| Message ID | 20190214013042.254790-5-mka@chromium.org |
|---|---|
| State | Deferred |
| Headers | show |
| Series | PM / devfreq: Refactor load monitoring | expand |
Hi Matthias,
As I commented on the first patch, it is not possible to call some codes
according to the intention of each governor between 'devfreq_moniotr_*()'
and some codes which are executed before or after 'devfreq_moniotr_*()'
For example, if some governor requires the following sequence,
after this patch, it is not possible.
case DEVFREQ_GOV_xxx:
/* execute some code before devfreq_monitor_xxx() */
devfreq_monitor_xxx()
/* execute some code after devfreq_monitor_xxx() */
2019년 2월 14일 (목) 오후 7:16, Matthias Kaehlcke <mka@chromium.org>님이 작성:
>
> devfreq expects governors to call devfreq_monitor_start/stop()
> in response to DEVFREQ_GOV_START/STOP events. Since the devfreq
> core itself generates these events and invokes the governor's event
> handler the start/stop of the load monitor can be done in the common
> code.
>
> Call devfreq_monitor_start/stop() when the governor reports a
> successful handling of DEVFREQ_GOV_START/STOP and remove these
> calls from the simpleondemand and tegra governors. Make
> devfreq_monitor_start/stop() static since these functions are now
> only used by the devfreq core. For better integration with the
> callers the functions must now be called with devfreq->lock held.
>
> Also stop manipulating the monitor workqueue directly, use
> devfreq_monitor_start/stop() instead.
>
> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> ---
> drivers/devfreq/devfreq.c | 45 +++++++++++------------
> drivers/devfreq/governor.h | 2 -
> drivers/devfreq/governor_simpleondemand.c | 8 ----
> drivers/devfreq/tegra-devfreq.c | 2 -
> 4 files changed, 22 insertions(+), 35 deletions(-)
>
> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
> index eb86461648d74..ac553c00f6790 100644
> --- a/drivers/devfreq/devfreq.c
> +++ b/drivers/devfreq/devfreq.c
> @@ -404,14 +404,14 @@ static void devfreq_monitor(struct work_struct *work)
> * devfreq_monitor_start() - Start load monitoring of devfreq instance
> * @devfreq: the devfreq instance.
> *
> - * Helper function for starting devfreq device load monitoing. By
> - * default delayed work based monitoring is supported. Function
> - * to be called from governor in response to DEVFREQ_GOV_START
> - * event when device is added to devfreq framework.
> + * Helper function for starting devfreq device load monitoring. By
> + * default delayed work based monitoring is supported. Must be called
> + * with devfreq->lock held.
> */
> -void devfreq_monitor_start(struct devfreq *devfreq)
> +static void devfreq_monitor_start(struct devfreq *devfreq)
> {
> - mutex_lock(&devfreq->lock);
> + WARN(!mutex_is_locked(&devfreq->lock),
> + "devfreq->lock must be held by the caller.\n");
>
> if (devfreq->profile->polling_ms) {
> INIT_DEFERRABLE_WORK(&devfreq->work, devfreq_monitor);
> @@ -420,28 +420,28 @@ void devfreq_monitor_start(struct devfreq *devfreq)
>
> devfreq->monitor_state = DEVFREQ_MONITOR_RUNNING;
> }
> -
> - mutex_unlock(&devfreq->lock);
> }
> -EXPORT_SYMBOL(devfreq_monitor_start);
>
> /**
> * devfreq_monitor_stop() - Stop load monitoring of a devfreq instance
> * @devfreq: the devfreq instance.
> *
> - * Helper function to stop devfreq device load monitoing. Function
> - * to be called from governor in response to DEVFREQ_GOV_STOP
> - * event when device is removed from devfreq framework.
> + * Helper function to stop devfreq device load monitoring. Must be
> + * called with devfreq->lock held.
> */
> -void devfreq_monitor_stop(struct devfreq *devfreq)
> +static void devfreq_monitor_stop(struct devfreq *devfreq)
> {
> + /* mutex must be held for symmetry with _start() */
> + WARN(!mutex_is_locked(&devfreq->lock),
> + "devfreq->lock must be held by the caller.\n");
> +
> + mutex_unlock(&devfreq->lock);
> cancel_delayed_work_sync(&devfreq->work);
>
> mutex_lock(&devfreq->lock);
> devfreq->monitor_state = DEVFREQ_MONITOR_STOPPED;
> mutex_unlock(&devfreq->lock);
> }
> -EXPORT_SYMBOL(devfreq_monitor_stop);
>
> /**
> * devfreq_monitor_suspend() - Suspend load monitoring of a devfreq instance
> @@ -518,27 +518,22 @@ void devfreq_interval_update(struct devfreq *devfreq, unsigned int *delay)
>
> /* if new delay is zero, stop polling */
> if (!new_delay) {
> + devfreq_monitor_stop(devfreq);
> mutex_unlock(&devfreq->lock);
> - cancel_delayed_work_sync(&devfreq->work);
> - devfreq->monitor_state == DEVFREQ_MONITOR_STOPPED;
> return;
> }
>
> /* if current delay is zero, start polling with new delay */
> if (!cur_delay) {
> - queue_delayed_work(devfreq_wq, &devfreq->work,
> - msecs_to_jiffies(devfreq->profile->polling_ms));
> + devfreq_monitor_start(devfreq);
> goto out;
> }
>
> /* if current delay is greater than new delay, restart polling */
> if (cur_delay > new_delay) {
> - mutex_unlock(&devfreq->lock);
> - cancel_delayed_work_sync(&devfreq->work);
> - mutex_lock(&devfreq->lock);
> + devfreq_monitor_stop(devfreq);
> if (devfreq->monitor_state != DEVFREQ_MONITOR_SUSPENDED)
> - queue_delayed_work(devfreq_wq, &devfreq->work,
> - msecs_to_jiffies(devfreq->profile->polling_ms));
> + devfreq_monitor_start(devfreq);
> }
> out:
> mutex_unlock(&devfreq->lock);
> @@ -601,6 +596,8 @@ static int devfreq_governor_start(struct devfreq *devfreq)
> return err;
> }
>
> + devfreq_monitor_start(devfreq);
> +
> return 0;
> }
>
> @@ -624,6 +621,8 @@ static int devfreq_governor_stop(struct devfreq *devfreq)
> return err;
> }
>
> + devfreq_monitor_stop(devfreq);
> +
> return 0;
> }
>
> diff --git a/drivers/devfreq/governor.h b/drivers/devfreq/governor.h
> index d136792c0cc91..47efe747b6f79 100644
> --- a/drivers/devfreq/governor.h
> +++ b/drivers/devfreq/governor.h
> @@ -57,8 +57,6 @@ struct devfreq_governor {
> unsigned int event, void *data);
> };
>
> -extern void devfreq_monitor_start(struct devfreq *devfreq);
> -extern void devfreq_monitor_stop(struct devfreq *devfreq);
> extern void devfreq_interval_update(struct devfreq *devfreq,
> unsigned int *delay);
>
> diff --git a/drivers/devfreq/governor_simpleondemand.c b/drivers/devfreq/governor_simpleondemand.c
> index 52eb0c734b312..e0f0944a9c7aa 100644
> --- a/drivers/devfreq/governor_simpleondemand.c
> +++ b/drivers/devfreq/governor_simpleondemand.c
> @@ -91,14 +91,6 @@ static int devfreq_simple_ondemand_handler(struct devfreq *devfreq,
> unsigned int event, void *data)
> {
> switch (event) {
> - case DEVFREQ_GOV_START:
> - devfreq_monitor_start(devfreq);
> - break;
> -
> - case DEVFREQ_GOV_STOP:
> - devfreq_monitor_stop(devfreq);
> - break;
> -
> case DEVFREQ_GOV_INTERVAL:
> devfreq_interval_update(devfreq, (unsigned int *)data);
> break;
> diff --git a/drivers/devfreq/tegra-devfreq.c b/drivers/devfreq/tegra-devfreq.c
> index 79efa1e51bd06..515fb852dbad6 100644
> --- a/drivers/devfreq/tegra-devfreq.c
> +++ b/drivers/devfreq/tegra-devfreq.c
> @@ -580,13 +580,11 @@ static int tegra_governor_event_handler(struct devfreq *devfreq,
>
> switch (event) {
> case DEVFREQ_GOV_START:
> - devfreq_monitor_start(devfreq);
> tegra_actmon_enable_interrupts(tegra);
> break;
>
> case DEVFREQ_GOV_STOP:
> tegra_actmon_disable_interrupts(tegra);
> - devfreq_monitor_stop(devfreq);
> break;
>
> case DEVFREQ_GOV_SUSPEND:
> --
> 2.20.1.791.gb4d0f1c61a-goog
>
Hi Matthias,
I have compiled and run your changes on Odroid xu3 and v5.0-rc6.
