Re: [Intel-gfx] [PATCH] rtc: cmos: Don't enable interrupts in the middle of the interrupt handler

From: Chris Wilson
Date: Wed Oct 19 2016 - 14:17:00 EST


On Wed, Oct 19, 2016 at 09:02:04PM +0300, ville.syrjala@xxxxxxxxxxxxxxx wrote:
> From: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
>
> Using spin_lock_irq()/spin_unlock_irq() from within the interrupt
> handler is a no-no. Let's save/restore the flags to avoid turning on
> interrupts prematurely.
>
> We hit this in a bunch of our CI systems, but for whatever reason I
> wasn't able to reproduce on my own machine, so this fix is just
> based on the backtrace.
>
> [ 202.634918] WARNING: CPU: 0 PID: 0 at kernel/locking/lockdep.c:2729 trace_hardirqs_on_caller+0x113/0x1b0
> [ 202.634919] DEBUG_LOCKS_WARN_ON(current->hardirq_context)
> [ 202.634929] Modules linked in: snd_hda_intel i915 x86_pkg_temp_thermal intel_powerclamp coretemp crct10dif_pclmul crc32_pclmul ghash_clmulni_intel lpc_ich snd_hda_codec_realtek snd_hda_codec_generic snd_hda_codec_hdmi snd_hda_codec snd_hwdep i2c_designware_platform i2c_designware_core snd_hda_core mei_me mei snd_pcm r8169 mii sdhci_acpi sdhci mmc_core i2c_hid [last unloaded: i915]
> [ 202.634930] CPU: 0 PID: 0 Comm: swapper/0 Tainted: G U 4.9.0-rc1-CI-CI_DRM_1734+ #1
> [ 202.634931] Hardware name: GIGABYTE M4HM87P-00/M4HM87P-00, BIOS F6 12/10/2014
> [ 202.634933] ffff88011ea03d68 ffffffff8142dce5 ffff88011ea03db8 0000000000000000
> [ 202.634934] ffff88011ea03da8 ffffffff8107e496 00000aa900000002 ffffffff81e249a0
> [ 202.634935] ffffffff81815637 ffffffff82e7c280 0000000000000000 0000000000000004
> [ 202.634936] Call Trace:
> [ 202.634939] <IRQ>
> [ 202.634939] [<ffffffff8142dce5>] dump_stack+0x67/0x92
> [ 202.634941] [<ffffffff8107e496>] __warn+0xc6/0xe0
> [ 202.634944] [<ffffffff81815637>] ? _raw_spin_unlock_irq+0x27/0x50
> [ 202.634945] [<ffffffff8107e4fa>] warn_slowpath_fmt+0x4a/0x50
> [ 202.634946] [<ffffffff810d6d83>] trace_hardirqs_on_caller+0x113/0x1b0
> [ 202.634948] [<ffffffff810d6e2d>] trace_hardirqs_on+0xd/0x10
> [ 202.634949] [<ffffffff81815637>] _raw_spin_unlock_irq+0x27/0x50
> [ 202.634951] [<ffffffff81672042>] rtc_handler+0x32/0xa0
> [ 202.634954] [<ffffffff814c08a3>] acpi_ev_fixed_event_detect+0xd4/0xfb
> [ 202.634956] [<ffffffff814c2ccb>] acpi_ev_sci_xrupt_handler+0xf/0x2d
> [ 202.634957] [<ffffffff814ab3ee>] acpi_irq+0x11/0x2c
> [ 202.634960] [<ffffffff810e5288>] __handle_irq_event_percpu+0x58/0x370
> [ 202.634961] [<ffffffff810e55be>] handle_irq_event_percpu+0x1e/0x50
> [ 202.634962] [<ffffffff810e5624>] handle_irq_event+0x34/0x60
> [ 202.634963] [<ffffffff810e8906>] handle_fasteoi_irq+0xa6/0x170
> [ 202.634966] [<ffffffff8101eef5>] handle_irq+0x15/0x20
> [ 202.634967] [<ffffffff8101e548>] do_IRQ+0x68/0x130
> [ 202.634968] [<ffffffff81816789>] common_interrupt+0x89/0x89
> [ 202.634970] <EOI>
> [ 202.634970] [<ffffffff81814c73>] ? mwait_idle+0x93/0x210
> [ 202.634971] [<ffffffff81814c6a>] ? mwait_idle+0x8a/0x210
> [ 202.634972] [<ffffffff81026b0a>] arch_cpu_idle+0xa/0x10
> [ 202.634973] [<ffffffff8181509e>] default_idle_call+0x1e/0x30
> [ 202.634974] [<ffffffff810cbf6c>] cpu_startup_entry+0x17c/0x1f0
> [ 202.634976] [<ffffffff8180ca87>] rest_init+0x127/0x130
> [ 202.634978] [<ffffffff81f77f08>] start_kernel+0x3f6/0x403
> [ 202.634980] [<ffffffff81f7728f>] x86_64_start_reservations+0x2a/0x2c
> [ 202.634981] [<ffffffff81f77404>] x86_64_start_kernel+0x173/0x186
> [ 202.634982] ---[ end trace 293c99618fa08d34 ]---
>
> Cc: Gabriele Mazzotta <gabriele.mzt@xxxxxxxxx>
> Cc: Alexandre Belloni <alexandre.belloni@xxxxxxxxxxxxxxxxxx>
> Fixes: 983bf1256edb ("rtc: cmos: Clear ACPI-driven alarms upon resume")
> Signed-off-by: Ville Syrjälä <ville.syrjala@xxxxxxxxxxxxxxx>
> ---
> drivers/rtc/rtc-cmos.c | 5 +++--
> 1 file changed, 3 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/rtc/rtc-cmos.c b/drivers/rtc/rtc-cmos.c
> index dd3d59806ffa..19cd49ad92dc 100644
> --- a/drivers/rtc/rtc-cmos.c
> +++ b/drivers/rtc/rtc-cmos.c
> @@ -996,8 +996,9 @@ static u32 rtc_handler(void *context)
> struct cmos_rtc *cmos = dev_get_drvdata(dev);
> unsigned char rtc_control = 0;
> unsigned char rtc_intr;
> + unsigned long flags;
>
> - spin_lock_irq(&rtc_lock);
> + spin_lock_irqsave(&rtc_lock, flags);

My reading of the acpi fixed event handlers suggests that this can only
be called from interrupt context.

Anyway using irqsave is correct (just may be optimised away to a plain
spin_lock() if the condition that this is only called in irq context is
satisfied).

Reviewed-by: Chris Wilson <chris@xxxxxxxxxxxxxxxxxx>
-Chris

--
Chris Wilson, Intel Open Source Technology Centre