2021-12-07 13:17:33

by Guo Zhengkui

Subject: [PATCH] phy: qcom: fix bug: application of sizeof to pointer

Fix following coccicheck error:
./drivers/phy/qualcomm/phy-qcom-edp.c:574:31-37:
ERROR: application of sizeof to pointer.

Use sizeof(*data) instead.

Signed-off-by: Guo Zhengkui <[email protected]>
---
drivers/phy/qualcomm/phy-qcom-edp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/phy/qualcomm/phy-qcom-edp.c b/drivers/phy/qualcomm/phy-qcom-edp.c
index 17d5653b661d..5fe4eab9cac1 100644
--- a/drivers/phy/qualcomm/phy-qcom-edp.c
+++ b/drivers/phy/qualcomm/phy-qcom-edp.c
@@ -571,7 +571,7 @@ static int qcom_edp_clks_register(struct qcom_edp *edp, struct device_node *np)
struct clk_init_data init = { };
int ret;

- data = devm_kzalloc(edp->dev, sizeof(data), GFP_KERNEL);
+ data = devm_kzalloc(edp->dev, sizeof(*data), GFP_KERNEL);
if (!data)
return -ENOMEM;

--
2.20.1

2021-12-07 15:12:21

by Bjorn Andersson

Subject: Re: [PATCH] phy: qcom: fix bug: application of sizeof to pointer

On Tue 07 Dec 05:16 PST 2021, Guo Zhengkui wrote:

> Fix following coccicheck error:
> ./drivers/phy/qualcomm/phy-qcom-edp.c:574:31-37:
> ERROR: application of sizeof to pointer.
>
> Use sizeof(*data) instead.
>
> Signed-off-by: Guo Zhengkui <[email protected]>
> ---
> drivers/phy/qualcomm/phy-qcom-edp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/phy/qualcomm/phy-qcom-edp.c b/drivers/phy/qualcomm/phy-qcom-edp.c
> index 17d5653b661d..5fe4eab9cac1 100644
> --- a/drivers/phy/qualcomm/phy-qcom-edp.c
> +++ b/drivers/phy/qualcomm/phy-qcom-edp.c
> @@ -571,7 +571,7 @@ static int qcom_edp_clks_register(struct qcom_edp *edp, struct device_node *np)
> struct clk_init_data init = { };
> int ret;
>
> - data = devm_kzalloc(edp->dev, sizeof(data), GFP_KERNEL);
> + data = devm_kzalloc(edp->dev, sizeof(*data), GFP_KERNEL);

Ouch, thanks for catching that!

But the clk_hw_onecell_data actually has a variable size array at the
end and as you can see further down I store 2 items in that array.

So that sizeof should be struct_size(data, hws, 2)

Would you be willing to update your patch with that?

And please add:
Fixes: f199223cb490 ("phy: qcom: Introduce new eDP PHY driver")

Regards,
Bjorn

> if (!data)
> return -ENOMEM;
>
> --
> 2.20.1
>

2021-12-08 06:56:50

by Guo Zhengkui

Subject: Re: [PATCH] phy: qcom: fix bug: application of sizeof to pointer

On 2021/12/7 23:13, Bjorn Andersson wrote:
> On Tue 07 Dec 05:16 PST 2021, Guo Zhengkui wrote:
>
>> Fix following coccicheck error:
>> ./drivers/phy/qualcomm/phy-qcom-edp.c:574:31-37:
>> ERROR: application of sizeof to pointer.
>>
>> Use sizeof(*data) instead.
>>
>> Signed-off-by: Guo Zhengkui <[email protected]>
>> ---
>> drivers/phy/qualcomm/phy-qcom-edp.c | 2 +-
>> 1 file changed, 1 insertion(+), 1 deletion(-)
>>
>> diff --git a/drivers/phy/qualcomm/phy-qcom-edp.c b/drivers/phy/qualcomm/phy-qcom-edp.c
>> index 17d5653b661d..5fe4eab9cac1 100644
>> --- a/drivers/phy/qualcomm/phy-qcom-edp.c
>> +++ b/drivers/phy/qualcomm/phy-qcom-edp.c
>> @@ -571,7 +571,7 @@ static int qcom_edp_clks_register(struct qcom_edp *edp, struct device_node *np)
>> struct clk_init_data init = { };
>> int ret;
>>
>> - data = devm_kzalloc(edp->dev, sizeof(data), GFP_KERNEL);
>> + data = devm_kzalloc(edp->dev, sizeof(*data), GFP_KERNEL);
>
> Ouch, thanks for catching that!
>
> But the clk_hw_onecell_data actually has a variable size array at the
> end and as you can see further down I store 2 items in that array.
>
> So that sizeof should be struct_size(data, hws, 2)
>
> Would you be willing to update your patch with that?
>

OK, I will commit another patch. Do you mind I add a "Suggested-by" tag
of your name in the new patch?

