drivers/net/phy/dp83867.c:167: possible bad if ?

6 messages, 3 authors, 2015-07-20 · open the first message on its own page

drivers/net/phy/dp83867.c:167: possible bad if ?

From: David Binderman <hidden>
Date: 2015-07-20 12:37:40

Hello there,

drivers/net/phy/dp83867.c:167:57: warning: logical ‘or’ of collectively exhaustive tests is always true [-Wlogical-op]

Source code is

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) ||
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {

Maybe

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) &&
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {


Regards

David Binderman

 		 	   		  

Re: drivers/net/phy/dp83867.c:167: possible bad if ?

From: Florian Fainelli <f.fainelli@gmail.com>
Date: 2015-07-20 17:30:32

Adding Dan,

On 20/07/15 05:37, David Binderman wrote:
Hello there,

drivers/net/phy/dp83867.c:167:57: warning: logical ‘or’ of collectively exhaustive tests is always true [-Wlogical-op]

Source code is

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) ||
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {

Maybe

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) &&
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {
Sounds like the former is the intended comparison that will make sure
that phydev->interface is between MODE_RGMII_ID and MODE_RGMII_RXID, and
not below or after.
-- 
Florian

Re: drivers/net/phy/dp83867.c:167: possible bad if ?

From: Dan Murphy <hidden>
Date: 2015-07-20 17:37:40

Florian

On 07/20/2015 12:28 PM, Florian Fainelli wrote:
Adding Dan,

On 20/07/15 05:37, David Binderman wrote:
quoted
Hello there,

drivers/net/phy/dp83867.c:167:57: warning: logical ‘or’ of collectively exhaustive tests is always true [-Wlogical-op]

Source code is

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) ||
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {

Maybe

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) &&
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {
Sounds like the former is the intended comparison that will make sure
that phydev->interface is between MODE_RGMII_ID and MODE_RGMII_RXID, and
not below or after.
That is correct.  The internal delay only needs to be set if this is declared
via the DT.  There can be one of 3 interface internal delay (RGMII_ID), RX internal delay (RGMII_RX_ID) or TX
internal delay (RGMII_TX_ID).

This internal delay is only applicable for RGMII and can be made specific to the
board.

The driver only needs to set the delay per the declaration in the DT.

Dan

-- 
------------------
Dan Murphy

Re: drivers/net/phy/dp83867.c:167: possible bad if ?

From: Florian Fainelli <f.fainelli@gmail.com>
Date: 2015-07-20 17:42:22

On 20/07/15 10:37, Dan Murphy wrote:
Florian

On 07/20/2015 12:28 PM, Florian Fainelli wrote:
quoted
Adding Dan,

On 20/07/15 05:37, David Binderman wrote:
quoted
Hello there,

drivers/net/phy/dp83867.c:167:57: warning: logical ‘or’ of collectively exhaustive tests is always true [-Wlogical-op]

Source code is

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) ||
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {

Maybe

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) &&
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {
Sounds like the former is the intended comparison that will make sure
that phydev->interface is between MODE_RGMII_ID and MODE_RGMII_RXID, and
not below or after.
That is correct.  The internal delay only needs to be set if this is declared
via the DT.  There can be one of 3 interface internal delay (RGMII_ID), RX internal delay (RGMII_RX_ID) or TX
internal delay (RGMII_TX_ID).
Sorry, I meant to write the latter instead of the former here, the
current code seems to be potentially too permissive, is not it? In that
case, it seems to me like David's proposed fix is relevant.
This internal delay is only applicable for RGMII and can be made specific to the
board.

The driver only needs to set the delay per the declaration in the DT.

Dan

-- 
Florian

Re: drivers/net/phy/dp83867.c:167: possible bad if ?

From: Dan Murphy <hidden>
Date: 2015-07-20 17:52:05

On 07/20/2015 12:40 PM, Florian Fainelli wrote:
On 20/07/15 10:37, Dan Murphy wrote:
quoted
Florian

On 07/20/2015 12:28 PM, Florian Fainelli wrote:
quoted
Adding Dan,

On 20/07/15 05:37, David Binderman wrote:
quoted
Hello there,

drivers/net/phy/dp83867.c:167:57: warning: logical ‘or’ of collectively exhaustive tests is always true [-Wlogical-op]

Source code is

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) ||
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {

Maybe

    if ((phydev->interface>= PHY_INTERFACE_MODE_RGMII_ID) &&
        (phydev->interface <= PHY_INTERFACE_MODE_RGMII_RXID)) {
Sounds like the former is the intended comparison that will make sure
that phydev->interface is between MODE_RGMII_ID and MODE_RGMII_RXID, and
not below or after.
That is correct.  The internal delay only needs to be set if this is declared
via the DT.  There can be one of 3 interface internal delay (RGMII_ID), RX internal delay (RGMII_RX_ID) or TX
internal delay (RGMII_TX_ID).
Sorry, I meant to write the latter instead of the former here, the
current code seems to be potentially too permissive, is not it? In that
case, it seems to me like David's proposed fix is relevant.
The proposed fix is logically correct.  And it matches the code for
phy_interface_is_rgmii with the '&&'.

Did you want me to patch and test?

Dan
quoted
This internal delay is only applicable for RGMII and can be made specific to the
board.

The driver only needs to set the delay per the declaration in the DT.

Dan

-- 
------------------
Dan Murphy

RE: drivers/net/phy/dp83867.c:167: possible bad if ?

From: David Binderman <hidden>
Date: 2015-07-20 18:00:26

Hello there,

----------------------------------------
The proposed fix is logically correct. And it matches the code for
phy_interface_is_rgmii with the '&&'.
As the original warning message indicates, I used compiler flag
-Wlogical-op to find this problem.

In order to detect this problem in future in your section of
the Linux kernel, it might be a wise idea to add compiler flag
-Wlogical-op to your builds.

If all goes well, this extra compiler flag should have nothing to say.

Just an idea.


Regards

David Binderman

 		 	   		  
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help