There are kernel warnings because of mutex not held in function
devfreq_monitor_[start|stop]() in use cases:
1) a few times during registration of new devices devfreq_add_device()
2) poking the device from sysfs
ad 1)
-----8<--------------------------------------------------------
[ 10.146986] exynos5-dmc 10c20000.memory-controller: Linked as a
consumer to regulator.39
[ 10.154619] ------------[ cut here ]------------
[ 10.158210] WARNING: CPU: 6 PID: 58 at drivers/devfreq/devfreq.c:414
devfreq_monitor_start+0xbc/0xdc
[ 10.167370] devfreq->lock must be held by the caller.
[ 10.172329] Modules linked in:
[ 10.175405] CPU: 6 PID: 58 Comm: kworker/6:1 Tainted: G W
5.0.0-rc6-00015-gd7c6e73 #43
[ 10.184565] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[ 10.190624] Workqueue: events deferred_probe_work_func
[ 10.195732] [<c0111a28>] (unwind_backtrace) from [<c010d8d4>]
(show_stack+0x10/0x14)
[ 10.203454] [<c010d8d4>] (show_stack) from [<c09efac0>]
(dump_stack+0x90/0xc8)
[ 10.210643] [<c09efac0>] (dump_stack) from [<c012591c>]
(__warn+0xf8/0x124)
[ 10.217573] [<c012591c>] (__warn) from [<c0125a08>]
(warn_slowpath_fmt+0x38/0x48)
[ 10.225029] [<c0125a08>] (warn_slowpath_fmt) from [<c07160ec>]
(devfreq_monitor_start+0xbc/0xdc)
[ 10.233786] [<c07160ec>] (devfreq_monitor_start) from [<c0716144>]
(devfreq_governor_start+0x38/0x84)
[ 10.242973] [<c0716144>] (devfreq_governor_start) from [<c07167b0>]
(devfreq_add_device+0x270/0x454)
[ 10.252072] [<c07167b0>] (devfreq_add_device) from [<c07169e0>]
(devm_devfreq_add_device+0x4c/0x80)
[ 10.261083] [<c07169e0>] (devm_devfreq_add_device) from [<c0719760>]
(exynos5_dmc_probe+0x490/0x6b4)
[ 10.270183] [<c0719760>] (exynos5_dmc_probe) from [<c054cb34>]
(platform_drv_probe+0x48/0x98)
[ 10.278672] [<c054cb34>] (platform_drv_probe) from [<c054a6b4>]
(really_probe+0x224/0x3f4)
[ 10.286903] [<c054a6b4>] (really_probe) from [<c054aad8>]
(driver_probe_device+0x70/0x1c4)
[ 10.295135] [<c054aad8>] (driver_probe_device) from [<c0548888>]
(bus_for_each_drv+0x44/0x8c)
[ 10.303628] [<c0548888>] (bus_for_each_drv) from [<c054a970>]
(__device_attach+0xa0/0x138)
[ 10.311859] [<c054a970>] (__device_attach) from [<c05497e8>]
(bus_probe_device+0x88/0x90)
[ 10.320005] [<c05497e8>] (bus_probe_device) from [<c0549d60>]
(deferred_probe_work_func+0x6c/0xbc)
[ 10.328933] [<c0549d60>] (deferred_probe_work_func) from [<c0143f1c>]
(process_one_work+0x200/0x740)
[ 10.338032] [<c0143f1c>] (process_one_work) from [<c0144484>]
(process_scheduled_works+0x28/0x38)
[ 10.346869] [<c0144484>] (process_scheduled_works) from [<c014469c>]
(worker_thread+0x208/0x4c8)
[ 10.355623] [<c014469c>] (worker_thread) from [<c014ac74>]
(kthread+0x128/0x164)
[ 10.362981] [<c014ac74>] (kthread) from [<c01010b4>]
(ret_from_fork+0x14/0x20)
[ 10.370171] Exception stack(0xee1d3fb0 to 0xee1d3ff8)
[ 10.375187] 3fa0: 00000000
00000000 00000000 00000000
[ 10.383348] 3fc0: 00000000 00000000 00000000 00000000 00000000
00000000 00000000 00000000
[ 10.391493] 3fe0: 00000000 00000000 00000000 00000000 00000013 00000000
[ 10.398132] irq event stamp: 32781
[ 10.401439] hardirqs last enabled at (32789): [<c018c030>]
console_unlock+0x474/0x700
[ 10.409382] hardirqs last disabled at (32806): [<c018bc7c>]
console_unlock+0xc0/0x700
[ 10.417180] softirqs last enabled at (32822): [<c01024cc>]
__do_softirq+0x3a4/0x66c
[ 10.424892] softirqs last disabled at (32833): [<c012d6b8>]
irq_exit+0x140/0x168
[ 10.432216] ---[ end trace 92e0ab098cadffbc ]---
[ 10.436841] exynos5-dmc 10c20000.memory-controller: DMC init for
prod_id=0xe5422001 pkg_id=0x2706832a
------->8------------------------------------------------------
ad 2)
-----8<--------------------------------------------------------
root@(none):/# echo simple_ondemand > /sys/class/devfreq/devfreq16/governor
[ 594.577414] ------------[ cut here ]------------
[ 594.581009] WARNING: CPU: 4 PID: 1 at drivers/devfreq/devfreq.c:436
devfreq_monitor_stop+0x58/0x64
[ 594.590032] devfreq->lock must be held by the caller.
[ 594.596509] Modules linked in:
[ 594.598149] CPU: 4 PID: 1 Comm: bash Tainted: G W
5.0.0-rc6-00015-gd7c6e73 #43
[ 594.606618] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[ 594.612682] [<c0111a28>] (unwind_backtrace) from [<c010d8d4>]
(show_stack+0x10/0x14)
[ 594.620398] [<c010d8d4>] (show_stack) from [<c09efac0>]
(dump_stack+0x90/0xc8)
[ 594.627586] [<c09efac0>] (dump_stack) from [<c012591c>]
(__warn+0xf8/0x124)
[ 594.634515] [<c012591c>] (__warn) from [<c0125a08>]
(warn_slowpath_fmt+0x38/0x48)
[ 594.641972] [<c0125a08>] (warn_slowpath_fmt) from [<c0715968>]
(devfreq_monitor_stop+0x58/0x64)
[ 594.650642] [<c0715968>] (devfreq_monitor_stop) from [<c07159ac>]
(devfreq_governor_stop+0x38/0x84)
[ 594.659656] [<c07159ac>] (devfreq_governor_stop) from [<c0716224>]
(governor_store+0x94/0x124)
[ 594.668242] [<c0716224>] (governor_store) from [<c0544f54>]
(dev_attr_store+0x18/0x24)
[ 594.676118] [<c0544f54>] (dev_attr_store) from [<c02fba24>]
(sysfs_kf_write+0x4c/0x50)
[ 594.683999] [<c02fba24>] (sysfs_kf_write) from [<c02fab70>]
(kernfs_fop_write+0xfc/0x1e4)
[ 594.692150] [<c02fab70>] (kernfs_fop_write) from [<c027f3ec>]
(__vfs_write+0x2c/0x160)
[ 594.700031] [<c027f3ec>] (__vfs_write) from [<c027f6a8>]
(vfs_write+0xa4/0x16c)
[ 594.707308] [<c027f6a8>] (vfs_write) from [<c027f884>]
(ksys_write+0x40/0x8c)
[ 594.714412] [<c027f884>] (ksys_write) from [<c0101000>]
(ret_fast_syscall+0x0/0x28)
[ 594.722037] Exception stack(0xee8fbfa8 to 0xee8fbff0)
[ 594.727053] bfa0: 00000010 000d0408 00000001
000d0408 00000010 00000000
[ 594.735213] bfc0: 00000010 000d0408 b6e84d60 00000004 00000000
000c8164 000c5730 000c5760
[ 594.743357] bfe0: 00000000 be8908ac b6de801b b6e23b46
[ 594.748470] irq event stamp: 411445
[ 594.751832] hardirqs last enabled at (411453): [<c018c030>]
console_unlock+0x474/0x700
[ 594.759875] hardirqs last disabled at (411476): [<c018bc7c>]
console_unlock+0xc0/0x700
[ 594.767749] softirqs last enabled at (411494): [<c01024cc>]
__do_softirq+0x3a4/0x66c
[ 594.775598] softirqs last disabled at (411505): [<c012d6b8>]
irq_exit+0x140/0x168
[ 594.783064] ---[ end trace 92e0ab098cadffbd ]---
[ 594.787535]
[ 594.788987] =====================================
[ 594.793668] WARNING: bad unlock balance detected!
