Re: [2/2] virtiofs: Improve error handling in virtio_fs_get_tree()

From: Matthew Wilcox <willy@infradead.org>
To: Markus Elfring <Markus.Elfring@web.de>
Cc: virtualization@lists.linux.dev, linux-fsdevel@vger.kernel.org,
	kernel-janitors@vger.kernel.org,
	Miklos Szeredi <miklos@szeredi.hu>,
	Stefan Hajnoczi <stefanha@redhat.com>,
	Vivek Goyal <vgoyal@redhat.com>,
	LKML <linux-kernel@vger.kernel.org>
Subject: Re: [2/2] virtiofs: Improve error handling in virtio_fs_get_tree()
Date: Tue, 2 Jan 2024 16:28:44 +0000	[thread overview]
Message-ID: <ZZQ5vKRcq9kkQxSD@casper.infradead.org> (raw)
In-Reply-To: <9b27d89d-c410-4898-b801-00d2a00fb693@web.de>

On Tue, Jan 02, 2024 at 11:47:38AM +0100, Markus Elfring wrote:
> > Do you consider more clarity in your argumentation?
> 
> It is probably clear that the function call “kfree(NULL)” does not perform
> data processing which is really useful for the caller.
> Such a call is kept in some cases because programmers did not like to invest
> development resources for its avoidance.

on the contrary, it is extremely useful for callers to not have to perform
the NULL check themselves.  It also mirrors userspace where free(NULL)
is valid according to ISO/ANSI C, so eases the transition for programmers
who are coming from userspace.  It costs nothing in the implementation
as it is part of the check for the ZERO_PTR.

And from a practical point of view, we can't take it out now.  We can
never find all the places which assume the current behaviour.  So since
we must keep kfree(NULL) working, we should take advantage of that to
simplify users.