| Message ID | 20210206013732.508552-1-edwin.peer@broadcom.com |
|---|---|
| State | New |
| Headers | show |
| Series | [net] net: watchdog: hold device global xmit lock during tx disable | expand |
On Fri, 5 Feb 2021 17:37:32 -0800 Edwin Peer wrote: > Prevent netif_tx_disable() running concurrently with dev_watchdog() by > taking the device global xmit lock. Otherwise, the recommended: > > netif_carrier_off(dev); > netif_tx_disable(dev); > > driver shutdown sequence can happen after the watchdog has already > checked carrier, resulting in possible false alarms. This is because > netif_tx_lock() only sets the frozen bit without maintaining the locks > on the individual queues. > > Fixes: c3f26a269c24 ("netdev: Fix lockdep warnings in multiqueue configurations.") > Signed-off-by: Edwin Peer <edwin.peer@broadcom.com> Reviewed-by: Jakub Kicinski <kuba@kernel.org> Even though using the dev->tx_global_lock as a barrier is not very clean in itself the thinking is that this restores previous semantics with deeper changes.
Hello: This patch was applied to netdev/net.git (refs/heads/master): On Fri, 5 Feb 2021 17:37:32 -0800 you wrote: > Prevent netif_tx_disable() running concurrently with dev_watchdog() by > taking the device global xmit lock. Otherwise, the recommended: > > netif_carrier_off(dev); > netif_tx_disable(dev); > > driver shutdown sequence can happen after the watchdog has already > checked carrier, resulting in possible false alarms. This is because > netif_tx_lock() only sets the frozen bit without maintaining the locks > on the individual queues. > > [...] Here is the summary with links: - [net] net: watchdog: hold device global xmit lock during tx disable https://git.kernel.org/netdev/net/c/3aa6bce9af0e You are awesome, thank you! -- Deet-doot-dot, I am a bot. https://korg.docs.kernel.org/patchwork/pwbot.html
diff --git a/include/linux/netdevice.h b/include/linux/netdevice.h index 259be67644e3..5ff27c12ce68 100644 --- a/include/linux/netdevice.h +++ b/include/linux/netdevice.h @@ -4352,6 +4352,7 @@ static inline void netif_tx_disable(struct net_device *dev) local_bh_disable(); cpu = smp_processor_id(); + spin_lock(&dev->tx_global_lock); for (i = 0; i < dev->num_tx_queues; i++) { struct netdev_queue *txq = netdev_get_tx_queue(dev, i); @@ -4359,6 +4360,7 @@ static inline void netif_tx_disable(struct net_device *dev) netif_tx_stop_queue(txq); __netif_tx_unlock(txq); } + spin_unlock(&dev->tx_global_lock); local_bh_enable(); }
Prevent netif_tx_disable() running concurrently with dev_watchdog() by taking the device global xmit lock. Otherwise, the recommended: netif_carrier_off(dev); netif_tx_disable(dev); driver shutdown sequence can happen after the watchdog has already checked carrier, resulting in possible false alarms. This is because netif_tx_lock() only sets the frozen bit without maintaining the locks on the individual queues. Fixes: c3f26a269c24 ("netdev: Fix lockdep warnings in multiqueue configurations.") Signed-off-by: Edwin Peer <edwin.peer@broadcom.com> --- include/linux/netdevice.h | 2 ++ 1 file changed, 2 insertions(+)