[ 594.798348] 5.0.0-rc6-00015-gd7c6e73 #43 Tainted: G W
[ 594.804674] -------------------------------------
[ 594.809353] bash/1 is trying to release lock (&devfreq->lock) at:
[ 594.815421] [<c0715930>] devfreq_monitor_stop+0x20/0x64
[ 594.820618] but there are no more locks to release!
[ 594.825470]
[ 594.825470] other info that might help us debug this:
[ 594.831972] 4 locks held by bash/1:
[ 594.835434] #0: a7707d8c (sb_writers#6){.+.+}, at: vfs_write+0x130/0x16c
[ 594.842196] #1: a7f34c9a (&of->mutex){+.+.}, at:
kernfs_fop_write+0xc0/0x1e4
[ 594.849301] #2: ce189988 (kn->count#35){.+.+}, at:
kernfs_fop_write+0xc8/0x1e4
[ 594.856581] #3: ab1648d8 (devfreq_list_lock){+.+.}, at:
governor_store+0x44/0x124
[ 594.864120]
[ 594.864120] stack backtrace:
[ 594.868453] CPU: 4 PID: 1 Comm: bash Tainted: G W
5.0.0-rc6-00015-gd7c6e73 #43
[ 594.876946] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[ 594.883014] [<c0111a28>] (unwind_backtrace) from [<c010d8d4>]
(show_stack+0x10/0x14)
[ 594.890727] [<c010d8d4>] (show_stack) from [<c09efac0>]
(dump_stack+0x90/0xc8)
[ 594.897920] [<c09efac0>] (dump_stack) from [<c017e094>]
(print_unlock_imbalance_bug+0xb0/0xe0)
[ 594.906499] [<c017e094>] (print_unlock_imbalance_bug) from
[<c01811b0>] (lock_release+0x190/0x380)
[ 594.915427] [<c01811b0>] (lock_release) from [<c0a0aa08>]
(__mutex_unlock_slowpath+0x30/0x274)
[ 594.924004] [<c0a0aa08>] (__mutex_unlock_slowpath) from [<c0715930>]
(devfreq_monitor_stop+0x20/0x64)
[ 594.933189] [<c0715930>] (devfreq_monitor_stop) from [<c07159ac>]
(devfreq_governor_stop+0x38/0x84)
[ 594.942201] [<c07159ac>] (devfreq_governor_stop) from [<c0716224>]
(governor_store+0x94/0x124)
[ 594.950781] [<c0716224>] (governor_store) from [<c0544f54>]
(dev_attr_store+0x18/0x24)
[ 594.958665] [<c0544f54>] (dev_attr_store) from [<c02fba24>]
(sysfs_kf_write+0x4c/0x50)
[ 594.966549] [<c02fba24>] (sysfs_kf_write) from [<c02fab70>]
(kernfs_fop_write+0xfc/0x1e4)
[ 594.974696] [<c02fab70>] (kernfs_fop_write) from [<c027f3ec>]
(__vfs_write+0x2c/0x160)
[ 594.982581] [<c027f3ec>] (__vfs_write) from [<c027f6a8>]
(vfs_write+0xa4/0x16c)
[ 594.989860] [<c027f6a8>] (vfs_write) from [<c027f884>]
(ksys_write+0x40/0x8c)
[ 594.996964] [<c027f884>] (ksys_write) from [<c0101000>]
(ret_fast_syscall+0x0/0x28)
[ 595.004588] Exception stack(0xee8fbfa8 to 0xee8fbff0)
[ 595.009613] bfa0: 00000010 000d0408 00000001
000d0408 00000010 00000000
[ 595.017761] bfc0: 00000010 000d0408 b6e84d60 00000004 00000000
000c8164 000c5730 000c5760
[ 595.025906] bfe0: 00000000 be8908ac b6de801b b6e23b46
[ 595.031013] ------------[ cut here ]------------
[ 595.035572] WARNING: CPU: 4 PID: 1 at drivers/devfreq/devfreq.c:414
devfreq_monitor_start+0xbc/0xdc
[ 595.044590] devfreq->lock must be held by the caller.
[ 595.049570] Modules linked in:
[ 595.052656] CPU: 4 PID: 1 Comm: bash Tainted: G W
5.0.0-rc6-00015-gd7c6e73 #43
[ 595.061106] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[ 595.067168] [<c0111a28>] (unwind_backtrace) from [<c010d8d4>]
(show_stack+0x10/0x14)
[ 595.074886] [<c010d8d4>] (show_stack) from [<c09efac0>]
(dump_stack+0x90/0xc8)
[ 595.082074] [<c09efac0>] (dump_stack) from [<c012591c>]
(__warn+0xf8/0x124)
[ 595.089005] [<c012591c>] (__warn) from [<c0125a08>]
(warn_slowpath_fmt+0x38/0x48)
[ 595.096461] [<c0125a08>] (warn_slowpath_fmt) from [<c07160ec>]
(devfreq_monitor_start+0xbc/0xdc)
[ 595.105217] [<c07160ec>] (devfreq_monitor_start) from [<c0716144>]
(devfreq_governor_start+0x38/0x84)
[ 595.114404] [<c0716144>] (devfreq_governor_start) from [<c0716248>]
(governor_store+0xb8/0x124)
[ 595.123069] [<c0716248>] (governor_store) from [<c0544f54>]
(dev_attr_store+0x18/0x24)
[ 595.130951] [<c0544f54>] (dev_attr_store) from [<c02fba24>]
(sysfs_kf_write+0x4c/0x50)
[ 595.138835] [<c02fba24>] (sysfs_kf_write) from [<c02fab70>]
(kernfs_fop_write+0xfc/0x1e4)
[ 595.146983] [<c02fab70>] (kernfs_fop_write) from [<c027f3ec>]
(__vfs_write+0x2c/0x160)
[ 595.154868] [<c027f3ec>] (__vfs_write) from [<c027f6a8>]
(vfs_write+0xa4/0x16c)
[ 595.162145] [<c027f6a8>] (vfs_write) from [<c027f884>]
(ksys_write+0x40/0x8c)
[ 595.169248] [<c027f884>] (ksys_write) from [<c0101000>]
(ret_fast_syscall+0x0/0x28)
[ 595.176874] Exception stack(0xee8fbfa8 to 0xee8fbff0)
[ 595.181893] bfa0: 00000010 000d0408 00000001
000d0408 00000010 00000000
[ 595.190049] bfc0: 00000010 000d0408 b6e84d60 00000004 00000000
000c8164 000c5730 000c5760
[ 595.198193] bfe0: 00000000 be8908ac b6de801b b6e23b46
[ 595.203283] irq event stamp: 411551
[ 595.206672] hardirqs last enabled at (411551): [<c018c030>]
console_unlock+0x474/0x700
[ 595.214690] hardirqs last disabled at (411550): [<c018bc7c>]
console_unlock+0xc0/0x700
[ 595.222572] softirqs last enabled at (411540): [<c01024cc>]
__do_softirq+0x3a4/0x66c
[ 595.230342] softirqs last disabled at (411527): [<c012d6b8>]
irq_exit+0x140/0x168
[ 595.237852] ---[ end trace 92e0ab098cadffbe ]---
root@(none):/# echo 50 > /sys/class/devfreq/devfreq16/polling_interval
[ 595.472717] ------------[ cut here ]------------
[ 595.475872] WARNING: CPU: 7 PID: 1 at drivers/devfreq/devfreq.c:414
devfreq_monitor_start+0xbc/0xdc
[ 595.484926] devfreq->lock must be held by the caller.
