Note: it's net/ only bits and doesn't include changes, which shoulf be merged separately and are posted separately. The full branch for convenience is at [1], and the patch is here:
https://lore.kernel.org/io-uring/7486ab32e99be1f614b3ef8d0e9bc77015b173f7.17...
Many modern NICs support configurable receive buffer lengths, and zcrx and memory providers can use buffers larger than 4K/PAGE_SIZE on x86 to improve performance. When paired with hw-gro larger rx buffer sizes can drastically reduce the number of buffers traversing the stack and save a lot of processing time. It also allows to give to users larger contiguous chunks of data. The idea was first floated around by Saeed during netdev conf 2024 and was asked about by a few folks.
Single stream benchmarks showed up to ~30% CPU util improvement. E.g. comparison for 4K vs 32K buffers using a 200Gbit NIC:
packets=23987040 (MB=2745098), rps=199559 (MB/s=22837) CPU %usr %nice %sys %iowait %irq %soft %idle 0 1.53 0.00 27.78 2.72 1.31 66.45 0.22 packets=24078368 (MB=2755550), rps=200319 (MB/s=22924) CPU %usr %nice %sys %iowait %irq %soft %idle 0 0.69 0.00 8.26 31.65 1.83 57.00 0.57
This series adds net infrastructure for memory providers configuring the size and implements it for bnxt. It's an opt-in feature for drivers, they should advertise support for the parameter in the qops and must check if the hardware supports the given size. It's limited to memory providers as it drastically simplifies implementation. It doesn't affect the fast path zcrx uAPI, and the sizes is defined in zcrx terms, which allows it to be flexible and adjusted in the future, see Patch 8 for details.
A liburing example can be found at [2]
full branch: [1] https://github.com/isilence/linux.git zcrx/large-buffers-v7 Liburing example: [2] https://github.com/isilence/liburing.git zcrx/rx-buf-len
v7: - Add xa_destroy - Rebase
v6: - Update docs and add a selftest
v5: https://lore.kernel.org/netdev/cover.1760440268.git.asml.silence@gmail.com/ - Remove all unnecessary bits like configuration via netlink, and multi-stage queue configuration.
v4: https://lore.kernel.org/all/cover.1760364551.git.asml.silence@gmail.com/ - Update fbnic qops - Propagate max buf len for hns3 - Use configured buf size in __bnxt_alloc_rx_netmem - Minor stylistic changes v3: https://lore.kernel.org/all/cover.1755499375.git.asml.silence@gmail.com/ - Rebased, excluded zcrx specific patches - Set agg_size_fac to 1 on warning v2: https://lore.kernel.org/all/cover.1754657711.git.asml.silence@gmail.com/ - Add MAX_PAGE_ORDER check on pp init - Applied comments rewording - Adjust pp.max_len based on order - Patch up mlx5 queue callbacks after rebase - Minor ->queue_mgmt_ops refactoring - Rebased to account for both fill level and agg_size_fac - Pass providers buf length in struct pp_memory_provider_params and apply it in __netdev_queue_confi(). - Use ->supported_ring_params to validate drivers support of set qcfg parameters.
Jakub Kicinski (1): eth: bnxt: adjust the fill level of agg queues with larger buffers
Pavel Begunkov (8): net: page pool: xa init with destroy on pp init net: page_pool: sanitise allocation order net: memzero mp params when closing a queue net: let pp memory provider to specify rx buf len eth: bnxt: store rx buffer size per queue eth: bnxt: allow providers to set rx buf size io_uring/zcrx: document area chunking parameter selftests: iou-zcrx: test large chunk sizes
Documentation/networking/iou-zcrx.rst | 20 +++ drivers/net/ethernet/broadcom/bnxt/bnxt.c | 118 ++++++++++++++---- drivers/net/ethernet/broadcom/bnxt/bnxt.h | 2 + drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.c | 6 +- drivers/net/ethernet/broadcom/bnxt/bnxt_xdp.h | 2 +- include/net/netdev_queues.h | 9 ++ include/net/page_pool/types.h | 1 + net/core/netdev_rx_queue.c | 14 ++- net/core/page_pool.c | 4 + .../selftests/drivers/net/hw/iou-zcrx.c | 72 +++++++++-- .../selftests/drivers/net/hw/iou-zcrx.py | 37 ++++++ 11 files changed, 236 insertions(+), 49 deletions(-)