[Deepin-Kernel-SIG] [linux 6.6-y] [Upstream] Sync tcp patches from upstream

[opsiff]

Merge branch 'tcp-refactor-bhash2'

Kuniyuki Iwashima says:

====================

tcp: Refactor bhash2 and remove sk_bind2_node.

This series refactors code around bhash2 and remove some bhash2-specific
fields; sock.sk_bind2_node, and inet_timewait_sock.tw_bind2_node.

patch 1 : optimise bind() for non-wildcard v4-mapped-v6 address
patch 2 - 4 : optimise bind() conflict tests
patch 5 - 12 : Link bhash2 to bhash and unlink sk from bhash2 to
remove sk_bind2_node

The patch 8 will trigger a false-positive error by checkpatch.

v2: resend of https://lore.kernel.org/netdev/[email protected]/

• Rebase on latest net-next
• Patch 11
  • Add change in inet_diag_dump_icsk() for recent bhash dump patch

v1: https://lore.kernel.org/netdev/[email protected]/

Merge branch 'tcp-scale-connect-under-pressure'

Eric Dumazet says:

====================
tcp: scale connect() under pressure

Adoption of bhash2 in linux-6.1 made some operations almost twice
more expensive, because of additional locks.

This series adds RCU in __inet_hash_connect() to help the
case where many attempts need to be made before finding
an available 4-tuple.

This brings a ~200 % improvement in this experiment:

Server:
ulimit -n 40000; neper/tcp_crr -T 200 -F 30000 -6 --nolog

Client:
ulimit -n 40000; neper/tcp_crr -T 200 -F 30000 -6 --nolog -c -H server

Before series:

utime_start=0.288582
utime_end=1.548707
stime_start=20.637138
stime_end=2002.489845
num_transactions=484453
latency_min=0.156279245
latency_max=20.922042756
latency_mean=1.546521274
latency_stddev=3.936005194
num_samples=312537
throughput=47426.00

perf top on the client:

49.54% [kernel] [k] _raw_spin_lock
25.87% [kernel] [k] _raw_spin_lock_bh
5.97% [kernel] [k] queued_spin_lock_slowpath
5.67% [kernel] [k] __inet_hash_connect
3.53% [kernel] [k] __inet6_check_established
3.48% [kernel] [k] inet6_ehashfn
0.64% [kernel] [k] rcu_all_qs

After this series:

utime_start=0.271607
utime_end=3.847111
stime_start=18.407684
stime_end=1997.485557
num_transactions=1350742
latency_min=0.014131929
latency_max=17.895073144
latency_mean=0.505675853 # Nice reduction of latency metrics
latency_stddev=2.125164772
num_samples=307884
throughput=139866.80 # 194 % increase

perf top on client:

56.86% [kernel] [k] __inet6_check_established
17.96% [kernel] [k] __inet_hash_connect
13.88% [kernel] [k] inet6_ehashfn
2.52% [kernel] [k] rcu_all_qs
2.01% [kernel] [k] __cond_resched
0.41% [kernel] [k] _raw_spin_lock

Link: https://patch.msgid.link/[email protected]
Signed-off-by: Jakub Kicinski [email protected]

Summary by Sourcery

Sync upstream TCP refactor and performance optimizations: restructure bhash2 binding, remove obsolete sk_bind2_node, add RCU fast-path for connect() under pressure, and introduce TCP_BOUND_INACTIVE support for diagnostics

New Features:

• Scale TCP connect path under high load using an RCU-based fast path in __inet_hash_connect
• Introduce a TCP_BOUND_INACTIVE pseudo-state and corresponding BPF_TCP_BOUND_INACTIVE support

Enhancements:

• Refactor bind buckets (bhash2) by removing sk_bind2_node, unifying inet_bind_bucket and inet_bind2_bucket structures, and switching to RCU-safe list operations
• Optimize bind() conflict checks and IPv4-mapped IPv6 address handling
• Simplify bind bucket creation and destruction interfaces by eliminating redundant cache parameters
• Update inet_diag to batch-process and report bound-inactive sockets

[sourcery-ai]

Reviewer's Guide

This PR refactors the secondary bind-hash (bhash2) into a unified, RCU-safe bind bucket model (removing sk_bind2_node and deathrow lists), introduces an RCU-based fast path in __inet_hash_connect to drastically improve connect() scalability under load, consolidates bind conflict checks, adds a cached quick-ACK metric to speed up ACK path decisions, and extends the inet_diag interface with a new TCP_BOUND_INACTIVE pseudo-state.

Sequence diagram for RCU-based fast path in __inet_hash_connect

sequenceDiagram
    participant App as Application
    participant Kernel as Kernel TCP Stack
    participant RCU as RCU Read Lock
    participant Hash as Bind Hash Bucket

    App->>Kernel: connect() syscall
    Kernel->>RCU: rcu_read_lock()
    Kernel->>Hash: hlist_for_each_entry_rcu (bind bucket lookup)
    alt Fast path: no conflict
        Kernel->>RCU: rcu_read_unlock()
        Kernel->>App: Success (no lock contention)
    else Conflict or not found
        Kernel->>RCU: rcu_read_unlock()
        Kernel->>Hash: spin_lock_bh (slow path)
        Kernel->>App: Continue with legacy path
    end

Loading

Class diagram for refactored bind bucket structures

classDiagram
    class inet_bind_bucket {
        +unsigned short port
        +unsigned short fastreuse
        +unsigned short fastreuseport
        +struct hlist_node node
        +struct hlist_head bhash2
        +struct rcu_head rcu
    }
    class inet_bind2_bucket {
        +struct net *ib_net
        +int l3mdev
        +unsigned short port
        +unsigned short addr_type
        +struct in6_addr v6_rcv_saddr
        +struct hlist_node node
        +struct hlist_node bhash_node
        +struct hlist_head owners
    }
    inet_bind_bucket "1" o-- "*" inet_bind2_bucket : bhash2

Loading

Class diagram for sock and inet_timewait_sock after sk_bind2_node removal

classDiagram
    class sock {
        ...
        -struct hlist_node sk_bind2_node
        ...
    }
    class inet_timewait_sock {
        ...
        -struct hlist_node tw_bind2_node
        ...
    }

Loading

Class diagram for inet_connection_sock quick-ACK caching

classDiagram
    class inet_connection_sock_ack_block {
        +unsigned int quick
        +unsigned int pingpong
        +unsigned int dst_quick_ack
    }
    class inet_connection_sock {
        +inet_connection_sock_ack_block icsk_ack
        ...
    }

Loading

Class diagram for new TCP state and flags

classDiagram
    class tcp_states_h {
        +enum TCP_BOUND_INACTIVE
        +enum TCPF_BOUND_INACTIVE
    }
    class bpf_h {
        +enum BPF_TCP_BOUND_INACTIVE
    }

Loading

Class diagram for inet_diag TCP_BOUND_INACTIVE support

classDiagram
    class inet_diag {
        +void inet_diag_dump_icsk(...)
        +TCP_BOUND_INACTIVE pseudo-state
    }

Loading

File-Level Changes

Change	Details	Files
Refactor bhash2 and remove sk_bind2_node with RCU support	Replace separate owners/deathrow lists with a single bhash2 list in bind buckets Use hlist_add_head_rcu/hlist_del_rcu and kfree_rcu for bind bucket lifecycle Update inet_bind2_bucket_init/create/destroy to accept the parent bind bucket pointer and eliminate redundant fields Remove sk_bind2_node, sk_add_bind2_node and __sk_del_bind2_node; use unified sk_add_bind_node Adjust timewait socket hashing to use the unified bind bucket list	net/ipv4/inet_hashtables.c include/net/inet_hashtables.h include/net/sock.h net/ipv4/inet_timewait_sock.c include/net/inet_timewait_sock.h net/ipv4/inet_connection_sock.c
Add RCU-based fast path in __inet_hash_connect	Extend check_established and __inet6_check_established signatures to take an rcu_lookup flag Introduce rcu_read_lock/unlock and hlist_for_each_entry_rcu in __inet_hash_connect to probe bind buckets without locks Branch off into an RCU path to skip busy buckets and avoid spin-lock contention Update header prototypes to reflect the new argument	net/ipv4/inet_hashtables.c net/ipv6/inet6_hashtables.c include/net/inet_hashtables.h
Consolidate bind conflict logic	Define sk_for_each_bound_bhash to iterate sockets across unified buckets Simplify inet_csk_bind_conflict to a single code path using the new macro Remove the dual-path conditional on inet_use_bhash2_on_bind Drop the obsolete sk_for_each_bound_bhash2 macro and related fields	net/ipv4/inet_connection_sock.c include/net/sock.h
Cache and apply dst_metric(RTAX_QUICKACK) for quick ACK decisions	Add dst_quick_ack bit field to inet_connection_sock Initialize dst_quick_ack in sk_setup_caps() based on dst_metric Update tcp_enter_quickack_mode() to consult the cached field	net/core/sock.c include/net/inet_connection_sock.h net/ipv4/tcp_input.c
Extend inet_diag and add TCP_BOUND_INACTIVE state	Introduce a new TCP_BOUND_INACTIVE pseudo-state and matching TCPF flag Implement batched SKARR_SZ walking of bound-inactive buckets in inet_diag_dump_icsk Add BUILD_BUG_ON check for the new bound_inactive state in tcp.c Update BPF and tcp_states headers to expose TCP_BOUND_INACTIVE	net/ipv4/inet_diag.c include/net/tcp_states.h include/uapi/linux/bpf.h net/ipv4/tcp.c

---

Tips and commands

Interacting with Sourcery

• Trigger a new review: Comment @sourcery-ai review on the pull request.
• Continue discussions: Reply directly to Sourcery's review comments.
• Generate a GitHub issue from a review comment: Ask Sourcery to create an
  issue from a review comment by replying to it. You can also reply to a
  review comment with @sourcery-ai issue to create an issue from it.
• Generate a pull request title: Write @sourcery-ai anywhere in the pull
  request title to generate a title at any time. You can also comment
  @sourcery-ai title on the pull request to (re-)generate the title at any time.
• Generate a pull request summary: Write @sourcery-ai summary anywhere in
  the pull request body to generate a PR summary at any time exactly where you
  want it. You can also comment @sourcery-ai summary on the pull request to
  (re-)generate the summary at any time.
• Generate reviewer's guide: Comment @sourcery-ai guide on the pull
  request to (re-)generate the reviewer's guide at any time.
• Resolve all Sourcery comments: Comment @sourcery-ai resolve on the
  pull request to resolve all Sourcery comments. Useful if you've already
  addressed all the comments and don't want to see them anymore.
• Dismiss all Sourcery reviews: Comment @sourcery-ai dismiss on the pull
  request to dismiss all existing Sourcery reviews. Especially useful if you
  want to start fresh with a new review - don't forget to comment
  @sourcery-ai review to trigger a new review!

Customizing Your Experience

Access your dashboard to:

• Enable or disable review features such as the Sourcery-generated pull request
  summary, the reviewer's guide, and others.
• Change the review language.
• Add, remove or edit custom review instructions.
• Adjust other review settings.

Getting Help

• Contact our support team for questions or feedback.
• Visit our documentation for detailed guides and information.
• Keep in touch with the Sourcery team by following us on X/Twitter, LinkedIn or GitHub.

[sourcery-ai]

Hey @opsiff - I've reviewed your changes - here's some feedback:

• Extract the duplicated RCU‐based port scanning logic in __inet_hash_connect and __inet6_check_established into a common helper to reduce code duplication and ease future maintenance.
• The resume/pagination state machine in inet_diag_dump_icsk is quite intricate; consider refactoring it into a dedicated iterator or helper to improve readability and avoid off‐by‐one errors.
• After switching inet_bind_bucket_destroy to kfree_rcu(tb, rcu), please audit all bind/unbind paths to ensure there are no use‐after‐free races under concurrent bind() or connect() workloads.

Prompt for AI Agents

Please address the comments from this code review:
## Overall Comments
- Extract the duplicated RCU‐based port scanning logic in __inet_hash_connect and __inet6_check_established into a common helper to reduce code duplication and ease future maintenance.
- The resume/pagination state machine in inet_diag_dump_icsk is quite intricate; consider refactoring it into a dedicated iterator or helper to improve readability and avoid off‐by‐one errors.
- After switching inet_bind_bucket_destroy to kfree_rcu(tb, rcu), please audit all bind/unbind paths to ensure there are no use‐after‐free races under concurrent bind() or connect() workloads.

---

Sourcery is free for open source - if you like our reviews please consider sharing them ✨

• X
• Mastodon
• LinkedIn
• Facebook

_{Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.}

[Avenger-285714]

/approve

[deepin-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: Avenger-285714

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• deepin/OWNERS [Avenger-285714]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[Copilot]

Pull Request Overview

This PR syncs upstream TCP refactor patches and adds performance optimizations, including RCU fast paths and a new pseudo-state for diagnostics.

• Refactor and unify bind bucket handling by removing sk_bind2_node and switching to RCU-safe lists.
• Introduce an RCU-based fast-path in __inet_hash_connect to scale connect() under load.
• Add TCP_BOUND_INACTIVE pseudo-state and batch-processing support in inet_diag_dump_icsk.

Reviewed Changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 1 comment.

Show a summary per file

File	Description
net/ipv6/inet6_hashtables.c	Added rcu_lookup parameter and RCU lookup path in established check.
net/ipv4/tcp_input.c	Replaced on-the-fly DST lookup with cached dst_quick_ack.
net/ipv4/tcp.c	Added BUILD_BUG_ON for new TCP_BOUND_INACTIVE state.
net/ipv4/inet_timewait_sock.c	Updated TW socket unhash to use __sk_del_bind_node and RCU destroy.
net/ipv4/inet_hashtables.c	Unified bind bucket API, removed legacy owners list, added RCU.
net/ipv4/inet_diag.c	Implemented TCP_BOUND_INACTIVE dump with batching (SKARR_SZ).
net/ipv4/inet_connection_sock.c	Revised bhash2 conflict logic and added sk_for_each_bound_bhash macro.
net/core/sock.c	Cached dst_quick_ack in inet_connection_sock.
include/uapi/linux/bpf.h	Defined BPF_TCP_BOUND_INACTIVE constant.
include/net/tcp_states.h	Defined TCP_BOUND_INACTIVE state and flag.
include/net/sock.h	Removed obsolete sk_bind2_node field and macros.
include/net/ipv6.h	Removed deprecated ipv6_addr_v4mapped_any.
include/net/inet_timewait_sock.h	Dropped tw_bind2_node and related macro.
include/net/inet_hashtables.h	Updated bind bucket structs and function prototypes.
include/net/inet_connection_sock.h	Added dst_quick_ack bit to inet_connection_sock.

Comments suppressed due to low confidence (1)

net/ipv4/inet_diag.c:1083

• Introduced a new diagnostic code path for bound-inactive sockets; please add unit or integration tests that exercise TCP_BOUND_INACTIVE reporting to ensure the batching logic and state filtering behave correctly.

#define SKARR_SZ 16

[Copilot]

[nitpick] The new batching loop for TCP_BOUND_INACTIVE in inet_diag_dump_icsk is quite large and complex. Consider extracting the bind-walk logic into a helper function to improve readability and reduce cyclomatic complexity.