* [PATCH] mm/slub: clamp slab_min_order to MAX_PAGE_ORDER
@ 2026-10-07 16:34 Sang-Heon Jeon
2026-10-07 17:47 ` Harry Yoo
0 siblings, 1 reply; 3+ messages in thread
From: Sang-Heon Jeon @ 2026-10-07 16:34 UTC (permalink / raw)
To: Andrew Morton, Harry Yoo, Vlastimil Babka
Cc: Christoph Lameter, David Rientjes, Hao Li, linux-mm,
Roman Gushchin, stable
setup_slub_max_order() clamps slab_max_order to MAX_PAGE_ORDER, but
setup_slub_min_order() has no such clamp.
The buddy allocator cannot allocate more than 2^MAX_PAGE_ORDER pages
at once, so slab_min_order above it makes no sense.
If slab_min_order is set above MAX_PAGE_ORDER, it can even panic at
boot or invoke undefined behavior in order_objects().
So clamp slab_min_order to MAX_PAGE_ORDER as well.
Fixes: 818cf5909701 ("slub: enforce MAX_ORDER")
Cc: stable@vger.kernel.org
Signed-off-by: Sang-Heon Jeon <ekffu200098@gmail.com>
---
QEMU-based test results
- Booted with "slab_min_order=20"
1) AS-IS (before fix)
[ 0.094926] mem auto-init: stack:all(zero), heap alloc:off, heap free:off
[ 0.097776] Kernel panic - not syncing: Creation of kmalloc slab kmem_cache_node size=64 failed. Reason -22
[ 0.099542] CPU: 0 UID: 0 PID: 0 Comm: swapper Not tainted 7.2.0 #1 PREEMPT(undef)
[ 0.100903] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
[ 0.102607] Call Trace:
[ 0.103040] <TASK>
[ 0.103417] vpanic+0x32a/0x4b0
[ 0.103976] panic+0x5e/0x60
[ 0.104491] create_boot_cache+0xe3/0xf0
[ 0.105194] kmem_cache_init+0x12d/0x410
[ 0.105890] ? mem_init+0x1fc/0x210
[ 0.106521] mm_core_init+0xba/0x140
[ 0.107153] start_kernel+0x441/0x8b0
[ 0.107814] x86_64_start_reservations+0x18/0x30
[ 0.108638] x86_64_start_kernel+0x10c/0x120
[ 0.109401] common_startup_64+0x13e/0x158
[ 0.110145] </TASK>
[ 0.110584] ---[ end Kernel panic - not syncing: Creation of kmalloc slab kmem_cache_node size=64 failed. Reason -22 ]---
2) TO-BE (after fix)
[ 0.008976] mem auto-init: stack:all(zero), heap alloc:off, heap free:off
[ 0.011366] SLUB: HWalign=64, Order=10-10, MinObjects=0, CPUs=4, Nodes=1
Also, I'm not sure about the Fixes commit. Please change it if you know
a better one.
---
mm/slub.c | 2 ++
1 file changed, 2 insertions(+)
diff --git a/mm/slub.c b/mm/slub.c
index 1559cbace469..8400df7c222e 100644
--- a/mm/slub.c
+++ b/mm/slub.c
@@ -8354,6 +8354,8 @@ static int __init setup_slub_min_order(const char *str, const struct kernel_para
if (ret)
return ret;
+ slub_min_order = min_t(unsigned int, slub_min_order, MAX_PAGE_ORDER);
+
if (slub_min_order > slub_max_order)
slub_max_order = slub_min_order;
base-commit: d755441b7672c74f29d1241ffdd2b067ecab1ca0
--
2.43.0
^ permalink raw reply related [flat|nested] 3+ messages in thread* Re: [PATCH] mm/slub: clamp slab_min_order to MAX_PAGE_ORDER
2026-10-07 16:34 [PATCH] mm/slub: clamp slab_min_order to MAX_PAGE_ORDER Sang-Heon Jeon
@ 2026-10-07 17:47 ` Harry Yoo
2026-10-08 15:06 ` Sang-Heon Jeon
0 siblings, 1 reply; 3+ messages in thread
From: Harry Yoo @ 2026-10-07 17:47 UTC (permalink / raw)
To: Sang-Heon Jeon
Cc: Andrew Morton, Vlastimil Babka, Christoph Lameter, David Rientjes,
Hao Li, linux-mm, Roman Gushchin, stable
On Thu, Oct 08, 2026 at 01:34:02AM +0900, Sang-Heon Jeon wrote:
> setup_slub_max_order() clamps slab_max_order to MAX_PAGE_ORDER, but
> setup_slub_min_order() has no such clamp.
