_res, prevents problems on some platforms that have _res as a global but
don't have getrrsetbyname(), eg IRIX 5.3. Found and tested by
georg.schwarz at freenet.de, ok djm@.
[authfile.c authfile.h sshconnect2.c ssh.c sshconnect1.c]
Prevent ssh from trying to open private keys with bad permissions more than
once or prompting for their passphrases (which it subsequently ignores
anyway), similar to a previous change in ssh-add. bz #1186, ok djm@
session.c ssh-rand-helper.c sshd.c openbsd-compat/bsd-cygwin_util.c
openbsd-compat/setproctitle.c] Convert malloc(foo*bar) -> calloc(foo,bar)
in Portable-only code; since calloc zeros, remove now-redundant memsets.
Also add a couple of sanity checks. With & ok djm@
[includes.h session.c sftp.c]
Switch from using pipes to socketpairs for communication between
sftp/scp and ssh, and between sshd and its subprocesses. This saves
a file descriptor per session and apparently makes userland ppp over
ssh work; ok markus@ deraadt@ (ID Sync only - portable makes this
decision on a per-platform basis)
[auth.h clientloop.c dispatch.c dispatch.h kex.h]
replace the last non-sig_atomic_t flag used in a signal handler with a
sig_atomic_t, unfortunately with some knock-on effects in other (non-
signal) contexts in which it is used; ok markus@
[bufaux.c bufbn.c]
Move Buffer bignum functions into their own file, bufbn.c. This means
that sftp and sftp-server (which use the Buffer functions in bufaux.c
but not the bignum ones) no longer need to be linked with libcrypto.
ok markus@
[sftp-client.c]
avoid making a tiny 4-byte write to send the packet length of sftp
commands, which would result in a separate tiny packet on the wire by
using atomiciov(writev, ...) to write the length and the command in one
pass; ok deraadt@
[atomicio.c atomicio.h]
introduce atomiciov() function that wraps readv/writev to retry
interrupted transfers like atomicio() does for read/write;
feedback deraadt@ dtucker@ stevesk@ ok deraadt@
[buffer.c buffer.h channels.c]
Fix condition where we could exit with a fatal error when an input
buffer became too large and the remote end had advertised a big window.
The problem was a mismatch in the backoff math between the channels code
and the buffer code, so make a buffer_check_alloc() function that the
channels code can use to propsectivly check whether an incremental
allocation will succeed. bz #1131, debugged with the assistance of
cove AT wildpackets.com; ok dtucker@ deraadt@
[gss-genr.c]
GSSAPI buffers shouldn't be nul-terminated, spotted in bugzilla #1066
by dleonard AT vintela.com. use xasprintf() to simplify code while in
there; "looks right" deraadt@
[contrib/redhat/sshd.init openbsd-compat/Makefile.in]
[openbsd-compat/openbsd-compat.h openbsd-compat/port-linux.c]
[openbsd-compat/port-linux.h] Add support for SELinux, setting
the execution and TTY contexts. based on patch from Daniel Walsh,
bz #880; ok dtucker@
[ssh.c ssh_config.5]
add percent escape chars to the IdentityFile option, bz #1159 based
on a patch by imaging AT math.ualberta.ca; feedback and ok dtucker@
[authfd.c bufaux.c deattack.c gss-serv.c mac.c misc.c misc.h]
[monitor_wrap.c msg.c packet.c sftp-client.c sftp-server.c ssh-agent.c]
replace {GET,PUT}_XXBIT macros with functionally similar functions,
silencing a heap of lint warnings. also allows them to use
__bounded__ checking which can't be applied to macros; requested
by and feedback from deraadt@
Damien Miller
Darren Tucker