: Add crypto support check (TLS 1.2 require sha256 support)
ssl: Dialyzer fixes
ssl: IDEA cipher is deprecated by TLS 1.2
ssl: Run relevant tests for all SSL/TLS versions
ssl: Add TLS version switches to openssl tests
ssl: Enable TLS 1.2
ssl: Enable mac_hash for TLS 1.2
ssl: Implement TLS 1.2 signature support
...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Replace _hash functions with {digest,_} argument
to existing sign/verify functions.
|
|
|
|
Remove the need for padding up to 16-byte multiple.
|
|
by using extra redundant information as part of the key
that will speed things up for OpenSSL.
Affects rsa_sign, rsa_private_encrypt and rsa_private_decrypt.
|
|
|
|
No test and doc
|
|
|
|
|
|
The current crypto module implementations require all of the data
being encrypted or authenticated to be in memory at one time. When
trying to encrypt or authenticate a large file (on order of GBs),
this is problematic.
The implementation of AES CTR uses the same underlying implementation
as aes_ctr_[en|de]crypt, but hands the state back to the client
after every operation.
The HMAC implementation differs from the previous implementations of
sha_mac and md5_mac. The old implementations did not utilize the
OpenSSL HMAC implementation. In order to ensure that I didn't
implement something incorrectly, I chose to use the OpenSSL HMAC
implementation directly, since it handles streaming as well. This
has the added side benefit of allowing other hash functions to be
used as desired (for instances, I added support for ripemd160
hashing).
While I haven't done this, it seems like the existing md5_mac and
sha_mac functions could either be depricated or redefined in terms
of the new hmac_ functions.
Update AES CTR and HMAC streaming with code review input
Ensure that memcpy operations in hmac operations are being size
checked properly. Rename aes_ctr_XXX_with_state to
aes_ctr_stream_XXX. Remove redundant hmac_init_[sha|md5|ripemd160]
functions. Fix documentation for hmac_final_n.
Fix possible error using negative value as a marker on an unsigned int
Now, use a separate marker and add a unit test to test specifically for
a case where HashLen is larger than the underlying resultant hash.
Revert "Fix possible error using negative value as a marker on an unsigned int"
This reverts commit 59cb177aa96444c0fd3ace6d01f7b8a70dd69cc9.
Resolve buffer overflow posibility on an unsigned int.
Change handling the marker for HashLen to use the fact that a second
parameter that has to be the the HashLen was passed. Also, ensure
that HashLen parameter is positive.
|
|
Added some checks in crypto.erl and crypto.c.
Changed ssh_bits to use strong_rand_mpint.
|
|
Also adds documentation and unit tests.
Thanks to Geoff Cant.
|
