[PATCH RFC] net: add helper eth_set_protocol

[PATCH RFC] net: add helper eth_set_protocol

[Heiner Kallweit]

In all cases I've seen eth_type_trans() is used as in the new helper.
Biggest benefit is improved readability when replacing statements like
the following:
desc->skb->protocol = eth_type_trans(desc->skb, priv->dev);

Coccinelle tells me that using the new helper tree-wide would touch
313 files. Therefore I'd like to check for feedback before bothering
100+ maintainers.

Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
---
 include/linux/etherdevice.h | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/include/linux/etherdevice.h b/include/linux/etherdevice.h
index 2e5debc03..c7f89b1bf 100644
--- a/include/linux/etherdevice.h
+++ b/include/linux/etherdevice.h
@@ -64,6 +64,11 @@ static const u8 eth_reserved_addr_base[ETH_ALEN] __aligned(2) =
 { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
 #define eth_stp_addr eth_reserved_addr_base
 
+static inline void eth_set_protocol(struct sk_buff *skb, struct net_device *dev)
+{
+	skb->protocol = eth_type_trans(skb, dev);
+}
+
 /**
  * is_link_local_ether_addr - Determine if given Ethernet address is link-local
  * @addr: Pointer to a six-byte array containing the Ethernet address
-- 
2.28.0

Re: [PATCH RFC] net: add helper eth_set_protocol

[Jakub Kicinski]

On Tue, 6 Oct 2020 22:10:00 +0200 Heiner Kallweit wrote:
> In all cases I've seen eth_type_trans() is used as in the new helper.
> Biggest benefit is improved readability when replacing statements like
> the following:
> desc->skb->protocol = eth_type_trans(desc->skb, priv->dev);
> 
> Coccinelle tells me that using the new helper tree-wide would touch
> 313 files. Therefore I'd like to check for feedback before bothering
> 100+ maintainers.
> 
> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>

FWIW in case you're planning to start sending conversion patches..

I'm not 100% sold on this. Maybe it's because I'm used to the call 
as it is. I don't feel like eth_set_protocol() expresses what
eth_type_trans() does well enough. Besides there's a whole bunch of
*_type_trans calls for different (old) L2s, are we going to leave those
as they are?

> diff --git a/include/linux/etherdevice.h b/include/linux/etherdevice.h
> index 2e5debc03..c7f89b1bf 100644
> --- a/include/linux/etherdevice.h
> +++ b/include/linux/etherdevice.h
> @@ -64,6 +64,11 @@ static const u8 eth_reserved_addr_base[ETH_ALEN] __aligned(2) =
>  { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
>  #define eth_stp_addr eth_reserved_addr_base
>  
> +static inline void eth_set_protocol(struct sk_buff *skb, struct net_device *dev)
> +{
> +	skb->protocol = eth_type_trans(skb, dev);
> +}
> +
>  /**
>   * is_link_local_ether_addr - Determine if given Ethernet address is link-local
>   * @addr: Pointer to a six-byte array containing the Ethernet address

Re: [PATCH RFC] net: add helper eth_set_protocol

[Heiner Kallweit]

On 12.10.2020 00:13, Jakub Kicinski wrote:
> On Tue, 6 Oct 2020 22:10:00 +0200 Heiner Kallweit wrote:
>> In all cases I've seen eth_type_trans() is used as in the new helper.
>> Biggest benefit is improved readability when replacing statements like
>> the following:
>> desc->skb->protocol = eth_type_trans(desc->skb, priv->dev);
>>
>> Coccinelle tells me that using the new helper tree-wide would touch
>> 313 files. Therefore I'd like to check for feedback before bothering
>> 100+ maintainers.
>>
>> Signed-off-by: Heiner Kallweit <hkallweit1@gmail.com>
> 
> FWIW in case you're planning to start sending conversion patches..
> 
> I'm not 100% sold on this. Maybe it's because I'm used to the call 
> as it is. I don't feel like eth_set_protocol() expresses what
> eth_type_trans() does well enough. Besides there's a whole bunch of
> *_type_trans calls for different (old) L2s, are we going to leave those
> as they are?
> 
Didn't look at the other *_type_trans functions yet. Seems like at least
eth_type_trans() has old functionality that I didn't expect. Based on
name and usage of the function I was under the assumption that it treats
the skb as const. But it does:
skb->dev = dev;
skb_reset_mac_header(skb);
Such (intentional) side effects are not too nice. But as this function is
used in hundreds of places, I think a refactoring would need significant
effort. For now I tend to leave the situation as it is.

>> diff --git a/include/linux/etherdevice.h b/include/linux/etherdevice.h
>> index 2e5debc03..c7f89b1bf 100644
>> --- a/include/linux/etherdevice.h
>> +++ b/include/linux/etherdevice.h
>> @@ -64,6 +64,11 @@ static const u8 eth_reserved_addr_base[ETH_ALEN] __aligned(2) =
>>  { 0x01, 0x80, 0xc2, 0x00, 0x00, 0x00 };
>>  #define eth_stp_addr eth_reserved_addr_base
>>  
>> +static inline void eth_set_protocol(struct sk_buff *skb, struct net_device *dev)
>> +{
>> +	skb->protocol = eth_type_trans(skb, dev);
>> +}
>> +
>>  /**
>>   * is_link_local_ether_addr - Determine if given Ethernet address is link-local
>>   * @addr: Pointer to a six-byte array containing the Ethernet address
>