openssh-4.1p1-2/cygwin-1.5.18-1: write(2) misbehaving?

["Dave Kilzer" <DKilzer at StudentLoan dot org>]

After upgrading my Cygwin packages recently (running cygwin-1.5.18-1), I
found that I could no longer log in to a variety of Linux servers
running various versions of sshd from OpenSSH (OpenSSH_3.5p1 on Red Hat
9, OpenSSH_3.8.1p1 on Debian 3.1, OpenSSH_4.1p1 on Debian
post-3.1-unstable).  Here's the output I'm getting from "ssh -v
hostname" (replaced real hostname with "hostname"; replaced real IP
address with "1.2.3.4"; replaced real home directory with "~"):

$ ssh -v hostname
OpenSSH_4.1p1, OpenSSL 0.9.8 05 Jul 2005
debug1: Connecting to hostname [1.2.3.4] port 22.
debug1: Connection established.
debug1: identity file ~/.ssh/identity type -1
debug1: identity file ~/.ssh/id_rsa type 1
debug1: identity file ~/.ssh/id_dsa type -1
debug1: Remote protocol version 2.0, remote software version
OpenSSH_3.8.1p1 Debian-8.sarge.4
debug1: match: OpenSSH_3.8.1p1 Debian-8.sarge.4 pat OpenSSH_3.*
debug1: Enabling compatibility mode for protocol 2.0
debug1: Local version string SSH-2.0-OpenSSH_4.1
debug1: SSH2_MSG_KEXINIT sent
debug1: SSH2_MSG_KEXINIT received
debug1: kex: server->client aes128-cbc hmac-md5 none
debug1: kex: client->server aes128-cbc hmac-md5 none
debug1: SSH2_MSG_KEX_DH_GEX_REQUEST(1024<1024<8192) sent
debug1: expecting SSH2_MSG_KEX_DH_GEX_GROUP
debug1: SSH2_MSG_KEX_DH_GEX_INIT sent
debug1: expecting SSH2_MSG_KEX_DH_GEX_REPLY
debug1: Host 'hostname' is known and matches the RSA host key.
debug1: Found key in ~/.ssh/known_hosts:1
debug1: ssh_rsa_verify: signature correct
debug1: SSH2_MSG_NEWKEYS sent
debug1: expecting SSH2_MSG_NEWKEYS
Write failed: No such file or directory

After some debugging, I found out that the error message after "Write
failed:" is simply the last errno entry populated in the course of
running ssh (in this case ENOENT), and doesn't actually represent the
error that is occurring.

I also discovered two distinct ways to fix this problem:

1. Configure openssh with the '--with-cflags="-DPACKET_DEBUG=1"' switch,
then build openssh.  This prevents the bug from occurring at the cost of
copious debugging output of packet information.  The fact that this
works would indicate to me that there is some kind of compiler (code
generation) issue or that some kind of race condition exists that is
related to the network layer.

2. Apply the attached patch and (re)build openssh.  After reading the
manpage for write(2), it said that write(2) returns -1 on error, thus
errno will not contain a "related" error if checked when write(2)
returns 0.  (There were other mitigating circumstances when writing to
"real" files according to the manpage, but this did not apply.)  The
fact that this patch works tells me that write(2) seems to be broken,
and is actually returning 0 when it should be returning a positive
integer (e.g., the value of "len" in this case).

Note that I tried running "ssh.exe" from openssh-3.9p1-3/cygwin-1.5.13-1
with cygwin-1.5.18-1, and it had the exact same problem.  (I have not
tried compiling openssh-4.1p1-2 on the system running cygwin-1.5.13-1
yet, though.)

Any idea what's going on?

Dave

Attachment: openssh-4.1p1-2-cygwin-fix-write-failed.diff
Description: openssh-4.1p1-2-cygwin-fix-write-failed.diff

--
Unsubscribe info:      http://cygwin.com/ml/#unsubscribe-simple
Problem reports:       http://cygwin.com/problems.html
Documentation:         http://cygwin.com/docs.html
FAQ:                   http://cygwin.com/faq/