[PATCH v2] net/af_packet: add multi-segment mbuf support for jumbo frames

Stephen Hemminger stephen at networkplumber.org
Mon Mar 16 17:02:03 CET 2026


On Tue, 10 Mar 2026 20:02:49 +0000
Sriram Yagnaraman <sriram.yagnaraman at ericsson.com> wrote:

> > -----Original Message-----
> > From: Stephen Hemminger <stephen at networkplumber.org>
> > Sent: Tuesday, 10 March 2026 15:03
> > To: Sriram Yagnaraman <sriram.yagnaraman at ericsson.com>
> > Cc: dev at dpdk.org; xavier.guillaume at ovhcloud.com
> > Subject: Re: [PATCH v2] net/af_packet: add multi-segment mbuf support for
> > jumbo frames
> > 
> > On Mon, 9 Mar 2026 22:02:10 +0100
> > Sriram Yagnaraman <sriram.yagnaraman at ericsson.com> wrote:
> >   
> > > Enable jumbo frame reception with default mbuf data room size by
> > > chaining multiple mbufs when packet exceeds single mbuf tailroom.
> > >
> > > The RX path now:
> > > - Copies first segment up to mbuf tailroom
> > > - Allocates and chains additional segments for remaining data
> > > - Properly maintains nb_segs, next pointers, and pkt_len
> > > - Silently drops packets on mbuf allocation failure
> > >
> > > This allows receiving 9KB jumbo frames using standard 2KB mbufs,
> > > chaining ~5 segments per jumbo packet. No need for custom mbuf pools
> > > with large data room.
> > >
> > > Advertises RTE_ETH_RX_OFFLOAD_SCATTER capability.
> > >
> > > Depends-on: series-34567 ("net/af_packet: fix MTU handling and add
> > > jumbo frame support")
> > >
> > > Note: Implementation generated with AI assistance.
> > >
> > > Signed-off-by: Sriram Yagnaraman <sriram.yagnaraman at ericsson.com>  
> > 
> > Using AI patch review finds several issues.  
> 
> Thanks for the review, I will fix them in v3.
> Perhaps, I will wait for Xavier's changes to be in main before posting the next version.
> 

You can use next-net for-main as baseline. But wait until that gets merged before resubmitting.


More information about the dev mailing list