Notice how this allow us get XDP statistic without affecting the XDP performance, as tracepoint is no-longer activated on a per packet basis.
Signed-off-by: Jesper Dangaard Brouer <bro...@redhat.com> --- include/linux/bpf.h | 6 ++++-- include/trace/events/xdp.h | 39 +++++++++++++++++++++++++++++++++++++++ kernel/bpf/devmap.c | 25 ++++++++++++++++++++----- net/core/filter.c | 2 +- 4 files changed, 64 insertions(+), 8 deletions(-) diff --git a/include/linux/bpf.h b/include/linux/bpf.h index fc1459bdcafc..ca7110b81793 100644 --- a/include/linux/bpf.h +++ b/include/linux/bpf.h @@ -489,7 +489,8 @@ struct xdp_buff; struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key); void __dev_map_insert_ctx(struct bpf_map *map, u32 index); void __dev_map_flush(struct bpf_map *map); -int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp); +int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, + struct net_device *dev_rx); struct bpf_cpu_map_entry *__cpu_map_lookup_elem(struct bpf_map *map, u32 key); void __cpu_map_insert_ctx(struct bpf_map *map, u32 index); @@ -575,7 +576,8 @@ static inline void __dev_map_flush(struct bpf_map *map) struct xdp_buff; struct bpf_dtab_netdev; static inline -int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp) +int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, + struct net_device *dev_rx) { return 0; } diff --git a/include/trace/events/xdp.h b/include/trace/events/xdp.h index 96104610d40e..2e9ef0650144 100644 --- a/include/trace/events/xdp.h +++ b/include/trace/events/xdp.h @@ -229,6 +229,45 @@ TRACE_EVENT(xdp_cpumap_enqueue, __entry->to_cpu) ); +TRACE_EVENT(xdp_devmap_xmit, + + TP_PROTO(const struct bpf_map *map, u32 map_index, + int sent, int drops, + const struct net_device *from_dev, + const struct net_device *to_dev), + + TP_ARGS(map, map_index, sent, drops, from_dev, to_dev), + + TP_STRUCT__entry( + __field(int, map_id) + __field(u32, act) + __field(u32, map_index) + __field(int, drops) + __field(int, sent) + __field(int, from_ifindex) + __field(int, to_ifindex) + ), + + TP_fast_assign( + __entry->map_id = map->id; + __entry->act = XDP_REDIRECT; + __entry->map_index = map_index; + __entry->drops = drops; + __entry->sent = sent; + __entry->from_ifindex = from_dev->ifindex; + __entry->to_ifindex = to_dev->ifindex; + ), + + TP_printk("ndo_xdp_xmit" + " map_id=%d map_index=%d action=%s" + " sent=%d drops=%d" + " from_ifindex=%d to_ifindex=%d", + __entry->map_id, __entry->map_index, + __print_symbolic(__entry->act, __XDP_ACT_SYM_TAB), + __entry->sent, __entry->drops, + __entry->from_ifindex, __entry->to_ifindex) +); + #endif /* _TRACE_XDP_H */ #include <trace/define_trace.h> diff --git a/kernel/bpf/devmap.c b/kernel/bpf/devmap.c index cab72c100bb5..6f84100723b0 100644 --- a/kernel/bpf/devmap.c +++ b/kernel/bpf/devmap.c @@ -50,6 +50,7 @@ #include <linux/bpf.h> #include <net/xdp.h> #include <linux/filter.h> +#include <trace/events/xdp.h> #define DEV_CREATE_FLAG_MASK \ (BPF_F_NUMA_NODE | BPF_F_RDONLY | BPF_F_WRONLY) @@ -57,6 +58,7 @@ #define DEV_MAP_BULK_SIZE 16 struct xdp_bulk_queue { struct xdp_frame *q[DEV_MAP_BULK_SIZE]; + struct net_device *dev_rx; unsigned int count; }; @@ -219,8 +221,8 @@ void __dev_map_insert_ctx(struct bpf_map *map, u32 bit) static int bq_xmit_all(struct bpf_dtab_netdev *obj, struct xdp_bulk_queue *bq) { - unsigned int processed = 0, drops = 0; struct net_device *dev = obj->dev; + int sent = 0, drops = 0; int i; if (unlikely(!bq->count)) @@ -241,10 +243,13 @@ static int bq_xmit_all(struct bpf_dtab_netdev *obj, drops++; xdp_return_frame(xdpf); } - processed++; + sent++; } bq->count = 0; + trace_xdp_devmap_xmit(&obj->dtab->map, obj->bit, + sent, drops, bq->dev_rx, dev); + bq->dev_rx = NULL; return 0; } @@ -301,18 +306,28 @@ struct bpf_dtab_netdev *__dev_map_lookup_elem(struct bpf_map *map, u32 key) /* Runs under RCU-read-side, plus in softirq under NAPI protection. * Thus, safe percpu variable access. */ -static int bq_enqueue(struct bpf_dtab_netdev *obj, struct xdp_frame *xdpf) +static int bq_enqueue(struct bpf_dtab_netdev *obj, struct xdp_frame *xdpf, + struct net_device *dev_rx) + { struct xdp_bulk_queue *bq = this_cpu_ptr(obj->bulkq); if (unlikely(bq->count == DEV_MAP_BULK_SIZE)) bq_xmit_all(obj, bq); + /* Ingress dev_rx will be the same for all xdp_frame's in + * bulk_queue, because bq stored per-CPU and must be flushed + * from net_device drivers NAPI func end. + */ + if (!bq->dev_rx) + bq->dev_rx = dev_rx; + bq->q[bq->count++] = xdpf; return 0; } -int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp) +int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp, + struct net_device *dev_rx) { struct net_device *dev = dst->dev; struct xdp_frame *xdpf; @@ -325,7 +340,7 @@ int dev_map_enqueue(struct bpf_dtab_netdev *dst, struct xdp_buff *xdp) if (unlikely(!xdpf)) return -EOVERFLOW; - err = bq_enqueue(dst, xdpf); + err = bq_enqueue(dst, xdpf, dev_rx); if (err) return err; diff --git a/net/core/filter.c b/net/core/filter.c index 1447ec94ef74..4a93423cc5ea 100644 --- a/net/core/filter.c +++ b/net/core/filter.c @@ -3063,7 +3063,7 @@ static int __bpf_tx_xdp_map(struct net_device *dev_rx, void *fwd, case BPF_MAP_TYPE_DEVMAP: { struct bpf_dtab_netdev *dst = fwd; - err = dev_map_enqueue(dst, xdp); + err = dev_map_enqueue(dst, xdp, dev_rx); if (err) return err; __dev_map_insert_ctx(map, index);