Re: [PATCH] ACPI / osl: add acpi_os_down_wait to avoid a schedule BUG

[Lan Tianyu <tianyu.lan@intel.com>]

On 2015年05月28日 14:33, Pan, XinhuiX wrote:
> acpi_os_wait_semaphore can be called in local/hard irq disabled path. like in cpu up/down callback.
> So when dirver try to acquire the semaphore, current code may call down_wait which might sleep.
> Then hit panic as we can't schedule here. So introduce acpi_os_down_wait to cover such case.
> acpi_os_down_wait use down_trylock, and use cpu_relax to wait the semaphore signalled if preempt is disabled.
> 
> below is the panic.

Hi Xinhui:

Does this issue happen in the latest upstream kernel?
In the latest code, acpi_cpu_soft_notify() doesn't detail with CPU_DYING
event and return directly. The issue should not take place.

> 
> [ 1148.230132, 1]smpboot: CPU 3 is now offline
> [ 1148.277288, 0]smpboot: CPU 2 is now offline
> [ 1148.322385, 1]BUG: scheduling while atomic: migration/1/13/0x00000002
> [ 1148.329604, 1]Modules linked in: hid_sensor_hub sens_col_core hid_heci_ish heci_ish heci vidt_driver atomisp_css2401a0_v21 lm3642 8723bs(O) cfg80211 gc2235 bt_lpm videobuf_vmalloc 6lowpan_iphc i        p6table_raw iptable_raw videobuf_core rfkill_gpio atmel_mxt_ts
> [ 1148.355276, 1]CPU: 1 PID: 13 Comm: migration/1 Tainted: G        W  O 3.14.37-x86_64-L1-R409-g73e8207 #25
> [ 1148.365983, 1]Hardware name: Intel Corporation CHERRYVIEW C0 PLATFORM/Cherry Trail CR, BIOS CH2TCR.X64.0004.R48.1504211851 04/21/2015
> [ 1148.379397, 1] ffff880077801140 ffff880073233a58 ffffffff819eec6c ffff8800732303d0
> [ 1148.387914, 1] ffff880073233a70 ffffffff819eb0e0 ffff88007ac92240 ffff880073233ad0
> [ 1148.396430, 1] ffffffff819f790a ffff8800732303d0 ffff880073233fd8 0000000000012240
> [ 1148.404948, 1]Call Trace:
> [ 1148.407912, 1] [<ffffffff819eec6c>] dump_stack+0x4e/0x7a
> [ 1148.413872, 1] [<ffffffff819eb0e0>] __schedule_bug+0x58/0x67
> [ 1148.420219, 1] [<ffffffff819f790a>] __schedule+0x67a/0x7b0
> [ 1148.426369, 1] [<ffffffff819f7a69>] schedule+0x29/0x70
> [ 1148.432123, 1] [<ffffffff819f6ce9>] schedule_timeout+0x269/0x310
> [ 1148.438860, 1] [<ffffffff810c519c>] ? update_group_power+0x16c/0x260
> [ 1148.445988, 1] [<ffffffff819fae59>] __down_common+0x91/0xd6
> [ 1148.452236, 1] [<ffffffff810bff00>] ? update_cfs_rq_blocked_load+0xc0/0x130
> [ 1148.460036, 1] [<ffffffff819faf11>] __down_timeout+0x16/0x18
> [ 1148.466380, 1] [<ffffffff810d21cc>] down_timeout+0x4c/0x60
> [ 1148.472534, 1] [<ffffffff813f1cf9>] acpi_os_wait_semaphore+0x43/0x57
> [ 1148.479658, 1] [<ffffffff81419ad8>] acpi_ut_acquire_mutex+0x48/0x88
> [ 1148.486683, 1] [<ffffffff813f54e7>] ? acpi_match_device+0x4d/0x4d
> [ 1148.493516, 1] [<ffffffff814119fe>] acpi_get_data+0x35/0x77
> [ 1148.499761, 1] [<ffffffff813f547d>] acpi_bus_get_device+0x21/0x3e
> [ 1148.506593, 1] [<ffffffff8141e52b>] acpi_cpu_soft_notify+0x3d/0xd3
> [ 1148.513522, 1] [<ffffffff81a00223>] notifier_call_chain+0x53/0xa0
> [ 1148.520356, 1] [<ffffffff8110b701>] ? cpu_stop_park+0x51/0x70
