From daf182d360e509a494db18666799f4e85d83dda0 Mon Sep 17 00:00:00 2001 From: Shyam Sundar S K Date: Wed, 27 Oct 2021 15:27:27 +0530 Subject: net: amd-xgbe: Toggle PLL settings during rate change For each rate change command submission, the FW has to do a phy power off sequence internally. For this to happen correctly, the PLL re-initialization control setting has to be turned off before sending mailbox commands and re-enabled once the command submission is complete. Without the PLL control setting, the link up takes longer time in a fixed phy configuration. Fixes: 47f164deab22 ("amd-xgbe: Add PCI device support") Co-developed-by: Sudheesh Mavila Signed-off-by: Sudheesh Mavila Signed-off-by: Shyam Sundar S K Acked-by: Tom Lendacky Signed-off-by: David S. Miller --- drivers/net/ethernet/amd/xgbe/xgbe-common.h | 8 ++++++++ drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c | 20 +++++++++++++++++++- 2 files changed, 27 insertions(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-common.h b/drivers/net/ethernet/amd/xgbe/xgbe-common.h index b2cd3bdba9f8..533b8519ec35 100644 --- a/drivers/net/ethernet/amd/xgbe/xgbe-common.h +++ b/drivers/net/ethernet/amd/xgbe/xgbe-common.h @@ -1331,6 +1331,10 @@ #define MDIO_VEND2_PMA_CDR_CONTROL 0x8056 #endif +#ifndef MDIO_VEND2_PMA_MISC_CTRL0 +#define MDIO_VEND2_PMA_MISC_CTRL0 0x8090 +#endif + #ifndef MDIO_CTRL1_SPEED1G #define MDIO_CTRL1_SPEED1G (MDIO_CTRL1_SPEED10G & ~BMCR_SPEED100) #endif @@ -1389,6 +1393,10 @@ #define XGBE_PMA_RX_RST_0_RESET_ON 0x10 #define XGBE_PMA_RX_RST_0_RESET_OFF 0x00 +#define XGBE_PMA_PLL_CTRL_MASK BIT(15) +#define XGBE_PMA_PLL_CTRL_ENABLE BIT(15) +#define XGBE_PMA_PLL_CTRL_DISABLE 0x0000 + /* Bit setting and getting macros * The get macro will extract the current bit field value from within * the variable diff --git a/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c b/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c index 18e48b3bc402..213769054391 100644 --- a/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c +++ b/drivers/net/ethernet/amd/xgbe/xgbe-phy-v2.c @@ -1977,12 +1977,26 @@ static void xgbe_phy_rx_reset(struct xgbe_prv_data *pdata) } } +static void xgbe_phy_pll_ctrl(struct xgbe_prv_data *pdata, bool enable) +{ + XMDIO_WRITE_BITS(pdata, MDIO_MMD_PMAPMD, MDIO_VEND2_PMA_MISC_CTRL0, + XGBE_PMA_PLL_CTRL_MASK, + enable ? XGBE_PMA_PLL_CTRL_ENABLE + : XGBE_PMA_PLL_CTRL_DISABLE); + + /* Wait for command to complete */ + usleep_range(100, 200); +} + static void xgbe_phy_perform_ratechange(struct xgbe_prv_data *pdata, unsigned int cmd, unsigned int sub_cmd) { unsigned int s0 = 0; unsigned int wait; + /* Disable PLL re-initialization during FW command processing */ + xgbe_phy_pll_ctrl(pdata, false); + /* Log if a previous command did not complete */ if (XP_IOREAD_BITS(pdata, XP_DRIVER_INT_RO, STATUS)) { netif_dbg(pdata, link, pdata->netdev, @@ -2003,7 +2017,7 @@ static void xgbe_phy_perform_ratechange(struct xgbe_prv_data *pdata, wait = XGBE_RATECHANGE_COUNT; while (wait--) { if (!XP_IOREAD_BITS(pdata, XP_DRIVER_INT_RO, STATUS)) - return; + goto reenable_pll; usleep_range(1000, 2000); } @@ -2013,6 +2027,10 @@ static void xgbe_phy_perform_ratechange(struct xgbe_prv_data *pdata, /* Reset on error */ xgbe_phy_rx_reset(pdata); + +reenable_pll: + /* Enable PLL re-initialization */ + xgbe_phy_pll_ctrl(pdata, true); } static void xgbe_phy_rrc(struct xgbe_prv_data *pdata) -- cgit v1.2.3 From f8d384a640dd32aaf0a05fec137ccbf0e986b09f Mon Sep 17 00:00:00 2001 From: Yinjun Zhang Date: Fri, 