| Message ID | 20210922075613.12186-1-magnus.karlsson@gmail.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | xsk: i40e: ice: introduce batching for Rx buffer allocation | expand |
Hello: This series was applied to bpf/bpf-next.git (refs/heads/master): On Wed, 22 Sep 2021 09:56:00 +0200 you wrote: > This patch set introduces a batched interface for Rx buffer allocation > in AF_XDP buffer pool. Instead of using xsk_buff_alloc(*pool), drivers > can now use xsk_buff_alloc_batch(*pool, **xdp_buff_array, > max). Instead of returning a pointer to an xdp_buff, it returns the > number of xdp_buffs it managed to allocate up to the maximum value of > the max parameter in the function call. Pointers to the allocated > xdp_buff:s are put in the xdp_buff_array supplied in the call. This > could be a SW ring that already exists in the driver or a new > structure that the driver has allocated. > > [...] Here is the summary with links: - [bpf-next,01/13] xsk: get rid of unused entry in struct xdp_buff_xsk https://git.kernel.org/bpf/bpf-next/c/10a5e009b93a - [bpf-next,02/13] xsk: batched buffer allocation for the pool https://git.kernel.org/bpf/bpf-next/c/47e4075df300 - [bpf-next,03/13] ice: use xdp_buf instead of rx_buf for xsk zero-copy https://git.kernel.org/bpf/bpf-next/c/57f7f8b6bc0b - [bpf-next,04/13] ice: use the xsk batched rx allocation interface https://git.kernel.org/bpf/bpf-next/c/db804cfc21e9 - [bpf-next,05/13] i40e: use the xsk batched rx allocation interface https://git.kernel.org/bpf/bpf-next/c/6aab0bb0c5cd - [bpf-next,06/13] xsk: optimize for aligned case https://git.kernel.org/bpf/bpf-next/c/94033cd8e73b - [bpf-next,07/13] selftests: xsk: fix missing initialization https://git.kernel.org/bpf/bpf-next/c/5b132056123d - [bpf-next,08/13] selftests: xsk: put the same buffer only once in the fill ring https://git.kernel.org/bpf/bpf-next/c/872a1184dbf2 - [bpf-next,09/13] selftests: xsk: fix socket creation retry https://git.kernel.org/bpf/bpf-next/c/89013b8a2928 - [bpf-next,10/13] selftests: xsk: introduce pacing of traffic https://git.kernel.org/bpf/bpf-next/c/1bf3649688c1 - [bpf-next,11/13] selftests: xsk: add single packet test https://git.kernel.org/bpf/bpf-next/c/96a40678ce53 - [bpf-next,12/13] selftests: xsk: change interleaving of packets in unaligned mode https://git.kernel.org/bpf/bpf-next/c/e4e9baf06a6e - [bpf-next,13/13] selftests: xsk: add frame_headroom test https://git.kernel.org/bpf/bpf-next/c/e34087fc00f4 You are awesome, thank you! -- Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
This patch set introduces a batched interface for Rx buffer allocation in AF_XDP buffer pool. Instead of using xsk_buff_alloc(*pool), drivers can now use xsk_buff_alloc_batch(*pool, **xdp_buff_array, max). Instead of returning a pointer to an xdp_buff, it returns the number of xdp_buffs it managed to allocate up to the maximum value of the max parameter in the function call. Pointers to the allocated xdp_buff:s are put in the xdp_buff_array supplied in the call. This could be a SW ring that already exists in the driver or a new structure that the driver has allocated. u32 xsk_buff_alloc_batch(struct xsk_buff_pool *pool, struct xdp_buff **xdp, u32 max); When using this interface, the driver should also use the new interface below to set the relevant fields in the struct xdp_buff. The reason for this is that xsk_buff_alloc_batch() does not fill in the data and data_meta fields for you as is the case with xsk_buff_alloc(). So it is not sufficient to just set data_end (effectively the size) anymore in the driver. The reason for this is performance as explained in detail in the commit message. void xsk_buff_set_size(struct xdp_buff *xdp, u32 size); Patch 6 also optimizes the buffer allocation in the aligned case. In this case, we can skip the reinitialization of most fields in the xdp_buff_xsk struct at allocation time. As the number of elements in the heads array is equal to the number of possible buffers in the umem, we can initialize them once and for all at bind time and then just point to the correct one in the xdp_buff_array that is returned to the driver. No reason to have a stack of free head entries. In the unaligned case, the buffers can reside anywhere in the umem, so this optimization is not possible as we still have to fill in the right information in the xdp_buff every single time one is allocated. I have updated i40e and ice to use this new batched interface. These are the throughput results on my 2.1 GHz Cascade Lake system: Aligned mode: ice: +11% / -9 cycles/pkt i40e: +12% / -9 cycles/pkt Unaligned mode: ice: +1.5% / -1 cycle/pkt i40e: +1% / -1 cycle/pkt For the aligned case, batching provides around 40% of the performance improvement and the aligned optimization the rest, around 60%. Would have expected a ~4% boost for unaligned with this data, but I only get around 1%. Do not know why. Note that memory consumption in aligned mode is also reduced by this patch set. Structure of the patch set: Patch 1: Removes an unused entry from xdp_buff_xsk. Patch 2: Introduce the batched buffer allocation API and implementation. Patch 3-4: Use the batched allocation interface for ice. Patch 5: Use the batched allocation interface for i40e. Patch 6: Optimize the buffer allocation for the aligned case. Patch 7-10: Fix some issues with the tests that were found while implementing the two new tests below. Patch 11-13: Implement two new tests: single packet and headroom validation. Thanks: Magnus Magnus Karlsson (13): xsk: get rid of unused entry in struct xdp_buff_xsk xsk: batched buffer allocation for the pool ice: use xdp_buf instead of rx_buf for xsk zero-copy ice: use the xsk batched rx allocation interface i40e: use the xsk batched rx allocation interface xsk: optimize for aligned case selftests: xsk: fix missing initialization selftests: xsk: put the same buffer only once in the fill ring selftests: xsk: fix socket creation retry selftests: xsk: introduce pacing of traffic selftests: xsk: add single packet test selftests: xsk: change interleaving of packets in unaligned mode selftests: xsk: add frame_headroom test drivers/net/ethernet/intel/i40e/i40e_xsk.c | 52 ++++---- drivers/net/ethernet/intel/ice/ice_txrx.h | 16 +-- drivers/net/ethernet/intel/ice/ice_xsk.c | 92 +++++++------- include/net/xdp_sock_drv.h | 22 ++++ include/net/xsk_buff_pool.h | 48 +++++++- net/xdp/xsk.c | 15 --- net/xdp/xsk_buff_pool.c | 131 +++++++++++++++++--- net/xdp/xsk_queue.h | 12 +- tools/testing/selftests/bpf/xdpxceiver.c | 133 ++++++++++++++++----- tools/testing/selftests/bpf/xdpxceiver.h | 11 +- 10 files changed, 376 insertions(+), 156 deletions(-) base-commit: 17b52c226a9a170f1611f69d12a71be05748aefd -- 2.29.0