Hello, Thomas,

Testing next-20250508 with lockdep enabled got a splat that is shown in
all its glory below.  Reverting this commit makes the problem go away:

88a4df117ad6 ("genirq/cpuhotplug: Convert to lock guards")

Alternatively, changing that commit's final "scoped_guard(raw_spinlock"
to "scoped_guard(raw_spinlock_irq" also fixes things, as shown in the
incremental patch at the very end of this email.  If this is a proper fix,
please feel free to fold it into your original.

							Thanx, Paul

------------------------------------------------------------------------

[    0.608026] ================================
[    0.608026] WARNING: inconsistent lock state
[    0.608026] 6.15.0-rc5-next-20250508 #4883 Not tainted
[    0.608026] --------------------------------
[    0.608026] inconsistent {IN-HARDIRQ-W} -> {HARDIRQ-ON-W} usage.
[    0.608026] cpuhp/1/21 [HC0[0]:SC0[0]:HE1:SE1] takes:
[    0.608026] ffff91f0010650c0 (&irq_desc_lock_class){?...}-{2:2}, at: irq_affinity_online_cpu+0x51/0x120
[    0.608026] {IN-HARDIRQ-W} state was registered at:
[    0.608026]   lock_acquire+0xbd/0x2d0
[    0.608026]   _raw_spin_lock+0x2b/0x40
[    0.608026]   handle_level_irq+0x18/0x160
[    0.608026]   __common_interrupt+0x4a/0xf0
[    0.608026]   common_interrupt+0x78/0x90
[    0.608026]   asm_common_interrupt+0x26/0x40
[    0.608026]   _raw_spin_unlock_irqrestore+0x34/0x50
[    0.608026]   __setup_irq+0x460/0x730
[    0.608026]   request_threaded_irq+0x10b/0x180
[    0.608026]   hpet_time_init+0x35/0x50
[    0.608026]   x86_late_time_init+0x16/0x40
[    0.608026]   start_kernel+0x6b6/0x810
[    0.608026]   x86_64_start_reservations+0x18/0x30
[    0.608026]   x86_64_start_kernel+0xc6/0xe0
[    0.608026]   common_startup_64+0x13e/0x148
[    0.608026] irq event stamp: 35
[    0.608026] hardirqs last  enabled at (35): [<ffffffffb8ede1fb>] finish_task_switch.isra.0+0xeb/0x300
[    0.608026] hardirqs last disabled at (34): [<ffffffffb9f01c69>] __schedule+0x909/0xfa0
[    0.608026] softirqs last  enabled at (0): [<ffffffffb8e9061b>] copy_process+0x98b/0x2070
[    0.608026] softirqs last disabled at (0): [<0000000000000000>] 0x0
[    0.608026]
[    0.608026] other info that might help us debug this:
[    0.608026]  Possible unsafe locking scenario:
[    0.608026]
[    0.608026]        CPU0
[    0.608026]        ----
[    0.608026]   lock(&irq_desc_lock_class);
[    0.608026]   <Interrupt>
[    0.608026]     lock(&irq_desc_lock_class);
[    0.608026]
[    0.608026]  *** DEADLOCK ***
[    0.608026]
[    0.608026] 3 locks held by cpuhp/1/21:
[    0.608026]  #0: ffffffffba86d630 (cpu_hotplug_lock){++++}-{0:0}, at: cpuhp_thread_fun+0x4d/0x200
[    0.608026]  #1: ffffffffba86fd40 (cpuhp_state-up){+.+.}-{0:0}, at: cpuhp_thread_fun+0x4d/0x200
[    0.608026]  #2: ffffffffba95d6c8 (sparse_irq_lock){+.+.}-{4:4}, at: irq_affinity_online_cpu+0x16/0x120
[    0.608026]
[    0.608026] stack backtrace:
[    0.608026] CPU: 1 UID: 0 PID: 21 Comm: cpuhp/1 Not tainted 6.15.0-rc5-next-20250508 #4883 PREEMPT(full)
[    0.608026] Hardware name: QEMU Standard PC (Q35 + ICH9, 2009), BIOS rel-1.14.0-0-g155821a1990b-prebuilt.qemu.org 04/01/2014
[    0.608026] Call Trace:
[    0.608026]  <TASK>
[    0.608026]  dump_stack_lvl+0x68/0xa0
[    0.608026]  print_usage_bug.part.0+0x22d/0x2d0
[    0.608026]  mark_lock+0x594/0x630
[    0.608026]  ? __lock_acquire+0x64d/0x18d0
[    0.608026]  __lock_acquire+0x408/0x18d0
[    0.608026]  ? mtree_load+0x81/0x5a0
[    0.608026]  ? find_held_lock+0x2b/0x80
[    0.608026]  lock_acquire+0xbd/0x2d0
[    0.608026]  ? irq_affinity_online_cpu+0x51/0x120
[    0.608026]  _raw_spin_lock+0x2b/0x40
[    0.608026]  ? irq_affinity_online_cpu+0x51/0x120
[    0.608026]  irq_affinity_online_cpu+0x51/0x120
[    0.608026]  ? __pfx_irq_affinity_online_cpu+0x10/0x10
[    0.608026]  cpuhp_invoke_callback+0x2e8/0x440
[    0.608026]  ? lock_release+0xc6/0x290
[    0.608026]  ? cpuhp_thread_fun+0x4d/0x200
[    0.608026]  cpuhp_thread_fun+0x181/0x200
[    0.608026]  ? __pfx_smpboot_thread_fn+0x10/0x10
[    0.608026]  smpboot_thread_fn+0xee/0x220
[    0.608026]  kthread+0x102/0x200
[    0.608026]  ? __pfx_kthread+0x10/0x10
[    0.608026]  ret_from_fork+0x165/0x1b0
[    0.608026]  ? __pfx_kthread+0x10/0x10
[    0.608026]  ret_from_fork_asm+0x1a/0x30
[    0.608026]  </TASK>

------------------------------------------------------------------------

diff --git a/kernel/irq/cpuhotplug.c b/kernel/irq/cpuhotplug.c
index 7bd4c2a5cef4b..e77ca6db5e11e 100644
--- a/kernel/irq/cpuhotplug.c
+++ b/kernel/irq/cpuhotplug.c
@@ -243,7 +243,7 @@ int irq_affinity_online_cpu(unsigned int cpu)
 	irq_lock_sparse();
 	for_each_active_irq(irq) {
 		desc = irq_to_desc(irq);
-		scoped_guard(raw_spinlock, &desc->lock)
+		scoped_guard(raw_spinlock_irq, &desc->lock)
 			irq_restore_affinity_of_irq(desc, cpu);
 	}
 	irq_unlock_sparse();

On Thu, May 08 2025 at 15:29, Paul E. McKenney wrote:
> Testing next-20250508 with lockdep enabled got a splat that is shown in
> all its glory below.  Reverting this commit makes the problem go away:
>
> 88a4df117ad6 ("genirq/cpuhotplug: Convert to lock guards")

https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=irq/core

On Fri, May 09, 2025 at 08:12:01AM +0200, Thomas Gleixner wrote:
> On Thu, May 08 2025 at 15:29, Paul E. McKenney wrote:
> > Testing next-20250508 with lockdep enabled got a splat that is shown in
> > all its glory below.  Reverting this commit makes the problem go away:
> >
> > 88a4df117ad6 ("genirq/cpuhotplug: Convert to lock guards")
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/tip/tip.git/commit/?h=irq/core

Very good, thank you, and feel free to add my Tested-by, though it
appears to already be set in stone.

							Thanx, Paul