Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20200124144248.11719-1-yu.c.chen@intel.com> <20200124192743.GL4675@bombadil.infradead.org>
In-Reply-To: <20200124192743.GL4675@bombadil.infradead.org>
From:   Chen Yu <yu.chen.surf@gmail.com>
Date:   Tue, 28 Jan 2020 10:09:07 +0800
Message-ID: <CADjb_WTBuhtKpko4dSh3Rj9D3=ie3hyh-nU6ib-XsCu9br1XZQ@mail.gmail.com>
Subject: Re: [PATCH][RFC] PCI: Add "pci=blacklist_dev=" parameter to blacklist
 specific devices
To:     Matthew Wilcox <willy@infradead.org>
Cc:     Chen Yu <yu.c.chen@intel.com>, linux-pci@vger.kernel.org,
        Jonathan Corbet <corbet@lwn.net>,
        Bjorn Helgaas <bhelgaas@google.com>, linux-doc@vger.kernel.org,
        Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
        Len Brown <lenb@kernel.org>,
        "Rafael J. Wysocki" <rjw@rjwysocki.net>
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi Matthew,
Thanks for repy.
On Sat, Jan 25, 2020 at 5:01 AM Matthew Wilcox <willy@infradead.org> wrote:
>
> On Fri, Jan 24, 2020 at 10:42:48PM +0800, Chen Yu wrote:
> > It was found that on some platforms the bogus pci device might bring
> > troubles to the system. For example, on a MacBookPro the system could
> > not be power off or suspended due to internal pci resource confliction
> > between bogus pci device and [io 0x1804]. Another case is that, once
> > resumed from hibernation on a VM, the pci config space of a pci device
> > is corrupt.
> >
> > To narrow down and benefit future debugging for such kind of issues,
> > introduce the command line blacklist_dev=<vendor:device_id>> to blacklist
> > such pci devices thus they will not be scanned thus not visible after
> > bootup. For example,
> >
> >  pci.blacklist_dev=8086:293e
> >
> > forbid the audio device to be exposed to the OS.
>
> This feels really unsafe to me.  Just because Linux ignores the device
> doesn't mean the device will ignore I/O requests.  I think we should
> call this pci.disable_dev and clear the device's I/O Space Enable,
> Memory Space Enable and Bus Master Enable bits (in the Command register,
> config space offset 4).
Oh right, the BIOS might already has enabled Memory/IO space
in the config during boot up and thus has already claimed the resource range
for this pci device.
I'll summarize the problem I found currently in Bjorn's reply and let's
discuss it there.

Thanks,
chenyu