[ENBD] 2.4.32 write & m_r problem

[Peter T. Breuer]

"Also sprach Peter T. Breuer:"
> Today I will try and shift the fairly adhoc fcntl locking code in
> enbd-client to go through one of the more standard (in enbd) locking
> abstractions

Incidentally, I don't prefectly "know" why the client is doing locking.
Each client daemon appears to be doing readers/writers locking of the
region on the enbd device that it is treating.

Perhaps if the server were very very slow to handle one request, so
that a second request for the same area were issued by the kernel
before the first request had yet been returned, that would be it ...

But the kernel is vulnerable to this in every device, surely?  A hard
disk may decide to be slow to return on one request, and then the kernel
may issue a new request on the same disk area before the first has
returned.

Anyone know if there is any protection against that? I have an
existential doubt.

Peter