Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755804Ab0DASXY (ORCPT ); Thu, 1 Apr 2010 14:23:24 -0400 Received: from www.tglx.de ([62.245.132.106]:36251 "EHLO www.tglx.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1755185Ab0DASXT (ORCPT ); Thu, 1 Apr 2010 14:23:19 -0400 Date: Thu, 1 Apr 2010 20:23:14 +0200 (CEST) From: Thomas Gleixner To: Xianghua Xiao cc: linux-kernel@vger.kernel.org Subject: Re: [ANNOUNCE] 2.6.33.1-rt11 - BUG? In-Reply-To: Message-ID: References: User-Agent: Alpine 2.00 (LFD 1167 2008-08-23) MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 2741 Lines: 78 On Thu, 1 Apr 2010, Xianghua Xiao wrote: > I'm testing 2.6.33.1 on ppc 834x with rt11 patch. After system is up I keep > seeing this same message from dmesg: > > BUG: sleeping function called from invalid context at kernel/rtmutex.c:684 > pcnt: 0 0 in_atomic(): 0, irqs_disabled(): 1, pid: 16, name: events/0 > Call Trace: > [cf85be80] [c00096cc] show_stack+0x6c/0x1a4 (unreliable) > [cf85beb0] [c001f928] __might_sleep+0x104/0x108 > [cf85bec0] [c03cb444] rt_spin_lock+0xa0/0xa4 > [cf85bed0] [c0270b6c] lock_tx_qs+0x40/0x68 > [cf85bef0] [c02712cc] adjust_link+0x4c/0x240 > [cf85bf20] [c026d5ec] phy_state_machine+0x3bc/0x628 > [cf85bf40] [c003e69c] worker_thread+0x148/0x1f0 > [cf85bfa0] [c0042ec0] kthread+0x84/0x88 > [cf85bff0] [c00137f8] kernel_thread+0x4c/0x68 > BUG: sleeping function called from invalid context at kernel/rtmutex.c:684 > pcnt: 0 0 in_atomic(): 0, irqs_disabled(): 1, pid: 16, name: events/0 > > #uname -a > Linux 2.6.33.1-rt11 #3 PREEMPT RT Thu Apr 1 11:58:25 CDT 2010 ppc unknown > > This does not hang/lock the system yet...google reported similar 'bug' on > older kernels. Is this a bug from 2.6.33.1-rt11 patch or I have to change > phy.c somehow to avoid it? This traces back to a mutex lock in phy.c. No, it traces back to a call to lock_tx_qs() which is a spinlock in mainline and gets converted to a "sleeping" spinlock in -RT. That means it can't be called with interrupts disabled. But the code in adjust_link does exaclty that. Does the patch below fix it ? Thanks, tglx --- Subject: net-gianfar-fix-rt-splat.patch From: Thomas Gleixner Date: Thu, 01 Apr 2010 20:20:57 +0200 Signed-off-by: Thomas Gleixner --- drivers/net/gianfar.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) Index: linux-2.6-tip/drivers/net/gianfar.c =================================================================== --- linux-2.6-tip.orig/drivers/net/gianfar.c +++ linux-2.6-tip/drivers/net/gianfar.c @@ -2717,7 +2717,7 @@ static void adjust_link(struct net_devic struct phy_device *phydev = priv->phydev; int new_state = 0; - local_irq_save(flags); + local_irq_save_nort(flags); lock_tx_qs(priv); if (phydev->link) { @@ -2785,7 +2785,7 @@ static void adjust_link(struct net_devic if (new_state && netif_msg_link(priv)) phy_print_status(phydev); unlock_tx_qs(priv); - local_irq_restore(flags); + local_irq_restore_nort(flags); } /* Update the hash table based on the current list of multicast -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/