On Thu, 2008-11-13 at 14:58 +0800, Eric W. Biederman wrote:
> Zhao Yakui <yakui.zhao@xxxxxxxxx> writes:
>
> >> Instead of just writing a plain 6. I think at least on some machines
> >> there is a requirement for a low to hight transition.
>
> > Maybe the 0xCF9 port is defined for NVidia, SiS and Several others
> > vendor. But we can't say that it is appropriate for most boxes.
>
> Let me get this straight. We survey all of the major chipsets in existence.
> We discover a register that is implemented the same way in all of them.
> You conclude that using that register is not appropriate for most boxes
I am sorry that I don't describe it very clearly. The value written into
0xCF9 I/O port maybe vary on different chipset vendors. In such case it
is difficult to write the generic 0xCF9 reboot mechanism. If the same
standard about 0xCF9 I/O port is followed by most chipset vendors, it
will be OK.
>
> > Of course IMO it is not a generic solution for reboot.
> > At the same time the value written to the 0xCF9 I/O port should not be a
> > fixed value. Maybe it varies on different vendors/chipset. Even the
> > value will vary on the boxes from different BIOS vendors although the
> > boxes are based on the same chipset.
>
> Reading in between the lines. Did you just say someone implemented
> a nasty hack on some laptop and attempting to reboot by writing to 0xcf9
> will cause nasty problems?
I have a laptop on which the ACPI reboot mechanism is not supported.(The
reset flag is not present). But there exists the definition of RESET_REG
& RESET_VALUE. The RESET_REG is 0xCF9 I/O port. (The box is based on
Intel i915 chipsets)
When writing 0x06 to 0xCF9 I/O port the box will hang and can't be
rebooted. (Hardware reset will fail)
When writing 0x04 to 0xCF9 I/O port the box can be
rebooted.(Software reset is OK)
On the laptop of bug 11942(kernel bugzilla) the acpi reboot is also
unsupported. But there exists the definition of 0xCF9 I/O port. (The box
is based on AMD platform. The CPU is Phenom 9500 Quad-core.)
After writing 0x06 to 0xCF9 I/O port, the box can't be rebooted.
>
> If you know of such a case please describe the failure mode. If there
> is a case where writing to 0xcf9 will makes things worse than we need
> to tread carefully, and we will happily deal with reality. If you can
> only tell us that we are being rude and circumventing a vendors vision
> of the then your input is not especially useful.
>
> > And the value should be provided by BIOS. If BIOS doesn't provide
> > the above info, it is not safe to use them.
>
> Balderdash. There are other ways to learn things than by talking to
> the BIOS. Usually BIOS's are a mediocre information source at best.
> The job of a BIOS is to boot and if it can do that solidly you can ship it,
> when the schedule gets tight. Everything else is a nice to have.
>
> Eric
>
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
