Re: [PATCH] pinctrl: renesas: rzt2h: fix invalid wait context

[Sebastian Andrzej Siewior <bigeasy@linutronix.de>]

On 2026-02-04 20:04:48 [+0200], Cosmin Tanislav wrote:
> The following invalid wait context BUG is observed when running with
> CONFIG_PROVE_LOCKING enabled.
> 
> This seems to stem from the fact that __setup_irq() takes a
> raw_spinlock_t, while we take a spinlock_t inside gpiod_get_direction().

From the backtrace yes. It is more that the irqchip used interrupt chip
needs to be able to mask interrupts in hardirq context which it needs a
raw_spinlock_t.

> raw_spinlock_t is a strictly spinning lock, while spinlock_t can sleep
> if PREEMPT_RT is enabled.
> 
> Switch to raw_spinlock_t to fix this.
> 
> [   20.641809] =============================
> [   20.646336] [ BUG: Invalid wait context ]
> [   20.650964] 6.19.0-rc8-next-20260202+ #417 Not tainted
> [   20.656892] -----------------------------
> [   20.661517] (udev-worker)/235 is trying to lock:
> [   20.666842] ffff000182d4ab58 (&pctrl->lock){....}-{3:3}, at: rzt2h_gpio_get_direction+0x3c/0x150
> [   20.676928] other info that might help us debug this:
> [   20.682560] context-{5:5}
> [   20.685572] 4 locks held by (udev-worker)/235:
> [   20.690701]  #0: ffff000180b398f8 (&dev->mutex){....}-{4:4}, at: __driver_attach+0x10c/0x210
> [   20.700480]  #1: ffff000186f41a60 (request_class){+.+.}-{4:4}, at: __setup_irq+0x10c/0x740
> [   20.709952]  #2: ffff000186f418c8 (lock_class){....}-{2:2}, at: __setup_irq+0x12c/0x740
> [   20.719057]  #3: ffff000182d4a690 (&gdev->srcu){.+.+}-{0:0}, at: gpiod_get_direction+0x40/0x24c
> [   20.729135] stack backtrace:
> [   20.732450] CPU: 2 UID: 0 PID: 235 Comm: (udev-worker) Not tainted 6.19.0-rc8-next-20260202+ #417 PREEMPT
> [   20.732459] Hardware name: Renesas RZ/T2H EVK Board based on r9a09g077m44 (DT)
> [   20.732464] Call trace:
> [   20.732467]  show_stack+0x18/0x30 (C)
> [   20.732479]  dump_stack_lvl+0x70/0x98
> [   20.732489]  dump_stack+0x18/0x24
> [   20.732497]  __lock_acquire+0x910/0x1594
> [   20.732508]  lock_acquire+0x2f0/0x42c
> [   20.732517]  _raw_spin_lock_irqsave+0x5c/0x94
> [   20.732529]  rzt2h_gpio_get_direction+0x3c/0x150
> [   20.732538]  gpiochip_get_direction+0x34/0x90
> [   20.732547]  gpiod_get_direction+0xd8/0x24c
> [   20.732556]  gpiochip_lock_as_irq+0x68/0x144
> [   20.732565]  gpiochip_irq_domain_activate+0x18/0x30
> [   20.732574]  __irq_domain_activate_irq+0x60/0xb0
> [   20.732583]  irq_domain_activate_irq+0x34/0x5c
> [   20.732593]  irq_activate+0x2c/0x38
> [   20.732600]  __setup_irq+0x2d8/0x740
> [   20.732610]  request_threaded_irq+0xb0/0x180
> [   20.732620]  request_any_context_irq+0x58/0xac
> [   20.732629]  devm_request_any_context_irq+0x78/0x160
> [   20.732637]  gpio_keys_probe+0x344/0xb40 [gpio_keys]
> [   20.732657]  platform_probe+0x5c/0xa4
> [   20.732666]  really_probe+0xbc/0x2ac
> [   20.732675]  __driver_probe_device+0x78/0x118
> [   20.732684]  driver_probe_device+0x3c/0x170
> [   20.732693]  __driver_attach+0x118/0x210
> [   20.732701]  bus_for_each_dev+0x7c/0xdc
> [   20.732710]  driver_attach+0x24/0x3c
> [   20.732718]  bus_add_driver+0xe4/0x204
> [   20.732727]  driver_register+0x68/0x124
> [   20.732736]  __platform_driver_register+0x20/0x2c
> [   20.732746]  gpio_keys_init+0x20/0x1000 [gpio_keys]
> [   20.732758]  do_one_initcall+0x7c/0x530
> [   20.732770]  do_init_module+0x54/0x23c
> [   20.732780]  load_module+0x181c/0x1ed0
> [   20.732789]  init_module_from_file+0xd8/0xfc
> [   20.732797]  idempotent_init_module+0x1e0/0x2d0
> [   20.732806]  __arm64_sys_finit_module+0x60/0xa4
> [   20.732814]  invoke_syscall.constprop.0+0x40/0xf0
> [   20.732824]  el0_svc_common.constprop.0+0xb8/0xd8
> [   20.732832]  do_el0_svc+0x1c/0x28
> [   20.732839]  el0_svc+0x4c/0x2dc
> [   20.732850]  el0t_64_sync_handler+0xa0/0xe4
> [   20.732860]  el0t_64_sync+0x198/0x19c

Please get rid of this backtrace. It adds no value to the description.

> Fixes: 829dde3369a9 ("pinctrl: renesas: rzt2h: Add GPIO IRQ chip to handle interrupts")
> Signed-off-by: Cosmin Tanislav <cosmin-gabriel.tanislav.xa@renesas.com>

Sebastian