Patchwork [V7,2/3] powerpc/ge_imp3a: Add ISA node to PCI primary bus

login
register
mail settings
Submitter Hongtao Jia
Date Aug. 15, 2012, 8:57 a.m.
Message ID <1345021026-10886-3-git-send-email-B38951@freescale.com>
Download mbox | patch
Permalink /patch/177593/
State Superseded
Headers show

Comments

Hongtao Jia - Aug. 15, 2012, 8:57 a.m.
PCI primary bus will be determined by looking for ISA node in device tree.
Also for board ge_imp3a primary bus is the second PCI controller detected.
So we add ISA node to ge_imp3a's device tree to fit the new determination.

Adding ISA node to other boards' device tree is not necessary. The situation
that the first bus is primary have already been handled. Please refer to the
following patch:
powerpc/pci: Make sure ISA IO base is not zero

Signed-off-by: Jia Hongtao <B38951@freescale.com>
Signed-off-by: Li Yang <leoli@freescale.com>
---
 arch/powerpc/boot/dts/ge_imp3a.dts |    4 ++++
 1 files changed, 4 insertions(+), 0 deletions(-)
Scott Wood - Aug. 15, 2012, 5:31 p.m.
On 08/15/2012 03:57 AM, Jia Hongtao wrote:
> PCI primary bus will be determined by looking for ISA node in device tree.
> Also for board ge_imp3a primary bus is the second PCI controller detected.
> So we add ISA node to ge_imp3a's device tree to fit the new determination.
> 
> Adding ISA node to other boards' device tree is not necessary. The situation
> that the first bus is primary have already been handled. Please refer to the
> following patch:
> powerpc/pci: Make sure ISA IO base is not zero
> 
> Signed-off-by: Jia Hongtao <B38951@freescale.com>
> Signed-off-by: Li Yang <leoli@freescale.com>
> ---
>  arch/powerpc/boot/dts/ge_imp3a.dts |    4 ++++
>  1 files changed, 4 insertions(+), 0 deletions(-)
> 
> diff --git a/arch/powerpc/boot/dts/ge_imp3a.dts b/arch/powerpc/boot/dts/ge_imp3a.dts
> index fefae41..aa2c4b5 100644
> --- a/arch/powerpc/boot/dts/ge_imp3a.dts
> +++ b/arch/powerpc/boot/dts/ge_imp3a.dts
> @@ -248,6 +248,10 @@
>  				  0x1000000 0x0 0x0
>  				  0x1000000 0x0 0x0
>  				  0x0 0x10000>;
> +
> +			isa@1e {
> +				device_type = "isa";
> +			};
>  		};
>  	};
>  };

I believe I already said it's not acceptable to break old device trees.

-Scott
Kumar Gala - Aug. 15, 2012, 9:01 p.m.
On Aug 15, 2012, at 12:31 PM, Scott Wood wrote:

> On 08/15/2012 03:57 AM, Jia Hongtao wrote:
>> PCI primary bus will be determined by looking for ISA node in device tree.
>> Also for board ge_imp3a primary bus is the second PCI controller detected.
>> So we add ISA node to ge_imp3a's device tree to fit the new determination.
>> 
>> Adding ISA node to other boards' device tree is not necessary. The situation
>> that the first bus is primary have already been handled. Please refer to the
>> following patch:
>> powerpc/pci: Make sure ISA IO base is not zero
>> 
>> Signed-off-by: Jia Hongtao <B38951@freescale.com>
>> Signed-off-by: Li Yang <leoli@freescale.com>
>> ---
>> arch/powerpc/boot/dts/ge_imp3a.dts |    4 ++++
>> 1 files changed, 4 insertions(+), 0 deletions(-)
>> 
>> diff --git a/arch/powerpc/boot/dts/ge_imp3a.dts b/arch/powerpc/boot/dts/ge_imp3a.dts
>> index fefae41..aa2c4b5 100644
>> --- a/arch/powerpc/boot/dts/ge_imp3a.dts
>> +++ b/arch/powerpc/boot/dts/ge_imp3a.dts
>> @@ -248,6 +248,10 @@
>> 				  0x1000000 0x0 0x0
>> 				  0x1000000 0x0 0x0
>> 				  0x0 0x10000>;
>> +
>> +			isa@1e {
>> +				device_type = "isa";
>> +			};
>> 		};
>> 	};
>> };
> 
> I believe I already said it's not acceptable to break old device trees.
> 
> -Scott
> 

I agree w/Scott this is NOT an acceptable solution, both because of old device tree and its just wrong to list ISA if it doesn't exist in the device.

- k
Scott Wood - Aug. 15, 2012, 9:05 p.m.
On 08/15/2012 04:01 PM, Kumar Gala wrote:
> 
> On Aug 15, 2012, at 12:31 PM, Scott Wood wrote:
> 
>> On 08/15/2012 03:57 AM, Jia Hongtao wrote:
>>> PCI primary bus will be determined by looking for ISA node in device tree.
>>> Also for board ge_imp3a primary bus is the second PCI controller detected.
>>> So we add ISA node to ge_imp3a's device tree to fit the new determination.
>>>
>>> Adding ISA node to other boards' device tree is not necessary. The situation
>>> that the first bus is primary have already been handled. Please refer to the
>>> following patch:
>>> powerpc/pci: Make sure ISA IO base is not zero
>>>
>>> Signed-off-by: Jia Hongtao <B38951@freescale.com>
>>> Signed-off-by: Li Yang <leoli@freescale.com>
>>> ---
>>> arch/powerpc/boot/dts/ge_imp3a.dts |    4 ++++
>>> 1 files changed, 4 insertions(+), 0 deletions(-)
>>>
>>> diff --git a/arch/powerpc/boot/dts/ge_imp3a.dts b/arch/powerpc/boot/dts/ge_imp3a.dts
>>> index fefae41..aa2c4b5 100644
>>> --- a/arch/powerpc/boot/dts/ge_imp3a.dts
>>> +++ b/arch/powerpc/boot/dts/ge_imp3a.dts
>>> @@ -248,6 +248,10 @@
>>> 				  0x1000000 0x0 0x0
>>> 				  0x1000000 0x0 0x0
>>> 				  0x0 0x10000>;
>>> +
>>> +			isa@1e {
>>> +				device_type = "isa";
>>> +			};
>>> 		};
>>> 	};
>>> };
>>
>> I believe I already said it's not acceptable to break old device trees.
>>
>> -Scott
>>
> 
> I agree w/Scott this is NOT an acceptable solution, both because of
> old device tree and its just wrong to list ISA if it doesn't exist in
> the device.

If there isn't actually ISA here, why do we care which bus is primary?
Other than the lack of hardware to sanity test the change on...

-Scott

Patch

diff --git a/arch/powerpc/boot/dts/ge_imp3a.dts b/arch/powerpc/boot/dts/ge_imp3a.dts
index fefae41..aa2c4b5 100644
--- a/arch/powerpc/boot/dts/ge_imp3a.dts
+++ b/arch/powerpc/boot/dts/ge_imp3a.dts
@@ -248,6 +248,10 @@ 
 				  0x1000000 0x0 0x0
 				  0x1000000 0x0 0x0
 				  0x0 0x10000>;
+
+			isa@1e {
+				device_type = "isa";
+			};
 		};
 	};
 };