On Thu, Sep 07, 2023 at 02:16:35PM +0200, Maxime Chevallier wrote:

Hello Russell,

On Thu, 7 Sep 2023 11:00:24 +0100
"Russell King (Oracle)" [off-list ref] wrote:

quoted

On Thu, Sep 07, 2023 at 11:24:02AM +0200, Maxime Chevallier wrote:

quoted

+#define PHY_MAX_ENTRIES	16
+
+struct phy_list_reply_data {
+	struct ethnl_reply_data		base;
+	u8 n_phys;
+	u32 phy_indices[PHY_MAX_ENTRIES];  

Please could you detail the decision making behind 16 entries - is this
arbitary or based on something?

Also, please consider what we should do if we happen to have more than
16 entries.

Ah indeed it was totally arbitrary, the idea was to have a fixed-size
reply struct, so that we can populate the
ethnl_request_ops.reply_data_size field and not do any manual memory
management. But I can store a pointer to the array of phy devices,
dynamically allocated and we won't have to deal with this fixed,
arbitrary-sized array anymore.

I think Jakub already commented on this somewhere, but netlink should
allow for arbitrary long lists.

      Andrew