[PATCH] net: broadcom: bcm4908enet: add platform_get_irq_byname error checking

[PATCH] net: broadcom: bcm4908enet: add platform_get_irq_byname error checking

[Yu Zhe]

The platform_get_irq_byname() function returns negative error codes on error,
check it.

Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
---
 drivers/net/ethernet/broadcom/bcm4908_enet.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/drivers/net/ethernet/broadcom/bcm4908_enet.c b/drivers/net/ethernet/broadcom/bcm4908_enet.c
index c131d8118489..d985056db6c2 100644
--- a/drivers/net/ethernet/broadcom/bcm4908_enet.c
+++ b/drivers/net/ethernet/broadcom/bcm4908_enet.c
@@ -705,6 +705,8 @@ static int bcm4908_enet_probe(struct platform_device *pdev)
 		return netdev->irq;
 
 	enet->irq_tx = platform_get_irq_byname(pdev, "tx");
+	if (enet->irq_tx < 0)
+		return enet->irq_tx;
 
 	err = dma_set_coherent_mask(dev, DMA_BIT_MASK(32));
 	if (err)
-- 
2.11.0

Re: [PATCH] net: broadcom: bcm4908enet: add platform_get_irq_byname error checking

[Dan Carpenter]

On Fri, Sep 09, 2022 at 02:25:45PM +0800, Yu Zhe wrote:
> The platform_get_irq_byname() function returns negative error codes on error,
> check it.
> 
> Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
> ---
>  drivers/net/ethernet/broadcom/bcm4908_enet.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/net/ethernet/broadcom/bcm4908_enet.c b/drivers/net/ethernet/broadcom/bcm4908_enet.c
> index c131d8118489..d985056db6c2 100644
> --- a/drivers/net/ethernet/broadcom/bcm4908_enet.c
> +++ b/drivers/net/ethernet/broadcom/bcm4908_enet.c
> @@ -705,6 +705,8 @@ static int bcm4908_enet_probe(struct platform_device *pdev)
>  		return netdev->irq;
>  
>  	enet->irq_tx = platform_get_irq_byname(pdev, "tx");
> +	if (enet->irq_tx < 0)
> +		return enet->irq_tx;
>  

If you read the driver, then you will see that this is deliberate.
Search for irq_tx and read the comments.  I'm not a subsystem expert so
I don't know if this an ideal way to write the code, but it's done
deliberately so please don't change it unless you can test it.

regards,
dan carpenter

Re: [PATCH] net: broadcom: bcm4908enet: add platform_get_irq_byname error checking

[Florian Fainelli]

On 9/12/22 04:29, Dan Carpenter wrote:
> On Fri, Sep 09, 2022 at 02:25:45PM +0800, Yu Zhe wrote:
>> The platform_get_irq_byname() function returns negative error codes on error,
>> check it.
>>
>> Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
>> ---
>>   drivers/net/ethernet/broadcom/bcm4908_enet.c | 2 ++
>>   1 file changed, 2 insertions(+)
>>
>> diff --git a/drivers/net/ethernet/broadcom/bcm4908_enet.c b/drivers/net/ethernet/broadcom/bcm4908_enet.c
>> index c131d8118489..d985056db6c2 100644
>> --- a/drivers/net/ethernet/broadcom/bcm4908_enet.c
>> +++ b/drivers/net/ethernet/broadcom/bcm4908_enet.c
>> @@ -705,6 +705,8 @@ static int bcm4908_enet_probe(struct platform_device *pdev)
>>   		return netdev->irq;
>>   
>>   	enet->irq_tx = platform_get_irq_byname(pdev, "tx");
>> +	if (enet->irq_tx < 0)
>> +		return enet->irq_tx;
>>   
> 
> If you read the driver, then you will see that this is deliberate.
> Search for irq_tx and read the comments.  I'm not a subsystem expert so
> I don't know if this an ideal way to write the code, but it's done
> deliberately so please don't change it unless you can test it.

Yup, the transmit interrupt is deemed optional, or at least was up to 
some point during the driver development. There is however a worthy bug 
you could fix:

   static int bcm4908_enet_stop(struct net_device *netdev)
   {
           struct bcm4908_enet *enet = netdev_priv(netdev);
           struct bcm4908_enet_dma_ring *tx_ring = &enet->tx_ring;
           struct bcm4908_enet_dma_ring *rx_ring = &enet->rx_ring;

           netif_stop_queue(netdev);
           netif_carrier_off(netdev);
           napi_disable(&rx_ring->napi);
           napi_disable(&tx_ring->napi);

           bcm4908_enet_dma_rx_ring_disable(enet, &enet->rx_ring);
           bcm4908_enet_dma_tx_ring_disable(enet, &enet->tx_ring);

           bcm4908_enet_dma_uninit(enet);

           free_irq(enet->irq_tx, enet);

