On Tue, Jul 22, 2025 at 03:50:01PM +0200, Andrew Lunn wrote:

On Tue, Jul 22, 2025 at 02:45:30PM +0800, Yibo Dong wrote:

quoted

On Mon, Jul 21, 2025 at 05:43:41PM +0200, Andrew Lunn wrote:

quoted

quoted

 #define MAX_VF_NUM (8)

quoted

+	hw->max_vfs = 7;

???

This is mistake, max vfs is 7. 8 is '7 vfs + 1 pf'.

So it seems like you need to add a new #define for MAX_FUNCS_NUM, and
set MAX_VF_NUM to 7. And then actually use MAX_VP_NUM. When reviewing
your own code, seeing the number 7, not a define, should of been a
warning, something is wrong....

Got it, I'll update this.

quoted

quoted

quoted

+static int mucse_obtain_mbx_lock_pf(struct mucse_hw *hw, enum MBX_ID mbx_id)
+{
+	struct mucse_mbx_info *mbx = &hw->mbx;
+	int try_cnt = 5000, ret;
+	u32 reg;
+
+	reg = (mbx_id == MBX_FW) ? PF2FW_MBOX_CTRL(mbx) :
+				   PF2VF_MBOX_CTRL(mbx, mbx_id);
+	while (try_cnt-- > 0) {
+		/* Take ownership of the buffer */
+		mbx_wr32(hw, reg, MBOX_PF_HOLD);
+		/* force write back before check */
+		wmb();
+		if (mbx_rd32(hw, reg) & MBOX_PF_HOLD)
+			return 0;
+		udelay(100);
+	}
+	return ret;

I've not compiled this, but isn't ret uninitialized here? I would also
expect it to return -ETIMEDOUT?

	Andrew

Yes, ret is uninitialized. I will fix this.

Did the compiler give a warning? Code should be warning free. We also
expect networking code to be W=1 warning free.

	Andrew

I can get this warning with 'make CC=clang-16 W=1' now.
I did't make with clang before, I'll add this step for future patches.

Thanks for your feedback.