Re: [RFC][PATCH] usb: dwc3: usb: dwc3: Force stop EP0 transfers during pullup disable

[Thinh Nguyen <Thinh.Nguyen@synopsys.com>]

Wesley Cheng wrote:
> Hi Thinh,
> 
> On 8/18/2021 5:40 PM, Thinh Nguyen wrote:
>> Wesley Cheng wrote:
>>> Hi Thinh/Felipe,
>>>
>>> On 8/16/2021 12:13 PM, Wesley Cheng wrote:
>>>> Hi Thinh,
>>>>
>>>> On 8/15/2021 5:33 PM, Thinh Nguyen wrote:
>>>>> Felipe Balbi wrote:
>>>>>>
>>>>>> Hi,
>>>>>>
>>>>>> Thinh Nguyen <Thinh.Nguyen@synopsys.com> writes:
>>>>>>>>>>>>>>> If this occurs, then the entire pullup disable routine is skipped and
>>>>>>>>>>>>>>> proper cleanup and halting of the controller does not complete.
>>>>>>>>>>>>>>> Instead of returning an error (which is ignored from the UDC
>>>>>>>>>>>>>>> perspective), do what is mentioned in the comments and force the
>>>>>>>>>>>>>>> transaction to complete and put the ep0state back to the SETUP phase.
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> Signed-off-by: Wesley Cheng <wcheng@codeaurora.org>
>>>>>>>>>>>>>>> ---
>>>>>>>>>>>>>>>  drivers/usb/dwc3/ep0.c    | 4 ++--
>>>>>>>>>>>>>>>  drivers/usb/dwc3/gadget.c | 6 +++++-
>>>>>>>>>>>>>>>  drivers/usb/dwc3/gadget.h | 3 +++
>>>>>>>>>>>>>>>  3 files changed, 10 insertions(+), 3 deletions(-)
>>>>>>>>>>>>>>>
>>>>>>>>>>>>>>> diff --git a/drivers/usb/dwc3/ep0.c b/drivers/usb/dwc3/ep0.c
>>>>>>>>>>>>>>> index 6587394..abfc42b 100644
>>>>>>>>>>>>>>> --- a/drivers/usb/dwc3/ep0.c
>>>>>>>>>>>>>>> +++ b/drivers/usb/dwc3/ep0.c
>>>>>>>>>>>>>>> @@ -218,7 +218,7 @@ int dwc3_gadget_ep0_queue(struct usb_ep *ep, struct usb_request *request,
>>>>>>>>>>>>>>>  	return ret;
>>>>>>>>>>>>>>>  }
>>>>>>>>>>>>>>>  
>>>>>>>>>>>>>>> -static void dwc3_ep0_stall_and_restart(struct dwc3 *dwc)
>>>>>>>>>>>>>>> +void dwc3_ep0_stall_and_restart(struct dwc3 *dwc)
>>>>>>>>>>>>>>>  {
>>>>>>>>>>>>>>>  	struct dwc3_ep		*dep;
>>>>>>>>>>>>>>>  
>>>>>>>>>>>>>>> @@ -1073,7 +1073,7 @@ void dwc3_ep0_send_delayed_status(struct dwc3 *dwc)
>>>>>>>>>>>>>>>  	__dwc3_ep0_do_control_status(dwc, dwc->eps[direction]);
>>>>>>>>>>>>>>>  }
>>>>>>>>>>>>>>>  
>>>>>>>>>>>>>>> -static void dwc3_ep0_end_control_data(struct dwc3 *dwc, struct dwc3_ep *dep)
>>>>>>>>>>>>>>> +void dwc3_ep0_end_control_data(struct dwc3 *dwc, struct dwc3_ep *dep)
>>>>>>>>>>>>>>>  {
>>>>>>>>>>>>>>>  	struct dwc3_gadget_ep_cmd_params params;
>>>>>>>>>>>>>>>  	u32			cmd;
>>>>>>>>>>>>>>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>>>>>>>>>>>>>>> index 54c5a08..a0e2e4d 100644
>>>>>>>>>>>>>>> --- a/drivers/usb/dwc3/gadget.c
>>>>>>>>>>>>>>> +++ b/drivers/usb/dwc3/gadget.c
>>>>>>>>>>>>>>> @@ -2437,7 +2437,11 @@ static int dwc3_gadget_pullup(struct usb_gadget *g, int is_on)
>>>>>>>>>>>>>>>  				msecs_to_jiffies(DWC3_PULL_UP_TIMEOUT));
>>>>>>>>>>>>>>>  		if (ret == 0) {
>>>>>>>>>>>>>>>  			dev_err(dwc->dev, "timed out waiting for SETUP phase\n");
>>>>>>>>>>>>>>> -			return -ETIMEDOUT;
>>>>>>>>>>>>>>> +			spin_lock_irqsave(&dwc->lock, flags);
>>>>>>>>>>>>>>> +			dwc3_ep0_end_control_data(dwc, dwc->eps[0]);
>>>>>>>>>>>>>>> +			dwc3_ep0_end_control_data(dwc, dwc->eps[1]);
>>>>>>>>>>>>>>
>>>>>>>>>>>>>> End transfer command takes time, need to wait for it to complete before
>>>>>>>>>>>>>> issuing Start transfer again. Also, why restart again when it's about to
>>>>>>>>>>>>>> be disconnected.
>>>>>>>>>>>>>
>>>>>>>>>>>>> I can try without restarting it again, and see if that works.  Instead
>>>>>>>>>>>>> of waiting for the command complete event, can we set the ForceRM bit,
>>>>>>>>>>>>> similar to what we do for dwc3_remove_requests()?
>>>>>>>>>>>>>
>>>>>>>>>>>>
>>>>>>>>>>>> ForceRM=1 means that the controller will ignore updating the TRBs
