netdev - Re: Problem with PHY state machine when using interrupts

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: <c81e12bc-0f89-5eab-547e-f0c1ec389ad8@free.fr>
Date:   Mon, 24 Jul 2017 23:20:55 +0200
From:   Mason <slash.tmp@...e.fr>
To:     Florian Fainelli <f.fainelli@...il.com>,
        Andrew Lunn <andrew@...n.ch>, Mans Rullgard <mans@...sr.com>
Cc:     netdev <netdev@...r.kernel.org>,
        Linux ARM <linux-arm-kernel@...ts.infradead.org>
Subject: Re: Problem with PHY state machine when using interrupts

On 24/07/2017 21:53, Florian Fainelli wrote:

> Well now that I see the possible interrupts generated, I indeed don't
> see how you can get a link down notification unless you somehow force
> the link down yourself, which would certainly happen in phy_suspend()
> when we set BMCR.pwrdwn, but that may be too late.
> 
> You should still expect the adjust_link() function to be called though
> with PHY_HALTED being set and that takes care of doing phydev->link = 0
> and netif_carrier_off(). If that still does not work, then see whether
> removing the call to phy_stop() does help (it really should).

The only functions setting phydev->state to PHY_HALTED
are phy_error() and phy_stop() AFAICT.

I am aware that when phy_state_machine() handles the
PHY_HALTED state, it will set phydev->link = 0;
and call netif_carrier_off() -- because that's where
I copied that code from.

My issue is that phy_state_machine() does not run when
I run 'ip set link dev eth0 down' from the command line.

If I'm reading the code right, phy_disconnect() actually
stops the state machine.

In interrupt mode, phy_state_machine() doesn't run
because no interrupt is generated.

In polling mode, phy_state_machine() doesn't run
because phy_disconnect() stops the state machine.

Introducing a sleep before phy_disconnect() gives
the state machine a chance to run in polling mode,
but it doesn't feel right, and doesn't fix the
other mode, which I'm using.

Looking at bcm_enet_stop() it calls phy_stop() and
phy_disconnect() just like the nb8800 driver...

I'm stumped.

Regards.