netdev - [PATCH AUTOSEL 5.7 227/274] ice: Fix Tx timeout when link is toggled on a VF's interface

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [day] [month] [year] [list]

Message-Id: <20200608230607.3361041-227-sashal@kernel.org>
Date:   Mon,  8 Jun 2020 19:05:20 -0400
From:   Sasha Levin <sashal@...nel.org>
To:     linux-kernel@...r.kernel.org, stable@...r.kernel.org
Cc:     Brett Creeley <brett.creeley@...el.com>,
        Andrew Bowers <andrewx.bowers@...el.com>,
        Jeff Kirsher <jeffrey.t.kirsher@...el.com>,
        Sasha Levin <sashal@...nel.org>,
        intel-wired-lan@...ts.osuosl.org, netdev@...r.kernel.org
Subject: [PATCH AUTOSEL 5.7 227/274] ice: Fix Tx timeout when link is toggled on a VF's interface

From: Brett Creeley <brett.creeley@...el.com>

[ Upstream commit 4dc926d3a59e73b8c4adf51b261f1a1bbd48a989 ]

Currently if the iavf is loaded and a VF link transitions from up to
down to up again a Tx timeout will be triggered. This happens because
Tx/Rx queue interrupts are only enabled when receiving the
VIRTCHNL_OP_CONFIG_MAP_IRQ message, which happens on reset or initial
iavf driver load, but not when bringing link up. This is problematic
because they are disabled on the VIRTCHNL_OP_DISABLE_QUEUES message,
which is part of bringing a VF's link down. However, they are not
enabled on the VIRTCHNL_OP_ENABLE_QUEUES message, which is part of
bringing a VF's link up.

Fix this by re-enabling the VF's Rx and Tx queue interrupts when they
were previously configured. This is done by first checking to make
sure the previous value in QINT_[R|T]QCTL.MSIX_INDX is not 0, which
is used to represent the OICR in the VF's interrupt space. If the
MSIX_INDX is non-zero then enable the interrupt by setting the
QINT_[R|T]CTL.CAUSE_ENA bit to 1.

Signed-off-by: Brett Creeley <brett.creeley@...el.com>
Tested-by: Andrew Bowers <andrewx.bowers@...el.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@...el.com>
Signed-off-by: Sasha Levin <sashal@...nel.org>
---
 .../net/ethernet/intel/ice/ice_virtchnl_pf.c  | 48 +++++++++++++++++++
 1 file changed, 48 insertions(+)

diff --git a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
index c9c281167873..f1fdb4d4c826 100644
--- a/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/ice/ice_virtchnl_pf.c
@@ -2118,6 +2118,52 @@ static bool ice_vc_validate_vqs_bitmaps(struct virtchnl_queue_select *vqs)
 	return true;
 }
 
+/**
+ * ice_vf_ena_txq_interrupt - enable Tx queue interrupt via QINT_TQCTL
+ * @vsi: VSI of the VF to configure
+ * @q_idx: VF queue index used to determine the queue in the PF's space
+ */
+static void ice_vf_ena_txq_interrupt(struct ice_vsi *vsi, u32 q_idx)
+{
+	struct ice_hw *hw = &vsi->back->hw;
+	u32 pfq = vsi->txq_map[q_idx];
+	u32 reg;
+
+	reg = rd32(hw, QINT_TQCTL(pfq));
+
+	/* MSI-X index 0 in the VF's space is always for the OICR, which means
+	 * this is most likely a poll mode VF driver, so don't enable an
+	 * interrupt that was never configured via VIRTCHNL_OP_CONFIG_IRQ_MAP
+	 */
+	if (!(reg & QINT_TQCTL_MSIX_INDX_M))
+		return;
+
+	wr32(hw, QINT_TQCTL(pfq), reg | QINT_TQCTL_CAUSE_ENA_M);
+}
+
+/**
+ * ice_vf_ena_rxq_interrupt - enable Tx queue interrupt via QINT_RQCTL
+ * @vsi: VSI of the VF to configure
+ * @q_idx: VF queue index used to determine the queue in the PF's space
+ */
+static void ice_vf_ena_rxq_interrupt(struct ice_vsi *vsi, u32 q_idx)
+{
+	struct ice_hw *hw = &vsi->back->hw;
+	u32 pfq = vsi->rxq_map[q_idx];
+	u32 reg;
+
+	reg = rd32(hw, QINT_RQCTL(pfq));
+
+	/* MSI-X index 0 in the VF's space is always for the OICR, which means
+	 * this is most likely a poll mode VF driver, so don't enable an
+	 * interrupt that was never configured via VIRTCHNL_OP_CONFIG_IRQ_MAP
+	 */
+	if (!(reg & QINT_RQCTL_MSIX_INDX_M))
+		return;
+
+	wr32(hw, QINT_RQCTL(pfq), reg | QINT_RQCTL_CAUSE_ENA_M);
+}
+
 /**
  * ice_vc_ena_qs_msg
  * @vf: pointer to the VF info
@@ -2178,6 +2224,7 @@ static int ice_vc_ena_qs_msg(struct ice_vf *vf, u8 *msg)
 			goto error_param;
 		}
 
+		ice_vf_ena_rxq_interrupt(vsi, vf_q_id);
 		set_bit(vf_q_id, vf->rxq_ena);
 	}
 
@@ -2193,6 +2240,7 @@ static int ice_vc_ena_qs_msg(struct ice_vf *vf, u8 *msg)
 		if (test_bit(vf_q_id, vf->txq_ena))
 			continue;
 
+		ice_vf_ena_txq_interrupt(vsi, vf_q_id);
 		set_bit(vf_q_id, vf->txq_ena);
 	}
 
-- 
2.25.1