From: Thomas Gleixner <tglx@xxxxxxxxxxxxx> enic_dev_wait() has a BUG_ON(in_interrupt()). Chasing the callers of enic_dev_wait() revealed the gems of enic_reset() and enic_tx_hang_reset() which are both invoked through work queues in order to be able to call rtnl_lock(). So far so good. After locking rtnl both functions acquire enic::enic_api_lock which serializes against the (ab)use from infiniband. This is where the trainwreck starts. enic::enic_api_lock is a spin_lock() which implicitly disables preemption, but both functions invoke a ton of functions under that lock which can sleep. The BUG_ON(in_interrupt()) does not trigger in that case because it can't detect the preempt disabled condition. This clearly has never been tested with any of the mandatory debug options for 7+ years, which would have caught that for sure. Cure it by adding a enic_api_busy member to struct enic, which is modified and evaluated with enic::enic_api_lock held. If enic_api_devcmd_proxy_by_index() observes enic::enic_api_busy as true, it drops enic::enic_api_lock and busy waits for enic::enic_api_busy to become false. It would be smarter to wait for a completion of that busy period, but enic_api_devcmd_proxy_by_index() is called with other spin locks held which obviously can't sleep. Remove the BUG_ON(in_interrupt()) check as well because it's incomplete and with proper debugging enabled the problem would have been caught from the debug checks in schedule_timeout(). Fixes: 0b038566c0ea ("drivers/net: enic: Add an interface for USNIC to interact with firmware") Signed-off-by: Thomas Gleixner <tglx@xxxxxxxxxxxxx> Cc: Christian Benvenuti <benve@xxxxxxxxx> Cc: Govindarajulu Varadarajan <_govind@xxxxxxx> Cc: Dave Miller <davem@xxxxxxxxxxxxx> Cc: Jakub Kicinski <kuba@xxxxxxxxxx> Cc: netdev@xxxxxxxxxxxxxxx --- drivers/net/ethernet/cisco/enic/enic.h | 1 + drivers/net/ethernet/cisco/enic/enic_api.c | 6 ++++++ drivers/net/ethernet/cisco/enic/enic_main.c | 27 +++++++++++++++++++++------ 3 files changed, 28 insertions(+), 6 deletions(-) --- a/drivers/net/ethernet/cisco/enic/enic.h +++ b/drivers/net/ethernet/cisco/enic/enic.h @@ -169,6 +169,7 @@ struct enic { u16 num_vfs; #endif spinlock_t enic_api_lock; + bool enic_api_busy; struct enic_port_profile *pp; /* work queue cache line section */ --- a/drivers/net/ethernet/cisco/enic/enic_api.c +++ b/drivers/net/ethernet/cisco/enic/enic_api.c @@ -34,6 +34,12 @@ int enic_api_devcmd_proxy_by_index(struc struct vnic_dev *vdev = enic->vdev; spin_lock(&enic->enic_api_lock); + while (enic->enic_api_busy) { + spin_unlock(&enic->enic_api_lock); + cpu_relax(); + spin_lock(&enic->enic_api_lock); + } + spin_lock_bh(&enic->devcmd_lock); vnic_dev_cmd_proxy_by_index_start(vdev, vf); --- a/drivers/net/ethernet/cisco/enic/enic_main.c +++ b/drivers/net/ethernet/cisco/enic/enic_main.c @@ -2106,8 +2106,6 @@ static int enic_dev_wait(struct vnic_dev int done; int err; - BUG_ON(in_interrupt()); - err = start(vdev, arg); if (err) return err; @@ -2295,6 +2293,13 @@ static int enic_set_rss_nic_cfg(struct e rss_hash_bits, rss_base_cpu, rss_enable); } +static void enic_set_api_busy(struct enic *enic, bool busy) +{ + spin_lock(&enic->enic_api_lock); + enic->enic_api_busy = busy; + spin_unlock(&enic->enic_api_lock); +} + static void enic_reset(struct work_struct *work) { struct enic *enic = container_of(work, struct enic, reset); @@ -2304,7 +2309,9 @@ static void enic_reset(struct work_struc rtnl_lock(); - spin_lock(&enic->enic_api_lock); + /* Stop any activity from infiniband */ + enic_set_api_state(enic, true); + enic_stop(enic->netdev); enic_dev_soft_reset(enic); enic_reset_addr_lists(enic); @@ -2312,7 +2319,10 @@ static void enic_reset(struct work_struc enic_set_rss_nic_cfg(enic); enic_dev_set_ig_vlan_rewrite_mode(enic); enic_open(enic->netdev); - spin_unlock(&enic->enic_api_lock); + + /* Allow infiniband to fiddle with the device again */ + enic_set_api_state(enic, false); + call_netdevice_notifiers(NETDEV_REBOOT, enic->netdev); rtnl_unlock(); @@ -2324,7 +2334,9 @@ static void enic_tx_hang_reset(struct wo rtnl_lock(); - spin_lock(&enic->enic_api_lock); + /* Stop any activity from infiniband */ + enic_set_api_state(enic, true); + enic_dev_hang_notify(enic); enic_stop(enic->netdev); enic_dev_hang_reset(enic); @@ -2333,7 +2345,10 @@ static void enic_tx_hang_reset(struct wo enic_set_rss_nic_cfg(enic); enic_dev_set_ig_vlan_rewrite_mode(enic); enic_open(enic->netdev); - spin_unlock(&enic->enic_api_lock); + + /* Allow infiniband to fiddle with the device again */ + enic_set_api_state(enic, false); + call_netdevice_notifiers(NETDEV_REBOOT, enic->netdev); rtnl_unlock();