[openib-general] Re: [PATCH] signal driven io support for ib_uverbs

[Gleb Natapov]

On Wed, Jul 06, 2005 at 10:38:46AM -0700, Roland Dreier wrote:
> Thanks for adding this.  A few questions:
> 
>     > +static int ib_uverbs_event_async (int fd, struct file *filp, int on)
>     > +{
>     > +	int ret;
>     > +	struct ib_uverbs_event_file *file = filp->private_data;
>     > +
>     > +	if (!file)
>     > +		return -EBADFD;
> 
> Can this function ever be called on a file with private_data not set?
> Why do we need this check?
I don't think it can, but it is slow path an check doesn't hurt.
> 
>     > +
>     > +	if ((ret = fasync_helper(fd, filp, on, &file->fasync)) < 0)
>     > +		return ret;
>     > +
>     > +	if (on)
>     > +		ret = f_setown(filp, current->pid, 0);
> 
> Why do we need the f_setown?  Doesn't the core kernel already handle
> this for us?
Same here.

> 
>     > +	kill_fasync (&file->async_file.fasync, SIGIO, POLL_IN);
>  
> We want this in both ib_uverbs_comp_handler() and also
> ib_uverbs_async_handler() too, right?
Yes, of cause. I am using signals for async events handling so I missed this
one.

--
			Gleb.