2016-03-22 11:37:18

by Shawn Lin

Subject: [PATCH] spi: rockchip: fix warning of static check

Let's improve the check with -EPROBE_DEFER, otherwise
we may pass on null pointer to PTR_ERR. That causes the
static checker warning: passing zero to 'PTR_ERR'.

Reported-by: Dan Carpenter <[email protected]>
Cc: Doug Anderson <[email protected]>
Cc: Dan Carpenter <[email protected]>
Signed-off-by: Shawn Lin <[email protected]>
---

drivers/spi/spi-rockchip.c | 10 +++++++---
1 file changed, 7 insertions(+), 3 deletions(-)

diff --git a/drivers/spi/spi-rockchip.c b/drivers/spi/spi-rockchip.c
index bfeb0d4..2cb36d9 100644
--- a/drivers/spi/spi-rockchip.c
+++ b/drivers/spi/spi-rockchip.c
@@ -726,20 +726,24 @@ static int rockchip_spi_probe(struct platform_device *pdev)
rs->dma_tx.ch = dma_request_slave_channel(rs->dev, "tx");
if (IS_ERR_OR_NULL(rs->dma_tx.ch)) {
/* Check tx to see if we need defer probing driver */
- if (PTR_ERR(rs->dma_tx.ch) == -EPROBE_DEFER) {
+ if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {
ret = -EPROBE_DEFER;
goto err_get_fifo_len;
}
dev_warn(rs->dev, "Failed to request TX DMA channel\n");
+ rs->dma_tx.ch = NULL;
}

rs->dma_rx.ch = dma_request_slave_channel(rs->dev, "rx");
- if (!rs->dma_rx.ch) {
- if (rs->dma_tx.ch) {
+ if (IS_ERR_OR_NULL(rs->dma_rx.ch)) {
+ if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {
dma_release_channel(rs->dma_tx.ch);
rs->dma_tx.ch = NULL;
+ ret = -EPROBE_DEFER;
+ goto err_get_fifo_len;
}
dev_warn(rs->dev, "Failed to request RX DMA channel\n");
+ rs->dma_rx.ch = NULL;
}

if (rs->dma_tx.ch && rs->dma_rx.ch) {
--
2.3.7

2016-03-22 12:56:34

by Vladimir Zapolskiy

Subject: Re: [PATCH] spi: rockchip: fix warning of static check

Hi Shawn,

On 22.03.2016 13:36, Shawn Lin wrote:
> Let's improve the check with -EPROBE_DEFER, otherwise
> we may pass on null pointer to PTR_ERR. That causes the
> static checker warning: passing zero to 'PTR_ERR'.
>
> Reported-by: Dan Carpenter <[email protected]>
> Cc: Doug Anderson <[email protected]>
> Cc: Dan Carpenter <[email protected]>
> Signed-off-by: Shawn Lin <[email protected]>
> ---
>
> drivers/spi/spi-rockchip.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/spi/spi-rockchip.c b/drivers/spi/spi-rockchip.c
> index bfeb0d4..2cb36d9 100644
> --- a/drivers/spi/spi-rockchip.c
> +++ b/drivers/spi/spi-rockchip.c
> @@ -726,20 +726,24 @@ static int rockchip_spi_probe(struct platform_device *pdev)
> rs->dma_tx.ch = dma_request_slave_channel(rs->dev, "tx");
> if (IS_ERR_OR_NULL(rs->dma_tx.ch)) {
> /* Check tx to see if we need defer probing driver */
> - if (PTR_ERR(rs->dma_tx.ch) == -EPROBE_DEFER) {
> + if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {

won't it be easier to change dma_request_slave_channel() to
dma_request_chan() or dma_request_slave_channel_reason() and replace strange
IS_ERR_OR_NULL() check with IS_ERR() one?

AFAIU that's almost what has been suggested by Doug and it is a better
fix than this one published.

At the moment check for IS_ERR_OR_NULL() is too aggressive and quite
confusing, because dma_request_slave_channel() never returns NULL.

> ret = -EPROBE_DEFER;
> goto err_get_fifo_len;
> }
> dev_warn(rs->dev, "Failed to request TX DMA channel\n");
> + rs->dma_tx.ch = NULL;
> }
>
> rs->dma_rx.ch = dma_request_slave_channel(rs->dev, "rx");
> - if (!rs->dma_rx.ch) {
> - if (rs->dma_tx.ch) {
> + if (IS_ERR_OR_NULL(rs->dma_rx.ch)) {
> + if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {

Same as above.

> dma_release_channel(rs->dma_tx.ch);
> rs->dma_tx.ch = NULL;
> + ret = -EPROBE_DEFER;
> + goto err_get_fifo_len;
> }
> dev_warn(rs->dev, "Failed to request RX DMA channel\n");
> + rs->dma_rx.ch = NULL;
> }
>
> if (rs->dma_tx.ch && rs->dma_rx.ch) {
>

