I have removed ANDIrand from my Solaris8 system and installed patch
112438-01 which provides native /dev/random and /dev/urandom devices.
Recompiled both openssl-0.9.6c and openssh-3.1p1, restarted sshd.
Now I seem unable to use scp, and connections to the local sshd appear
to be very slow.
$ scp php-4.2.0.tar.gz user at remote:/WWWserv/src
local at bastion's password:
php-4.2.0.tar.gz 6% |* | 224 KB - stalled -
When I'm trussing this scp process, I find that the following
output repeats over and over at some stage:
...
Received signal #14, SIGALRM, in write() [caught]
write(7, " AF8 d >FF z y 9FF02 z x".., 8192) = 5120
ioctl(1, TIOCGSID, 0xFFBEDE34) = 0
getsid(0) = 508
ioctl(1, TIOCGPGRP, 0xFFBEDE9C) = 0
ioctl(1, TIOCGWINSZ, 0xFFBEDEF8) = 0
write(1, "\r p h p - 4 . 2 . 0 . t".., 80) = 80
sigaction(SIGALRM, 0xFFBEDFF8, 0xFFBEE078) = 0
alarm(1) = 0
setcontext(0xFFBEE1E8)
...
Has anyone seens this before? Is openssh at fault here, or is it openssl?
openssh configure: --prefix=/opt/ssh --without-rpath --sysconfdir=/etc .
(Still not on this list; please Cc: replies :)
On Wed, 1 May 2002, Lars Hecking wrote:> > I have removed ANDIrand from my Solaris8 system and installed patch > 112438-01 which provides native /dev/random and /dev/urandom devices. > Recompiled both openssl-0.9.6c and openssh-3.1p1, restarted sshd. > > Now I seem unable to use scp, and connections to the local sshd appear > to be very slow. > > $ scp php-4.2.0.tar.gz user at remote:/WWWserv/src > local at bastion's password: > php-4.2.0.tar.gz 6% |* | 224 KB - stalled - > > When I'm trussing this scp process, I find that the following > output repeats over and over at some stage: > > ... > Received signal #14, SIGALRM, in write() [caught] > write(7, " AF8 d >FF z y 9FF02 z x".., 8192) = 5120 > ioctl(1, TIOCGSID, 0xFFBEDE34) = 0 > getsid(0) = 508 > ioctl(1, TIOCGPGRP, 0xFFBEDE9C) = 0 > ioctl(1, TIOCGWINSZ, 0xFFBEDEF8) = 0 > write(1, "\r p h p - 4 . 2 . 0 . t".., 80) = 80 > sigaction(SIGALRM, 0xFFBEDFF8, 0xFFBEE078) = 0 > alarm(1) = 0 > setcontext(0xFFBEE1E8) > ...This is just the progress meter periodically updating, which will happen regardless of data flow. You need to see why the connection stalls. Does a FTP connection stall between the two hosts? If no, you should try "scp -v -v -v xxxx yy:zzzz" to turn on full debugging. -d
Lars Hecking wrote:> I have removed ANDIrand from my Solaris8 system and installed patch > 112438-01 which provides native /dev/random and /dev/urandom devices. > Recompiled both openssl-0.9.6c and openssh-3.1p1, restarted sshd.You did reboot it, right? The patch installs a kernel driver and touches /reconfigure. Try ls -l /dev/*rand* and see if the devices really are there. -Daz.
Darren Tucker writes:> Lars Hecking wrote: > > I have removed ANDIrand from my Solaris8 system and installed patch > > 112438-01 which provides native /dev/random and /dev/urandom devices. > > Recompiled both openssl-0.9.6c and openssh-3.1p1, restarted sshd. > > You did reboot it, right? The patch installs a kernel driver and touches > /reconfigure. Try ls -l /dev/*rand* and see if the devices really are > there.Yes and yes. I am quite familiar with patch-application and related procedures :-) In this case, I was applying the recommended patch cluster plus 112438-01 after an upgrade to Sol8 02/02.