We might attempt to free an invalid interrupt here ^^

           free_irq(enet->netdev->irq, enet);

           return 0;
-- 
Florian

Re: [PATCH] net: broadcom: bcm4908enet: add platform_get_irq_byname error checking

[Dan Carpenter]

On Mon, Sep 12, 2022 at 11:14:42AM -0700, Florian Fainelli wrote:
> On 9/12/22 04:29, Dan Carpenter wrote:
> > On Fri, Sep 09, 2022 at 02:25:45PM +0800, Yu Zhe wrote:
> > > The platform_get_irq_byname() function returns negative error codes on error,
> > > check it.
> > > 
> > > Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
> > > ---
> > >   drivers/net/ethernet/broadcom/bcm4908_enet.c | 2 ++
> > >   1 file changed, 2 insertions(+)
> > > 
> > > diff --git a/drivers/net/ethernet/broadcom/bcm4908_enet.c b/drivers/net/ethernet/broadcom/bcm4908_enet.c
> > > index c131d8118489..d985056db6c2 100644
> > > --- a/drivers/net/ethernet/broadcom/bcm4908_enet.c
> > > +++ b/drivers/net/ethernet/broadcom/bcm4908_enet.c
> > > @@ -705,6 +705,8 @@ static int bcm4908_enet_probe(struct platform_device *pdev)
> > >   		return netdev->irq;
> > >   	enet->irq_tx = platform_get_irq_byname(pdev, "tx");
> > > +	if (enet->irq_tx < 0)
> > > +		return enet->irq_tx;
> > 
> > If you read the driver, then you will see that this is deliberate.
> > Search for irq_tx and read the comments.  I'm not a subsystem expert so
> > I don't know if this an ideal way to write the code, but it's done
> > deliberately so please don't change it unless you can test it.
> 
> Yup, the transmit interrupt is deemed optional, or at least was up to some
> point during the driver development. There is however a worthy bug you could
> fix:
> 
>   static int bcm4908_enet_stop(struct net_device *netdev)
>   {
>           struct bcm4908_enet *enet = netdev_priv(netdev);
>           struct bcm4908_enet_dma_ring *tx_ring = &enet->tx_ring;
>           struct bcm4908_enet_dma_ring *rx_ring = &enet->rx_ring;
> 
>           netif_stop_queue(netdev);
>           netif_carrier_off(netdev);
>           napi_disable(&rx_ring->napi);
>           napi_disable(&tx_ring->napi);
> 
>           bcm4908_enet_dma_rx_ring_disable(enet, &enet->rx_ring);
>           bcm4908_enet_dma_tx_ring_disable(enet, &enet->tx_ring);
> 
>           bcm4908_enet_dma_uninit(enet);
> 
>           free_irq(enet->irq_tx, enet);
> 
> We might attempt to free an invalid interrupt here ^^

Freeing a negative IRQ does not cause an issue.  The irq_to_desc()
function will return NULL so it becomes a no-op.

It's ugly code because you have to read in a couple different files to
verify that it works...

regards,
dan carpenter

Re: [PATCH] net: broadcom: bcm4908enet: add platform_get_irq_byname error checking

[Dan Carpenter]

On Mon, Sep 12, 2022 at 02:29:17PM +0300, Dan Carpenter wrote:
> On Fri, Sep 09, 2022 at 02:25:45PM +0800, Yu Zhe wrote:
> > The platform_get_irq_byname() function returns negative error codes on error,
> > check it.
> > 
> > Signed-off-by: Yu Zhe <yuzhe@nfschina.com>
> > ---
> >  drivers/net/ethernet/broadcom/bcm4908_enet.c | 2 ++
> >  1 file changed, 2 insertions(+)
> > 
> > diff --git a/drivers/net/ethernet/broadcom/bcm4908_enet.c b/drivers/net/ethernet/broadcom/bcm4908_enet.c
> > index c131d8118489..d985056db6c2 100644
> > --- a/drivers/net/ethernet/broadcom/bcm4908_enet.c
> > +++ b/drivers/net/ethernet/broadcom/bcm4908_enet.c
> > @@ -705,6 +705,8 @@ static int bcm4908_enet_probe(struct platform_device *pdev)
> >  		return netdev->irq;
> >  
> >  	enet->irq_tx = platform_get_irq_byname(pdev, "tx");
> > +	if (enet->irq_tx < 0)
> > +		return enet->irq_tx;
> >  
> 
> If you read the driver, then you will see that this is deliberate.
> Search for irq_tx and read the comments.  I'm not a subsystem expert so
> I don't know if this an ideal way to write the code, but it's done
> deliberately so please don't change it unless you can test it.

Btw, my comment of "don't change it unless you can test it" only applies
to this specific IRQ...  We constantly apply untested patches.  I have
submitted my share of buggy patches.  I don't want to discourage you
from your static analysis work.

regards,
dan carpenter