Skip to contentSkip to conversation

usbserial: fix inconsistent lock state

PATCH In review Borislav Petkov 2 participants 2 messages Oct 28, 2007

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 usbserial: fix inconsistent lock state

Hi,

i get

Oct 28 12:50:53 zmei kernel: [ 6656.113371] =================================
Oct 28 12:50:53 zmei kernel: [ 6656.113378] [ INFO: inconsistent lock state ]
Oct 28 12:50:53 zmei kernel: [ 6656.113382] 2.6.24-rc1 #13
Oct 28 12:50:53 zmei kernel: [ 6656.113384] ---------------------------------
Oct 28 12:50:53 zmei kernel: [ 6656.113387] inconsistent {hardirq-on-W} -> {in-hardirq-W} usage.
Oct 28 12:50:53 zmei kernel: [ 6656.113392] swapper/0 [HC1[1]:SC0[0]:HE0:SE1] takes:
Oct 28 12:50:53 zmei kernel: [ 6656.113395]  (&port->lock){+...}, at: [<e09f0658>] usb_serial_generic_read_bulk_callback+0xee/0x14c [usbserial]
Oct 28 12:50:53 zmei kernel: [ 6656.113413] {hardirq-on-W} state was registered at:
Oct 28 12:50:53 zmei kernel: [ 6656.113416]   [<c014220e>] __lock_acquire+0x44d/0xbdd
Oct 28 12:50:53 zmei kernel: [ 6656.113427]   [<c0142d81>] lock_acquire+0x5f/0x78
Oct 28 12:50:53 zmei kernel: [ 6656.113433]   [<c02f92be>] _spin_lock_bh+0x3a/0x47
Oct 28 12:50:53 zmei kernel: [ 6656.113441]   [<e09f08c0>] usb_serial_generic_write+0x8a/0x252 [usbserial]
Oct 28 12:50:53 zmei kernel: [ 6656.113452]   [<e09ee5e2>] serial_write+0xa3/0xb2 [usbserial]
Oct 28 12:50:53 zmei kernel: [ 6656.113462]   [<c021a6a0>] write_chan+0x1e4/0x2ac
Oct 28 12:50:53 zmei kernel: [ 6656.113471]   [<c02182f4>] tty_write+0x13e/0x1ba
Oct 28 12:50:53 zmei kernel: [ 6656.113478]   [<c016e466>] vfs_write+0xa8/0x131
Oct 28 12:50:53 zmei kernel: [ 6656.113486]   [<c016ea06>] sys_write+0x3d/0x61
Oct 28 12:50:53 zmei kernel: [ 6656.113493]   [<c010411a>] syscall_call+0x7/0xb
Oct 28 12:50:53 zmei kernel: [ 6656.113500]   [<ffffffff>] 0xffffffff
Oct 28 12:50:53 zmei kernel: [ 6656.113510] irq event stamp: 68037894
Oct 28 12:50:53 zmei kernel: [ 6656.113513] hardirqs last  enabled at (68037893): [<c0102b1f>] default_idle+0x3e/0x59
Oct 28 12:50:53 zmei kernel: [ 6656.113520] hardirqs last disabled at (68037894): [<c0104b20>] common_interrupt+0x24/0x34
Oct 28 12:50:53 zmei kernel: [ 6656.113526] softirqs last  enabled at (68037884): [<c012a56e>] __do_softirq+0xe5/0xed
Oct 28 12:50:53 zmei kernel: [ 6656.113535] softirqs last disabled at (68037873): [<c012a5af>] do_softirq+0x39/0x55
Oct 28 12:50:53 zmei kernel: [ 6656.113542] 
Oct 28 12:50:53 zmei kernel: [ 6656.113543] other info that might help us debug this:
Oct 28 12:50:53 zmei kernel: [ 6656.113546] no locks held by swapper/0.
Oct 28 12:50:53 zmei kernel: [ 6656.113548] 
Oct 28 12:50:53 zmei kernel: [ 6656.113549] stack backtrace:
Oct 28 12:50:53 zmei kernel: [ 6656.113553]  [<c010515b>] show_trace_log_lvl+0x1a/0x2f
Oct 28 12:50:53 zmei kernel: [ 6656.113559]  [<c0105b4d>] show_trace+0x12/0x14
Oct 28 12:50:53 zmei kernel: [ 6656.113566]  [<c0105c6d>] dump_stack+0x16/0x18
Oct 28 12:50:53 zmei kernel: [ 6656.113571]  [<c0140de6>] print_usage_bug+0x143/0x14d
Oct 28 12:50:53 zmei kernel: [ 6656.113576]  [<c01413cc>] mark_lock+0xc5/0x46a
Oct 28 12:50:53 zmei kernel: [ 6656.113580]  [<c01421a1>] __lock_acquire+0x3e0/0xbdd
Oct 28 12:50:53 zmei kernel: [ 6656.113585]  [<c0142d81>] lock_acquire+0x5f/0x78
Oct 28 12:50:53 zmei kernel: [ 6656.113590]  [<c02f9277>] _spin_lock+0x35/0x42
Oct 28 12:50:53 zmei kernel: [ 6656.113594]  [<e09f0658>] usb_serial_generic_read_bulk_callback+0xee/0x14c [usbserial]
Oct 28 12:50:53 zmei kernel: [ 6656.113603]  [<c024f161>] usb_hcd_giveback_urb+0x61/0x93
Oct 28 12:50:53 zmei kernel: [ 6656.113609]  [<e0846902>] ehci_urb_done+0x7d/0x8a [ehci_hcd]
Oct 28 12:50:53 zmei kernel: [ 6656.113617]  [<e0847897>] qh_completions+0x3fb/0x499 [ehci_hcd]
Oct 28 12:50:53 zmei kernel: [ 6656.113624]  [<e08479f6>] ehci_work+0xc1/0x6c7 [ehci_hcd]
Oct 28 12:50:53 zmei kernel: [ 6656.113630]  [<e084c0c9>] ehci_irq+0x24d/0x275 [ehci_hcd]
Oct 28 12:50:53 zmei kernel: [ 6656.113637]  [<c024fa75>] usb_hcd_irq+0x4f/0x58
Oct 28 12:50:53 zmei kernel: [ 6656.113641]  [<c014e459>] handle_IRQ_event+0x1a/0x46
Oct 28 12:50:53 zmei kernel: [ 6656.113648]  [<c014f53c>] handle_fasteoi_irq+0xab/0xb8
Oct 28 12:50:53 zmei kernel: [ 6656.113653]  [<c0106a9e>] do_IRQ+0x7c/0x96
Oct 28 12:50:53 zmei kernel: [ 6656.113657]  [<c0104b2a>] common_interrupt+0x2e/0x34
Oct 28 12:50:53 zmei kernel: [ 6656.113662]  [<c01025b3>] cpu_idle+0x9e/0xd3
Oct 28 12:50:53 zmei kernel: [ 6656.113666]  [<c02f6ea9>] rest_init+0x5d/0x5f
Oct 28 12:50:53 zmei kernel: [ 6656.113672]  [<c041393a>] start_kernel+0x2e3/0x2eb
Oct 28 12:50:53 zmei kernel: [ 6656.113679]  [<00000000>] 0x0
Oct 28 12:50:53 zmei kernel: [ 6656.113683]  =======================

