2018-10-31 14:43:42

by Rob Herring (Arm)

Subject: [PATCH] of: Fix cpu node iterator to not ignore disabled cpu nodes

In most cases, nodes with 'status = "disabled";' are treated as if the
node is not present though it is a common bug to forget to check that.
However, cpu nodes are different in that "disabled" simply means offline
and the OS can bring the CPU core online. Commit f1f207e43b8a ("of: Add
cpu node iterator for_each_of_cpu_node()") followed the common behavior
of ignoring disabled cpu nodes. This breaks some powerpc systems (at
least NXP P50XX/e5500). Fix this by dropping the status check.

Fixes: 651d44f9679c ("of: use for_each_of_cpu_node iterator")
Fixes: f1f207e43b8a ("of: Add cpu node iterator for_each_of_cpu_node()")
Reported-by: Michael Ellerman <[email protected]>
Cc: Christian Zigotzky <[email protected]>
Cc: Frank Rowand <[email protected]>
Signed-off-by: Rob Herring <[email protected]>
---
drivers/of/base.c | 2 --
1 file changed, 2 deletions(-)

diff --git a/drivers/of/base.c b/drivers/of/base.c
index cc62da278663..e47c5ce6cd58 100644
--- a/drivers/of/base.c
+++ b/drivers/of/base.c
@@ -776,8 +776,6 @@ struct device_node *of_get_next_cpu_node(struct device_node *prev)
if (!(of_node_name_eq(next, "cpu") ||
(next->type && !of_node_cmp(next->type, "cpu"))))
continue;
- if (!__of_device_is_available(next))
- continue;
if (of_node_get(next))
break;
}
--
2.19.1

2018-10-31 16:41:38

by Christian Zigotzky

Subject: Re: [PATCH] of: Fix cpu node iterator to not ignore disabled cpu nodes

Thank you. I will test this patch tomorrow.

— Christian

Sent from my iPhone

> On 31. Oct 2018, at 15:42, Rob Herring <[email protected]> wrote:
>
> In most cases, nodes with 'status = "disabled";' are treated as if the
> node is not present though it is a common bug to forget to check that.
> However, cpu nodes are different in that "disabled" simply means offline
> and the OS can bring the CPU core online. Commit f1f207e43b8a ("of: Add
> cpu node iterator for_each_of_cpu_node()") followed the common behavior
> of ignoring disabled cpu nodes. This breaks some powerpc systems (at
> least NXP P50XX/e5500). Fix this by dropping the status check.
>
> Fixes: 651d44f9679c ("of: use for_each_of_cpu_node iterator")
> Fixes: f1f207e43b8a ("of: Add cpu node iterator for_each_of_cpu_node()")
> Reported-by: Michael Ellerman <[email protected]>
> Cc: Christian Zigotzky <[email protected]>
> Cc: Frank Rowand <[email protected]>
> Signed-off-by: Rob Herring <[email protected]>
> ---
> drivers/of/base.c | 2 --
> 1 file changed, 2 deletions(-)
>
> diff --git a/drivers/of/base.c b/drivers/of/base.c
> index cc62da278663..e47c5ce6cd58 100644
> --- a/drivers/of/base.c
> +++ b/drivers/of/base.c
> @@ -776,8 +776,6 @@ struct device_node *of_get_next_cpu_node(struct device_node *prev)
> if (!(of_node_name_eq(next, "cpu") ||
> (next->type && !of_node_cmp(next->type, "cpu"))))
> continue;
> - if (!__of_device_is_available(next))
> - continue;
> if (of_node_get(next))
> break;
> }
> --
> 2.19.1
>

2018-11-01 11:10:48

by Michael Ellerman

Subject: Re: [PATCH] of: Fix cpu node iterator to not ignore disabled cpu nodes

Rob Herring <[email protected]> writes:

