mbox series

[net-next,v12,0/4] ethtool: provide the dim profile fine-tuning channel

Message ID 20240504064447.129622-1-hengqi@linux.alibaba.com (mailing list archive)
Headers show
Series ethtool: provide the dim profile fine-tuning channel | expand

Message

Heng Qi May 4, 2024, 6:44 a.m. UTC
The NetDIM library provides excellent acceleration for many modern
network cards. However, the default profiles of DIM limits its maximum
capabilities for different NICs, so providing a way which the NIC can
be custom configured is necessary.

Currently, the way is based on the commonly used "ethtool -C".

Please review, thank you very much!

Changelog
=====
v11->v12:
  - Remove the use of IS_ENABLED(DIMLIB).
  - Update Simon's htmldoc hint.

v10->v11:
  - Fix and clean up some issues from Kuba, thanks.
  - Rebase net-next/main

v9->v10:
  - Collect dim related flags/mode/work into one place.
  - Use rx_profile + tx_profile instead of four profiles.
  - Add several helps.
  - Update commit logs.

v8->v9:
  - Fix the compilation error of conflicting names of rx_profile in
    dim.h and ice driver: in dim.h, rx_profile is replaced with
    dim_rx_profile. So does tx_profile.

v7->v8:
  - Use kmemdup() instead of kzalloc()/memcpy() in dev_dim_profile_init().

v6->v7:
  - A new wrapper struct pointer is used in struct net_device.
  - Add IS_ENABLED(CONFIG_DIMLIB) to avoid compiler warnings.
  - Profile fields changed from u16 to u32.

v5->v6:
  - Place the profile in netdevice to bypass the driver.
    The interaction code of ethtool <-> kernel has not changed at all,
    only the interaction part of kernel <-> driver has changed.

v4->v5:
  - Update some snippets from Kuba.

v3->v4:
  - Some tiny updates and patch 1 only add a new comment.

v2->v3:
  - Break up the attributes to avoid the use of raw c structs.
  - Use per-device profile instead of global profile in the driver.

v1->v2:
  - Use ethtool tool instead of net-sysfs.

Heng Qi (4):
  linux/dim: move useful macros to .h file
  ethtool: provide customized dim profile management
  dim: add new interfaces for initialization and getting results
  virtio-net: support dim profile fine-tuning

 Documentation/netlink/specs/ethtool.yaml     |  31 +++
 Documentation/networking/ethtool-netlink.rst |   4 +
 drivers/net/virtio_net.c                     |  47 +++-
 include/linux/dim.h                          | 114 ++++++++
 include/linux/ethtool.h                      |   4 +-
 include/linux/netdevice.h                    |   3 +
 include/uapi/linux/ethtool_netlink.h         |  22 ++
 lib/dim/net_dim.c                            | 145 ++++++++++-
 net/ethtool/coalesce.c                       | 259 ++++++++++++++++++-
 9 files changed, 613 insertions(+), 16 deletions(-)

Comments

Heng Qi May 8, 2024, 2:12 a.m. UTC | #1
On Sat,  4 May 2024 14:44:43 +0800, Heng Qi <hengqi@linux.alibaba.com> wrote:
> The NetDIM library provides excellent acceleration for many modern
> network cards. However, the default profiles of DIM limits its maximum
> capabilities for different NICs, so providing a way which the NIC can
> be custom configured is necessary.
> 
> Currently, the way is based on the commonly used "ethtool -C".
> 
> Please review, thank you very much!

Hi,

I would like to confirm if there are still comments on the current version,
since the current series and the just merged "Remove RTNL lock protection of
CVQ" conflict with a line of code with the fourth patch, if I can collect
other comments or ack/review tags, then release the new version seems better.

Thank you very much!

