| 49ebea12f4954f3dbcd16a33a8fa98532d29f8c9 |
|
22-Dec-2015 |
Alex Klyubin <klyubin@google.com> |
Fix sanity checks around direct ByteBuffer memory access.
This fixes the sanity checks around access to memory backing direct
BytBuffer instances. The previous checks would've erronously failed
if pointers crossed the 2^63 boundary. There is no need for check for
pointer overflow.
Bug: 24674857
Change-Id: Ic8b5a651418c401d32eb0c8053217988963cd326
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 67cf2fcd35de497bb4dd7a2cc9bf2a02f0b06349 |
|
06-Nov-2015 |
Alex Klyubin <klyubin@google.com> |
Adjust names of digest-related NativeCrypto methods.
This adjusts the names of digest-related NativeCrypto methods to match
the names of underlying BoringSSL functions. This makes it easier to
reason about the functionality being invoked via NativeCrypto.
Change-Id: I04e2148ba818ae3e9ad60871b046052fcfffec4d
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 75d8933cc4b5053080612f44f9d975f659d933d2 |
|
02-Nov-2015 |
Alex Klyubin <klyubin@google.com> |
Speed up digesting by avoiding unnecessary operations.
Conscrypt's MessageDigest implementations at the end of computing a
digest create and initialize a new EVP_MD_CTX and then also intialize
the digest struct there. This is done because the MessageDigest
instance could be reused for a new digesting session.
This change implements three optimizations:
1. MessageDigestSpi now reuses its EVP_MD_CTX instead of creating a
new one for each new digesting session.
2. MessageDigestSpi now defers the initialization of the digest struct
in EVP_MD_CTX till the first invocation of
engineUpdate/engineDigest.
3. MessagDigestSpi (and SignatureSpi) no longer invoke EVP_MD_CTX_init
after EVP_MD_CTX_create because EVP_MD_CTX_create initializes the
EVP_MD_CTX it creates.
libcore's MessageDigestBenchmark on Nexus 5 shows:
* 10-15% faster performance for a single digest of 8192 bytes.
* 15-20% faster performance for reusing a MessageDigest instance to
compute a digest of 8192 bytes ten times.
Change-Id: I8a0697310ef7efcd4db6870e54eb46102fd4a941
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| c97e9654f16d2830a7ca268496f195d27cbc1b34 |
|
31-Oct-2015 |
Alex Klyubin <klyubin@google.com> |
Revert "Speed up digesting by avoiding unnecessary operations."
This reverts commit 5041dd13c9499e4154436ef1f105a3d5d46caa19.
This change broke a libcore test:
org.apache.harmony.security.tests.java.security.DigestInputStream2Test#test_onZ
java.lang.RuntimeException: Unable to copy EVP_MD_CTX
at com.android.org.conscrypt.NativeCrypto.EVP_MD_CTX_copy(Native Method)
at com.android.org.conscrypt.OpenSSLMessageDigestJDK.clone(OpenSSLMessageDigestJDK.java:200)
at java.security.MessageDigest$MessageDigestImpl.clone(MessageDigest.java:428)
at org.apache.harmony.security.tests.java.security.DigestInputStream2Test.test_onZ(DigestInputStream2Test.java:62)
org.apache.harmony.security.tests.java.security.DigestInputStream2Test#test_onZ FAIL (EXEC_FAILED)
Change-Id: Ib925bc0aadb633bbad4240f1d52bdb6676afc56f
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 5041dd13c9499e4154436ef1f105a3d5d46caa19 |
|
29-Oct-2015 |
Alex Klyubin <klyubin@google.com> |
Speed up digesting by avoiding unnecessary operations.
Conscrypt's MessageDigest implementations at the end of computing a
digest create and initialize a new EVP_MD_CTX and then also intialize
the digest struct there. This is done because the MessageDigest
instance could be reused for a new digesting session.
This change implements two optimizations:
1. MessageDigestImpl instance now reuses its EVP_MD_CTX instead of
creating a new one for each new digesting session.
