According to the specification a new “Diversion” header field should be inserted as the topmost Diversion header field in the message, that means before any other existing Diversion header field in the message. In addition to that, the add_diversion function can be called several times and each time it should insert the new Diversion header field as the topmost one.

In order to implement this, add_diversion function creates the anchor in data_lump lists as a static variable to ensure that the next call of the function will use the same anchor and would insert new Diversion headers before the one created in the previous execution. To my knowledge this is the only way of inserting the diversion header field before any other created in previous runs of the function.

The anchor kept this way is only valid for a single message and we have to invalidate it when another message is being processed. For this reason, the function also stores the id of the message in another static variable and compares the value of that variable with the id of the SIP message being processed. If they differ then the anchor will be invalidated and the function creates a new one.

The following code snippet shows the code that invalidates the anchor, new anchor will be created when the anchor variable is set to 0.

static inline int add_diversion_helper(struct sip_msg* msg, str* s)
{
    static struct lump* anchor = 0;
    static int msg_id = 0;

    if (msg_id != msg->id) {
        msg_id = msg->id;
        anchor = 0;
    }
...
}