Re: [NBD] Use per-device semaphore instead of BKL

[Paul Clements <paul.clements@steeleye.com>]

Herbert Xu wrote:
> On Sun, Nov 20, 2005 at 12:19:17PM -0500, Paul Clements wrote:
> 
>>The dropping of the lock in nbd_do_it is actually critical to the way 
>>nbd functions. nbd_do_it runs for the lifetime of the nbd device, so if 
>>nbd_do_it were holding some lock (BKL or otherwise), we'd have big problems.
> 
> 
> Why would you want to issue an ioctl from a different process while
> nbd-client is still running?

nbd-client -d (disconnect) is the main reason -- this functionality 
would be broken if we didn't allow ioctls anymore

> Allow ioctl's while nbd_do_it is in progress is a *serious* bug.  For a

Certain ioctls, yes. There's no harm in NBD_DISCONNECT, though.

> start, if someone else clears the socket then nbd_read_stat will crash.

I agree, NBD_CLEAR_SOCK from another process while nbd_do_it is running 
would cause problems. But, if the user-level tools are coded properly, 
this is not an issue.

Perhaps your ioctl lock scheme, with an exemption for NBD_DISCONNECT 
would work?

--
Paul