From patchwork Thu Feb 18 09:04:02 2016 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Kirsher, Jeffrey T" X-Patchwork-Id: 584570 X-Patchwork-Delegate: davem@davemloft.net Return-Path: X-Original-To: patchwork-incoming@ozlabs.org Delivered-To: patchwork-incoming@ozlabs.org Received: from vger.kernel.org (vger.kernel.org [209.132.180.67]) by ozlabs.org (Postfix) with ESMTP id 2C351140271 for ; Thu, 18 Feb 2016 20:05:30 +1100 (AEDT) Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1425762AbcBRJFU (ORCPT ); Thu, 18 Feb 2016 04:05:20 -0500 Received: from mga03.intel.com ([134.134.136.65]:36842 "EHLO mga03.intel.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1425712AbcBRJEZ (ORCPT ); Thu, 18 Feb 2016 04:04:25 -0500 Received: from fmsmga004.fm.intel.com ([10.253.24.48]) by orsmga103.jf.intel.com with ESMTP; 18 Feb 2016 01:04:17 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.22,464,1449561600"; d="scan'208";a="50035342" Received: from jtkirshe-linux.jf.intel.com ([134.134.3.161]) by fmsmga004.fm.intel.com with ESMTP; 18 Feb 2016 01:04:16 -0800 From: Jeff Kirsher To: davem@davemloft.net Cc: Anjali Singhai Jain , netdev@vger.kernel.org, nhorman@redhat.com, sassmann@redhat.com, jogreene@redhat.com, Jeff Kirsher Subject: [net-next 05/15] i40e: Add a SW workaround for lost interrupts Date: Thu, 18 Feb 2016 01:04:02 -0800 Message-Id: <1455786252-71881-6-git-send-email-jeffrey.t.kirsher@intel.com> X-Mailer: git-send-email 2.5.0 In-Reply-To: <1455786252-71881-1-git-send-email-jeffrey.t.kirsher@intel.com> References: <1455786252-71881-1-git-send-email-jeffrey.t.kirsher@intel.com> Sender: netdev-owner@vger.kernel.org Precedence: bulk List-ID: X-Mailing-List: netdev@vger.kernel.org From: Anjali Singhai Jain This patch adds a workaround for cases where we might have interrupts that got lost but WB happened. If that happens without this patch we will see a tx_timeout. To work around it, this patch goes ahead and reschedules NAPI in that situation, if NAPI is not already scheduled. We also add a counter in ethtool to keep track of when we detect a case of tx_lost_interrupt. Note: napi_reschedule() can be safely called from process/service_task context and is done in other drivers as well without an issue. Change-ID: I00f98f1ce3774524d9421227652bef20fcbd0d20 Signed-off-by: Anjali Singhai Jain Tested-by: Andrew Bowers Signed-off-by: Jeff Kirsher --- drivers/net/ethernet/intel/i40e/i40e.h | 1 + drivers/net/ethernet/intel/i40e/i40e_ethtool.c | 1 + drivers/net/ethernet/intel/i40e/i40e_main.c | 32 ++++++++++++++++---------- drivers/net/ethernet/intel/i40e/i40e_txrx.c | 10 +++++--- drivers/net/ethernet/intel/i40e/i40e_txrx.h | 3 ++- drivers/net/ethernet/intel/i40evf/i40e_txrx.c | 10 +++++--- drivers/net/ethernet/intel/i40evf/i40e_txrx.h | 3 ++- 7 files changed, 40 insertions(+), 20 deletions(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e.h b/drivers/net/ethernet/intel/i40e/i40e.h index 5ea431d..5d8e159 100644 --- a/drivers/net/ethernet/intel/i40e/i40e.h +++ b/drivers/net/ethernet/intel/i40e/i40e.h @@ -512,6 +512,7 @@ struct i40e_vsi { u32 tx_busy; u64 tx_linearize; u64 tx_force_wb; + u64 tx_lost_interrupt; u32 rx_buf_failed; u32 rx_page_failed; diff --git a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c index 230fa40..5778f9e 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_ethtool.c +++ b/drivers/net/ethernet/intel/i40e/i40e_ethtool.c @@ -89,6 +89,7 @@ static const struct i40e_stats i40e_gstrings_misc_stats[] = { I40E_VSI_STAT("rx_unknown_protocol", eth_stats.rx_unknown_protocol), I40E_VSI_STAT("tx_linearize", tx_linearize), I40E_VSI_STAT("tx_force_wb", tx_force_wb), + I40E_VSI_STAT("tx_lost_interrupt", tx_lost_interrupt), I40E_VSI_STAT("rx_alloc_fail", rx_buf_failed), I40E_VSI_STAT("rx_pg_alloc_fail", rx_page_failed), }; diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index 161603b..aa1c7ca 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -819,6 +819,7 @@ static void i40e_update_vsi_stats(struct i40e_vsi *vsi) struct i40e_eth_stats *oes; struct i40e_eth_stats *es; /* device's eth stats */ u32 tx_restart, tx_busy; + u64 tx_lost_interrupt; struct i40e_ring *p; u32 rx_page, rx_buf; u64 bytes, packets; @@ -844,6 +845,7 @@ static void i40e_update_vsi_stats(struct i40e_vsi *vsi) rx_b = rx_p = 0; tx_b = tx_p = 0; tx_restart = tx_busy = tx_linearize = tx_force_wb = 0; + tx_lost_interrupt = 0; rx_page = 0; rx_buf = 0; rcu_read_lock(); @@ -862,6 +864,7 @@ static void i40e_update_vsi_stats(struct i40e_vsi *vsi) tx_busy += p->tx_stats.tx_busy; tx_linearize += p->tx_stats.tx_linearize; tx_force_wb += p->tx_stats.tx_force_wb; + tx_lost_interrupt += p->tx_stats.tx_lost_interrupt; /* Rx queue is part of the same block as Tx queue */ p = &p[1]; @@ -880,6 +883,7 @@ static void i40e_update_vsi_stats(struct i40e_vsi *vsi) vsi->tx_busy = tx_busy; vsi->tx_linearize = tx_linearize; vsi->tx_force_wb = tx_force_wb; + vsi->tx_lost_interrupt = tx_lost_interrupt; vsi->rx_page_failed = rx_page; vsi->rx_buf_failed = rx_buf; @@ -4349,7 +4353,7 @@ static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi) { struct i40e_ring *tx_ring = NULL; struct i40e_pf *pf; - u32 head, val, tx_pending; + u32 head, val, tx_pending_hw; int i; pf = vsi->back; @@ -4375,16 +4379,9 @@ static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi) else val = rd32(&pf->hw, I40E_PFINT_DYN_CTL0); - /* Bail out if interrupts are disabled because napi_poll - * execution in-progress or will get scheduled soon. - * napi_poll cleans TX and RX queues and updates 'next_to_clean'. - */ - if (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK)) - return; - head = i40e_get_head(tx_ring); - tx_pending = i40e_get_tx_pending(tx_ring); + tx_pending_hw = i40e_get_tx_pending(tx_ring, false); /* HW is done executing descriptors, updated HEAD write back, * but SW hasn't processed those descriptors. If interrupt is @@ -4392,12 +4389,12 @@ static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi) * dev_watchdog detecting timeout on those netdev_queue, * hence proactively trigger SW interrupt. */ - if (tx_pending) { + if (tx_pending_hw && (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) { /* NAPI Poll didn't run and clear since it was set */ if (test_and_clear_bit(I40E_Q_VECTOR_HUNG_DETECT, &tx_ring->q_vector->hung_detected)) { - netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n", - vsi->seid, q_idx, tx_pending, + netdev_info(vsi->netdev, "VSI_seid %d, Hung TX queue %d, tx_pending_hw: %d, NTC:0x%x, HWB: 0x%x, NTU: 0x%x, TAIL: 0x%x\n", + vsi->seid, q_idx, tx_pending_hw, tx_ring->next_to_clean, head, tx_ring->next_to_use, readl(tx_ring->tail)); @@ -4410,6 +4407,17 @@ static void i40e_detect_recover_hung_queue(int q_idx, struct i40e_vsi *vsi) &tx_ring->q_vector->hung_detected); } } + + /* This is the case where we have interrupts missing, + * so the tx_pending in HW will most likely be 0, but we + * will have tx_pending in SW since the WB happened but the + * interrupt got lost. + */ + if ((!tx_pending_hw) && i40e_get_tx_pending(tx_ring, true) && + (!(val & I40E_PFINT_DYN_CTLN_INTENA_MASK))) { + if (napi_reschedule(&tx_ring->q_vector->napi)) + tx_ring->tx_stats.tx_lost_interrupt++; + } } /** diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index 0ffa9a8..5b43585 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -610,15 +610,19 @@ void i40e_free_tx_resources(struct i40e_ring *tx_ring) /** * i40e_get_tx_pending - how many tx descriptors not processed * @tx_ring: the ring of descriptors + * @in_sw: is tx_pending being checked in SW or HW * * Since there is no access to the ring head register * in XL710, we need to use our local copies **/ -u32 i40e_get_tx_pending(struct i40e_ring *ring) +u32 i40e_get_tx_pending(struct i40e_ring *ring, bool in_sw) { u32 head, tail; - head = i40e_get_head(ring); + if (!in_sw) + head = i40e_get_head(ring); + else + head = ring->next_to_clean; tail = readl(ring->tail); if (head != tail) @@ -741,7 +745,7 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget) * them to be written back in case we stay in NAPI. * In this mode on X722 we do not enable Interrupt. */ - j = i40e_get_tx_pending(tx_ring); + j = i40e_get_tx_pending(tx_ring, false); if (budget && ((j / (WB_STRIDE + 1)) == 0) && (j != 0) && diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.h b/drivers/net/ethernet/intel/i40e/i40e_txrx.h index ae22c4e..3acc924 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.h @@ -203,6 +203,7 @@ struct i40e_tx_queue_stats { u64 tx_done_old; u64 tx_linearize; u64 tx_force_wb; + u64 tx_lost_interrupt; }; struct i40e_rx_queue_stats { @@ -338,7 +339,7 @@ int i40e_tx_prepare_vlan_flags(struct sk_buff *skb, struct i40e_ring *tx_ring, u32 *flags); #endif void i40e_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector); -u32 i40e_get_tx_pending(struct i40e_ring *ring); +u32 i40e_get_tx_pending(struct i40e_ring *ring, bool in_sw); /** * i40e_get_head - Retrieve head from head writeback diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c index 76bad75..0f73a4f 100644 --- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.c @@ -129,15 +129,19 @@ void i40evf_free_tx_resources(struct i40e_ring *tx_ring) /** * i40evf_get_tx_pending - how many Tx descriptors not processed * @tx_ring: the ring of descriptors + * @in_sw: is tx_pending being checked in SW or HW * * Since there is no access to the ring head register * in XL710, we need to use our local copies **/ -u32 i40evf_get_tx_pending(struct i40e_ring *ring) +u32 i40evf_get_tx_pending(struct i40e_ring *ring, bool in_sw) { u32 head, tail; - head = i40e_get_head(ring); + if (!in_sw) + head = i40e_get_head(ring); + else + head = ring->next_to_clean; tail = readl(ring->tail); if (head != tail) @@ -259,7 +263,7 @@ static bool i40e_clean_tx_irq(struct i40e_ring *tx_ring, int budget) * them to be written back in case we stay in NAPI. * In this mode on X722 we do not enable Interrupt. */ - j = i40evf_get_tx_pending(tx_ring); + j = i40evf_get_tx_pending(tx_ring, false); if (budget && ((j / (WB_STRIDE + 1)) == 0) && (j > 0) && diff --git a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h index 5467fcdf..81c9661 100644 --- a/drivers/net/ethernet/intel/i40evf/i40e_txrx.h +++ b/drivers/net/ethernet/intel/i40evf/i40e_txrx.h @@ -202,6 +202,7 @@ struct i40e_tx_queue_stats { u64 tx_done_old; u64 tx_linearize; u64 tx_force_wb; + u64 tx_lost_interrupt; }; struct i40e_rx_queue_stats { @@ -326,7 +327,7 @@ void i40evf_free_tx_resources(struct i40e_ring *tx_ring); void i40evf_free_rx_resources(struct i40e_ring *rx_ring); int i40evf_napi_poll(struct napi_struct *napi, int budget); void i40evf_force_wb(struct i40e_vsi *vsi, struct i40e_q_vector *q_vector); -u32 i40evf_get_tx_pending(struct i40e_ring *ring); +u32 i40evf_get_tx_pending(struct i40e_ring *ring, bool in_sw); /** * i40e_get_head - Retrieve head from head writeback