Re: netpoll: raspberrypi [4 5] driver locking woes

[Sean Anderson <sean.anderson@linux.dev>]

[-- Attachment #1: Type: text/plain, Size: 2058 bytes --]

Hi Mike,

On 8/28/25 13:26, Mike Galbraith wrote:
> On Thu, 2025-08-28 at 10:57 -0400, Sean Anderson wrote:
>> Hi Mike,
>> 
>> On 8/27/25 12:02, Mike Galbraith wrote:
>> > Unexpected addendum to done deal datapoint, so off list.
>> > 
>> > On Tue, 2025-08-26 at 11:49 +0200, Mike Galbraith wrote:
>> > > 
>> > > The pi5 gripe fix is equally trivial, but submitting that is pointless
>> > > given there's something else amiss in fingered commit.  This is all of
>> > > the crash info that escapes the box w/wo gripes silenced.
>> > > 
>> > > [   51.688868] sysrq: Trigger a crash
>> > > [   51.688892] Kernel panic - not syncing: sysrq triggered crash
>> > > [   51.698066] CPU: 0 UID: 0 PID: 0 Comm: swapper/0 Not tainted 6.17.0-rc2-v8-lockdep #533 PREEMPTLAZY
>> > > [   51.707234] Hardware name: Raspberry Pi 5 Model B Rev 1.0 (DT)
>> > > [   51.713085] Call trace:
>> > > [   51.715532]  show_stack+0x20/0x38 (C)
>> > > [   51.719206]  dump_stack_lvl+0x38/0xd0
>> > > [   51.722878]  dump_stack+0x18/0x28
>> > > 
>> > > That aspect is a punt and run atm (time.. and dash of laziness:).
>> > 
>> > Plan was to end datapoint thread, but after booting pi5's 6.12 kernel,
>> > for some reason I fired up netconsole.. and box promptly exhibited the
>> > netpoll locking bug warning, indicating presence of 138badbc21a0. 
>> > Instead of saying to self "nope, just walk away", I poked SysRq-C.. and
>> > the bloody damn monitoring box received a 100% complete death rattle. 
>> > Well bugger.
>> 
>> Did you get a backtrace for this?
> 
> Yes, logs for 6.12.41 and 6.17.0-rc2 attached.
> 
> Since a patch has meanwhile landed, also a log of patched 6.17.0-rc3
> now gripe free (yay) but with aforementioned broken output, followed by
> addition of the e6a532185daa revert to confirm it still cures that.
> 
>> And to be clear, the steps to reproduce this are to boot a kernel with
>> lockdep enabled with netconsole on macb and then hit sysrq?
> 
> Yup.

Looks like the tx completion path can also be called from netpoll. Can
you try the attached patch?

--Sean

[-- Attachment #2: 0001-net-macb-Fix-tx_ptr_lock-locking.patch --]
[-- Type: text/x-patch, Size: 4585 bytes --]

From 3fe439760955370b9f005ef2728b5fc9b14eeea5 Mon Sep 17 00:00:00 2001
From: Sean Anderson <sean.anderson@linux.dev>
Date: Thu, 28 Aug 2025 11:55:01 -0400
Subject: [PATCH net v2] net: macb: Fix tx_ptr_lock locking

macb_start_xmit can be called with bottom-halves disabled (e.g.
transmitting from softirqs) as well as with interrupts disabled (with
netpoll). Because of this, all other functions taking tx_ptr_lock must
disable IRQs, and macb_start_xmit must only re-enable IRQs if they
were already enabled.

Fixes: 138badbc21a0 ("net: macb: use NAPI for TX completion path")
Reported-by: Mike Galbraith <efault@gmx.de>
Signed-off-by: Sean Anderson <sean.anderson@linux.dev>
---

Changes in v2:
- Use irqsave/restore for all accesses, since they can also also be
  called from netpoll.

 drivers/net/ethernet/cadence/macb_main.c | 28 ++++++++++++++----------
 1 file changed, 16 insertions(+), 12 deletions(-)

diff --git a/drivers/net/ethernet/cadence/macb_main.c b/drivers/net/ethernet/cadence/macb_main.c
index 16d28a8b3b56..c769b7dbd3ba 100644
--- a/drivers/net/ethernet/cadence/macb_main.c
+++ b/drivers/net/ethernet/cadence/macb_main.c
@@ -1223,12 +1223,13 @@ static int macb_tx_complete(struct macb_queue *queue, int budget)
 {
 	struct macb *bp = queue->bp;
 	u16 queue_index = queue - bp->queues;
+	unsigned long flags;
 	unsigned int tail;
 	unsigned int head;
 	int packets = 0;
 	u32 bytes = 0;
 
