linux-kernel - Re: [PATCH net-next v3 3/7] net: macb: parse PHY nodes found under an MDIO node

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]

Message-ID: <89383e2e-5232-b983-ec67-565ff49821d9@microchip.com>
Date:   Mon, 27 Jul 2020 09:13:38 +0000
From:   <Codrin.Ciubotariu@...rochip.com>
To:     <f.fainelli@...il.com>, <netdev@...r.kernel.org>,
        <devicetree@...r.kernel.org>,
        <linux-arm-kernel@...ts.infradead.org>,
        <linux-kernel@...r.kernel.org>
CC:     <Nicolas.Ferre@...rochip.com>, <Claudiu.Beznea@...rochip.com>,
        <davem@...emloft.net>, <kuba@...nel.org>, <andrew@...n.ch>,
        <robh+dt@...nel.org>, <alexandre.belloni@...tlin.com>,
        <Ludovic.Desroches@...rochip.com>
Subject: Re: [PATCH net-next v3 3/7] net: macb: parse PHY nodes found under an
 MDIO node

On 24.07.2020 20:40, Florian Fainelli wrote:
> EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe
> 
> On 7/24/20 3:50 AM, Codrin Ciubotariu wrote:
>> The MACB embeds an MDIO bus controller. For this reason, the PHY nodes
>> were represented as sub-nodes in the MACB node. Generally, the
>> Ethernet controller is different than the MDIO controller, so the PHYs
>> are probed by a separate MDIO driver. Since adding the PHY nodes directly
>> under the ETH node became deprecated, we adjust the MACB driver to look
>> for an MDIO node and register the subnode MDIO devices.
>>
>> Signed-off-by: Codrin Ciubotariu <codrin.ciubotariu@...rochip.com>
>> ---
>>
>> Changes in v3:
>>   - moved the check for the mdio node at the beginnging of
>>     macb_mdiobus_register(). This way, the mdio devices will be probed even
>>     if macb is a fixed-link
>>
>> Changes in v2:
>>   - readded newline removed by mistake;
>>
>>   drivers/net/ethernet/cadence/macb_main.c | 10 ++++++++++
>>   1 file changed, 10 insertions(+)
>>
>> diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
>> index 89fe7af5e408..cb0b3637651c 100644
>> --- a/drivers/net/ethernet/cadence/macb_main.c
>> +++ b/drivers/net/ethernet/cadence/macb_main.c
>> @@ -740,6 +740,16 @@ static int macb_mii_probe(struct net_device *dev)
>>   static int macb_mdiobus_register(struct macb *bp)
>>   {
>>        struct device_node *child, *np = bp->pdev->dev.of_node;
>> +     struct device_node *mdio_node;
>> +     int ret;
>> +
>> +     /* if an MDIO node is present, it should contain the PHY nodes */
>> +     mdio_node = of_get_child_by_name(np, "mdio");
>> +     if (mdio_node) {
>> +             ret = of_mdiobus_register(bp->mii_bus, mdio_node);
>> +             of_node_put(mdio_node);
>> +             return ret;
>> +     }
> 
> This does take care of registering the MDIO bus controller when present
> as a sub-node, however if you also plan on making use of fixed-link, we
> will have already returned.
> 
>>
>>        if (of_phy_is_fixed_link(np))
>>                return mdiobus_register(bp->mii_bus);
>>
> 
> Really not sure what this is achieving, because we start off assuming
> that we have an OF driven configuration, but later on we register the
> MDIO bus with mdiobus_register() (and not of_mdiobus_register()), so no
> scanning of the MDIO bus will happen.

I agree that returning mdiobus_register() here makes no sense since 
there are no other PHY devices to scan for and probe. I can replace this 
with NULL.
The reason for fixed-link check here is to skip the following loop:
for_each_available_child_of_node(np, child)
		if (of_mdiobus_child_is_phy(child)) {
			of_node_put(child);
			return of_mdiobus_register(bp->mii_bus, np);
		}
of_mdiobus_child_is_phy() returns true when it finds the fixed-link 
node, because it has no compatible.

> 
> How does the driver currently support being provided a fixed-link
> property? Should not we at least have this pattern:
> 
>           */
>          if (of_phy_is_fixed_link(dn)) {
>                  ret = of_phy_register_fixed_link(dn);
>                  if (ret)
>                          return ret;
> 
>                  priv->phy_dn = dn;
>          }
> 
> It does not look like you are breaking anything here, because it does
> not look like this works at all.

 From what I understand, the new phylink(_create) handles the fixed-link 
case, so there is not reason to explicitly register the fixed-link.

Best regards,
Codrin

> --
> Florian
>