usb/83863: Communication problem between opensc/openct via USB with e-gate smart-card.

Thu Aug 4 19:10:17 GMT 2005

The following reply was made to PR usb/83863; it has been noted by GNATS.

From: Hans Petter Selasky <hselasky at c2i.net>
To: Mohacsi Janos <mohacsi at niif.hu>
Cc: bug-followup at freebsd.org,
 janos.mohacsi at bsd.hu
Subject: Re: usb/83863: Communication problem between opensc/openct via USB with e-gate smart-card.
Date: Thu, 4 Aug 2005 21:04:19 +0200

 --Boundary-00=_1am8Czaqof+Hkbv
 Content-Type: text/plain;
   charset="iso-8859-1"
 Content-Transfer-Encoding: 7bit
 Content-Disposition: inline

 On Thursday 04 August 2005 16:39, Mohacsi Janos wrote:
 > Hi Hans,
 >   Here it is: - (Sorry, it is large... - should I omit
 > bug-followup  next time?)

 I think it is ok. It is not so much.

 Stalled means that the data-toggle has the wrong value. The data toggle can 
 have two values, either 1 or 0, and is marked "D=" in the debugging output. 
 Control transfers always start with D=0, and end with D=1.

 I cannot see anything wrong. The 8-bytes are the request header, and that is 
 accepted. Then the 5-bytes are the data to write, and that is not accepted.

 > uhci_check_transfer: xfer=0xc1bb3000 active Aug  4 15:48:10 scone kernel:
 > uhci_non_isoc_done: xfer=0xc1bb3000 pipe=0xc1ca3004 transfer done Aug  4
 > 15:48:10 scone kernel: uhci_dump_td: TD(0xc1bb31f0) at 002071f4 =
 > link=0x00207214 status=0x1c000007 token=0x00e07e2d buffer=0x002071d4 Aug  4
 > 15:48:10 scone kernel: uhci_dump_td: 207214<VF>
 > 1c000007<LS>,errcnt=3,actlen=8 pid=2d,addr=126,endpt=0,D=0,maxlen=8 Aug  4
 > 15:48:10 scone kernel: uhci_dump_td: TD(0xc1bb3210) at 00207214 =
 > link=0x00207234 status=0x3c400004 token=0x00887ee1 buffer=0x002071dc Aug  4
 > 15:48:10 scone kernel: uhci_dump_td: 207234<VF>
 > 3c400004<STALLED,LS,SPD>,errcnt=3,actlen=5
 > pid=e1,addr=126,endpt=0,D=1,maxlen=5 Aug  4 15:48:10 scone kernel:
 > uhci_dump_td: TD(0xc1bb3230) at 00207234 = link=0x00000001
 > status=0x3d8003ff token=0xffe87e69 buffer=0x00000000 Aug  4 15:48:10 scone
 > kernel: uhci_dump_td: 1<T> 3d8003ff<ACTIVE,IOC,LS,SPD>,errcnt=3,actlen=0
 > pid=69,addr=126,endpt=0,D=1,maxlen=0 Aug  4 15:48:10 scone kernel:
 > uhci_non_isoc_done: actlen=13
 > Aug  4 15:48:10 scone kernel: uhci_non_isoc_done: error, addr=126,
 > endpt=0x00, status 400000<STALLED> Aug  4 15:48:10 scone kernel:
 > uhci_device_done: xfer=0xc1bb3000, pipe=0xc1ca3004 length=13 error=22 Aug 
 > 4 15:48:10 scone kernel: _uhci_remove_qh: 0xc1bb3270 from 0xc1bb3270 Aug  4
 > 15:48:10 scone kernel: uhci_device_done: xfer=0xc1bb3000, pipe=0xc1ca3004
 > length=13 error=5 Aug  4 15:48:10 scone kernel: _uhci_remove_qh: 0xc1bb3270
 > from 0xdd6c5000 Aug  4 15:48:10 scone ifdhandler[808]: usb_do_request
 > failed: Input/output error (5) Aug  4 15:48:11 scone kernel:
 > uhci_root_intr_start: xfer=0xc1be6c00 len=32
 >
 >
 > 3. the usb handler routine  is attached sys-bsd.c. You will find the
 > usb_do_request ioctl in the function ifd_sysdep_usb_control() - I am not
 > sure usb device handled properly there - timeout? USBD_SHORT_XFER_OK
 > option?