>
> The buddy allocator cannot allocate more than 2^MAX_PAGE_ORDER pages
> at once, so slab_min_order above it makes no sense.
>
> If slab_min_order is set above MAX_PAGE_ORDER, it can even panic at
> boot or invoke undefined behavior in order_objects().
>
> So clamp slab_min_order to MAX_PAGE_ORDER as well.
>
> Fixes: 818cf5909701 ("slub: enforce MAX_ORDER")
I think Fixes: should be the commit that introduced slub_min_order.
> Cc: stable@vger.kernel.org
And no need for Cc: stable? It's been there for decades and people
should realize the kernel parameter is broken when it doesn't boot.
Otherwise looks good to me.
> Signed-off-by: Sang-Heon Jeon <ekffu200098@gmail.com>
> ---
> QEMU-based test results
>
> - Booted with "slab_min_order=20"
>
> 1) AS-IS (before fix)
> [ 0.094926] mem auto-init: stack:all(zero), heap alloc:off, heap free:off
> [ 0.097776] Kernel panic - not syncing: Creation of kmalloc slab kmem_cache_node size=64 failed. Reason -22
> [ 0.099542] CPU: 0 UID: 0 PID: 0 Comm: swapper Not tainted 7.2.0 #1 PREEMPT(undef)
> [ 0.100903] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> [ 0.102607] Call Trace:
> [ 0.103040] <TASK>
> [ 0.103417] vpanic+0x32a/0x4b0
> [ 0.103976] panic+0x5e/0x60
> [ 0.104491] create_boot_cache+0xe3/0xf0
> [ 0.105194] kmem_cache_init+0x12d/0x410
> [ 0.105890] ? mem_init+0x1fc/0x210
> [ 0.106521] mm_core_init+0xba/0x140
> [ 0.107153] start_kernel+0x441/0x8b0
> [ 0.107814] x86_64_start_reservations+0x18/0x30
> [ 0.108638] x86_64_start_kernel+0x10c/0x120
> [ 0.109401] common_startup_64+0x13e/0x158
> [ 0.110145] </TASK>
> [ 0.110584] ---[ end Kernel panic - not syncing: Creation of kmalloc slab kmem_cache_node size=64 failed. Reason -22 ]---
>
> 2) TO-BE (after fix)
> [ 0.008976] mem auto-init: stack:all(zero), heap alloc:off, heap free:off
> [ 0.011366] SLUB: HWalign=64, Order=10-10, MinObjects=0, CPUs=4, Nodes=1
>
> Also, I'm not sure about the Fixes commit. Please change it if you know
> a better one.
>
> ---
> mm/slub.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/mm/slub.c b/mm/slub.c
> index 1559cbace469..8400df7c222e 100644
> --- a/mm/slub.c
> +++ b/mm/slub.c
> @@ -8354,6 +8354,8 @@ static int __init setup_slub_min_order(const char *str, const struct kernel_para
> if (ret)
> return ret;
>
> + slub_min_order = min_t(unsigned int, slub_min_order, MAX_PAGE_ORDER);
> +
> if (slub_min_order > slub_max_order)
> slub_max_order = slub_min_order;
>
>
> base-commit: d755441b7672c74f29d1241ffdd2b067ecab1ca0
> --
> 2.43.0
>
--
Cheers,
Harry / Hyeonggon
^ permalink raw reply [flat|nested] 3+ messages in thread* Re: [PATCH] mm/slub: clamp slab_min_order to MAX_PAGE_ORDER
2026-10-07 17:47 ` Harry Yoo
@ 2026-10-08 15:06 ` Sang-Heon Jeon
0 siblings, 0 replies; 3+ messages in thread
From: Sang-Heon Jeon @ 2026-10-08 15:06 UTC (permalink / raw)
To: Harry Yoo
Cc: Andrew Morton, Vlastimil Babka, Christoph Lameter, David Rientjes,
Hao Li, linux-mm, Roman Gushchin, stable
Hi, Harry
2026년 10월 8일 (목) 오전 2:47, Harry Yoo <harry@kernel.org>님이 작성:
>
> On Thu, Oct 08, 2026 at 01:34:02AM +0900, Sang-Heon Jeon wrote:
> > setup_slub_max_order() clamps slab_max_order to MAX_PAGE_ORDER, but
> > setup_slub_min_order() has no such clamp.