29 Oct 2021 13:29:02 +0200 Subject: nfp: fix NULL pointer access when scheduling dim work Each rx/tx ring has a related dim work, when rx/tx ring number is decreased by `ethtool -L`, the corresponding rx_ring or tx_ring is assigned NULL, while its related work is not destroyed. When scheduled, the work will access NULL pointer. Fixes: 9d32e4e7e9e1 ("nfp: add support for coalesce adaptive feature") Signed-off-by: Yinjun Zhang Signed-off-by: Louis Peens Signed-off-by: Simon Horman Signed-off-by: David S. Miller --- drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index 5bfa22accf2c..f8b880c8e514 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -2067,7 +2067,7 @@ static int nfp_net_poll(struct napi_struct *napi, int budget) if (napi_complete_done(napi, pkts_polled)) nfp_net_irq_unmask(r_vec->nfp_net, r_vec->irq_entry); - if (r_vec->nfp_net->rx_coalesce_adapt_on) { + if (r_vec->nfp_net->rx_coalesce_adapt_on && r_vec->rx_ring) { struct dim_sample dim_sample = {}; unsigned int start; u64 pkts, bytes; @@ -2082,7 +2082,7 @@ static int nfp_net_poll(struct napi_struct *napi, int budget) net_dim(&r_vec->rx_dim, dim_sample); } - if (r_vec->nfp_net->tx_coalesce_adapt_on) { + if (r_vec->nfp_net->tx_coalesce_adapt_on && r_vec->tx_ring) { struct dim_sample dim_sample = {}; unsigned int start; u64 pkts, bytes; -- cgit v1.2.3 From 17e712c6a1bade9dac02a7bf2b464746faa7e9a0 Mon Sep 17 00:00:00 2001 From: Yinjun Zhang Date: Fri, 29 Oct 2021 13:29:03 +0200 Subject: nfp: fix potential deadlock when canceling dim work When port is linked down, the process which has acquired rtnl_lock will wait for the in-progress dim work to finish, and the work also acquires rtnl_lock, which may cause deadlock. Currently IRQ_MOD registers can be configured by `ethtool -C` and dim work, and which will take effect depends on the execution order, rtnl_lock is useless here, so remove them. Fixes: 9d32e4e7e9e1 ("nfp: add support for coalesce adaptive feature") Signed-off-by: Yinjun Zhang Signed-off-by: Louis Peens Signed-off-by: Simon Horman Signed-off-by: David S. Miller --- drivers/net/ethernet/netronome/nfp/nfp_net_common.c | 4 ---- 1 file changed, 4 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c index f8b880c8e514..850bfdf83d0a 100644 --- a/drivers/net/ethernet/netronome/nfp/nfp_net_common.c +++ b/drivers/net/ethernet/netronome/nfp/nfp_net_common.c @@ -3016,10 +3016,8 @@ static void nfp_net_rx_dim_work(struct work_struct *work) /* copy RX interrupt coalesce parameters */ value = (moder.pkts << 16) | (factor * moder.usec); - rtnl_lock(); nn_writel(nn, NFP_NET_CFG_RXR_IRQ_MOD(r_vec->rx_ring->idx), value); (void)nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_IRQMOD); - rtnl_unlock(); dim->state = DIM_START_MEASURE; } @@ -3047,10 +3045,8 @@ static void nfp_net_tx_dim_work(struct work_struct *work) /* copy TX interrupt coalesce parameters */ value = (moder.pkts << 16) | (factor * moder.usec); - rtnl_lock(); nn_writel(nn, NFP_NET_CFG_TXR_IRQ_MOD(r_vec->tx_ring->idx), value); (void)nfp_net_reconfig(nn, NFP_NET_CFG_UPDATE_IRQMOD); - rtnl_unlock(); dim->state = DIM_START_MEASURE; } -- cgit v1.2.3 From fd8d9731bcdfb22d28e45bce789bcb211c868c78 Mon Sep 17 00:00:00 2001 From: "Russell King (Oracle)" Date: Thu, 28 Oct 2021 15:55:34 +0100 Subject: net: phylink: avoid mvneta warning when setting pause parameters mvneta does not support asymetric pause modes, and it flags this by the lack of AsymPause in the supported field. When setting pause modes, we check that pause->rx_pause == pause->tx_pause, but only when pause autoneg is enabled. When pause autoneg is disabled, we still allow pause->rx_pause != pause->tx_pause, which is incorrect when the MAC does not support asymetric pause, and causes mvneta to issue a warning. Fix this by removing the test for pause->autoneg, so we always check that pause->rx_pause == pause->tx_pause for network devices that do not support AsymPause. Fixes: 9525ae83959b ("phylink: add phylink infrastructure") Signed-off-by: Russell King (Oracle) Signed-off-by: David S. Miller --- drivers/net/phy/phylink.c | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) (limited to 'drivers') diff --git a/drivers/net/phy/phylink.c b/drivers/net/phy/phylink.c index 0a0abe8e4be0..5defc721dd05 100644 --- a/drivers/net/phy/phylink.c +++ b/drivers/net/phy/phylink.c @@ -1724,7 +1724,7 @@ int phylink_ethtool_set_pauseparam(struct phylink *pl, return -EOPNOTSUPP; if (!phylink_test(pl->supported, Asym_Pause) && - !pause->autoneg && pause->rx_pause != pause->tx_pause) + pause->rx_pause != pause->tx_pause) return -EINVAL; pause_state = 0; -- cgit v1.2.3 From 8878e46fcfd46b19964bd90e13b25dd94cbfc9be Mon Sep 17 00:00:00 2001 From: Sukadev Bhattiprolu Date: Fri, 29 Oct 2021 15:03:14 -0700 Subject: ibmvnic: don't stop queue in xmit If adapter's resetting bit is on, discard the packet but don't stop the transmit queue - instead leave that to the reset code. With this change, it is possible that we may get several calls to ibmvnic_xmit() that simply discard packets and return. But if we stop the queue here, we might end up doing so just after __ibmvnic_open() started the queues (during a hard/soft reset) and before the ->resetting bit was cleared. If that happens, there will be no one to restart queue and transmissions will be blocked indefinitely. This can cause a TIMEOUT reset and with auto priority failover enabled, an unnecessary FAILOVER reset to less favored backing device and then a FAILOVER back to the most favored backing device. If we hit the window repeatedly, we can get stuck in a loop of TIMEOUT, FAILOVER, FAILOVER resets leaving the adapter unusable for extended periods of time. Fixes: 7f5b030830fe ("ibmvnic: Free skb's in cases of failure in transmit") Reported-by: Abdul Haleem Reported-by: Vaishnavi Bhat Signed-off-by: Sukadev Bhattiprolu Reviewed-by: Dany Madden Signed-off-by: David S. Miller --- drivers/net/ethernet/ibm/ibmvnic.c | 2 -- 1 file changed, 2 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/ibm/ibmvnic.c b/drivers/net/ethernet/ibm/ibmvnic.c index 6aa6ff89a765..7438138c3766 100644 --- a/drivers/net/ethernet/ibm/ibmvnic.c +++ b/drivers/net/ethernet/ibm/ibmvnic.c @@ -1724,8 +1724,6 @@ static netdev_tx_t ibmvnic_xmit(struct sk_buff *skb, struct net_device *netdev) ind_bufp = &tx_scrq->ind_buf; if (test_bit(0, &adapter->resetting)) { - if (!netif_subqueue_stopped(netdev, skb)) - netif_stop_subqueue(netdev, queue_num); dev_kfree_skb_any(skb); tx_send_failed++; -- cgit v1.2.3 From 6e20d00158f31f7631d68b86996b7e951c4451c8 Mon Sep 17 00:00:00 2001 From: Sukadev Bhattiprolu Date: Fri, 29 Oct 2021 15:03:15 -0700 Subject: ibmvnic: Process crqs after enabling interrupts Soon after registering a CRQ it is possible that we get a fail over or maybe a CRQ_INIT from the VIOS while interrupts were disabled. Look for any such CRQs after enabling interrupts. Otherwise we can intermittently fail to bring up ibmvnic adapters during boot, specially in kexec/kdump kernels. Fixes: 032c5e82847a ("Driver for IBM System i/p VNIC