Zhengkui

> And please add:
> Fixes: f199223cb490 ("phy: qcom: Introduce new eDP PHY driver")
>
> Regards,
> Bjorn
>
>> if (!data)
>> return -ENOMEM;
>>
>> --
>> 2.20.1
>>

2021-12-08 16:27:40

by Bjorn Andersson

Subject: Re: [PATCH] phy: qcom: fix bug: application of sizeof to pointer

On Tue 07 Dec 22:56 PST 2021, Guo Zhengkui wrote:

> On 2021/12/7 23:13, Bjorn Andersson wrote:
> > On Tue 07 Dec 05:16 PST 2021, Guo Zhengkui wrote:
> >
> > > Fix following coccicheck error:
> > > ./drivers/phy/qualcomm/phy-qcom-edp.c:574:31-37:
> > > ERROR: application of sizeof to pointer.
> > >
> > > Use sizeof(*data) instead.
> > >
> > > Signed-off-by: Guo Zhengkui <[email protected]>
> > > ---
> > > drivers/phy/qualcomm/phy-qcom-edp.c | 2 +-
> > > 1 file changed, 1 insertion(+), 1 deletion(-)
> > >
> > > diff --git a/drivers/phy/qualcomm/phy-qcom-edp.c b/drivers/phy/qualcomm/phy-qcom-edp.c
> > > index 17d5653b661d..5fe4eab9cac1 100644
> > > --- a/drivers/phy/qualcomm/phy-qcom-edp.c
> > > +++ b/drivers/phy/qualcomm/phy-qcom-edp.c
> > > @@ -571,7 +571,7 @@ static int qcom_edp_clks_register(struct qcom_edp *edp, struct device_node *np)
> > > struct clk_init_data init = { };
> > > int ret;
> > > - data = devm_kzalloc(edp->dev, sizeof(data), GFP_KERNEL);
> > > + data = devm_kzalloc(edp->dev, sizeof(*data), GFP_KERNEL);
> >
> > Ouch, thanks for catching that!
> >
> > But the clk_hw_onecell_data actually has a variable size array at the
> > end and as you can see further down I store 2 items in that array.
> >
> > So that sizeof should be struct_size(data, hws, 2)
> >
> > Would you be willing to update your patch with that?
> >
>
> OK, I will commit another patch. Do you mind I add a "Suggested-by" tag of
> your name in the new patch?
>

I don't mind.

Regards,
Bjorn

> Zhengkui
>
> > And please add:
> > Fixes: f199223cb490 ("phy: qcom: Introduce new eDP PHY driver")
> >
> > Regards,
> > Bjorn
> >
> > > if (!data)
> > > return -ENOMEM;
> > > --
> > > 2.20.1
> > >

2021-12-09 03:21:45

by Guo Zhengkui

Subject: [PATCH] phy: qcom: use struct_size instead of sizeof

Use struct_size() to get the accurate size of `clk_hw_onecell_data`
with a variable size array, instead of sizeof(data) to get the size
of a pointer.

Suggested-by: Bjorn Andersson <[email protected]>
Signed-off-by: Guo Zhengkui <[email protected]>
---
drivers/phy/qualcomm/phy-qcom-edp.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/phy/qualcomm/phy-qcom-edp.c b/drivers/phy/qualcomm/phy-qcom-edp.c
index 17d5653b661d..a8ecd2e8442d 100644
--- a/drivers/phy/qualcomm/phy-qcom-edp.c
+++ b/drivers/phy/qualcomm/phy-qcom-edp.c
@@ -571,7 +571,7 @@ static int qcom_edp_clks_register(struct qcom_edp *edp, struct device_node *np)
struct clk_init_data init = { };
int ret;

- data = devm_kzalloc(edp->dev, sizeof(data), GFP_KERNEL);
+ data = devm_kzalloc(edp->dev, struct_size(data, hws, 2), GFP_KERNEL);
if (!data)
return -ENOMEM;

--
2.20.1

2021-12-09 04:48:44

by Bjorn Andersson

Subject: Re: [PATCH] phy: qcom: use struct_size instead of sizeof

On Wed 08 Dec 19:21 PST 2021, Guo Zhengkui wrote:

> Use struct_size() to get the accurate size of `clk_hw_onecell_data`
> with a variable size array, instead of sizeof(data) to get the size
> of a pointer.
>

Fixes: f199223cb490 ("phy: qcom: Introduce new eDP PHY driver")
Reviewed-by: Bjorn Andersson <[email protected]>

Thanks,
Bjorn

> Suggested-by: Bjorn Andersson <[email protected]>
> Signed-off-by: Guo Zhengkui <[email protected]>
> ---
> drivers/phy/qualcomm/phy-qcom-edp.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/phy/qualcomm/phy-qcom-edp.c b/drivers/phy/qualcomm/phy-qcom-edp.c
> index 17d5653b661d..a8ecd2e8442d 100644
> --- a/drivers/phy/qualcomm/phy-qcom-edp.c
> +++ b/drivers/phy/qualcomm/phy-qcom-edp.c
> @@ -571,7 +571,7 @@ static int qcom_edp_clks_register(struct qcom_edp *edp, struct device_node *np)
> struct clk_init_data init = { };
> int ret;
>
> - data = devm_kzalloc(edp->dev, sizeof(data), GFP_KERNEL);
> + data = devm_kzalloc(edp->dev, struct_size(data, hws, 2), GFP_KERNEL);
> if (!data)
> return -ENOMEM;
>
> --
> 2.20.1
>

2021-12-09 11:55:20

by Vinod Koul

Subject: Re: [PATCH] phy: qcom: use struct_size instead of sizeof

On 09-12-21, 11:21, Guo Zhengkui wrote:
> Use struct_size() to get the accurate size of `clk_hw_onecell_data`
> with a variable size array, instead of sizeof(data) to get the size
> of a pointer.

Applied, thanks

--
~Vinod