[Qemu-devel] IO port 0x61 is multiple used

[Qemu-devel] IO port 0x61 is multiple used

[Weidong Huang]

Hi ALL:

IO port 0x61 is used by pc speaker and NMI reason port.

This is ambiguous. How to distinguish in QEMU?

I see that NMI reason port is not realized in QEMU.

But linux guest will access this port to determine what the NMI reason code means.

It will access pc speaker's io port. Is this a bug?

Re: [Qemu-devel] IO port 0x61 is multiple used

[Gonglei]

On 2015/3/15 20:49, Weidong Huang wrote:
> Hi ALL:
> 
> IO port 0x61 is used by pc speaker and NMI reason port.
> 
> This is ambiguous. How to distinguish in QEMU?
> 
> I see that NMI reason port is not realized in QEMU.
> 
> But linux guest will access this port to determine what the NMI reason code means.
> 
> It will access pc speaker's io port. Is this a bug?
> 
Gerd, any thoughts?

Thanks,
-Gonglei

Re: [Qemu-devel] IO port 0x61 is multiple used

[Gerd Hoffmann]

On Di, 2015-03-17 at 15:11 +0800, Gonglei wrote:
> On 2015/3/15 20:49, Weidong Huang wrote:
> > Hi ALL:
> > 
> > IO port 0x61 is used by pc speaker and NMI reason port.
> > 
> > This is ambiguous. How to distinguish in QEMU?
> > 
> > I see that NMI reason port is not realized in QEMU.
> > 
> > But linux guest will access this port to determine what the NMI reason code means.
> > 
> > It will access pc speaker's io port. Is this a bug?
> > 
> Gerd, any thoughts?

Not my area of expertise, but can qemu send NMIs to the guest in the
first place?

cheers,
  Gerd

Re: [Qemu-devel] IO port 0x61 is multiple used

[Jan Kiszka]

On 2015-03-19 12:41, Gerd Hoffmann wrote:
> On Di, 2015-03-17 at 15:11 +0800, Gonglei wrote:
>> On 2015/3/15 20:49, Weidong Huang wrote:
>>> Hi ALL:
>>>
>>> IO port 0x61 is used by pc speaker and NMI reason port.
>>>
>>> This is ambiguous. How to distinguish in QEMU?

That register is a mess functionality-wise because it serves multiple
hardware units that are at least logically separated - legacy. In QEMU,
the speaker emulation adopted it as it is the only unit that actually
has some functionality behind the bits (see below).

>>>
>>> I see that NMI reason port is not realized in QEMU.
>>>
>>> But linux guest will access this port to determine what the NMI reason code means.
>>>
>>> It will access pc speaker's io port. Is this a bug?
>>>
>> Gerd, any thoughts?
> 
> Not my area of expertise, but can qemu send NMIs to the guest in the
> first place?

QEMU doesn't generate the kind of NMI events that would normally be
reported via the status bits in port 0x61. Therefore, bits 6 and 7 will
always remain unset for now.

Jan

PS: The port name shown in "info mtree" is wrong. Will send a trivial
fix to help identifying the device model behind it.

-- 
Siemens AG, Corporate Technology, CT RTC ITP SES-DE
Corporate Competence Center Embedded Linux