Re: [PATCH] net: usb: r8152: Check used MAC passthrough address

[Aaron Ma <aaron.ma@canonical.com>]

On 1/5/22 15:32, Henning Schild wrote:
> Am Wed, 5 Jan 2022 08:23:55 +0100
> schrieb Henning Schild <henning.schild@siemens.com>:
> 
>> Hi Aaron,
>>
>> if this or something similar goes in, please add another patch to
>> remove the left-over defines.
>>

Sure, I will do it.

>> Am Wed,  5 Jan 2022 14:17:47 +0800
>> schrieb Aaron Ma <aaron.ma@canonical.com>:
>>
>>> When plugin multiple r8152 ethernet dongles to Lenovo Docks
>>> or USB hub, MAC passthrough address from BIOS should be
>>> checked if it had been used to avoid using on other dongles.
>>>
>>> Currently builtin r8152 on Dock still can't be identified.
>>> First detected r8152 will use the MAC passthrough address.
>>>
>>> Signed-off-by: Aaron Ma <aaron.ma@canonical.com>
>>> ---
>>>   drivers/net/usb/r8152.c | 10 ++++++++++
>>>   1 file changed, 10 insertions(+)
>>>
>>> diff --git a/drivers/net/usb/r8152.c b/drivers/net/usb/r8152.c
>>> index f9877a3e83ac..77f11b3f847b 100644
>>> --- a/drivers/net/usb/r8152.c
>>> +++ b/drivers/net/usb/r8152.c
>>> @@ -1605,6 +1605,7 @@ static int
>>> vendor_mac_passthru_addr_read(struct r8152 *tp, struct sockaddr
>>> *sa) char *mac_obj_name; acpi_object_type mac_obj_type;
>>>   	int mac_strlen;
>>> +	struct net_device *ndev;
>>>   
>>>   	if (tp->lenovo_macpassthru) {
>>>   		mac_obj_name = "\\MACA";
>>> @@ -1662,6 +1663,15 @@ static int
>>> vendor_mac_passthru_addr_read(struct r8152 *tp, struct sockaddr *sa)
>>> ret = -EINVAL; goto amacout;
>>>   	}
>>> +	rcu_read_lock();
>>> +	for_each_netdev_rcu(&init_net, ndev) {
>>> +		if (strncmp(buf, ndev->dev_addr, 6) == 0) {
>>> +			rcu_read_unlock();
>>> +			goto amacout;
>>
>> Since the original PCI netdev will always be there, that would disable
>> inheritance would it not?
>> I guess a strncmp(MODULE_NAME, info->driver, strlen(MODULE_NAME)) is
>> needed as well.
>>

PCI ethernet could be a builtin one on dock since there will be TBT4 dock.

>> Maybe leave here with
>> netif_info()
>>

Not good to print in rcu lock.

>> And move the whole block up, we can skip the whole ACPI story if we
>> find the MAC busy.
> 
> That is wrong, need to know that MAC so can not move up too much. But
> maybe above the is_valid_ether_addr

The MAC passthough address is read from ACPI.
ACPI read only happens once during r8152 driver probe.
To keep the lock less time, do it after is_valid_ether_addr.

> 
> Henning
> 
>>> +		}
>>> +	}
>>> +	rcu_read_unlock();
>>
>> Not sure if this function is guaranteed to only run once at a time,
>> otherwise i think that is a race. Multiple instances could make it to
>> this very point at the same time.
>>

Run once for one device.
So add a safe lock.

Aaron

>> Henning
>>
>>>   	memcpy(sa->sa_data, buf, 6);
>>>   	netif_info(tp, probe, tp->netdev,
>>>   		   "Using pass-thru MAC addr %pM\n", sa->sa_data);
>>>   
>>
>