Notice how this allow us get XDP statistic without affecting the XDP
performance, as tracepoint is no-longer activated on a per packet basis.
The xdp_monitor sample/tool is updated to use this new tracepoint.
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 +-
samples/bpf/xdp_monitor_kern.c | 39 +++++++++++++++++++++++++++++++++++
samples/bpf/xdp_monitor_user.c | 44 +++++++++++++++++++++++++++++++++++++++-
6 files changed, 146 insertions(+), 9 deletions(-)
diff --git a/include/linux/bpf.h b/include/linux/bpf.h
index 6457343f6dfa..066b5c67c71f 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);
@@ -574,7 +575,8 @@ static inline void __dev_map_flush(struct bpf_map *map)
struct xdp_buff;
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 258d243099e1..8b7924368dc1 100644
--- a/net/core/filter.c
+++ b/net/core/filter.c
@@ -3022,7 +3022,7 @@ 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);
diff --git a/samples/bpf/xdp_monitor_kern.c b/samples/bpf/xdp_monitor_kern.c
index 211db8ded0de..2854aa0665ea 100644
--- a/samples/bpf/xdp_monitor_kern.c
+++ b/samples/bpf/xdp_monitor_kern.c
@@ -208,3 +208,42 @@ int trace_xdp_cpumap_kthread(struct cpumap_kthread_ctx
*ctx)
return 0;
}
+
+struct bpf_map_def SEC("maps") devmap_xmit_cnt = {
+ .type = BPF_MAP_TYPE_PERCPU_ARRAY,
+ .key_size = sizeof(u32),
+ .value_size = sizeof(struct datarec),
+ .max_entries = 1,
+};
+
+/* Tracepoint: /sys/kernel/debug/tracing/events/xdp/xdp_devmap_xmit/format
+ * Code in: kernel/include/trace/events/xdp.h
+ */
+struct devmap_xmit_ctx {
+ u64 __pad; // First 8 bytes are not accessible by bpf code
+ int map_id; // offset:8; size:4; signed:1;
+ u32 act; // offset:12; size:4; signed:0;
+ u32 map_index; // offset:16; size:4; signed:0;
+ int drops; // offset:20; size:4; signed:1;
+ int sent; // offset:24; size:4; signed:1;
+ int from_ifindex; // offset:28; size:4; signed:1;
+ int to_ifindex; // offset:32; size:4; signed:1;
+};
+
+SEC("tracepoint/xdp/xdp_devmap_xmit")
+int trace_xdp_devmap_xmit(struct devmap_xmit_ctx *ctx)
+{
+ struct datarec *rec;
+ u32 key = 0;
+
+ rec = bpf_map_lookup_elem(&devmap_xmit_cnt, &key);
+ if (!rec)
+ return 0;
+ rec->processed += ctx->sent;
+ rec->dropped += ctx->drops;
+
+ /* Record bulk events, then userspace can calc average bulk size */
+ rec->info += 1;
+
+ return 1;
+}
diff --git a/samples/bpf/xdp_monitor_user.c b/samples/bpf/xdp_monitor_user.c
index 894bc64c2cac..4aaf1ab1927d 100644
--- a/samples/bpf/xdp_monitor_user.c
+++ b/samples/bpf/xdp_monitor_user.c
@@ -141,6 +141,7 @@ struct stats_record {
struct record_u64 xdp_exception[XDP_ACTION_MAX];
struct record xdp_cpumap_kthread;
struct record xdp_cpumap_enqueue[MAX_CPUS];
+ struct record xdp_devmap_xmit;
};
static bool map_collect_record(int fd, __u32 key, struct record *rec)
@@ -397,7 +398,7 @@ static void stats_print(struct stats_record *stats_rec,
info = calc_info(r, p, t);
if (info > 0)
i_str = "sched";
- if (pps > 0)
+ if (pps > 0 || drop > 0)
printf(fmt1, "cpumap-kthread",
i, pps, drop, info, i_str);
}
@@ -409,6 +410,42 @@ static void stats_print(struct stats_record *stats_rec,
printf(fmt2, "cpumap-kthread", "total", pps, drop, info, i_str);
}
+ /* devmap ndo_xdp_xmit stats */
+ {
+ char *fmt1 = "%-15s %-7d %'-12.0f %'-12.0f %'-10.2f %s\n";
+ char *fmt2 = "%-15s %-7s %'-12.0f %'-12.0f %'-10.2f %s\n";
+ struct record *rec, *prev;
+ double drop, info;
+ char *i_str = "";
+
+ rec = &stats_rec->xdp_devmap_xmit;
+ prev = &stats_prev->xdp_devmap_xmit;
+ t = calc_period(rec, prev);
+ for (i = 0; i < nr_cpus; i++) {
+ struct datarec *r = &rec->cpu[i];
+ struct datarec *p = &prev->cpu[i];
+
+ pps = calc_pps(r, p, t);
+ drop = calc_drop(r, p, t);
+ info = calc_info(r, p, t);
+ if (info > 0) {
+ i_str = "bulk-average";
+ info = (pps+drop) / info; /* calc avg bulk */
+ }
+ if (pps > 0 || drop > 0)
+ printf(fmt1, "devmap-xmit",
+ i, pps, drop, info, i_str);
+ }
+ pps = calc_pps(&rec->total, &prev->total, t);
+ drop = calc_drop(&rec->total, &prev->total, t);
+ info = calc_info(&rec->total, &prev->total, t);
+ if (info > 0) {
+ i_str = "bulk-average";
+ info = (pps+drop) / info; /* calc avg bulk */
+ }
+ printf(fmt2, "devmap-xmit", "total", pps, drop, info, i_str);
+ }
+
printf("\n");
}
@@ -437,6 +474,9 @@ static bool stats_collect(struct stats_record *rec)
fd = map_data[3].fd; /* map3: cpumap_kthread_cnt */
map_collect_record(fd, 0, &rec->xdp_cpumap_kthread);
+ fd = map_data[4].fd; /* map4: devmap_xmit_cnt */
+ map_collect_record(fd, 0, &rec->xdp_devmap_xmit);
+
return true;
}
@@ -480,6 +520,7 @@ static struct stats_record *alloc_stats_record(void)
rec_sz = sizeof(struct datarec);
rec->xdp_cpumap_kthread.cpu = alloc_rec_per_cpu(rec_sz);
+ rec->xdp_devmap_xmit.cpu = alloc_rec_per_cpu(rec_sz);
for (i = 0; i < MAX_CPUS; i++)
rec->xdp_cpumap_enqueue[i].cpu = alloc_rec_per_cpu(rec_sz);
@@ -498,6 +539,7 @@ static void free_stats_record(struct stats_record *r)
free(r->xdp_exception[i].cpu);
free(r->xdp_cpumap_kthread.cpu);
+ free(r->xdp_devmap_xmit.cpu);
for (i = 0; i < MAX_CPUS; i++)
free(r->xdp_cpumap_enqueue[i].cpu);
