From 6b95c3e9697254dab0c8eafc6ab9d5e10d2eca4e Mon Sep 17 00:00:00 2001 From: Michael Chan Date: Mon, 3 Sep 2018 04:23:17 -0400 Subject: bnxt_en: Fix firmware signaled resource change logic in open. When the driver detects that resources have changed during open, it should reset the rx and tx rings to 0. This will properly setup the init sequence to initialize the default rings again. We also need to signal the RDMA driver to stop and clear its interrupts. We then call the RoCE driver to restart if a new set of default rings is successfully reserved. Fixes: 25e1acd6b92b ("bnxt_en: Notify firmware about IF state changes.") Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 8bb1e38b1681..6a1baf375ac3 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -6684,6 +6684,8 @@ static int bnxt_hwrm_if_change(struct bnxt *bp, bool up) hw_resc->resv_rx_rings = 0; hw_resc->resv_hw_ring_grps = 0; hw_resc->resv_vnics = 0; + bp->tx_nr_rings = 0; + bp->rx_nr_rings = 0; } return rc; } @@ -8769,20 +8771,25 @@ static int bnxt_init_dflt_ring_mode(struct bnxt *bp) if (bp->tx_nr_rings) return 0; + bnxt_ulp_irq_stop(bp); + bnxt_clear_int_mode(bp); rc = bnxt_set_dflt_rings(bp, true); if (rc) { netdev_err(bp->dev, "Not enough rings available.\n"); - return rc; + goto init_dflt_ring_err; } rc = bnxt_init_int_mode(bp); if (rc) - return rc; + goto init_dflt_ring_err; + bp->tx_nr_rings_per_tc = bp->tx_nr_rings; if (bnxt_rfs_supported(bp) && bnxt_rfs_capable(bp)) { bp->flags |= BNXT_FLAG_RFS; bp->dev->features |= NETIF_F_NTUPLE; } - return 0; +init_dflt_ring_err: + bnxt_ulp_irq_restart(bp, rc); + return rc; } int bnxt_restore_pf_fw_resources(struct bnxt *bp) -- cgit v1.2.1 From ad95c27bdb930105f3eea02621bda157caf2862d Mon Sep 17 00:00:00 2001 From: Michael Chan Date: Mon, 3 Sep 2018 04:23:18 -0400 Subject: bnxt_en: Clean up unused functions. Remove unused bnxt_subtract_ulp_resources(). Change bnxt_get_max_func_irqs() to static since it is only locally used. Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 6a1baf375ac3..6472ce447f87 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -5918,7 +5918,7 @@ void bnxt_set_max_func_cp_rings(struct bnxt *bp, unsigned int max) bp->hw_resc.max_cp_rings = max; } -unsigned int bnxt_get_max_func_irqs(struct bnxt *bp) +static unsigned int bnxt_get_max_func_irqs(struct bnxt *bp) { struct bnxt_hw_resc *hw_resc = &bp->hw_resc; -- cgit v1.2.1 From 00fe9c326d2027f2437dea38ef0e82f9d02d94c0 Mon Sep 17 00:00:00 2001 From: Michael Chan Date: Mon, 3 Sep 2018 04:23:19 -0400 Subject: bnxt_en: Do not adjust max_cp_rings by the ones used by RDMA. Currently, the driver adjusts the bp->hw_resc.max_cp_rings by the number of MSIX vectors used by RDMA. There is one code path in open that needs to check the true max_cp_rings including any used by RDMA. This code is now checking for the reduced max_cp_rings which will fail when the number of cp rings is very small. To fix this in a clean way, we don't adjust max_cp_rings anymore. Instead, we add a helper bnxt_get_max_func_cp_rings_for_en() to get the reduced max_cp_rings when appropriate. Fixes: ec86f14ea506 ("bnxt_en: Add ULP calls to stop and restart IRQs.") Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 6472ce447f87..cecbb1d1f587 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -5913,9 +5913,9 @@ unsigned int bnxt_get_max_func_cp_rings(struct bnxt *bp) return bp->hw_resc.max_cp_rings; } -void bnxt_set_max_func_cp_rings(struct bnxt *bp, unsigned int max) +unsigned int bnxt_get_max_func_cp_rings_for_en(struct bnxt *bp) { - bp->hw_resc.max_cp_rings = max; + return bp->hw_resc.max_cp_rings - bnxt_get_ulp_msix_num(bp); } static unsigned int bnxt_get_max_func_irqs(struct bnxt *bp) @@ -8631,7 +8631,8 @@ static void _bnxt_get_max_rings(struct bnxt *bp, int *max_rx, int *max_tx, *max_tx = hw_resc->max_tx_rings; *max_rx = hw_resc->max_rx_rings; - *max_cp = min_t(int, hw_resc->max_irqs, hw_resc->max_cp_rings); + *max_cp = min_t(int, bnxt_get_max_func_cp_rings_for_en(bp), + hw_resc->max_irqs); *max_cp = min_t(int, *max_cp, hw_resc->max_stat_ctxs); max_ring_grps = hw_resc->max_hw_ring_grps; if (BNXT_CHIP_TYPE_NITRO_A0(bp) && BNXT_PF(bp)) { -- cgit v1.2.1 From 28ea334bd1657f3c43485b4a8592672fc6835fac Mon Sep 17 00:00:00 2001 From: Michael Chan Date: Fri, 14 Sep 2018 15:41:29 -0400 Subject: bnxt_en: Fix VF mac address regression. The recent commit to always forward the VF MAC address to the PF for approval may not work if the PF driver or the firmware is older. This will cause the VF driver to fail during probe: bnxt_en 0000:00:03.0 (unnamed net_device) (uninitialized): hwrm req_type 0xf seq id 0x5 error 0xffff bnxt_en 0000:00:03.0 (unnamed net_device) (uninitialized): VF MAC address 00:00:17:02:05:d0 not approved by the PF bnxt_en 0000:00:03.0: Unable to initialize mac address. bnxt_en: probe of 0000:00:03.0 failed with error -99 We fix it by treating the error as fatal only if the VF MAC address is locally generated by the VF. Fixes: 707e7e966026 ("bnxt_en: Always forward VF MAC address to the PF.") Reported-by: Seth Forshee Reported-by: Siwei Liu Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 9 +++++++-- 1 file changed, 7 insertions(+), 2 deletions(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index cecbb1d1f587..177587f9c3f1 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -8027,7 +8027,7 @@ static int bnxt_change_mac_addr(struct net_device *dev, void *p) if (ether_addr_equal(addr->sa_data, dev->dev_addr)) return 0; - rc = bnxt_approve_mac(bp, addr->sa_data); + rc = bnxt_approve_mac(bp, addr->sa_data, true); if (rc) return rc; @@ -8827,14 +8827,19 @@ static int bnxt_init_mac_addr(struct bnxt *bp) } else { #ifdef CONFIG_BNXT_SRIOV struct bnxt_vf_info *vf = &bp->vf; + bool strict_approval = true; if (is_valid_ether_addr(vf->mac_addr)) { /* overwrite netdev dev_addr with admin VF MAC */ memcpy(bp->dev->dev_addr, vf->mac_addr, ETH_ALEN); + /* Older PF driver or firmware may not approve this + * correctly. + */ + strict_approval = false; } else { eth_hw_addr_random(bp->dev); } - rc = bnxt_approve_mac(bp, bp->dev->dev_addr); + rc = bnxt_approve_mac(bp, bp->dev->dev_addr, strict_approval); #endif } return rc; -- cgit v1.2.1 From 58e0e22bff638055278ea73e34d0d07a95260790 Mon Sep 17 00:00:00 2001 From: Eric Dumazet Date: Fri, 21 Sep 2018 15:27:50 -0700 Subject: bnxt: remove ndo_poll_controller As diagnosed by Song Liu, ndo_poll_controller() can be very dangerous on loaded hosts, since the cpu calling ndo_poll_controller() might steal all NAPI contexts (for all RX/TX queues of the NIC). This capture can last for unlimited amount of time, since one cpu is generally not able to drain all the queues under load. bnxt uses NAPI for TX completions, so we better let core networking stack call the napi->poll() to avoid the capture. Signed-off-by: Eric Dumazet Cc: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 18 ------------------ 1 file changed, 18 deletions(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 177587f9c3f1..61957b0bbd8c 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -7672,21 +7672,6 @@ static void bnxt_tx_timeout(struct net_device *dev) bnxt_queue_sp_work(bp); } -#ifdef CONFIG_NET_POLL_CONTROLLER -static void bnxt_poll_controller(struct net_device *dev) -{ - struct bnxt *bp = netdev_priv(dev); - int i; - - /* Only process tx rings/combined rings in netpoll mode. */ - for (i = 0; i < bp->tx_nr_rings; i++) { - struct bnxt_tx_ring_info *txr = &bp->tx_ring[i]; - - napi_schedule(&txr->bnapi->napi); - } -} -#endif - static void bnxt_timer(struct timer_list *t) { struct bnxt *bp = from_timer(bp, t, timer); @@ -8519,9 +8504,6 @@ static const struct net_device_ops bnxt_netdev_ops = { .ndo_set_vf_link_state = bnxt_set_vf_link_state, .ndo_set_vf_spoofchk = bnxt_set_vf_spoofchk, .ndo_set_vf_trust = bnxt_set_vf_trust, -#endif -#ifdef CONFIG_NET_POLL_CONTROLLER - .ndo_poll_controller = bnxt_poll_controller, #endif .ndo_setup_tc = bnxt_setup_tc, #ifdef CONFIG_RFS_ACCEL -- cgit v1.2.1 From 73f21c653f930f438d53eed29b5e4c65c8a0f906 Mon Sep 17 00:00:00 2001 From: Michael Chan Date: Wed, 26 Sep 2018 00:41:04 -0400 Subject: bnxt_en: Fix TX timeout during netpoll. The current netpoll implementation in the bnxt_en driver has problems that may miss TX completion events. bnxt_poll_work() in effect is only handling at most 1 TX packet before exiting. In addition, there may be in flight TX completions that ->poll() may miss even after we fix bnxt_poll_work() to handle all visible TX completions. netpoll may not call ->poll() again and HW may not generate IRQ because the driver does not ARM the IRQ when the budget (0 for netpoll) is reached. We fix it by handling all TX completions and to always ARM the IRQ when we exit ->poll() with 0 budget. Also, the logic to ACK the completion ring in case it is almost filled with TX completions need to be adjusted to take care of the 0 budget case, as discussed with Eric Dumazet Reported-by: Song Liu Reviewed-by: Song Liu Tested-by: Song Liu Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 13 ++++++++++--- 1 file changed, 10 insertions(+), 3 deletions(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 61957b0bbd8c..0478e562abac 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -1884,8 +1884,11 @@ static int bnxt_poll_work(struct bnxt *bp, struct bnxt_napi *bnapi, int budget) if (TX_CMP_TYPE(txcmp) == CMP_TYPE_TX_L2_CMP) { tx_pkts++; /* return full budget so NAPI will complete. */ - if (unlikely(tx_pkts > bp->tx_wake_thresh)) + if (unlikely(tx_pkts > bp->tx_wake_thresh)) { rx_pkts = budget; + raw_cons = NEXT_RAW_CMP(raw_cons); + break; + } } else if ((TX_CMP_TYPE(txcmp) & 0x30) == 0x10) { if (likely(budget)) rc = bnxt_rx_pkt(bp, bnapi, &raw_cons, &event); @@ -1913,7 +1916,7 @@ static int bnxt_poll_work(struct bnxt *bp, struct bnxt_napi *bnapi, int budget) } raw_cons = NEXT_RAW_CMP(raw_cons); - if (rx_pkts == budget) + if (rx_pkts && rx_pkts == budget) break; } @@ -2027,8 +2030,12 @@ static int bnxt_poll(struct napi_struct *napi, int budget) while (1) { work_done += bnxt_poll_work(bp, bnapi, budget - work_done); - if (work_done >= budget) + if (work_done >= budget) { + if (!budget) + BNXT_CP_DB_REARM(cpr->cp_doorbell, + cpr->cp_raw_cons); break; + } if (!bnxt_has_work(bp, cpr)) { if (napi_complete_done(napi, work_done)) -- cgit v1.2.1 From dbe80d446c859873820eedfff4abc61c71f1927b Mon Sep 17 00:00:00 2001 From: Michael Chan Date: Fri, 5 Oct 2018 00:26:00 -0400 Subject: bnxt_en: Fix VNIC reservations on the PF. The enables bit for VNIC was set wrong when calling the HWRM_FUNC_CFG firmware call to reserve VNICs. This has the effect that the firmware will keep a large number of VNICs for the PF, and having very few for VFs. DPDK driver running on the VFs, which requires more VNICs, may not work properly as a result. Fixes: 674f50a5b026 ("bnxt_en: Implement new method to reserve rings.") Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 0478e562abac..2564a92dcb02 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -4650,7 +4650,7 @@ __bnxt_hwrm_reserve_pf_rings(struct bnxt *bp, struct hwrm_func_cfg_input *req, FUNC_CFG_REQ_ENABLES_NUM_STAT_CTXS : 0; enables |= ring_grps ? FUNC_CFG_REQ_ENABLES_NUM_HW_RING_GRPS : 0; - enables |= vnics ? FUNC_VF_CFG_REQ_ENABLES_NUM_VNICS : 0; + enables |= vnics ? FUNC_CFG_REQ_ENABLES_NUM_VNICS : 0; req->num_rx_rings = cpu_to_le16(rx_rings); req->num_hw_ring_grps = cpu_to_le16(ring_grps); -- cgit v1.2.1 From a2bf74f4e1b82395dad2b08d2a911d9151db71c1 Mon Sep 17 00:00:00 2001 From: Venkat Duvvuru Date: Fri, 5 Oct 2018 00:26:02 -0400 Subject: bnxt_en: free hwrm resources, if driver probe fails. When the driver probe fails, all the resources that were allocated prior to the failure must be freed. However, hwrm dma response memory is not getting freed. This patch fixes the problem described above. Fixes: c0c050c58d84 ("bnxt_en: New Broadcom ethernet driver.") Signed-off-by: Venkat Duvvuru Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 10 ++++++---- 1 file changed, 6 insertions(+), 4 deletions(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 2564a92dcb02..3718984a8185 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -3017,10 +3017,11 @@ static void bnxt_free_hwrm_resources(struct bnxt *bp) { struct pci_dev *pdev = bp->pdev; - dma_free_coherent(&pdev->dev, PAGE_SIZE, bp->hwrm_cmd_resp_addr, - bp->hwrm_cmd_resp_dma_addr); - - bp->hwrm_cmd_resp_addr = NULL; + if (bp->hwrm_cmd_resp_addr) { + dma_free_coherent(&pdev->dev, PAGE_SIZE, bp->hwrm_cmd_resp_addr, + bp->hwrm_cmd_resp_dma_addr); + bp->hwrm_cmd_resp_addr = NULL; + } } static int bnxt_alloc_hwrm_resources(struct bnxt *bp) @@ -9057,6 +9058,7 @@ init_err_cleanup_tc: bnxt_clear_int_mode(bp); init_err_pci_clean: + bnxt_free_hwrm_resources(bp); bnxt_cleanup_pci(bp); init_err_free: -- cgit v1.2.1 From c78fe058879bdea919d44f23e21da26f603e9166 Mon Sep 17 00:00:00 2001 From: Vasundhara Volam Date: Fri, 5 Oct 2018 00:26:03 -0400 Subject: bnxt_en: get the reduced max_irqs by the ones used by RDMA When getting the max rings supported, get the reduced max_irqs by the ones used by RDMA. If the number MSIX is the limiting factor, this bug may cause the max ring count to be higher than it should be when RDMA driver is loaded and may result in ring allocation failures. Fixes: 30f529473ec9 ("bnxt_en: Do not modify max IRQ count after RDMA driver requests/frees IRQs.") Signed-off-by: Vasundhara Volam Signed-off-by: Michael Chan Signed-off-by: David S. Miller --- drivers/net/ethernet/broadcom/bnxt/bnxt.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers/net/ethernet/broadcom/bnxt/bnxt.c') diff --git a/drivers/net/ethernet/broadcom/bnxt/bnxt.c b/drivers/net/ethernet/broadcom/bnxt/bnxt.c index 3718984a8185..e2d92548226a 100644 --- a/drivers/net/ethernet/broadcom/bnxt/bnxt.c +++ b/drivers/net/ethernet/broadcom/bnxt/bnxt.c @@ -8622,7 +8622,7 @@ static void _bnxt_get_max_rings(struct bnxt *bp, int *max_rx, int *max_tx, *max_tx = hw_resc->max_tx_rings; *max_rx = hw_resc->max_rx_rings; *max_cp = min_t(int, bnxt_get_max_func_cp_rings_for_en(bp), - hw_resc->max_irqs); + hw_resc->max_irqs - bnxt_get_ulp_msix_num(bp)); *max_cp = min_t(int, *max_cp, hw_resc->max_stat_ctxs); max_ring_grps = hw_resc->max_hw_ring_grps; if (BNXT_CHIP_TYPE_NITRO_A0(bp) && BNXT_PF(bp)) { -- cgit v1.2.1