* [PATCH] fork: Honor task_struct's declared alignment
@ 2026-07-10 12:39 Karl Mehltretter
2026-07-16 17:21 ` Bradley Morgan
2026-07-17 8:14 ` Vlastimil Babka (SUSE)
0 siblings, 2 replies; 3+ messages in thread
From: Karl Mehltretter @ 2026-07-10 12:39 UTC (permalink / raw)
To: Ingo Molnar, Peter Zijlstra, Andrew Morton
Cc: Kees Cook, Vlastimil Babka, linux-mm, linux-kernel, Karl Mehltretter
Since commit cb7ca40a3882 ("x86/fpu: Make task_struct::thread constant
size"), struct task_struct is declared __attribute__((aligned(64))) on
all architectures.
But fork_init() still sets the task_struct slab cache's alignment to
align = max(L1_CACHE_BYTES, ARCH_MIN_TASKALIGN)
which is smaller than 64 on architectures whose cache lines are below
64 bytes: e.g. 32 on ARMv5.
In practice plain SLUB happens to hand out 64-byte-aligned objects
anyway. With CONFIG_SLUB_DEBUG_ON the red-zone padding shifts objects
to the requested alignment.
With CONFIG_UBSAN_ALIGNMENT=y a boot on QEMU versatilepb (ARM926EJ-S,
v7.2-rc2, gcc 13.3) floods the console with reports like:
UBSAN: misaligned-access in include/linux/sched.h:2087:9
member access within misaligned address c295d7e0 for type 'struct task_struct'
which requires 64 byte alignment
CPU: 0 UID: 0 PID: 15 Comm: pr/ttyAMA-1 Not tainted 7.2.0-rc2 #1 VOLUNTARY
Set the slab alignment to at least the type's declared alignment.
Fixes: cb7ca40a3882 ("x86/fpu: Make task_struct::thread constant size")
Assisted-by: Claude:claude-fable-5
Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
---
kernel/fork.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
diff --git a/kernel/fork.c b/kernel/fork.c
index f0e2e131a9a5..b85cf1343329 100644
--- a/kernel/fork.c
+++ b/kernel/fork.c
@@ -857,7 +857,8 @@ void __init fork_init(void)
#ifndef ARCH_MIN_TASKALIGN
#define ARCH_MIN_TASKALIGN 0
#endif
- int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
+ int align = max3(L1_CACHE_BYTES, ARCH_MIN_TASKALIGN,
+ __alignof__(struct task_struct));
unsigned long useroffset, usersize;
/* create a slab on which task_structs can be allocated */
--
2.39.5 (Apple Git-154)
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] fork: Honor task_struct's declared alignment
2026-07-10 12:39 [PATCH] fork: Honor task_struct's declared alignment Karl Mehltretter
@ 2026-07-16 17:21 ` Bradley Morgan
2026-07-17 8:14 ` Vlastimil Babka (SUSE)
1 sibling, 0 replies; 3+ messages in thread
From: Bradley Morgan @ 2026-07-16 17:21 UTC (permalink / raw)
To: kmehltretter; +Cc: akpm, kees, linux-kernel, linux-mm, mingo, peterz, vbabka
> Since commit cb7ca40a3882 ("x86/fpu: Make task_struct::thread constant
> size"), struct task_struct is declared __attribute__((aligned(64))) on
> all architectures.
>
> But fork_init() still sets the task_struct slab cache's alignment to
>
> align = max(L1_CACHE_BYTES, ARCH_MIN_TASKALIGN)
>
> which is smaller than 64 on architectures whose cache lines are below
> 64 bytes: e.g. 32 on ARMv5.
>
> In practice plain SLUB happens to hand out 64-byte-aligned objects
> anyway. With CONFIG_SLUB_DEBUG_ON the red-zone padding shifts objects
> to the requested alignment.
>
> With CONFIG_UBSAN_ALIGNMENT=y a boot on QEMU versatilepb (ARM926EJ-S,
> v7.2-rc2, gcc 13.3) floods the console with reports like:
>
> UBSAN: misaligned-access in include/linux/sched.h:2087:9
> member access within misaligned address c295d7e0 for type 'struct task_struct'
> which requires 64 byte alignment
> CPU: 0 UID: 0 PID: 15 Comm: pr/ttyAMA-1 Not tainted 7.2.0-rc2 #1 VOLUNTARY
>
> Set the slab alignment to at least the type's declared alignment.
