Re: [PATCH] drm/bridge: nxp-ptn3460: fix i2c_master_send() error checking

[Neil Armstrong <neil.armstrong@linaro.org>]

On 04/12/2023 17:59, Dan Carpenter wrote:
> On Mon, Dec 04, 2023 at 02:53:05PM +0100, Neil Armstrong wrote:
>> On 04/12/2023 13:29, Dan Carpenter wrote:
>>> The i2c_master_send/recv() functions return negative error codes or the
>>> number of bytes that were able to be sent/received.  This code has
>>> two problems.  1)  Instead of checking if all the bytes were sent or
>>> received, it checks that at least one byte was sent or received.
>>> 2) If there was a partial send/receive then we should return a negative
>>> error code but this code returns success.
>>>
>>> Fixes: a9fe713d7d45 ("drm/bridge: Add PTN3460 bridge driver")
>>> Cc: stable@vger.kernel.org
>>> Signed-off-by: Dan Carpenter <dan.carpenter@linaro.org>
>>> ---
>>> This is from static analysis and code review.  It's always a concern
>>> when you add stricter error handling that something will break.
>>>
>>>    drivers/gpu/drm/bridge/nxp-ptn3460.c | 10 +++++-----
>>>    1 file changed, 5 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/bridge/nxp-ptn3460.c b/drivers/gpu/drm/bridge/nxp-ptn3460.c
>>> index d81920227a8a..9b7eb8c669c1 100644
>>> --- a/drivers/gpu/drm/bridge/nxp-ptn3460.c
>>> +++ b/drivers/gpu/drm/bridge/nxp-ptn3460.c
>>> @@ -56,13 +56,13 @@ static int ptn3460_read_bytes(struct ptn3460_bridge *ptn_bridge, char addr,
>>>    	ret = i2c_master_send(ptn_bridge->client, &addr, 1);
>>>    	if (ret <= 0) {
>>>    		DRM_ERROR("Failed to send i2c command, ret=%d\n", ret);
>>> -		return ret;
>>> +		return ret ?: -EIO;
>>>    	}
>>>    	ret = i2c_master_recv(ptn_bridge->client, buf, len);
>>> -	if (ret <= 0) {
>>> +	if (ret != len) {
>>
>> This is impossible, i2c_transfer_buffer_flags() returns len as-is if no error, so
>> ret can only be negative or equal to len. The original code is right.
> 
> It works, but it's not "right".  The <= 0 could be changed to < 0.  The
> "len" variable is EDID_LENGTH (128).

So indeed, switching to < 0 is the most reasonable, no need to change the ret value in this case.

Neil

> 
> regards,
> dan carpenter
>