[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20231101224855.GJ1957730@ZenIV>
Date: Wed, 1 Nov 2023 22:48:55 +0000
From: Al Viro <viro@...iv.linux.org.uk>
To: Oleg Nesterov <oleg@...hat.com>
Cc: David Howells <dhowells@...hat.com>,
Marc Dionne <marc.dionne@...istor.com>,
"David S. Miller" <davem@...emloft.net>,
Eric Dumazet <edumazet@...gle.com>,
Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
Chuck Lever <chuck.lever@...cle.com>, linux-afs@...ts.infradead.org,
netdev@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] rxrpc_find_service_conn_rcu: use read_seqbegin() rather
than read_seqbegin_or_lock()
On Wed, Nov 01, 2023 at 10:52:15PM +0100, Oleg Nesterov wrote:
> > Why would you want to force that "switch to locked on the second pass" policy
> > on every possible caller?
>
> Because this is what (I think) read_seqbegin_or_lock() is supposed to do.
> It should take the lock for writing if the lockless access failed. At least
> according to the documentation.
Not really - it's literally seqbegin or lock, depending upon what the caller
tells it... IMO the mistake in docs is the insistence on using do-while
loop for its users.
Take a look at d_walk() and try to shoehorn that into your variant. Especially
the D_WALK_NORETRY handling...
Powered by blists - more mailing lists