> >
> > The buddy allocator cannot allocate more than 2^MAX_PAGE_ORDER pages
> > at once, so slab_min_order above it makes no sense.
> >
> > If slab_min_order is set above MAX_PAGE_ORDER, it can even panic at
> > boot or invoke undefined behavior in order_objects().
> >
> > So clamp slab_min_order to MAX_PAGE_ORDER as well.
> >
> > Fixes: 818cf5909701 ("slub: enforce MAX_ORDER")
>
> I think Fixes: should be the commit that introduced slub_min_order.
Thanks for your guidance. I'll switch this to 81819f0fc828 ("SLUB core") in v2.
> > Cc: stable@vger.kernel.org
>
> And no need for Cc: stable? It's been there for decades and people
> should realize the kernel parameter is broken when it doesn't boot.
Okay, I'll remove it in v2.
> Otherwise looks good to me.
Thanks for the review :)
> > Signed-off-by: Sang-Heon Jeon <ekffu200098@gmail.com>
> > ---
> > QEMU-based test results
> >
> > - Booted with "slab_min_order=20"
> >
> > 1) AS-IS (before fix)
> > [ 0.094926] mem auto-init: stack:all(zero), heap alloc:off, heap free:off
> > [ 0.097776] Kernel panic - not syncing: Creation of kmalloc slab kmem_cache_node size=64 failed. Reason -22
> > [ 0.099542] CPU: 0 UID: 0 PID: 0 Comm: swapper Not tainted 7.2.0 #1 PREEMPT(undef)
> > [ 0.100903] Hardware name: QEMU Standard PC (i440FX + PIIX, 1996), BIOS 1.16.3-debian-1.16.3-2 04/01/2014
> > [ 0.102607] Call Trace:
> > [ 0.103040] <TASK>
> > [ 0.103417] vpanic+0x32a/0x4b0
> > [ 0.103976] panic+0x5e/0x60
> > [ 0.104491] create_boot_cache+0xe3/0xf0
> > [ 0.105194] kmem_cache_init+0x12d/0x410
> > [ 0.105890] ? mem_init+0x1fc/0x210
> > [ 0.106521] mm_core_init+0xba/0x140
> > [ 0.107153] start_kernel+0x441/0x8b0
> > [ 0.107814] x86_64_start_reservations+0x18/0x30
> > [ 0.108638] x86_64_start_kernel+0x10c/0x120
> > [ 0.109401] common_startup_64+0x13e/0x158
> > [ 0.110145] </TASK>
> > [ 0.110584] ---[ end Kernel panic - not syncing: Creation of kmalloc slab kmem_cache_node size=64 failed. Reason -22 ]---
> >
> > 2) TO-BE (after fix)
> > [ 0.008976] mem auto-init: stack:all(zero), heap alloc:off, heap free:off
> > [ 0.011366] SLUB: HWalign=64, Order=10-10, MinObjects=0, CPUs=4, Nodes=1
> >
> > Also, I'm not sure about the Fixes commit. Please change it if you know
> > a better one.
> >
> > ---
> > mm/slub.c | 2 ++
> > 1 file changed, 2 insertions(+)
> >
> > diff --git a/mm/slub.c b/mm/slub.c
> > index 1559cbace469..8400df7c222e 100644
> > --- a/mm/slub.c
> > +++ b/mm/slub.c
> > @@ -8354,6 +8354,8 @@ static int __init setup_slub_min_order(const char *str, const struct kernel_para
> > if (ret)
> > return ret;
> >
> > + slub_min_order = min_t(unsigned int, slub_min_order, MAX_PAGE_ORDER);
> > +
> > if (slub_min_order > slub_max_order)
> > slub_max_order = slub_min_order;
> >
> >
> > base-commit: d755441b7672c74f29d1241ffdd2b067ecab1ca0
> > --
> > 2.43.0
> >
>
> --
> Cheers,
> Harry / Hyeonggon
Best regards,
Sang-Heon Jeon
^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2026-10-08 15:07 UTC | newest]
Thread overview: 3+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-10-07 16:34 [PATCH] mm/slub: clamp slab_min_order to MAX_PAGE_ORDER Sang-Heon Jeon
2026-10-07 17:47 ` Harry Yoo
2026-10-08 15:06 ` Sang-Heon Jeon
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox