Kernel Janitor TODO: dev_kfree_skb

Kernel Janitor TODO: dev_kfree_skb

[John Daiker]

I was running through the kernel janitors TODO list 
(http://janitor.kernelnewbies.org/TODO) and thought I'd try to find 
something easy for my first attempt at kernel hacking/maintenance.

The "Balancing Function" header suggests that all net_device interrupt 
functions should use dev_kfree_skb_irq as opposed to dev_kfree_skb.
Should we ever be calling dev_kfree_skb directly, or would the better 
practice be to always call dev_kfree_skb_any (thus letting the kernel 
decide to call dev_kfree_skb or dev_kfree_skb_irq)?

dev_kfree_skb_any is defined in 'net/core/dev.c:1118'

John Daiker

Re: Kernel Janitor TODO: dev_kfree_skb

[Herbert Xu]

John Daiker <jdaiker@osdl.org> wrote:
> 
> The "Balancing Function" header suggests that all net_device interrupt 
> functions should use dev_kfree_skb_irq as opposed to dev_kfree_skb.
> Should we ever be calling dev_kfree_skb directly, or would the better 
> practice be to always call dev_kfree_skb_any (thus letting the kernel 
> decide to call dev_kfree_skb or dev_kfree_skb_irq)?

It's best if you write code in a way so that you always know whether it's
IRQ context or not.  You should only use dev_kfree_skb_any when you fail
to write code like that.

Now once you've achieved that, then the next objective is to strive to
not have to call kfree_skb in IRQ context at all.  Have a look at the TG3
driver for an example where IRQ context is kept to a minimum.

Cheers,
-- 
Visit Openswan at http://www.openswan.org/
Email: Herbert Xu ~{PmV>HI~} <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt