[PATCH] smc91x: fix broken irq-context in PREEMPT_RT

[PATCH] smc91x: fix broken irq-context in PREEMPT_RT

[Yeoreum Yun]

When smc91x.c is built with PREEMPT_RT, the following splat occurs
in arm FVP_RevC:

[   13.055000] smc91x LNRO0003:00 eth0: link up, 10Mbps, half-duplex, lpa 0x0000
[   13.062137] BUG: workqueue leaked atomic, lock or RCU: kworker/2:1[106]
[   13.062137]      preempt=0x00000000 lock=0->0 RCU=0->1 workfn=mld_ifc_work
[   13.062266] C
** replaying previous printk message **
[   13.062266] CPU: 2 UID: 0 PID: 106 Comm: kworker/2:1 Not tainted 6.18.0-dirty #179 PREEMPT_{RT,(full)}
[   13.062353] Hardware name:  , BIOS
[   13.062382] Workqueue: mld mld_ifc_work
[   13.062469] Call trace:
[   13.062494]  show_stack+0x24/0x40 (C)
[   13.062602]  __dump_stack+0x28/0x48
[   13.062710]  dump_stack_lvl+0x7c/0xb0
[   13.062818]  dump_stack+0x18/0x34
[   13.062926]  process_scheduled_works+0x294/0x450
[   13.063043]  worker_thread+0x260/0x3d8
[   13.063124]  kthread+0x1c4/0x228
[   13.063235]  ret_from_fork+0x10/0x20

This happens because smc_special_trylock() disables IRQs even on PREEMPT_RT,
but smc_special_unlock() does not restore IRQs on PREEMPT_RT.
The reason is that smc_special_unlock() calls spin_unlock_irqrestore(),
and rcu_read_unlock_bh() in __dev_queue_xmit() cannot invoke
rcu_read_unlock() through __local_bh_enable_ip() when current->softirq_disable_cnt becomes zero.

To address this issue, replace smc_special_trylock() with spin_trylock_irqsave().

Fixes: 8ff499e43c53 ("smc91x: let smc91x work well under netpoll")
Signed-off-by: Yeoreum Yun <yeoreum.yun@arm.com>
---
This patch based on v6.18
---
 drivers/net/ethernet/smsc/smc91x.c | 11 ++---------
 1 file changed, 2 insertions(+), 9 deletions(-)

diff --git a/drivers/net/ethernet/smsc/smc91x.c b/drivers/net/ethernet/smsc/smc91x.c
index 9d1a83a5fa7e..b7fef6ce8615 100644
--- a/drivers/net/ethernet/smsc/smc91x.c
+++ b/drivers/net/ethernet/smsc/smc91x.c
@@ -516,15 +516,7 @@ static inline void  smc_rcv(struct net_device *dev)
  * any other concurrent access and C would always interrupt B. But life
  * isn't that easy in a SMP world...
  */
-#define smc_special_trylock(lock, flags)				\
-({									\
-	int __ret;							\
-	local_irq_save(flags);						\
-	__ret = spin_trylock(lock);					\
-	if (!__ret)							\
-		local_irq_restore(flags);				\
-	__ret;								\
-})
+#define smc_special_trylock(lock, flags)	spin_trylock_irqsave(lock, flags)
 #define smc_special_lock(lock, flags)		spin_lock_irqsave(lock, flags)
 #define smc_special_unlock(lock, flags) 	spin_unlock_irqrestore(lock, flags)
 #else
@@ -658,6 +650,7 @@ smc_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		return NETDEV_TX_OK;
 	}

+	pr_err("[LEVI:%s:%d] before xmit_one %d\n", __func__, __LINE__, irqs_disabled());
 	smc_special_lock(&lp->lock, flags);

 	/* now, try to allocate the memory */
--
LEVI:{C3F47F37-75D8-414A-A8BA-3980EC8A46D7}

Re: [PATCH] smc91x: fix broken irq-context in PREEMPT_RT

[Yeoreum Yun]

> When smc91x.c is built with PREEMPT_RT, the following splat occurs
> in arm FVP_RevC:
>
> [   13.055000] smc91x LNRO0003:00 eth0: link up, 10Mbps, half-duplex, lpa 0x0000
> [   13.062137] BUG: workqueue leaked atomic, lock or RCU: kworker/2:1[106]
> [   13.062137]      preempt=0x00000000 lock=0->0 RCU=0->1 workfn=mld_ifc_work
> [   13.062266] C
> ** replaying previous printk message **
> [   13.062266] CPU: 2 UID: 0 PID: 106 Comm: kworker/2:1 Not tainted 6.18.0-dirty #179 PREEMPT_{RT,(full)}
> [   13.062353] Hardware name:  , BIOS
> [   13.062382] Workqueue: mld mld_ifc_work
> [   13.062469] Call trace:
> [   13.062494]  show_stack+0x24/0x40 (C)
> [   13.062602]  __dump_stack+0x28/0x48
> [   13.062710]  dump_stack_lvl+0x7c/0xb0
> [   13.062818]  dump_stack+0x18/0x34
> [   13.062926]  process_scheduled_works+0x294/0x450
> [   13.063043]  worker_thread+0x260/0x3d8
> [   13.063124]  kthread+0x1c4/0x228
> [   13.063235]  ret_from_fork+0x10/0x20
>
> This happens because smc_special_trylock() disables IRQs even on PREEMPT_RT,
> but smc_special_unlock() does not restore IRQs on PREEMPT_RT.
> The reason is that smc_special_unlock() calls spin_unlock_irqrestore(),
> and rcu_read_unlock_bh() in __dev_queue_xmit() cannot invoke
> rcu_read_unlock() through __local_bh_enable_ip() when current->softirq_disable_cnt becomes zero.
>
> To address this issue, replace smc_special_trylock() with spin_trylock_irqsave().
>
> Fixes: 8ff499e43c53 ("smc91x: let smc91x work well under netpoll")
> Signed-off-by: Yeoreum Yun <yeoreum.yun@arm.com>
> ---
> This patch based on v6.18
> ---
>  drivers/net/ethernet/smsc/smc91x.c | 11 ++---------
>  1 file changed, 2 insertions(+), 9 deletions(-)
>
> diff --git a/drivers/net/ethernet/smsc/smc91x.c b/drivers/net/ethernet/smsc/smc91x.c
> index 9d1a83a5fa7e..b7fef6ce8615 100644
> --- a/drivers/net/ethernet/smsc/smc91x.c
> +++ b/drivers/net/ethernet/smsc/smc91x.c
> @@ -516,15 +516,7 @@ static inline void  smc_rcv(struct net_device *dev)
>   * any other concurrent access and C would always interrupt B. But life
>   * isn't that easy in a SMP world...
>   */
> -#define smc_special_trylock(lock, flags)				\
> -({									\
> -	int __ret;							\
> -	local_irq_save(flags);						\
> -	__ret = spin_trylock(lock);					\
> -	if (!__ret)							\
> -		local_irq_restore(flags);				\
> -	__ret;								\
> -})
> +#define smc_special_trylock(lock, flags)	spin_trylock_irqsave(lock, flags)
>  #define smc_special_lock(lock, flags)		spin_lock_irqsave(lock, flags)
>  #define smc_special_unlock(lock, flags) 	spin_unlock_irqrestore(lock, flags)
>  #else
> @@ -658,6 +650,7 @@ smc_hard_start_xmit(struct sk_buff *skb, struct net_device *dev)
>  		return NETDEV_TX_OK;
>  	}
>
> +	pr_err("[LEVI:%s:%d] before xmit_one %d\n", __func__, __LINE__, irqs_disabled());
>  	smc_special_lock(&lp->lock, flags);

Ah sorry. I'll send again with removing debug log...

>
>  	/* now, try to allocate the memory */
> --
> LEVI:{C3F47F37-75D8-414A-A8BA-3980EC8A46D7}
>

--
Sincerely,
Yeoreum Yun