lists.openwall.net   lists  /  announce  owl-users  owl-dev  john-users  john-dev  passwdqc-users  yescrypt  popa3d-users  /  oss-security  kernel-hardening  musl  sabotage  tlsify  passwords  /  crypt-dev  xvendor  /  Bugtraq  Full-Disclosure  linux-kernel  linux-netdev  linux-ext4  linux-hardening  linux-cve-announce  PHC 
Open Source and information security mailing list archives
 
Hash Suite: Windows password security audit tool. GUI, reports in PDF.
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date:   Mon, 3 May 2021 19:45:44 -0700
From:   Wesley Cheng <wcheng@...eaurora.org>
To:     Thinh Nguyen <Thinh.Nguyen@...opsys.com>,
        "balbi@...nel.org" <balbi@...nel.org>,
        "gregkh@...uxfoundation.org" <gregkh@...uxfoundation.org>
Cc:     "linux-usb@...r.kernel.org" <linux-usb@...r.kernel.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        "jackp@...eaurora.org" <jackp@...eaurora.org>
Subject: Re: [PATCH v2] usb: dwc3: gadget: Avoid canceling current request for
 queuing error



On 5/3/2021 7:20 PM, Thinh Nguyen wrote:
> Hi,
> 
> Wesley Cheng wrote:
>> If an error is received when issuing a start or update transfer
>> command, the error handler will stop all active requests (including
>> the current USB request), and call dwc3_gadget_giveback() to notify
>> function drivers of the requests which have been stopped.  Avoid
>> having to cancel the current request which is trying to be queued, as
>> the function driver will handle the EP queue error accordingly.
>> Simply unmap the request as it was done before, and allow previously
>> started transfers to be cleaned up.
>>

Hi Thinh,

> 
> It looks like you're still letting dwc3 stopping and cancelling all the
> active requests instead letting the function driver doing the dequeue.
> 

Yeah, main issue isn't due to the function driver doing dequeue, but
having cleanup (ie USB request free) if there is an error during
usb_ep_queue().

The function driver in question at the moment is the f_fs driver in AIO
mode.  When async IO is enabled in the FFS driver, every time it queues
a packet, it will allocate a io_data struct beforehand.  If the
usb_ep_queue() fails it will free this io_data memory.  Problem is that,
since the DWC3 gadget calls the completion with -ECONNRESET, the FFS
driver will also schedule a work item (within io_data struct) to handle
the completion.  So you end up with a flow like below

allocate io_data (ffs)
 --> usb_ep_queue()
   --> __dwc3_gadget_kick_transfer()
   --> dwc3_send_gadget_ep_cmd(EINVAL)
   --> dwc3_gadget_ep_cleanup_cancelled_requests()
   --> dwc3_gadget_giveback(ECONNRESET)
ffs completion callback
queue work item within io_data
 --> usb_ep_queue returns EINVAL
ffs frees io_data
...

work scheduled
 --> NULL pointer/memory fault as io_data is freed

> BTW, what kinds of command and error do you see in your setup and for
> what type endpoint? I'm thinking of letting the function driver to
> dequeue the requests instead of letting dwc3 automatically
> ending/cancelling the queued requests. However, it's a bit tricky to do
> that if the error is -ETIMEDOUT since we're not sure if the controller
> had already cached the TRBs.
> 

Happens on bulk EPs so far, but I think it wouldn't matter as long as
its over the FFS interface. (and using async IO transfers)

> This seems to add more complexity and I don't have a good solution to
> it. Since you're already cancelling all the active request anyway, what
> do you think of always letting dwc3_gadget_ep_queue() to go through with
> success, but report failure through request completion?
> 

We do have something similar as well downstream (returning success
always on dwc3_gadget_ep_queue()) and its been working for us also.
Problem is we don't test the ISOC path much, so this is the only type of
EP that might come into question...

Coming up with a way to address the concerns you brought up was a bit
difficult as there were scenarios we needed to consider.  next_request()
doesn't always have to be the request being queued (even if ep queue
triggered it).  There was no easy way to determine if kick transfer was
due to ep queue, but even if there was, we'd need to remember the
previous point as well.