[ 595.489903] Modules linked in:
[ 595.492963] CPU: 7 PID: 1 Comm: bash Tainted: G W
5.0.0-rc6-00015-gd7c6e73 #43
[ 595.501439] Hardware name: SAMSUNG EXYNOS (Flattened Device Tree)
[ 595.507502] [<c0111a28>] (unwind_backtrace) from [<c010d8d4>]
(show_stack+0x10/0x14)
[ 595.515218] [<c010d8d4>] (show_stack) from [<c09efac0>]
(dump_stack+0x90/0xc8)
[ 595.522409] [<c09efac0>] (dump_stack) from [<c012591c>]
(__warn+0xf8/0x124)
[ 595.529339] [<c012591c>] (__warn) from [<c0125a08>]
(warn_slowpath_fmt+0x38/0x48)
[ 595.536795] [<c0125a08>] (warn_slowpath_fmt) from [<c07160ec>]
(devfreq_monitor_start+0xbc/0xdc)
[ 595.545550] [<c07160ec>] (devfreq_monitor_start) from [<c0716434>]
(devfreq_interval_update+0x78/0x90)
[ 595.554825] [<c0716434>] (devfreq_interval_update) from [<c0717970>]
(devfreq_simple_ondemand_handler+0x14/0)
[ 595.564966] [<c0717970>] (devfreq_simple_ondemand_handler) from
[<c071575c>] (polling_interval_store+0x50/0x)
[ 595.575019] [<c071575c>] (polling_interval_store) from [<c0544f54>]
(dev_attr_store+0x18/0x24)
[ 595.583592] [<c0544f54>] (dev_attr_store) from [<c02fba24>]
(sysfs_kf_write+0x4c/0x50)
[ 595.591474] [<c02fba24>] (sysfs_kf_write) from [<c02fab70>]
(kernfs_fop_write+0xfc/0x1e4)
[ 595.599621] [<c02fab70>] (kernfs_fop_write) from [<c027f3ec>]
(__vfs_write+0x2c/0x160)
[ 595.607506] [<c027f3ec>] (__vfs_write) from [<c027f6a8>]
(vfs_write+0xa4/0x16c)
[ 595.614783] [<c027f6a8>] (vfs_write) from [<c027f884>]
(ksys_write+0x40/0x8c)
[ 595.621887] [<c027f884>] (ksys_write) from [<c0101000>]
(ret_fast_syscall+0x0/0x28)
[ 595.629512] Exception stack(0xee8fbfa8 to 0xee8fbff0)
[ 595.634531] bfa0: 00000003 000d0408 00000001
000d0408 00000003 00000000
[ 595.642686] bfc0: 00000003 000d0408 b6e84d60 00000004 00000000
000c8164 000c5730 000c5760
[ 595.650830] bfe0: 00000000 be8908ac b6de801b b6e23b46
[ 595.655877] irq event stamp: 411551
[ 595.659309] hardirqs last enabled at (411551): [<c018c030>]
console_unlock+0x474/0x700
[ 595.667324] hardirqs last disabled at (411550): [<c018bc7c>]
console_unlock+0xc0/0x700
[ 595.675210] softirqs last enabled at (411540): [<c01024cc>]
__do_softirq+0x3a4/0x66c
[ 595.683009] softirqs last disabled at (411527): [<c012d6b8>]
irq_exit+0x140/0x168
[ 595.690430] ---[ end trace 92e0ab098cadffbf ]---
------->8------------------------------------------------------
Regards,
Lukasz
On 2/14/19 2:30 AM, Matthias Kaehlcke wrote:
> devfreq expects governors to call devfreq_monitor_start/stop()
> in response to DEVFREQ_GOV_START/STOP events. Since the devfreq
> core itself generates these events and invokes the governor's event
> handler the start/stop of the load monitor can be done in the common
> code.
>
> Call devfreq_monitor_start/stop() when the governor reports a
> successful handling of DEVFREQ_GOV_START/STOP and remove these
> calls from the simpleondemand and tegra governors. Make
> devfreq_monitor_start/stop() static since these functions are now
> only used by the devfreq core. For better integration with the
> callers the functions must now be called with devfreq->lock held.
>
> Also stop manipulating the monitor workqueue directly, use
> devfreq_monitor_start/stop() instead.
>
> Signed-off-by: Matthias Kaehlcke <mka@chromium.org>
> ---
> drivers/devfreq/devfreq.c | 45 +++++++++++------------
> drivers/devfreq/governor.h | 2 -
> drivers/devfreq/governor_simpleondemand.c | 8 ----
> drivers/devfreq/tegra-devfreq.c | 2 -
> 4 files changed, 22 insertions(+), 35 deletions(-)
>
> diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c
> index eb86461648d74..ac553c00f6790 100644
> --- a/drivers/devfreq/devfreq.c
> +++ b/drivers/devfreq/devfreq.c
> @@ -404,14 +404,14 @@ static void devfreq_monitor(struct work_struct *work)
> * devfreq_monitor_start() - Start load monitoring of devfreq instance
> * @devfreq: the devfreq instance.
> *
> - * Helper function for starting devfreq device load monitoing. By
> - * default delayed work based monitoring is supported. Function
> - * to be called from governor in response to DEVFREQ_GOV_START
> - * event when device is added to devfreq framework.
> + * Helper function for starting devfreq device load monitoring. By
> + * default delayed work based monitoring is supported. Must be called
> + * with devfreq->lock held.
> */
> -void devfreq_monitor_start(struct devfreq *devfreq)
> +static void devfreq_monitor_start(struct devfreq *devfreq)
> {
> - mutex_lock(&devfreq->lock);
> + WARN(!mutex_is_locked(&devfreq->lock),
> + "devfreq->lock must be held by the caller.\n");
>
> if (devfreq->profile->polling_ms) {
> INIT_DEFERRABLE_WORK(&devfreq->work, devfreq_monitor);
> @@ -420,28 +420,28 @@ void devfreq_monitor_start(struct devfreq *devfreq)
>
> devfreq->monitor_state = DEVFREQ_MONITOR_RUNNING;
> }
> -
> - mutex_unlock(&devfreq->lock);
> }
> -EXPORT_SYMBOL(devfreq_monitor_start);
>
> /**
> * devfreq_monitor_stop() - Stop load monitoring of a devfreq instance
> * @devfreq: the devfreq instance.
> *
> - * Helper function to stop devfreq device load monitoing. Function
> - * to be called from governor in response to DEVFREQ_GOV_STOP
> - * event when device is removed from devfreq framework.
> + * Helper function to stop devfreq device load monitoring. Must be
> + * called with devfreq->lock held.
