non-blocking BIOs so that it is sane -- so that, in other words, -1 with
a meaningful library error code (WANT_READ or WANT_WRITE) is returned
when we would block for I/O. Without this change, you have to sleep or
spin -- you can't know how to put the underlying socket in your select
or poll set.
Patch from http://marc.info/?l=openssl-dev&m=115154030723033&w=2 and
rationale at http://marc.info/?l=openssl-dev&m=115153998821797&w=2 where
sadly they were overlooked by the OpenSSL team for some time. It is hoped
that now that we've brought this change to their attention they will
integrate it into their sources and we can lose the local change in
NetBSD.
- Fix flaw if server key exchange message is omitted from a TLS handshake
which could lead to a silent crash.
- Fix double free in TLS server name extensions which could lead to a
remote crash.
Fixes CVE-2008-1672.
The BN_from_montgomery function in crypto/bn/bn_mont.c in OpenSSL 0.9.8e and
earlier does not properly perform Montgomery multiplication, which might
allow local users to conduct a side-channel attack and retrieve RSA
private keys.
OpenSSL:
1) Fix extremely misleading text in crypto.4 manual page so it does not
appear to claim that a new cloned file descriptor is required for every
session.
2) Fix severe performance problem (and fd leak!) in openssl cryptodev
engine resulting from misunderstanding probably caused by said manual
page text.
3) Check for session-ID wraparound in kernel cryptodev provider. Also,
start allocating sessions at 1, not 0 -- this will be necessary when
we add ioctls for the creation of multiple sessions at once, so we
can tell which if any creations failed.
OpenSSL 0.9.7 before 0.9.7l and 0.9.8 before 0.9.8d allows
remote attackers to cause a denial of service (inifnite loop
and memory consumption) via malformed ASN.1 structures that
trigger an improperly handled error condition.
http://www.cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-2940
OpenSSL 0.9.7 before 0.9.7l, 0.9.8 before 0.9.8d, and earlier
versions allows attackers to cause a denial of service (CPU
consumption) via certain public keys that require extra time
to process.
http://www.cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-3738
Buffer overflow in the SSL_get_shared_ciphers function in
OpenSSL 0.9.7 before 0.9.7l, 0.9.8 before 0.9.8d, and earlier
versions has unspecified impact and remote attack vectors
involving a long list of ciphers.
http://www.cve.mitre.org/cgi-bin/cvename.cgi?name=CVE-2006-4343
Unspecified vulnerability in the SSLv2 client code in OpenSSL
0.9.7 before 0.9.7l, 0.9.8 before 0.9.8d, and earlier versions
allows remote servers to cause a denial of service (client
crash) via unknown vectors.
- Rollback the updates for rsa.h, rsa_eay.c and rsa_err.c as they were
not necessary to address this vulnerability.
- Small update to the patch for rsa_sign.c for backward compatability so
the same patch can be applied to 0.9.[6-9]
Daniel Bleichenbacher recently described an attack on PKCS #1 v1.5
signatures. If an RSA key with exponent 3 is used it may be possible
to forge a PKCS #1 v1.5 signature signed by that key. Implementations
may incorrectly verify the certificate if they are not checking for
excess data in the RSA exponentiation result of the signature.
Since there are CAs using exponent 3 in wide use, and PKCS #1 v1.5 is
used in X.509 certificates, all software that uses OpenSSL to verify
X.509 certificates is potentially vulnerable, as well as any other use
of PKCS #1 v1.5. This includes software that uses OpenSSL for SSL or
TLS.
> use explicitly sized types for U_LLONG U_LONG and LONG; otherwise bn
> breaks on 64 bit platforms. The "LONG" openssl wants is really a 32 bit int.
Instead define SIXTY_FOUR_BIT_LONG where apropriate.
Regression tests still pass on sparc64 and i386. Furthermore this allows
us to finaly close PR 28935 (thanks to christos for removing the local
hacks on last import).
1. Instead of trying to cleanup the ugly ifdefs, we leave them alone so that
there are going to be fewer conflicts in the future.
2. Where we make changes to override things #ifdef __NetBSD__ around them
so that it is clear what we are changing. This is still missing in some
places, notably in opensslconf.h because it would make things messier.
