[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <38443fee-bb05-49a1-9962-f5309be5dddb@nvidia.com>
Date: Wed, 3 Jul 2024 10:33:24 -0700
From: William Tu <witu@...dia.com>
To: Geetha sowjanya <gakula@...vell.com>, netdev@...r.kernel.org,
linux-kernel@...r.kernel.org
Cc: kuba@...nel.org, davem@...emloft.net, pabeni@...hat.com,
edumazet@...gle.com, sgoutham@...vell.com, sbhatta@...vell.com,
hkelam@...vell.com
Subject: Re: [net-next PATCH v7 02/10] octeontx2-pf: RVU representor driver
On 6/28/24 6:35 AM, Geetha sowjanya wrote:
> External email: Use caution opening links or attachments
>
>
> Adds basic driver for the RVU representor.
>
> Driver on probe does pci specific initialization and
> does hw resources configuration. Introduces RVU_ESWITCH
> kernel config to enable/disable the driver. Representor
> and NIC shares the code but representors netdev support
> subset of NIC functionality. Hence "otx2_rep_dev" API
> helps to skip the features initialization that are not
> supported by the representors.
Hi Geetha,
I'm trying to summarize how all representor device works, here:
https://lore.kernel.org/netdev/39dbf7f6-76e0-4319-97d8-24b54e788435@nvidia.com/
So in the RVU representor case, IIUC,
- it has its own PCI func id?
- it has its own receive queue (unlike ice which shares the rxq with
PF), does representor netdev support multiple rx queues?
- it has its own send queue (always use qidx=0)
> Signed-off-by: Geetha sowjanya <gakula@...vell.com>
> ---
> .../net/ethernet/marvell/octeontx2/Kconfig | 8 +
> .../ethernet/marvell/octeontx2/af/Makefile | 3 +-
> .../net/ethernet/marvell/octeontx2/af/mbox.h | 8 +
> .../net/ethernet/marvell/octeontx2/af/rvu.h | 11 +
> .../ethernet/marvell/octeontx2/af/rvu_nix.c | 21 +-
> .../ethernet/marvell/octeontx2/af/rvu_rep.c | 47 ++++
> .../ethernet/marvell/octeontx2/nic/Makefile | 2 +
> .../marvell/octeontx2/nic/otx2_common.h | 12 +-
> .../ethernet/marvell/octeontx2/nic/otx2_pf.c | 17 +-
> .../marvell/octeontx2/nic/otx2_txrx.c | 23 +-
> .../net/ethernet/marvell/octeontx2/nic/rep.c | 223 ++++++++++++++++++
> .../net/ethernet/marvell/octeontx2/nic/rep.h | 31 +++
> 12 files changed, 388 insertions(+), 18 deletions(-)
> create mode 100644 drivers/net/ethernet/marvell/octeontx2/af/rvu_rep.c
> create mode 100644 drivers/net/ethernet/marvell/octeontx2/nic/rep.c
> create mode 100644 drivers/net/ethernet/marvell/octeontx2/nic/rep.h
>
> diff --git a/drivers/net/ethernet/marvell/octeontx2/Kconfig b/drivers/net/ethernet/marvell/octeontx2/Kconfig
> index a32d85d6f599..ff86a5f267c3 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/Kconfig
> +++ b/drivers/net/ethernet/marvell/octeontx2/Kconfig
> @@ -46,3 +46,11 @@ config OCTEONTX2_VF
> depends on OCTEONTX2_PF
> help
> This driver supports Marvell's OcteonTX2 NIC virtual function.
