2020-09-15 08:04:53

by Wei Yongjun

Subject: [PATCH -next] soc: ti: pruss: Fix return value check

In case of error, the function of_device_get_match_data() returns NULL
pointer not ERR_PTR(). The IS_ERR() test in the return value check
should be replaced with NULL test.

Fixes: ba59c9b43c86 ("soc: ti: pruss: support CORECLK_MUX and IEPCLK_MUX")
Reported-by: Hulk Robot <[email protected]>
Signed-off-by: Wei Yongjun <[email protected]>
---
drivers/soc/ti/pruss.c | 4 ++--
1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/soc/ti/pruss.c b/drivers/soc/ti/pruss.c
index cc0b4ad7a3d3..582f48051c30 100644
--- a/drivers/soc/ti/pruss.c
+++ b/drivers/soc/ti/pruss.c
@@ -126,7 +126,7 @@ static int pruss_clk_init(struct pruss *pruss, struct device_node *cfg_node)
int ret = 0;

data = of_device_get_match_data(dev);
- if (IS_ERR(data))
+ if (!data)
return -ENODEV;

clks_np = of_get_child_by_name(cfg_node, "clocks");
@@ -175,7 +175,7 @@ static int pruss_probe(struct platform_device *pdev)
const char *mem_names[PRUSS_MEM_MAX] = { "dram0", "dram1", "shrdram2" };

data = of_device_get_match_data(&pdev->dev);
- if (IS_ERR(data)) {
+ if (!data) {
dev_err(dev, "missing private data\n");
return -ENODEV;
}

2020-09-17 18:23:10

by Grzegorz Jaszczyk

Subject: Re: [PATCH -next] soc: ti: pruss: Fix return value check

Hi

On Tue, 15 Sep 2020 at 09:58, Wei Yongjun <[email protected]> wrote:
>
> In case of error, the function of_device_get_match_data() returns NULL
> pointer not ERR_PTR(). The IS_ERR() test in the return value check
> should be replaced with NULL test.
>
> Fixes: ba59c9b43c86 ("soc: ti: pruss: support CORECLK_MUX and IEPCLK_MUX")
> Reported-by: Hulk Robot <[email protected]>
> Signed-off-by: Wei Yongjun <[email protected]>
> ---
> drivers/soc/ti/pruss.c | 4 ++--
> 1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/soc/ti/pruss.c b/drivers/soc/ti/pruss.c
> index cc0b4ad7a3d3..582f48051c30 100644
> --- a/drivers/soc/ti/pruss.c
> +++ b/drivers/soc/ti/pruss.c
> @@ -126,7 +126,7 @@ static int pruss_clk_init(struct pruss *pruss, struct device_node *cfg_node)
> int ret = 0;
>
> data = of_device_get_match_data(dev);
> - if (IS_ERR(data))
> + if (!data)
> return -ENODEV;
>
> clks_np = of_get_child_by_name(cfg_node, "clocks");
> @@ -175,7 +175,7 @@ static int pruss_probe(struct platform_device *pdev)
> const char *mem_names[PRUSS_MEM_MAX] = { "dram0", "dram1", "shrdram2" };
>
> data = of_device_get_match_data(&pdev->dev);
> - if (IS_ERR(data)) {
> + if (!data) {
> dev_err(dev, "missing private data\n");
> return -ENODEV;
> }

First of all thank you for reporting this issue. Indeed the IS_ERR is
wrongly used and is leftover from a bit different former internal
implementation. Nevertheless with your fix the driver will not be
functional anymore for all devices without match data (e.g.
"ti,am3356-pruss") which is not what we want.

The proper fix would be removing the mentioned error checks, since the
"data" in both cases is checked later, before usage:
if (data && data->...).

Please let me know if you want to improve this patch by yourself or
want me to push a proper fix with your e.g. "Reported-by:" tag. For me
both solutions are ok.

Thank you,
Grzegorz