| Message ID | 20240903175141.4132898-1-sean.anderson@linux.dev (mailing list archive) |
|---|---|
| State | Superseded |
| Commit | 858430db28a5f5a11f8faa3a6fa805438e6f0851 |
| Delegated to: | Netdev Maintainers |
| Headers | show |
| Series | [net] net: xilinx: axienet: Fix race in axienet_stop | expand |
Hello: This patch was applied to netdev/net.git (main) by Jakub Kicinski <kuba@kernel.org>: On Tue, 3 Sep 2024 13:51:41 -0400 you wrote: > axienet_dma_err_handler can race with axienet_stop in the following > manner: > > CPU 1 CPU 2 > ====================== ================== > axienet_stop() > napi_disable() > axienet_dma_stop() > axienet_dma_err_handler() > napi_disable() > axienet_dma_stop() > axienet_dma_start() > napi_enable() > cancel_work_sync() > free_irq() > > [...] Here is the summary with links: - [net] net: xilinx: axienet: Fix race in axienet_stop https://git.kernel.org/netdev/net/c/858430db28a5 You are awesome, thank you!
diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet.h b/drivers/net/ethernet/xilinx/xilinx_axienet.h index 09c9f9787180..1223fcc1a8da 100644 --- a/drivers/net/ethernet/xilinx/xilinx_axienet.h +++ b/drivers/net/ethernet/xilinx/xilinx_axienet.h @@ -436,6 +436,8 @@ struct skbuf_dma_descriptor { * @tx_bytes: TX byte count for statistics * @tx_stat_sync: Synchronization object for TX stats * @dma_err_task: Work structure to process Axi DMA errors + * @stopping: Set when @dma_err_task shouldn't do anything because we are + * about to stop the device. * @tx_irq: Axidma TX IRQ number * @rx_irq: Axidma RX IRQ number * @eth_irq: Ethernet core IRQ number @@ -507,6 +509,7 @@ struct axienet_local { struct u64_stats_sync tx_stat_sync; struct work_struct dma_err_task; + bool stopping; int tx_irq; int rx_irq; diff --git a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c index 9aeb7b9f3ae4..9eb300fc3590 100644 --- a/drivers/net/ethernet/xilinx/xilinx_axienet_main.c +++ b/drivers/net/ethernet/xilinx/xilinx_axienet_main.c @@ -1460,6 +1460,7 @@ static int axienet_init_legacy_dma(struct net_device *ndev) struct axienet_local *lp = netdev_priv(ndev); /* Enable worker thread for Axi DMA error handling */ + lp->stopping = false; INIT_WORK(&lp->dma_err_task, axienet_dma_err_handler); napi_enable(&lp->napi_rx); @@ -1580,6 +1581,9 @@ static int axienet_stop(struct net_device *ndev) dev_dbg(&ndev->dev, "axienet_close()\n"); if (!lp->use_dmaengine) { + WRITE_ONCE(lp->stopping, true); + flush_work(&lp->dma_err_task); + napi_disable(&lp->napi_tx); napi_disable(&lp->napi_rx); } @@ -2154,6 +2158,10 @@ static void axienet_dma_err_handler(struct work_struct *work) dma_err_task); struct net_device *ndev = lp->ndev; + /* Don't bother if we are going to stop anyway */ + if (READ_ONCE(lp->stopping)) + return; + napi_disable(&lp->napi_tx); napi_disable(&lp->napi_rx);
axienet_dma_err_handler can race with axienet_stop in the following manner: CPU 1 CPU 2 ====================== ================== axienet_stop() napi_disable() axienet_dma_stop() axienet_dma_err_handler() napi_disable() axienet_dma_stop() axienet_dma_start() napi_enable() cancel_work_sync() free_irq() Fix this by setting a flag in axienet_stop telling axienet_dma_err_handler not to bother doing anything. I chose not to use disable_work_sync to allow for easier backporting. Signed-off-by: Sean Anderson <sean.anderson@linux.dev> Fixes: 8a3b7a252dca ("drivers/net/ethernet/xilinx: added Xilinx AXI Ethernet driver") --- drivers/net/ethernet/xilinx/xilinx_axienet.h | 3 +++ drivers/net/ethernet/xilinx/xilinx_axienet_main.c | 8 ++++++++ 2 files changed, 11 insertions(+)