Re: [PATCH] coresight: cti: remove pm_runtime_get_sync() from CPU hotplug

From: Mike Leach
Date: Tue Aug 18 2020 - 08:47:45 EST


On Tue, 18 Aug 2020 at 12:11, Tingwei Zhang <tingwei@xxxxxxxxxxxxxx> wrote:
>
> Below BUG is triggered by call pm_runtime_get_sync() in
> cti_cpuhp_enable_hw(). It's in CPU hotplug callback with interrupt
> disabled. Pm_runtime_get_sync() calls clock driver to enable clock
> which could sleep. Remove pm_runtime_get_sync() in cti_cpuhp_enable_hw()
> since pm_runtime_get_sync() is called in cti_enabld and pm_runtime_put()
> is called in cti_disabled. No need to increase pm count when CPU gets
> online since it's not decreased when CPU is offline.
>
> [ 105.800279] BUG: scheduling while atomic: swapper/1/0/0x00000002
> [ 105.800290] Modules linked in:
> [ 105.800327] CPU: 1 PID: 0 Comm: swapper/1 Tainted: G W
> 5.9.0-rc1-gff1304be0a05-dirty #21
> [ 105.800337] Hardware name: Thundercomm Dragonboard 845c (DT)
> [ 105.800353] Call trace:
> [ 105.800414] dump_backtrace+0x0/0x1d4
> [ 105.800439] show_stack+0x14/0x1c
> [ 105.800462] dump_stack+0xc0/0x100
> [ 105.800490] __schedule_bug+0x58/0x74
> [ 105.800523] __schedule+0x590/0x65c
> [ 105.800538] schedule+0x78/0x10c
> [ 105.800553] schedule_timeout+0x188/0x250
> [ 105.800585] qmp_send.constprop.10+0x12c/0x1b0
> [ 105.800599] qmp_qdss_clk_prepare+0x18/0x20
> [ 105.800622] clk_core_prepare+0x48/0xd4
> [ 105.800639] clk_prepare+0x20/0x34
> [ 105.800663] amba_pm_runtime_resume+0x54/0x90
> [ 105.800695] __rpm_callback+0xdc/0x138
> [ 105.800709] rpm_callback+0x24/0x78
> [ 105.800724] rpm_resume+0x328/0x47c
> [ 105.800739] __pm_runtime_resume+0x50/0x74
> [ 105.800768] cti_starting_cpu+0x40/0xa4
> [ 105.800795] cpuhp_invoke_callback+0x84/0x1e0
> [ 105.800814] notify_cpu_starting+0x9c/0xb8
> [ 105.800834] secondary_start_kernel+0xd8/0x164
> [ 105.800933] CPU1: Booted secondary processor 0x0000000100 [0x517f803c]
>
> Fixes: e9b880581d55 ("coresight: cti: Add CPU Hotplug handling to CTI driver")
> Signed-off-by: Tingwei Zhang <tingwei@xxxxxxxxxxxxxx>
> ---
> drivers/hwtracing/coresight/coresight-cti.c | 3 ---
> 1 file changed, 3 deletions(-)
>
> diff --git a/drivers/hwtracing/coresight/coresight-cti.c b/drivers/hwtracing/coresight/coresight-cti.c
> index 869569eb8c7f..baba6af83440 100644
> --- a/drivers/hwtracing/coresight/coresight-cti.c
> +++ b/drivers/hwtracing/coresight/coresight-cti.c
> @@ -126,9 +126,7 @@ static int cti_enable_hw(struct cti_drvdata *drvdata)
> static void cti_cpuhp_enable_hw(struct cti_drvdata *drvdata)
> {
> struct cti_config *config = &drvdata->config;
> - struct device *dev = &drvdata->csdev->dev;
>
> - pm_runtime_get_sync(dev->parent);
> spin_lock(&drvdata->spinlock);
> config->hw_powered = true;
>
> @@ -148,7 +146,6 @@ static void cti_cpuhp_enable_hw(struct cti_drvdata *drvdata)
> /* did not re-enable due to no claim / no request */
> cti_hp_not_enabled:
> spin_unlock(&drvdata->spinlock);
> - pm_runtime_put(dev->parent);
> }
>
> /* disable hardware */
> --
> The Qualcomm Innovation Center, Inc. is a member of the Code Aurora Forum,
> a Linux Foundation Collaborative Project
>
Reviewed-by Mike Leach <mike.leach@xxxxxxxxxx>

--
Mike Leach
Principal Engineer, ARM Ltd.
Manchester Design Centre. UK