| Message ID | 20230131204506.219292-1-maciej.fijalkowski@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | ice: add XDP mbuf support | expand |
From: Maciej Fijalkowski <maciej.fijalkowski@intel.com> Date: Tue, 31 Jan 2023 21:44:53 +0100 > Hi there, > > although this work started as an effort to add multi-buffer XDP support > to ice driver, as usual it turned out that some other side stuff needed > to be addressed, so let me give you an overview. > > First patch adjusts legacy-rx in a way that it will be possible to refer > to skb_shared_info being at the end of the buffer when gathering up > frame fragments within xdp_buff. > > Then, patches 2-9 prepare ice driver in a way that actual multi-buffer > patches will be easier to swallow. > > 10 and 11 are the meat. What is worth mentioning is that this set > actually *fixes* things as patch 11 removes the logic based on > next_dd/rs and we previously stepped away from this for ice_xmit_zc(). > Currently, AF_XDP ZC XDP_TX workload is off as there are two cleaning > sides that can be triggered and two of them work on different internal > logic. This set unifies that and allows us to improve the performance by > 2x with a trick on the last (13) patch. > > 12th is a simple cleanup of no longer fields from Tx ring. > > I might be wrong but I have not seen anyone reporting performance impact > among patches that add XDP multi-buffer support to a particular driver. > Numbers below were gathered via xdp_rxq_info and xdp_redirect_map on > 1500 MTU: > > XDP_DROP +1% > XDP_PASS -1,2% > XDP_TX -0,5% > XDP_REDIRECT -3,3% > > Cherry on top, which is not directly related to mbuf support (last > patch): > XDP_TX ZC +126% > > Target the we agreed on was to not degrade performance for any action by > anything that would be over 5%, so our goal was met. Basically this set > keeps the performance where it was. Redirect is slower due to more > frequent tail bumps. > > Thanks! You forgot to add my Reviewed-by: Alexander Lobakin <alexandr.lobakin@intel.com> for the whole series :D > > > Maciej Fijalkowski (13): > ice: prepare legacy-rx for upcoming XDP multi-buffer support > ice: add xdp_buff to ice_rx_ring struct > ice: store page count inside ice_rx_buf > ice: pull out next_to_clean bump out of ice_put_rx_buf() > ice: inline eop check > ice: centrallize Rx buffer recycling > ice: use ice_max_xdp_frame_size() in ice_xdp_setup_prog() > ice: do not call ice_finalize_xdp_rx() unnecessarily > ice: use xdp->frame_sz instead of recalculating truesize > ice: add support for XDP multi-buffer on Rx side > ice: add support for XDP multi-buffer on Tx side > ice: remove next_{dd,rs} fields from ice_tx_ring > ice: xsk: do not convert to buff to frame for XDP_TX > > drivers/net/ethernet/intel/ice/ice_base.c | 21 +- > drivers/net/ethernet/intel/ice/ice_ethtool.c | 4 +- > drivers/net/ethernet/intel/ice/ice_lib.c | 8 +- > drivers/net/ethernet/intel/ice/ice_main.c | 47 +- > drivers/net/ethernet/intel/ice/ice_txrx.c | 408 ++++++++++-------- > drivers/net/ethernet/intel/ice/ice_txrx.h | 54 ++- > drivers/net/ethernet/intel/ice/ice_txrx_lib.c | 236 ++++++---- > drivers/net/ethernet/intel/ice/ice_txrx_lib.h | 75 +++- > drivers/net/ethernet/intel/ice/ice_xsk.c | 192 +++++---- > 9 files changed, 629 insertions(+), 416 deletions(-) Thanks, Olek
Hello: This series was applied to bpf/bpf-next.git (master) by Daniel Borkmann <daniel@iogearbox.net>: On Tue, 31 Jan 2023 21:44:53 +0100 you wrote: > Hi there, > > although this work started as an effort to add multi-buffer XDP support > to ice driver, as usual it turned out that some other side stuff needed > to be addressed, so let me give you an overview. > > First patch adjusts legacy-rx in a way that it will be possible to refer > to skb_shared_info being at the end of the buffer when gathering up > frame fragments within xdp_buff. > > [...] Here is the summary with links: - [bpf-next,01/13] ice: prepare legacy-rx for upcoming XDP multi-buffer support https://git.kernel.org/bpf/bpf-next/c/c61bcebde72d - [bpf-next,02/13] ice: add xdp_buff to ice_rx_ring struct https://git.kernel.org/bpf/bpf-next/c/cb0473e0e9dc - [bpf-next,03/13] ice: store page count inside ice_rx_buf https://git.kernel.org/bpf/bpf-next/c/ac0753391195 - [bpf-next,04/13] ice: pull out next_to_clean bump out of ice_put_rx_buf() https://git.kernel.org/bpf/bpf-next/c/d7956d81f150 - [bpf-next,05/13] ice: inline eop check https://git.kernel.org/bpf/bpf-next/c/e44f4790a2ba - [bpf-next,06/13] ice: centrallize Rx buffer recycling https://git.kernel.org/bpf/bpf-next/c/1dc1a7e7f410 - [bpf-next,07/13] ice: use ice_max_xdp_frame_size() in ice_xdp_setup_prog() https://git.kernel.org/bpf/bpf-next/c/60bc72b3c4e9 - [bpf-next,08/13] ice: do not call ice_finalize_xdp_rx() unnecessarily https://git.kernel.org/bpf/bpf-next/c/9070fe3da0b1 - [bpf-next,09/13] ice: use xdp->frame_sz instead of recalculating truesize https://git.kernel.org/bpf/bpf-next/c/8a11b334ec9b - [bpf-next,10/13] ice: add support for XDP multi-buffer on Rx side https://git.kernel.org/bpf/bpf-next/c/2fba7dc5157b - [bpf-next,11/13] ice: add support for XDP multi-buffer on Tx side https://git.kernel.org/bpf/bpf-next/c/3246a10752a7 - [bpf-next,12/13] ice: remove next_{dd,rs} fields from ice_tx_ring https://git.kernel.org/bpf/bpf-next/c/f4db7b314dd5 - [bpf-next,13/13] ice: xsk: do not convert to buff to frame for XDP_TX https://git.kernel.org/bpf/bpf-next/c/a24b4c6e9aab You are awesome, thank you!
Hi there, although this work started as an effort to add multi-buffer XDP support to ice driver, as usual it turned out that some other side stuff needed to be addressed, so let me give you an overview. First patch adjusts legacy-rx in a way that it will be possible to refer to skb_shared_info being at the end of the buffer when gathering up frame fragments within xdp_buff. Then, patches 2-9 prepare ice driver in a way that actual multi-buffer patches will be easier to swallow. 10 and 11 are the meat. What is worth mentioning is that this set actually *fixes* things as patch 11 removes the logic based on next_dd/rs and we previously stepped away from this for ice_xmit_zc(). Currently, AF_XDP ZC XDP_TX workload is off as there are two cleaning sides that can be triggered and two of them work on different internal logic. This set unifies that and allows us to improve the performance by 2x with a trick on the last (13) patch. 12th is a simple cleanup of no longer fields from Tx ring. I might be wrong but I have not seen anyone reporting performance impact among patches that add XDP multi-buffer support to a particular driver. Numbers below were gathered via xdp_rxq_info and xdp_redirect_map on 1500 MTU: XDP_DROP +1% XDP_PASS -1,2% XDP_TX -0,5% XDP_REDIRECT -3,3% Cherry on top, which is not directly related to mbuf support (last patch): XDP_TX ZC +126% Target the we agreed on was to not degrade performance for any action by anything that would be over 5%, so our goal was met. Basically this set keeps the performance where it was. Redirect is slower due to more frequent tail bumps. Thanks! Maciej Fijalkowski (13): ice: prepare legacy-rx for upcoming XDP multi-buffer support ice: add xdp_buff to ice_rx_ring struct ice: store page count inside ice_rx_buf ice: pull out next_to_clean bump out of ice_put_rx_buf() ice: inline eop check ice: centrallize Rx buffer recycling ice: use ice_max_xdp_frame_size() in ice_xdp_setup_prog() ice: do not call ice_finalize_xdp_rx() unnecessarily ice: use xdp->frame_sz instead of recalculating truesize ice: add support for XDP multi-buffer on Rx side ice: add support for XDP multi-buffer on Tx side ice: remove next_{dd,rs} fields from ice_tx_ring ice: xsk: do not convert to buff to frame for XDP_TX drivers/net/ethernet/intel/ice/ice_base.c | 21 +- drivers/net/ethernet/intel/ice/ice_ethtool.c | 4 +- drivers/net/ethernet/intel/ice/ice_lib.c | 8 +- drivers/net/ethernet/intel/ice/ice_main.c | 47 +- drivers/net/ethernet/intel/ice/ice_txrx.c | 408 ++++++++++-------- drivers/net/ethernet/intel/ice/ice_txrx.h | 54 ++- drivers/net/ethernet/intel/ice/ice_txrx_lib.c | 236 ++++++---- drivers/net/ethernet/intel/ice/ice_txrx_lib.h | 75 +++- drivers/net/ethernet/intel/ice/ice_xsk.c | 192 +++++---- 9 files changed, 629 insertions(+), 416 deletions(-)