multithread CIFS server and ECHO request

NTFSD
1

Hi,

I’ve developed a multithread CIFS server which runs on NT box. I am
using
NetBIOS API for communication with clients. And I’ve got a problem in
the
following scenario:

  1. Client sends write request to CIFS server.
  2. CIFS server gets the request and puts it in a queue for processing.
  3. CIFS server enters receive procedure again (NetBIOS receive).
  4. Job thread starts request processing and does not answer for a long time
    ~30 minutes
  5. After that client’s behavior is different.
    Some clients send ECHO request (CIFS server immediately answers) and after
    that
    clients are patiently waiting, other clients just reset connection after a
    timeout.

I wonder if there is a way to force client to send ECHO instead of
connection
reset.

For my server it is legal to process a request for a very long time (30
minutes
or even more). SESSTIMEOUT is not good solution, because every client
has to
change the registry.

Any ideas would be greatly appreciated.

Gena

2

No , you should not take that long to send an SMB response. The default
behaviour of the redirector on NT and OS/2 is to disconnect the session
if an SMB request times out. The value if I remember correctly is 45
seconds.
You possibly could see a SMB_ECHO keepalive for multi packet commands
like MPX, RAW or TRANSACT.

Roy.

-----Original Message-----
From: xxxxx@lists.osr.com
[mailto:xxxxx@lists.osr.com]On Behalf Of NTFSD
Sent: Wednesday, January 08, 2003 11:33 PM
To: File Systems Developers
Subject: [ntfsd] multithread CIFS server and ECHO request

Hi,

I’ve developed a multithread CIFS server which runs on NT box. I am
using
NetBIOS API for communication with clients. And I’ve got a problem in
the
following scenario:

  1. Client sends write request to CIFS server.
  2. CIFS server gets the request and puts it in a queue for processing.
  3. CIFS server enters receive procedure again (NetBIOS receive).
  4. Job thread starts request processing and does not answer for a long time
    ~30 minutes
  5. After that client’s behavior is different.
    Some clients send ECHO request (CIFS server immediately answers) and after
    that
    clients are patiently waiting, other clients just reset connection after a
    timeout.

I wonder if there is a way to force client to send ECHO instead of
connection
reset.

For my server it is legal to process a request for a very long time (30
minutes
or even more). SESSTIMEOUT is not good solution, because every client
has to
change the registry.

Any ideas would be greatly appreciated.

Gena

You are currently subscribed to ntfsd as: xxxxx@mwti.net
To unsubscribe send a blank email to xxxxx@lists.osr.com