re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Dan Carpenter]

Hello Olivier Sobrie,

The patch 080f40a6fa28: "can: kvaser_usb: Add support for Kvaser
CAN/USB devices" from Nov 21, 2012, leads to the following static
checker warning:

	drivers/net/can/usb/kvaser_usb.c:949 kvaser_usb_rx_error()
	0x08 | 0x18 has 0x08 set on both sides

drivers/net/can/usb/kvaser_usb.c
   941          switch (dev->family) {
   942          case KVASER_LEAF:
   943                  if (es->leaf.error_factor) {
   944                          cf->can_id |= CAN_ERR_BUSERROR | CAN_ERR_PROT;
   945  
   946                          if (es->leaf.error_factor & M16C_EF_ACKE)
   947                                  cf->data[3] |= (CAN_ERR_PROT_LOC_ACK);
   948                          if (es->leaf.error_factor & M16C_EF_CRCE)
   949                                  cf->data[3] |= (CAN_ERR_PROT_LOC_CRC_SEQ |
   950                                                  CAN_ERR_PROT_LOC_CRC_DEL);

CAN_ERR_PROT_LOC_CRC_SEQ is 0x08
CAN_ERR_PROT_LOC_CRC_DEL is 0x18

It's weird that the bits overlap.  Was that intentional?  Why isn't it
enough to just say?:
					cf->data[3] |= CAN_ERR_PROT_LOC_CRC_DEL;

The static checker complains about most places where
CAN_ERR_PROT_LOC_CRC_SEQ is used.  I see us setting the flag but not I
don't see where we test for this so I'm not sure.

   951                          if (es->leaf.error_factor & M16C_EF_FORME)
   952                                  cf->data[2] |= CAN_ERR_PROT_FORM;
   953                          if (es->leaf.error_factor & M16C_EF_STFE)
   954                                  cf->data[2] |= CAN_ERR_PROT_STUFF;
   955                          if (es->leaf.error_factor & M16C_EF_BITE0)
   956                                  cf->data[2] |= CAN_ERR_PROT_BIT0;
   957                          if (es->leaf.error_factor & M16C_EF_BITE1)
   958                                  cf->data[2] |= CAN_ERR_PROT_BIT1;
   959                          if (es->leaf.error_factor & M16C_EF_TRE)
   960                                  cf->data[2] |= CAN_ERR_PROT_TX;
   961                  }
   962                  break;
   963          case KVASER_USBCAN:

regards,
dan carpenter

Re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Marc Kleine-Budde]

[-- Attachment #1: Type: text/plain, Size: 2695 bytes --]

On 11/19/2015 01:42 PM, Dan Carpenter wrote:
> Hello Olivier Sobrie,
> 
> The patch 080f40a6fa28: "can: kvaser_usb: Add support for Kvaser
> CAN/USB devices" from Nov 21, 2012, leads to the following static
> checker warning:
> 
> 	drivers/net/can/usb/kvaser_usb.c:949 kvaser_usb_rx_error()
> 	0x08 | 0x18 has 0x08 set on both sides
> 
> drivers/net/can/usb/kvaser_usb.c
>    941          switch (dev->family) {
>    942          case KVASER_LEAF:
>    943                  if (es->leaf.error_factor) {
>    944                          cf->can_id |= CAN_ERR_BUSERROR | CAN_ERR_PROT;
>    945  
>    946                          if (es->leaf.error_factor & M16C_EF_ACKE)
>    947                                  cf->data[3] |= (CAN_ERR_PROT_LOC_ACK);
>    948                          if (es->leaf.error_factor & M16C_EF_CRCE)
>    949                                  cf->data[3] |= (CAN_ERR_PROT_LOC_CRC_SEQ |
>    950                                                  CAN_ERR_PROT_LOC_CRC_DEL);
> 
> CAN_ERR_PROT_LOC_CRC_SEQ is 0x08
> CAN_ERR_PROT_LOC_CRC_DEL is 0x18
> 
> It's weird that the bits overlap.  Was that intentional?  Why isn't it
> enough to just say?:
> 					cf->data[3] |= CAN_ERR_PROT_LOC_CRC_DEL;

Looking at include/uapi/linux/can/error.h I'd say, they are not meant to
be used bitwise. Oliver?

