diff mbox series

usb: dwc3: gadget: Fail request submission if it was already queued

Message ID 20190111060212.7390-1-mgautam@codeaurora.org (mailing list archive)
State New, archived
Headers show
Series usb: dwc3: gadget: Fail request submission if it was already queued | expand

Commit Message

Manu Gautam Jan. 11, 2019, 6:02 a.m. UTC
If a function driver tries to re-submit an already queued request,
it can results in corruption of pending/started request lists.
Catch such conditions and fail the request submission to DCD.

Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
---
 drivers/usb/dwc3/gadget.c | 6 ++++++
 1 file changed, 6 insertions(+)

Comments

Felipe Balbi Jan. 11, 2019, 7:43 a.m. UTC | #1
Hi,

Manu Gautam <mgautam@codeaurora.org> writes:
> If a function driver tries to re-submit an already queued request,
> it can results in corruption of pending/started request lists.
> Catch such conditions and fail the request submission to DCD.
>
> Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
> ---
>  drivers/usb/dwc3/gadget.c | 6 ++++++
>  1 file changed, 6 insertions(+)
>
> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
> index 679c12e14522..51716c6b286a 100644
> --- a/drivers/usb/dwc3/gadget.c
> +++ b/drivers/usb/dwc3/gadget.c
> @@ -1290,6 +1290,12 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
>  				&req->request, req->dep->name))
>  		return -EINVAL;
>  
> +	if (req->request.status == -EINPROGRESS) {

this test is really not enough. What if gadget driver set status to
EINPROGRESS before submission? A better check would involve making sure
req isn't part of dep->pending_list or dep->started_list or
dep->cancelled_list. It's clear that this won't work very well as the
amount of requests grow.

Anyway, which gadget driver did this? Why is it only affecting dwc3?
Manu Gautam Jan. 11, 2019, 8:24 a.m. UTC | #2
Hi,

On 1/11/2019 1:13 PM, Felipe Balbi wrote:
> Hi,
>
> Manu Gautam <mgautam@codeaurora.org> writes:
>> If a function driver tries to re-submit an already queued request,
>> it can results in corruption of pending/started request lists.
>> Catch such conditions and fail the request submission to DCD.
>>
>> Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
>> ---
>>  drivers/usb/dwc3/gadget.c | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>> index 679c12e14522..51716c6b286a 100644
>> --- a/drivers/usb/dwc3/gadget.c
>> +++ b/drivers/usb/dwc3/gadget.c
>> @@ -1290,6 +1290,12 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
>>  				&req->request, req->dep->name))
>>  		return -EINVAL;
>>  
>> +	if (req->request.status == -EINPROGRESS) {
> this test is really not enough. What if gadget driver set status to
> EINPROGRESS before submission? A better check would involve making sure
> req isn't part of dep->pending_list or dep->started_list or
> dep->cancelled_list. It's clear that this won't work very well as the
> amount of requests grow.

Thanks for quick review.
'request.status' check can be replaced:
+if (!list_empty(&req->list) {

And replace list_del with list_del_init from dwc3_gadget_giveback()


>
> Anyway, which gadget driver did this? Why is it only affecting dwc3?

Function driver is not in upstream (f_mtp.c).