--
With best wishes,
Vladimir

2016-03-22 13:00:36

by Vladimir Zapolskiy

Subject: Re: [PATCH] spi: rockchip: fix warning of static check

On 22.03.2016 14:56, Vladimir Zapolskiy wrote:
> Hi Shawn,
>
> On 22.03.2016 13:36, Shawn Lin wrote:
>> Let's improve the check with -EPROBE_DEFER, otherwise
>> we may pass on null pointer to PTR_ERR. That causes the
>> static checker warning: passing zero to 'PTR_ERR'.
>>
>> Reported-by: Dan Carpenter <[email protected]>
>> Cc: Doug Anderson <[email protected]>
>> Cc: Dan Carpenter <[email protected]>
>> Signed-off-by: Shawn Lin <[email protected]>
>> ---
>>
>> drivers/spi/spi-rockchip.c | 10 +++++++---
>> 1 file changed, 7 insertions(+), 3 deletions(-)
>>
>> diff --git a/drivers/spi/spi-rockchip.c b/drivers/spi/spi-rockchip.c
>> index bfeb0d4..2cb36d9 100644
>> --- a/drivers/spi/spi-rockchip.c
>> +++ b/drivers/spi/spi-rockchip.c
>> @@ -726,20 +726,24 @@ static int rockchip_spi_probe(struct platform_device *pdev)
>> rs->dma_tx.ch = dma_request_slave_channel(rs->dev, "tx");
>> if (IS_ERR_OR_NULL(rs->dma_tx.ch)) {
>> /* Check tx to see if we need defer probing driver */
>> - if (PTR_ERR(rs->dma_tx.ch) == -EPROBE_DEFER) {
>> + if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {
>
> won't it be easier to change dma_request_slave_channel() to
> dma_request_chan() or dma_request_slave_channel_reason() and replace strange
> IS_ERR_OR_NULL() check with IS_ERR() one?
>
> AFAIU that's almost what has been suggested by Doug and it is a better
> fix than this one published.
>
> At the moment check for IS_ERR_OR_NULL() is too aggressive and quite
> confusing, because dma_request_slave_channel() never returns NULL.

Sorry for a typo, never returns IS_ERR() of course.

>> ret = -EPROBE_DEFER;
>> goto err_get_fifo_len;
>> }
>> dev_warn(rs->dev, "Failed to request TX DMA channel\n");
>> + rs->dma_tx.ch = NULL;
>> }
>>
>> rs->dma_rx.ch = dma_request_slave_channel(rs->dev, "rx");
>> - if (!rs->dma_rx.ch) {
>> - if (rs->dma_tx.ch) {
>> + if (IS_ERR_OR_NULL(rs->dma_rx.ch)) {
>> + if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {
>
> Same as above.
>
>> dma_release_channel(rs->dma_tx.ch);
>> rs->dma_tx.ch = NULL;
>> + ret = -EPROBE_DEFER;
>> + goto err_get_fifo_len;
>> }
>> dev_warn(rs->dev, "Failed to request RX DMA channel\n");
>> + rs->dma_rx.ch = NULL;
>> }
>>
>> if (rs->dma_tx.ch && rs->dma_rx.ch) {
>>
>

--
With best wishes,
Vladimir

2016-03-22 16:23:39

by Doug Anderson

Subject: Re: [PATCH] spi: rockchip: fix warning of static check

Hi,

On Tue, Mar 22, 2016 at 4:36 AM, Shawn Lin <[email protected]> wrote:
> Let's improve the check with -EPROBE_DEFER, otherwise
> we may pass on null pointer to PTR_ERR. That causes the
> static checker warning: passing zero to 'PTR_ERR'.
>
> Reported-by: Dan Carpenter <[email protected]>
> Cc: Doug Anderson <[email protected]>
> Cc: Dan Carpenter <[email protected]>
> Signed-off-by: Shawn Lin <[email protected]>
> ---
>
> drivers/spi/spi-rockchip.c | 10 +++++++---
> 1 file changed, 7 insertions(+), 3 deletions(-)
>
> diff --git a/drivers/spi/spi-rockchip.c b/drivers/spi/spi-rockchip.c
> index bfeb0d4..2cb36d9 100644
> --- a/drivers/spi/spi-rockchip.c
> +++ b/drivers/spi/spi-rockchip.c
> @@ -726,20 +726,24 @@ static int rockchip_spi_probe(struct platform_device *pdev)
> rs->dma_tx.ch = dma_request_slave_channel(rs->dev, "tx");

Please change from:
dma_request_slave_channel()
to:
dma_request_chan()

The first will never return -EPROBE_DEFER. The second will.

> if (IS_ERR_OR_NULL(rs->dma_tx.ch)) {

Once you've changed to dma_request_chan(), change IS_ERR_OR_NULL() to
just IS_ERR(). dma_request_chan() never returns NULL.

> /* Check tx to see if we need defer probing driver */
> - if (PTR_ERR(rs->dma_tx.ch) == -EPROBE_DEFER) {
> + if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {

Once you've changed to use dma_request_chan(), no need to change this
line. Leave it as it was.

> ret = -EPROBE_DEFER;
> goto err_get_fifo_len;
> }
> dev_warn(rs->dev, "Failed to request TX DMA channel\n");
> + rs->dma_tx.ch = NULL;
> }
>
> rs->dma_rx.ch = dma_request_slave_channel(rs->dev, "rx");
> - if (!rs->dma_rx.ch) {
> - if (rs->dma_tx.ch) {
> + if (IS_ERR_OR_NULL(rs->dma_rx.ch)) {
> + if (rs->dma_tx.ch == ERR_PTR(-EPROBE_DEFER)) {
> dma_release_channel(rs->dma_tx.ch);
> rs->dma_tx.ch = NULL;
> + ret = -EPROBE_DEFER;
> + goto err_get_fifo_len;
> }
> dev_warn(rs->dev, "Failed to request RX DMA channel\n");
> + rs->dma_rx.ch = NULL;

Same comments apply to the rx path.