[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <a42eca0f-8a13-443e-a380-86609d88b4ee@flourine.local>
Date: Thu, 8 May 2025 12:06:46 +0200
From: Daniel Wagner <dwagner@...e.de>
To: Hannes Reinecke <hare@...e.de>
Cc: Daniel Wagner <wagi@...nel.org>,
James Smart <james.smart@...adcom.com>, Christoph Hellwig <hch@....de>, Sagi Grimberg <sagi@...mberg.me>,
Chaitanya Kulkarni <kch@...dia.com>, Keith Busch <kbusch@...nel.org>, linux-nvme@...ts.infradead.org,
linux-kernel@...r.kernel.org
Subject: Re: [PATCH v6 01/14] nvmet-fcloop: track ref counts for nports
On Wed, May 07, 2025 at 03:50:30PM +0200, Hannes Reinecke wrote:
> > + spin_lock_irqsave(&fcloop_lock, flags);
> > + tport = __unlink_target_port(nport);
> > spin_unlock_irqrestore(&fcloop_lock, flags);
> > - if (!nport)
> > - return -ENOENT;
> > + if (!tport) {
> > + ret = -ENOENT;
> > + goto out_nport_put;
> > + }
> > ret = __targetport_unreg(nport, tport);
>
> The lock needs to extend across both lookup and unlink,
> ie don't drop the lock in between.
No, this wont wor. There will be a nested lock in
nvmet_fc_unregister_targetport. The lock only protects the list
insert/delete/iterate operations not anything else.
FWIW, this is hasn't changed the unregister step was already done uside
the lock.
BTW, I've played with turning the spin lock into a mutex as there
doesn't seem to be any necessity to use a spin lock then we could in
theory keep the lock over the whole section but this is something I
would leave for the future.
Powered by blists - more mailing lists