> */
> -void devfreq_monitor_stop(struct devfreq *devfreq)
> +static void devfreq_monitor_stop(struct devfreq *devfreq)
> {
> + /* mutex must be held for symmetry with _start() */
> + WARN(!mutex_is_locked(&devfreq->lock),
> + "devfreq->lock must be held by the caller.\n");
> +
> + mutex_unlock(&devfreq->lock);
> cancel_delayed_work_sync(&devfreq->work);
>
> mutex_lock(&devfreq->lock);
> devfreq->monitor_state = DEVFREQ_MONITOR_STOPPED;
> mutex_unlock(&devfreq->lock);
> }
> -EXPORT_SYMBOL(devfreq_monitor_stop);
>
> /**
> * devfreq_monitor_suspend() - Suspend load monitoring of a devfreq instance
> @@ -518,27 +518,22 @@ void devfreq_interval_update(struct devfreq *devfreq, unsigned int *delay)
>
> /* if new delay is zero, stop polling */
> if (!new_delay) {
> + devfreq_monitor_stop(devfreq);
> mutex_unlock(&devfreq->lock);
> - cancel_delayed_work_sync(&devfreq->work);
> - devfreq->monitor_state == DEVFREQ_MONITOR_STOPPED;
> return;
> }
>
> /* if current delay is zero, start polling with new delay */
> if (!cur_delay) {
> - queue_delayed_work(devfreq_wq, &devfreq->work,
> - msecs_to_jiffies(devfreq->profile->polling_ms));
> + devfreq_monitor_start(devfreq);
> goto out;
> }
>
> /* if current delay is greater than new delay, restart polling */
> if (cur_delay > new_delay) {
> - mutex_unlock(&devfreq->lock);
> - cancel_delayed_work_sync(&devfreq->work);
> - mutex_lock(&devfreq->lock);
> + devfreq_monitor_stop(devfreq);
> if (devfreq->monitor_state != DEVFREQ_MONITOR_SUSPENDED)
> - queue_delayed_work(devfreq_wq, &devfreq->work,
> - msecs_to_jiffies(devfreq->profile->polling_ms));
> + devfreq_monitor_start(devfreq);
> }
> out:
> mutex_unlock(&devfreq->lock);
> @@ -601,6 +596,8 @@ static int devfreq_governor_start(struct devfreq *devfreq)
> return err;
> }
>
> + devfreq_monitor_start(devfreq);
> +
> return 0;
> }
>
> @@ -624,6 +621,8 @@ static int devfreq_governor_stop(struct devfreq *devfreq)
> return err;
> }
>
> + devfreq_monitor_stop(devfreq);
> +
> return 0;
> }
>
> diff --git a/drivers/devfreq/governor.h b/drivers/devfreq/governor.h
> index d136792c0cc91..47efe747b6f79 100644
> --- a/drivers/devfreq/governor.h
> +++ b/drivers/devfreq/governor.h
> @@ -57,8 +57,6 @@ struct devfreq_governor {
> unsigned int event, void *data);
> };
>
> -extern void devfreq_monitor_start(struct devfreq *devfreq);
> -extern void devfreq_monitor_stop(struct devfreq *devfreq);
> extern void devfreq_interval_update(struct devfreq *devfreq,
> unsigned int *delay);
>
> diff --git a/drivers/devfreq/governor_simpleondemand.c b/drivers/devfreq/governor_simpleondemand.c
> index 52eb0c734b312..e0f0944a9c7aa 100644
> --- a/drivers/devfreq/governor_simpleondemand.c
> +++ b/drivers/devfreq/governor_simpleondemand.c
> @@ -91,14 +91,6 @@ static int devfreq_simple_ondemand_handler(struct devfreq *devfreq,
> unsigned int event, void *data)
> {
> switch (event) {
> - case DEVFREQ_GOV_START:
> - devfreq_monitor_start(devfreq);
> - break;
> -
> - case DEVFREQ_GOV_STOP:
> - devfreq_monitor_stop(devfreq);
> - break;
> -
> case DEVFREQ_GOV_INTERVAL:
> devfreq_interval_update(devfreq, (unsigned int *)data);
> break;
> diff --git a/drivers/devfreq/tegra-devfreq.c b/drivers/devfreq/tegra-devfreq.c
> index 79efa1e51bd06..515fb852dbad6 100644
> --- a/drivers/devfreq/tegra-devfreq.c
> +++ b/drivers/devfreq/tegra-devfreq.c
> @@ -580,13 +580,11 @@ static int tegra_governor_event_handler(struct devfreq *devfreq,
>
> switch (event) {
> case DEVFREQ_GOV_START:
> - devfreq_monitor_start(devfreq);
> tegra_actmon_enable_interrupts(tegra);
> break;
>
> case DEVFREQ_GOV_STOP:
> tegra_actmon_disable_interrupts(tegra);
> - devfreq_monitor_stop(devfreq);
> break;
>
> case DEVFREQ_GOV_SUSPEND:
>
Hi Lukasz, On Thu, Feb 14, 2019 at 07:01:36PM +0100, Lukasz Luba wrote: > Hi Matthias, > > I have compiled and run your changes on Odroid xu3 and v5.0-rc6. > There are kernel warnings because of mutex not held in function > devfreq_monitor_[start|stop]() in use cases: > 1) a few times during registration of new devices devfreq_add_device() > 2) poking the device from sysfs Thanks testing! I messed up carrying over changes from my v4.19 device kernel to the tree used for upstreaming :( Do you still see warnings with the below patch? Thanks Matthias diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c index a42c37543c190..866fe711b43ca 100644 --- a/drivers/devfreq/devfreq.c +++ b/drivers/devfreq/devfreq.c @@ -440,7 +440,6 @@ static void devfreq_monitor_stop(struct devfreq *devfreq) mutex_lock(&devfreq->lock); devfreq->monitor_state = DEVFREQ_MONITOR_STOPPED; - mutex_unlock(&devfreq->lock); } /** @@ -596,7 +595,9 @@ static int governor_start(struct devfreq *devfreq) return err; } + mutex_lock(&devfreq->lock); devfreq_monitor_start(devfreq); + mutex_unlock(&devfreq->lock); return 0; } @@ -614,7 +615,9 @@ static int governor_stop(struct devfreq *devfreq) "mutex must *not* be held by the caller\n")) return -EINVAL; + mutex_lock(&devfreq->lock); devfreq_monitor_stop(devfreq); + mutex_unlock(&devfreq->lock); err = devfreq->governor->event_handler(devfreq, DEVFREQ_GOV_STOP, NULL); if (err) {
Hi Chanwoo, On Thu, Feb 14, 2019 at 11:17:36PM +0900, Chanwoo Choi wrote: > Hi Matthias, > > As I commented on the first patch, it is not possible to call some codes > according to the intention of each governor between 'devfreq_moniotr_*()' > and some codes which are executed before or after 'devfreq_moniotr_*()' > > For example, if some governor requires the following sequence, > after this patch, it is not possible. > > case DEVFREQ_GOV_xxx: > /* execute some code before devfreq_monitor_xxx() */ > devfreq_monitor_xxx() > /* execute some code after devfreq_monitor_xxx() */ As for the suspend/resume case I agree that the patch introduces this limitation, but I'm not convinced that this is an actual problem. For governor_start(): why can't the governor execute the code before polling started, does it make any difference to the governor that a work is scheduled? For governor_stop(): why would the governor require polling to be active during stop? If it needs update_devfreq() to run (called by devfreq_monitor()) it can call it directly, instead of waiting for the monitor to run at some later time. Cheers Matthias
Hi Matthias, On 19. 2. 15. 오전 4:28, Matthias Kaehlcke wrote: > Hi Chanwoo, > > On Thu, Feb 14, 2019 at 11:17:36PM +0900, Chanwoo Choi wrote: >> Hi Matthias, >> >> As I commented on the first patch, it is not possible to call some codes >> according to the intention of each governor between 'devfreq_moniotr_*()' >> and some codes which are executed before or after 'devfreq_moniotr_*()' >> >> For example, if some governor requires the following sequence, >> after this patch, it is not possible. >> >> case DEVFREQ_GOV_xxx: >> /* execute some code before devfreq_monitor_xxx() */ >> devfreq_monitor_xxx() >> /* execute some code after devfreq_monitor_xxx() */ > > As for the suspend/resume case I agree that the patch introduces this > limitation, but I'm not convinced that this is an actual problem. > > For governor_start(): why can't the governor execute the code > before polling started, does it make any difference to the governor > that a work is scheduled? The some governors might want to do something before starting the work and do something after work started. I think that the existing style is more flexible. And, It has one more problem when changing the governor on the fly from simple_ondemand to other governors like performance, powersave and so on. Even if other governors don't need to monitor the utilization, the work timer will be executed continually because the devfreq device has the polling_ms value. It is not necessary of the other governors such as performance, powersave. It means that only specific governor like simple_ondemand have to call the devfreq_monitor_start() by itself instead of calling it on devfreq core. > > For governor_stop(): why would the governor require polling to be > active during stop? If it needs update_devfreq() to run (called by > devfreq_monitor()) it can call it directly, instead of waiting for the > monitor to run at some later time. As I knew, if the current governors are performance/powersave/ userspace, the monitoring is already stopped and not used. Because they don't need to handle the codes related to the work like queue_delayed_work(), cancel_delayed_work_sync(). And, In case of the existing style for calling devfreq_monitor_*(), other governors like performance/powersave/userspace/passice don't need to call the devfreq_monitor_stop() because they didn't use the work timer. > > Cheers > > Matthias > > > > >
