[pnfs] [PATCH 9/9] pnfs client prevent race in sequence slot

[Trond Myklebust]

On Thu, 2007-09-27 at 12:07 -0400, William A. (Andy) Adamson wrote:

> 
> There is no way for the server to rewind a non-idempotent RPC that it
> has processed, so canceling an outstanding RPC would not work. I agree
> that -EINTR (-ERESTARTSYS ?) needs to be returned upstream. The client
> needs to respond differently depending on where the interupted RPC is
> in the execute state machine. If it has not sent an rpc, then re-use
> the sequence number for the next RPC. If an RPC has been sent to the
> server, then the client must finish processing the RPC in the normal
> fashion.

That would be a regression w.r.t. current behaviour. No deal :-)

Trond