[PATCH v2 bpf-next] xdp: report rx queue index in xdp_frame

[PATCH v2 bpf-next] xdp: report rx queue index in xdp_frame

[Lorenzo Bianconi]

Report rx queue index in xdp_frame according to the xdp_buff xdp_rxq_info
pointer. xdp_frame queue_index is currently used in cpumap code to convert
the xdp_frame into a xdp_buff and allow the ebpf program attached to the
map entry to differentiate traffic according to the receiving hw queue.
xdp_frame size is not increased adding queue_index since an alignment
padding in the structure is used to insert queue_index field.

Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
---
Changes since v1:
- rebase on top of bpf-next
---
 include/net/xdp.h   | 2 ++
 kernel/bpf/cpumap.c | 2 +-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/include/net/xdp.h b/include/net/xdp.h
index 04c852c7a77f..3567866b0af5 100644
--- a/include/net/xdp.h
+++ b/include/net/xdp.h
@@ -172,6 +172,7 @@ struct xdp_frame {
 	struct xdp_mem_info mem;
 	struct net_device *dev_rx; /* used by cpumap */
 	u32 flags; /* supported values defined in xdp_buff_flags */
+	u32 queue_index;
 };
 
 static __always_inline bool xdp_frame_has_frags(struct xdp_frame *frame)
@@ -301,6 +302,7 @@ struct xdp_frame *xdp_convert_buff_to_frame(struct xdp_buff *xdp)
 
 	/* rxq only valid until napi_schedule ends, convert to xdp_mem_info */
 	xdp_frame->mem = xdp->rxq->mem;
+	xdp_frame->queue_index = xdp->rxq->queue_index;
 
 	return xdp_frame;
 }
diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
index b5ba34ddd4b6..48003450c98c 100644
--- a/kernel/bpf/cpumap.c
+++ b/kernel/bpf/cpumap.c
@@ -228,7 +228,7 @@ static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu,
 
 		rxq.dev = xdpf->dev_rx;
 		rxq.mem = xdpf->mem;
-		/* TODO: report queue_index to xdp_rxq_info */
+		rxq.queue_index = xdpf->queue_index;
 
 		xdp_convert_frame_to_buff(xdpf, &xdp);
 
-- 
2.37.2

Re: [PATCH v2 bpf-next] xdp: report rx queue index in xdp_frame

[Jesper Dangaard Brouer]

On 17/08/2022 09.40, Lorenzo Bianconi wrote:
> Report rx queue index in xdp_frame according to the xdp_buff xdp_rxq_info
> pointer. xdp_frame queue_index is currently used in cpumap code to convert
> the xdp_frame into a xdp_buff and allow the ebpf program attached to the
> map entry to differentiate traffic according to the receiving hw queue.
> xdp_frame size is not increased adding queue_index since an alignment
> padding in the structure is used to insert queue_index field.
> 
> Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>

(Sorry, I replied to v1 and not this v2.)

I'm still unsure about this change, because the XDP-hints will also
contain the rx_queue number.  And placing it in XDP-hints automatically
makes it avail for AF_XDP consumers.

I do think it is relevant for the BPF-prog to get access to the rx_queue
index, because it can be used for scaling the workload.


> ---
> Changes since v1:
> - rebase on top of bpf-next
> ---
>   include/net/xdp.h   | 2 ++
>   kernel/bpf/cpumap.c | 2 +-
>   2 files changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/include/net/xdp.h b/include/net/xdp.h
> index 04c852c7a77f..3567866b0af5 100644
> --- a/include/net/xdp.h
> +++ b/include/net/xdp.h
> @@ -172,6 +172,7 @@ struct xdp_frame {
>   	struct xdp_mem_info mem;
>   	struct net_device *dev_rx; /* used by cpumap */
>   	u32 flags; /* supported values defined in xdp_buff_flags */
> +	u32 queue_index;
>   };
>   
>   static __always_inline bool xdp_frame_has_frags(struct xdp_frame *frame)
> @@ -301,6 +302,7 @@ struct xdp_frame *xdp_convert_buff_to_frame(struct xdp_buff *xdp)
>   
>   	/* rxq only valid until napi_schedule ends, convert to xdp_mem_info */
>   	xdp_frame->mem = xdp->rxq->mem;
> +	xdp_frame->queue_index = xdp->rxq->queue_index;
>   
>   	return xdp_frame;
>   }
> diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
> index b5ba34ddd4b6..48003450c98c 100644
> --- a/kernel/bpf/cpumap.c
> +++ b/kernel/bpf/cpumap.c
> @@ -228,7 +228,7 @@ static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu,
>   
>   		rxq.dev = xdpf->dev_rx;
>   		rxq.mem = xdpf->mem;
> -		/* TODO: report queue_index to xdp_rxq_info */
> +		rxq.queue_index = xdpf->queue_index;
>   
>   		xdp_convert_frame_to_buff(xdpf, &xdp);
>   