>>>>>>>>>>>> (including not clearing the HWO and remain transfer size). The driver
>>>>>>>>>>>> still needs to wait for the command to complete before issuing Start
>>>>>>>>>>>> Transfer command. Otherwise Start Transfer won't go through. If we know
>>>>>>>>>>>> that we're not going to issue Start Transfer any time soon, then we may
>>>>>>>>>>>> be able to get away with ignoring End Transfer command completion.
>>>>>>>>>>>>
>>>>>>>>>>>
>>>>>>>>>>> I see.  Currently, in the place that we do use
>>>>>>>>>>> dwc3_ep0_end_control_data(), its followed by
>>>>>>>>>>> dwc3_ep0_stall_and_restart() which would execute start transfer.  For
>>>>>>>>>>
>>>>>>>>>> That doesn't look right. You can try to see if it can recover from a
>>>>>>>>>> control write request. Often time we do control read and not write.
>>>>>>>>>> (i.e. try to End Transfer and immediately Start Transfer on the same
>>>>>>>>>> direction control endpoint).
>>>>>>>>>>
>>>>>>>>> OK, I can try, but just to clarify, I was referring to how it was being
>>>>>>>>> done in:
>>>>>>>>>
>>>>>>>>> static void dwc3_ep0_xfernotready(struct dwc3 *dwc,
>>>>>>>>> 		const struct dwc3_event_depevt *event)
>>>>>>>>> {
>>>>>>>>> ...
>>>>>>>>> 		if (dwc->ep0_expect_in != event->endpoint_number) {
>>>>>>>>> 			struct dwc3_ep	*dep = dwc->eps[dwc->ep0_expect_in];
>>>>>>>>>
>>>>>>>>> 			dev_err(dwc->dev, "unexpected direction for Data Phase\n");
>>>>>>>>> 			dwc3_ep0_end_control_data(dwc, dep);
>>>>>>>>> 			dwc3_ep0_stall_and_restart(dwc);
>>>>>>>>> 			return;
>>>>>>>>> 		}
>>>>>>>>>
>>>>>>>
>>>>>>> Looking at this snippet again, it looks wrong. For control write
>>>>>>> unexpected direction, if the driver hasn't setup and started the DATA
>>>>>>> phase yet, then it's fine, but there is a problem if it did.
>>>>>>>
>>>>>>> Since dwc3_ep0_end_control_data() doesn't issue End Transfer command to
>>>>>>> ep0 due to the resource_index check, it doesn't follow the control
>>>>>>
>>>>>> IIRC resource_index is always non-zero, so the command should be
>>>>>
>>>>> No, resource_index for ep0out is 0, ep0in is 1. You can check from any
>>>>> of the driver tracepoint log for the return value of Start Transfer
>>>>> command for the resource index of ep0. There could be a mixed up with
>>>>> the undocumented return value of Set Endpoint Transfer Resource command
>>>>> before when this code was written, don't mix up with that.
>>>>>
>>>>>> triggered. If you have access to a Lecroy USB Trainer, could you script
>>>>>> this very scenario for verification?
>>>>>
>>>>> For anyone who wants to work on this, we don't need a LeCroy USB
>>>>> trainer. If you use xhci host, just modify the xhci-ring.c to queue a
>>>>> wrong direction DATA phase TRB of a particular control write request
>>>>> test, and continue with the next control requests.
>>>>>
>>>> Let me give this a try since I already have a modified (broken :)) XHCI
>>>> stack.
>>>>
>>>> Thanks
>>>> Wesley Cheng
>>>
>>> Sorry for the late response.  I was trying to get a reliable change to
>>> get the issue to reproduce.  I think I was able to find a set up which
>>> will generate the unexpected direction issue.  I'll try my best to
>>> summarize the traces here:
>>>
>>> Set up:
>>> - Modified XHCI stack to queue an IN TRB for a three stage CONTROL OUT
>>> transaction (OUT data stage)
>>> - Device is using RNDIS, as that has interface specific commands (ie
>>> SEND_ENCAPSUALTED messages)
>>>
>>> Kernel Log:
>>> [ 1255.312870] msm-usb-hsphy 88e3000.hsphy: Avail curr from USB = 900
>>> [ 1255.315300] dwc3_ep0_xfernotready event status = 1
>>> [ 1255.315429] dwc3_ep0_xfernotready event status = 2
