[Intel-wired-lan] [PATCH] igb: Use a sperate mutex insead of rtnl_lock()

Paul Menzel pmenzel at molgen.mpg.de
Thu Mar 26 11:16:10 UTC 2020


Dear Kai-Heng,


Thank you.

There is a small typo in the commit summary: s*epa*rate.

Am 26.03.20 um 11:39 schrieb Kai-Heng Feng:
> Commit 9474933caf21 ("igb: close/suspend race in netif_device_detach")
> fixed race condition between close and power management ops by using
> rtnl_lock().
> 
> This fix is a preparation for next patch, to prevent a dead lock under
> rtnl_lock() when calling runtime resume routine.

Do you refer with *this fix* to the referenced commit? Or do you mean 
the patch you just sent?

How can the issue be reproduced?

> However, we can't use device_lock() in igb_close() because when module
> is getting removed, the lock is already held for igb_remove(), and
> igb_close() gets called during unregistering the netdev, hence causing a
> deadlock. So let's introduce a new mutex so we don't cause a deadlock
> with driver core or netdev core.

Is there a bug report with more details?

If this fixes a regression, please add the appropriate `Fixes:` tag.

> Signed-off-by: Kai-Heng Feng <kai.heng.feng at canonical.com>
> ---
>   drivers/net/ethernet/intel/igb/igb_main.c | 19 +++++++++++++------
>   1 file changed, 13 insertions(+), 6 deletions(-)
> 
> diff --git a/drivers/net/ethernet/intel/igb/igb_main.c b/drivers/net/ethernet/intel/igb/igb_main.c
> index b46bff8fe056..dc7ed5dd216b 100644
> --- a/drivers/net/ethernet/intel/igb/igb_main.c
> +++ b/drivers/net/ethernet/intel/igb/igb_main.c
> @@ -288,6 +288,8 @@ static const struct igb_reg_info igb_reg_info_tbl[] = {
>   	{}
>   };
>   
> +static DEFINE_MUTEX(igb_mutex);
> +
>   /* igb_regdump - register printout routine */
>   static void igb_regdump(struct e1000_hw *hw, struct igb_reg_info *reginfo)
>   {
> @@ -4026,9 +4028,14 @@ static int __igb_close(struct net_device *netdev, bool suspending)
>   
>   int igb_close(struct net_device *netdev)
>   {
> +	int err = 0;
> +
> +	mutex_lock(&igb_mutex);
>   	if (netif_device_present(netdev) || netdev->dismantle)
> -		return __igb_close(netdev, false);
> -	return 0;
> +		err = __igb_close(netdev, false);
> +	mutex_unlock(&igb_mutex);
> +
> +	return err;
>   }
>   
>   /**
> @@ -8760,7 +8767,7 @@ static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
>   	u32 wufc = runtime ? E1000_WUFC_LNKC : adapter->wol;
>   	bool wake;
>   
> -	rtnl_lock();
> +	mutex_lock(&igb_mutex);
>   	netif_device_detach(netdev);
>   
>   	if (netif_running(netdev))
> @@ -8769,7 +8776,7 @@ static int __igb_shutdown(struct pci_dev *pdev, bool *enable_wake,
>   	igb_ptp_suspend(adapter);
>   
>   	igb_clear_interrupt_scheme(adapter);
> -	rtnl_unlock();
> +	mutex_unlock(&igb_mutex);
>   
>   	status = rd32(E1000_STATUS);
>   	if (status & E1000_STATUS_LU)
> @@ -8897,13 +8904,13 @@ static int __maybe_unused igb_resume(struct device *dev)
>   
>   	wr32(E1000_WUS, ~0);
>   
> -	rtnl_lock();
> +	mutex_lock(&igb_mutex);
>   	if (!err && netif_running(netdev))
>   		err = __igb_open(netdev, true);
>   
>   	if (!err)
>   		netif_device_attach(netdev);
> -	rtnl_unlock();
> +	mutex_unlock(&igb_mutex);
>   
>   	return err;
>   }

The rest looks fine.


Kind regards,

Paul


More information about the Intel-wired-lan mailing list