| Message ID | 20231027232252.36111-1-michael.chan@broadcom.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | bnxt_en: TX path improvements | expand |
On Fri, 27 Oct 2023 16:22:39 -0700 Michael Chan wrote: > All patches in this patchset are related to improving the TX path. > There are 2 areas of improvements: > > 1. The TX interrupt logic currently counts the number of TX completions > to determine the number of TX SKBs to free. We now change it so that > the TX completion will now contain the hardware consumer index > information. The driver will keep track of the latest hardware > consumer index from the last TX completion and clean up all TX SKBs > up to that index. This scheme aligns better with future chips and > allows xmit_more code path to be more optimized. > > 2. The current driver logic requires an additional MSIX for each > additional MQPRIO TX ring. This scheme uses too many MSIX vectors if > the user enables a large number of MQPRIO TCs. We now use a new scheme > that will use the same MSIX for all the MQPRIO TX rings for each > ethtool channel. Each ethtool TX channel can have up to 8 MQPRIO > TX rings and now they all will share the same MSIX. ## Form letter - net-next-closed The merge window for v6.7 has begun and we have already posted our pull request. Therefore net-next is closed for new drivers, features, code refactoring and optimizations. We are currently accepting bug fixes only. Please repost when net-next reopens after Nov 12th. RFC patches sent for review only are obviously welcome at any time. See: https://www.kernel.org/doc/html/next/process/maintainer-netdev.html#development-cycle
All patches in this patchset are related to improving the TX path. There are 2 areas of improvements: 1. The TX interrupt logic currently counts the number of TX completions to determine the number of TX SKBs to free. We now change it so that the TX completion will now contain the hardware consumer index information. The driver will keep track of the latest hardware consumer index from the last TX completion and clean up all TX SKBs up to that index. This scheme aligns better with future chips and allows xmit_more code path to be more optimized. 2. The current driver logic requires an additional MSIX for each additional MQPRIO TX ring. This scheme uses too many MSIX vectors if the user enables a large number of MQPRIO TCs. We now use a new scheme that will use the same MSIX for all the MQPRIO TX rings for each ethtool channel. Each ethtool TX channel can have up to 8 MQPRIO TX rings and now they all will share the same MSIX. Michael Chan (13): bnxt_en: Put the TX producer information in the TX BD opaque field bnxt_en: Add completion ring pointer in TX and RX ring structures bnxt_en: Restructure cp_ring_arr in struct bnxt_cp_ring_info bnxt_en: Add completion ring pointer in TX and RX ring structures bnxt_en: Remove BNXT_RX_HDL and BNXT_TX_HDL bnxt_en: Refactor bnxt_tx_int() bnxt_en: New encoding for the TX opaque field bnxt_en: Refactor bnxt_hwrm_set_coal() bnxt_en: Support up to 8 TX rings per MSIX bnxt_en: Add helper to get the number of CP rings required for TX rings bnxt_en: Add macros related to TC and TX rings bnxt_en: Use existing MSIX vectors for all mqprio TX rings bnxt_en: Optimize xmit_more TX path drivers/net/ethernet/broadcom/bnxt/bnxt.c | 513 +++++++++++------- drivers/net/ethernet/broadcom/bnxt/bnxt.h | 59 +- .../net/ethernet/broadcom/bnxt/bnxt_ethtool.c | 11 +- drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c | 23 +- 4 files changed, 385 insertions(+), 221 deletions(-)