| Message ID | 20240511062009.652918-1-wei.fang@nxp.com (mailing list archive) |
|---|---|
| State | Accepted |
| Commit | c2e0c58b25a0a0c37ec643255558c5af4450c9f5 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] net: fec: remove .ndo_poll_controller to avoid deadlocks | expand |
Hello: This patch was applied to netdev/net.git (main) by Jakub Kicinski <kuba@kernel.org>: On Sat, 11 May 2024 14:20:09 +0800 you wrote: > There is a deadlock issue found in sungem driver, please refer to the > commit ac0a230f719b ("eth: sungem: remove .ndo_poll_controller to avoid > deadlocks"). The root cause of the issue is that netpoll is in atomic > context and disable_irq() is called by .ndo_poll_controller interface > of sungem driver, however, disable_irq() might sleep. After analyzing > the implementation of fec_poll_controller(), the fec driver should have > the same issue. Due to the fec driver uses NAPI for TX completions, the > .ndo_poll_controller is unnecessary to be implemented in the fec driver, > so fec_poll_controller() can be safely removed. > > [...] Here is the summary with links: - [net] net: fec: remove .ndo_poll_controller to avoid deadlocks https://git.kernel.org/netdev/net/c/c2e0c58b25a0 You are awesome, thank you!
diff --git a/drivers/net/ethernet/freescale/fec_main.c b/drivers/net/ethernet/freescale/fec_main.c index 8bd213da8fb6..a72d8a2eb0b3 100644 --- a/drivers/net/ethernet/freescale/fec_main.c +++ b/drivers/net/ethernet/freescale/fec_main.c @@ -3674,29 +3674,6 @@ fec_set_mac_address(struct net_device *ndev, void *p) return 0; } -#ifdef CONFIG_NET_POLL_CONTROLLER -/** - * fec_poll_controller - FEC Poll controller function - * @dev: The FEC network adapter - * - * Polled functionality used by netconsole and others in non interrupt mode - * - */ -static void fec_poll_controller(struct net_device *dev) -{ - int i; - struct fec_enet_private *fep = netdev_priv(dev); - - for (i = 0; i < FEC_IRQ_NUM; i++) { - if (fep->irq[i] > 0) { - disable_irq(fep->irq[i]); - fec_enet_interrupt(fep->irq[i], dev); - enable_irq(fep->irq[i]); - } - } -} -#endif - static inline void fec_enet_set_netdev_features(struct net_device *netdev, netdev_features_t features) { @@ -4003,9 +3980,6 @@ static const struct net_device_ops fec_netdev_ops = { .ndo_tx_timeout = fec_timeout, .ndo_set_mac_address = fec_set_mac_address, .ndo_eth_ioctl = phy_do_ioctl_running, -#ifdef CONFIG_NET_POLL_CONTROLLER - .ndo_poll_controller = fec_poll_controller, -#endif .ndo_set_features = fec_set_features, .ndo_bpf = fec_enet_bpf, .ndo_xdp_xmit = fec_enet_xdp_xmit,
There is a deadlock issue found in sungem driver, please refer to the commit ac0a230f719b ("eth: sungem: remove .ndo_poll_controller to avoid deadlocks"). The root cause of the issue is that netpoll is in atomic context and disable_irq() is called by .ndo_poll_controller interface of sungem driver, however, disable_irq() might sleep. After analyzing the implementation of fec_poll_controller(), the fec driver should have the same issue. Due to the fec driver uses NAPI for TX completions, the .ndo_poll_controller is unnecessary to be implemented in the fec driver, so fec_poll_controller() can be safely removed. Fixes: 7f5c6addcdc0 ("net/fec: add poll controller function for fec nic") Signed-off-by: Wei Fang <wei.fang@nxp.com> --- drivers/net/ethernet/freescale/fec_main.c | 26 ----------------------- 1 file changed, 26 deletions(-)