Re: [PATCH v2 14/26] usb: gadget: s3c-hsudc: remove usage of list iterator past the loop body

[Jakob Koschel]

> On 9. Mar 2022, at 18:25, Krzysztof Kozlowski <krzysztof.kozlowski@xxxxxxxxxxxxx> wrote:
> 
> On 08/03/2022 18:18, Jakob Koschel wrote:
>> If the list representing the request queue does not contain the expected
>> request, the value of the list_for_each_entry() iterator will not point
>> to a valid structure. To avoid type confusion in such case, the list
>> iterator scope will be limited to the list_for_each_entry() loop.
>> 
>> In preparation to limiting scope of the list iterator to the list traversal
>> loop, use a dedicated pointer to point to the found request object [1].
>> 
>> Link: https://lore.kernel.org/all/YhdfEIwI4EdtHdym@xxxxxxxxx/
>> Signed-off-by: Jakob Koschel <jakobkoschel@xxxxxxxxx>
>> ---
>> drivers/usb/gadget/udc/s3c-hsudc.c | 12 +++++++-----
>> 1 file changed, 7 insertions(+), 5 deletions(-)
>> 
>> diff --git a/drivers/usb/gadget/udc/s3c-hsudc.c b/drivers/usb/gadget/udc/s3c-hsudc.c
>> index 89f1f8c9f02e..bf803e013458 100644
>> --- a/drivers/usb/gadget/udc/s3c-hsudc.c
>> +++ b/drivers/usb/gadget/udc/s3c-hsudc.c
>> @@ -877,7 +877,7 @@ static int s3c_hsudc_dequeue(struct usb_ep *_ep, struct usb_request *_req)
>> {
>> 	struct s3c_hsudc_ep *hsep = our_ep(_ep);
>> 	struct s3c_hsudc *hsudc = hsep->dev;
>> -	struct s3c_hsudc_req *hsreq;
>> +	struct s3c_hsudc_req *hsreq = NULL, *iter;
>> 	unsigned long flags;
>> 
>> 	hsep = our_ep(_ep);
>> @@ -886,11 +886,13 @@ static int s3c_hsudc_dequeue(struct usb_ep *_ep, struct usb_request *_req)
>> 
>> 	spin_lock_irqsave(&hsudc->lock, flags);
>> 
>> -	list_for_each_entry(hsreq, &hsep->queue, queue) {
>> -		if (&hsreq->req == _req)
>> -			break;
>> +	list_for_each_entry(iter, &hsep->queue, queue) {
>> +		if (&iter->req != _req)
>> +			continue;
>> +		hsreq = iter;
>> +		break;
> 
> You have in the loop both continue and break, looks a bit complicated.
> Why not simpler:
> 
> if (&iter->req == _req) {
> 	hsreq = iter;
> 	break;
> }

Ultimately I changed this based on the feedback from Linus
(Link: https://lore.kernel.org/linux-kernel/CAHk-=wheru6rEfzC2wuO9k03PRF6s3nhxryCAnwR5bzKwPV2ww@xxxxxxxxxxxxxx/).

> 
> ?
> Less code, typical (expected) code flow when looking for something in
> the list. Is your approach related to some speculative execution?

no relation to speculative execution.

I have no preference for one over the other, so I'll just change it to
however it is desired.

It would just be great to have a (somewhat) consistent way so I can prepare
the rest of the patch sets accordingly.

> 
>> 	}
>> -	if (&hsreq->req != _req) {
>> +	if (!hsreq) {
>> 		spin_unlock_irqrestore(&hsudc->lock, flags);
>> 		return -EINVAL;
>> 	}
> 
> 
> Best regards,
> Krzysztof

Jakob