re: net: phy: continue searching for C45 MMDs even if first returned ffff:ffff

re: net: phy: continue searching for C45 MMDs even if first returned ffff:ffff

[Colin Ian King]

Hi,

Static analysis by Coverity has found a potential issue with the
following commit in /drivers/net/phy/phy_device.c:

commit bba238ed037c60242332dd1e4c5778af9eba4d9b
Author: Vladimir Oltean <vladimir.oltean@nxp.com>
Date:   Sun Jul 12 19:48:15 2020 +0300

    net: phy: continue searching for C45 MMDs even if first returned
ffff:ffff

The analysis is as follows:

735         * for 802.3 c45 complied PHYs, so don't probe it at first.
736         */

dead_error_condition: The condition (devs_in_pkg & 0x1fffffffU) ==
0x1fffffffU cannot be true.

737        for (i = 1; i < MDIO_MMD_NUM && devs_in_pkg == 0 &&

const: At condition (devs_in_pkg & 0x1fffffffU) == 0x1fffffffU, the
value of devs_in_pkg must be equal to 0.

738             (devs_in_pkg & 0x1fffffff) == 0x1fffffff; i++) {

Logically dead code (DEADCODE)dead_error_line: Execution cannot reach
this statement: if (i == 30 || i == 31) {

To summarize, if devs_in_pkg is zero, then (devs_in_pkg & 0x1fffffffU)
== 0x1fffffffU can never be true, so the loop is never iterated over.

Colin

Re: net: phy: continue searching for C45 MMDs even if first returned ffff:ffff

[Vladimir Oltean]

On Mon, Jul 20, 2020 at 05:21:13PM +0100, Colin Ian King wrote:
> Hi,
> 
> Static analysis by Coverity has found a potential issue with the
> following commit in /drivers/net/phy/phy_device.c:
> 
> commit bba238ed037c60242332dd1e4c5778af9eba4d9b
> Author: Vladimir Oltean <vladimir.oltean@nxp.com>
> Date:   Sun Jul 12 19:48:15 2020 +0300
> 
>     net: phy: continue searching for C45 MMDs even if first returned
> ffff:ffff
> 
> The analysis is as follows:
> 
> 735         * for 802.3 c45 complied PHYs, so don't probe it at first.
> 736         */
> 
> dead_error_condition: The condition (devs_in_pkg & 0x1fffffffU) ==
> 0x1fffffffU cannot be true.
> 
> 737        for (i = 1; i < MDIO_MMD_NUM && devs_in_pkg == 0 &&
> 
> const: At condition (devs_in_pkg & 0x1fffffffU) == 0x1fffffffU, the
> value of devs_in_pkg must be equal to 0.
> 
> 738             (devs_in_pkg & 0x1fffffff) == 0x1fffffff; i++) {
> 
> Logically dead code (DEADCODE)dead_error_line: Execution cannot reach
> this statement: if (i == 30 || i == 31) {
> 
> To summarize, if devs_in_pkg is zero, then (devs_in_pkg & 0x1fffffffU)
> == 0x1fffffffU can never be true, so the loop is never iterated over.
> 
> Colin

You are absolutely correct. The check should have been || and not &&.
I have a patch in my tree where I am fixing that. I was giving it some
more thorough testing to understand why it was working, though, and how
I could've missed it. One hypothesis I can't rule out is that I tested
it using || but submitted it using && somehow (although I don't remember
doing that).

Thanks,
-Vladimir