>
> Fixes: cb7ca40a3882 ("x86/fpu: Make task_struct::thread constant size")
Nice!
Please add
Reviewed-by: Bradley Morgan <include@grrlz.net>
> Assisted-by: Claude:claude-fable-5
> Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
> ---
> kernel/fork.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/fork.c b/kernel/fork.c
> index f0e2e131a9a5..b85cf1343329 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -857,7 +857,8 @@ void __init fork_init(void)
> #ifndef ARCH_MIN_TASKALIGN
> #define ARCH_MIN_TASKALIGN 0
> #endif
> - int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
> + int align = max3(L1_CACHE_BYTES, ARCH_MIN_TASKALIGN,
> + __alignof__(struct task_struct));
> unsigned long useroffset, usersize;
>
> /* create a slab on which task_structs can be allocated */
> --
> 2.39.5 (Apple Git-154)
Thanks!
^ permalink raw reply [flat|nested] 3+ messages in thread
* Re: [PATCH] fork: Honor task_struct's declared alignment
2026-07-10 12:39 [PATCH] fork: Honor task_struct's declared alignment Karl Mehltretter
2026-07-16 17:21 ` Bradley Morgan
@ 2026-07-17 8:14 ` Vlastimil Babka (SUSE)
1 sibling, 0 replies; 3+ messages in thread
From: Vlastimil Babka (SUSE) @ 2026-07-17 8:14 UTC (permalink / raw)
To: Karl Mehltretter, Ingo Molnar, Peter Zijlstra, Andrew Morton
Cc: Kees Cook, linux-mm, linux-kernel, Harry Yoo (Oracle)
On 7/10/26 14:39, Karl Mehltretter wrote:
> Since commit cb7ca40a3882 ("x86/fpu: Make task_struct::thread constant
> size"), struct task_struct is declared __attribute__((aligned(64))) on
> all architectures.
>
> But fork_init() still sets the task_struct slab cache's alignment to
>
> align = max(L1_CACHE_BYTES, ARCH_MIN_TASKALIGN)
>
> which is smaller than 64 on architectures whose cache lines are below
> 64 bytes: e.g. 32 on ARMv5.
>
> In practice plain SLUB happens to hand out 64-byte-aligned objects
> anyway. With CONFIG_SLUB_DEBUG_ON the red-zone padding shifts objects
> to the requested alignment.
>
> With CONFIG_UBSAN_ALIGNMENT=y a boot on QEMU versatilepb (ARM926EJ-S,
> v7.2-rc2, gcc 13.3) floods the console with reports like:
>
> UBSAN: misaligned-access in include/linux/sched.h:2087:9
> member access within misaligned address c295d7e0 for type 'struct task_struct'
> which requires 64 byte alignment
> CPU: 0 UID: 0 PID: 15 Comm: pr/ttyAMA-1 Not tainted 7.2.0-rc2 #1 VOLUNTARY
>
> Set the slab alignment to at least the type's declared alignment.
>
> Fixes: cb7ca40a3882 ("x86/fpu: Make task_struct::thread constant size")
> Assisted-by: Claude:claude-fable-5
> Signed-off-by: Karl Mehltretter <kmehltretter@gmail.com>
LGTM. Additionally I think you can replace the manual L1_CACHE_BYTES
handling with a SLAB_HWCACHE_ALIGN flag. That will apply cache_line_size().
> ---
> kernel/fork.c | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/fork.c b/kernel/fork.c
> index f0e2e131a9a5..b85cf1343329 100644
> --- a/kernel/fork.c
> +++ b/kernel/fork.c
> @@ -857,7 +857,8 @@ void __init fork_init(void)
> #ifndef ARCH_MIN_TASKALIGN
> #define ARCH_MIN_TASKALIGN 0
> #endif
> - int align = max_t(int, L1_CACHE_BYTES, ARCH_MIN_TASKALIGN);
> + int align = max3(L1_CACHE_BYTES, ARCH_MIN_TASKALIGN,
> + __alignof__(struct task_struct));
> unsigned long useroffset, usersize;
>
> /* create a slab on which task_structs can be allocated */
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-07-17 8:14 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2026-07-10 12:39 [PATCH] fork: Honor task_struct's declared alignment Karl Mehltretter
2026-07-16 17:21 ` Bradley Morgan
2026-07-17 8:14 ` Vlastimil Babka (SUSE)
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox