From 45bb006d3c924b1201ed43c87a96b437662dcaa8 Mon Sep 17 00:00:00 2001 From: Przemyslaw Patynowski Date: Tue, 9 Aug 2022 10:57:44 +0200 Subject: [PATCH 1/3] i40e: Fix ADQ rate limiting for PF Fix HW rate limiting for ADQ. Fallback to kernel queue selection for ADQ, as it is network stack that decides which queue to use for transmit with ADQ configured. Reset PF after creation of VMDq2 VSIs required for ADQ, as to reprogram TX queue contexts in i40e_configure_tx_ring. Without this patch PF would limit TX rate only according to TC0. Fixes: a9ce82f744dc ("i40e: Enable 'channel' mode in mqprio for TC configs") Signed-off-by: Przemyslaw Patynowski Signed-off-by: Jan Sokolowski Tested-by: Bharathi Sreenivas Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/i40e/i40e_main.c | 3 +++ drivers/net/ethernet/intel/i40e/i40e_txrx.c | 3 ++- 2 files changed, 5 insertions(+), 1 deletion(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c b/drivers/net/ethernet/intel/i40e/i40e_main.c index 9f1d5de7bf16..10c1e1ea83a1 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_main.c +++ b/drivers/net/ethernet/intel/i40e/i40e_main.c @@ -6659,6 +6659,9 @@ static int i40e_configure_queue_channels(struct i40e_vsi *vsi) vsi->tc_seid_map[i] = ch->seid; } } + + /* reset to reconfigure TX queue contexts */ + i40e_do_reset(vsi->back, I40E_PF_RESET_FLAG, true); return ret; err_free: diff --git a/drivers/net/ethernet/intel/i40e/i40e_txrx.c b/drivers/net/ethernet/intel/i40e/i40e_txrx.c index d4226161a3ef..69e67eb6aea7 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_txrx.c +++ b/drivers/net/ethernet/intel/i40e/i40e_txrx.c @@ -3688,7 +3688,8 @@ u16 i40e_lan_select_queue(struct net_device *netdev, u8 prio; /* is DCB enabled at all? */ - if (vsi->tc_config.numtc == 1) + if (vsi->tc_config.numtc == 1 || + i40e_is_tc_mqprio_enabled(vsi->back)) return netdev_pick_tx(netdev, skb, sb_dev); prio = skb->priority; From fb8396aeda5872369a8ed6d2301e2c86e303c520 Mon Sep 17 00:00:00 2001 From: Ivan Vecera Date: Tue, 16 Aug 2022 18:22:30 +0200 Subject: [PATCH 2/3] i40e: Fix kernel crash during module removal The driver incorrectly frees client instance and subsequent i40e module removal leads to kernel crash. Reproducer: 1. Do ethtool offline test followed immediately by another one host# ethtool -t eth0 offline; ethtool -t eth0 offline 2. Remove recursively irdma module that also removes i40e module host# modprobe -r irdma Result: [ 8675.035651] i40e 0000:3d:00.0 eno1: offline testing starting [ 8675.193774] i40e 0000:3d:00.0 eno1: testing finished [ 8675.201316] i40e 0000:3d:00.0 eno1: offline testing starting [ 8675.358921] i40e 0000:3d:00.0 eno1: testing finished [ 8675.496921] i40e 0000:3d:00.0: IRDMA hardware initialization FAILED init_state=2 status=-110 [ 8686.188955] i40e 0000:3d:00.1: i40e_ptp_stop: removed PHC on eno2 [ 8686.943890] i40e 0000:3d:00.1: Deleted LAN device PF1 bus=0x3d dev=0x00 func=0x01 [ 8686.952669] i40e 0000:3d:00.0: i40e_ptp_stop: removed PHC on eno1 [ 8687.761787] BUG: kernel NULL pointer dereference, address: 0000000000000030 [ 8687.768755] #PF: supervisor read access in kernel mode [ 8687.773895] #PF: error_code(0x0000) - not-present page [ 8687.779034] PGD 0 P4D 0 [ 8687.781575] Oops: 0000 [#1] PREEMPT SMP NOPTI [ 8687.785935] CPU: 51 PID: 172891 Comm: rmmod Kdump: loaded Tainted: G W I 5.19.0+ #2 [ 8687.794800] Hardware name: Intel Corporation S2600WFD/S2600WFD, BIOS SE5C620.86B.0X.02.0001.051420190324 05/14/2019 [ 8687.805222] RIP: 0010:i40e_lan_del_device+0x13/0xb0 [i40e] [ 8687.810719] Code: d4 84 c0 0f 84 b8 25 01 00 e9 9c 25 01 00 41 bc f4 ff ff ff eb 91 90 0f 1f 44 00 00 41 54 55 53 48 8b 87 58 08 00 00 48 89 fb <48> 8b 68 30 48 89 ef e8 21 8a 0f d5 48 89 ef e8 a9 78 0f d5 48 8b [ 8687.829462] RSP: 0018:ffffa604072efce0 EFLAGS: 00010202 [ 8687.834689] RAX: 0000000000000000 RBX: ffff8f43833b2000 RCX: 0000000000000000 [ 8687.841821] RDX: 0000000000000000 RSI: ffff8f4b0545b298 RDI: ffff8f43833b2000 [ 8687.848955] RBP: ffff8f43833b2000 R08: 0000000000000001 R09: 0000000000000000 [ 8687.856086] R10: 0000000000000000 R11: 000ffffffffff000 R12: ffff8f43833b2ef0 [ 8687.863218] R13: ffff8f43833b2ef0 R14: ffff915103966000 R15: ffff8f43833b2008 [ 8687.870342] FS: 00007f79501c3740(0000) GS:ffff8f4adffc0000(0000) knlGS:0000000000000000 [ 8687.878427] CS: 0010 DS: 0000 ES: 0000 CR0: 0000000080050033 [ 8687.884174] CR2: 0000000000000030 CR3: 000000014276e004 CR4: 00000000007706e0 [ 8687.891306] DR0: 0000000000000000 DR1: 0000000000000000 DR2: 0000000000000000 [ 8687.898441] DR3: 0000000000000000 DR6: 00000000fffe0ff0 DR7: 0000000000000400 [ 8687.905572] PKRU: 55555554 [ 8687.908286] Call Trace: [ 8687.910737] [ 8687.912843] i40e_remove+0x2c0/0x330 [i40e] [ 8687.917040] pci_device_remove+0x33/0xa0 [ 8687.920962] device_release_driver_internal+0x1aa/0x230 [ 8687.926188] driver_detach+0x44/0x90 [ 8687.929770] bus_remove_driver+0x55/0xe0 [ 8687.933693] pci_unregister_driver+0x2a/0xb0 [ 8687.937967] i40e_exit_module+0xc/0xf48 [i40e] Two offline tests cause IRDMA driver failure (ETIMEDOUT) and this failure is indicated back to i40e_client_subtask() that calls i40e_client_del_instance() to free client instance referenced by pf->cinst and sets this pointer to NULL. During the module removal i40e_remove() calls i40e_lan_del_device() that dereferences pf->cinst that is NULL -> crash. Do not remove client instance when client open callbacks fails and just clear __I40E_CLIENT_INSTANCE_OPENED bit. The driver also needs to take care about this situation (when netdev is up and client is NOT opened) in i40e_notify_client_of_netdev_close() and calls client close callback only when __I40E_CLIENT_INSTANCE_OPENED is set. Fixes: 0ef2d5afb12d ("i40e: KISS the client interface") Signed-off-by: Ivan Vecera Tested-by: Helena Anna Dubel Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/i40e/i40e_client.