[sshd.c]
use kill(0, ...) instead of killpg(0, ...); on most operating systems
they are equivalent, but SUSv2 describes the latter as having undefined
behaviour; from portable; ok dtucker
(Id sync only; change is already in portable)
[Makefile.in digest.c digest.h hostfile.c kex.h mac.c hmac.c hmac.h]
replace openssl HMAC with an implementation based on our ssh_digest_*
ok and feedback djm@
different symbols for 'read' when various compiler flags are
in use, causing atomicio.c comparisons against it to break and
read/write operations to hang; ok dtucker
RLIMIT_NOFILE pseudo-sandbox on FreeBSD. In some configurations,
libc will attempt to open additional file descriptors for crypto
offload and crash if they cannot be opened.
[cipher.c cipher.h kex.c kex.h kexgexc.c]
Add a special case for the DH group size for 3des-cbc, which has an
effective strength much lower than the key size. This causes problems
with some cryptlib implementations, which don't support group sizes larger
than 4k but also don't use the largest group size it does support as
specified in the RFC. Based on a patch from Petr Lautrbach at Redhat,
reduced by me with input from Markus. ok djm@ markus@
platform hardening options: include some long long int arithmatic
to detect missing support functions for -ftrapv in libgcc and
equivalents, actually test linking when -ftrapv is supplied and
set either both -pie/-fPIE or neither. feedback and ok dtucker@
to on platforms where it's known to be reliably detected and off elsewhere.
Works around platforms such as FreeBSD 9.1 where it does not interop with
-ftrapv (it seems to work but fails when trying to link ssh). ok djm@
[canohost.c addrmatch.c]
Cast socklen_t when comparing to size_t and use socklen_t to iterate over
the ip options, both to prevent signed/unsigned comparison warnings.
Patch from vinschen at redhat via portable openssh, begrudging ok deraadt.
[session.c]
explicitly define USE_PIPES to 1 to prevent redefinition warnings in
portable on platforms that use pipes for everything. From redhat @
redhat.
optind) are defined in getopt.h already. Unfortunately they are defined as
"declspec(dllimport)" for historical reasons, because the GNU linker didn't
allow auto-import on PE/COFF targets way back when. The problem is the
dllexport attributes collide with the definitions in the various source
files in OpenSSH, which obviousy define the variables without
declspec(dllimport). The least intrusive way to get rid of these warnings
is to disable warnings for GCC compiler attributes when building on Cygwin.
Patch from vinschen at redhat.com.
declarations that stopped being included when we stopped including
<windows.h> from openbsd-compat/bsd-cygwin_util.h. Patch from vinschen at
redhat.com.
openbsd-compat/bsd-statvfs.h] Implement enough of statvfs on top of statfs
to be useful (and for the regression tests to pass) on platforms that
have statfs and fstatfs. ok djm@
[sandbox-null.c sandbox-rlimit.c sandbox-seccomp-filter.c]
[sandbox-systrace.c ssh-sandbox.h sshd.c] Support preauth sandboxing
using the Capsicum API introduced in FreeBSD 10. Patch by Dag-Erling
Smorgrav, updated by Loganaden Velvindron @ AfriNIC; ok dtucker@
hardening flags including -fstack-protector-strong. These default to on
if the toolchain supports them, but there is a configure-time knob
(--without-hardening) to disable them if necessary. ok djm@
[bufaux.c buffer.h kex.c kex.h kexc25519.c kexc25519c.c kexc25519s.c]
[kexdhc.c kexdhs.c kexecdhc.c kexecdhs.c kexgexc.c kexgexs.c]
avoid use of OpenSSL BIGNUM type and functions for KEX with
Curve25519 by adding a buffer_put_bignum2_from_string() that stores
a string using the bignum encoding rules. Will make it easier to
build a reduced-feature OpenSSH without OpenSSL in the future;
ok markus@
[sshconnect.c sshd.c]
ban clients/servers that suffer from SSH_BUG_DERIVEKEY, they are ancient,
deranged and might make some attacks on KEX easier; ok markus@
[digest.c digest.h hostfile.c kex.c kex.h kexc25519.c kexc25519c.c]
[kexc25519s.c kexdh.c kexecdh.c kexecdhc.c kexecdhs.c kexgex.c kexgexc.c]
[kexgexs.c key.c key.h roaming_client.c roaming_common.c schnorr.c]
[schnorr.h ssh-dss.c ssh-ecdsa.c ssh-rsa.c sshconnect2.c]
Introduce digest API and use it to perform all hashing operations
rather than calling OpenSSL EVP_Digest* directly. Will make it easier
to build a reduced-feature OpenSSH without OpenSSL in future;
feedback, ok markus@
[sftp-common.c]
When formating the time for "ls -l"-style output, show dates in the future
with the year, and rearrange a comparison to avoid a potentional signed
arithmetic overflow that would give the wrong result.
ok djm@
[auth2-hostbased.c auth2-pubkey.c compat.c compat.h ssh-rsa.c]
[sshconnect.c sshconnect2.c sshd.c]
refuse RSA keys from old proprietary clients/servers that use the
obsolete RSA+MD5 signature scheme. it will still be possible to connect
with these clients/servers but only DSA keys will be accepted, and we'll
deprecate them entirely in a future release. ok markus@
[ssh-dss.c ssh-ecdsa.c ssh-rsa.c]
make the original RSA and DSA signing/verification code look more like
the ECDSA/Ed25519 ones: use key_type_plain() when checking the key type
rather than tediously listing all variants, use __func__ for debug/
error messages
[ssh-agent.c]
bz#2186: don't crash (NULL deref) when deleting PKCS#11 keys from an agent
that has a mix of normal and PKCS#11 keys; fix from jay AT slushpupie.com;
ok dtucker
[channels.c]
bz#2147: fix multiple remote forwardings with dynamically assigned
listen ports. In the s->c message to open the channel we were sending
zero (the magic number to request a dynamic port) instead of the actual
listen port. The client therefore had no way of discriminating between
them.
Diagnosis and fix by ronf AT timeheart.net
[serverloop.c]
Cast client_alive_interval to u_int64_t before assinging to
max_time_milliseconds to avoid potential integer overflow in the timeout.
bz#2170, patch from Loganaden Velvindron, ok djm@
Damien Miller
Tim Rice
Darren Tucker