| Message ID | 20231223025554.2316836-1-aleksander.lobakin@intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Christmas 3-serie XDP for idpf (+generic stuff) | expand |
Alexander Lobakin wrote: > I was highly asked to send this WIP before the holidays to trigger > some discussions at least for the generic parts. > > This all depends on libie[0] and WB-on-ITR fix[1]. The RFC does not > guarantee to work perfectly, but at least regular XDP seems to work > for me... > > In fact, here are 3 separate series: > * 01-08: convert idpf to libie and make it more sane; > * 09-25: add XDP to idpf; > * 26-34: add XSk to idpf. > > Most people may want to be interested only in the following generic > changes: > * 11: allow attaching already registered memory models to XDP RxQ info; > * 12-13: generic helpers for adding a frag to &xdp_buff and converting > it to an skb; > * 14: get rid of xdp_frame::mem.id, allow mixing pages from different > page_pools within one &xdp_buff/&xdp_frame; > * 15: some Page Pool helper; > * 18: it's for libie, but I wanted to talk about XDP_TX bulking; > * 26: same as 13, but for converting XSK &xdp_buff to skb. > > The rest is up to you, driver-specific stuff is pretty boring sometimes. > > I'll be polishing and finishing this all starting January 3rd and then > preparing and sending sane series, some early feedback never hurts tho. > > Merry Yule! > > [0] https://lore.kernel.org/netdev/20231213112835.2262651-1-aleksander.lobakin@intel.com > [1] https://lore.kernel.org/netdev/20231215193721.425087-1-michal.kubiak@intel.com This is great. Thanks for sharing the entire series. Which SHA1 should we apply this to? I'm having a hard time applying cleanly. The libie v7 series applied cleanly on bc044ae9d64b. Which I chose only based on the follow-on page pool patch. But that base commit causes too many conflicts when applying this. Patch 6 had a trivial one in idpf_rx_singleq_clean (`skb = rx_q->skb`). But patch 14 has so many conflicts in page_pool.c, that I'm clearly on the wrong track trying to fix up manually.
From: Willem De Bruijn <willemdebruijn.kernel@gmail.com> Date: Tue, 26 Dec 2023 15:23:41 -0500 > Alexander Lobakin wrote: >> I was highly asked to send this WIP before the holidays to trigger >> some discussions at least for the generic parts. >> >> This all depends on libie[0] and WB-on-ITR fix[1]. The RFC does not >> guarantee to work perfectly, but at least regular XDP seems to work >> for me... >> >> In fact, here are 3 separate series: >> * 01-08: convert idpf to libie and make it more sane; >> * 09-25: add XDP to idpf; >> * 26-34: add XSk to idpf. >> >> Most people may want to be interested only in the following generic >> changes: >> * 11: allow attaching already registered memory models to XDP RxQ info; >> * 12-13: generic helpers for adding a frag to &xdp_buff and converting >> it to an skb; >> * 14: get rid of xdp_frame::mem.id, allow mixing pages from different >> page_pools within one &xdp_buff/&xdp_frame; >> * 15: some Page Pool helper; >> * 18: it's for libie, but I wanted to talk about XDP_TX bulking; >> * 26: same as 13, but for converting XSK &xdp_buff to skb. >> >> The rest is up to you, driver-specific stuff is pretty boring sometimes. >> >> I'll be polishing and finishing this all starting January 3rd and then >> preparing and sending sane series, some early feedback never hurts tho. >> >> Merry Yule! >> >> [0] https://lore.kernel.org/netdev/20231213112835.2262651-1-aleksander.lobakin@intel.com >> [1] https://lore.kernel.org/netdev/20231215193721.425087-1-michal.kubiak@intel.com > > This is great. Thanks for sharing the entire series. > > Which SHA1 should we apply this to? I'm having a hard time applying > cleanly. > > The libie v7 series applied cleanly on bc044ae9d64b. Which I chose > only based on the follow-on page pool patch. > > But that base commit causes too many conflicts when applying this. > Patch 6 had a trivial one in idpf_rx_singleq_clean (`skb = rx_q->skb`). > But patch 14 has so many conflicts in page_pool.c, that I'm clearly > on the wrong track trying to fix up manually. net-next was updated while I was preparing the series. I also did a couple changes in the basic libie code, but a new rev wasn't sent. Please just use my open GH[0]. [0] https://github.com/alobakin/linux/tree/idpf-libie Thanks, Olek
