Re: mysterious dead-locks with system ssh

2002-07-21 Thread Marc Recht

 I'm getting strange dead-locks/complete lookups when I use the system
 ssh with port forwarding. Using something like: 
 ssh -L8080:remote:8080 account@remote
 to forward a remote apache to my local box. When I access
 http://localhost:8080/ not later than the third click on link (or
 pressing reload) my box locks up completely. I don't even get into the
 debugger or get a crash dump. The box is just plain dead. Though
 normal SSH works perfectly and accessing a local (not forwarded)
 apache works, too. The world and kernel is of today and IIRC this
 actually did work last week (or the week before).
I've found the reason for this dead-lock. I had turned ZERO_COPY_SOCKETS on.
ESound/zero_copy_sockets causes the same dead-lock...

Marc



msg41231/pgp0.pgp
Description: PGP signature


Re: mysterious dead-locks with system ssh

2002-07-21 Thread Kenneth D. Merry

On Sun, Jul 21, 2002 at 22:29:55 +0200, Marc Recht wrote:
  I'm getting strange dead-locks/complete lookups when I use the system
  ssh with port forwarding. Using something like: 
  ssh -L8080:remote:8080 account@remote
  to forward a remote apache to my local box. When I access
  http://localhost:8080/ not later than the third click on link (or
  pressing reload) my box locks up completely. I don't even get into the
  debugger or get a crash dump. The box is just plain dead. Though
  normal SSH works perfectly and accessing a local (not forwarded)
  apache works, too. The world and kernel is of today and IIRC this
  actually did work last week (or the week before).
 I've found the reason for this dead-lock. I had turned ZERO_COPY_SOCKETS on.
 ESound/zero_copy_sockets causes the same dead-lock...

Hmm, nothing has changed in the zero copy code, although lots of things are
changing in the VM code and elsewhere in the kernel.

Can you track down whether the problem is with zero copy send or receive?

To figure this out, try disabling zero copy send and receive individually:

sysctl -w kern.ipc.zero_copy.send=0
sysctl -w kern.ipc.zero_copy.receive=1

then

sysctl -w kern.ipc.zero_copy.send=1
sysctl -w kern.ipc.zero_copy.receive=0

That'll help isolate the problem somewhat.  It's also possible that it only
happens with both turned on.

Also, what sort of network interfaces do you have on your machine?  What is
the MTU?

Are you running with SMP or not, and is WITNESS turned on or not?

Would you be able to try a binary search to try to figure out where things
broke for you?

Ken
-- 
Kenneth Merry
[EMAIL PROTECTED]

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-current in the body of the message



mysterious dead-locks with system ssh

2002-07-14 Thread Marc Recht

Hi!

I'm getting strange dead-locks/complete lookups when I use the system
ssh with port forwarding. Using something like: 
ssh -L8080:remote:8080 account@remote
to forward a remote apache to my local box. When I access
http://localhost:8080/ not later than the third click on link (or
pressing reload) my box locks up completely. I don't even get into the
debugger or get a crash dump. The box is just plain dead. Though
normal SSH works perfectly and accessing a local (not forwarded)
apache works, too. The world and kernel is of today and IIRC this
actually did work last week (or the week before).

Marc









To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-current in the body of the message



Re: mysterious dead-locks with system ssh

2002-07-14 Thread Cyrille Lefevre

On Sun, Jul 14, 2002 at 11:08:07PM +0200, Marc Recht wrote:
 I'm getting strange dead-locks/complete lookups when I use the system
 ssh with port forwarding. Using something like: 
 ssh -L8080:remote:8080 account@remote
 to forward a remote apache to my local box. When I access
 http://localhost:8080/ not later than the third click on link (or
 pressing reload) my box locks up completely. I don't even get into the
 debugger or get a crash dump. The box is just plain dead. Though

I had a similar problem w/ netscape-4 configured to use a local
(non started) proxy in -stable. w/ the proxy, no problem, w/o it,
plain dead-lock. don't know where is the lockup ?

Cyrille.
-- 
Cyrille Lefevre mailto:[EMAIL PROTECTED]

To Unsubscribe: send mail to [EMAIL PROTECTED]
with unsubscribe freebsd-current in the body of the message