protocol") Reported-by: Vaishnavi Bhat Signed-off-by: Sukadev Bhattiprolu Reviewed-by: Dany Madden Signed-off-by: David S. Miller --- drivers/net/ethernet/ibm/ibmvnic.c | 3 +++ 1 file changed, 3 insertions(+) (limited to 'drivers') diff --git a/drivers/net/ethernet/ibm/ibmvnic.c b/drivers/net/ethernet/ibm/ibmvnic.c index 7438138c3766..84961a83803b 100644 --- a/drivers/net/ethernet/ibm/ibmvnic.c +++ b/drivers/net/ethernet/ibm/ibmvnic.c @@ -5412,6 +5412,9 @@ static int init_crq_queue(struct ibmvnic_adapter *adapter) crq->cur = 0; spin_lock_init(&crq->lock); + /* process any CRQs that were queued before we enabled interrupts */ + tasklet_schedule(&adapter->tasklet); + return retrc; req_irq_failed: -- cgit v1.2.3 From 6b278c0cb378079f3c0c61ae4a369c09ff1a4188 Mon Sep 17 00:00:00 2001 From: Sukadev Bhattiprolu Date: Fri, 29 Oct 2021 15:03:16 -0700 Subject: ibmvnic: delay complete() If we get CRQ_INIT, we set errno to -EIO and first call complete() to notify the waiter. Then we try to schedule a FAILOVER reset. If this occurs while adapter is in PROBING state, ibmvnic_reset() changes the error code to EAGAIN and returns without scheduling the FAILOVER. The purpose of setting error code to EAGAIN is to ask the waiter to retry. But due to the earlier complete() call, the waiter may already have seen the -EIO response and decided not to retry. This can cause intermittent failures when bringing up ibmvnic adapters during boot, specially in in kexec/kdump kernels. Defer the complete() call until after scheduling the reset. Also streamline the error code to EAGAIN. Don't see why we need EIO sometimes. All 3 callers of ibmvnic_reset_init() can handle EAGAIN. Fixes: 17c8705838a5 ("ibmvnic: Return error code if init interrupted by transport event") Reported-by: Vaishnavi Bhat Signed-off-by: Sukadev Bhattiprolu Reviewed-by: Dany Madden Signed-off-by: David S. Miller --- drivers/net/ethernet/ibm/ibmvnic.c | 16 +++++++++------- 1 file changed, 9 insertions(+), 7 deletions(-) (limited to 'drivers') diff --git a/drivers/net/ethernet/ibm/ibmvnic.c b/drivers/net/ethernet/ibm/ibmvnic.c index 84961a83803b..352ffe982d84 100644 --- a/drivers/net/ethernet/ibm/ibmvnic.c +++ b/drivers/net/ethernet/ibm/ibmvnic.c @@ -2565,7 +2565,7 @@ static int ibmvnic_reset(struct ibmvnic_adapter *adapter, if (adapter->state == VNIC_PROBING) { netdev_warn(netdev, "Adapter reset during probe\n"); - adapter->init_done_rc = EAGAIN; + adapter->init_done_rc = -EAGAIN; ret = EAGAIN; goto err; } @@ -5067,11 +5067,6 @@ static void ibmvnic_handle_crq(union ibmvnic_crq *crq, */ adapter->login_pending = false; - if (!completion_done(&adapter->init_done)) { - complete(&adapter->init_done); - adapter->init_done_rc = -EIO; - } - if (adapter->state == VNIC_DOWN) rc = ibmvnic_reset(adapter, VNIC_RESET_PASSIVE_INIT); else @@ -5092,6 +5087,13 @@ static void ibmvnic_handle_crq(union ibmvnic_crq *crq, rc); adapter->failover_pending = false; } + + if (!completion_done(&adapter->init_done)) { + complete(&adapter->init_done); + if (!adapter->init_done_rc) + adapter->init_done_rc = -EAGAIN; + } + break; case IBMVNIC_CRQ_INIT_COMPLETE: dev_info(dev, "Partner initialization complete\n"); @@ -5559,7 +5561,7 @@ static int ibmvnic_probe(struct vio_dev *dev, const struct vio_device_id *id) } rc = ibmvnic_reset_init(adapter, false); - } while (rc == EAGAIN); + } while (rc == -EAGAIN); /* We are ignoring the error from ibmvnic_reset_init() assuming that the * partner is not ready. CRQ is not active. When the partner becomes -- cgit v1.2.3