2023-08-11 13:05:57

by marshall.shao

Subject: [PATCH] Fix kernel panic issue after removing igb driver

From: Marshall Shao <[email protected]>

This patch fixes a kernel panic issue after removing the igb driver
from the usermode.

A delayed work will be schedule in igb_ptp_init(),

if (adapter->ptp_flags & IGB_PTP_OVERFLOW_CHECK)
INIT_DELAYED_WORK(&adapter->ptp_overflow_work,
igb_ptp_overflow_check);

If CONFIG_PTP_1588_CLOCK is not enabled, the delayed work cannot be
cancelled when igb_ptp_suspend is called.

Signed-off-by: Marshall Shao <[email protected]>
---
drivers/net/ethernet/intel/igb/igb_ptp.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/net/ethernet/intel/igb/igb_ptp.c b/drivers/net/ethernet/intel/igb/igb_ptp.c
index 405886ee5261..b21822ea1c7d 100644
--- a/drivers/net/ethernet/intel/igb/igb_ptp.c
+++ b/drivers/net/ethernet/intel/igb/igb_ptp.c
@@ -1435,12 +1435,12 @@ void igb_ptp_sdp_init(struct igb_adapter *adapter)
*/
void igb_ptp_suspend(struct igb_adapter *adapter)
{
- if (!(adapter->ptp_flags & IGB_PTP_ENABLED))
- return;
-
if (adapter->ptp_flags & IGB_PTP_OVERFLOW_CHECK)
cancel_delayed_work_sync(&adapter->ptp_overflow_work);

+ if (!(adapter->ptp_flags & IGB_PTP_ENABLED))
+ return;
+
cancel_work_sync(&adapter->ptp_tx_work);
if (adapter->ptp_tx_skb) {
dev_kfree_skb_any(adapter->ptp_tx_skb);
--
2.34.1

2023-08-11 16:46:20

by Vadim Fedorenko

Subject: Re: [PATCH] Fix kernel panic issue after removing igb driver

On 11/08/2023 13:02, [email protected] wrote:
> From: Marshall Shao <[email protected]>
>
> This patch fixes a kernel panic issue after removing the igb driver
> from the usermode.
>
> A delayed work will be schedule in igb_ptp_init(),
>
> if (adapter->ptp_flags & IGB_PTP_OVERFLOW_CHECK)
> INIT_DELAYED_WORK(&adapter->ptp_overflow_work,
> igb_ptp_overflow_check);
>
> If CONFIG_PTP_1588_CLOCK is not enabled, the delayed work cannot be
> cancelled when igb_ptp_suspend is called.

But should this work be inited in case there is no PTP support in the
kernel? Maybe it's better to avoid any driver threads in this case and
move the initialization to the if block where we know that PHC device
is created and working?

>
> Signed-off-by: Marshall Shao <[email protected]>
> ---
> drivers/net/ethernet/intel/igb/igb_ptp.c | 6 +++---
> 1 file changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/net/ethernet/intel/igb/igb_ptp.c b/drivers/net/ethernet/intel/igb/igb_ptp.c
> index 405886ee5261..b21822ea1c7d 100644
> --- a/drivers/net/ethernet/intel/igb/igb_ptp.c
> +++ b/drivers/net/ethernet/intel/igb/igb_ptp.c
> @@ -1435,12 +1435,12 @@ void igb_ptp_sdp_init(struct igb_adapter *adapter)
> */
> void igb_ptp_suspend(struct igb_adapter *adapter)
> {
> - if (!(adapter->ptp_flags & IGB_PTP_ENABLED))
> - return;
> -
> if (adapter->ptp_flags & IGB_PTP_OVERFLOW_CHECK)
> cancel_delayed_work_sync(&adapter->ptp_overflow_work);
>
> + if (!(adapter->ptp_flags & IGB_PTP_ENABLED))
> + return;
> +
> cancel_work_sync(&adapter->ptp_tx_work);
> if (adapter->ptp_tx_skb) {
> dev_kfree_skb_any(adapter->ptp_tx_skb);