netdev - Re: [net-next PATCH 2/2] net: phylink: Fix issues with link balancing w/ BMC present

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <Z//k4PUTWXo3+IBh@shell.armlinux.org.uk>
Date: Wed, 16 Apr 2025 18:12:00 +0100
From: "Russell King (Oracle)" <linux@...linux.org.uk>
To: Alexander Duyck <alexander.duyck@...il.com>
Cc: netdev@...r.kernel.org, andrew@...n.ch, hkallweit1@...il.com,
	davem@...emloft.net, kuba@...nel.org, pabeni@...hat.com
Subject: Re: [net-next PATCH 2/2] net: phylink: Fix issues with link
 balancing w/ BMC present

On Wed, Apr 16, 2025 at 05:01:09PM +0100, Russell King (Oracle) wrote:
> On Wed, Apr 16, 2025 at 08:29:00AM -0700, Alexander Duyck wrote:
> > From: Alexander Duyck <alexanderduyck@...com>
> > 
> > This change is meant to address the fact that there are link imbalances
> > introduced when using phylink on a system with a BMC. Specifically there
> > are two issues.
> > 
> > The first issue is that if we lose link after the first call to
> > phylink_start but before it gets to the phylink_resolve we will end up with
> > the phylink interface assuming the link was always down and not calling
> > phylink_link_down resulting in a stuck interface.
> 
> That is intentional.
> 
> phylink strictly orders .mac_link_down and .mac_link_up, and starts from
> an initial position that the link _will_ be considered to be down. So,
> it is intentional that .mac_link_down will _never_ be called after
> phylink_start().
> 
> > The second issue is that when a BMC is present we are currently forcing the
> > link down. This results in us bouncing the link for a fraction of a second
> > and that will result in dropped packets for the BMC.
> 
> ... but you don't explain how that happens.
> 
> > The third issue is just an extra "Link Down" message that is seen when
> > calling phylink_resume. This is addressed by identifying that the link
> > isn't balanced and just not displaying the down message in such a case.
> 
> Hmm, this one is an error, but is not as simple as "don't print the
> message" as it results in a violation of the rule I mentioned above.
> We need phylink_suspend() to record the state of the link at that
> point, and avoid calling phylink_link_down() if the link was down
> prior to suspend.

Okay, confirmed on nvidia Jetson Xavier NX:

[   11.838132] dwc-eth-dwmac 2490000.ethernet eth0: Adding VLAN ID 0 is not supported
[   15.299757] dwc-eth-dwmac 2490000.ethernet eth0: Link is Up - 1Gbps/Full - flow control rx/tx

LAN cable was unplugged:
[   50.436587] dwc-eth-dwmac 2490000.ethernet eth0: Link is Down

Then the system was suspended using rtcwake for 3 seconds:

[   54.736849] dwc-eth-dwmac 2490000.ethernet eth0: No Safety Features support found
[   54.736898] dwc-eth-dwmac 2490000.ethernet eth0: IEEE 1588-2008 Advanced Timestamp supported
[   54.741078] dwc-eth-dwmac 2490000.ethernet eth0: Link is Down

This shouldn't happen. With the patch I posted, this second "Link is Down"
message is not printed, and .mac_link_down() will not be called.

-- 
RMK's Patch system: https://www.armlinux.org.uk/developer/patches/
FTTP is here! 80Mbps down 10Mbps up. Decent connectivity at last!