Re: [PATCH v3 04/06] 9p fid refcount: add p9_fid_get/put wrappers

From: Dominique Martinet <asmadeus@codewreck.org>
To: Christian Schoenebeck <linux_oss@crudebyte.com>
Cc: Eric Van Hensbergen <ericvh@gmail.com>,
	Latchesar Ionkov <lucho@ionkov.net>,
	"David S. Miller" <davem@davemloft.net>,
	Jakub Kicinski <kuba@kernel.org>, Paolo Abeni <pabeni@redhat.com>,
	Tyler Hicks <tyhicks@linux.microsoft.com>,
	v9fs-developer@lists.sourceforge.net,
	linux-kernel@vger.kernel.org, netdev@vger.kernel.org
Subject: Re: [PATCH v3 04/06] 9p fid refcount: add p9_fid_get/put wrappers
Date: Wed, 15 Jun 2022 22:47:05 +0900	[thread overview]
Message-ID: <Yqni2QQg0uX9Lx13@codewreck.org> (raw)
In-Reply-To: <21498866.auqpVWlHDa@silver>

Christian Schoenebeck wrote on Wed, Jun 15, 2022 at 03:00:19PM +0200:
> > -	if (!fid || IS_ERR(fid)) {
> > -		pr_warn("%s (%d): Trying to clunk with invalid fid\n",
> > -			__func__, task_pid_nr(current));
> > -		dump_stack();
> > -		return 0;
> > -	}
> > -	if (!refcount_dec_and_test(&fid->count))
> > -		return 0;
> > -
> 
> I probably would have moved (and that way preserved) that sanity warning to 
> p9_fid_put(), but anyway LGTM.

The existing code was careful not to call clunk on error, but I consider
put() calls to be kind of like free in that it's better to make these
easy to call: this allowed patch 6 reworked most fs/ functions getting a
ref to just initialize fids to NULL and inconditionally call
p9_fid_put() before return.

I guess it's just a matter of preference ultimately, but I think that'll
make it a bit easier to not leak fids. Time will tell if this works :)

> Reviewed-by: Christian Schoenebeck <linux_oss@crudebyte.com>

Thanks for this and other reviews!

--
Dominique