Re: [PATCH] net: dsa: mv88e6xxx: initialize return variable on declaration

On Mon, 6 Dec 2021 14:21:03 +0100
Andrew Lunn [off-list ref] wrote:

On Mon, Dec 06, 2021 at 04:32:19PM +0500, Ameer Hamza wrote:

quoted

Uninitialized err variable defined in mv88e6393x_serdes_power
function may cause undefined behaviour if it is called from
mv88e6xxx_serdes_power_down context.

Addresses-Coverity: 1494644 ("Uninitialized scalar variable")

Signed-off-by: Ameer Hamza <redacted>
---
 drivers/net/dsa/mv88e6xxx/serdes.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/net/dsa/mv88e6xxx/serdes.c b/drivers/net/dsa/mv88e6xxx/serdes.c
index 55273013bfb5..33727439724a 100644
--- a/drivers/net/dsa/mv88e6xxx/serdes.c
+++ b/drivers/net/dsa/mv88e6xxx/serdes.c

@@ -1507,7 +1507,7 @@ int mv88e6393x_serdes_power(struct mv88e6xxx_chip *chip, int port, int lane,
 			    bool on)
 {
 	u8 cmode = chip->ports[port].cmode;
-	int err;
+	int err = 0;
 
 	if (port != 0 && port != 9 && port != 10)
 		return -EOPNOTSUPP;  

Hi Marek

This warning likely comes from cmode not being a SERDES mode, and that
is not handles in the switch statementing. Do we want an

default:
	err = EINVAL;

?

	Andrew

Hi Andrew,

currently all the .serdes_power() methods return 0 for non-serdes ports.
This is because the way it is written, these methods are not called if
there is not a serdes lane for a given port.

For this issue with err variable undefined, to fix it we should simply
set int err=0 at the beginning of mv88e6393x_serdes_power(), to make it
behave like other serdes_power() methods do in serdes.c.



But a refactor may be needed for serdes_power() methods, at least
because they are a little weird. But it should be unrelated to this fix.

In serdes.h we have static inline functions
  mv88e6xxx_serdes_power_up(chip, port, lane)
  mv88e6xxx_serdes_power_down(chip, port, lane)

  (These simply call the serdes_power() method of chip ops, with
   additional boolean argument to specify powerup/powerdown.
   Also for these we first need to determine lane for a port. If lane
   does not exists, these should not be called.)

In chip.c we have function
  mv88e6xxx_serdes_power(chip, port, on)
  
  (This finds if the port has a lane, and if so, calls, if on=true
   mv88e6xxx_serdes_power_up()
     from serdes.h, and then
   mv88e6xxx_serdes_irq_request()
     also from serdes.h

   and if on=false, calls _irq_free() & _serdes_power_down()
  )

So if I call
  mv88e6xxx_serdes_power(chip, port, true)
it goes
    mv88e6xxx_serdes_power_up(chip, port, lane)
      chip->info->ops->serdes_power(chip, port, lane, true)
so the `on` argument is used in some places, but in other places there
are two functions instead.

Which I find a little weird.

Marek