From mboxrd@z Thu Jan  1 00:00:00 1970
From: Bjorn Helgaas <helgaas@kernel.org>
Subject: Re: [PATCH] pci: fix unavailable irq number 255 reported by BIOS
Date: Fri, 22 Jan 2016 11:53:32 -0600
Message-ID: <20160122175332.GA11004@localhost>
References: <1453167913-16248-1-git-send-email-chen.fan.fnst@cn.fujitsu.com>
 <6239489.QX2A2kAHWp@vostro.rjw.lan>
 <20160120002431.GA7973@localhost>
 <56A0EE00.80709@cn.fujitsu.com>
 <CAJZ5v0iq29mq_sFzk=mEbjPGJV9vPL9MCgU4WamrGV_fWcpEMA@mail.gmail.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Return-path: <linux-acpi-owner@vger.kernel.org>
Received: from mail.kernel.org ([198.145.29.136]:44825 "EHLO mail.kernel.org"
	rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP
	id S1754415AbcAVRxg (ORCPT <rfc822;linux-acpi@vger.kernel.org>);
	Fri, 22 Jan 2016 12:53:36 -0500
Content-Disposition: inline
In-Reply-To: <CAJZ5v0iq29mq_sFzk=mEbjPGJV9vPL9MCgU4WamrGV_fWcpEMA@mail.gmail.com>
Sender: linux-acpi-owner@vger.kernel.org
List-Id: linux-acpi@vger.kernel.org
To: "Rafael J. Wysocki" <rafael@kernel.org>
Cc: Cao jin <caoj.fnst@cn.fujitsu.com>, "Rafael J. Wysocki" <rjw@rjwysocki.net>, Chen Fan <chen.fan.fnst@cn.fujitsu.com>, ACPI Devel Maling List <linux-acpi@vger.kernel.org>, Linux Kernel Mailing List <linux-kernel@vger.kernel.org>, Len Brown <lenb@kernel.org>, izumi.taku@jp.fujitsu.com, wency@cn.fujitsu.com, Bjorn Helgaas <bhelgaas@google.com>, Linux PCI <linux-pci@vger.kernel.org>, Jiang Liu <jiang.liu@linux.intel.com>

On Thu, Jan 21, 2016 at 11:58:26PM +0100, Rafael J. Wysocki wrote:
> On Thu, Jan 21, 2016 at 3:41 PM, Cao jin <caoj.fnst@cn.fujitsu.com> wrote:
> > Hi,
> >
> >     IMHO, I think maybe modification on i801_smbus driver is easier.
> >
> >     Because when i801_smbus request_irq using pci_dev->irq, this
> > pci_dev->irq seems still holds the value read from register(
> > pci_setup_device->pci_read_irq), if the value is 255, it is invalid in
> > register,
> 
> Right.
> 
> Which is why the PCI core should not leak it into the driver's ->probe callback.

Is there a reserved IRQ value we could use to mean "invalid"?

I guess we have NR_IRQS as a ceiling, so the range of valid IRQs would be
[0 .. NR_IRQS - 1].  It looks like irq_desc() and a few drivers already
rely on NR_IRQS being the bound:

  lpc32xx_kscan_probe
  lpc32xx_nand_probe
  pcmcia_setup_isa_irq
  lpc32xx_rtc_probe
  apbuart_verify_port
  ar933x_uart_verify_port
  lqasc_verify_port

So I guess we could use ~0 as "invalid IRQ", and maybe the PCI core could
set dev->irq to ~0 in these cases, and drivers like i801_smbus could check
for that.  Maybe a wrapper like irq_valid() would be useful.

Bjorn