[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20180106215347.GB8258@zzz.localdomain>
Date: Sat, 6 Jan 2018 13:53:47 -0800
From: Eric Biggers <ebiggers3@...il.com>
To: Al Viro <viro@...IV.linux.org.uk>
Cc: linux-fsdevel@...r.kernel.org, linux-kernel@...r.kernel.org,
Andrew Morton <akpm@...ux-foundation.org>,
Davide Libenzi <davidel@...ilserver.org>
Subject: Re: [PATCH 0/3] eventfd: clean up unneeded cruft
On Sat, Jan 06, 2018 at 07:00:48PM +0000, Al Viro wrote:
> On Sat, Jan 06, 2018 at 06:46:19PM +0000, Al Viro wrote:
> > On Sat, Jan 06, 2018 at 09:45:41AM -0800, Eric Biggers wrote:
> > > This series removes some cruft (mainly exported functions) from
> > > fs/eventfd.c which seems to have been added while the KVM irqfd feature
> > > was under development some years ago, but was never actually used.
> >
> > Applied.
>
>
> Said that, is there any reason for keeping e.g. vhost_dev->log_file around?
> We already keep ->log_ctx, so... Ditto for vq->call and vq->error,
> and if we did it for vq->kick as well, we'd be able to get rid of
> separate eventfd_fget() as well.
We can remove vhost_dev->log_file, vq->call, and vq->error since for those only
the eventfd_ctx is needed (they are only used to send notifications). But
vq->kick is different because it doesn't use the eventfd_ctx but rather
->poll()s the file. I'll probably leave that part as-is, but perhaps the
VHOST_SET_VRING_KICK ioctl should be changed to allow any file with a ->poll()
method, since it doesn't actually use any eventfd-specific functionality of the
file.
Eric
Powered by blists - more mailing lists