> The static checker complains about most places where
> CAN_ERR_PROT_LOC_CRC_SEQ is used.  I see us setting the flag but not I
> don't see where we test for this so I'm not sure.
> 
>    951                          if (es->leaf.error_factor & M16C_EF_FORME)
>    952                                  cf->data[2] |= CAN_ERR_PROT_FORM;
>    953                          if (es->leaf.error_factor & M16C_EF_STFE)
>    954                                  cf->data[2] |= CAN_ERR_PROT_STUFF;
>    955                          if (es->leaf.error_factor & M16C_EF_BITE0)
>    956                                  cf->data[2] |= CAN_ERR_PROT_BIT0;
>    957                          if (es->leaf.error_factor & M16C_EF_BITE1)
>    958                                  cf->data[2] |= CAN_ERR_PROT_BIT1;
>    959                          if (es->leaf.error_factor & M16C_EF_TRE)
>    960                                  cf->data[2] |= CAN_ERR_PROT_TX;
>    961                  }
>    962                  break;
>    963          case KVASER_USBCAN:

Marc

-- 
Pengutronix e.K.                  | Marc Kleine-Budde           |
Industrial Linux Solutions        | Phone: +49-231-2826-924     |
Vertretung West/Dortmund          | Fax:   +49-5121-206917-5555 |
Amtsgericht Hildesheim, HRA 2686  | http://www.pengutronix.de   |


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 455 bytes --]

Re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Oliver Hartkopp]

> Marc Kleine-Budde <mkl@pengutronix.de> hat am 19. November 2015 um 14:23
> geschrieben:
> 
> 
> On 11/19/2015 01:42 PM, Dan Carpenter wrote:
> > Hello Olivier Sobrie,
> > 
> > The patch 080f40a6fa28: "can: kvaser_usb: Add support for Kvaser
> > CAN/USB devices" from Nov 21, 2012, leads to the following static
> > checker warning:
> > 
> > 	drivers/net/can/usb/kvaser_usb.c:949 kvaser_usb_rx_error()
> > 	0x08 | 0x18 has 0x08 set on both sides
> > 
> > drivers/net/can/usb/kvaser_usb.c
> >    941          switch (dev->family) {
> >    942          case KVASER_LEAF:
> >    943                  if (es->leaf.error_factor) {
> >    944                          cf->can_id |= CAN_ERR_BUSERROR |
> > CAN_ERR_PROT;
> >    945  
> >    946                          if (es->leaf.error_factor & M16C_EF_ACKE)
> >    947                                  cf->data[3] |=
> > (CAN_ERR_PROT_LOC_ACK);
> >    948                          if (es->leaf.error_factor & M16C_EF_CRCE)
> >    949                                  cf->data[3] |=
> > (CAN_ERR_PROT_LOC_CRC_SEQ |
> >    950
> >                                                  CAN_ERR_PROT_LOC_CRC_DEL);
> > 
> > CAN_ERR_PROT_LOC_CRC_SEQ is 0x08
> > CAN_ERR_PROT_LOC_CRC_DEL is 0x18
> > 
> > It's weird that the bits overlap.  Was that intentional?  Why isn't it
> > enough to just say?:
> > 					cf->data[3] |= CAN_ERR_PROT_LOC_CRC_DEL;
> 
> Looking at include/uapi/linux/can/error.h I'd say, they are not meant to
> be used bitwise. Oliver?

You are right!

The values reflect more or less the incident position (location) where the error
in bit stream processing occurred.

Instead of 

cf->data[3] |= (CAN_ERR_PROT_LOC_CRC_SEQ | ...

it should be

cf->data[3] = CAN_ERR_PROT_LOC_CRC_SEQ;

There's no need for bitwise OR'ed values in data[3].

Regards,
Oliver

Re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Olivier Sobrie]

Hello Dan,

On Thu, Nov 19, 2015 at 03:42:19PM +0300, Dan Carpenter wrote:
> Hello Olivier Sobrie,
> 
> The patch 080f40a6fa28: "can: kvaser_usb: Add support for Kvaser
> CAN/USB devices" from Nov 21, 2012, leads to the following static
> checker warning:
> 
> 	drivers/net/can/usb/kvaser_usb.c:949 kvaser_usb_rx_error()
> 	0x08 | 0x18 has 0x08 set on both sides
> 
> drivers/net/can/usb/kvaser_usb.c
>    941          switch (dev->family) {
>    942          case KVASER_LEAF:
>    943                  if (es->leaf.error_factor) {
>    944                          cf->can_id |= CAN_ERR_BUSERROR | CAN_ERR_PROT;
>    945  
>    946                          if (es->leaf.error_factor & M16C_EF_ACKE)
>    947                                  cf->data[3] |= (CAN_ERR_PROT_LOC_ACK);
>    948                          if (es->leaf.error_factor & M16C_EF_CRCE)
>    949                                  cf->data[3] |= (CAN_ERR_PROT_LOC_CRC_SEQ |
>    950                                                  CAN_ERR_PROT_LOC_CRC_DEL);
> 
> CAN_ERR_PROT_LOC_CRC_SEQ is 0x08
> CAN_ERR_PROT_LOC_CRC_DEL is 0x18
> 
> It's weird that the bits overlap.  Was that intentional?  Why isn't it
> enough to just say?:
> 					cf->data[3] |= CAN_ERR_PROT_LOC_CRC_DEL;
> 
> The static checker complains about most places where
> CAN_ERR_PROT_LOC_CRC_SEQ is used.  I see us setting the flag but not I
> don't see where we test for this so I'm not sure.