> +
> +config RVU_ESWITCH
> + tristate "Marvell RVU E-Switch support"
> + depends on OCTEONTX2_PF
> + default m
> + help
> + This driver supports Marvell's RVU E-Switch that
> + provides internal SRIOV packet steering and switching for the
> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/Makefile b/drivers/net/ethernet/marvell/octeontx2/af/Makefile
> index 3cf4c8285c90..ccea37847df8 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/Makefile
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/Makefile
> @@ -11,4 +11,5 @@ rvu_mbox-y := mbox.o rvu_trace.o
> rvu_af-y := cgx.o rvu.o rvu_cgx.o rvu_npa.o rvu_nix.o \
> rvu_reg.o rvu_npc.o rvu_debugfs.o ptp.o rvu_npc_fs.o \
> rvu_cpt.o rvu_devlink.o rpm.o rvu_cn10k.o rvu_switch.o \
> - rvu_sdp.o rvu_npc_hash.o mcs.o mcs_rvu_if.o mcs_cnf10kb.o
> + rvu_sdp.o rvu_npc_hash.o mcs.o mcs_rvu_if.o mcs_cnf10kb.o \
> + rvu_rep.o
> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/mbox.h b/drivers/net/ethernet/marvell/octeontx2/af/mbox.h
> index e6d7d6e862c0..befb327e8aff 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/mbox.h
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/mbox.h
> @@ -144,6 +144,7 @@ M(LMTST_TBL_SETUP, 0x00a, lmtst_tbl_setup, lmtst_tbl_setup_req, \
> msg_rsp) \
> M(SET_VF_PERM, 0x00b, set_vf_perm, set_vf_perm, msg_rsp) \
> M(PTP_GET_CAP, 0x00c, ptp_get_cap, msg_req, ptp_get_cap_rsp) \
> +M(GET_REP_CNT, 0x00d, get_rep_cnt, msg_req, get_rep_cnt_rsp) \
> /* CGX mbox IDs (range 0x200 - 0x3FF) */ \
> M(CGX_START_RXTX, 0x200, cgx_start_rxtx, msg_req, msg_rsp) \
> M(CGX_STOP_RXTX, 0x201, cgx_stop_rxtx, msg_req, msg_rsp) \
> @@ -1525,6 +1526,13 @@ struct ptp_get_cap_rsp {
> u64 cap;
> };
>
> +struct get_rep_cnt_rsp {
> + struct mbox_msghdr hdr;
> + u16 rep_cnt;
> + u16 rep_pf_map[64];
> + u64 rsvd;
> +};
> +
> struct flow_msg {
> unsigned char dmac[6];
> unsigned char smac[6];
> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> index 30efa5607c58..cbdc7aeaccfc 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu.h
> @@ -594,6 +594,9 @@ struct rvu {
> spinlock_t cpt_intr_lock;
>
> struct mutex mbox_lock; /* Serialize mbox up and down msgs */
> + u16 rep_pcifunc;
does rep_pcifunc mean the VF/SF's pcifunc, or rep itself has its own
pcifunc?
> + int rep_cnt;
does rep_cnt mean number of rx/tx queues in representor?
> + u16 *rep2pfvf_map;
> };
>
> static inline void rvu_write64(struct rvu *rvu, u64 block, u64 offset, u64 val)
> @@ -822,6 +825,14 @@ bool is_sdp_pfvf(u16 pcifunc);
> bool is_sdp_pf(u16 pcifunc);
> bool is_sdp_vf(struct rvu *rvu, u16 pcifunc);
>
> +static inline bool is_rep_dev(struct rvu *rvu, u16 pcifunc)
> +{
> + if (rvu->rep_pcifunc && rvu->rep_pcifunc == pcifunc)
> + return true;
> +
> + return false;
> +}
> +
> /* CGX APIs */
> static inline bool is_pf_cgxmapped(struct rvu *rvu, u8 pf)
> {
> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
> index 785ef71a5ead..02d83c4958d9 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_nix.c