Hi Chanwoo, On Fri, Feb 15, 2019 at 08:42:30AM +0900, Chanwoo Choi wrote: > Hi Matthias, > > On 19. 2. 15. 오전 4:28, Matthias Kaehlcke wrote: > > Hi Chanwoo, > > > > On Thu, Feb 14, 2019 at 11:17:36PM +0900, Chanwoo Choi wrote: > >> Hi Matthias, > >> > >> As I commented on the first patch, it is not possible to call some codes > >> according to the intention of each governor between 'devfreq_moniotr_*()' > >> and some codes which are executed before or after 'devfreq_moniotr_*()' > >> > >> For example, if some governor requires the following sequence, > >> after this patch, it is not possible. > >> > >> case DEVFREQ_GOV_xxx: > >> /* execute some code before devfreq_monitor_xxx() */ > >> devfreq_monitor_xxx() > >> /* execute some code after devfreq_monitor_xxx() */ > > > > As for the suspend/resume case I agree that the patch introduces this > > limitation, but I'm not convinced that this is an actual problem. > > > > For governor_start(): why can't the governor execute the code > > before polling started, does it make any difference to the governor > > that a work is scheduled? > > The some governors might want to do something before starting > the work and do something after work started. I think that > the existing style is more flexible. Could you provide a practical example that answers my question above: "why can't the governor execute the code before polling started, does it make any difference to the governor that a work is scheduled?" > And, > It has one more problem when changing the governor on the fly > from simple_ondemand to other governors like performance, > powersave and so on. > > Even if other governors don't need to monitor the utilization, > the work timer will be executed continually because the devfreq > device has the polling_ms value. It is not necessary > of the other governors such as performance, powersave. > > It means that only specific governor like simple_ondemand > have to call the devfreq_monitor_start() by itself > instead of calling it on devfreq core. yes, I noticed that too, it can be easily fixed with a flag in the governor. > > For governor_stop(): why would the governor require polling to be > > active during stop? If it needs update_devfreq() to run (called by > > devfreq_monitor()) it can call it directly, instead of waiting for the > > monitor to run at some later time. > > As I knew, if the current governors are performance/powersave/ > userspace, the monitoring is already stopped and not used. > Because they don't need to handle the codes related to the work > like queue_delayed_work(), cancel_delayed_work_sync(). > > And, > In case of the existing style for calling devfreq_monitor_*(), > other governors like performance/powersave/userspace/passice > don't need to call the devfreq_monitor_stop() because they > didn't use the work timer. As per above, the governor could have a flag indicating that it doesn't need load monitoring. I think it should be avoided to expect the governors to do the right thing if certain actions are mandatory and common for all governors (unless the feature in question is not used). It should be handled in the core code, unless there are good reasons not to do this. With thispatch set the amount of code remains essentially the same, and no new code needs to be added for governors that don't do anything special in start/stop/suspend/resume. TBH I think probably the entire event handler multiplexing should go away and be changed to a struct devfreq_ops with optional callbacks for start, stop, suspend, resume and update interval. As far as I can tell the multiplexing isn't needed and separate ops would result in cleaner code, in both the devfreq core and the governors. Cheers Matthias
Hi Matthias, On 19. 2. 15. 오전 9:19, Matthias Kaehlcke wrote: > Hi Chanwoo, > > On Fri, Feb 15, 2019 at 08:42:30AM +0900, Chanwoo Choi wrote: >> Hi Matthias, >> >> On 19. 2. 15. 오전 4:28, Matthias Kaehlcke wrote: >>> Hi Chanwoo, >>> >>> On Thu, Feb 14, 2019 at 11:17:36PM +0900, Chanwoo Choi wrote: >>>> Hi Matthias, >>>> >>>> As I commented on the first patch, it is not possible to call some codes >>>> according to the intention of each governor between 'devfreq_moniotr_*()' >>>> and some codes which are executed before or after 'devfreq_moniotr_*()' >>>> >>>> For example, if some governor requires the following sequence, >>>> after this patch, it is not possible. >>>> >>>> case DEVFREQ_GOV_xxx: >>>> /* execute some code before devfreq_monitor_xxx() */ >>>> devfreq_monitor_xxx() >>>> /* execute some code after devfreq_monitor_xxx() */ >>> >>> As for the suspend/resume case I agree that the patch introduces this >>> limitation, but I'm not convinced that this is an actual problem. >>> >>> For governor_start(): why can't the governor execute the code >>> before polling started, does it make any difference to the governor >>> that a work is scheduled? >> >> The some governors might want to do something before starting >> the work and do something after work started. I think that >> the existing style is more flexible. > > Could you provide a practical example that answers my question above: > > "why can't the governor execute the code before polling started, does > it make any difference to the governor that a work is scheduled?" Actually, as for now, I don't know the correct practice and now. I want to say that the existing style is more flexible for the new governor in the future. If there are no any special benefits, I think we don't need to harm the flexibility. > >> And, >> It has one more problem when changing the governor on the fly >> from simple_ondemand to other governors like performance, >> powersave and so on. >> >> Even if other governors don't need to monitor the utilization, >> the work timer will be executed continually because the devfreq >> device has the polling_ms value. It is not necessary >> of the other governors such as performance, powersave. >> >> It means that only specific governor like simple_ondemand >> have to call the devfreq_monitor_start() by itself >> instead of calling it on devfreq core. > > yes, I noticed that too, it can be easily fixed with a flag in the > governor. Right, If we add some codes like flag, it is easy. Actually, it is just difference how to support them. I think that the existing style has not any problem and is not complicated to develop the new governor. If there are no some benefits, IMO, we better to keep the flexibility. It can give the flexibility to the developer of governor. > >>> For governor_stop(): why would the governor require polling to be >>> active during stop? If it needs update_devfreq() to run (called by >>> devfreq_monitor()) it can call it directly, instead of waiting for the >>> monitor to run at some later time. >> >> As I knew, if the current governors are performance/powersave/ >> userspace, the monitoring is already stopped and not used. >> Because they don't need to handle the codes related to the work >> like queue_delayed_work(), cancel_delayed_work_sync(). >> >> And, >> In case of the existing style for calling devfreq_monitor_*(), >> other governors like performance/powersave/userspace/passice >> don't need to call the devfreq_monitor_stop() because they >> didn't use the work timer. > > As per above, the governor could have a flag indicating that it > doesn't need load monitoring. About this, I commented above. Pleas check it. > > I think it should be avoided to expect the governors to do the right > thing if certain actions are mandatory and common for all governors > (unless the feature in question is not used). It should be handled in > the core code, unless there are good reasons not to do this. I agree to distinguish for common code and specific code depends on governor. But, it is harm for the flexibility. We know that there are no problem after applying this patchset. But, as I commented, I'm not sure that it is meaningful to harm the flexibility. > > With thispatch set the amount of code remains essentially the same, > and no new code needs to be added for governors that don't do anything > special in start/stop/suspend/resume. > > TBH I think probably the entire event handler multiplexing should go > away and be changed to a struct devfreq_ops with optional callbacks > for start, stop, suspend, resume and update interval. As far as I can > tell the multiplexing isn't needed and separate ops would result in > cleaner code, in both the devfreq core and the governors. > > Cheers > > Matthias > >
