Re: [PATCH 2/2] arm/virt/acpi: remove _ADR from devices identified by _HID

From: Igor Mammedov <imammedo@redhat.com>
To: Guoheyi <guoheyi@huawei.com>
Cc: Peter Maydell <peter.maydell@linaro.org>,
	"Michael S. Tsirkin" <mst@redhat.com>,
	qemu-devel@nongnu.org, Shannon Zhao <shannon.zhaosl@gmail.com>,
	qemu-arm@nongnu.org, Corey Minyard <cminyard@mvista.com>,
	wanghaibin.wang@huawei.com
Subject: Re: [PATCH 2/2] arm/virt/acpi: remove _ADR from devices identified by _HID
Date: Thu, 16 Jan 2020 14:25:08 +0100	[thread overview]
Message-ID: <20200116142508.1d82af31@redhat.com> (raw)
In-Reply-To: <a4992b63-e8e7-7f54-341e-f7dd3d7e8880@huawei.com>

On Thu, 16 Jan 2020 19:56:19 +0800
Guoheyi <guoheyi@huawei.com> wrote:

> 在 2020/1/5 20:53, Michael S. Tsirkin 写道:
> > On Sun, Jan 05, 2020 at 07:34:01AM -0500, Michael S. Tsirkin wrote:  
> >> On Thu, Dec 19, 2019 at 02:47:59PM +0800, Heyi Guo wrote:  
> >>> According to ACPI spec, _ADR should be used for device which is on a
> >>> bus that has a standard enumeration algorithm. It does not make sense
> >>> to have a _ADR object for devices which already have _HID and will be
> >>> enumerated by OSPM.
> >>>
> >>> Signed-off-by: Heyi Guo <guoheyi@huawei.com>  
> >> Are you sure? I would think this depends on the ID and the device
> >> really. E.g. PCI devices all are expected to have _ADR and some of them
> >> have a _HID.  
> >
> > To clarify I am not commenting on patches.
> > The spec says this:
> > 	6.1.5 _HID (Hardware ID)
> >
> > 	This object is used to supply OSPM with the device’s PNP ID or ACPI ID. 1
> >
> > 	When describing a platform, use of any _HID objects is optional. However, a _HID object must be
> >
> > 	used to describe any device that will be enumerated by OSPM. OSPM only enumerates a device
> >
> > 	when no bus enumerator can detect the device ID. For example, devices on an ISA bus are
> >
> > 	enumerated by OSPM. Use the _ADR object to describe devices enumerated by bus enumerators
> >
> > 	other than OSPM.
> >
> >
> > Note: "detect the device ID" not "enumerate the device" which I think
> > means there's a driver matching this vendor/device ID.
> >
> > So it seems fine to have _ADR so device is enumerated, and still have
> > _HID e.g. so ACPI driver can be loaded as fallback if there's
> > no bus driver.
> >
> >
> > Note I am not saying the patch itself is not correct.
> > Maybe these devices are not on any standard bus and that
> > is why they should not have _ADR? I have not looked.
> >
> > I am just saying that spec does not seem to imply _HID and _ADR
> > can't coexist.  
> 
> More reading on the spec, I found a statement as below 
> (https://uefi.org/sites/default/files/resources/ACPI_6_3_May16.pdf, 
> section 6.1, on top of page 343):

I'd replace 'It does not make sense ...' sentence with pointer to spec
and quote below in commit message.

> A device object must contain either an _HID object or an _ADR object, 
> but should not contain both

[...]