> @@ -329,7 +329,9 @@ static bool is_valid_txschq(struct rvu *rvu, int blkaddr,
>
> /* TLs aggegating traffic are shared across PF and VFs */
> if (lvl >= hw->cap.nix_tx_aggr_lvl) {
> - if (rvu_get_pf(map_func) != rvu_get_pf(pcifunc))
> + if ((nix_get_tx_link(rvu, map_func) !=
> + nix_get_tx_link(rvu, pcifunc)) &&
> + (rvu_get_pf(map_func) != rvu_get_pf(pcifunc)))
> return false;
> else
> return true;
> @@ -1634,6 +1636,12 @@ int rvu_mbox_handler_nix_lf_alloc(struct rvu *rvu,
> cfg = NPC_TX_DEF_PKIND;
> rvu_write64(rvu, blkaddr, NIX_AF_LFX_TX_PARSE_CFG(nixlf), cfg);
>
> + if (is_rep_dev(rvu, pcifunc)) {
> + pfvf->tx_chan_base = RVU_SWITCH_LBK_CHAN;
> + pfvf->tx_chan_cnt = 1;
> + goto exit;
> + }
> +
> intf = is_lbk_vf(rvu, pcifunc) ? NIX_INTF_TYPE_LBK : NIX_INTF_TYPE_CGX;
> if (is_sdp_pfvf(pcifunc))
> intf = NIX_INTF_TYPE_SDP;
> @@ -1704,6 +1712,9 @@ int rvu_mbox_handler_nix_lf_free(struct rvu *rvu, struct nix_lf_free_req *req,
> if (nixlf < 0)
> return NIX_AF_ERR_AF_LF_INVALID;
>
> + if (is_rep_dev(rvu, pcifunc))
> + goto free_lf;
> +
> if (req->flags & NIX_LF_DISABLE_FLOWS)
> rvu_npc_disable_mcam_entries(rvu, pcifunc, nixlf);
> else
> @@ -1715,6 +1726,7 @@ int rvu_mbox_handler_nix_lf_free(struct rvu *rvu, struct nix_lf_free_req *req,
>
> nix_interface_deinit(rvu, pcifunc, nixlf);
>
> +free_lf:
> /* Reset this NIX LF */
> err = rvu_lf_reset(rvu, block, nixlf);
> if (err) {
> @@ -2010,7 +2022,8 @@ static void nix_get_txschq_range(struct rvu *rvu, u16 pcifunc,
> struct rvu_hwinfo *hw = rvu->hw;
> int pf = rvu_get_pf(pcifunc);
>
> - if (is_lbk_vf(rvu, pcifunc)) { /* LBK links */
> + /* LBK links */
> + if (is_lbk_vf(rvu, pcifunc) || is_rep_dev(rvu, pcifunc)) {
> *start = hw->cap.nix_txsch_per_cgx_lmac * link;
> *end = *start + hw->cap.nix_txsch_per_lbk_lmac;
> } else if (is_pf_cgxmapped(rvu, pf)) { /* CGX links */
> @@ -4523,7 +4536,7 @@ int rvu_mbox_handler_nix_set_hw_frs(struct rvu *rvu, struct nix_frs_cfg *req,
> if (!nix_hw)
> return NIX_AF_ERR_INVALID_NIXBLK;
>
> - if (is_lbk_vf(rvu, pcifunc))
> + if (is_lbk_vf(rvu, pcifunc) || is_rep_dev(rvu, pcifunc))
> rvu_get_lbk_link_max_frs(rvu, &max_mtu);
> else
> rvu_get_lmac_link_max_frs(rvu, &max_mtu);
> @@ -4551,6 +4564,8 @@ int rvu_mbox_handler_nix_set_hw_frs(struct rvu *rvu, struct nix_frs_cfg *req,
> /* For VFs of PF0 ingress is LBK port, so config LBK link */
> pfvf = rvu_get_pfvf(rvu, pcifunc);
> link = hw->cgx_links + pfvf->lbkid;
> + } else if (is_rep_dev(rvu, pcifunc)) {
> + link = hw->cgx_links + 0;
> }
>
> if (link < 0)
> diff --git a/drivers/net/ethernet/marvell/octeontx2/af/rvu_rep.c b/drivers/net/ethernet/marvell/octeontx2/af/rvu_rep.c
> new file mode 100644
> index 000000000000..cf13c5f0a3c5
> --- /dev/null
> +++ b/drivers/net/ethernet/marvell/octeontx2/af/rvu_rep.c
> @@ -0,0 +1,47 @@
> +// SPDX-License-Identifier: GPL-2.0
> +/* Marvell RVU Admin Function driver
> + *
> + * Copyright (C) 2024 Marvell.
