[PATCH] powerpc/smp: parse_thread_groups() return error on missing property

[PATCH] powerpc/smp: parse_thread_groups() return error on missing property

[John Ogness]

parse_thread_groups() is supposed to return -EINVAL if the property
does not exist and -ENODATA if property does not have a value.
However, it is currently assuming it exists and blindly calling
kzalloc_objs() with whatever of_property_count_u32_elems() returns.

Since of_property_count_u32_elems() returns exactly these same error
values in those cases, simply return with the returned error value.

Signed-off-by: John Ogness <john.ogness@linutronix.de>
---
 arch/powerpc/kernel/smp.c | 2 ++
 1 file changed, 2 insertions(+)

diff --git a/arch/powerpc/kernel/smp.c b/arch/powerpc/kernel/smp.c
index 7a25eaa59e626..58824f5f4c800 100644
--- a/arch/powerpc/kernel/smp.c
+++ b/arch/powerpc/kernel/smp.c
@@ -824,6 +824,8 @@ static int parse_thread_groups(struct device_node *dn,
 	int i = 0;
 
 	count = of_property_count_u32_elems(dn, "ibm,thread-groups");
+	if (count < 0)
+		return count;
 	thread_group_array = kzalloc_objs(u32, count);
 	if (!thread_group_array)
 		return -ENOMEM;
-- 
2.47.3

Re: [PATCH] powerpc/smp: parse_thread_groups() return error on missing property

[Sebastian Andrzej Siewior]

+ Kees, Vlastimil

On 2026-11-14 08:44:57 [+0106], John Ogness wrote:
> parse_thread_groups() is supposed to return -EINVAL if the property
> does not exist and -ENODATA if property does not have a value.
> However, it is currently assuming it exists and blindly calling
> kzalloc_objs() with whatever of_property_count_u32_elems() returns.
> 
> Since of_property_count_u32_elems() returns exactly these same error
> values in those cases, simply return with the returned error value.
> 
> Signed-off-by: John Ogness <john.ogness@linutronix.de>

Reviewed-by: Sebastian Andrzej Siewior <bigeasy@linutronix.de>
Fixes: 0af298e62e5cf ("treewide: refresh kmalloc_obj() conversions")

But… The difference in the commit was:

|-       thread_group_array = kcalloc(count, sizeof(u32), GFP_KERNEL);
|+       thread_group_array = kzalloc_objs(u32, count);
|        if (!thread_group_array)
|                return -ENOMEM;

kcalloc() returned NULL on an overflow and did not continue with the
allocation. So the check below worked.

kzalloc_objs() on the other hand returns for the overflow SIZE_MAX and
passes this down to the allocator. We still get a NULL pointer but with
a warning (from ___kmalloc_large_node()).

Is the warning on purpose now?

Sebastian

Re: [PATCH] powerpc/smp: parse_thread_groups() return error on missing property

[Ritesh Harjani]

John Ogness <john.ogness@linutronix.de> writes:

> parse_thread_groups() is supposed to return -EINVAL if the property
> does not exist and -ENODATA if property does not have a value.
> However, it is currently assuming it exists and blindly calling
> kzalloc_objs() with whatever of_property_count_u32_elems() returns.
>
> Since of_property_count_u32_elems() returns exactly these same error
> values in those cases, simply return with the returned error value.
>
> Signed-off-by: John Ogness <john.ogness@linutronix.de>
> ---
>  arch/powerpc/kernel/smp.c | 2 ++
>  1 file changed, 2 insertions(+)

Thanks! However, the fix for this is already queued to powerpc's fixes branch.

https://git.kernel.org/pub/scm/linux/kernel/git/powerpc/linux.git/commit/?h=fixes&id=126b47221b7b98d83c118781ce724c0515d86d9c


-ritesh