linux-kernel - Re: [RFC PATCH net] net: phy: Fix "link partner" information disappear issue

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: <670bd6ac-54ee-ecfa-c108-fcf48a3a7dc8@huawei.com>
Date:   Fri, 11 Oct 2019 13:55:30 +0800
From:   Yonglong Liu <liuyonglong@...wei.com>
To:     Heiner Kallweit <hkallweit1@...il.com>, <davem@...emloft.net>,
        <andrew@...n.ch>, Florian Fainelli <f.fainelli@...il.com>
CC:     <netdev@...r.kernel.org>, <linux-kernel@...r.kernel.org>,
        <linuxarm@...wei.com>, <salil.mehta@...wei.com>,
        <yisen.zhuang@...wei.com>, <shiju.jose@...wei.com>
Subject: Re: [RFC PATCH net] net: phy: Fix "link partner" information
 disappear issue



On 2019/10/11 3:17, Heiner Kallweit wrote:
> On 10.10.2019 11:30, Yonglong Liu wrote:
>> Some drivers just call phy_ethtool_ksettings_set() to set the
>> links, for those phy drivers that use genphy_read_status(), if
>> autoneg is on, and the link is up, than execute "ethtool -s
>> ethx autoneg on" will cause "link partner" information disappear.
>>
>> The call trace is phy_ethtool_ksettings_set()->phy_start_aneg()
>> ->linkmode_zero(phydev->lp_advertising)->genphy_read_status(),
>> the link didn't change, so genphy_read_status() just return, and
>> phydev->lp_advertising is zero now.
>>
> I think that clearing link partner advertising info in
> phy_start_aneg() is questionable. If advertising doesn't change
> then phy_config_aneg() basically is a no-op. Instead we may have
> to clear the link partner advertising info in genphy_read_lpa()
> if aneg is disabled or aneg isn't completed (basically the same
> as in genphy_c45_read_lpa()). Something like:
> 
> if (!phydev->autoneg_complete) { /* also covers case that aneg is disabled */
> 	linkmode_zero(phydev->lp_advertising);
> } else if (phydev->autoneg == AUTONEG_ENABLE) {
> 	...
> }
> 

If clear the link partner advertising info in genphy_read_lpa() and
genphy_c45_read_lpa(), for the drivers that use genphy_read_status()
is ok, but for those drivers that use there own read_status() may
have problem, like aqr_read_status(), it will update lp_advertising
first, and than call genphy_c45_read_status(), so will cause
lp_advertising lost.

Another question, please see genphy_c45_read_status(), if clear the
link partner advertising info in genphy_c45_read_lpa(), if autoneg is
off, phydev->lp_advertising will not clear.

>> This patch call genphy_read_lpa() before the link state judgement
>> to fix this problem.
>>
>> Fixes: 88d6272acaaa ("net: phy: avoid unneeded MDIO reads in genphy_read_status")
>> Signed-off-by: Yonglong Liu <liuyonglong@...wei.com>
>> ---
>>  drivers/net/phy/phy_device.c | 8 ++++----
>>  1 file changed, 4 insertions(+), 4 deletions(-)
>>
>> diff --git a/drivers/net/phy/phy_device.c b/drivers/net/phy/phy_device.c
>> index 9d2bbb1..ef3073c 100644
>> --- a/drivers/net/phy/phy_device.c
>> +++ b/drivers/net/phy/phy_device.c
>> @@ -1839,6 +1839,10 @@ int genphy_read_status(struct phy_device *phydev)
>>  	if (err)
>>  		return err;
>>  
>> +	err = genphy_read_lpa(phydev);
>> +	if (err < 0)
>> +		return err;
>> +
>>  	/* why bother the PHY if nothing can have changed */
>>  	if (phydev->autoneg == AUTONEG_ENABLE && old_link && phydev->link)
>>  		return 0;
>> @@ -1848,10 +1852,6 @@ int genphy_read_status(struct phy_device *phydev)
>>  	phydev->pause = 0;
>>  	phydev->asym_pause = 0;
>>  
>> -	err = genphy_read_lpa(phydev);
>> -	if (err < 0)
>> -		return err;
>> -
>>  	if (phydev->autoneg == AUTONEG_ENABLE && phydev->autoneg_complete) {
>>  		phy_resolve_aneg_linkmode(phydev);
>>  	} else if (phydev->autoneg == AUTONEG_DISABLE) {
>>
> 
> 
> .
>