From: Markus Armbruster Subject: Re: [PATCH] Re: Re: [PATCH 2/2] PV framebuffer Date: Thu, 14 Dec 2006 13:37:25 +0100 Message-ID: <87ejr21wmy.fsf@pike.pond.sub.org> References: Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Return-path: In-Reply-To: (Keir Fraser's message of "Thu, 14 Dec 2006 11:30:13 +0000") List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Keir Fraser Cc: Atsushi SAKAI , xen-devel@lists.xensource.com, xen-ia64-devel@lists.xensource.com List-Id: xen-devel@lists.xenproject.org Keir Fraser writes: > On 14/12/06 10:58, "Atsushi SAKAI" wrote: > >> I searched the code and found >> the code in smp_call_function line337@linux/arch/ia64/kernel/smp.c >> WARN_ON(irqs_disabled()); >> >> The call trace problem caused by spin_lock_irqsave() >> in xenfb_update_screen@linux/drivers/xen/fbfront/xenfb.c Thanks for debugging this, Atsushi! >> Why you need to use spin_lock_irqsave in xenfb_update()? >> I understand the locking but why irqsave is required at this point? > > The lock is used in fb_ops, vm_ops and kthread contexts. None of these are > executed in IRQ context so irqsave is not needed. I'll remove it. I'm not > sure whether you can call zap_page_range() even with a spinlock held > though... > > -- Keir You can't. This was spotted just yesterday in our internal review. I'm working on a fix. If you could hold the removal of irqsave for a bit, that would avoid merge conflicts.