net_sched: sch_fq: add three drop_reason

Add three new drop_reason, more precise than generic QDISC_DROP:

"tc -s qd" show aggregate counters, it might be more useful
to use drop_reason infrastructure for bug hunting.

1) SKB_DROP_REASON_FQ_BAND_LIMIT
   Whenever a packet is added while its band limit is hit.
   Corresponding value in "tc -s qd" is bandX_drops XXXX

2) SKB_DROP_REASON_FQ_HORIZON_LIMIT
   Whenever a packet has a timestamp too far in the future.
   Corresponding value in "tc -s qd" is horizon_drops XXXX

3) SKB_DROP_REASON_FQ_FLOW_LIMIT
   Whenever a flow has reached its limit.
   Corresponding value in "tc -s qd" is flows_plimit XXXX

Tested:
tc qd replace dev eth1 root fq flow_limit 10 limit 100000
perf record -a -e skb:kfree_skb sleep 1; perf script

      udp_stream   12329 [004]   216.929492: skb:kfree_skb: skbaddr=0xffff888eabe17e00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_FLOW_LIMIT
      udp_stream   12385 [006]   216.929593: skb:kfree_skb: skbaddr=0xffff888ef8827f00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_FLOW_LIMIT
      udp_stream   12389 [005]   216.929871: skb:kfree_skb: skbaddr=0xffff888ecb9ba500 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_FLOW_LIMIT
      udp_stream   12316 [009]   216.930398: skb:kfree_skb: skbaddr=0xffff888eca286b00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_FLOW_LIMIT
      udp_stream   12400 [008]   216.930490: skb:kfree_skb: skbaddr=0xffff888eabf93d00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_FLOW_LIMIT

tc qd replace dev eth1 root fq flow_limit 100 limit 10000
perf record -a -e skb:kfree_skb sleep 1; perf script

      udp_stream   18074 [001]  1058.318040: skb:kfree_skb: skbaddr=0xffffa23c881fc000 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_BAND_LIMIT
      udp_stream   18126 [005]  1058.320651: skb:kfree_skb: skbaddr=0xffffa23c6aad4000 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_BAND_LIMIT
      udp_stream   18118 [006]  1058.321065: skb:kfree_skb: skbaddr=0xffffa23df0d48a00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_BAND_LIMIT
      udp_stream   18074 [001]  1058.321126: skb:kfree_skb: skbaddr=0xffffa23c881ffa00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_BAND_LIMIT
      udp_stream   15815 [003]  1058.321224: skb:kfree_skb: skbaddr=0xffffa23c9835db00 rx_sk=(nil) protocol=34525 location=__dev_queue_xmit+0x9d9 reason: FQ_BAND_LIMIT

tc -s -d qd sh dev eth1
qdisc fq 8023: root refcnt 257 limit 10000p flow_limit 100p buckets 1024 orphan_mask 1023
 bands 3 priomap 1 2 2 2 1 2 0 0 1 1 1 1 1 1 1 1 weights 589824 196608 65536 quantum 18Kb
 initial_quantum 92120b low_rate_threshold 550Kbit refill_delay 40ms
 timer_slack 10us horizon 10s horizon_drop
 Sent 492439603330 bytes 336953991 pkt (dropped 61724094, overlimits 0 requeues 4463)
 backlog 14611228b 9995p requeues 4463
  flows 2965 (inactive 1151 throttled 0) band0_pkts 0 band1_pkts 9993 band2_pkts 0
  gc 6347 highprio 0 fastpath 30 throttled 5 latency 2.32us flows_plimit 7403693
 band1_drops 54320401

Signed-off-by: Eric Dumazet <edumazet@google.com>
Reviewed-by: Victor Nogueira <victor@mojatatu.com>
Reviewed-by: Toke Høiland-Jørgensen <toke@redhat.com>
Acked-by: Jamal Hadi Salim <jhs@mojatatu.com>
Link: https://patch.msgid.link/20241204171950.89829-1-edumazet@google.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