> [ 1148.526801, 1] [<ffffffff810b082e>] __raw_notifier_call_chain+0xe/0x10
> [ 1148.534118, 1] [<ffffffff81088963>] cpu_notify+0x23/0x50
> [ 1148.540075, 1] [<ffffffff819e64f7>] take_cpu_down+0x27/0x40
> [ 1148.546322, 1] [<ffffffff8110b831>] multi_cpu_stop+0xc1/0x110
> [ 1148.552763, 1] [<ffffffff8110b770>] ? cpu_stop_should_run+0x50/0x50
> [ 1148.559776, 1] [<ffffffff8110ba48>] cpu_stopper_thread+0x78/0x150
> [ 1148.566608, 1] [<ffffffff819fc1ee>] ? _raw_spin_unlock_irq+0x1e/0x40
> [ 1148.573730, 1] [<ffffffff810b4257>] ? finish_task_switch+0x57/0xd0
> [ 1148.580646, 1] [<ffffffff819f760e>] ? __schedule+0x37e/0x7b0
> [ 1148.586991, 1] [<ffffffff810b2f7d>] smpboot_thread_fn+0x17d/0x2b0
> [ 1148.593819, 1] [<ffffffff810b2e00>] ? SyS_setgroups+0x160/0x160
> [ 1148.600455, 1] [<ffffffff810ab9b4>] kthread+0xe4/0x100
> [ 1148.606208, 1] [<ffffffff810ab8d0>] ? kthread_create_on_node+0x190/0x190
> [ 1148.613721, 1] [<ffffffff81a044c8>] ret_from_fork+0x58/0x90
> [ 1148.619967, 1] [<ffffffff810ab8d0>] ? kthread_create_on_node+0x190/0x190
> 
> Signed-off-by: Pan Xinhui <xinhuix.pan@intel.com>
> ---
>   drivers/acpi/osl.c | 28 +++++++++++++++++++++++++++-
>   1 file changed, 27 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/acpi/osl.c b/drivers/acpi/osl.c
> index 7ccba39..57a1812 100644
> --- a/drivers/acpi/osl.c
> +++ b/drivers/acpi/osl.c
> @@ -1195,6 +1195,32 @@ void acpi_os_wait_events_complete(void)
>         flush_workqueue(kacpi_notify_wq);
>   }
> 
> +static int acpi_os_down_wait(struct semaphore *sem, long jiffies_timeout)
> +{
> +       unsigned long deadline_time;
> +       int ret = 0;
> +
> +       if (down_trylock(sem)) {
> +               if (unlikely(preempt_count())) {
> +                       deadline_time = jiffies + jiffies_timeout;
> +                       while (true) {
> +                               cpu_relax();
> +
> +                               if (!down_trylock(sem))
> +                                       break;
> +
> +                               if (time_after(jiffies, deadline_time)) {
> +                                       ret = -ETIME;
> +                                       break;
> +                               }
> +                       }
> +               } else
> +                       ret = down_timeout(sem, jiffies_timeout);
> +       }
> +
> +       return ret;
> +}
> +
>   struct acpi_hp_work {
>         struct work_struct work;
>         struct acpi_device *adev;
> @@ -1309,7 +1335,7 @@ acpi_status acpi_os_wait_semaphore(acpi_handle handle, u32 units, u16 timeout)
>         else
>                 jiffies = msecs_to_jiffies(timeout);
> 
> -       ret = down_timeout(sem, jiffies);
> +       ret = acpi_os_down_wait(sem, jiffies);
>         if (ret)
>                 status = AE_TIME;
> 
> --
> 1.9.1
> --
> To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 


-- 
Best regards
Tianyu Lan
--
To unsubscribe from this list: send the line "unsubscribe linux-acpi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html