Does CIFS client retry?

[Richard Sharpe]

On Wed, 26 Jun 2002, Cuong Tran wrote:

> Hi, I hope the experts could shed some light on my questions here:
>  
> Say a CIFS client sends a number of delayed writes to a server. The server
> acknowledges the writes but it does not commit the writes to persistent
> storage yet. 

You mean, when the client sends a WriteX Request, the server responds with 
a WriteX response with a status of OK?
  
> Due to some reason, the server "loses" the writes by the CIFS client. 
> How does the server recover from this problem?

You mean that the server has violated its agreement with the client?
 
>     (a) Can it ask the client to resend the writes which the server has yet
> committed to persistent storage? 

There is nothing in the protocol, AFAIK, that allows the server to say:

  Oh, by the way, you know that data I told you was safe, well, I screwed
  up, and you have to send it to me again.

>     (b) What happens on the CIFS client if the server  closes its connection
> to the client? How much is the application on the CIFS client "shielded"
> from this type of errors?

Clients implement reconnect, but it is not clear that all of these issues 
have been well thought out. If the client still has data that has not been 
written to the server, it will try to reconnect and flush the data, I 
believe.

Regards
-----
Richard Sharpe, rsharpe at ns.aus.com, rsharpe at samba.org, 
sharpe at ethereal.com