linux-kernel - [PATCH net v2] net: usbnet: Avoid potential RCU stall on LINK

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives
Hash Suite for Android: free password hash cracker in your pocket
[<prev] [next>] [day] [month] [year] [list]
Message-ID: <20250723102526.1305339-1-john.ernberg@actia.se>
Date: Wed, 23 Jul 2025 10:25:35 +0000
From: John Ernberg <john.ernberg@...ia.se>
To: Oliver Neukum <oneukum@...e.com>, Andrew Lunn <andrew+netdev@...n.ch>,
	"David S . Miller" <davem@...emloft.net>, Eric Dumazet <edumazet@...gle.com>,
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>
CC: Greg Kroah-Hartman <gregkh@...uxfoundation.org>, "netdev@...r.kernel.org"
	<netdev@...r.kernel.org>, "linux-usb@...r.kernel.org"
	<linux-usb@...r.kernel.org>, "linux-kernel@...r.kernel.org"
	<linux-kernel@...r.kernel.org>, John Ernberg <john.ernberg@...ia.se>,
	"stable@...r.kernel.org" <stable@...r.kernel.org>
Subject: [PATCH net v2] net: usbnet: Avoid potential RCU stall on LINK_CHANGE
 event

The Gemalto Cinterion PLS83-W modem (cdc_ether) is emitting confusing link
up and down events when the WWAN interface is activated on the modem-side.

Interrupt URBs will in consecutive polls grab:
* Link Connected
* Link Disconnected
* Link Connected

Where the last Connected is then a stable link state.

When the system is under load this may cause the unlink_urbs() work in
__handle_link_change() to not complete before the next usbnet_link_change()
call turns the carrier on again, allowing rx_submit() to queue new SKBs.

In that event the URB queue is filled faster than it can drain, ending up
in a RCU stall:

    rcu: INFO: rcu_sched detected expedited stalls on CPUs/tasks: { 0-.... } 33108 jiffies s: 201 root: 0x1/.
    rcu: blocking rcu_node structures (internal RCU debug):
    Sending NMI from CPU 1 to CPUs 0:
    NMI backtrace for cpu 0

    Call trace:
     arch_local_irq_enable+0x4/0x8
     local_bh_enable+0x18/0x20
     __netdev_alloc_skb+0x18c/0x1cc
     rx_submit+0x68/0x1f8 [usbnet]
     rx_alloc_submit+0x4c/0x74 [usbnet]
     usbnet_bh+0x1d8/0x218 [usbnet]
     usbnet_bh_tasklet+0x10/0x18 [usbnet]
     tasklet_action_common+0xa8/0x110
     tasklet_action+0x2c/0x34
     handle_softirqs+0x2cc/0x3a0
     __do_softirq+0x10/0x18
     ____do_softirq+0xc/0x14
     call_on_irq_stack+0x24/0x34
     do_softirq_own_stack+0x18/0x20
     __irq_exit_rcu+0xa8/0xb8
     irq_exit_rcu+0xc/0x30
     el1_interrupt+0x34/0x48
     el1h_64_irq_handler+0x14/0x1c
     el1h_64_irq+0x68/0x6c
     _raw_spin_unlock_irqrestore+0x38/0x48
     xhci_urb_dequeue+0x1ac/0x45c [xhci_hcd]
     unlink1+0xd4/0xdc [usbcore]
     usb_hcd_unlink_urb+0x70/0xb0 [usbcore]
     usb_unlink_urb+0x24/0x44 [usbcore]
     unlink_urbs.constprop.0.isra.0+0x64/0xa8 [usbnet]
     __handle_link_change+0x34/0x70 [usbnet]
     usbnet_deferred_kevent+0x1c0/0x320 [usbnet]
     process_scheduled_works+0x2d0/0x48c
     worker_thread+0x150/0x1dc
     kthread+0xd8/0xe8
     ret_from_fork+0x10/0x20

Get around the problem by delaying the carrier on to the scheduled work.

This needs a new flag to keep track of the necessary action.

The carrier ok check cannot be removed as it remains required for the
LINK_RESET event flow.

Fixes: 4b49f58fff00 ("usbnet: handle link change")
Cc: stable@...r.kernel.org
Signed-off-by: John Ernberg <john.ernberg@...ia.se>

---

I've been testing this quite aggressively over a night, and seems equally
stable to my first approach. I'm a little bit concerned that the bit stuff
can now race (although much smaller) in the opposite direction, that a
carrier off can occur between test_and_clear_bit() and the carrier on
action in the handler. Leaving the carrier on when it shouldn't be.

v2:
 - target tree in patch description.
 - Drop Ming Lei from address list as their address bounces.
 - Rework solution based on feedback by Jakub (let me know if you want a
     Suggested-by tag, if we're keeping this direction)

v1: https://lore.kernel.org/netdev/20250710085028.1070922-1-john.ernberg@actia.se/

Tested on 6.12.20 and forward ported. Stack trace from 6.12.20.
---
 drivers/net/usb/usbnet.c   | 11 ++++++++---
 include/linux/usb/usbnet.h |  1 +
 2 files changed, 9 insertions(+), 3 deletions(-)

diff --git a/drivers/net/usb/usbnet.c b/drivers/net/usb/usbnet.c
index c04e715a4c2a..bc1d8631ffe0 100644
--- a/drivers/net/usb/usbnet.c
+++ b/drivers/net/usb/usbnet.c
@@ -1122,6 +1122,9 @@ static void __handle_link_change(struct usbnet *dev)
 		 * tx queue is stopped by netcore after link becomes off
 		 */
 	} else {
+		if (test_and_clear_bit(EVENT_LINK_CARRIER_ON, &dev->flags))
+			netif_carrier_on(dev->net);
+
 		/* submitting URBs for reading packets */
 		tasklet_schedule(&dev->bh);
 	}
@@ -2009,10 +2012,12 @@ EXPORT_SYMBOL(usbnet_manage_power);
 void usbnet_link_change(struct usbnet *dev, bool link, bool need_reset)
 {
 	/* update link after link is reseted */
-	if (link && !need_reset)
-		netif_carrier_on(dev->net);
-	else
+	if (link && !need_reset) {
+		set_bit(EVENT_LINK_CARRIER_ON, &dev->flags);
+	} else {
+		clear_bit(EVENT_LINK_CARRIER_ON, &dev->flags);
 		netif_carrier_off(dev->net);
+	}
 
 	if (need_reset && link)
 		usbnet_defer_kevent(dev, EVENT_LINK_RESET);
diff --git a/include/linux/usb/usbnet.h b/include/linux/usb/usbnet.h
index 0b9f1e598e3a..4bc6bb01a0eb 100644
--- a/include/linux/usb/usbnet.h
+++ b/include/linux/usb/usbnet.h
@@ -76,6 +76,7 @@ struct usbnet {
 #		define EVENT_LINK_CHANGE	11
 #		define EVENT_SET_RX_MODE	12
 #		define EVENT_NO_IP_ALIGN	13
+#		define EVENT_LINK_CARRIER_ON	14
 /* This one is special, as it indicates that the device is going away
  * there are cyclic dependencies between tasklet, timer and bh
  * that must be broken
-- 
2.49.0