From b1370658804510f6a1c0517a8ff1c9534d371edd Mon Sep 17 00:00:00 2001 From: Geert Uytterhoeven Date: Wed, 9 Sep 2015 13:42:53 +0200 Subject: [PATCH] irqchip/renesas-irqc: Use a separate lockdep class The renesas-irqc interrupt controller is cascaded to the GIC. Hence when propagating wake-up settings to its parent interrupt controller, the following lockdep warning is printed: ============================================= [ INFO: possible recursive locking detected ] 4.2.0-ape6evm-10725-g50fcd7643c034198 #280 Not tainted --------------------------------------------- s2ram/1072 is trying to acquire lock: (&irq_desc_lock_class){-.-...}, at: [] __irq_get_desc_lock+0x58/0x98 but task is already holding lock: (&irq_desc_lock_class){-.-...}, at: [] __irq_get_desc_lock+0x58/0x98 other info that might help us debug this: Possible unsafe locking scenario: CPU0 ---- lock(&irq_desc_lock_class); lock(&irq_desc_lock_class); *** DEADLOCK *** May be due to missing lock nesting notation 6 locks held by s2ram/1072: #0: (sb_writers#7){.+.+.+}, at: [] __sb_start_write+0xa0/0xa8 #1: (&of->mutex){+.+.+.}, at: [] kernfs_fop_write+0x4c/0x1bc #2: (s_active#24){.+.+.+}, at: [] kernfs_fop_write+0x54/0x1bc #3: (pm_mutex){+.+.+.}, at: [] pm_suspend+0x10c/0x510 #4: (&dev->mutex){......}, at: [] __device_suspend+0xdc/0x2cc #5: (&irq_desc_lock_class){-.-...}, at: [] __irq_get_desc_lock+0x58/0x98 stack backtrace: CPU: 0 PID: 1072 Comm: s2ram Not tainted 4.2.0-ape6evm-10725-g50fcd7643c034198 #280 Hardware name: Generic R8A73A4 (Flattened Device Tree) [] (unwind_backtrace) from [] (show_stack+0x10/0x14) [] (show_stack) from [] (dump_stack+0x88/0x98) [] (dump_stack) from [] (__lock_acquire+0x15cc/0x20e4) [] (__lock_acquire) from [] (lock_acquire+0xac/0x12c) [] (lock_acquire) from [] (_raw_spin_lock_irqsave+0x40/0x54) [] (_raw_spin_lock_irqsave) from [] (__irq_get_desc_lock+0x58/0x98) [] (__irq_get_desc_lock) from [] (irq_set_irq_wake+0x20/0xf8) [] (irq_set_irq_wake) from [] (irqc_irq_set_wake+0x20/0x4c) [] (irqc_irq_set_wake) from [] (irq_set_irq_wake+0x8c/0xf8) [] (irq_set_irq_wake) from [] (gpio_keys_suspend+0x74/0xc0) [] (gpio_keys_suspend) from [] (dpm_run_callback+0x54/0x124) Avoid this false positive by using a separate lockdep class for IRQC interrupts. Signed-off-by: Geert Uytterhoeven Cc: Grygorii Strashko Cc: Magnus Damm Cc: Jason Cooper Link: http://lkml.kernel.org/r/1441798974-25716-2-git-send-email-geert%2Brenesas@glider.be Signed-off-by: Thomas Gleixner --- drivers/irqchip/irq-renesas-irqc.c | 7 +++++++ 1 file changed, 7 insertions(+) diff --git a/drivers/irqchip/irq-renesas-irqc.c b/drivers/irqchip/irq-renesas-irqc.c index 2aa3add711a6..4652cc385d39 100644 --- a/drivers/irqchip/irq-renesas-irqc.c +++ b/drivers/irqchip/irq-renesas-irqc.c @@ -150,6 +150,12 @@ static irqreturn_t irqc_irq_handler(int irq, void *dev_id) return IRQ_NONE; } +/* + * This lock class tells lockdep that IRQC irqs are in a different + * category than their parents, so it won't report false recursion. + */ +static struct lock_class_key irqc_irq_lock_class; + static int irqc_irq_domain_map(struct irq_domain *h, unsigned int virq, irq_hw_number_t hw) { @@ -157,6 +163,7 @@ static int irqc_irq_domain_map(struct irq_domain *h, unsigned int virq, irqc_dbg(&p->irq[hw], "map"); irq_set_chip_data(virq, h->host_data); + irq_set_lockdep_class(virq, &irqc_irq_lock_class); irq_set_chip_and_handler(virq, &p->irq_chip, handle_level_irq); return 0; }