Hi Matthias, On 2/14/19 8:07 PM, Matthias Kaehlcke wrote: > Hi Lukasz, > > On Thu, Feb 14, 2019 at 07:01:36PM +0100, Lukasz Luba wrote: >> Hi Matthias, >> >> I have compiled and run your changes on Odroid xu3 and v5.0-rc6. >> There are kernel warnings because of mutex not held in function >> devfreq_monitor_[start|stop]() in use cases: >> 1) a few times during registration of new devices devfreq_add_device() >> 2) poking the device from sysfs > > Thanks testing! > > I messed up carrying over changes from my v4.19 device kernel > to the tree used for upstreaming :( > > Do you still see warnings with the below patch? No, the warnings are no longer present in the log. The patch solves it. Regards, Lukasz > > Thanks > > Matthias > > diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c > index a42c37543c190..866fe711b43ca 100644 > --- a/drivers/devfreq/devfreq.c > +++ b/drivers/devfreq/devfreq.c > @@ -440,7 +440,6 @@ static void devfreq_monitor_stop(struct devfreq *devfreq) > > mutex_lock(&devfreq->lock); > devfreq->monitor_state = DEVFREQ_MONITOR_STOPPED; > - mutex_unlock(&devfreq->lock); > } > > /** > @@ -596,7 +595,9 @@ static int governor_start(struct devfreq *devfreq) > return err; > } > > + mutex_lock(&devfreq->lock); > devfreq_monitor_start(devfreq); > + mutex_unlock(&devfreq->lock); > > return 0; > } > @@ -614,7 +615,9 @@ static int governor_stop(struct devfreq *devfreq) > "mutex must *not* be held by the caller\n")) > return -EINVAL; > > + mutex_lock(&devfreq->lock); > devfreq_monitor_stop(devfreq); > + mutex_unlock(&devfreq->lock); > > err = devfreq->governor->event_handler(devfreq, DEVFREQ_GOV_STOP, NULL); > if (err) { > >
Hi Chanwoo, On Fri, Feb 15, 2019 at 09:33:05AM +0900, Chanwoo Choi wrote: > Hi Matthias, > > On 19. 2. 15. 오전 9:19, Matthias Kaehlcke wrote: > > Hi Chanwoo, > > > > On Fri, Feb 15, 2019 at 08:42:30AM +0900, Chanwoo Choi wrote: > >> Hi Matthias, > >> > >> On 19. 2. 15. 오전 4:28, Matthias Kaehlcke wrote: > >>> Hi Chanwoo, > >>> > >>> On Thu, Feb 14, 2019 at 11:17:36PM +0900, Chanwoo Choi wrote: > >>>> Hi Matthias, > >>>> > >>>> As I commented on the first patch, it is not possible to call some codes > >>>> according to the intention of each governor between 'devfreq_moniotr_*()' > >>>> and some codes which are executed before or after 'devfreq_moniotr_*()' > >>>> > >>>> For example, if some governor requires the following sequence, > >>>> after this patch, it is not possible. > >>>> > >>>> case DEVFREQ_GOV_xxx: > >>>> /* execute some code before devfreq_monitor_xxx() */ > >>>> devfreq_monitor_xxx() > >>>> /* execute some code after devfreq_monitor_xxx() */ > >>> > >>> As for the suspend/resume case I agree that the patch introduces this > >>> limitation, but I'm not convinced that this is an actual problem. > >>> > >>> For governor_start(): why can't the governor execute the code > >>> before polling started, does it make any difference to the governor > >>> that a work is scheduled? > >> > >> The some governors might want to do something before starting > >> the work and do something after work started. I think that > >> the existing style is more flexible. > > > > Could you provide a practical example that answers my question above: > > > > "why can't the governor execute the code before polling started, does > > it make any difference to the governor that a work is scheduled?" > > Actually, as for now, I don't know the correct practice and now. > I want to say that the existing style is more flexible for the > new governor in the future. If you try submitting 'flexible' code in other parts of the kernel without users of this flexibility and no solid arguments why this flexibility is needed it would most likely be rejected. Unnecessary flexibility can be a burden, rather than an advantage. > If there are no any special benefits I think we don't need to harm > the flexibility. There are multiple benefits, the following shortcomings of the current approach are eliminated: - it is error prone and allows governors to do the wrong thing, e.g. - start monitoring before the governor is fully ready - keep monitoring when the governor is partially 'stopped' - omit mandatory calls - delegates tasks to the governors which are responsibility of the core - code is harder to read - switch from common code to governor code and back - unecessary boilerplate code in governors - option to replace ->event_handler() with ->start(), ->stop(), etc, which is cleaner I'm easily convinced if the flexibility is really needed. I'm not even expecting a real world example, just be creative and make something up that sounds reasonable. start/resume() { // prepare governor // start polling => what needs to be done here that couldn't have been done => before polling was started? } stop/suspend() { => what needs to be done here that couldn't be done after polling => is stopped? // stop polling // 'unprepare' governor } > >> And, > >> It has one more problem when changing the governor on the fly > >> from simple_ondemand to other governors like performance, > >> powersave and so on. > >> > >> Even if other governors don't need to monitor the utilization, > >> the work timer will be executed continually because the devfreq > >> device has the polling_ms value. It is not necessary > >> of the other governors such as performance, powersave. > >> > >> It means that only specific governor like simple_ondemand > >> have to call the devfreq_monitor_start() by itself > >> instead of calling it on devfreq core. > > > > yes, I noticed that too, it can be easily fixed with a flag in the > > governor. > > Right, If we add some codes like flag, it is easy. > Actually, it is just difference how to support them. > > I think that the existing style has not any problem and is not > complicated to develop the new governor. If there are no > some benefits, IMO, we better to keep the flexibility. > It can give the flexibility to the developer of governor. I listed several benefits, please comment on the specific items if you disagree, instead of just saying there are no benefits. OTOH so far we haven't seen an even hypothetical example that shows that the flexibility *could* be needed. And even if such a rare case that we currently can't imagine came up, it could be easily addressed with notifiers, a standard mechanism in the kernel to inform drivers about events they are interested in. Cheers Matthias
Hi Matthias, On 19. 2. 16. 오전 7:56, Matthias Kaehlcke wrote: > Hi Chanwoo, > > On Fri, Feb 15, 2019 at 09:33:05AM +0900, Chanwoo Choi wrote: >> Hi Matthias, >> >> On 19. 2. 15. 오전 9:19, Matthias Kaehlcke wrote: >>> Hi Chanwoo, >>> >>> On Fri, Feb 15, 2019 at 08:42:30AM +0900, Chanwoo Choi wrote: >>>> Hi Matthias, >>>> >>>> On 19. 2. 15. 오전 4:28, Matthias Kaehlcke wrote: >>>>> Hi Chanwoo, >>>>> >>>>> On Thu, Feb 14, 2019 at 11:17:36PM +0900, Chanwoo Choi wrote: >>>>>> Hi Matthias, >>>>>> >>>>>> As I commented on the first patch, it is not possible to call some codes >>>>>> according to the intention of each governor between 'devfreq_moniotr_*()' >>>>>> and some codes which are executed before or after 'devfreq_moniotr_*()' >>>>>> >>>>>> For example, if some governor requires the following sequence, >>>>>> after this patch, it is not possible. >>>>>> >>>>>> case DEVFREQ_GOV_xxx: >>>>>> /* execute some code before devfreq_monitor_xxx() */ >>>>>> devfreq_monitor_xxx() >>>>>> /* execute some code after devfreq_monitor_xxx() */ >>>>> >>>>> As for the suspend/resume case I agree that the patch introduces this >>>>> limitation, but I'm not convinced that this is an actual problem. >>>>> >>>>> For governor_start(): why can't the governor execute the code >>>>> before polling started, does it make any difference to the governor >>>>> that a work is scheduled? >>>> >>>> The some governors might want to do something before starting >>>> the work and do something after work started. I think that >>>> the existing style is more flexible. >>> >>> Could you provide a practical example that answers my question above: >>> >>> "why can't the governor execute the code before polling started, does >>> it make any difference to the governor that a work is scheduled?" >> >> Actually, as for now, I don't know the correct practice and now. >> I want to say that the existing style is more flexible for the >> new governor in the future. > > If you try submitting 'flexible' code in other parts of the kernel > without users of this flexibility and no solid arguments why this > flexibility is needed it would most likely be rejected. > > Unnecessary flexibility can be a burden, rather than an advantage. > >> If there are no any special benefits I think we don't need to harm >> the flexibility. > > There are multiple benefits, the following shortcomings of the current > approach are eliminated: > > - it is error prone and allows governors to do the wrong thing, e.g. > - start monitoring before the governor is fully ready > - keep monitoring when the governor is partially 'stopped' > - omit mandatory calls > - delegates tasks to the governors which are responsibility of the > core > - code is harder to read > - switch from common code to governor code and back > - unecessary boilerplate code in governors > - option to replace ->event_handler() with ->start(), ->stop(), etc, > which is cleaner > > I'm easily convinced if the flexibility is really needed. I'm not even > expecting a real world example, just be creative and make something > up that sounds reasonable. > > start/resume() > { > // prepare governor > > // start polling > > => what needs to be done here that couldn't have