Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
MIME-Version: 1.0
References: <20210311072311.2969-1-xie.he.0141@gmail.com> <20210311124309.5ee0ef02@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
In-Reply-To: <20210311124309.5ee0ef02@kicinski-fedora-pc1c0hjn.dhcp.thefacebook.com>
From:   Xie He <xie.he.0141@gmail.com>
Date:   Thu, 11 Mar 2021 13:12:25 -0800
Message-ID: <CAJht_EMToKj2OeeE1fMfwAVYvhbgZpENkv0C7ac+XHnWcTe2Tg@mail.gmail.com>
Subject: Re: [PATCH net] net: lapbether: Prevent racing when checking whether
 the netif is running
To:     Jakub Kicinski <kuba@kernel.org>
Cc:     Martin Schiller <ms@dev.tdt.de>,
        "David S. Miller" <davem@davemloft.net>,
        Linux X25 <linux-x25@vger.kernel.org>,
        Linux Kernel Network Developers <netdev@vger.kernel.org>,
        LKML <linux-kernel@vger.kernel.org>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Mar 11, 2021 at 12:43 PM Jakub Kicinski <kuba@kernel.org> wrote:
>
> Is this a theoretical issues or do you see a path where it triggers?
>
> Who are the callers sending frames to a device which went down?

This is a theoretical issue. I didn't see this issue in practice.

When "__dev_queue_xmit" and "sch_direct_xmit" call
"dev_hard_start_xmit", there appears to be no locking mechanism
preventing the netif from going down while "dev_hard_start_xmit" is
doing its work.

David once confirmed in an email that a driver's "ndo_stop" function
would indeed race with its "ndo_start_xmit" function:
https://lore.kernel.org/netdev/20190520.200922.2277656639346033061.davem@davemloft.net/