From mboxrd@z Thu Jan 1 00:00:00 1970 From: Jan Kiszka Subject: Re: [PATCH v3 2/4] genirq: Inform handler about line sharing state Date: Fri, 17 Dec 2010 11:48:45 +0100 Message-ID: <4D0B400D.8010903@siemens.com> References: <4D0A75E3.3090900@web.de> <4D0B1CD9.5060601@web.de> <4D0B3C11.4090307@siemens.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: Jan Kiszka , Avi Kivity , Marcelo Tosatti , "linux-kernel@vger.kernel.org" , kvm , Tom Lyon , Alex Williamson , "Michael S. Tsirkin" To: Thomas Gleixner Return-path: In-Reply-To: Sender: linux-kernel-owner@vger.kernel.org List-Id: kvm.vger.kernel.org Am 17.12.2010 11:41, Thomas Gleixner wrote: > On Fri, 17 Dec 2010, Jan Kiszka wrote: >> Am 17.12.2010 11:23, Thomas Gleixner wrote: >>> OTOH, if we have to disable anyway, then we could simply keep it >>> disabled across the installation of a new handler. That would make the >>> notification business go away, wouldn't it ? >> >> No, the notification is still necessary in case the registered handler >> keeps the line off after returning from both hard and threaded handler. > > And how should that happen? If it is in oneshot mode then the line is > reenabled when the thread handler returns. disable_irq_nosync is called by the handler before returning. And it's the handler's job to revert this, properly synchronizing it internally. Jan -- Siemens AG, Corporate Technology, CT T DE IT 1 Corporate Competence Center Embedded Linux