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;
Date:   Thu, 10 Sep 2020 12:20:33 -0700 (PDT)
Message-Id: <20200910.122033.2205909020498878557.davem@davemloft.net>
To:     geert+renesas@glider.be
Cc:     hkallweit1@gmail.com, f.fainelli@gmail.com, andrew@lunn.ch,
        kuba@kernel.org, gaku.inami.xh@renesas.com,
        yoshihiro.shimoda.uh@renesas.com, netdev@vger.kernel.org,
        linux-kernel@vger.kernel.org
Subject: Re: [PATCH] Revert "net: linkwatch: add check for netdevice being
 present to linkwatch_do_dev"
From:   David Miller <davem@davemloft.net>
In-Reply-To: <20200901150237.15302-1-geert+renesas@glider.be>
References: <20200901150237.15302-1-geert+renesas@glider.be>
Mime-Version: 1.0
Content-Type: Text/Plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

From: Geert Uytterhoeven <geert+renesas@glider.be>
Date: Tue,  1 Sep 2020 17:02:37 +0200

> This reverts commit 124eee3f6955f7aa19b9e6ff5c9b6d37cb3d1e2c.
> 
> Inami-san reported that this commit breaks bridge support in a Xen
> environment, and that reverting it fixes this.
> 
> During system resume, bridge ports are no longer enabled, as that relies
> on the receipt of the NETDEV_CHANGE notification.  This notification is
> not sent, as netdev_state_change() is no longer called.
> 
> Note that the condition this commit intended to fix never existed
> upstream, as the patch triggering it and referenced in the commit was
> never applied upstream.  Hence I can confirm s2ram on r8a73a4/ape6evm
> and sh73a0/kzm9g works fine before/after this revert.
> 
> Reported-by Gaku Inami <gaku.inami.xh@renesas.com>
> Signed-off-by: Geert Uytterhoeven <geert+renesas@glider.be>

Maybe you cannot reproduce it, but the problem is there and it still
looks very real to me.

netdev_state_change() does two things:

1) Emit the NETDEV_CHANGE notification

2) Emit an rtmsg_ifinfo() netlink message, which in turn tries to access
   the device statistics via ->ndo_get_stats*().

It is absolutely wrong to do #2 when netif_device_present() is false.

So I cannot apply this patch as-is, sorry.