Thanks
Wesley Cheng

> Hi Felipe, can you also chime in?
> 
> Thanks,
> Thinh
> 
> 
>> Signed-off-by: Wesley Cheng <wcheng@...eaurora.org>
>> ---
>> Changes in v2:
>>  - Addressed feedback received by making sure the logic only applies to a req
>>    which is being queued, decrementing the enqueue pointer, and only clearing
>>    the HWO bit.
>>
>>  drivers/usb/dwc3/gadget.c | 75 +++++++++++++++++++++++++++++++++++++++++------
>>  1 file changed, 66 insertions(+), 9 deletions(-)
>>
>> diff --git a/drivers/usb/dwc3/gadget.c b/drivers/usb/dwc3/gadget.c
>> index dd80e5c..c8ddbe1 100644
>> --- a/drivers/usb/dwc3/gadget.c
>> +++ b/drivers/usb/dwc3/gadget.c
>> @@ -140,6 +140,29 @@ int dwc3_gadget_set_link_state(struct dwc3 *dwc, enum dwc3_link_state state)
>>  }
>>  
>>  /**
>> + * dwc3_ep_dec_trb - decrement a trb index.
>> + * @index: Pointer to the TRB index to increment.
>> + *
>> + * The index should never point to the link TRB. After decrementing,
>> + * if index is zero, wrap around to the TRB before the link TRB.
>> + */
>> +static void dwc3_ep_dec_trb(u8 *index)
>> +{
>> +	(*index)--;
>> +	if (*index < 0)
>> +		*index = DWC3_TRB_NUM - 1;
>> +}
>> +
>> +/**
>> + * dwc3_ep_dec_enq - decrement endpoint's enqueue pointer
>> + * @dep: The endpoint whose enqueue pointer we're decrementing
>> + */
>> +static void dwc3_ep_dec_enq(struct dwc3_ep *dep)
>> +{
>> +	dwc3_ep_dec_trb(&dep->trb_enqueue);
>> +}
>> +
>> +/**
>>   * dwc3_ep_inc_trb - increment a trb index.
>>   * @index: Pointer to the TRB index to increment.
>>   *
>> @@ -1352,7 +1375,26 @@ static int dwc3_prepare_trbs(struct dwc3_ep *dep)
>>  
>>  static void dwc3_gadget_ep_cleanup_cancelled_requests(struct dwc3_ep *dep);
>>  
>> -static int __dwc3_gadget_kick_transfer(struct dwc3_ep *dep)
>> +static void dwc3_gadget_ep_revert_trbs(struct dwc3_ep *dep, struct dwc3_request *req)
>> +{
>> +	int i;
>> +
>> +	if (!req->trb)
>> +		return;
>> +
>> +	for (i = 0; i < req->num_trbs; i++) {
>> +		struct dwc3_trb *trb;
>> +
>> +		trb = &dep->trb_pool[dep->trb_enqueue];
>> +		trb->ctrl &= ~DWC3_TRB_CTRL_HWO;
>> +		dwc3_ep_dec_enq(dep);
>> +	}
>> +
>> +	req->num_trbs = 0;
>> +}
>> +
>> +static int __dwc3_gadget_kick_transfer(struct dwc3_ep *dep,
>> +				       struct dwc3_request *queued_req)
>>  {
>>  	struct dwc3_gadget_ep_cmd_params params;
>>  	struct dwc3_request		*req;
>> @@ -1410,8 +1452,23 @@ static int __dwc3_gadget_kick_transfer(struct dwc3_ep *dep)
>>  
>>  		dwc3_stop_active_transfer(dep, true, true);
>>  
>> -		list_for_each_entry_safe(req, tmp, &dep->started_list, list)
>> -			dwc3_gadget_move_cancelled_request(req, DWC3_REQUEST_STATUS_DEQUEUED);
>> +		/*
>> +		 * In order to ensure the logic is applied to a request being
>> +		 * queued by dwc3_gadget_ep_queue(), it needs to explicitly
>> +		 * check that req is the same as queued_req (request being