Re: [PATCH v2 bpf-next] xdp: report rx queue index in xdp_frame

[Lorenzo Bianconi]

[-- Attachment #1: Type: text/plain, Size: 2494 bytes --]

> 
> On 17/08/2022 09.40, Lorenzo Bianconi wrote:
> > Report rx queue index in xdp_frame according to the xdp_buff xdp_rxq_info
> > pointer. xdp_frame queue_index is currently used in cpumap code to convert
> > the xdp_frame into a xdp_buff and allow the ebpf program attached to the
> > map entry to differentiate traffic according to the receiving hw queue.
> > xdp_frame size is not increased adding queue_index since an alignment
> > padding in the structure is used to insert queue_index field.
> > 
> > Signed-off-by: Lorenzo Bianconi <lorenzo@kernel.org>
> 
> (Sorry, I replied to v1 and not this v2.)
> 
> I'm still unsure about this change, because the XDP-hints will also
> contain the rx_queue number.  And placing it in XDP-hints automatically
> makes it avail for AF_XDP consumers.
> 
> I do think it is relevant for the BPF-prog to get access to the rx_queue
> index, because it can be used for scaling the workload.

ack, I agree. Then we can implement it with xdp hw-hints.

Regards,
Lorenzo

> 
> 
> > ---
> > Changes since v1:
> > - rebase on top of bpf-next
> > ---
> >   include/net/xdp.h   | 2 ++
> >   kernel/bpf/cpumap.c | 2 +-
> >   2 files changed, 3 insertions(+), 1 deletion(-)
> > 
> > diff --git a/include/net/xdp.h b/include/net/xdp.h
> > index 04c852c7a77f..3567866b0af5 100644
> > --- a/include/net/xdp.h
> > +++ b/include/net/xdp.h
> > @@ -172,6 +172,7 @@ struct xdp_frame {
> >   	struct xdp_mem_info mem;
> >   	struct net_device *dev_rx; /* used by cpumap */
> >   	u32 flags; /* supported values defined in xdp_buff_flags */
> > +	u32 queue_index;
> >   };
> >   static __always_inline bool xdp_frame_has_frags(struct xdp_frame *frame)
> > @@ -301,6 +302,7 @@ struct xdp_frame *xdp_convert_buff_to_frame(struct xdp_buff *xdp)
> >   	/* rxq only valid until napi_schedule ends, convert to xdp_mem_info */
> >   	xdp_frame->mem = xdp->rxq->mem;
> > +	xdp_frame->queue_index = xdp->rxq->queue_index;
> >   	return xdp_frame;
> >   }
> > diff --git a/kernel/bpf/cpumap.c b/kernel/bpf/cpumap.c
> > index b5ba34ddd4b6..48003450c98c 100644
> > --- a/kernel/bpf/cpumap.c
> > +++ b/kernel/bpf/cpumap.c
> > @@ -228,7 +228,7 @@ static int cpu_map_bpf_prog_run_xdp(struct bpf_cpu_map_entry *rcpu,
> >   		rxq.dev = xdpf->dev_rx;
> >   		rxq.mem = xdpf->mem;
> > -		/* TODO: report queue_index to xdp_rxq_info */
> > +		rxq.queue_index = xdpf->queue_index;
> >   		xdp_convert_frame_to_buff(xdpf, &xdp);
> 

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 228 bytes --]