[PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[Laurent Pinchart]

Memory allocated for the MDIO bus with the devm_kzalloc() API is
associated with the network device. While this will cause memory to be
freed at the right time, it doesn't allow allocating memory before the
network device is initialized.

Replace the network device with the parent platform device for memory
allocation to remove that dependency. This also improves consistency
with the other devm_* calls in the driver that all use the platform
device.

Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
---
 drivers/net/ethernet/renesas/sh_eth.c | 7 +++----
 1 file changed, 3 insertions(+), 4 deletions(-)

diff --git a/drivers/net/ethernet/renesas/sh_eth.c b/drivers/net/ethernet/renesas/sh_eth.c
index be7211d..b80abb6 100644
--- a/drivers/net/ethernet/renesas/sh_eth.c
+++ b/drivers/net/ethernet/renesas/sh_eth.c
@@ -2615,10 +2615,10 @@ static int sh_mdio_init(struct net_device *ndev, int id,
 	int ret, i;
 	struct bb_info *bitbang;
 	struct sh_eth_private *mdp = netdev_priv(ndev);
+	struct device *dev = &mdp->pdev->dev;
 
 	/* create bit control struct for PHY */
-	bitbang = devm_kzalloc(&ndev->dev, sizeof(struct bb_info),
-			       GFP_KERNEL);
+	bitbang = devm_kzalloc(dev, sizeof(struct bb_info), GFP_KERNEL);
 	if (!bitbang) {
 		ret = -ENOMEM;
 		goto out;
@@ -2647,8 +2647,7 @@ static int sh_mdio_init(struct net_device *ndev, int id,
 		 mdp->pdev->name, id);
 
 	/* PHY IRQ */
-	mdp->mii_bus->irq = devm_kzalloc(&ndev->dev,
-					 sizeof(int) * PHY_MAX_ADDR,
+	mdp->mii_bus->irq = devm_kzalloc(dev, sizeof(int) * PHY_MAX_ADDR,
 					 GFP_KERNEL);
 	if (!mdp->mii_bus->irq) {
 		ret = -ENOMEM;
-- 
1.8.3.2

Re: [PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[Sergei Shtylyov]

Hello.

On 03/19/2014 02:25 AM, Laurent Pinchart wrote:

> Memory allocated for the MDIO bus with the devm_kzalloc() API is
> associated with the network device. While this will cause memory to be
> freed at the right time, it doesn't allow allocating memory before the
> network device is initialized.

> Replace the network device with the parent platform device for memory
> allocation to remove that dependency. This also improves consistency
> with the other devm_* calls in the driver that all use the platform
> device.

> Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@ideasonboard.com>
> ---
>   drivers/net/ethernet/renesas/sh_eth.c | 7 +++----
>   1 file changed, 3 insertions(+), 4 deletions(-)

> diff --git a/drivers/net/ethernet/renesas/sh_eth.c b/drivers/net/ethernet/renesas/sh_eth.c
> index be7211d..b80abb6 100644
> --- a/drivers/net/ethernet/renesas/sh_eth.c
> +++ b/drivers/net/ethernet/renesas/sh_eth.c
> @@ -2615,10 +2615,10 @@ static int sh_mdio_init(struct net_device *ndev, int id,
>   	int ret, i;
>   	struct bb_info *bitbang;
>   	struct sh_eth_private *mdp = netdev_priv(ndev);

    I'd suggest to also declare:

	struct platform_device *pdev = &mdp->pdev;

so that you could simplify 'mdev->pdev->name' derefs in the sprintf() call and 
to get rid of the 'id' parameter to sh_mdio_init() by using 'pdev->id' 
directly in the same call.

