netdev - Re: [PATCH] [RFC] net: phy: phy drivers should not set SUPPORTED_Pause or SUPPORTED_Asym

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] [day] [month] [year] [list]

Date:   Wed, 9 Nov 2016 09:55:06 -0800
From:   Florian Fainelli <f.fainelli@...il.com>
To:     Timur Tabi <timur@...eaurora.org>,
        David Miller <davem@...emloft.net>, netdev@...r.kernel.org
Subject: Re: [PATCH] [RFC] net: phy: phy drivers should not set
 SUPPORTED_Pause or SUPPORTED_Asym_Pause

On 11/09/2016 09:34 AM, Timur Tabi wrote:
> On 11/09/2016 11:06 AM, Florian Fainelli wrote:
> 
>>> 1) PHY drivers and/or phylib sets the SUPPORTED_Pause |
>>> SUPPORTED_AsymPause bits in phydev->supported.  This indicates that the
>>> PHY supports pause frames.
>>
>> Agreed.
> 
> So I'm still not sure *where* phylib should set the bits in
> phydev->supported.  I previously thought that phy_sanitize_settings()
> would be a candidate, but that function is called only when
> autonegotiation is disabled.
> 
> So the only thing I can think of is to do this:
> 
> /* A mapping of all SUPPORTED settings to speed/duplex */
> static const struct phy_setting settings[] = {
>     {
>         .speed = SPEED_10000,
>         .duplex = DUPLEX_FULL,
> -        .setting = SUPPORTED_10000baseKR_Full,
> +        .setting = SUPPORTED_10000baseKR_Full |
> +            SUPPORTED_Pause | SUPPORTED_Asym_Pause,
> 
> ... for all of the entries in settings[]
> 
> But this seems excessive, and I don't know if all of these SPEED_xxx
> standards actually support pause frames.

You could set these bits in phy_probe() and later let the
phy_sanitize_settings() make sure that these bits do not get advertised
unless we have a full duplex link.

> 
>>> >3) When the link state changes, the MAC driver checks
>>> >phydev->advertising, and if the bits are set, then it enables those
>>> >features in the MAC.
> 
>> Almost, we need to see what the link partner advertised, by looking at
>> phydev->pause / phydev->asym_pause and checking the local knobs whether
>> flow control can be enabled, but other than that, this looks correct
>> to me.
> 
> Yes, that's what I meant to write.  Fortunately, my EMAC driver already
> does this.

OK, just making sure ;)
-- 
Florian