> + *
> + */
> +
> +#include <linux/types.h>
> +#include <linux/device.h>
> +#include <linux/module.h>
> +#include <linux/pci.h>
> +
> +#include "rvu.h"
> +#include "rvu_reg.h"
> +
> +int rvu_mbox_handler_get_rep_cnt(struct rvu *rvu, struct msg_req *req,
> + struct get_rep_cnt_rsp *rsp)
> +{
> + int pf, vf, numvfs, hwvf, rep = 0;
> + u16 pcifunc;
> +
> + rvu->rep_pcifunc = req->hdr.pcifunc;
> + rsp->rep_cnt = rvu->cgx_mapped_pfs + rvu->cgx_mapped_vfs;
> + rvu->rep_cnt = rsp->rep_cnt;
> +
> + rvu->rep2pfvf_map = devm_kzalloc(rvu->dev, rvu->rep_cnt *
> + sizeof(u16), GFP_KERNEL);
> + if (!rvu->rep2pfvf_map)
> + return -ENOMEM;
> +
> + for (pf = 0; pf < rvu->hw->total_pfs; pf++) {
> + if (!is_pf_cgxmapped(rvu, pf))
> + continue;
> + pcifunc = pf << RVU_PFVF_PF_SHIFT;
> + rvu->rep2pfvf_map[rep] = pcifunc;
> + rsp->rep_pf_map[rep] = pcifunc;
> + rep++;
> + rvu_get_pf_numvfs(rvu, pf, &numvfs, &hwvf);
> + for (vf = 0; vf < numvfs; vf++) {
> + rvu->rep2pfvf_map[rep] = pcifunc |
> + ((vf + 1) & RVU_PFVF_FUNC_MASK);
> + rsp->rep_pf_map[rep] = rvu->rep2pfvf_map[rep];
> + rep++;
> + }
> + }
> + return 0;
> +}
> diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/Makefile b/drivers/net/ethernet/marvell/octeontx2/nic/Makefile
> index 64a97a0a10ed..dbc971266865 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/nic/Makefile
> +++ b/drivers/net/ethernet/marvell/octeontx2/nic/Makefile
> @@ -5,11 +5,13 @@
>
> obj-$(CONFIG_OCTEONTX2_PF) += rvu_nicpf.o otx2_ptp.o
> obj-$(CONFIG_OCTEONTX2_VF) += rvu_nicvf.o otx2_ptp.o
> +obj-$(CONFIG_RVU_ESWITCH) += rvu_rep.o
>
> rvu_nicpf-y := otx2_pf.o otx2_common.o otx2_txrx.o otx2_ethtool.o \
> otx2_flows.o otx2_tc.o cn10k.o otx2_dmac_flt.o \
> otx2_devlink.o qos_sq.o qos.o
> rvu_nicvf-y := otx2_vf.o
> +rvu_rep-y := rep.o
>
> rvu_nicpf-$(CONFIG_DCB) += otx2_dcbnl.o
> rvu_nicpf-$(CONFIG_MACSEC) += cn10k_macsec.o
> diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h
> index 772fe01bdf98..d297138c356e 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h
> +++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_common.h
> @@ -29,6 +29,7 @@
> #include "otx2_devlink.h"
> #include <rvu_trace.h>
> #include "qos.h"
> +#include "rep.h"
>
> /* IPv4 flag more fragment bit */
> #define IPV4_FLAG_MORE 0x20
> @@ -439,6 +440,7 @@ struct otx2_nic {
> #define OTX2_FLAG_PTP_ONESTEP_SYNC BIT_ULL(15)
> #define OTX2_FLAG_ADPTV_INT_COAL_ENABLED BIT_ULL(16)
> #define OTX2_FLAG_TC_MARK_ENABLED BIT_ULL(17)
> +#define OTX2_FLAG_REP_MODE_ENABLED BIT_ULL(18)
> u64 flags;
> u64 *cq_op_addr;
>
> @@ -506,11 +508,19 @@ struct otx2_nic {
> #if IS_ENABLED(CONFIG_MACSEC)
> struct cn10k_mcs_cfg *macsec_cfg;
> #endif
> +
> +#if IS_ENABLED(CONFIG_RVU_ESWITCH)
> + struct rep_dev **reps;
> + int rep_cnt;
> + u16 rep_pf_map[RVU_MAX_REP];
