[PATCH net v2 1/2] net: apple: mace: don't call dev_kfree_skb() under spin_lock_irqsave()

[PATCH net v2 1/2] net: apple: mace: don't call dev_kfree_skb() under spin_lock_irqsave()

[Yang Yingliang]

It is not allowed to call consume_skb() from hardware interrupt context
or with interrupts being disabled. So replace dev_kfree_skb() with
dev_consume_skb_irq() under spin_lock_irqsave().

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
---
v1 -> v2:
  Add a fix tag.
---
 drivers/net/ethernet/apple/mace.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/apple/mace.c b/drivers/net/ethernet/apple/mace.c
index d0a771b65e88..77b4ed05140b 100644
--- a/drivers/net/ethernet/apple/mace.c
+++ b/drivers/net/ethernet/apple/mace.c
@@ -846,7 +846,7 @@ static void mace_tx_timeout(struct timer_list *t)
     if (mp->tx_bad_runt) {
 	mp->tx_bad_runt = 0;
     } else if (i != mp->tx_fill) {
-	dev_kfree_skb(mp->tx_bufs[i]);
+	dev_consume_skb_irq(mp->tx_bufs[i]);
 	if (++i >= N_TX_RING)
 	    i = 0;
 	mp->tx_empty = i;
-- 
2.25.1

[PATCH net v2 2/2] net: apple: bmac: don't call dev_kfree_skb() under spin_lock_irqsave()

[Yang Yingliang]

It is not allowed to call consume_skb() from hardware interrupt context
or with interrupts being disabled. So replace dev_kfree_skb() with
dev_consume_skb_irq() under spin_lock_irqsave().

Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
---
v1 -> v2:
  Add a fix tag.
---
 drivers/net/ethernet/apple/bmac.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/ethernet/apple/bmac.c b/drivers/net/ethernet/apple/bmac.c
index 334de0d93c89..3e6a629bcf21 100644
--- a/drivers/net/ethernet/apple/bmac.c
+++ b/drivers/net/ethernet/apple/bmac.c
@@ -1510,7 +1510,7 @@ static void bmac_tx_timeout(struct timer_list *t)
 	i = bp->tx_empty;
 	++dev->stats.tx_errors;
 	if (i != bp->tx_fill) {
-		dev_kfree_skb(bp->tx_bufs[i]);
+		dev_consume_skb_irq(bp->tx_bufs[i]);
 		bp->tx_bufs[i] = NULL;
 		if (++i >= N_TX_RING) i = 0;
 		bp->tx_empty = i;
-- 
2.25.1

Re: [PATCH net v2 1/2] net: apple: mace: don't call dev_kfree_skb() under spin_lock_irqsave()

[Leon Romanovsky]

On Wed, Dec 07, 2022 at 09:29:58AM +0800, Yang Yingliang wrote:
> It is not allowed to call consume_skb() from hardware interrupt context
> or with interrupts being disabled. So replace dev_kfree_skb() with
> dev_consume_skb_irq() under spin_lock_irqsave().
> 
> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
> ---
> v1 -> v2:
>   Add a fix tag.
> ---
>  drivers/net/ethernet/apple/mace.c | 2 +-
>  1 file changed, 1 insertion(+), 1 deletion(-)
> 
> diff --git a/drivers/net/ethernet/apple/mace.c b/drivers/net/ethernet/apple/mace.c
> index d0a771b65e88..77b4ed05140b 100644
> --- a/drivers/net/ethernet/apple/mace.c
> +++ b/drivers/net/ethernet/apple/mace.c
> @@ -846,7 +846,7 @@ static void mace_tx_timeout(struct timer_list *t)
>      if (mp->tx_bad_runt) {
>  	mp->tx_bad_runt = 0;
>      } else if (i != mp->tx_fill) {
> -	dev_kfree_skb(mp->tx_bufs[i]);
> +	dev_consume_skb_irq(mp->tx_bufs[i]);

Same question, why did you chose dev_consume_skb_irq and not dev_kfree_skb_irq?

Thanks

>  	if (++i >= N_TX_RING)
>  	    i = 0;
>  	mp->tx_empty = i;
> -- 
> 2.25.1
> 

Re: [PATCH net v2 1/2] net: apple: mace: don't call dev_kfree_skb() under spin_lock_irqsave()

[Yang Yingliang]

On 2022/12/8 15:58, Leon Romanovsky wrote:
> On Wed, Dec 07, 2022 at 09:29:58AM +0800, Yang Yingliang wrote:
>> It is not allowed to call consume_skb() from hardware interrupt context
>> or with interrupts being disabled. So replace dev_kfree_skb() with
>> dev_consume_skb_irq() under spin_lock_irqsave().
>>
>> Fixes: 1da177e4c3f4 ("Linux-2.6.12-rc2")
>> Signed-off-by: Yang Yingliang <yangyingliang@huawei.com>
>> ---
>> v1 -> v2:
>>    Add a fix tag.
>> ---
>>   drivers/net/ethernet/apple/mace.c | 2 +-
>>   1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/net/ethernet/apple/mace.c b/drivers/net/ethernet/apple/mace.c
>> index d0a771b65e88..77b4ed05140b 100644
>> --- a/drivers/net/ethernet/apple/mace.c
>> +++ b/drivers/net/ethernet/apple/mace.c
>> @@ -846,7 +846,7 @@ static void mace_tx_timeout(struct timer_list *t)
>>       if (mp->tx_bad_runt) {
>>   	mp->tx_bad_runt = 0;
>>       } else if (i != mp->tx_fill) {
>> -	dev_kfree_skb(mp->tx_bufs[i]);
>> +	dev_consume_skb_irq(mp->tx_bufs[i]);
> Same question, why did you chose dev_consume_skb_irq and not dev_kfree_skb_irq?
I chose dev_consume_skb_irq(), because dev_kfree_skb() is consume_skb().

Thanks,
Yang
>
> Thanks
>
>>   	if (++i >= N_TX_RING)
>>   	    i = 0;
>>   	mp->tx_empty = i;
>> -- 
>> 2.25.1
>>
> .

Re: [PATCH net v2 1/2] net: apple: mace: don't call dev_kfree_skb() under spin_lock_irqsave()

[Jakub Kicinski]

On Thu, 8 Dec 2022 16:39:03 +0800 Yang Yingliang wrote:
> >> @@ -846,7 +846,7 @@ static void mace_tx_timeout(struct timer_list *t)
> >>       if (mp->tx_bad_runt) {
> >>   	mp->tx_bad_runt = 0;
> >>       } else if (i != mp->tx_fill) {
> >> -	dev_kfree_skb(mp->tx_bufs[i]);
> >> +	dev_consume_skb_irq(mp->tx_bufs[i]);  
> > Same question, why did you chose dev_consume_skb_irq and not dev_kfree_skb_irq?  
> I chose dev_consume_skb_irq(), because dev_kfree_skb() is consume_skb().

kfree_skb() should be used on error paths, when packet is dropped.
consume_skb() on normal paths, when packet left the system successfully.

dev_kfree* helpers probably default to consume_skb() to avoid spamming
drop monitor, but switching to dev_consume explicitly is not right.