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 for Android: free password hash cracker in your pocket
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date:   Thu, 17 Jan 2019 19:55:54 +0100
From:   Heiner Kallweit <hkallweit1@...il.com>
To:     Andrew Lunn <andrew@...n.ch>
Cc:     Florian Fainelli <f.fainelli@...il.com>,
        David Miller <davem@...emloft.net>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: [PATCH v2 net-next 1/3] net: phy: check that PHY is stopped when
 entering phy_disconnect

On 17.01.2019 07:19, Heiner Kallweit wrote:
> On 16.01.2019 22:48, Andrew Lunn wrote:
>> On Wed, Jan 16, 2019 at 09:25:15PM +0100, Heiner Kallweit wrote:
>>> Every driver should have called phy_stop() before calling
>>> phy_disconnect(). Let's check for this and ensure PHY is stopped
>>> when starting with the actual work in phy_disconnect().
>>
>> Hi Heiner
>>
>> Looking at the patch, i think why must the MAC driver call phy_stop()
>> before phy_disconnect()? It keeps is symmetrical, you need
>> phy_connect() and then phy_start(). But if the core can detect that
>> phy_stop() has not been called, and can call phy_stop() when needed,
>> we can probably simplify the MAC drivers by removing many of the
>> phy_stop() calls.
>>
>> I think it might come down to where the phy_connect()/phy_disconnect()
>> is performed. Sometimes it is in probe()/remove(), sometimes it is in
>> open()/close(). If phy_disconnect() is in remove(), phy_stop() is
>> needed in close(). But if phy_disconnect() is called in close() the
>> phy_stop() could be skipped?
>>
> Right, there may be cases where this is possible. However typically I
> see the following in network drivers in close():
> - first different things are stopped / cleaned up, where order is
>   critical (stopping PHY, tx queues, ..)
> - then resources are released (free interrupt, ..)
> Therefore I assume that in most cases the split to phy_stop() and
> phy_disconnect() is needed.
> 
After thinking about it again I think I'll go with the proposed
approach to leave it to the driver whether he wants to call phy_stop()
separately or not. So I will just remove the warning.

>> Before we start adding warning, we probably should first document the
>> expectations. Documentation/networking/phy.txt seems like a good
>> place.
>>
> Indeed, that's something we should do.
> 
>> 	Andrew
>>
> Heiner
> 

Powered by blists - more mailing lists