Re: [PATCH][RFC] sigurg/sigio cleanup for 2.5.31

[James Morris <jmorris@intercode.com.au>]

On Thu, 15 Aug 2002, Matthew Wilcox wrote:

> In general, this is good... I think it could be better:
> 
> > +	lock_kernel();
> > +	error = f_setown(filp, current->pid);
> > +	unlock_kernel();
> 
> There are a lot of these, and you even batch it up as sock_setown()
> later.  May I suggest renaming f_setown to __setown and sock_setown
> to f_setown?

Sounds like a good idea.

> this one's particularly silly -- now you've done the good job of wrapping
> the security_ops up inside f_setown this can simply be:
> 
> 			lock_kernel();
> 			err = f_setown(filp, arg);
> 			unlock_kernel();
> 			break;

Yep.

> Might make more sense to refactor as:
> 
> void sk_send_sigurg(struct sock *sk)
> {
> 	if (!sk->socket || !sk->socket->file)
> 		return;
> 	if (send_sigurg(&sk->socket->file->f_owner))
> 		sk_wake_async(sk, 3, POLL_PRI);
> }
> 

Possibly.  I guess it's up to the networking guys -- is there any point in 
keeping these separate?  I can't see any with the current code.

Thanks for the feedback.


- James
-- 
James Morris
<jmorris@intercode.com.au>