On Mon, Nov 27, 2023 at 02:37:41PM +0100, Andrew Lunn wrote:

quoted

 #define PHY_ID_LAN867X_REVB1 0x0007C162
+#define PHY_ID_LAN867X_REVC1 0x0007C164

So there is a gap in the revisions. Maybe a B2 exists?

The datasheet lists A0, B1 and C1, seems like Microchip removes the
application notes for old revisions, so no way that I can see to add the
init-fixup for A0.

I'm guessing there is a rev.c0 that was never released to the public.

quoted

+	const u16 magic_or = 0xE0;
+	const u16 magic_reg_mask = 0x1F;
+	const u16 magic_check_mask = 0x10;

Reverse christmass tree please. Longest first, shorted last.

My bad, I was just thinking 'christmas tree' forgot about the reverse.
I'll fix that.

quoted

+	int err;
+	int regval;
+	u16 override0;
+	u16 override1;
+	const u16 override_addr0 = 0x4;
+	const u16 override_addr1 = 0x8;
+	const u8 index_to_override0 = 2;
+	const u8 index_to_override1 = 3;

Same here.

I'll fix this.

quoted

+
+	err = lan867x_wait_for_reset_complete(phydev);
+	if (err)
+		return err;
+
+	/* The application note specifies a super convenient process
+	 * where 2 of the fixup regs needs a write with a value that is
+	 * a modified result of another reg read.
+	 * Enjoy the magic show.
+	 */

I really do hope that by revision D1 they get the firmware sorted out
so none of this undocumented magic is needed.

	Andrew

Really do hope so..