RE: [PATCH] net: use hardware buffer pool to allocate skb

["Jiafei.Pan@freescale.com" <Jiafei.Pan@freescale.com>]

> -----Original Message-----
> From: Eric Dumazet [mailto:eric.dumazet@gmail.com]
> Sent: Wednesday, October 15, 2014 5:16 PM
> To: Pan Jiafei-B37022
> Cc: David Miller; jkosina@suse.cz; netdev@vger.kernel.org; Li Yang-Leo-R58472;
> linux-doc@vger.kernel.org
> Subject: Re: [PATCH] net: use hardware buffer pool to allocate skb
> 
> On Wed, 2014-10-15 at 05:34 +0000, Jiafei.Pan@freescale.com wrote:
> 
> > Yes, for this matter, in order to do this we should modify the Ethernet
> > drivers. For example, driver A want to driver B, C, D.. to support driver
> > A's Hardware block access functions, so we have to modify driver B, C, D...
> > It will be so complex for this matter.
> >
> > But by using my patch, I just modify a Ethernet device (I don't care
> > Which driver it is used) flag in driver A in order to implement this
> > Ethernet device using hardware block access functions provided by
> > Driver A.
> 
> We care a lot of all the bugs added by your patches. You have little
> idea of how many of them were added. We do not want to spend days of
> work explaining everything or fixing all the details for you.
> 
> Carefully read net/core/skbuff.c, net/core/dev.c, GRO layer, you'll see
> how many spots you missed.
> 
> You cannot control how skbs are cooked before reaching your driver
> ndo_start_xmit(). You are not going to add hooks in UDP , TCP, or other
> drivers RX path. This would be absolutely insane.
> 

Thanks for your comments and suggestion. In my case, I want to build skb
from hardware block specified memory, I only can see two ways, one is modified
net card driver replace common skb allocation function with my specially
functions, another way is to hack common skb allocation function in which
redirect to my specially functions. My patch is just for the second way.
Except these two ways, would you please give me some advice for some other
ways for my case? Thanks.

> Trying to control how skb are cooked in RX path is absolutely something
> drivers do, using page frags that are read-only by all the stack.
> 
> Fix your driver to use existing infra, your suggestion is not going to
> be accepted.
>