Re: [PATCHv3 1/4] net: phy: decouple PHY id and PHY address in fixed PHY driver

On Tue, 4 Mar 2014 20:04:24 +0100, Thomas Petazzoni [off-list ref] wrote:

Dear Florian Fainelli,

On Tue, 4 Mar 2014 10:43:12 -0800, Florian Fainelli wrote:

quoted

quoted

 struct fixed_phy {
-       int id;
+       int addr;
        u16 regs[MII_REGS_NUM];
        struct phy_device *phydev;
        struct fixed_phy_status status;

@@ -104,8 +104,8 @@ static int fixed_phy_update_regs(struct fixed_phy *fp)
        if (fp->status.asym_pause)
                lpa |= LPA_PAUSE_ASYM;

-       fp->regs[MII_PHYSID1] = fp->id >> 16;
-       fp->regs[MII_PHYSID2] = fp->id;
+       fp->regs[MII_PHYSID1] = 0xdead;
+       fp->regs[MII_PHYSID2] = 0xbeef;

I am still scratching my head as to whether we want that change to be
in this particular version of changes, or if we want that to happen at
a later time when (if?) we can finally get some proper OUI number
allocation.

Technically we are presenting some sort of ABI to user-space, although
detecting a fixed PHY device by reading its MII_PHYSID1/2 and matching
it against its PHY address on the fixed MDIO bus would have been a
little "weak" (especially when you can check that the parent device in
sysfs is the fixed-0 bus).

Well the problem is that fp->id really isn't an id, it's the fake
address of the PHY on the fake fixed MDIO bus. So it would mean that
the MII_PHYSID of the first fixed PHY would probably be 0x0, then the
second would have 0x1, then 0x2, and so on.

I think the change is reasonable, but I'm not thrilled with the 0xdeadbeaf
constant. I'd rather PHYSID simply be left as 0.

g.