-	spin_lock(&queue->tx_ptr_lock);
+	spin_lock_irqsave(&queue->tx_ptr_lock, flags);
 	head = queue->tx_head;
 	for (tail = queue->tx_tail; tail != head && packets < budget; tail++) {
 		struct macb_tx_skb	*tx_skb;
@@ -1291,7 +1292,7 @@ static int macb_tx_complete(struct macb_queue *queue, int budget)
 	    CIRC_CNT(queue->tx_head, queue->tx_tail,
 		     bp->tx_ring_size) <= MACB_TX_WAKEUP_THRESH(bp))
 		netif_wake_subqueue(bp->dev, queue_index);
-	spin_unlock(&queue->tx_ptr_lock);
+	spin_unlock_irqrestore(&queue->tx_ptr_lock, flags);
 
 	return packets;
 }
@@ -1707,8 +1708,9 @@ static void macb_tx_restart(struct macb_queue *queue)
 {
 	struct macb *bp = queue->bp;
 	unsigned int head_idx, tbqp;
+	unsigned long flags;
 
-	spin_lock(&queue->tx_ptr_lock);
+	spin_lock_irqsave(&queue->tx_ptr_lock, flags);
 
 	if (queue->tx_head == queue->tx_tail)
 		goto out_tx_ptr_unlock;
@@ -1720,19 +1722,20 @@ static void macb_tx_restart(struct macb_queue *queue)
 	if (tbqp == head_idx)
 		goto out_tx_ptr_unlock;
 
-	spin_lock_irq(&bp->lock);
+	spin_lock(&bp->lock);
 	macb_writel(bp, NCR, macb_readl(bp, NCR) | MACB_BIT(TSTART));
-	spin_unlock_irq(&bp->lock);
+	spin_unlock(&bp->lock);
 
 out_tx_ptr_unlock:
-	spin_unlock(&queue->tx_ptr_lock);
+	spin_unlock_irqrestore(&queue->tx_ptr_lock, flags);
 }
 
 static bool macb_tx_complete_pending(struct macb_queue *queue)
 {
 	bool retval = false;
+	unsigned long flags;
 
-	spin_lock(&queue->tx_ptr_lock);
+	spin_lock_irqsave(&queue->tx_ptr_lock, flags);
 	if (queue->tx_head != queue->tx_tail) {
 		/* Make hw descriptor updates visible to CPU */
 		rmb();
@@ -1740,7 +1743,7 @@ static bool macb_tx_complete_pending(struct macb_queue *queue)
 		if (macb_tx_desc(queue, queue->tx_tail)->ctrl & MACB_BIT(TX_USED))
 			retval = true;
 	}
-	spin_unlock(&queue->tx_ptr_lock);
+	spin_unlock_irqrestore(&queue->tx_ptr_lock, flags);
 	return retval;
 }
 
@@ -2308,6 +2311,7 @@ static netdev_tx_t macb_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	struct macb_queue *queue = &bp->queues[queue_index];
 	unsigned int desc_cnt, nr_frags, frag_size, f;
 	unsigned int hdrlen;
+	unsigned long flags;
 	bool is_lso;
 	netdev_tx_t ret = NETDEV_TX_OK;
 
@@ -2368,7 +2372,7 @@ static netdev_tx_t macb_start_xmit(struct sk_buff *skb, struct net_device *dev)
 		desc_cnt += DIV_ROUND_UP(frag_size, bp->max_tx_length);
 	}
 
-	spin_lock_bh(&queue->tx_ptr_lock);
+	spin_lock_irqsave(&queue->tx_ptr_lock, flags);
 
 	/* This is a hard error, log it. */
 	if (CIRC_SPACE(queue->tx_head, queue->tx_tail,
@@ -2392,15 +2396,15 @@ static netdev_tx_t macb_start_xmit(struct sk_buff *skb, struct net_device *dev)
 	netdev_tx_sent_queue(netdev_get_tx_queue(bp->dev, queue_index),
 			     skb->len);
 
-	spin_lock_irq(&bp->lock);
+	spin_lock(&bp->lock);
 	macb_writel(bp, NCR, macb_readl(bp, NCR) | MACB_BIT(TSTART));
-	spin_unlock_irq(&bp->lock);
+	spin_unlock(&bp->lock);
 
 	if (CIRC_SPACE(queue->tx_head, queue->tx_tail, bp->tx_ring_size) < 1)
 		netif_stop_subqueue(dev, queue_index);
 
 unlock:
-	spin_unlock_bh(&queue->tx_ptr_lock);
+	spin_unlock_irqrestore(&queue->tx_ptr_lock, flags);
 
 	return ret;
 }
-- 
2.35.1.1320.gc452695387.dirty