diff --git a/include/net/dropreason-core.h b/include/net/dropreason-core.h
index 6c5a1ea209a22d8702f8c982762ca5f69791b8eb..c29282fabae6cdf9dd79f698b92b4b8f57156b1e 100644 (file)
--- a/include/net/dropreason-core.h
+++ b/include/net/dropreason-core.h
@@ -58,6 +58,9 @@
        FN(TC_EGRESS)                   \
        FN(SECURITY_HOOK)               \
        FN(QDISC_DROP)                  \
+       FN(FQ_BAND_LIMIT)               \
+       FN(FQ_HORIZON_LIMIT)            \
+       FN(FQ_FLOW_LIMIT)               \
        FN(CPU_BACKLOG)                 \
        FN(XDP)                         \
        FN(TC_INGRESS)                  \
@@ -311,6 +314,21 @@ enum skb_drop_reason {
         * failed to enqueue to current qdisc)
         */
        SKB_DROP_REASON_QDISC_DROP,
+       /**
+        * @SKB_DROP_REASON_FQ_BAND_LIMIT: dropped by fq qdisc when per band
+        * limit is reached.
+        */
+       SKB_DROP_REASON_FQ_BAND_LIMIT,
+       /**
+        * @SKB_DROP_REASON_FQ_HORIZON_LIMIT: dropped by fq qdisc when packet
+        * timestamp is too far in the future.
+        */
+       SKB_DROP_REASON_FQ_HORIZON_LIMIT,
+       /**
+        * @SKB_DROP_REASON_FQ_FLOW_LIMIT: dropped by fq qdisc when a flow
+        * exceeds its limits.
+        */
+       SKB_DROP_REASON_FQ_FLOW_LIMIT,
        /**
         * @SKB_DROP_REASON_CPU_BACKLOG: failed to enqueue the skb to the per CPU
         * backlog queue. This can be caused by backlog queue full (see

diff --git a/include/net/sch_generic.h b/include/net/sch_generic.h
index 5d74fa7e694cc85be91dbf01f0876b9feaa29115..8074322dd636a9706923dcef469cb6473c68d16b 100644 (file)
--- a/include/net/sch_generic.h
+++ b/include/net/sch_generic.h
@@ -1245,6 +1245,14 @@ static inline int qdisc_drop(struct sk_buff *skb, struct Qdisc *sch,
        return NET_XMIT_DROP;
 }
 
+static inline int qdisc_drop_reason(struct sk_buff *skb, struct Qdisc *sch,
+                                   struct sk_buff **to_free,
+                                   enum skb_drop_reason reason)
+{
+       tcf_set_drop_reason(skb, reason);
+       return qdisc_drop(skb, sch, to_free);
+}
+
 static inline int qdisc_drop_all(struct sk_buff *skb, struct Qdisc *sch,
                                 struct sk_buff **to_free)
 {

diff --git a/net/sched/sch_fq.c b/net/sched/sch_fq.c
index a5e87f9ea9861cbedb7ce858fbbcabb5d67cf821..2ca5332cfcc5c52bf30e6f8337814a656b919b17 100644 (file)
--- a/net/sched/sch_fq.c
+++ b/net/sched/sch_fq.c
@@ -537,6 +537,8 @@ static bool fq_packet_beyond_horizon(const struct sk_buff *skb,
        return unlikely((s64)skb->tstamp > (s64)(now + q->horizon));
 }
 
+#define FQDR(reason) SKB_DROP_REASON_FQ_##reason
+
 static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
                      struct sk_buff **to_free)
 {
@@ -548,7 +550,8 @@ static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
        band = fq_prio2band(q->prio2band, skb->priority & TC_PRIO_MAX);
        if (unlikely(q->band_pkt_count[band] >= sch->limit)) {
                q->stat_band_drops[band]++;
-               return qdisc_drop(skb, sch, to_free);
+               return qdisc_drop_reason(skb, sch, to_free,
+                                        FQDR(BAND_LIMIT));
        }
 
        now = ktime_get_ns();
@@ -558,8 +561,9 @@ static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
                /* Check if packet timestamp is too far in the future. */
                if (fq_packet_beyond_horizon(skb, q, now)) {
                        if (q->horizon_drop) {
-                                       q->stat_horizon_drops++;
-                                       return qdisc_drop(skb, sch, to_free);
+                               q->stat_horizon_drops++;
+                               return qdisc_drop_reason(skb, sch, to_free,
+                                                        FQDR(HORIZON_LIMIT));
                        }
                        q->stat_horizon_caps++;
                        skb->tstamp = now + q->horizon;
@@ -572,7 +576,8 @@ static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
        if (f != &q->internal) {
                if (unlikely(f->qlen >= q->flow_plimit)) {
                        q->stat_flows_plimit++;
-                       return qdisc_drop(skb, sch, to_free);
+                       return qdisc_drop_reason(skb, sch, to_free,
+                                                FQDR(FLOW_LIMIT));
                }
 
                if (fq_flow_is_detached(f)) {
@@ -597,6 +602,7 @@ static int fq_enqueue(struct sk_buff *skb, struct Qdisc *sch,
 
        return NET_XMIT_SUCCESS;
 }
+#undef FQDR
 
 static void fq_check_throttled(struct fq_sched_data *q, u64 now)
 {