> In most cases, nodes with 'status = "disabled";' are treated as if the
> node is not present though it is a common bug to forget to check that.
> However, cpu nodes are different in that "disabled" simply means offline
> and the OS can bring the CPU core online. Commit f1f207e43b8a ("of: Add
> cpu node iterator for_each_of_cpu_node()") followed the common behavior
> of ignoring disabled cpu nodes. This breaks some powerpc systems (at
> least NXP P50XX/e5500). Fix this by dropping the status check.
>
> Fixes: 651d44f9679c ("of: use for_each_of_cpu_node iterator")
> Fixes: f1f207e43b8a ("of: Add cpu node iterator for_each_of_cpu_node()")
> Reported-by: Michael Ellerman <[email protected]>
> Cc: Christian Zigotzky <[email protected]>
> Cc: Frank Rowand <[email protected]>
> Signed-off-by: Rob Herring <[email protected]>
> ---
> drivers/of/base.c | 2 --
> 1 file changed, 2 deletions(-)

This fixes my machine, thanks.

Tested-by: Michael Ellerman <[email protected]>

It was actually originally reported to me by Christian, so also:

Reported-by: Christian Zigotzky <[email protected]>


cheers

> diff --git a/drivers/of/base.c b/drivers/of/base.c
> index cc62da278663..e47c5ce6cd58 100644
> --- a/drivers/of/base.c
> +++ b/drivers/of/base.c
> @@ -776,8 +776,6 @@ struct device_node *of_get_next_cpu_node(struct device_node *prev)
> if (!(of_node_name_eq(next, "cpu") ||
> (next->type && !of_node_cmp(next->type, "cpu"))))
> continue;
> - if (!__of_device_is_available(next))
> - continue;
> if (of_node_get(next))
> break;
> }
> --
> 2.19.1

2018-11-01 12:47:02

by Christian Zigotzky

Subject: Re: [PATCH] of: Fix cpu node iterator to not ignore disabled cpu nodes

On 01 November 2018 at 12:08PM, Michael Ellerman wrote:
> Rob Herring <[email protected]> writes:
>
>> In most cases, nodes with 'status = "disabled";' are treated as if the
>> node is not present though it is a common bug to forget to check that.
>> However, cpu nodes are different in that "disabled" simply means offline
>> and the OS can bring the CPU core online. Commit f1f207e43b8a ("of: Add
>> cpu node iterator for_each_of_cpu_node()") followed the common behavior
>> of ignoring disabled cpu nodes. This breaks some powerpc systems (at
>> least NXP P50XX/e5500). Fix this by dropping the status check.
>>
>> Fixes: 651d44f9679c ("of: use for_each_of_cpu_node iterator")
>> Fixes: f1f207e43b8a ("of: Add cpu node iterator for_each_of_cpu_node()")
>> Reported-by: Michael Ellerman <[email protected]>
>> Cc: Christian Zigotzky <[email protected]>
>> Cc: Frank Rowand <[email protected]>
>> Signed-off-by: Rob Herring <[email protected]>
>> ---
>> drivers/of/base.c | 2 --
>> 1 file changed, 2 deletions(-)
> This fixes my machine, thanks.
>
> Tested-by: Michael Ellerman <[email protected]>
>
> It was actually originally reported to me by Christian, so also:
>
> Reported-by: Christian Zigotzky <[email protected]>
>
>
> cheers
>
>> diff --git a/drivers/of/base.c b/drivers/of/base.c
>> index cc62da278663..e47c5ce6cd58 100644
>> --- a/drivers/of/base.c
>> +++ b/drivers/of/base.c
>> @@ -776,8 +776,6 @@ struct device_node *of_get_next_cpu_node(struct device_node *prev)
>> if (!(of_node_name_eq(next, "cpu") ||
>> (next->type && !of_node_cmp(next->type, "cpu"))))
>> continue;
>> - if (!__of_device_is_available(next))
>> - continue;
>> if (of_node_get(next))
>> break;
>> }
>> --
>> 2.19.1

Hi All,

Many thanks for your help! This patch fixes my P5020 board (dual core)
and my virtual QEMU ppce500 machine (quad core).

Have a nice day!

Cheers,
Christian