Netdev List
 help / color / mirror / Atom feed
From: Jesper Dangaard Brouer <hawk@kernel.org>
To: netdev@vger.kernel.org, "Eric Dumazet" <eric.dumazet@gmail.com>,
	"David S. Miller" <davem@davemloft.net>,
	"Paolo Abeni" <pabeni@redhat.com>,
	"Toke Høiland-Jørgensen" <toke@toke.dk>
Cc: Jesper Dangaard Brouer <hawk@kernel.org>,
	bpf@vger.kernel.org, Jakub Kicinski <kuba@kernel.org>,
	horms@kernel.org, jiri@resnulli.us, edumazet@google.com,
	xiyou.wangcong@gmail.com, jhs@mojatatu.com, atenart@redhat.com,
	carges@cloudflare.com, kernel-team@cloudflare.com
Subject: [PATCH net-next v5 0/5] net: sched: refactor qdisc drop reasons into dedicated tracepoint
Date: Thu, 26 Feb 2026 14:44:06 +0100	[thread overview]
Message-ID: <177211325634.3011628.9343837509740374154.stgit@firesoul> (raw)

This series refactors qdisc drop reason handling by introducing a dedicated
enum qdisc_drop_reason and trace_qdisc_drop tracepoint, providing qdisc
layer drop diagnostics with direct qdisc context visibility.

Background:
-----------
Identifying which qdisc dropped a packet via skb_drop_reason is difficult.
Normally, the kfree_skb tracepoint caller "location" hints at the dropping
code, but qdisc drops happen at a central point (__dev_queue_xmit), making
this unusable. As a workaround, commits 5765c7f6e317 ("net_sched: sch_fq:
add three drop_reason") and a42d71e322a8 ("net_sched: sch_cake: Add drop
reasons") encoded qdisc names directly in the drop reason enums.

This series provides a cleaner solution by creating a dedicated qdisc
tracepoint that naturally includes qdisc context (handle, parent, kind).

Solution:
---------
Create a new tracepoint trace_qdisc_drop that builds on top of existing
trace_qdisc_enqueue infrastructure. It includes qdisc handle, parent,
qdisc kind (name), and device information directly.

The existing SKB_DROP_REASON_QDISC_DROP is retained for backwards
compatibility via kfree_skb_reason(). The qdisc-specific drop reasons
(QDISC_DROP_*) provide fine-grained detail via the new tracepoint.
The enum uses subsystem encoding (offset by SKB_DROP_REASON_SUBSYS_QDISC)
to catch type mismatches during debugging.

This implements the alternative approach described in:
https://lore.kernel.org/all/6be17a08-f8aa-4f91-9bd0-d9e1f0a92d90@kernel.org/

Changes since v4:
-----------------
v4: https://lore.kernel.org/all/177125575993.2500844.8718737675156179479.stgit@firesoul/
- Patch 1: Fix clang -Wimplicit-enum-enum-cast warning in sch_dualpi2.c (Jakub)
  - DualPI2 dequeue path: use SKB_DROP_REASON_QDISC_DROP (correct enum type)
    instead of QDISC_DROP_CONGESTED for drop_and_retry() which takes
    enum skb_drop_reason. Patch 5 properly converts this.

Changes since v3:
-----------------
v3: https://lore.kernel.org/all/177039500964.2258217.2989656069254156812.stgit@firesoul/
- Folded patch 3 (CAKE cobalt_should_drop) into patch 1 to avoid
  temporary enum type confusion between patches (Toke, Paolo)
- Split tcf_kfree_skb_list() into inline wrapper + __tcf_kfree_skb_list()
  for fastpath optimization when skb is NULL (Paolo)
- Series is now 5 patches instead of 6
- Added Reviewed-by from Toke Høiland-Jørgensen

Changes since v2:
-----------------
Link v2: https://lore.kernel.org/all/177032644012.1975497.16411100029657607833.stgit@firesoul/
- Patch 1: Fix tcf_kfree_skb_list() to handle TC classifier drop reasons
  - syzbot triggered WARN_ON_ONCE revealing TC classifier's SKB_DROP_REASON_TC_*
    reasons arrive via shared tc_skb_cb->drop_reason storage
  - Use SKB_DROP_REASON_SUBSYS_MASK to distinguish qdisc vs TC classifier reasons
  - Pass TC classifier reasons through to kfree_skb_reason() unchanged
  - Reported-by: syzbot+ci335f183b8446659a@syzkaller.appspotmail.com
- Patch 5 (was 6): Fix compile error - add missing QDISC_DROP_L4S_STEP_NON_ECN
  - Reported-by: kernel test robot <lkp@intel.com>
  - Closes: https://lore.kernel.org/oe-kbuild-all/202602061240.CRDtcOeX-lkp@intel.com/

Changes since RFC v1:
---------------------
Link RFC-v1: https://lore.kernel.org/all/177022452988.1827734.5121740962985640333.stgit@firesoul/
- Added SKB_DROP_REASON_SUBSYS_QDISC subsystem encoding for enum debugging
- Added WARN_ON_ONCE to catch wrong enum type (skb vs qdisc drop reason)
- Renamed QDISC_DROP_CAKE_FLOOD to QDISC_DROP_FLOOD_PROTECTION (Toke)
- Renamed QDISC_DROP_FQ_BAND_LIMIT to QDISC_DROP_BAND_LIMIT
- Renamed QDISC_DROP_FQ_HORIZON_LIMIT to QDISC_DROP_HORIZON_LIMIT
- Updated SKB_DROP_REASON_QDISC_DROP doc to reference qdisc:qdisc_drop
- Split rename changes into separate patches for clarity