> 
> Changelog
> =====
> v11->v12:
>   - Remove the use of IS_ENABLED(DIMLIB).
>   - Update Simon's htmldoc hint.
> 
> v10->v11:
>   - Fix and clean up some issues from Kuba, thanks.
>   - Rebase net-next/main
> 
> v9->v10:
>   - Collect dim related flags/mode/work into one place.
>   - Use rx_profile + tx_profile instead of four profiles.
>   - Add several helps.
>   - Update commit logs.
> 
> v8->v9:
>   - Fix the compilation error of conflicting names of rx_profile in
>     dim.h and ice driver: in dim.h, rx_profile is replaced with
>     dim_rx_profile. So does tx_profile.
> 
> v7->v8:
>   - Use kmemdup() instead of kzalloc()/memcpy() in dev_dim_profile_init().
> 
> v6->v7:
>   - A new wrapper struct pointer is used in struct net_device.
>   - Add IS_ENABLED(CONFIG_DIMLIB) to avoid compiler warnings.
>   - Profile fields changed from u16 to u32.
> 
> v5->v6:
>   - Place the profile in netdevice to bypass the driver.
>     The interaction code of ethtool <-> kernel has not changed at all,
>     only the interaction part of kernel <-> driver has changed.
> 
> v4->v5:
>   - Update some snippets from Kuba.
> 
> v3->v4:
>   - Some tiny updates and patch 1 only add a new comment.
> 
> v2->v3:
>   - Break up the attributes to avoid the use of raw c structs.
>   - Use per-device profile instead of global profile in the driver.
> 
> v1->v2:
>   - Use ethtool tool instead of net-sysfs.
> 
> Heng Qi (4):
>   linux/dim: move useful macros to .h file
>   ethtool: provide customized dim profile management
>   dim: add new interfaces for initialization and getting results
>   virtio-net: support dim profile fine-tuning
> 
>  Documentation/netlink/specs/ethtool.yaml     |  31 +++
>  Documentation/networking/ethtool-netlink.rst |   4 +
>  drivers/net/virtio_net.c                     |  47 +++-
>  include/linux/dim.h                          | 114 ++++++++
>  include/linux/ethtool.h                      |   4 +-
>  include/linux/netdevice.h                    |   3 +
>  include/uapi/linux/ethtool_netlink.h         |  22 ++
>  lib/dim/net_dim.c                            | 145 ++++++++++-
>  net/ethtool/coalesce.c                       | 259 ++++++++++++++++++-
>  9 files changed, 613 insertions(+), 16 deletions(-)
> 
> -- 
> 2.32.0.3.g01195cf9f
> 
>
Jakub Kicinski May 8, 2024, 2:47 a.m. UTC | #2
On Wed, 8 May 2024 10:12:35 +0800 Heng Qi wrote:
> I would like to confirm if there are still comments on the current version,
> since the current series and the just merged "Remove RTNL lock protection of
> CVQ" conflict with a line of code with the fourth patch, if I can collect
> other comments or ack/review tags, then release the new version seems better.

Looking now!

Please note that I merged a patch today which makes DIMLIB a tri-state
config, meaning it can be a module now. So please double check that
didn't break things, especially referring to dim symbols from the core
code.
Jakub Kicinski May 8, 2024, 3:01 a.m. UTC | #3
On Sat,  4 May 2024 14:44:43 +0800 Heng Qi wrote:
> The NetDIM library provides excellent acceleration for many modern
> network cards. However, the default profiles of DIM limits its maximum
> capabilities for different NICs, so providing a way which the NIC can
> be custom configured is necessary.
> 
> Currently, the way is based on the commonly used "ethtool -C".
> 
> Please review, thank you very much!

Good progress! Please make sure to also update
Documentation/networking/net_dim.rst in the next version.
Heng Qi May 8, 2024, 3:53 p.m. UTC | #4
On Tue, 7 May 2024 19:47:07 -0700, Jakub Kicinski <kuba@kernel.org> wrote:
> On Wed, 8 May 2024 10:12:35 +0800 Heng Qi wrote:
> > I would like to confirm if there are still comments on the current version,
> > since the current series and the just merged "Remove RTNL lock protection of
> > CVQ" conflict with a line of code with the fourth patch, if I can collect
> > other comments or ack/review tags, then release the new version seems better.
> 
> Looking now!
> 
> Please note that I merged a patch today which makes DIMLIB a tri-state
> config, meaning it can be a module now. So please double check that
> didn't break things, especially referring to dim symbols from the core
> code.


The transition from bool-state to tri-state has no impact on this series,
since functions in the DIM library used by external modules require
EXPORT_SYMBOL in both conditions.

Thanks.