[PATCH sched_ext/for-7.2] sched_ext: Fix ops_cid layout assert

Tejun Heo posted 1 patch 1 month ago
kernel/sched/ext.c          | 6 ++----
kernel/sched/ext_internal.h | 3 +++
2 files changed, 5 insertions(+), 4 deletions(-)
[PATCH sched_ext/for-7.2] sched_ext: Fix ops_cid layout assert
Posted by Tejun Heo 1 month ago
ca1d48a86fab ("sched_ext: Use offsetofend on both sides of the ops_cid
layout assert") replaced sizeof() with offsetofend() to dodge 32-bit
PPC trailing padding, but the resulting check is tautological: with
CID_OFFSET_MATCH(priv, priv) already enforcing offsetof(priv) equality
and @priv being the same type in both structs, the two offsetofends
are equal by construction. The original protection - catching a stray
field added past @priv in sched_ext_ops_cid - is gone.

Anchor on a zero-size __end[] marker appended after @priv. Its offset
sits flush after @priv regardless of trailing struct padding; if a
field is inserted past @priv, __end shifts and the assert fires.

Closes: https://lore.kernel.org/all/20260508215211.0C03AC2BCB0@smtp.kernel.org/
Signed-off-by: Tejun Heo <tj@kernel.org>
---
 kernel/sched/ext.c          | 6 ++----
 kernel/sched/ext_internal.h | 3 +++
 2 files changed, 5 insertions(+), 4 deletions(-)

diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
index f86ee15be7cb..b685f45b4fd0 100644
--- a/kernel/sched/ext.c
+++ b/kernel/sched/ext.c
@@ -10380,11 +10380,9 @@ static int __init scx_init(void)
 	/*
 	 * cid-form must end exactly at @priv - validate_ops() skips
 	 * cpu_acquire/cpu_release for cid-form because reading those fields
-	 * past the BPF allocation would be UB. offsetofend() on both sides
-	 * instead of sizeof() on sched_ext_ops_cid to sidestep trailing
-	 * struct padding (e.g. 32-bit PPC tail-pads ops_cid past @priv).
+	 * past the BPF allocation would be UB.
 	 */
-	BUILD_BUG_ON(offsetofend(struct sched_ext_ops_cid, priv) !=
+	BUILD_BUG_ON(offsetof(struct sched_ext_ops_cid, __end) !=
 		     offsetofend(struct sched_ext_ops, priv));
 #undef CID_OFFSET_MATCH
 
diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
index 0ed79bd891c7..cd33984cffcf 100644
--- a/kernel/sched/ext_internal.h
+++ b/kernel/sched/ext_internal.h
@@ -938,6 +938,9 @@ struct sched_ext_ops_cid {
 
 	/* internal use only, must be NULL */
 	void __rcu *priv;
+
+	/* layout end anchor for the BUILD_BUG_ON in scx_init(); keep last */
+	char __end[0];
 };
 
 enum scx_opi {
-- 
2.54.0
Re: [PATCH sched_ext/for-7.2] sched_ext: Fix ops_cid layout assert
Posted by Tejun Heo 1 month ago
Hello,

Applied to sched_ext/for-7.2.

Thanks.

--
tejun
Re: [PATCH sched_ext/for-7.2] sched_ext: Fix ops_cid layout assert
Posted by Emil Tsalapatis 1 month ago
On Fri May 8, 2026 at 7:59 PM EDT, Tejun Heo wrote:
> ca1d48a86fab ("sched_ext: Use offsetofend on both sides of the ops_cid
> layout assert") replaced sizeof() with offsetofend() to dodge 32-bit
> PPC trailing padding, but the resulting check is tautological: with
> CID_OFFSET_MATCH(priv, priv) already enforcing offsetof(priv) equality
> and @priv being the same type in both structs, the two offsetofends
> are equal by construction. The original protection - catching a stray
> field added past @priv in sched_ext_ops_cid - is gone.
>
> Anchor on a zero-size __end[] marker appended after @priv. Its offset
> sits flush after @priv regardless of trailing struct padding; if a
> field is inserted past @priv, __end shifts and the assert fires.
>
> Closes: https://lore.kernel.org/all/20260508215211.0C03AC2BCB0@smtp.kernel.org/
> Signed-off-by: Tejun Heo <tj@kernel.org>

Reviewed-by: Emil Tsalapatis <emil@etsalapatis.com>

> ---
>  kernel/sched/ext.c          | 6 ++----
>  kernel/sched/ext_internal.h | 3 +++
>  2 files changed, 5 insertions(+), 4 deletions(-)
>
> diff --git a/kernel/sched/ext.c b/kernel/sched/ext.c
> index f86ee15be7cb..b685f45b4fd0 100644
> --- a/kernel/sched/ext.c
> +++ b/kernel/sched/ext.c
> @@ -10380,11 +10380,9 @@ static int __init scx_init(void)
>  	/*
>  	 * cid-form must end exactly at @priv - validate_ops() skips
>  	 * cpu_acquire/cpu_release for cid-form because reading those fields
> -	 * past the BPF allocation would be UB. offsetofend() on both sides
> -	 * instead of sizeof() on sched_ext_ops_cid to sidestep trailing
> -	 * struct padding (e.g. 32-bit PPC tail-pads ops_cid past @priv).
> +	 * past the BPF allocation would be UB.
>  	 */
> -	BUILD_BUG_ON(offsetofend(struct sched_ext_ops_cid, priv) !=
> +	BUILD_BUG_ON(offsetof(struct sched_ext_ops_cid, __end) !=
>  		     offsetofend(struct sched_ext_ops, priv));
>  #undef CID_OFFSET_MATCH
>  
> diff --git a/kernel/sched/ext_internal.h b/kernel/sched/ext_internal.h
> index 0ed79bd891c7..cd33984cffcf 100644
> --- a/kernel/sched/ext_internal.h
> +++ b/kernel/sched/ext_internal.h
> @@ -938,6 +938,9 @@ struct sched_ext_ops_cid {
>  
>  	/* internal use only, must be NULL */
>  	void __rcu *priv;
> +
> +	/* layout end anchor for the BUILD_BUG_ON in scx_init(); keep last */
> +	char __end[0];
>  };
>  
>  enum scx_opi {