>
Felipe Balbi Jan. 11, 2019, 9:21 a.m. UTC | #3
Hi,

Manu Gautam <mgautam@codeaurora.org> writes:
>> Manu Gautam <mgautam@codeaurora.org> writes:
>>> If a function driver tries to re-submit an already queued request,
>>> it can results in corruption of pending/started request lists.
>>> Catch such conditions and fail the request submission to DCD.
>>>
>>> Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
>>> ---
>>>  drivers/usb/dwc3/gadget.c | 6 ++++++
>>>  1 file changed, 6 insertions(+)
>>>
>>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>>> index 679c12e14522..51716c6b286a 100644
>>> --- a/drivers/usb/dwc3/gadget.c
>>> +++ b/drivers/usb/dwc3/gadget.c
>>> @@ -1290,6 +1290,12 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
>>>  				&req->request, req->dep->name))
>>>  		return -EINVAL;
>>>  
>>> +	if (req->request.status == -EINPROGRESS) {
>> this test is really not enough. What if gadget driver set status to
>> EINPROGRESS before submission? A better check would involve making sure
>> req isn't part of dep->pending_list or dep->started_list or
>> dep->cancelled_list. It's clear that this won't work very well as the
>> amount of requests grow.
>
> Thanks for quick review.
> 'request.status' check can be replaced:
> +if (!list_empty(&req->list) {
>
> And replace list_del with list_del_init from dwc3_gadget_giveback()

I would rather avoid this. We could start tracking our own internal
dwc3_request status. Something along the lines of:

diff --git a/drivers/usb/dwc3/core.h b/drivers/usb/dwc3/core.h
index df876418cb78..5c3ee741541f 100644
--- a/drivers/usb/dwc3/core.h
+++ b/drivers/usb/dwc3/core.h
@@ -863,6 +863,7 @@ struct dwc3_hwparams {
  * @num_pending_sgs: counter to pending sgs
  * @num_queued_sgs: counter to the number of sgs which already got queued
  * @remaining: amount of data remaining
+ * @status: internal dwc3 request status tracking
  * @epnum: endpoint number to which this request refers
  * @trb: pointer to struct dwc3_trb
  * @trb_dma: DMA address of @trb
@@ -883,6 +884,14 @@ struct dwc3_request {
 	unsigned		num_pending_sgs;
 	unsigned int		num_queued_sgs;
 	unsigned		remaining;
+
+	unsigned int		status;
+#define DWC3_REQUEST_STATUS_QUEUED	0
+#define DWC3_REQUEST_STATUS_STARTED	1
+#define DWC3_REQUEST_STATUS_CANCELLED	2
+#define DWC3_REQUEST_STATUS_COMPLETED	3
+#define DWC3_REQUEST_STATUS_UNKNOWN	-1
+
 	u8			epnum;
 	struct dwc3_trb		*trb;
 	dma_addr_t		trb_dma;
diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 07bd31bb2f8a..74db274786bc 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -208,6 +208,7 @@ void dwc3_gadget_giveback(struct dwc3_ep *dep, struct dwc3_request *req,
 	struct dwc3			*dwc = dep->dwc;
 
 	dwc3_gadget_del_and_unmap_request(dep, req, status);
+	req->status = DWC3_REQUEST_STATUS_COMPLETED;
 
 	spin_unlock(&dwc->lock);
 	usb_gadget_giveback_request(&dep->endpoint, &req->request);
@@ -846,6 +847,7 @@ static struct usb_request *dwc3_gadget_ep_alloc_request(struct usb_ep *ep,
 	req->direction	= dep->direction;
 	req->epnum	= dep->number;
 	req->dep	= dep;
+	req->status	= DWC3_REQUEST_STATUS_UNKNOWN;
 
 	trace_dwc3_alloc_request(req);
 
@@ -1434,6 +1436,11 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
 				&req->request, req->dep->name))
 		return -EINVAL;
 
+	if (WARN(req->status < DWC3_REQUEST_STATUS_COMPLETED,
+				"%s: request %pK already in flight\n",
+				dep->name, &req->request))
+		return -EINVAL;
+
 	pm_runtime_get(dwc->dev);
 
 	req->request.actual	= 0;
@@ -1442,6 +1449,7 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
 	trace_dwc3_ep_queue(req);
 
 	list_add_tail(&req->list, &dep->pending_list);
+	req->status = DWC3_REQUEST_STATUS_QUEUED;
 
