[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <Z-3z0MUidedMwpLt@mini-arch>
Date: Wed, 2 Apr 2025 19:34:56 -0700
From: Stanislav Fomichev <stfomichev@...il.com>
To: Jakub Kicinski <kuba@...nel.org>
Cc: Stanislav Fomichev <sdf@...ichev.me>, netdev@...r.kernel.org,
davem@...emloft.net, edumazet@...gle.com, pabeni@...hat.com,
Cosmin Ratiu <cratiu@...dia.com>
Subject: Re: [PATCH net v5 02/11] net: hold instance lock during
NETDEV_REGISTER/UP
On 04/02, Jakub Kicinski wrote:
> On Wed, 2 Apr 2025 17:02:20 -0700 Jakub Kicinski wrote:
> > Is there a reason we don't hold the instance lock over
> > unlist_netdevice() in unregister_netdevice_many_notify()
> > but we do here? We need a separate fix for that..
No particular reason. It felt better to flip list/unlist and reg_state
'atomically', but I agree that it doesn't make much sense since the
other paths don't look at the device listed/unlisted state.
> I deleted too much here. I meant to say that we need a fix
> for netns changing while netdev_get_by_index_lock() is
> grabbing the device.
A fix for somebody grabbing the device in between two locks?
Powered by blists - more mailing lists