Re: Re-add ibft tree to linux-next please

Re: Re-add ibft tree to linux-next please

[H. Peter Anvin]

On 12/12/2011 06:53 AM, Konrad Rzeszutek Wilk wrote:
> Hey Stephen,
> 
> I've one patch for the iBFT and since I am the co-maintainer of said tree,
> please add this tree to your awesome build system:
> 
> git://git.kernel.org/pub/scm/linux/kernel/git/konrad/ibft.git
> 
> The branch is "linux-next"
> 
> Signed-off-by: Konrad Rzeszutek Wilk <konrad@kernel.org>
> CC: Peter Jones <pjones@redhat.com>
> Thanks!
> 
> P.S.
> After kernel.org went down, I choose not to resurface the iBFT tree b/c I
> figured that the driver was sooo stable that there would be no need for any
> patches. Well, not the case

iBFT is basically an ACPI table with nonstandard discovery.  Any way we
could get this merged into the ACPI core eventually?

	-hpa

Re: Re-add ibft tree to linux-next please

[Konrad Rzeszutek Wilk]

On Mon, Dec 12, 2011 at 03:25:11PM -0800, H. Peter Anvin wrote:
> On 12/12/2011 06:53 AM, Konrad Rzeszutek Wilk wrote:
> > Hey Stephen,
> > 
> > I've one patch for the iBFT and since I am the co-maintainer of said tree,
> > please add this tree to your awesome build system:
> > 
> > git://git.kernel.org/pub/scm/linux/kernel/git/konrad/ibft.git
> > 
> > The branch is "linux-next"
> > 
> > Signed-off-by: Konrad Rzeszutek Wilk <konrad@kernel.org>
> > CC: Peter Jones <pjones@redhat.com>
> > Thanks!
> > 
> > P.S.
> > After kernel.org went down, I choose not to resurface the iBFT tree b/c I
> > figured that the driver was sooo stable that there would be no need for any
> > patches. Well, not the case
> 
> iBFT is basically an ACPI table with nonstandard discovery.  Any way we
> could get this merged into the ACPI core eventually?

Part of the driver has been exported to the SCSI tree already (the software and
broadcam iSCSI one). The only two remainig pieces are the SysFS skeleton
(which should probably still sit in drivers/firmware) and iBFT finder.

The remaining piece that could be moved to drivers/acpi is the "finder" of
the IBFT which can find it via ACPI, similar to how MADT tables are found
(so perhaps arch/x86/acpi?).

The trick is that we also have to retain the code functionality if
"# CONFIG_ACPI is not set" to scan the memory for the iBFT. That part
should definitly _not_ be moved to drivers/acpi. It could be moved to
arch/x86/ .. but it could also stay in drivers/firmware. <shrugs>

Re: Re-add ibft tree to linux-next please

[Len Brown]

> The trick is that we also have to retain the code functionality if
> "# CONFIG_ACPI is not set" to scan the memory for the iBFT. That part
> should definitly _not_ be moved to drivers/acpi. It could be moved to
> arch/x86/ .. but it could also stay in drivers/firmware. <shrugs>


Are there really hardware/firmware configurations that support IBFT
and do not support ACPI?

If yes, do they require an OS that has CONFIG_ACPI=n?
Or would finding the table with CONFIG_ACPI=y and acpi_disabled=1
be sufficient?

-Len

Re: Re-add ibft tree to linux-next please

[Konrad Rzeszutek Wilk]

On Tue, Jan 17, 2012 at 09:08:30AM -0500, Len Brown wrote:
> 
> > The trick is that we also have to retain the code functionality if
> > "# CONFIG_ACPI is not set" to scan the memory for the iBFT. That part
> > should definitly _not_ be moved to drivers/acpi. It could be moved to
> > arch/x86/ .. but it could also stay in drivers/firmware. <shrugs>
> 
> 
> Are there really hardware/firmware configurations that support IBFT
> and do not support ACPI?

gPXE can be used in legacy environments. And it (gPXE) does not update
the ACPI tables.

> 
> If yes, do they require an OS that has CONFIG_ACPI=n?

I can't think of anybody nowadays doing CONFIG_ACPI=n for x86.
Is there anybody doing it?


> Or would finding the table with CONFIG_ACPI=y and acpi_disabled=1
> be sufficient?

I think that would work. And also if 'acpi_disabled=0' as you might
have an IBFT table in memory that is _not_ hooked up to the ACPI tables.

Re: Re-add ibft tree to linux-next please

[H. Peter Anvin]

On 01/17/2012 08:49 AM, Konrad Rzeszutek Wilk wrote:
>>
>> Are there really hardware/firmware configurations that support IBFT
>> and do not support ACPI?
> 
> gPXE can be used in legacy environments. And it (gPXE) does not update
> the ACPI tables.
> 

That's irrelevant.  THe point is that iBFT *is* an ACPI table.

> I think that would work. And also if 'acpi_disabled=0' as you might
> have an IBFT table in memory that is _not_ hooked up to the ACPI tables.

That is *always* the case, the "finder" is nonstandard (because the ACPI
tables are largely non-extensible, says the one who has tried to
actually extend ACPI tables in memory.)

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.