[ENBD] proxing ioctl's

[Peter T. Breuer]

"A month of sundays ago Daniel Shane wrote:"
> All we need to do is mark the page undeletable, and when we close
> the connection to a server, just destroy the reserved buffers.
> 
> It looks like a 4 or 5 lines of code max to implement client 
> side journaling. I even think it would be more efficient that
> the current solution. 

Well, I don't know if this is "right", but it _seems_ to be working for
me... as far as I have tested, which is not much. I may simply be
looking at the ordinary kernel caching effect. I know the code was
called.

In the function nbd_do_request, which takes stuff off the kernel queue,
try changing the

             nbd_enqueue(lo,req);

to

             if ((req->cmd & 0x03) == WRITE && buffer_writes) {
                    struct buffer_head * bh;
                    // PTB take it off the queue and ack it now
                    blkdev_dequeue_request(req);
                    // PTB go through and protect the written buffers
                    for (bh = req->bh; bh; bh = bh->b_reqnext)
                       mark_buffer_protected(bh);
                    // PTB now vamoosh it normally
                    nbd_end_request(req);
             } else {
                    nbd_enqueue(lo,req);
             }

yah ... I added a MODULE_PARM(buffer_writes,"i"); and a static int
buffer_writes = 0; up top. If you have a better name, tell me.

I also added a stanza to write_proc allowing it to be set dynamically.

What happens to these buffers when one removes the driver? Don't I have
to do something about them? Do you think the protected bit is all that
is required?

Peter