2. MessageDigestImpl instance now defers the initialization of the
digest struct in EVP_MD_CTX till the first invocation of
engineUpdate/engineDigest.
libcore's MessageDigestBenchmark on Nexus 5 shows:
* 10-15% faster performance for a single digest of 8192 bytes.
* 15-20% faster performance for reusing a MessageDigest instance to
compute a digest of 8192 bytes ten times.
Change-Id: I0e476381321127642315355f848a1ba90114fe7d
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 91fd701b4587210b794fa1a537c38f306a3241d6 |
|
15-Oct-2015 |
Alex Klyubin <klyubin@google.com> |
Zero-copy digesting for direct ByteBuffer input.
Prior to this change, Conscrypt's MessageDigest.update(ByteBuffer)
invoked for a direct ByteBuffer resulted in the creation of a new
byte[] of size ByteBuffer.remaining() and the copying of the
ByteBuffer's contents into that array.
This change implements an optimization which avoids the allocation
and copying, by making BoringSSL EVP_DigestUpdate read directly from
the memory region represented by the direct ByteBuffer.
Change-Id: I112d318128402d1d78e226df9dfe54af55955953
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 4bff0a15ae03c8a3e1ae95590cc8c4240837bff6 |
|
25-Nov-2014 |
Kenny Root <kroot@google.com> |
Convert EVP_MD_CTX to new style
To avoid conflicts in the language spec and how Conscrypt does native
calls, we need to wrap all native references in a Java object reference.
Calling NativeCrypto's static native methods with a raw pointer doesn't
guarantee that the calling object won't be finalized during the method
running.
Bug: 16656908
Change-Id: I165e041a8fe056770d6ce6d6cd064c411575b7c4
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 88b56b0666fe3a702da4edc7b5b58266033c5871 |
|
16-May-2014 |
Kenny Root <kroot@google.com> |
OpenSSLMessageDigestJDK: support clone
Add the ability to clone the MessageDigest state.
Bug: 14821275
Change-Id: Ifa1b48db708448b971afe1e7360876f3fbe47588
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 652ff53bd48ed61389337a42d8e50cdb7ace0fec |
|
24-Feb-2014 |
Kenny Root <kroot@google.com> |
Fix up concurrent use of APIs
Code that is incorrectly using MessageDigest, Signature, or Mac in
multiple threads simultaneously could cause a SEGV if OpenSSL is
clearing out the MD_CTX at the same time another thread is trying to
write to it. Make sure we initialize a new MD_CTX after each run to
avoid crashing. The program using the instances concurrently is still
wrong and will most likely get inconsistent results.
Switch to using a context object instance to make sure we can hold a
reference to the object during the native call.
Bug: 8787753
Change-Id: I2518613a47cf03c811a29d17040804fc708394dd
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| f06338c01394610174fe2b3532beac56d61d9e26 |
|
07-Nov-2013 |
Kenny Root <kroot@google.com> |
Random cleanups of old code style
Add @Override annotation, remove unused imports, and remove unnecessary
casts. Also make sure annotations are on a line by themselves.
Change-Id: I294b43353d7b1e77fd1c9d031af7b7062f024eee
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| d2db2c558ef6afc14d59f4a6b547598ff3973597 |
|
26-Sep-2013 |
Kenny Root <kroot@google.com> |
Conscrypt: add SHA-224 with tests
SHA-224 has made a comeback in the latest StandardNames documentation.
This change adds tests for SHA-224 and also Conscrypt providers for
things we have code paths to support.
Change-Id: I8c200082ff76ee4ae38b6efaa16e6741b33b7f5b
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|
| 860d2707ce126ef8f66e3eac7ceeab6d24218cd8 |
|
24-Apr-2013 |
Kenny Root <kroot@google.com> |
Move JSSE to new package
To help with shipping the JSSE with apps that want to bundle it, move
it to a new package so that the tangles in other parts of the library
can be untangled.
Change-Id: I810b6861388635301e28aee5b9b47b8e6b35b430
/external/conscrypt/src/main/java/org/conscrypt/OpenSSLMessageDigestJDK.java
|