clocksource: exynos_mct: Avoid blocking calls in the cpu hotplug notifier
authorDamian Eppel <d.eppel@samsung.com>
Fri, 26 Jun 2015 13:23:04 +0000 (15:23 +0200)
committerThomas Gleixner <tglx@linutronix.de>
Fri, 26 Jun 2015 19:53:01 +0000 (21:53 +0200)
Whilst testing cpu hotplug events on kernel configured with
DEBUG_PREEMPT and DEBUG_ATOMIC_SLEEP we get following BUG message,
caused by calling request_irq() and free_irq() in the context of
hotplug notification (which is in this case atomic context).

[   40.785859] CPU1: Software reset
[   40.786660] BUG: sleeping function called from invalid context at mm/slub.c:1241
[   40.786668] in_atomic(): 1, irqs_disabled(): 128, pid: 0, name: swapper/1
[   40.786678] Preemption disabled at:[<  (null)>]   (null)
[   40.786681]
[   40.786692] CPU: 1 PID: 0 Comm: swapper/1 Not tainted 3.19.0-rc4-00024-g7dca860 #36
[   40.786698] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[   40.786728] [<c0014a00>] (unwind_backtrace) from [<c0011980>] (show_stack+0x10/0x14)
[   40.786747] [<c0011980>] (show_stack) from [<c0449ba0>] (dump_stack+0x70/0xbc)
[   40.786767] [<c0449ba0>] (dump_stack) from [<c00c6124>] (kmem_cache_alloc+0xd8/0x170)
[   40.786785] [<c00c6124>] (kmem_cache_alloc) from [<c005d6f8>] (request_threaded_irq+0x64/0x128)
[   40.786804] [<c005d6f8>] (request_threaded_irq) from [<c0350b8c>] (exynos4_local_timer_setup+0xc0/0x13c)
[   40.786820] [<c0350b8c>] (exynos4_local_timer_setup) from [<c0350ca8>] (exynos4_mct_cpu_notify+0x30/0xa8)
[   40.786838] [<c0350ca8>] (exynos4_mct_cpu_notify) from [<c003b330>] (notifier_call_chain+0x44/0x84)
[   40.786857] [<c003b330>] (notifier_call_chain) from [<c0022fd4>] (__cpu_notify+0x28/0x44)
[   40.786873] [<c0022fd4>] (__cpu_notify) from [<c0013714>] (secondary_start_kernel+0xec/0x150)
[   40.786886] [<c0013714>] (secondary_start_kernel) from [<40008764>] (0x40008764)

Interrupts cannot be requested/freed in the CPU_STARTING/CPU_DYING
notifications which run on the hotplugged cpu with interrupts and
preemption disabled.

To avoid the issue, request the interrupts for all possible cpus in
the boot code. The interrupts are marked NO_AUTOENABLE to avoid a racy
request_irq/disable_irq() sequence. The flag prevents the
request_irq() code from enabling the interrupt immediately.

The interrupt is then enabled in the CPU_STARTING notifier of the
hotplugged cpu and again disabled with disable_irq_nosync() in the
CPU_DYING notifier.

[ tglx: Massaged changelog to match the patch ]

Fixes: 7114cd749a12 ("clocksource: exynos_mct: use (request/free)_irq calls for local timer registration")
Reported-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Reviewed-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Tested-by: Krzysztof Kozlowski <k.kozlowski@samsung.com>
Tested-by: Marcin Jabrzyk <m.jabrzyk@samsung.com>
Signed-off-by: Damian Eppel <d.eppel@samsung.com>
Cc: m.szyprowski@samsung.com
Cc: kyungmin.park@samsung.com
Cc: daniel.lezcano@linaro.org
Cc: kgene@kernel.org
Cc: linux-arm-kernel@lists.infradead.org
Link: http://lkml.kernel.org/r/1435324984-7328-1-git-send-email-d.eppel@samsung.com
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>
Cc: <stable@vger.kernel.org>
drivers/clocksource/exynos_mct.c

index 935b05936dbdd9588764b0c04bf84db32b425af0..9064ff743598a42515333f073cbd3d86193a2524 100644 (file)
@@ -462,15 +462,12 @@ static int exynos4_local_timer_setup(struct clock_event_device *evt)
        exynos4_mct_write(TICK_BASE_CNT, mevt->base + MCT_L_TCNTB_OFFSET);
 
        if (mct_int_type == MCT_INT_SPI) {
-               evt->irq = mct_irqs[MCT_L0_IRQ + cpu];
-               if (request_irq(evt->irq, exynos4_mct_tick_isr,
-                               IRQF_TIMER | IRQF_NOBALANCING,
-                               evt->name, mevt)) {
-                       pr_err("exynos-mct: cannot register IRQ %d\n",
-                               evt->irq);
+
+               if (evt->irq == -1)
                        return -EIO;
-               }
-               irq_force_affinity(mct_irqs[MCT_L0_IRQ + cpu], cpumask_of(cpu));
+
+               irq_force_affinity(evt->irq, cpumask_of(cpu));
+               enable_irq(evt->irq);
        } else {
                enable_percpu_irq(mct_irqs[MCT_L0_IRQ], 0);
        }
@@ -483,10 +480,12 @@ static int exynos4_local_timer_setup(struct clock_event_device *evt)
 static void exynos4_local_timer_stop(struct clock_event_device *evt)
 {
        evt->set_mode(CLOCK_EVT_MODE_UNUSED, evt);
-       if (mct_int_type == MCT_INT_SPI)
-               free_irq(evt->irq, this_cpu_ptr(&percpu_mct_tick));
-       else
+       if (mct_int_type == MCT_INT_SPI) {
+               if (evt->irq != -1)
+                       disable_irq_nosync(evt->irq);
+       } else {
                disable_percpu_irq(mct_irqs[MCT_L0_IRQ]);
+       }
 }
 
 static int exynos4_mct_cpu_notify(struct notifier_block *self,
@@ -518,7 +517,7 @@ static struct notifier_block exynos4_mct_cpu_nb = {
 
 static void __init exynos4_timer_resources(struct device_node *np, void __iomem *base)
 {
-       int err;
+       int err, cpu;
        struct mct_clock_event_device *mevt = this_cpu_ptr(&percpu_mct_tick);
        struct clk *mct_clk, *tick_clk;
 
@@ -545,7 +544,25 @@ static void __init exynos4_timer_resources(struct device_node *np, void __iomem
                WARN(err, "MCT: can't request IRQ %d (%d)\n",
                     mct_irqs[MCT_L0_IRQ], err);
        } else {
-               irq_set_affinity(mct_irqs[MCT_L0_IRQ], cpumask_of(0));
+               for_each_possible_cpu(cpu) {
+                       int mct_irq = mct_irqs[MCT_L0_IRQ + cpu];
+                       struct mct_clock_event_device *pcpu_mevt =
+                               per_cpu_ptr(&percpu_mct_tick, cpu);
+
+                       pcpu_mevt->evt.irq = -1;
+
+                       irq_set_status_flags(mct_irq, IRQ_NOAUTOEN);
+                       if (request_irq(mct_irq,
+                                       exynos4_mct_tick_isr,
+                                       IRQF_TIMER | IRQF_NOBALANCING,
+                                       pcpu_mevt->name, pcpu_mevt)) {
+                               pr_err("exynos-mct: cannot register IRQ (cpu%d)\n",
+                                                                       cpu);
+
+                               continue;
+                       }
+                       pcpu_mevt->evt.irq = mct_irq;
+               }
        }
 
        err = register_cpu_notifier(&exynos4_mct_cpu_nb);