>>> [ 1255.316390] android_work: sent uevent USB_STATE=CONFIGURED
>>> [ 1255.317467] dwc3_ep0_xfernotready event status = 1
>>> [ 1255.317588] dwc3_ep0_xfernotready event status = 2
>>> [ 1255.334196] dwc3_ep0_xfernotready event status = 1
>>> [ 1255.334217] dwc3 a600000.dwc3: unexpected direction for Data Phase
>>> [ 1255.334311] rndis_msg_parser: unknown RNDIS message 0x0052033A len
>>> 4456526
>>> [ 1255.334328] RNDIS command error -524, 0/24
>>> [ 1255.334369] ------------[ cut here ]------------
>>> [ 1255.334377] dwc3 a600000.dwc3: No resource for ep0out
>>> [ 1255.334440] WARNING: CPU: 0 PID: 8364 at
>>> drivers/usb/dwc3/gadget.c:360 dwc3_send_gadget_ep_cmd+0x3c4/0x96c
>>> ...
>>>  1255.336163] WARNING: CPU: 0 PID: 8364 at drivers/usb/dwc3/ep0.c:281
>>> dwc3_ep0_out_start+0x108/0x144
>>>
>>> So the kernel log does indeed show the concern mentioned by Thinh, where
>>> after ending the transfer, we do see the dwc3_ep0_out_start() fail due
>>> to no xfer resource.
>>>
>>
>> No, there's no end transfer command seen, and it's expected from code
>> review.
>>
> Sorry, yes that's correct.
>>> ftrace:
>>> <...>-8364    [000] d..1  1255.333988: dwc3_ctrl_req: Get Interface
>>> Status(Intf = 0, Length = 24)
>>> <...>-8364    [000] d..1  1255.334115: dwc3_prepare_trb: ep0out: trb
>>> ffffffc01bffd000 (E1:D0) buf 00000000efb76000 size 24 ctrl 00000455
>>> (HlCs:Sc:data)
>>> <...>-8364    [000] d..1  1255.334128: dwc3_prepare_trb: ep0out: trb
>>> ffffffc01bffd010 (E1:D0) buf 00000000efff9000 size 488 ctrl 00000c53
>>> (HLcs:SC:data)
>>> <...>-8364    [000] d..1  1255.334166: dwc3_gadget_ep_cmd: ep0out: cmd
>>> 'Start Transfer' [406] params 00000000 efffa000 00000000 --> status:
>>> Successful
>>> <...>-8364    [000] d..1  1255.334239: dwc3_gadget_ep_cmd: ep0out: cmd
>>> 'Set Stall' [404] params 00000000 00000000 00000000 --> status: Successful
>>> <...>-8364    [000] d..1  1255.334291: dwc3_gadget_giveback: ep0out: req
>>> ffffff8891724e00 length 0/24 zsI ==> -104
>>
>> It detected wrong direction and sets STALL here, but no End Transfer
>> command.
>>
>>> <...>-8364    [000] d..1  1255.334339: dwc3_prepare_trb: ep0out: trb
>>> ffffffc01bffd010 (E1:D0) buf 00000000efffa000 size 8 ctrl 00000c23
>>> (HLcs:SC:setup)
>>
>> The driver overwrote the active TRB with a new SETUP TRB.
>>
>>> <...>-8364    [000] d..1  1255.336099: dwc3_gadget_ep_cmd: ep0out: cmd
>>> 'Start Transfer' [406] params 00000000 efffa000 00000000 --> status: No
>>> Resource
>>
>> The dep->flags DWC3_EP_TRANSFER_STARTED got cleared. When it tries to
>> issue a Start Transfer a SETUP transfer, the command will fail with no
>> resource because the endpoint never ended properly.
>>
>>> <...>-8364    [000] d..1  1255.357594: dwc3_ctrl_req: 00 60 b7 ef 00 00
>>> 00 00
>>
>> Here is iffy because the behavior is undefined. The driver overwrote the
>> previous TRB. The Start Transfer command didn't go through, so the
>> controller still has the old TRB setup in its cache. It gets the next
>> SETUP request completion anyway because the SETUP stage is a short
>> packet. The driver updated its state that it's expecting the SETUP
>> stage, and it doesn't check the TRB write back buffer size for more or
>> less than 8 bytes or whether this is valid data.
>>
>> Wesley, is this SETUP packet the correct RNDIS control request?
>>
> That doesn't look to be a RNDIS control packet.  I collected a bus
> analyzer log as well w/ this snippet, and nothing was transmitted from
> the host side during the data stage.  Subsequent SETUP transactions were
> standard USB descriptors (GET string descriptors).