 I think USBD_SHORT_XFER_OK is valid for control transfers.
 Setting the timeout is valid too.

 >
 > 4. The relevant part of the egate driver also included. For some reason
 > it does not like the EGATE_CMD_SEND_APDU (0x80):
 > ifd_usb_control() actually after some debug printing and setting timeout
 > to 10000 if <0, invokes ifd_sysdep_usb_control()
 >

 The best I can come up with, is that we issue a clear stall request on the 
 control endpoint, and then repeat the transfer. See if you can apply the 
 attached patch:

 cd directory_where_files_are
 cat sys-bsd.c.diff | patch

 Post debugging output again.

 --HPS

 --Boundary-00=_1am8Czaqof+Hkbv
 Content-Type: text/x-diff;
   charset="iso-8859-1";
   name="sys-bsd.c.diff"
 Content-Transfer-Encoding: 7bit
 Content-Disposition: attachment;
 	filename="sys-bsd.c.diff"

 *** sys-bsd.c.ref	Thu Aug  4 20:33:40 2005
 --- sys-bsd.c	Thu Aug  4 20:59:59 2005
 ***************
 *** 215,220 ****
 --- 215,251 ----
       return 0;
   }

 + static void
 + ifd_sysdep_usb_clearstall(ifd_device_t *dev, u_int8_t endpoint)
 + {
 + 	struct usb_ctl_request ctrl;
 + 	int rc, val;
 + 
 + 	bzero(&ctrl, sizeof(ctrl));
 + 
 +         ctrl.ucr_request.bmRequestType = UT_WRITE_ENDPOINT;
 +         ctrl.ucr_request.bRequest = UR_CLEAR_FEATURE;
 +         USETW(ctrl.ucr_request.wValue, UF_ENDPOINT_HALT);
 +         USETW(ctrl.ucr_request.wIndex, endpoint);
 +         USETW(ctrl.ucr_request.wLength, 0);
 + 
 + 	val = 1000;
 + 	if ((rc = ioctl(dev->fd, USB_SET_TIMEOUT, &val)) < 0) {
 + 		ifd_debug(1,"USB_SET_TIMEOUT failed: %d", rc);
 + 		ct_error("usb_set_timeout failed: %s(%d)",
 + 				strerror(errno), errno);
 +         	return;
 + 	}
 +  
 +         if ((rc = ioctl(dev->fd, USB_DO_REQUEST, &ctrl)) < 0) {
 + 	        ifd_debug(1, "USB_DO_REQUEST failed: %d", rc);
 +                 ct_error("usb_do_request failed: %s (%d)",
 + 			 strerror(errno), errno);
 +                 return;
 +         }
 +         return;
 + }
 + 
   /*
    * USB control command
    */
 ***************
 *** 228,233 ****
 --- 259,265 ----
   {
   	struct usb_ctl_request ctrl;
   	int		rc,val;
 + 	int count = 1;

   	ifd_debug(1, "BSD: ifd_sysdep_usb_control(0x%x)", request);
   	memset(&ctrl, 0, sizeof(ctrl));
 ***************
 *** 247,252 ****
 --- 279,286 ----
   	if(len)
   		ifd_debug(5, "BSD: CTRL SEND data %s", ct_hexdump(data,len));

 + repeat:
 +  
   	val = timeout;
   	if ((rc = ioctl(dev->fd, USB_SET_TIMEOUT, &val)) < 0) {
   		ifd_debug(1,"USB_SET_TIMEOUT failed: %d", rc);
 ***************
 *** 259,264 ****
 --- 293,307 ----
   		ifd_debug(1, "USB_DO_REQUEST failed: %d", rc);
   		ct_error("usb_do_request failed: %s (%d)",
   				strerror(errno), errno);
 + 		if(count)
 + 		{
 + 		    count--;
 + 
 + 		    ifd_debug(1, "trying to clear stall ...");
 + 
 + 		    ifd_sysdep_usb_clearstall(dev, 0);
 + 		    goto repeat;
 + 		}
   		return IFD_ERROR_COMM_ERROR;
           }

 --Boundary-00=_1am8Czaqof+Hkbv--