net: macb: gate TX stall watchdog on netif_carrier_ok, use warn_ratelimited

Commit 79dc190 ("net: macb: add TX stall watchdog as
defence-in-depth safety net") added a per-queue 1 Hz delayed_work
that fires macb_tx_restart() if tx_tail has not advanced.  Two
operational follow-ups:

1. Boot-time false positive.  Between macb_open() and link autoneg
   completion, queue->tx_head can advance from kernel-queued packets
   while tx_tail stays at 0 because no TCOMPs have arrived yet.  The
   first watchdog tick at +1000 ms then trips its head!=tail check
   and fires a spurious TX-stall warning + re-kick.  Observed at
   ~25% rate during fleet rolling reboots (6 of 24 nodes, all
   tail=0 head=5-7).  The re-kick itself is harmless on a healthy
   ring (macb_tx_restart() verifies TBQP vs head before re-asserting
   TSTART), but the warning is noise and confuses operators who
   read it as a real silent-stall event.

   Gate the stall check on netif_carrier_ok().  No carrier means no
   completion is possible, so skipping the check is safe; the
   watchdog still ticks so it picks up immediately once carrier
   comes up.

2. netdev_warn_once() limits visibility to one event per netdev
   lifetime, which is fine for a warn-once defensive log but defeats
   any operator accounting of real stall events on long-running
   nodes.  Switch to pr_warn_ratelimited() with an explicit
   netdev_name() prefix: bounded log noise, the events stay visible
   in the journal for accounting, and it avoids
   netdev_warn_ratelimited() which is not available in this tree.

This is the rpi-6.18.y portion of the v2 follow-up to the netdev RFC
submission; the mainline v2 is on lore:

  https://lore.kernel.org/netdev/cover.1777064117.git.lukasz@raczylo.com/T/

Co-authored-by: John Laur <johnlaur@gmail.com>
Signed-off-by: Lukasz Raczylo <lukasz@raczylo.com>

Author: lukaszraczylo

drivers/net/ethernet/cadence/macb_main.c

@@ -2075,6 +2075,16 @@ static void macb_tx_stall_watchdog(struct work_struct *work)
 	if (!netif_running(bp->dev))
 		return;
 
+	/* No carrier => no completion is possible.  Skip the stall
+	 * check (otherwise queue->tx_head can advance from kernel-
+	 * queued packets between macb_open() and link autoneg
+	 * completion while tx_tail stays unchanged, tripping a false
+	 * positive), but keep the watchdog ticking so it picks up
+	 * once carrier comes up.
+	 */
+	if (!netif_carrier_ok(bp->dev))
+		goto reschedule;
+
 	spin_lock_irqsave(&queue->tx_ptr_lock, flags);
 	cur_tail = queue->tx_tail;
 	cur_head = queue->tx_head;
@@ -2084,13 +2094,14 @@ static void macb_tx_stall_watchdog(struct work_struct *work)
 	spin_unlock_irqrestore(&queue->tx_ptr_lock, flags);
 
 	if (stalled) {
-		netdev_warn_once(bp->dev,
-				 "TX stall detected on queue %u (tail=%u head=%u); re-kicking TSTART\n",
-				 (unsigned int)(queue - bp->queues),
-				 cur_tail, cur_head);
+		pr_warn_ratelimited("%s: TX stall detected on queue %u (tail=%u head=%u); re-kicking TSTART\n",
+				    netdev_name(bp->dev),
+				    (unsigned int)(queue - bp->queues),
+				    cur_tail, cur_head);
 		macb_tx_restart(queue);
 	}
 
+reschedule:
 	schedule_delayed_work(&queue->tx_stall_watchdog_work,
 			      msecs_to_jiffies(MACB_TX_STALL_INTERVAL_MS));
 }