>> +		 * queued).  If so, then just unmap and decrement the enqueue
>> +		 * pointer, as the usb_ep_queue() error handler in the function
>> +		 * driver will handle cleaning up the USB request.
>> +		 */
>> +		list_for_each_entry_safe(req, tmp, &dep->started_list, list) {
>> +			if (req == queued_req) {
>> +				dwc3_gadget_ep_revert_trbs(dep, req);
>> +				dwc3_gadget_del_and_unmap_request(dep, req, ret);
>> +			} else {
>> +				dwc3_gadget_move_cancelled_request(req,
>> +								   DWC3_REQUEST_STATUS_DEQUEUED);
>> +			}
>> +		}
>>  
>>  		/* If ep isn't started, then there's no end transfer pending */
>>  		if (!(dep->flags & DWC3_EP_END_TRANSFER_PENDING))
>> @@ -1546,7 +1603,7 @@ static int dwc3_gadget_start_isoc_quirk(struct dwc3_ep *dep)
>>  	dep->start_cmd_status = 0;
>>  	dep->combo_num = 0;
>>  
>> -	return __dwc3_gadget_kick_transfer(dep);
>> +	return __dwc3_gadget_kick_transfer(dep, NULL);
>>  }
>>  
>>  static int __dwc3_gadget_start_isoc(struct dwc3_ep *dep)
>> @@ -1593,7 +1650,7 @@ static int __dwc3_gadget_start_isoc(struct dwc3_ep *dep)
>>  	for (i = 0; i < DWC3_ISOC_MAX_RETRIES; i++) {
>>  		dep->frame_number = DWC3_ALIGN_FRAME(dep, i + 1);
>>  
>> -		ret = __dwc3_gadget_kick_transfer(dep);
>> +		ret = __dwc3_gadget_kick_transfer(dep, NULL);
>>  		if (ret != -EAGAIN)
>>  			break;
>>  	}
>> @@ -1684,7 +1741,7 @@ static int __dwc3_gadget_ep_queue(struct dwc3_ep *dep, struct dwc3_request *req)
>>  		}
>>  	}
>>  
>> -	return __dwc3_gadget_kick_transfer(dep);
>> +	return __dwc3_gadget_kick_transfer(dep, req);
>>  }
>>  
>>  static int dwc3_gadget_ep_queue(struct usb_ep *ep, struct usb_request *request,
>> @@ -1893,7 +1950,7 @@ int __dwc3_gadget_ep_set_halt(struct dwc3_ep *dep, int value, int protocol)
>>  
>>  		if ((dep->flags & DWC3_EP_DELAY_START) &&
>>  		    !usb_endpoint_xfer_isoc(dep->endpoint.desc))
>> -			__dwc3_gadget_kick_transfer(dep);
>> +			__dwc3_gadget_kick_transfer(dep, NULL);
>>  
>>  		dep->flags &= ~DWC3_EP_DELAY_START;
>>  	}
>> @@ -2992,7 +3049,7 @@ static bool dwc3_gadget_endpoint_trbs_complete(struct dwc3_ep *dep,
>>  		(list_empty(&dep->pending_list) || status == -EXDEV))
>>  		dwc3_stop_active_transfer(dep, true, true);
>>  	else if (dwc3_gadget_ep_should_continue(dep))
>> -		if (__dwc3_gadget_kick_transfer(dep) == 0)
>> +		if (__dwc3_gadget_kick_transfer(dep, NULL) == 0)
>>  			no_started_trb = false;
>>  
>>  out:
>> @@ -3106,7 +3163,7 @@ static void dwc3_gadget_endpoint_command_complete(struct dwc3_ep *dep,
>>  
>>  	if ((dep->flags & DWC3_EP_DELAY_START) &&
>>  	    !usb_endpoint_xfer_isoc(dep->endpoint.desc))
>> -		__dwc3_gadget_kick_transfer(dep);
>> +		__dwc3_gadget_kick_transfer(dep, NULL);
>>  
>>  	dep->flags &= ~DWC3_EP_DELAY_START;
>>  }
>>
> 

-- 
The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
a Linux Foundation Collaborative Project

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