WBR, Sergei

Re: [PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[Laurent Pinchart]

Hi Sergei,

On Wednesday 19 March 2014 20:19:51 Sergei Shtylyov wrote:
> On 03/19/2014 02:25 AM, Laurent Pinchart wrote:
> > Memory allocated for the MDIO bus with the devm_kzalloc() API is
> > associated with the network device. While this will cause memory to be
> > freed at the right time, it doesn't allow allocating memory before the
> > network device is initialized.
> > 
> > Replace the network device with the parent platform device for memory
> > allocation to remove that dependency. This also improves consistency
> > with the other devm_* calls in the driver that all use the platform
> > device.
> > 
> > Signed-off-by: Laurent Pinchart
> > <laurent.pinchart+renesas@ideasonboard.com>
> > ---
> > 
> >   drivers/net/ethernet/renesas/sh_eth.c | 7 +++----
> >   1 file changed, 3 insertions(+), 4 deletions(-)
> > 
> > diff --git a/drivers/net/ethernet/renesas/sh_eth.c
> > b/drivers/net/ethernet/renesas/sh_eth.c index be7211d..b80abb6 100644
> > --- a/drivers/net/ethernet/renesas/sh_eth.c
> > +++ b/drivers/net/ethernet/renesas/sh_eth.c
> > @@ -2615,10 +2615,10 @@ static int sh_mdio_init(struct net_device *ndev,
> > int id,> 
> >   	int ret, i;
> >   	struct bb_info *bitbang;
> >   	struct sh_eth_private *mdp = netdev_priv(ndev);
> 
> I'd suggest to also declare:
> 
> 	struct platform_device *pdev = &mdp->pdev;
> 
> so that you could simplify 'mdev->pdev->name' derefs in the sprintf() call
> and to get rid of the 'id' parameter to sh_mdio_init() by using 'pdev->id'
> directly in the same call.

Good idea. I'll do that in patch 2/5 for v2.

-- 
Regards,

Laurent Pinchart

Re: [PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[Laurent Pinchart]

On Thursday 20 March 2014 14:42:30 Laurent Pinchart wrote:
> On Wednesday 19 March 2014 20:19:51 Sergei Shtylyov wrote:
> > On 03/19/2014 02:25 AM, Laurent Pinchart wrote:
> > > Memory allocated for the MDIO bus with the devm_kzalloc() API is
> > > associated with the network device. While this will cause memory to be
> > > freed at the right time, it doesn't allow allocating memory before the
> > > network device is initialized.
> > > 
> > > Replace the network device with the parent platform device for memory
> > > allocation to remove that dependency. This also improves consistency
> > > with the other devm_* calls in the driver that all use the platform
> > > device.
> > > 
> > > Signed-off-by: Laurent Pinchart
> > > <laurent.pinchart+renesas@ideasonboard.com>
> > > ---
> > > 
> > >   drivers/net/ethernet/renesas/sh_eth.c | 7 +++----
> > >   1 file changed, 3 insertions(+), 4 deletions(-)
> > > 
> > > diff --git a/drivers/net/ethernet/renesas/sh_eth.c
> > > b/drivers/net/ethernet/renesas/sh_eth.c index be7211d..b80abb6 100644
> > > --- a/drivers/net/ethernet/renesas/sh_eth.c
> > > +++ b/drivers/net/ethernet/renesas/sh_eth.c
> > > @@ -2615,10 +2615,10 @@ static int sh_mdio_init(struct net_device *ndev,
> > > int id,>
> > > 
> > >   	int ret, i;
> > >   	struct bb_info *bitbang;
> > >   	struct sh_eth_private *mdp = netdev_priv(ndev);
> > 
> > I'd suggest to also declare:
> > 	struct platform_device *pdev = &mdp->pdev;
> > 
> > so that you could simplify 'mdev->pdev->name' derefs in the sprintf() call
> > and to get rid of the 'id' parameter to sh_mdio_init() by using 'pdev->id'
> > directly in the same call.
> 
> Good idea. I'll do that in patch 2/5 for v2.

I meant 3/5, sorry.

-- 
Regards,

Laurent Pinchart

Re: [PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[Sergei Shtylyov]

Hello.

On 20-03-2014 17:43, Laurent Pinchart wrote:

>>> On 03/19/2014 02:25 AM, Laurent Pinchart wrote:
>>>> Memory allocated for the MDIO bus with the devm_kzalloc() API is
>>>> associated with the network device. While this will cause memory to be
>>>> freed at the right time, it doesn't allow allocating memory before the
>>>> network device is initialized.

>>>> Replace the network device with the parent platform device for memory
>>>> allocation to remove that dependency. This also improves consistency
>>>> with the other devm_* calls in the driver that all use the platform
>>>> device.

