Re: [KVM PATCH v10] kvm: add support for irqfd

From: "Michael S. Tsirkin" <mst@redhat.com>
To: Gregory Haskins <ghaskins@novell.com>
Cc: kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
	avi@redhat.com, davidel@xmailserver.org, mtosatti@redhat.com
Subject: Re: [KVM PATCH v10] kvm: add support for irqfd
Date: Thu, 11 Jun 2009 16:36:43 +0300	[thread overview]
Message-ID: <20090611133643.GA16335@redhat.com> (raw)
In-Reply-To: <20090611131647.GA15743@redhat.com>

On Thu, Jun 11, 2009 at 04:16:47PM +0300, Michael S. Tsirkin wrote:
> > +
> > +	ret = file->f_op->poll(file, &irqfd->pt);
> > +	if (ret < 0)
> > +		goto fail;

Looking at it some more, we have:
struct file_operations {
....
	unsigned int (*poll) (struct file *, struct poll_table_struct *);

So the comparison above does not seem to make sense:
it seems that the return value from poll can not be negative.

Will the callback be executed if someone did a write to eventfd
before we attached it? If no, maybe we should call it here
if ret != 0.

> > +
> > +	irqfd->file = file;
> > +
> > +	mutex_lock(&kvm->lock);
> > +	list_add_tail(&irqfd->list, &kvm->irqfds);
> > +	mutex_unlock(&kvm->lock);
> > +
> > +	return 0;
> > +
> > +fail:
> > +	if (irqfd->wqh)
> > +		remove_wait_queue(irqfd->wqh, &irqfd->wait);
> 
> Why are these 2 lines here? Either we might get a callback even though
> poll failed - and then this test without lock is probably racy -
> or we can't, and then we can replace the above with BUG_ON(irqfd->wqh).
> 
> Which is it? I think the later ...
> 
> 
> > +
> > +	if (file && !IS_ERR(file))
> > +		fput(file);
> > +
> > +	kfree(irqfd);
> > +	return ret;
> > +}
> > +