On Nov 06 2017 or thereabouts, Sébastien Szymanski wrote:

On 11/06/2017 09:11 AM, Benjamin Tissoires wrote:

quoted

On Nov 02 2017 or thereabouts, Sébastien Szymanski wrote:

quoted

When everything goes smoothly, ret is set to 0 which makes the function
to return EIO error.

Fixes: 8e9faa15469e ("HID: cp2112: fix gpio-callback error handling")
Signed-off-by: Sébastien Szymanski <redacted>
---
 drivers/hid/hid-cp2112.c | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/drivers/hid/hid-cp2112.c b/drivers/hid/hid-cp2112.c
index 28e3c18..f7754a6 100644
--- a/drivers/hid/hid-cp2112.c
+++ b/drivers/hid/hid-cp2112.c

@@ -205,12 +205,13 @@ static int cp2112_gpio_direction_input(struct gpio_chip *chip, unsigned offset)
 	ret = hid_hw_raw_request(hdev, CP2112_GPIO_CONFIG, buf,
 				 CP2112_GPIO_CONFIG_LENGTH, HID_FEATURE_REPORT,
 				 HID_REQ_SET_REPORT);
-	if (ret < 0) {
+	if (ret != CP2112_GPIO_CONFIG_LENGTH) {

Ack for this.

As explained in the interface specification, the device doesn't answer
to set reports, so the transfer should be CP2112_GPIO_CONFIG_LENGTH (5)
bytes.

quoted

quoted

 		hid_err(hdev, "error setting GPIO config: %d\n", ret);
 		goto exit;
 	}
 
-	ret = 0;
+	mutex_unlock(&dev->lock);
+	return 0;

Wouldn't it be better to just turn
- return ret < 0 ? ret : -EIO;
into
+ return ret <= 0 ? ret : -EIO;
at the end of the function?

Well, the commit I mentioned in the Fixes tag, changes from

- return ret <= 0 ? ret : -EIO;

to

+ return ret < 0 ? ret : -EIO;

because ret being 0 could mean that one of the hid_hw_raw_request
returned 0.

True.

So, commit 8e9faa15469e basically makes all transfert return -EIO and
should be reverted, right?

IMO, your commit should be:
if (ret != CP2112_GPIO_CONFIG_LENGTH) {
	hid_err(hdev, "error setting GPIO config: %d\n", ret);
	if (ret >= 0)
		ret = -EIO;
	goto exit;
}

ret = 0;
exit:
mutex_unlock(&dev->lock);
return ret;

How does that sound?

Cheers,
Benjamin

Regards,

quoted

I'd rather keep the same exit path in both cases, error or success.

Cheers,
Benjamin

quoted

 
 exit:
 	mutex_unlock(&dev->lock);
-- 
2.7.3

-- 
Sébastien Szymanski
Software engineer, Armadeus Systems
Tel: +33 (0)9 72 29 41 44
Fax: +33 (0)9 72 28 79 26