> + u16 esw_mode;
> +#endif
> };
>
> static inline bool is_otx2_lbkvf(struct pci_dev *pdev)
> {
> - return pdev->device == PCI_DEVID_OCTEONTX2_RVU_AFVF;
> + return (pdev->device == PCI_DEVID_OCTEONTX2_RVU_AFVF) ||
> + (pdev->device == PCI_DEVID_RVU_REP);
> }
>
> static inline bool is_96xx_A0(struct pci_dev *pdev)
> diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c
> index 2b2afcc4b921..8078d1c1fff9 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c
> +++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_pf.c
> @@ -1502,10 +1502,11 @@ int otx2_init_hw_resources(struct otx2_nic *pf)
> hw->sqpool_cnt = otx2_get_total_tx_queues(pf);
> hw->pool_cnt = hw->rqpool_cnt + hw->sqpool_cnt;
>
> - /* Maximum hardware supported transmit length */
> - pf->tx_max_pktlen = pf->netdev->max_mtu + OTX2_ETH_HLEN;
> -
> - pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
> + if (!otx2_rep_dev(pf->pdev)) {
> + /* Maximum hardware supported transmit length */
> + pf->tx_max_pktlen = pf->netdev->max_mtu + OTX2_ETH_HLEN;
> + pf->rbsize = otx2_get_rbuf_size(pf, pf->netdev->mtu);
> + }
>
> mutex_lock(&mbox->lock);
> /* NPA init */
> @@ -1634,11 +1635,12 @@ void otx2_free_hw_resources(struct otx2_nic *pf)
> otx2_pfc_txschq_stop(pf);
> #endif
>
> - otx2_clean_qos_queues(pf);
> + if (!otx2_rep_dev(pf->pdev))
> + otx2_clean_qos_queues(pf);
>
> mutex_lock(&mbox->lock);
> /* Disable backpressure */
> - if (!(pf->pcifunc & RVU_PFVF_FUNC_MASK))
> + if (!is_otx2_lbkvf(pf->pdev))
> otx2_nix_config_bp(pf, false);
> mutex_unlock(&mbox->lock);
>
> @@ -1670,7 +1672,8 @@ void otx2_free_hw_resources(struct otx2_nic *pf)
> otx2_free_cq_res(pf);
>
> /* Free all ingress bandwidth profiles allocated */
> - cn10k_free_all_ipolicers(pf);
> + if (!otx2_rep_dev(pf->pdev))
> + cn10k_free_all_ipolicers(pf);
>
> mutex_lock(&mbox->lock);
> /* Reset NIX LF */
> diff --git a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c
> index fbd9fe98259f..5dcdd8b65837 100644
> --- a/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c
> +++ b/drivers/net/ethernet/marvell/octeontx2/nic/otx2_txrx.c
> @@ -375,11 +375,13 @@ static void otx2_rcv_pkt_handler(struct otx2_nic *pfvf,
> }
> start += sizeof(*sg);
> }
> - otx2_set_rxhash(pfvf, cqe, skb);
>
> - skb_record_rx_queue(skb, cq->cq_idx);
> - if (pfvf->netdev->features & NETIF_F_RXCSUM)
> - skb->ip_summed = CHECKSUM_UNNECESSARY;
> + if (!(pfvf->flags & OTX2_FLAG_REP_MODE_ENABLED)) {
> + otx2_set_rxhash(pfvf, cqe, skb);
> + skb_record_rx_queue(skb, cq->cq_idx);
> + if (pfvf->netdev->features & NETIF_F_RXCSUM)
> + skb->ip_summed = CHECKSUM_UNNECESSARY;
> + }
Does this mean representor have only one rxq, and it doesn't support
checksum offload?
Thanks
William
Powered by blists - more mailing lists