[net-next,v4,0/5] HTB offload

Message

Maxim Mikityanskiy Jan. 19, 2021, 12:08 p.m. UTC

This series adds support for HTB offload to the HTB qdisc, and adds
usage to mlx5 driver.

The previous RFCs are available at [1], [2].

The feature is intended to solve the performance bottleneck caused by
the single lock of the HTB qdisc, which prevents it from scaling well.
The HTB algorithm itself is offloaded to the device, eliminating the
need to take the root lock of HTB on every packet. Classification part
is done in clsact (still in software) to avoid acquiring the lock, which
imposes a limitation that filters can target only leaf classes.

The speedup on Mellanox ConnectX-6 Dx was 14.2 times in the UDP
multi-stream test, compared to software HTB implementation (more details
in the mlx5 patch).

[1]: https://www.spinics.net/lists/netdev/msg628422.html
[2]: https://www.spinics.net/lists/netdev/msg663548.html

v2 changes:

Fixed sparse and smatch warnings. Formatted HTB patches to 80 chars per
line.

v3 changes:

Fixed the CI failure on parisc with 16-bit xchg by replacing it with
WRITE_ONCE. Fixed the capability bits in mlx5_ifc.h and the value of
MLX5E_QOS_MAX_LEAF_NODES.

v4 changes:

Check if HTB is root when offloading. Add extack for hardware errors.
Rephrase explanations of how it works in the commit message. Remove %hu
from format strings. Add resiliency when leaf_del_last fails to create a
new leaf node.

Maxim Mikityanskiy (5):
  net: sched: Add multi-queue support to sch_tree_lock
  net: sched: Add extack to Qdisc_class_ops.delete
  sch_htb: Hierarchical QoS hardware offload
  sch_htb: Stats for offloaded HTB
  net/mlx5e: Support HTB offload

 .../net/ethernet/mellanox/mlx5/core/Makefile  |   6 +-
 drivers/net/ethernet/mellanox/mlx5/core/en.h  |  27 +-
 .../ethernet/mellanox/mlx5/core/en/params.h   |   2 +
 .../net/ethernet/mellanox/mlx5/core/en/ptp.c  |   2 +-
 .../net/ethernet/mellanox/mlx5/core/en/qos.c  | 984 ++++++++++++++++++
 .../net/ethernet/mellanox/mlx5/core/en/qos.h  |  44 +
 .../ethernet/mellanox/mlx5/core/en_ethtool.c  |  21 +
 .../net/ethernet/mellanox/mlx5/core/en_main.c | 176 +++-
 .../ethernet/mellanox/mlx5/core/en_stats.c    | 100 ++
 .../ethernet/mellanox/mlx5/core/en_stats.h    |   2 +
 .../net/ethernet/mellanox/mlx5/core/en_tx.c   |  47 +-
 .../net/ethernet/mellanox/mlx5/core/en_txrx.c |  26 +
 drivers/net/ethernet/mellanox/mlx5/core/qos.c |  85 ++
 drivers/net/ethernet/mellanox/mlx5/core/qos.h |  30 +
 include/linux/mlx5/mlx5_ifc.h                 |  13 +-
 include/linux/netdevice.h                     |   1 +
 include/net/pkt_cls.h                         |  36 +
 include/net/sch_generic.h                     |  17 +-
 include/uapi/linux/pkt_sched.h                |   1 +
 net/sched/sch_api.c                           |   7 +-
 net/sched/sch_atm.c                           |   3 +-
 net/sched/sch_cbq.c                           |   3 +-
 net/sched/sch_drr.c                           |   3 +-
 net/sched/sch_dsmark.c                        |   3 +-
 net/sched/sch_hfsc.c                          |   3 +-
 net/sched/sch_htb.c                           | 557 +++++++++-
 net/sched/sch_qfq.c                           |   3 +-
 net/sched/sch_sfb.c                           |   3 +-
 tools/include/uapi/linux/pkt_sched.h          |   1 +
 29 files changed, 2113 insertions(+), 93 deletions(-)
 create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en/qos.c
 create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/en/qos.h
 create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/qos.c
 create mode 100644 drivers/net/ethernet/mellanox/mlx5/core/qos.h

Comments

[e-mail] Jan. 23, 2021, 6:10 a.m. UTC | #1

Hello:

This series was applied to netdev/net-next.git (refs/heads/master):

On Tue, 19 Jan 2021 14:08:10 +0200 you wrote:
> This series adds support for HTB offload to the HTB qdisc, and adds
> usage to mlx5 driver.
> 
> The previous RFCs are available at [1], [2].
> 
> The feature is intended to solve the performance bottleneck caused by
> the single lock of the HTB qdisc, which prevents it from scaling well.
> The HTB algorithm itself is offloaded to the device, eliminating the
> need to take the root lock of HTB on every packet. Classification part
> is done in clsact (still in software) to avoid acquiring the lock, which
> imposes a limitation that filters can target only leaf classes.
> 
> [...]

Here is the summary with links:
  - [net-next,v4,1/5] net: sched: Add multi-queue support to sch_tree_lock
    https://git.kernel.org/netdev/net-next/c/ca1e4ab19993
  - [net-next,v4,2/5] net: sched: Add extack to Qdisc_class_ops.delete
    https://git.kernel.org/netdev/net-next/c/4dd78a73738a
  - [net-next,v4,3/5] sch_htb: Hierarchical QoS hardware offload
    https://git.kernel.org/netdev/net-next/c/d03b195b5aa0
  - [net-next,v4,4/5] sch_htb: Stats for offloaded HTB
    https://git.kernel.org/netdev/net-next/c/83271586249c
  - [net-next,v4,5/5] net/mlx5e: Support HTB offload
    https://git.kernel.org/netdev/net-next/c/214baf22870c

You are awesome, thank you!
--
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html