Alexander Lobakin wrote: > From: Willem De Bruijn <willemdebruijn.kernel@gmail.com> > Date: Tue, 26 Dec 2023 15:23:41 -0500 > > > Alexander Lobakin wrote: > >> I was highly asked to send this WIP before the holidays to trigger > >> some discussions at least for the generic parts. > >> > >> This all depends on libie[0] and WB-on-ITR fix[1]. The RFC does not > >> guarantee to work perfectly, but at least regular XDP seems to work > >> for me... > >> > >> In fact, here are 3 separate series: > >> * 01-08: convert idpf to libie and make it more sane; > >> * 09-25: add XDP to idpf; > >> * 26-34: add XSk to idpf. > >> > >> Most people may want to be interested only in the following generic > >> changes: > >> * 11: allow attaching already registered memory models to XDP RxQ info; > >> * 12-13: generic helpers for adding a frag to &xdp_buff and converting > >> it to an skb; > >> * 14: get rid of xdp_frame::mem.id, allow mixing pages from different > >> page_pools within one &xdp_buff/&xdp_frame; > >> * 15: some Page Pool helper; > >> * 18: it's for libie, but I wanted to talk about XDP_TX bulking; > >> * 26: same as 13, but for converting XSK &xdp_buff to skb. > >> > >> The rest is up to you, driver-specific stuff is pretty boring sometimes. > >> > >> I'll be polishing and finishing this all starting January 3rd and then > >> preparing and sending sane series, some early feedback never hurts tho. > >> > >> Merry Yule! > >> > >> [0] https://lore.kernel.org/netdev/20231213112835.2262651-1-aleksander.lobakin@intel.com > >> [1] https://lore.kernel.org/netdev/20231215193721.425087-1-michal.kubiak@intel.com > > > > This is great. Thanks for sharing the entire series. > > > > Which SHA1 should we apply this to? I'm having a hard time applying > > cleanly. > > > > The libie v7 series applied cleanly on bc044ae9d64b. Which I chose > > only based on the follow-on page pool patch. > > > > But that base commit causes too many conflicts when applying this. > > Patch 6 had a trivial one in idpf_rx_singleq_clean (`skb = rx_q->skb`). > > But patch 14 has so many conflicts in page_pool.c, that I'm clearly > > on the wrong track trying to fix up manually. > > net-next was updated while I was preparing the series. I also did a > couple changes in the basic libie code, but a new rev wasn't sent. > Please just use my open GH[0]. > > [0] https://github.com/alobakin/linux/tree/idpf-libie Even better, thanks. I'll use that to run my basic XSK tests.
I was highly asked to send this WIP before the holidays to trigger some discussions at least for the generic parts. This all depends on libie[0] and WB-on-ITR fix[1]. The RFC does not guarantee to work perfectly, but at least regular XDP seems to work for me... In fact, here are 3 separate series: * 01-08: convert idpf to libie and make it more sane; * 09-25: add XDP to idpf; * 26-34: add XSk to idpf. Most people may want to be interested only in the following generic changes: * 11: allow attaching already registered memory models to XDP RxQ info; * 12-13: generic helpers for adding a frag to &xdp_buff and converting it to an skb; * 14: get rid of xdp_frame::mem.id, allow mixing pages from different page_pools within one &xdp_buff/&xdp_frame; * 15: some Page Pool helper; * 18: it's for libie, but I wanted to talk about XDP_TX bulking; * 26: same as 13, but for converting XSK &xdp_buff to skb. The rest is up to you, driver-specific stuff is pretty boring sometimes. I'll be polishing and finishing this all starting January 3rd and then preparing and sending sane series, some early feedback never hurts tho. Merry Yule! [0] https://lore.kernel.org/netdev/20231213112835.2262651-1-aleksander.lobakin@intel.com [1] https://lore.kernel.org/netdev/20231215193721.425087-1-michal.kubiak@intel.com Alexander Lobakin (23): idpf: reuse libie's definitions