 	/*
 	 * NOTICE: Isochronous endpoints should NEVER be prestarted. We must
diff --git a/drivers/usb/dwc3/gadget.h b/drivers/usb/dwc3/gadget.h
index 023a473648eb..6aebe8c0eae1 100644
--- a/drivers/usb/dwc3/gadget.h
+++ b/drivers/usb/dwc3/gadget.h
@@ -76,6 +76,7 @@ static inline void dwc3_gadget_move_started_request(struct dwc3_request *req)
 	struct dwc3_ep		*dep = req->dep;
 
 	req->started = true;
+	req->status = DWC3_REQUEST_STATUS_STARTED;
 	list_move_tail(&req->list, &dep->started_list);
 }
 
@@ -91,6 +92,7 @@ static inline void dwc3_gadget_move_cancelled_request(struct dwc3_request *req)
 	struct dwc3_ep		*dep = req->dep;
 
 	req->started = false;
+	req->status = DWC3_REQUEST_STATUS_CANCELLED;
 	list_move_tail(&req->list, &dep->cancelled_list);
 }
 
With this, we can remove some of the other request flags, such as "started".

>> Anyway, which gadget driver did this? Why is it only affecting dwc3?
>
> Function driver is not in upstream (f_mtp.c).

So this could happen with any UDC, right? Why is f_mtp.c queueing the
same request twice?
Manu Gautam Jan. 16, 2019, 4:34 a.m. UTC | #4
Hi,

On 1/11/2019 2:51 PM, Felipe Balbi wrote:
> Hi,
>
> Manu Gautam <mgautam@codeaurora.org> writes:
>>> Manu Gautam <mgautam@codeaurora.org> writes:
>>>> If a function driver tries to re-submit an already queued request,
>>>> it can results in corruption of pending/started request lists.
>>>> Catch such conditions and fail the request submission to DCD.
>>>>
>>>> Signed-off-by: Manu Gautam <mgautam@codeaurora.org>
>>>> ---
>>>>  drivers/usb/dwc3/gadget.c | 6 ++++++
>>>>  1 file changed, 6 insertions(+)
>>>>
>>>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>>>> index 679c12e14522..51716c6b286a 100644
>>>> --- a/drivers/usb/dwc3/gadget.c
>>>> +++ b/drivers/usb/dwc3/gadget.c
>>>> @@ -1290,6 +1290,12 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
>>>>  				&req->request, req->dep->name))
>>>>  		return -EINVAL;
>>>>  
>>>> +	if (req->request.status == -EINPROGRESS) {
>>> this test is really not enough. What if gadget driver set status to
>>> EINPROGRESS before submission? A better check would involve making sure
>>> req isn't part of dep->pending_list or dep->started_list or
>>> dep->cancelled_list. It's clear that this won't work very well as the
>>> amount of requests grow.
>> Thanks for quick review.
>> 'request.status' check can be replaced:
>> +if (!list_empty(&req->list) {
>>
>> And replace list_del with list_del_init from dwc3_gadget_giveback()
> I would rather avoid this. We could start tracking our own internal
> dwc3_request status. Something along the lines of:

Thanks for this quick patch.

[snip]

>  
> With this, we can remove some of the other request flags, such as "started".
>
>>> Anyway, which gadget driver did this? Why is it only affecting dwc3?
>> Function driver is not in upstream (f_mtp.c).
> So this could happen with any UDC, right? Why is f_mtp.c queueing the
> same request twice?


Looks like chipidea UDC already has such a check.
It is not yet clear that why f_mtp queued same request twice.
However, having a safeguard check in UDC should be helpful.


>
diff mbox series

Patch

diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
index 679c12e14522..51716c6b286a 100644
--- a/drivers/usb/dwc3/gadget.c
+++ b/drivers/usb/dwc3/gadget.c
@@ -1290,6 +1290,12 @@  static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
 				&req->request, req->dep->name))
 		return -EINVAL;
 
+	if (req->request.status == -EINPROGRESS) {
+		dev_err(dwc->dev, "%s: %pK request already in queue\n",
+					dep->name, req);
+		return -EBUSY;
+	}
+
 	pm_runtime_get(dwc->dev);
 
 	req->request.actual	= 0;