been done > => before polling was started? > } > > stop/suspend() > { > => what needs to be done here that couldn't be done after polling > => is stopped? > > // stop polling > > // 'unprepare' governor > } > >>>> And, >>>> It has one more problem when changing the governor on the fly >>>> from simple_ondemand to other governors like performance, >>>> powersave and so on. >>>> >>>> Even if other governors don't need to monitor the utilization, >>>> the work timer will be executed continually because the devfreq >>>> device has the polling_ms value. It is not necessary >>>> of the other governors such as performance, powersave. >>>> >>>> It means that only specific governor like simple_ondemand >>>> have to call the devfreq_monitor_start() by itself >>>> instead of calling it on devfreq core. >>> >>> yes, I noticed that too, it can be easily fixed with a flag in the >>> governor. >> >> Right, If we add some codes like flag, it is easy. >> Actually, it is just difference how to support them. >> >> I think that the existing style has not any problem and is not >> complicated to develop the new governor. If there are no >> some benefits, IMO, we better to keep the flexibility. >> It can give the flexibility to the developer of governor. > > I listed several benefits, please comment on the specific items if you > disagree, instead of just saying there are no benefits. > > OTOH so far we haven't seen an even hypothetical example that shows > that the flexibility *could* be needed. > > And even if such a rare case that we currently can't imagine came up, > it could be easily addressed with notifiers, a standard mechanism in > the kernel to inform drivers about events they are interested in. As I said on previous mail, I didn't know the correct practice. When we develop the something, I think that there are no need to consider too much flexibility. But, if already developed code had the more flexibility than the new refactoring code, I think that keep the existing code if there are no any special benefits. Also as I said, it is just refactoring without any changes or improvement of feature. If the existing code have some bug, I will agree them without any question. Actually, the monitoring feature are used on only simple_ondemand governor and tegra-devfreq.c driver. The devfreq core don't need to consider the start/stop monitoring because each governor/driver can handle them enough with the existing method. IMHO, I think that it is just different way how to implement them because the existing code doesn't have the bug. I have no any strong objection. But, in my case, it is not easy to like this. You better to wait the comments from devfreq maintainer. > > Cheers > > Matthias > >
diff --git a/drivers/devfreq/devfreq.c b/drivers/devfreq/devfreq.c index eb86461648d74..ac553c00f6790 100644 --- a/drivers/devfreq/devfreq.c +++ b/drivers/devfreq/devfreq.c @@ -404,14 +404,14 @@ static void devfreq_monitor(struct work_struct *work) * devfreq_monitor_start() - Start load monitoring of devfreq instance * @devfreq: the devfreq instance. * - * Helper function for starting devfreq device load monitoing. By - * default delayed work based monitoring is supported. Function - * to be called from governor in response to DEVFREQ_GOV_START - * event when device is added to devfreq framework. + * Helper function for starting devfreq device load monitoring. By + * default delayed work based monitoring is supported. Must be called + * with devfreq->lock held. */ -void devfreq_monitor_start(struct devfreq *devfreq) +static void devfreq_monitor_start(struct devfreq *devfreq) { - mutex_lock(&devfreq->lock); + WARN(!mutex_is_locked(&devfreq->lock), + "devfreq->lock must be held by the caller.\n"); if (devfreq->profile->polling_ms) { INIT_DEFERRABLE_WORK(&devfreq->work, devfreq_monitor); @@ -420,28 +420,28 @@ void devfreq_monitor_start(struct devfreq *devfreq) devfreq->monitor_state = DEVFREQ_MONITOR_RUNNING; } - - mutex_unlock(&devfreq->lock); } -EXPORT_SYMBOL(devfreq_monitor_start); /** * devfreq_monitor_stop() - Stop load monitoring of a devfreq instance * @devfreq: the devfreq instance. * - * Helper function to stop devfreq device load monitoing. Function - * to be called from governor in response to DEVFREQ_GOV_STOP - * event when device is removed from devfreq framework. + * Helper function to stop devfreq device load monitoring. Must be + * called with devfreq->lock held. */ -void devfreq_monitor_stop(struct devfreq *devfreq) +static void devfreq_monitor_stop(struct devfreq *devfreq) { + /* mutex must be held for symmetry with _start() */ + WARN(!mutex_is_locked(&devfreq->lock), + "devfreq->lock must be held by the caller.\n"); + + mutex_unlock(&devfreq->lock); cancel_delayed_work_sync(&devfreq->work); mutex_lock(&devfreq->lock); devfreq->monitor_state = DEVFREQ_MONITOR_STOPPED; mutex_unlock(&devfreq->lock); } -EXPORT_SYMBOL(devfreq_monitor_stop); /** * devfreq_monitor_suspend() - Suspend load monitoring of a devfreq instance @@ -518,27 +518,22 @@ void devfreq_interval_update(struct devfreq *devfreq, unsigned int *delay) /* if new delay is zero, stop polling */ if (!new_delay) { + devfreq_monitor_stop(devfreq); mutex_unlock(&devfreq->lock); - cancel_delayed_work_sync(&devfreq->work); - devfreq->monitor_state == DEVFREQ_MONITOR_STOPPED; return; } /* if current delay is zero, start polling with new delay */ if (!cur_delay) { - queue_delayed_work(devfreq_wq, &devfreq->work, - msecs_to_jiffies(devfreq->profile->polling_ms)); + devfreq_monitor_start(devfreq); goto out; } /* if current delay is greater than new delay, restart polling */ if (cur_delay > new_delay) { - mutex_unlock(&devfreq->lock); - cancel_delayed_work_sync(&devfreq->work); - mutex_lock(&devfreq->lock); + devfreq_monitor_stop(devfreq); if (devfreq->monitor_state != DEVFREQ_MONITOR_SUSPENDED) - queue_delayed_work(devfreq_wq, &devfreq->work, - msecs_to_jiffies(devfreq->profile->polling_ms)); + devfreq_monitor_start(devfreq); } out: mutex_unlock(&devfreq->lock); @@ -601,6 +596,8 @@ static int devfreq_governor_start(struct devfreq *devfreq) return err; } + devfreq_monitor_start(devfreq); + return 0; } @@ -624,6 +621,8 @@ static int devfreq_governor_stop(struct devfreq *devfreq) return err; } + devfreq_monitor_stop(devfreq); + return 0; } diff --git a/drivers/devfreq/governor.h b/drivers/devfreq/governor.h index d136792c0cc91..47efe747b6f79 100644 --- a/drivers/devfreq/governor.h +++ b/drivers/devfreq/governor.h @@ -57,8 +57,6 @@ struct devfreq_governor { unsigned int event, void *data); }; -extern void devfreq_monitor_start(struct devfreq *devfreq); -extern void devfreq_monitor_stop(struct devfreq *devfreq); extern void devfreq_interval_update(struct devfreq *devfreq, unsigned int *delay); diff --git a/drivers/devfreq/governor_simpleondemand.c b/drivers/devfreq/governor_simpleondemand.c index 52eb0c734b312..e0f0944a9c7aa 100644 --- a/drivers/devfreq/governor_simpleondemand.c +++ b/drivers/devfreq/governor_simpleondemand.c @@ -91,14 +91,6 @@ static int devfreq_simple_ondemand_handler(struct devfreq *devfreq, unsigned int event, void *data) { switch (event) { - case DEVFREQ_GOV_START: - devfreq_monitor_start(devfreq); - break; - - case DEVFREQ_GOV_STOP: - devfreq_monitor_stop(devfreq); - break; - case DEVFREQ_GOV_INTERVAL: devfreq_interval_update(devfreq, (unsigned int *)data); break; diff --git a/drivers/devfreq/tegra-devfreq.c b/drivers/devfreq/tegra-devfreq.c index 79efa1e51bd06..515fb852dbad6 100644 --- a/drivers/devfreq/tegra-devfreq.c +++ b/drivers/devfreq/tegra-devfreq.c @@ -580,13 +580,11 @@ static int tegra_governor_event_handler(struct devfreq *devfreq, switch (event) { case DEVFREQ_GOV_START: - devfreq_monitor_start(devfreq); tegra_actmon_enable_interrupts(tegra); break; case DEVFREQ_GOV_STOP: tegra_actmon_disable_interrupts(tegra); - devfreq_monitor_stop(devfreq); break; case DEVFREQ_GOV_SUSPEND:
devfreq expects governors to call devfreq_monitor_start/stop() in response to DEVFREQ_GOV_START/STOP events. Since the devfreq core itself generates these events and invokes the governor's event handler the start/stop of the load monitor can be done in the common code. Call devfreq_monitor_start/stop() when the governor reports a successful handling of DEVFREQ_GOV_START/STOP and remove these calls from the simpleondemand and tegra governors. Make devfreq_monitor_start/stop() static since these functions are now only used by the devfreq core. For better integration with the callers the functions must now be called with devfreq->lock held. Also stop manipulating the monitor workqueue directly, use devfreq_monitor_start/stop() instead. Signed-off-by: Matthias Kaehlcke <mka@chromium.org> --- drivers/devfreq/devfreq.c | 45 +++++++++++------------ drivers/devfreq/governor.h | 2 - drivers/devfreq/governor_simpleondemand.c | 8 ---- drivers/devfreq/tegra-devfreq.c | 2 - 4 files changed, 22 insertions(+), 35 deletions(-)