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/drivers/net/ethernet/intel/i40e/i40e_client.c b/drivers/net/ethernet/intel/i40e/i40e_client.c index ea2bb0140a6e..10d7a982a5b9 100644 --- a/drivers/net/ethernet/intel/i40e/i40e_client.c +++ b/drivers/net/ethernet/intel/i40e/i40e_client.c @@ -177,6 +177,10 @@ void i40e_notify_client_of_netdev_close(struct i40e_vsi *vsi, bool reset) "Cannot locate client instance close routine\n"); return; } + if (!test_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state)) { + dev_dbg(&pf->pdev->dev, "Client is not open, abort close\n"); + return; + } cdev->client->ops->close(&cdev->lan_info, cdev->client, reset); clear_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state); i40e_client_release_qvlist(&cdev->lan_info); @@ -429,7 +433,6 @@ void i40e_client_subtask(struct i40e_pf *pf) /* Remove failed client instance */ clear_bit(__I40E_CLIENT_INSTANCE_OPENED, &cdev->state); - i40e_client_del_instance(pf); return; } } From aa626da947e9cd30c4cf727493903e1adbb2c0a0 Mon Sep 17 00:00:00 2001 From: Ivan Vecera Date: Tue, 30 Aug 2022 10:16:27 +0200 Subject: [PATCH 3/3] iavf: Detach device during reset task iavf_reset_task() takes crit_lock at the beginning and holds it during whole call. The function subsequently calls iavf_init_interrupt_scheme() that grabs RTNL. Problem occurs when userspace initiates during the reset task any ndo callback that runs under RTNL like iavf_open() because some of that functions tries to take crit_lock. This leads to classic A-B B-A deadlock scenario. To resolve this situation the device should be detached in iavf_reset_task() prior taking crit_lock to avoid subsequent ndos running under RTNL and reattach the device at the end. Fixes: 62fe2a865e6d ("i40evf: add missing rtnl_lock() around i40evf_set_interrupt_capability") Cc: Jacob Keller Cc: Patryk Piotrowski Cc: SlawomirX Laba Tested-by: Vitaly Grinberg Signed-off-by: Ivan Vecera Tested-by: Konrad Jankowski Signed-off-by: Tony Nguyen --- drivers/net/ethernet/intel/iavf/iavf_main.c | 14 +++++++++++--- 1 file changed, 11 insertions(+), 3 deletions(-) diff --git a/drivers/net/ethernet/intel/iavf/iavf_main.c b/drivers/net/ethernet/intel/iavf/iavf_main.c index f39440ad5c50..10aa99dfdcdb 100644 --- a/drivers/net/ethernet/intel/iavf/iavf_main.c +++ b/drivers/net/ethernet/intel/iavf/iavf_main.c @@ -2877,6 +2877,11 @@ static void iavf_reset_task(struct work_struct *work) int i = 0, err; bool running; + /* Detach interface to avoid subsequent NDO callbacks */ + rtnl_lock(); + netif_device_detach(netdev); + rtnl_unlock(); + /* When device is being removed it doesn't make sense to run the reset * task, just return in such a case. */ @@ -2884,7 +2889,7 @@ static void iavf_reset_task(struct work_struct *work) if (adapter->state != __IAVF_REMOVE) queue_work(iavf_wq, &adapter->reset_task); - return; + goto reset_finish; } while (!mutex_trylock(&adapter->client_lock)) @@ -2954,7 +2959,6 @@ continue_reset: if (running) { netif_carrier_off(netdev); - netif_tx_stop_all_queues(netdev); adapter->link_up = false; iavf_napi_disable_all(adapter); } @@ -3084,7 +3088,7 @@ continue_reset: mutex_unlock(&adapter->client_lock); mutex_unlock(&adapter->crit_lock); - return; + goto reset_finish; reset_err: if (running) { set_bit(__IAVF_VSI_DOWN, adapter->vsi.state); @@ -3095,6 +3099,10 @@ reset_err: mutex_unlock(&adapter->client_lock); mutex_unlock(&adapter->crit_lock); dev_err(&adapter->pdev->dev, "failed to allocate resources during reinit\n"); +reset_finish: + rtnl_lock(); + netif_device_attach(netdev); + rtnl_unlock(); } /**