<snip>

when inserting my UTMS usb modem and the following patch fixes it:

---
In commit acd2a847e7fee7df11817f67dba75a2802793e5d usb_serial_generic_write() disables interrupts when taking &port->lock which is also taken in usb_serial_generic_read_bulk_callback() resulting in an inconsistent lock state due to the latter not disabling interrupts on the local cpu. Fix that by disabling interrupts in the latter call site also.

--

drivers/usb/serial/generic.c +3 -2
index 9eb4a65..d415311 100644
@@ -327,6 +327,7 @@ void usb_serial_generic_read_bulk_callback (struct urb *urb)
 	struct usb_serial_port *port = (struct usb_serial_port *)urb->context;
 	unsigned char *data = urb->transfer_buffer;
 	int status = urb->status;
+	unsigned long flags;
 
 	dbg("%s - port %d", __FUNCTION__, port->number);
 
@@ -339,11 +340,11 @@ void usb_serial_generic_read_bulk_callback (struct urb *urb)
 	usb_serial_debug_data(debug, &port->dev, __FUNCTION__, urb->actual_length, data);
 
 	/* Throttle the device if requested by tty */
-	spin_lock(&port->lock);
+	spin_lock_irqsave(&port->lock, flags);
 	if (!(port->throttled = port->throttle_req))
 		/* Handle data and continue reading from device */
 		flush_and_resubmit_read_urb(port);
-	spin_unlock(&port->lock);
+	spin_unlock_irqrestore(&port->lock, flags);
 }
 EXPORT_SYMBOL_GPL(usb_serial_generic_read_bulk_callback);
 
Signature · 2 lines
Regards/Gruß,
    Boris.
Acked-by on v1 1/1
#1 lore raw
Quoted 7 lines · Borislav Petkov

---
In commit acd2a847e7fee7df11817f67dba75a2802793e5d usb_serial_generic_write() disables interrupts when taking &port->lock which is also taken in usb_serial_generic_read_bulk_callback() resulting in an inconsistent lock state due to the latter not disabling interrupts on the local cpu. Fix that by disabling interrupts in the latter call site also.

You are right, I missed the usb_serial_generic_read_bulk_callback() one, thanks for catching this.

Signature · 2 lines
Jiri Kosina
SUSE Labs