Series overview:
----------------
Patch 1: Introduce the qdisc drop reason infrastructure
  - New enum qdisc_drop_reason with subsystem encoding
  - New trace_qdisc_drop tracepoint with qdisc context
  - tcf_kfree_skb_list() handles both qdisc and TC classifier drop reasons
  - Split into inline wrapper + __tcf_kfree_skb_list() for fastpath
  - Convert qdiscs with drop reasons to use the new infrastructure
  - Fix CAKE cobalt_should_drop() return type for type correctness

Patch 2: Convert SFQ as example for reviewers
  - Demonstrates how to convert a flow-based qdisc
  - Adds QDISC_DROP_MAXFLOWS for flow table exhaustion
  - Rename FQ_FLOW_LIMIT to generic FLOW_LIMIT, now shared by FQ and SFQ

Patch 3: Rename QDISC_DROP_FQ_* to generic names
  - QDISC_DROP_FQ_BAND_LIMIT -> QDISC_DROP_BAND_LIMIT
  - QDISC_DROP_FQ_HORIZON_LIMIT -> QDISC_DROP_HORIZON_LIMIT
  - Remove local FQDR() macro, use full QDISC_DROP_* names

Patch 4: Rename QDISC_DROP_CAKE_FLOOD to QDISC_DROP_FLOOD_PROTECTION
  - Generic name without embedding qdisc name

Patch 5: DualPI2 use qdisc_dequeue_drop() for dequeue drops
  - Demonstrates dequeue-time drop handling with trace_qdisc_drop
  - Adds QDISC_DROP_L4S_STEP_NON_ECN for L4S step AQM drops

---

Jesper Dangaard Brouer (5):
      net: sched: introduce qdisc-specific drop reason tracing
      net: sched: sfq: convert to qdisc drop reasons
      net: sched: rename QDISC_DROP_FQ_* to generic names
      net: sched: rename QDISC_DROP_CAKE_FLOOD to QDISC_DROP_FLOOD_PROTECTION
      net: sched: sch_dualpi2: use qdisc_dequeue_drop() for dequeue drops


 include/net/dropreason-core.h  |  48 ++------------
 include/net/dropreason-qdisc.h | 114 +++++++++++++++++++++++++++++++++
 include/net/dropreason.h       |   6 ++
 include/net/sch_generic.h      |  43 +++++++++----
 include/trace/events/qdisc.h   |  51 +++++++++++++++
 net/core/dev.c                 |   8 +--
 net/sched/sch_cake.c           |  26 ++++----
 net/sched/sch_codel.c          |   5 +-
 net/sched/sch_dualpi2.c        |  18 +++---
 net/sched/sch_fq.c             |  10 +--
 net/sched/sch_fq_codel.c       |   4 +-
 net/sched/sch_fq_pie.c         |   4 +-
 net/sched/sch_generic.c        |  29 ++++++++-
 net/sched/sch_gred.c           |   4 +-
 net/sched/sch_pie.c            |   4 +-
 net/sched/sch_red.c            |   4 +-
 net/sched/sch_sfb.c            |   4 +-
 net/sched/sch_sfq.c            |   8 +--
 18 files changed, 278 insertions(+), 112 deletions(-)
 create mode 100644 include/net/dropreason-qdisc.h

--
Ideology: "If you cannot measure it, you cannot improve it"
--Jesper



             reply	other threads:[~2026-02-26 13:44 UTC|newest]

Thread overview: 7+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2026-02-26 13:44 Jesper Dangaard Brouer [this message]
2026-02-26 13:44 ` [PATCH net-next v5 1/5] net: sched: introduce qdisc-specific drop reason tracing Jesper Dangaard Brouer
2026-02-26 13:44 ` [PATCH net-next v5 2/5] net: sched: sfq: convert to qdisc drop reasons Jesper Dangaard Brouer
2026-02-26 13:44 ` [PATCH net-next v5 3/5] net: sched: rename QDISC_DROP_FQ_* to generic names Jesper Dangaard Brouer
2026-02-26 13:44 ` [PATCH net-next v5 4/5] net: sched: rename QDISC_DROP_CAKE_FLOOD to QDISC_DROP_FLOOD_PROTECTION Jesper Dangaard Brouer
2026-02-26 13:45 ` [PATCH net-next v5 5/5] net: sched: sch_dualpi2: use qdisc_dequeue_drop() for dequeue drops Jesper Dangaard Brouer
2026-03-01  0:00 ` [PATCH net-next v5 0/5] net: sched: refactor qdisc drop reasons into dedicated tracepoint patchwork-bot+netdevbpf

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=177211325634.3011628.9343837509740374154.stgit@firesoul \
    --to=hawk@kernel.org \
    --cc=atenart@redhat.com \
    --cc=bpf@vger.kernel.org \
    --cc=carges@cloudflare.com \
    --cc=davem@davemloft.net \
    --cc=edumazet@google.com \
    --cc=eric.dumazet@gmail.com \
    --cc=horms@kernel.org \
    --cc=jhs@mojatatu.com \
    --cc=jiri@resnulli.us \
    --cc=kernel-team@cloudflare.com \
    --cc=kuba@kernel.org \
    --cc=netdev@vger.kernel.org \
    --cc=pabeni@redhat.com \
    --cc=toke@toke.dk \
    --cc=xiyou.wangcong@gmail.com \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox