Re: init scripts problems ?

"Also sprach denis bonnenfant:"
> without -e ,

One thing I've now checked is that the "openers patch"  works on
its own for me.  You should _back out_ the hunk of the original "hack
patch" that set "do_reset = 1" in the no-show_errs branch of an if.

You should, in that hunk, get back to


               if (lo->aslot <= 0) {
                        // PTB we were the last client alive, diasable device
                        if (atomic_read (&lo->flags) & ENBD_SHOW_ERRS) {
                                // PTB soft_reset will invalidate_buffers
                                do_reset = 1;
                        }
               } if (!(atomic_read (&lo->flags) & ENBD_ENABLED)) {
                        // PTB must not call reenable as that clears the queue
                        do_enable = 1;
               }

The reason for backing it out is that the forced reset, although useful
for killing existing requests, is inadvertently generating a few more
requests by invalidating the partition table data (or lack of one),
causing the kernel to try and revalidate the table (by reading it!) when
the device is reenabled a moment or two later as the new client daemon
takes up its role.

The extra requests are making me nervous - it's a race at a bad time.
And I don't like to have to kill them. So lets stayenabled and
validated if e can.


Peter