2018-06-21 13:39:01

by Christoph Hellwig

[permalink] [raw]
Subject: [PATCH] base: fix order of OF initialization

From: "Wesley W. Terpstra" <[email protected]>

This fixes: [ 0.010000] cpu cpu0: Error -2 creating of_node link
... which you get for every CPU on all architectures that use
CONFIG_GENERIC_CPU_DEVICES.

In that case, driver_init() calls cpu_dev_init() before calling
of_core_init(). Then we get the callchain:

cpu_dev_init()
-> cpu_dev_register_generic()
-> register_cpu(cpu, i)
-> device_register(&cpu->dev)
-> device_add(dev)
-> device_add_class_symlinks(dev)

... in device_add_class_symlinks, we we dev->of_node, and call
sysfs_create_link(), which fails because we haven't called
of_core_init() to register the sysfs devicetree directory yet.

Signed-off-by: Wesley W. Terpstra <[email protected]>
[hch: updated the changelog based on review feedback]
Signed-off-by: Christoph Hellwig <[email protected]>
Acked-by: Mark Rutland <[email protected]>
Acked-by: Frank Rowand <[email protected]>
---

In case anyone is wondering why I send out this patch - I'm digging
through the RISC-V tree and trying to the patches required to let
a RISC-V kernel boot on qemu upstream. I also got the previously
missing signoff from Wes offlist.

drivers/base/init.c | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/base/init.c b/drivers/base/init.c
index dd85b05a6a16..908e6520e804 100644
--- a/drivers/base/init.c
+++ b/drivers/base/init.c
@@ -30,9 +30,9 @@ void __init driver_init(void)
/* These are also core pieces, but must come after the
* core core pieces.
*/
+ of_core_init();
platform_bus_init();
cpu_dev_init();
memory_dev_init();
container_dev_init();
- of_core_init();
}
--
2.17.1



2018-06-22 01:06:51

by Frank Rowand

[permalink] [raw]
Subject: Re: [PATCH] base: fix order of OF initialization

On 06/21/18 06:37, Christoph Hellwig wrote:
> From: "Wesley W. Terpstra" <[email protected]>
>
> This fixes: [ 0.010000] cpu cpu0: Error -2 creating of_node link
> ... which you get for every CPU on all architectures that use
> CONFIG_GENERIC_CPU_DEVICES.
>
> In that case, driver_init() calls cpu_dev_init() before calling
> of_core_init(). Then we get the callchain:
>
> cpu_dev_init()
> -> cpu_dev_register_generic()
> -> register_cpu(cpu, i)
> -> device_register(&cpu->dev)
> -> device_add(dev)
> -> device_add_class_symlinks(dev)
>
> ... in device_add_class_symlinks, we we dev->of_node, and call
> sysfs_create_link(), which fails because we haven't called
> of_core_init() to register the sysfs devicetree directory yet.
>
> Signed-off-by: Wesley W. Terpstra <[email protected]>
> [hch: updated the changelog based on review feedback]
> Signed-off-by: Christoph Hellwig <[email protected]>
> Acked-by: Mark Rutland <[email protected]>
> Acked-by: Frank Rowand <[email protected]>
> ---
>
> In case anyone is wondering why I send out this patch - I'm digging
> through the RISC-V tree and trying to the patches required to let
> a RISC-V kernel boot on qemu upstream. I also got the previously
> missing signoff from Wes offlist.
>
> drivers/base/init.c | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/drivers/base/init.c b/drivers/base/init.c
> index dd85b05a6a16..908e6520e804 100644
> --- a/drivers/base/init.c
> +++ b/drivers/base/init.c
> @@ -30,9 +30,9 @@ void __init driver_init(void)
> /* These are also core pieces, but must come after the
> * core core pieces.
> */
> + of_core_init();
> platform_bus_init();
> cpu_dev_init();
> memory_dev_init();
> container_dev_init();
> - of_core_init();
> }
>

The updated patch comment looks good. It would be good to get this into
-next while we are still early in the release cycle.

For those trying to find the previous version of this patch, it was cced
to the devicetree list on 06/07/17.

-Frank