>>>> Signed-off-by: Laurent Pinchart
>>>> <laurent.pinchart+renesas@ideasonboard.com>
>>>> ---
>>>>
>>>>    drivers/net/ethernet/renesas/sh_eth.c | 7 +++----
>>>>    1 file changed, 3 insertions(+), 4 deletions(-)

>>>> diff --git a/drivers/net/ethernet/renesas/sh_eth.c
>>>> b/drivers/net/ethernet/renesas/sh_eth.c index be7211d..b80abb6 100644
>>>> --- a/drivers/net/ethernet/renesas/sh_eth.c
>>>> +++ b/drivers/net/ethernet/renesas/sh_eth.c
>>>> @@ -2615,10 +2615,10 @@ static int sh_mdio_init(struct net_device *ndev,
>>>> int id,>
>>>>
>>>>    	int ret, i;
>>>>    	struct bb_info *bitbang;
>>>>    	struct sh_eth_private *mdp = netdev_priv(ndev);

>>> I'd suggest to also declare:
>>> 	struct platform_device *pdev = &mdp->pdev;

>>> so that you could simplify 'mdev->pdev->name' derefs in the sprintf() call
>>> and to get rid of the 'id' parameter to sh_mdio_init() by using 'pdev->id'
>>> directly in the same call.

>> Good idea. I'll do that in patch 2/5 for v2.

> I meant 3/5, sorry.

    In fact, I think it would fit better in either this patch or a separate 
patch (at least removal of 'id' parameter probably deserves a separate patch).

WBR, Sergei

Re: [PATCH/RFC 1/5] sh_eth: Use the platform device for memory allocation

[Laurent Pinchart]

Hi Sergei,

On Thursday 20 March 2014 19:28:44 Sergei Shtylyov wrote:
> On 20-03-2014 17:43, Laurent Pinchart wrote:
> >>> On 03/19/2014 02:25 AM, Laurent Pinchart wrote:
> >>>> Memory allocated for the MDIO bus with the devm_kzalloc() API is
> >>>> associated with the network device. While this will cause memory to be
> >>>> freed at the right time, it doesn't allow allocating memory before the
> >>>> network device is initialized.
> >>>> 
> >>>> Replace the network device with the parent platform device for memory
> >>>> allocation to remove that dependency. This also improves consistency
> >>>> with the other devm_* calls in the driver that all use the platform
> >>>> device.
> >>>> 
> >>>> Signed-off-by: Laurent Pinchart
> >>>> <laurent.pinchart+renesas@ideasonboard.com>
> >>>> ---
> >>>> 
> >>>>    drivers/net/ethernet/renesas/sh_eth.c | 7 +++----
> >>>>    1 file changed, 3 insertions(+), 4 deletions(-)
> >>>> 
> >>>> diff --git a/drivers/net/ethernet/renesas/sh_eth.c
> >>>> b/drivers/net/ethernet/renesas/sh_eth.c index be7211d..b80abb6 100644
> >>>> --- a/drivers/net/ethernet/renesas/sh_eth.c
> >>>> +++ b/drivers/net/ethernet/renesas/sh_eth.c
> >>>> @@ -2615,10 +2615,10 @@ static int sh_mdio_init(struct net_device
> >>>> *ndev,
> >>>> int id,>
> >>>> 
> >>>>    	int ret, i;
> >>>>    	struct bb_info *bitbang;
> >>>>    	struct sh_eth_private *mdp = netdev_priv(ndev);
> >>> 
> >>> I'd suggest to also declare:
> >>> 	struct platform_device *pdev = &mdp->pdev;
> >>> 
> >>> so that you could simplify 'mdev->pdev->name' derefs in the sprintf()
> >>> call
> >>> and to get rid of the 'id' parameter to sh_mdio_init() by using
> >>> 'pdev->id'
> >>> directly in the same call.
> >> 
> >> Good idea. I'll do that in patch 2/5 for v2.
> > 
> > I meant 3/5, sorry.
> 
> In fact, I think it would fit better in either this patch or a separate
> patch (at least removal of 'id' parameter probably deserves a separate
> patch).

David has been faster than me and has already applied the series to his tree, 
sorry.

-- 
Regards,

Laurent Pinchart