mbox

[pull,request,net-next,00/16] mlx5 updates 2022-09-27

Message ID 20220927203611.244301-1-saeed@kernel.org (mailing list archive)
State Accepted
Headers show

Pull-request

git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5-updates-2022-09-27

Message

Saeed Mahameed Sept. 27, 2022, 8:35 p.m. UTC
From: Saeed Mahameed <saeedm@nvidia.com>

XSK buffer improvements, This is part #1 of 4 parts series.
For more information please see tag log below.

Please pull and let me know if there is any problem.

Thanks,
Saeed.

The following changes since commit 7bcd9683e51575c72c9289c05213150245d1c186:

  selftests/net: enable io_uring sendzc testing (2022-09-27 07:59:57 -0700)

are available in the Git repository at:

  git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux.git tags/mlx5-updates-2022-09-27

for you to fetch changes up to 48670cfdb211a57b41d870098bae659be5b453f7:

  net/mlx5e: Use runtime values of striding RQ parameters in datapath (2022-09-27 13:32:42 -0700)

----------------------------------------------------------------
mlx5-updates-2022-09-27

This is Part #1 of 4 parts series to align mlx5's implementation of
XSK (AF_XDP) RX-Qs indexing and management with other vendors:

Maxim Says:
===========

xsk: Bug fixes for frame mapping on striding RQ

Striding RQ relies on the driver mapping RX buffers into the NIC's
virtual memory space. Currently, regadless of the XSK frame size, mlx5e
maps them using MTT, and each mapping's length is PAGE_SIZE. As the
result, the stride size used by striding RQ is also equal to PAGE_SIZE.

This decision has the following issues:

1. In the XSK aligned mode with frame size smaller than PAGE_SIZE, it's
suboptimal. Using 2K strides and 2K pages allows to post twice as fewer
WQEs.

2. MTT is not suitable for unaligned frames, as it requires natural
alignment theoretically, in practice at least 8-byte alignment.

3. Using mapping and stride bigger than the frame has risk of writing
over the bounds of the XSK frame upon receiving packets bigger than MTU,
which is possible in some specific configurations.

This series addresses issues 1 and 2 and alleviates issue 3. Where
possible, page and stride size will match the XSK frame size (firmware
upgrade may be needed to have effect for 2K frames). Unaligned mode will
use KSM instead of MTT, which allows to drop the partial workaround [1].

[1]: https://lore.kernel.org/netdev/YufYFQ6JN91lQbso@boxer/T/

===========

----------------------------------------------------------------
Maxim Mikityanskiy (16):
      net/mlx5: Add the log_min_mkey_entity_size capability
      net/mlx5e: Convert mlx5e_get_max_sq_wqebbs to u8
      net/mlx5e: Remove unused fields from datapath structs
      net/mlx5e: Make mlx5e_verify_rx_mpwqe_strides static
      net/mlx5e: Validate striding RQ before enabling XDP
      net/mlx5e: Let mlx5e_get_sw_max_sq_mpw_wqebbs accept mdev
      net/mlx5e: Use mlx5e_stop_room_for_max_wqe where appropriate
      net/mlx5e: Fix a typo in mlx5e_xdp_mpwqe_is_full
      net/mlx5e: Use the aligned max TX MPWQE size
      net/mlx5e: kTLS, Check ICOSQ WQE size in advance
      net/mlx5e: Simplify stride size calculation for linear RQ
      net/mlx5e: xsk: Remove dead code in validation
      net/mlx5e: xsk: Fix SKB headroom calculation in validation
      net/mlx5e: Improve the MTU change shortcut
      net/mlx5e: Make dma_info array dynamic in struct mlx5e_mpw_info
      net/mlx5e: Use runtime values of striding RQ parameters in datapath

 drivers/net/ethernet/mellanox/mlx5/core/en.h       |  37 ++----
 .../net/ethernet/mellanox/mlx5/core/en/params.c    | 141 +++++++++++----------
 .../net/ethernet/mellanox/mlx5/core/en/params.h    |  15 ++-
 drivers/net/ethernet/mellanox/mlx5/core/en/txrx.h  |  13 +-
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.c   |   2 +-
 drivers/net/ethernet/mellanox/mlx5/core/en/xdp.h   |   2 +-
 .../net/ethernet/mellanox/mlx5/core/en/xsk/rx.c    |   2 +-
 .../net/ethernet/mellanox/mlx5/core/en/xsk/setup.c |   6 +-
 .../ethernet/mellanox/mlx5/core/en_accel/ktls.c    |  18 +++
 .../ethernet/mellanox/mlx5/core/en_accel/ktls.h    |   5 +-
 .../net/ethernet/mellanox/mlx5/core/en_ethtool.c   |  12 +-
 drivers/net/ethernet/mellanox/mlx5/core/en_main.c  |  85 ++++++++-----
 drivers/net/ethernet/mellanox/mlx5/core/en_rx.c    |  60 +++++----
 include/linux/mlx5/mlx5_ifc.h                      |   8 +-
 14 files changed, 230 insertions(+), 176 deletions(-)

Comments

Jakub Kicinski Sept. 29, 2022, 2:35 a.m. UTC | #1
On Tue, 27 Sep 2022 13:35:55 -0700 Saeed Mahameed wrote:
> XSK buffer improvements, This is part #1 of 4 parts series.
> For more information please see tag log below.
> 
> Please pull and let me know if there is any problem.

You're missing your s-o-bs in your tree.
I'll apply from the list, you'd have to rebase anyway.
Saeed Mahameed Sept. 29, 2022, 7:20 a.m. UTC | #2
On 28 Sep 19:35, Jakub Kicinski wrote:
>On Tue, 27 Sep 2022 13:35:55 -0700 Saeed Mahameed wrote:
>> XSK buffer improvements, This is part #1 of 4 parts series.
>> For more information please see tag log below.
>>
>> Please pull and let me know if there is any problem.
>
>You're missing your s-o-bs in your tree.
>I'll apply from the list, you'd have to rebase anyway.

Thanks ! I changed laptop and some of my scripts broke !

I will submit the 2nd part as patchset to avoid such mistakes
for now, it includes 2 simple non-mlx5 changes anyhow.
I would appreciate it if you expedite acceptance so I can post part3
and 4 in this cycle, and get mlx5's XSK issues behind me once and for all.
Jakub Kicinski Sept. 29, 2022, 3:33 p.m. UTC | #3
On Thu, 29 Sep 2022 00:20:43 -0700 Saeed Mahameed wrote:
> I would appreciate it if you expedite acceptance so I can post part3
> and 4 in this cycle, and get mlx5's XSK issues behind me once and for all.

Don't think it's gonna make it if Linus cuts final on Sunday :(
I'll look at the mlx5 patches as soon as you post but I don't
see any real reason to circumvent the normal list wait time.