Found by Barry K. Nathan.
Quoting from http://gcc.gnu.org/onlinedocs/gcc/Extended-Asm.html:
"You can put multiple assembler instructions together in a single asm
template, separated by the characters normally used in assembly code for the
system. A combination that works in most places is a newline to break the
line, plus a tab character to move to the instruction field (written as
‘\n\t’). Sometimes semicolons can be used, if the assembler allows semicolons
as a line-breaking character. Note that some assembler dialects use semicolons
to start a comment."
The MIPS32 bn_mul asm code causes segfaults on MIPS64 and failing
tests. Until someone has time to fix this up, MIPS64 platforms should
fall back to the C implementation (which works fine).
All symmetric cipher algorithms and hash algorithms now include support
for a POLARSSL_XXX_ALT flag that prevents the definition of the
algorithm context structure and all 'core' functions.
PKCS#8 encrypted key file support has been added to x509parse_key() with
support for some PCKS#12 PBE functions (pbeWithSHAAnd128BitRC4,
pbeWithSHAAnd3-KeyTripleDES-CBC and pbeWithSHAAnd2-KeyTripleDES-CBC)
Split up x509parse_key() into a (PEM) handler function and specific
DER parser functions for the PKCS#1 (x509parse_key_pkcs1_der()) and
unencrypted PKCS#8 (x509parse_key_pkcs8_unencrypted_der()) private
key formats.
Because of new pem_read_buffer() handling of when it writes use_len,
x509parse_crt() is able to better handle situations where a PEM blob
results in an error but the other blobs can still be parsed.
Rationale: The HAVEGE random generator has too many caveats to be a
standard generator that people rely on. The HAVEGE random generator is not
suitable for virtualized environments. In addition the HAVEGE random
generator is dependent on timing and specific processor traits that
cannot be guaranteed by default on compile time.
Our advice: only use HAVEGE as an additional random source for your
entropy pool, never as your primary source.
The ciphersuites parameter in the ssl_session structure changed from
'int *' to 'int **' and is now malloced in ssl_init() and freed in
ssl_free().
The new function ssl_set_ciphersuite_for_version() sets specific entries
inside this array. ssl_set_ciphersuite() sets all entries to the same
value.
The real peer certificate is copied into a x509_buf in the
ssl_cache_entry and reinstated upon cache retrieval. The information
about the rest of the certificate chain is lost in the process.
As the handshake (and certificate verification) has already been
performed, no issue is foreseen.
If the define POLARSSL_SSL_SRV_SUPPORT_SSLV2_CLIENT_HELLO is enabled,
the SSL Server module can handle the old SSLv2 Client Hello messages.
It has been updated to deny SSLv2 Client Hello messages during
renegotiation.
Split rsa_pkcs1_encrypt() into rsa_rsaes_oaep_encrypt() and
rsa_rsaes_pkcs1_v15_encrypt()
Split rsa_pkcs1_decrypt() into rsa_rsaes_oaep_decrypt() and
rsa_rsaes_pkcs1_v15_decrypt()
Split rsa_pkcs1_sign() into rsa_rsassa_pss_sign() and
rsa_rsassa_pkcs1_v15_sign()
Split rsa_pkcs1_verify() into rsa_rsassa_pss_verify() and
rsa_rsassa_pkcs1_v15_verify()
The original functions exist as generic wrappers to these functions.
New padding checking is unbiased on correct or incorrect padding and
has no branch prediction timing differences.
The additional MAC checks further straighten out the timing differences.
The flag POLARSSL_SSL_ALERT_MESSAGES switched between enabling and
disabling the sending of alert messages that give adversaries intel
about the result of their action. PolarSSL can still communicate with
other parties if they are disabled, but debugging of issues might be
harder.
Enable a dummy error function to make use of error_strerror() in
third party libraries easier.
Disable if you run into name conflicts and want to really remove the
error_strerror()
process
Single stepping the handshake process allows for better support of
non-blocking network stacks and for getting information from specific
handshake messages if wanted.
