[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Yqd0OJupifpplENB@slm.duckdns.org>
Date: Mon, 13 Jun 2022 07:30:32 -1000
From: Tejun Heo <tj@...nel.org>
To: Imran Khan <imran.f.khan@...cle.com>
Cc: viro@...iv.linux.org.uk, gregkh@...uxfoundation.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH RESEND v4 1/4] kernfs: make ->attr.open RCU protected.
On Mon, Jun 13, 2022 at 01:56:43PM +1000, Imran Khan wrote:
> Agree. I did over think this and went for the safest interface that I could
> think of in each of the use cases. I will remove
> kernfs_check_open_node_protected and use kernfs_deref_open_node_protected in its
> place as well. This will cover all accesses under kernfs_open_file_mutex.
>
> But we will still need kernfs_deref_open_node for cases where
> !list_empty(&of->list) ensures safe access of ->attr.open and where we can't
> ensure holding of kernfs_open_file_mutex. So we will need 2 deref accessors.
> Right? Just asking this because you mentioned above to come up with one deref
> accessor that can be used in all three use cases
>
> Please let me if this sounds okay. I can send updated patch-set with these
> changes in place.
Just merge all three into one accessor. You can list both the !list_empty
condition and lock held conditions on the same rcu_dereference_check() call.
Thanks.
--
tejun
Powered by blists - more mailing lists