of parsed ptype structures idpf: pack &idpf_queue way more efficiently idpf: remove legacy Page Pool Ethtool stats libie: support different types of buffers for Rx idpf: convert header split mode to libie + napi_build_skb() idpf: use libie Rx buffer management for payload buffer libie: add Tx buffer completion helpers idpf: convert to libie Tx buffer completion bpf, xdp: constify some bpf_prog * function arguments xdp: constify read-only arguments of some static inline helpers xdp: allow attaching already registered memory model to xdp_rxq_info xdp: add generic xdp_buff_add_frag() xdp: add generic xdp_build_skb_from_buff() xdp: get rid of xdp_frame::mem.id page_pool: add inline helper to sync VA for device (for XDP_TX) jump_label: export static_key_slow_{inc,dec}_cpuslocked() libie: support native XDP and register memory model libie: add a couple of XDP helpers idpf: stop using macros for accessing queue descriptors idpf: use generic functions to build xdp_buff and skb idpf: add support for XDP on Rx idpf: add support for .ndo_xdp_xmit() xdp: add generic XSk xdp_buff -> skb conversion Michal Kubiak (11): idpf: make complq cleaning dependent on scheduling mode idpf: prepare structures to support xdp idpf: implement XDP_SETUP_PROG in ndo_bpf for splitq idpf: add support for sw interrupt idpf: add relative queue id member to idpf_queue idpf: add vc functions to manage selected queues idpf: move search rx and tx queues to header idpf: add XSk pool initialization idpf: implement Tx path for AF_XDP idpf: implement Rx path for AF_XDP idpf: enable XSk features and ndo_xsk_wakeup .../net/ethernet/freescale/dpaa/dpaa_eth.c | 2 +- drivers/net/ethernet/intel/Kconfig | 3 +- drivers/net/ethernet/intel/idpf/Makefile | 3 + drivers/net/ethernet/intel/idpf/idpf.h | 91 +- drivers/net/ethernet/intel/idpf/idpf_dev.c | 3 + .../net/ethernet/intel/idpf/idpf_ethtool.c | 74 +- .../net/ethernet/intel/idpf/idpf_lan_txrx.h | 6 +- drivers/net/ethernet/intel/idpf/idpf_lib.c | 40 +- drivers/net/ethernet/intel/idpf/idpf_main.c | 1 + .../ethernet/intel/idpf/idpf_singleq_txrx.c | 221 ++- drivers/net/ethernet/intel/idpf/idpf_txrx.c | 1142 ++++++++-------- drivers/net/ethernet/intel/idpf/idpf_txrx.h | 451 +++---- drivers/net/ethernet/intel/idpf/idpf_vf_dev.c | 3 + .../net/ethernet/intel/idpf/idpf_virtchnl.c | 1132 ++++++++++------ drivers/net/ethernet/intel/idpf/idpf_xdp.c | 522 ++++++++ drivers/net/ethernet/intel/idpf/idpf_xdp.h | 38 + drivers/net/ethernet/intel/idpf/idpf_xsk.c | 1181 +++++++++++++++++ drivers/net/ethernet/intel/idpf/idpf_xsk.h | 30 + drivers/net/ethernet/intel/libie/Makefile | 3 + drivers/net/ethernet/intel/libie/rx.c | 135 +- drivers/net/ethernet/intel/libie/tx.c | 16 + drivers/net/ethernet/intel/libie/xdp.c | 50 + drivers/net/ethernet/intel/libie/xsk.c | 49 + drivers/net/veth.c | 4 +- include/linux/bpf.h | 12 +- include/linux/filter.h | 9 +- include/linux/net/intel/libie/rx.h | 25 +- include/linux/net/intel/libie/tx.h | 94 ++ include/linux/net/intel/libie/xdp.h | 586 ++++++++ include/linux/net/intel/libie/xsk.h | 172 +++ include/linux/netdevice.h | 6 +- include/linux/skbuff.h | 14 +- include/net/page_pool/helpers.h | 32 + include/net/page_pool/types.h | 6 +- include/net/xdp.h | 109 +- kernel/bpf/cpumap.c | 2 +- kernel/bpf/devmap.c | 8 +- kernel/jump_label.c | 2 + net/bpf/test_run.c | 6 +- net/core/dev.c | 8 +- net/core/filter.c | 27 +- net/core/page_pool.c | 31 +- net/core/xdp.c | 189 ++- 43 files changed, 4971 insertions(+), 1567 deletions(-) create mode 100644 drivers/net/ethernet/intel/idpf/idpf_xdp.c create mode 100644 drivers/net/ethernet/intel/idpf/idpf_xdp.h create mode 100644 drivers/net/ethernet/intel/idpf/idpf_xsk.c create mode 100644 drivers/net/ethernet/intel/idpf/idpf_xsk.h create mode 100644 drivers/net/ethernet/intel/libie/tx.c create mode 100644 drivers/net/ethernet/intel/libie/xdp.c create mode 100644 drivers/net/ethernet/intel/libie/xsk.c create mode 100644 include/linux/net/intel/libie/tx.h create mode 100644 include/linux/net/intel/libie/xdp.h create mode 100644 include/linux/net/intel/libie/xsk.h