Ok.

>>> <...>-8364    [000] d..1  1255.357680: dwc3_gadget_ep_cmd: ep0out: cmd
>>> 'Set Stall' [404] params 00000000 00000000 00000000 --> status: Successful
>>
>> The device Set Stall on an unrecognized request, probably from the
>> application, which can be normal.
>>
>>> <...>-8364    [000] d..1  1255.357696: dwc3_prepare_trb: ep0out: trb
>>> ffffffc01bffd010 (E1:D0) buf 00000000efffa000 size 8 ctrl 00000c23
>>> (HLcs:SC:setup)
>>> <...>-8364    [000] d..1  1255.357722: dwc3_gadget_ep_cmd: ep0out: cmd
>>> 'Start Transfer' [406] params 00000000 efffa000 00000000 --> status:
>>> Successful
>>> <...>-224     [005] d..1  1266.313014: dwc3_gadget_ep_cmd: ep2out: cmd
>>> 'End Transfer' [40c08] params 00000000 00000000 00000000 --> status:
>>> Timed Out
>>>
>>> Thanks
>>> Wesley Cheng
>>>
>>
>> Thanks for the test Wesley
>>
>> BR,
>> Thinh
>>
> 
> So back to the original issue, which was the SETUP timeout during pullup
> disable, I went ahead and collected the ftrace w/ a change to just
> remove the return statement:
> 
> ret = wait_for_completion_timeout(&dwc->ep0_in_setup,
> 				msecs_to_jiffies(DWC3_PULL_UP_TIMEOUT));
> if (ret == 0) {
> 	dev_err(dwc->dev, "timed out waiting for SETUP phase\n");
> 	return -ETIMEDOUT;
> }
> 
> This would allow the __dwc3_gadget_stop() to disable EP0/1, which would
> call dwc3_gadget_stop_active_transfer() to send the end xfer command:
> 
> //Packet which will not queue a data stage (injected failure)
> <...>-7098    [003] d..1   346.560711: dwc3_ctrl_req: Get String
> Descriptor(Index = 3, Length = 2)
> 
> //Prepare IN data stage TRB
> <...>-7098    [003] d..1   346.560865: dwc3_prepare_trb: ep0in: trb
> ffffffc011edd000 (E0:D0) buf 00000000effcc000 size 2 ctrl 00000c53
> (HLcs:SC:data)
> <...>-7098    [003] d..1   346.560915: dwc3_gadget_ep_cmd: ep0in: cmd
> 'Start Transfer' [406] params 00000000 efffa000 00000000 --> status:
> Successful
> 
> //Pullup disable here kicks in - __dwc3_gadget_stop()
> <...>-224     [006] d..1   348.607367: dwc3_gadget_ep_disable: ep0out:
> mps 64/512 streams 0 burst 1 ring 0/0 flags E:swbp:>
> <...>-224     [006] d..1   348.607430: dwc3_gadget_giveback: ep0out: req
> ffffff884e5f5500 length 0/2 zsI ==> -108
> <...>-224     [006] d..1   348.607444: dwc3_gadget_ep_disable: ep0in:
> mps 64/512 streams 0 burst 1 ring 0/0 flags E:swBp:<
> 
> //End transfer on the pending EP0 in TRB queued previously
> <...>-224     [006] d..1   348.607484: dwc3_gadget_ep_cmd: ep0in: cmd
> 'End Transfer' [10c08] params 00000000 00000000 00000000 --> status:
> Successful
> <...>-224     [006] dN.1   348.607788: usb_gadget_vbus_draw: speed 3/6
> state 6 0mA [sg:self-powered:activated:disconnected] --> 0
> <...>-224     [006] ....   348.616888: usb_gadget_disconnect: speed 3/6
> state 6 0mA [sg:self-powered:activated:disconnected] --> 0
> 
> Does this look ok, Thinh?
> 

It should be fine.

The way we're handling soft-connect and soft-disconnect in dwc3 is
different than what the programming guide suggested. We're not doing
soft-reset on soft-connect. Our HW testing didn't show a problem in
quick soft-connect/disconnect succession before, but can you test doing
soft-connect immediately after this scenario?

Also, maybe we can change dev_err() print to dev_warn() on timeout instead?

Thanks,
Thinh