No it wasn't intentional. I think that it was inspired from other CAN
drivers. I see the same in net/can/c_can/c_can.c and
net/can/m_can/m_can.c. Sorry for the error.
Btw, which static checker are you using?

As suggested by Oliver, I assume we can transform this in
  cf->data[3] = CAN_ERR_PROT_LOC_CRC_SEQ;
and fix all the other places where bitwise operations are done
for errors in data[3].

Should I send a patch to fix this? Or do you or someone else plan
to send a patch?

Thanks,

Olivier

Re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Dan Carpenter]

Could you fix this and give me a reported-by tag?  I can't test it and
it's weird to me that CAN_ERR_PROT_LOC_CRC_SEQ is always used as a
bitmap.

The static checker is a Smatch thing I'm working on but haven't
published yet.  It only finds this bug and a couple false positives.

regards,
dan carpenter

Re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Oliver Hartkopp]

Hello Olivier,

On 11/20/2015 09:19 AM, Olivier Sobrie wrote:

> No it wasn't intentional. I think that it was inspired from other CAN
> drivers. I see the same in net/can/c_can/c_can.c and
> net/can/m_can/m_can.c. Sorry for the error.
> Btw, which static checker are you using?
> 
> As suggested by Oliver, I assume we can transform this in
>   cf->data[3] = CAN_ERR_PROT_LOC_CRC_SEQ;
> and fix all the other places where bitwise operations are done
> for errors in data[3].
> 
> Should I send a patch to fix this? Or do you or someone else plan
> to send a patch?

I already prepared a patch to fix the data[3] assignment issue.

It's a bunch of drivers that do it wrong:

drivers/net/can/c_can/c_can.c:
drivers/net/can/cc770/cc770.c:
drivers/net/can/flexcan.c:
drivers/net/can/m_can/m_can.c:
drivers/net/can/pch_can.c:
drivers/net/can/rcar_can.c:
drivers/net/can/ti_hecc.c:
drivers/net/can/usb/kvaser_usb.c:
drivers/net/can/usb/usb_8dev.c:
drivers/net/can/xilinx_can.c:

So if it's ok for you I'll post it this weekend :-)

Best regards,
Oliver

Re: can: kvaser_usb: Add support for Kvaser CAN/USB devices

[Olivier Sobrie]

On Sat, Nov 21, 2015 at 01:40:05PM +0100, Oliver Hartkopp wrote:
> Hello Olivier,
> 
> On 11/20/2015 09:19 AM, Olivier Sobrie wrote:
> 
> > No it wasn't intentional. I think that it was inspired from other CAN
> > drivers. I see the same in net/can/c_can/c_can.c and
> > net/can/m_can/m_can.c. Sorry for the error.
> > Btw, which static checker are you using?
> > 
> > As suggested by Oliver, I assume we can transform this in
> >   cf->data[3] = CAN_ERR_PROT_LOC_CRC_SEQ;
> > and fix all the other places where bitwise operations are done
> > for errors in data[3].
> > 
> > Should I send a patch to fix this? Or do you or someone else plan
> > to send a patch?
> 
> I already prepared a patch to fix the data[3] assignment issue.
> 
> It's a bunch of drivers that do it wrong:
> 
> drivers/net/can/c_can/c_can.c:
> drivers/net/can/cc770/cc770.c:
> drivers/net/can/flexcan.c:
> drivers/net/can/m_can/m_can.c:
> drivers/net/can/pch_can.c:
> drivers/net/can/rcar_can.c:
> drivers/net/can/ti_hecc.c:
> drivers/net/can/usb/kvaser_usb.c:
> drivers/net/can/usb/usb_8dev.c:
> drivers/net/can/xilinx_can.c:
> 
> So if it's ok for you I'll post it this weekend :-)

Of course it's ok for me :-)
Thank you,

Olivier