Re: [PATCH] debugobjects: Don't call fill_pool() in early boot non-task context of RT kernel

[Waiman Long <longman@redhat.com>]

On 5/20/26 12:43 PM, Waiman Long wrote:
> When booting a debug PREEMPT_RT kernel on an arm64 system with grace
> processor, the following lockdep splat was reported during early boot.
>
>    ================================
>    WARNING: inconsistent lock state
>    7.1.0-rc4-test+ #1 Not tainted
>    --------------------------------
>    inconsistent {HARDIRQ-ON-W} -> {IN-HARDIRQ-W} usage.
>    swapper/0/0 [HC1[1]:SC0[0]:HE0:SE1] takes:
>    ffff0000803346a0 (&n->list_lock){?.+.}-{3:3}, at: get_from_partial_node+0x74/0xa0
>    {HARDIRQ-ON-W} state was registered at:
>      __lock_acquire+0x3d4/0xb70
>      lock_acquire.part.0+0x178/0x2e0
>      lock_acquire+0xa0/0x240
>      rt_spin_lock+0xa0/0x400
>      __refill_objects_node+0x8c/0x638
>      refill_objects+0x60/0x120
>      __pcs_replace_empty_main+0x11c/0x3a8
>      __kmalloc_noprof+0x550/0x5e0
>      __alloc_workqueue+0x7a4/0xb68
>      alloc_workqueue_noprof+0xc0/0x118
>      kmem_cache_init_late+0x3c/0xd8
>      start_kernel+0x360/0x460
>      __primary_switched+0x8c/0xa0
>    irq event stamp: 12818
>    hardirqs last  enabled at (12817): [<ffffda0f4322be98>] __raw_spin_unlock_irqrestore+0xb8/0xe8
>    hardirqs last disabled at (12818): [<ffffda0f45a611f4>] el1_interrupt+0x34/0xb0
>    softirqs last  enabled at (0): [<0000000000000000>] 0x0
>    softirqs last disabled at (0): [<0000000000000000>] 0x0
>      :
>    Call trace:
>     show_stack+0x20/0x40 (C)
>     dump_stack_lvl+0x7c/0x160
>     dump_stack+0x1c/0x48
>     print_usage_bug.part.0+0x248/0x270
>     mark_lock_irq+0x410/0x608
>     mark_lock+0x1ec/0x3a8
>     mark_usage+0x138/0x170
>     __lock_acquire+0x3d4/0xb70
>     lock_acquire.part.0+0x178/0x2e0
>     lock_acquire+0xa0/0x240
>     rt_spin_lock+0xa0/0x400
>     get_from_partial_node+0x74/0xa0
>     ___slab_alloc+0x94/0x4f8
>     kmem_cache_alloc_noprof+0x2d4/0x598
>     kmem_alloc_batch+0x54/0x170
>     fill_pool+0x12c/0x438
>     debug_objects_fill_pool.part.0+0x88/0x100
>     debug_objects_fill_pool+0x58/0x60
>     debug_object_activate+0xfc/0x3d0
>     add_timer_on+0x250/0x3a0
>     add_interrupt_randomness+0x2d4/0x340
>     handle_percpu_devid_irq+0x2e0/0x4e0
>     handle_irq_desc+0xc0/0x120
>     generic_handle_domain_irq+0x20/0x40
>     __gic_handle_irq_from_irqson.isra.0+0x3c4/0x708
>     gic_handle_irq+0x7c/0xe0
>     call_on_irq_stack+0x30/0x48
>     do_interrupt_handler+0x134/0x158
>     el1_interrupt+0x48/0xb0
>     el1h_64_irq_handler+0x18/0x28
>     el1h_64_irq+0x80/0x88
>
> The {IN-HARDIRQ-W} usage happens when debug_objects_fill_pool() calls
> fill_pool() in the hardirq context during early boot. It is because of
> the "system_state < SYSTEM_SCHEDULING" check in debug_objects_fill_pool()
> which allows fill_pool() to be called from any context during early
> boot.
>
> It shouldn't really be a problem as fill_poll() will not called from
> non-preemptible context after early boot, but the lockdep warning can
> still cause confusion and anxiety. Fix that by further restricting the
> call to only in_task() context during early boot.

Sorry, the above paragraph isn't right. Will send out a v2.

Cheers,
Longman

> Fixes: 06e0ae988f6e ("debugobjects: Allow to refill the pool before SYSTEM_SCHEDULING")
> Signed-off-by: Waiman Long <longman@redhat.com>
> ---
>   lib/debugobjects.c | 7 +++++--
>   1 file changed, 5 insertions(+), 2 deletions(-)
>
> diff --git a/lib/debugobjects.c b/lib/debugobjects.c
> index 12e2e42e6a31..236ea5e716df 100644
> --- a/lib/debugobjects.c
> +++ b/lib/debugobjects.c
> @@ -727,11 +727,14 @@ static void debug_objects_fill_pool(void)
>   
>   	/*
>   	 * On RT enabled kernels the pool refill must happen in preemptible
> -	 * context -- for !RT kernels we rely on the fact that spinlock_t and
> +	 * context or in task context during early boot.
> +	 *
> +	 * For !RT kernels we rely on the fact that spinlock_t and
>   	 * raw_spinlock_t are basically the same type and this lock-type
>   	 * inversion works just fine.
>   	 */
> -	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible() || system_state < SYSTEM_SCHEDULING) {
> +	if (!IS_ENABLED(CONFIG_PREEMPT_RT) || preemptible() ||
> +	   (system_state < SYSTEM_SCHEDULING && in_task())) {
>   		/*
>   		 * Annotate away the spinlock_t inside raw_spinlock_t warning
>   		 * by temporarily raising the wait-type to LD_WAIT_CONFIG, matching