From: Vitor Soares vitor.soares@toradex.com
When the mcp251xfd_start_xmit() function fails, the driver stops processing messages, and the interrupt routine does not return, running indefinitely even after killing the running application.
Error messages: [ 441.298819] mcp251xfd spi2.0 can0: ERROR in mcp251xfd_start_xmit: -16 [ 441.306498] mcp251xfd spi2.0 can0: Transmit Event FIFO buffer not empty. (seq=0x000017c7, tef_tail=0x000017cf, tef_head=0x000017d0, tx_head=0x000017d3). ... and repeat forever.
The issue can be triggered when multiple devices share the same SPI interface. And there is concurrent access to the bus.
The problem occurs because tx_ring->head increments even if mcp251xfd_start_xmit() fails. Consequently, the driver skips one TX package while still expecting a response in mcp251xfd_handle_tefif_one().
This patch resolves the issue by starting a workqueue to write the tx obj synchronously if err = -EBUSY. In case of another error, it decrements tx_ring->head, removes skb from the echo stack, and drops the message.
Fixes: 55e5b97f003e ("can: mcp25xxfd: add driver for Microchip MCP25xxFD SPI CAN") Cc: stable@vger.kernel.org Signed-off-by: Vitor Soares vitor.soares@toradex.com ---
V4->V5: - Start a workqueue to write tx obj with spi_sync() when spi_async() == -EBUSY.
V3->V4: - Leave can_put_echo_skb() and stop the queue if needed, before mcp251xfd_tx_obj_write(). - Re-sync head and remove echo skb if mcp251xfd_tx_obj_write() fails. - Revert -> return NETDEV_TX_BUSY if mcp251xfd_tx_obj_write() == -EBUSY.
V2->V3: - Add tx_dropped stats. - netdev_sent_queue() only if can_put_echo_skb() succeed.
V1->V2: - Return NETDEV_TX_BUSY if mcp251xfd_tx_obj_write() == -EBUSY. - Rework the commit message to address the change above. - Change can_put_echo_skb() to be called after mcp251xfd_tx_obj_write() succeed. Otherwise, we get Kernel NULL pointer dereference error.
.../net/can/spi/mcp251xfd/mcp251xfd-core.c | 13 ++++- drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c | 51 ++++++++++++++++--- drivers/net/can/spi/mcp251xfd/mcp251xfd.h | 5 ++ 3 files changed, 60 insertions(+), 9 deletions(-)
diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c b/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c index 1d9057dc44f2..6cca853f2b1e 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c @@ -2141,15 +2141,25 @@ static int mcp251xfd_probe(struct spi_device *spi) if (err) goto out_free_candev;
+ priv->tx_work_obj = NULL; + priv->wq = alloc_workqueue("mcp251xfd_wq", WQ_FREEZABLE, 0); + if (!priv->wq) { + err = -ENOMEM; + goto out_can_rx_offload_del; + } + INIT_WORK(&priv->tx_work, mcp251xfd_tx_obj_write_sync); + err = mcp251xfd_register(priv); if (err) { dev_err_probe(&spi->dev, err, "Failed to detect %s.\n", mcp251xfd_get_model_str(priv)); - goto out_can_rx_offload_del; + goto out_can_free_wq; }
return 0;
+ out_can_free_wq: + destroy_workqueue(priv->wq); out_can_rx_offload_del: can_rx_offload_del(&priv->offload); out_free_candev: @@ -2165,6 +2175,7 @@ static void mcp251xfd_remove(struct spi_device *spi) struct mcp251xfd_priv *priv = spi_get_drvdata(spi); struct net_device *ndev = priv->ndev;
+ destroy_workqueue(priv->wq); can_rx_offload_del(&priv->offload); mcp251xfd_unregister(priv); spi->max_speed_hz = priv->spi_max_speed_hz_orig; diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c b/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c index 160528d3cc26..1e7ddf316643 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c @@ -131,6 +131,41 @@ mcp251xfd_tx_obj_from_skb(const struct mcp251xfd_priv *priv, tx_obj->xfer[0].len = len; }
+static void mcp251xfd_tx_failure_drop(const struct mcp251xfd_priv *priv, + struct mcp251xfd_tx_ring *tx_ring, + int err) +{ + struct net_device *ndev = priv->ndev; + struct net_device_stats *stats = &ndev->stats; + unsigned int frame_len = 0; + u8 tx_head; + + tx_ring->head--; + stats->tx_dropped++; + tx_head = mcp251xfd_get_tx_head(tx_ring); + can_free_echo_skb(ndev, tx_head, &frame_len); + netdev_completed_queue(ndev, 1, frame_len); + netif_wake_queue(ndev); + + if (net_ratelimit()) + netdev_err(priv->ndev, "ERROR in %s: %d\n", __func__, err); +} + +void mcp251xfd_tx_obj_write_sync(struct work_struct *ws) +{ + struct mcp251xfd_priv *priv = container_of(ws, struct mcp251xfd_priv, + tx_work); + struct mcp251xfd_tx_obj *tx_obj = priv->tx_work_obj; + struct mcp251xfd_tx_ring *tx_ring = priv->tx; + int err; + + err = spi_sync(priv->spi, &tx_obj->msg); + if (err) + mcp251xfd_tx_failure_drop(priv, tx_ring, err); + + priv->tx_work_obj = NULL; +} + static int mcp251xfd_tx_obj_write(const struct mcp251xfd_priv *priv, struct mcp251xfd_tx_obj *tx_obj) { @@ -175,7 +210,7 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, if (can_dev_dropped_skb(ndev, skb)) return NETDEV_TX_OK;
- if (mcp251xfd_tx_busy(priv, tx_ring)) + if (mcp251xfd_tx_busy(priv, tx_ring) || priv->tx_work_obj) return NETDEV_TX_BUSY;
tx_obj = mcp251xfd_get_tx_obj_next(tx_ring); @@ -193,13 +228,13 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, netdev_sent_queue(priv->ndev, frame_len);
err = mcp251xfd_tx_obj_write(priv, tx_obj); - if (err) - goto out_err; - - return NETDEV_TX_OK; - - out_err: - netdev_err(priv->ndev, "ERROR in %s: %d\n", __func__, err); + if (err == -EBUSY) { + priv->tx_work_obj = tx_obj; + netif_stop_queue(ndev); + queue_work(priv->wq, &priv->tx_work); + } else if (err) { + mcp251xfd_tx_failure_drop(priv, tx_ring, err); + }
return NETDEV_TX_OK; } diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd.h b/drivers/net/can/spi/mcp251xfd/mcp251xfd.h index 24510b3b8020..4e27a33f4030 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd.h +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd.h @@ -633,6 +633,10 @@ struct mcp251xfd_priv { struct mcp251xfd_rx_ring *rx[MCP251XFD_FIFO_RX_NUM]; struct mcp251xfd_tx_ring tx[MCP251XFD_FIFO_TX_NUM];
+ struct workqueue_struct *wq; + struct work_struct tx_work; + struct mcp251xfd_tx_obj *tx_work_obj; + DECLARE_BITMAP(flags, __MCP251XFD_FLAGS_SIZE__);
u8 rx_ring_num; @@ -952,6 +956,7 @@ void mcp251xfd_skb_set_timestamp(const struct mcp251xfd_priv *priv, void mcp251xfd_timestamp_init(struct mcp251xfd_priv *priv); void mcp251xfd_timestamp_stop(struct mcp251xfd_priv *priv);
+void mcp251xfd_tx_obj_write_sync(struct work_struct *ws); netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, struct net_device *ndev);
On 14.05.2024 11:58:22, Vitor Soares wrote:
From: Vitor Soares vitor.soares@toradex.com
When the mcp251xfd_start_xmit() function fails, the driver stops processing messages, and the interrupt routine does not return, running indefinitely even after killing the running application.
Error messages: [ 441.298819] mcp251xfd spi2.0 can0: ERROR in mcp251xfd_start_xmit: -16 [ 441.306498] mcp251xfd spi2.0 can0: Transmit Event FIFO buffer not empty. (seq=0x000017c7, tef_tail=0x000017cf, tef_head=0x000017d0, tx_head=0x000017d3). ... and repeat forever.
The issue can be triggered when multiple devices share the same SPI interface. And there is concurrent access to the bus.
The problem occurs because tx_ring->head increments even if mcp251xfd_start_xmit() fails. Consequently, the driver skips one TX package while still expecting a response in mcp251xfd_handle_tefif_one().
This patch resolves the issue by starting a workqueue to write the tx obj synchronously if err = -EBUSY. In case of another error, it decrements tx_ring->head, removes skb from the echo stack, and drops the message.
This looks quite good! Good work!
I think you better move the allocation/destroy of the wq into the open() and stop() callbacks. You have to destroy the workqueue before putting the interface to sleep.
Fixes: 55e5b97f003e ("can: mcp25xxfd: add driver for Microchip MCP25xxFD SPI CAN") Cc: stable@vger.kernel.org Signed-off-by: Vitor Soares vitor.soares@toradex.com
V4->V5:
- Start a workqueue to write tx obj with spi_sync() when spi_async() == -EBUSY.
V3->V4:
- Leave can_put_echo_skb() and stop the queue if needed, before mcp251xfd_tx_obj_write().
- Re-sync head and remove echo skb if mcp251xfd_tx_obj_write() fails.
- Revert -> return NETDEV_TX_BUSY if mcp251xfd_tx_obj_write() == -EBUSY.
V2->V3:
- Add tx_dropped stats.
- netdev_sent_queue() only if can_put_echo_skb() succeed.
V1->V2:
- Return NETDEV_TX_BUSY if mcp251xfd_tx_obj_write() == -EBUSY.
- Rework the commit message to address the change above.
- Change can_put_echo_skb() to be called after mcp251xfd_tx_obj_write() succeed. Otherwise, we get Kernel NULL pointer dereference error.
.../net/can/spi/mcp251xfd/mcp251xfd-core.c | 13 ++++- drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c | 51 ++++++++++++++++--- drivers/net/can/spi/mcp251xfd/mcp251xfd.h | 5 ++ 3 files changed, 60 insertions(+), 9 deletions(-)
diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c b/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c index 1d9057dc44f2..6cca853f2b1e 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c @@ -2141,15 +2141,25 @@ static int mcp251xfd_probe(struct spi_device *spi) if (err) goto out_free_candev;
- priv->tx_work_obj = NULL;
- priv->wq = alloc_workqueue("mcp251xfd_wq", WQ_FREEZABLE, 0);
The m_can driver uses a ordered workqueue and you can add the name of the spi device to the wq's name :)
priv->wq = alloc_ordered_workqueue("%s-mcp251xfd_wq", WQ_FREEZABLE | WQ_MEM_RECLAIM, dev_name(&spi->dev));
- if (!priv->wq) {
err = -ENOMEM;goto out_can_rx_offload_del;- }
- INIT_WORK(&priv->tx_work, mcp251xfd_tx_obj_write_sync);
- err = mcp251xfd_register(priv); if (err) { dev_err_probe(&spi->dev, err, "Failed to detect %s.\n", mcp251xfd_get_model_str(priv));
goto out_can_rx_offload_del;
goto out_can_free_wq;
nitpick: out_destroy_workqueue;
to match the function call.
} return 0;
- out_can_free_wq:
- destroy_workqueue(priv->wq); out_can_rx_offload_del: can_rx_offload_del(&priv->offload); out_free_candev:
@@ -2165,6 +2175,7 @@ static void mcp251xfd_remove(struct spi_device *spi) struct mcp251xfd_priv *priv = spi_get_drvdata(spi); struct net_device *ndev = priv->ndev;
- destroy_workqueue(priv->wq); can_rx_offload_del(&priv->offload); mcp251xfd_unregister(priv); spi->max_speed_hz = priv->spi_max_speed_hz_orig;
diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c b/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c index 160528d3cc26..1e7ddf316643 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c @@ -131,6 +131,41 @@ mcp251xfd_tx_obj_from_skb(const struct mcp251xfd_priv *priv, tx_obj->xfer[0].len = len; } +static void mcp251xfd_tx_failure_drop(const struct mcp251xfd_priv *priv,
struct mcp251xfd_tx_ring *tx_ring,int err)+{
- struct net_device *ndev = priv->ndev;
- struct net_device_stats *stats = &ndev->stats;
- unsigned int frame_len = 0;
- u8 tx_head;
- tx_ring->head--;
- stats->tx_dropped++;
- tx_head = mcp251xfd_get_tx_head(tx_ring);
- can_free_echo_skb(ndev, tx_head, &frame_len);
- netdev_completed_queue(ndev, 1, frame_len);
- netif_wake_queue(ndev);
- if (net_ratelimit())
netdev_err(priv->ndev, "ERROR in %s: %d\n", __func__, err);+}
+void mcp251xfd_tx_obj_write_sync(struct work_struct *ws) +{
- struct mcp251xfd_priv *priv = container_of(ws, struct mcp251xfd_priv,
tx_work);- struct mcp251xfd_tx_obj *tx_obj = priv->tx_work_obj;
- struct mcp251xfd_tx_ring *tx_ring = priv->tx;
- int err;
- err = spi_sync(priv->spi, &tx_obj->msg);
- if (err)
mcp251xfd_tx_failure_drop(priv, tx_ring, err);- priv->tx_work_obj = NULL;
Race condition: - after spi_sync() the CAN frame is send - after the TX complete IRQ the TX queue is restarted - the xmit handler might get BUSY - fill the tx_work_obj again
+}
static int mcp251xfd_tx_obj_write(const struct mcp251xfd_priv *priv, struct mcp251xfd_tx_obj *tx_obj) { @@ -175,7 +210,7 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, if (can_dev_dropped_skb(ndev, skb)) return NETDEV_TX_OK;
- if (mcp251xfd_tx_busy(priv, tx_ring))
- if (mcp251xfd_tx_busy(priv, tx_ring) || priv->tx_work_obj)
This should not happen, but better save than sorry.
return NETDEV_TX_BUSY;tx_obj = mcp251xfd_get_tx_obj_next(tx_ring); @@ -193,13 +228,13 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, netdev_sent_queue(priv->ndev, frame_len); err = mcp251xfd_tx_obj_write(priv, tx_obj);
- if (err)
goto out_err;- return NETDEV_TX_OK;
- out_err:
- netdev_err(priv->ndev, "ERROR in %s: %d\n", __func__, err);
- if (err == -EBUSY) {
priv->tx_work_obj = tx_obj;netif_stop_queue(ndev);queue_work(priv->wq, &priv->tx_work);
nitpick: I would do "netif_stop_queue(ndev);" first.
My mental idea of the code flow is: - stop the queue - do everything to start the workqueue.
- } else if (err) {
mcp251xfd_tx_failure_drop(priv, tx_ring, err);- }
return NETDEV_TX_OK; } diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd.h b/drivers/net/can/spi/mcp251xfd/mcp251xfd.h index 24510b3b8020..4e27a33f4030 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd.h +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd.h @@ -633,6 +633,10 @@ struct mcp251xfd_priv { struct mcp251xfd_rx_ring *rx[MCP251XFD_FIFO_RX_NUM]; struct mcp251xfd_tx_ring tx[MCP251XFD_FIFO_TX_NUM];
- struct workqueue_struct *wq;
- struct work_struct tx_work;
- struct mcp251xfd_tx_obj *tx_work_obj;
- DECLARE_BITMAP(flags, __MCP251XFD_FLAGS_SIZE__);
u8 rx_ring_num; @@ -952,6 +956,7 @@ void mcp251xfd_skb_set_timestamp(const struct mcp251xfd_priv *priv, void mcp251xfd_timestamp_init(struct mcp251xfd_priv *priv); void mcp251xfd_timestamp_stop(struct mcp251xfd_priv *priv); +void mcp251xfd_tx_obj_write_sync(struct work_struct *ws); netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, struct net_device *ndev); -- 2.34.1
regards, Marc
Hi Marc,
Appreciate your feedback.
On Tue, 2024-05-14 at 14:23 +0200, Marc Kleine-Budde wrote:
On 14.05.2024 11:58:22, Vitor Soares wrote:
From: Vitor Soares vitor.soares@toradex.com
When the mcp251xfd_start_xmit() function fails, the driver stops processing messages, and the interrupt routine does not return, running indefinitely even after killing the running application.
Error messages: [ 441.298819] mcp251xfd spi2.0 can0: ERROR in mcp251xfd_start_xmit: -16 [ 441.306498] mcp251xfd spi2.0 can0: Transmit Event FIFO buffer not empty. (seq=0x000017c7, tef_tail=0x000017cf, tef_head=0x000017d0, tx_head=0x000017d3). ... and repeat forever.
The issue can be triggered when multiple devices share the same SPI interface. And there is concurrent access to the bus.
The problem occurs because tx_ring->head increments even if mcp251xfd_start_xmit() fails. Consequently, the driver skips one TX package while still expecting a response in mcp251xfd_handle_tefif_one().
This patch resolves the issue by starting a workqueue to write the tx obj synchronously if err = -EBUSY. In case of another error, it decrements tx_ring->head, removes skb from the echo stack, and drops the message.
This looks quite good! Good work!
I think you better move the allocation/destroy of the wq into the open() and stop() callbacks. You have to destroy the workqueue before putting the interface to sleep.
Fixes: 55e5b97f003e ("can: mcp25xxfd: add driver for Microchip MCP25xxFD SPI CAN") Cc: stable@vger.kernel.org Signed-off-by: Vitor Soares vitor.soares@toradex.com
V4->V5: - Start a workqueue to write tx obj with spi_sync() when spi_async() == - EBUSY.
V3->V4: - Leave can_put_echo_skb() and stop the queue if needed, before mcp251xfd_tx_obj_write(). - Re-sync head and remove echo skb if mcp251xfd_tx_obj_write() fails. - Revert -> return NETDEV_TX_BUSY if mcp251xfd_tx_obj_write() == -EBUSY.
V2->V3: - Add tx_dropped stats. - netdev_sent_queue() only if can_put_echo_skb() succeed.
V1->V2: - Return NETDEV_TX_BUSY if mcp251xfd_tx_obj_write() == -EBUSY. - Rework the commit message to address the change above. - Change can_put_echo_skb() to be called after mcp251xfd_tx_obj_write() succeed. Otherwise, we get Kernel NULL pointer dereference error.
.../net/can/spi/mcp251xfd/mcp251xfd-core.c | 13 ++++- drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c | 51 ++++++++++++++++--- drivers/net/can/spi/mcp251xfd/mcp251xfd.h | 5 ++ 3 files changed, 60 insertions(+), 9 deletions(-)
diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c b/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c index 1d9057dc44f2..6cca853f2b1e 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd-core.c @@ -2141,15 +2141,25 @@ static int mcp251xfd_probe(struct spi_device *spi) if (err) goto out_free_candev; + priv->tx_work_obj = NULL; + priv->wq = alloc_workqueue("mcp251xfd_wq", WQ_FREEZABLE, 0);
The m_can driver uses a ordered workqueue and you can add the name of the spi device to the wq's name :)
priv->wq = alloc_ordered_workqueue("%s-mcp251xfd_wq", WQ_FREEZABLE | WQ_MEM_RECLAIM, dev_name(&spi->dev));
+ if (!priv->wq) { + err = -ENOMEM; + goto out_can_rx_offload_del; + } + INIT_WORK(&priv->tx_work, mcp251xfd_tx_obj_write_sync);
err = mcp251xfd_register(priv); if (err) { dev_err_probe(&spi->dev, err, "Failed to detect %s.\n", mcp251xfd_get_model_str(priv)); - goto out_can_rx_offload_del; + goto out_can_free_wq;
nitpick: out_destroy_workqueue;
to match the function call.
} return 0;
- out_can_free_wq:
+ destroy_workqueue(priv->wq); out_can_rx_offload_del: can_rx_offload_del(&priv->offload); out_free_candev: @@ -2165,6 +2175,7 @@ static void mcp251xfd_remove(struct spi_device *spi) struct mcp251xfd_priv *priv = spi_get_drvdata(spi); struct net_device *ndev = priv->ndev; + destroy_workqueue(priv->wq); can_rx_offload_del(&priv->offload); mcp251xfd_unregister(priv); spi->max_speed_hz = priv->spi_max_speed_hz_orig; diff --git a/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c b/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c index 160528d3cc26..1e7ddf316643 100644 --- a/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c +++ b/drivers/net/can/spi/mcp251xfd/mcp251xfd-tx.c @@ -131,6 +131,41 @@ mcp251xfd_tx_obj_from_skb(const struct mcp251xfd_priv *priv, tx_obj->xfer[0].len = len; } +static void mcp251xfd_tx_failure_drop(const struct mcp251xfd_priv *priv, + struct mcp251xfd_tx_ring *tx_ring, + int err) +{ + struct net_device *ndev = priv->ndev; + struct net_device_stats *stats = &ndev->stats; + unsigned int frame_len = 0; + u8 tx_head;
+ tx_ring->head--; + stats->tx_dropped++; + tx_head = mcp251xfd_get_tx_head(tx_ring); + can_free_echo_skb(ndev, tx_head, &frame_len); + netdev_completed_queue(ndev, 1, frame_len); + netif_wake_queue(ndev);
+ if (net_ratelimit()) + netdev_err(priv->ndev, "ERROR in %s: %d\n", __func__, err); +}
+void mcp251xfd_tx_obj_write_sync(struct work_struct *ws) +{ + struct mcp251xfd_priv *priv = container_of(ws, struct mcp251xfd_priv, + tx_work); + struct mcp251xfd_tx_obj *tx_obj = priv->tx_work_obj; + struct mcp251xfd_tx_ring *tx_ring = priv->tx; + int err;
+ err = spi_sync(priv->spi, &tx_obj->msg); + if (err) + mcp251xfd_tx_failure_drop(priv, tx_ring, err);
+ priv->tx_work_obj = NULL;
Race condition:
- after spi_sync() the CAN frame is send
- after the TX complete IRQ the TX queue is restarted
- the xmit handler might get BUSY
- fill the tx_work_obj again
+}
static int mcp251xfd_tx_obj_write(const struct mcp251xfd_priv *priv, struct mcp251xfd_tx_obj *tx_obj) { @@ -175,7 +210,7 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, if (can_dev_dropped_skb(ndev, skb)) return NETDEV_TX_OK; - if (mcp251xfd_tx_busy(priv, tx_ring)) + if (mcp251xfd_tx_busy(priv, tx_ring) || priv->tx_work_obj)
This should not happen, but better save than sorry.
As there is the race condition you mentioned above, on this condition: priv->tx_work_obj = tx_obj --> xmit will return NETDEV_TX_BUSY
or
priv->tx_work_obj = NULL --> It goes through the rest of the code or the workqueue may sleep after setting tx_work_obj to NULL. Should I use work_busy() here instead or do you have another suggestion?
Everything else is clear to me and I'm addressing it for the v6.
Best regards, Vitor Soares
On 14.05.2024 15:34:01, Vitor Soares wrote:
+void mcp251xfd_tx_obj_write_sync(struct work_struct *ws) +{
struct mcp251xfd_priv *priv = container_of(ws, structmcp251xfd_priv,
tx_work);struct mcp251xfd_tx_obj *tx_obj = priv->tx_work_obj;struct mcp251xfd_tx_ring *tx_ring = priv->tx;int err;err = spi_sync(priv->spi, &tx_obj->msg);if (err)mcp251xfd_tx_failure_drop(priv, tx_ring, err);priv->tx_work_obj = NULL;Race condition:
- after spi_sync() the CAN frame is send
- after the TX complete IRQ the TX queue is restarted
- the xmit handler might get BUSY
- fill the tx_work_obj again
You can avoid the race condition by moving "priv->tx_work_obj = NULL;" in front of the "spi_sync();". Right?
+}
static int mcp251xfd_tx_obj_write(const struct mcp251xfd_priv *priv, struct mcp251xfd_tx_obj *tx_obj) { @@ -175,7 +210,7 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, if (can_dev_dropped_skb(ndev, skb)) return NETDEV_TX_OK;
if (mcp251xfd_tx_busy(priv, tx_ring))
if (mcp251xfd_tx_busy(priv, tx_ring) || priv->tx_work_obj)This should not happen, but better save than sorry.
As there is the race condition you mentioned above, on this condition: priv->tx_work_obj = tx_obj --> xmit will return NETDEV_TX_BUSY
or
priv->tx_work_obj = NULL --> It goes through the rest of the code or the workqueue may sleep after setting tx_work_obj to NULL. Should I use work_busy() here instead or do you have another suggestion?
Yes, introduce mcp251xfd_work_busy().
I'm not sure what happens if the xmit is called between the "priv->tx_work_obj = NULL" and the end of the work. Will queue_work() return false, as the queue is still running?
Everything else is clear to me and I'm addressing it for the v6.
Marc
On Wed, 2024-05-15 at 13:12 +0200, Marc Kleine-Budde wrote:
On 14.05.2024 15:34:01, Vitor Soares wrote:
+void mcp251xfd_tx_obj_write_sync(struct work_struct *ws) +{ + struct mcp251xfd_priv *priv = container_of(ws, struct mcp251xfd_priv, + tx_work); + struct mcp251xfd_tx_obj *tx_obj = priv->tx_work_obj; + struct mcp251xfd_tx_ring *tx_ring = priv->tx; + int err;
+ err = spi_sync(priv->spi, &tx_obj->msg); + if (err) + mcp251xfd_tx_failure_drop(priv, tx_ring, err);
+ priv->tx_work_obj = NULL;
Race condition:
- after spi_sync() the CAN frame is send
- after the TX complete IRQ the TX queue is restarted
- the xmit handler might get BUSY
- fill the tx_work_obj again
You can avoid the race condition by moving "priv->tx_work_obj = NULL;" in front of the "spi_sync();". Right?
+}
static int mcp251xfd_tx_obj_write(const struct mcp251xfd_priv *priv, struct mcp251xfd_tx_obj *tx_obj) { @@ -175,7 +210,7 @@ netdev_tx_t mcp251xfd_start_xmit(struct sk_buff *skb, if (can_dev_dropped_skb(ndev, skb)) return NETDEV_TX_OK; - if (mcp251xfd_tx_busy(priv, tx_ring)) + if (mcp251xfd_tx_busy(priv, tx_ring) || priv->tx_work_obj)
This should not happen, but better save than sorry.
As there is the race condition you mentioned above, on this condition: priv->tx_work_obj = tx_obj --> xmit will return NETDEV_TX_BUSY
or
priv->tx_work_obj = NULL --> It goes through the rest of the code or the workqueue may sleep after setting tx_work_obj to NULL. Should I use work_busy() here instead or do you have another suggestion?
Yes, introduce mcp251xfd_work_busy().
I'll implement it.
I'm not sure what happens if the xmit is called between the "priv->tx_work_obj = NULL" and the end of the work. Will queue_work() return false, as the queue is still running?
From the test I did so far, my understanding is the following:
If mcp251xfd_tx_obj_write doesn't fail, everything is OK.
if mcp251xfd_tx_obj_write fails with EBUSY - stop netif queue - fill the tx_work_obj - start worker
queue_work() doesn't return false even when work_busy() = true. - xmit handler return, and wait netif_wake_queue() - the work handler waits until the previous job gets done before starting the next one. - after the TX completes IRQ, the TX queue is restarted
If the TX queue is restarted immediately after queue_work(), tx_work_obj is filled, making the xmit handler return NETDEV_TX_BUSY.
The tests were done with a delay after priv->tx_work_obj = NULL.
Best regards, Vitor Soares
linux-stable-mirror@lists.linaro.org
-
Marc Kleine-Budde -
Vitor Soares