diff mbox

[v3,12/12] drm/i915: Listen for PMIC bus access notifications

Message ID 20170210102802.20898-13-hdegoede@redhat.com (mailing list archive)
State New, archived
Headers show

Commit Message

Hans de Goede Feb. 10, 2017, 10:28 a.m. UTC
Listen for PMIC bus access notifications and get FORCEWAKE_ALL while
the bus is accessed to avoid needing to do any forcewakes, which need
PMIC bus access, while the PMIC bus is busy:

This fixes errors like these showing up in dmesg, usually followed
by a gfx or system freeze:

[drm:fw_domains_get [i915]] *ERROR* render: timed out waiting for forcewake ack request.
[drm:fw_domains_get [i915]] *MEDIA* render: timed out waiting for forcewake ack request.
i2c_designware 808622C1:06: punit semaphore timed out, resetting
i2c_designware 808622C1:06: PUNIT SEM: 2
i2c_designware 808622C1:06: couldn't acquire bus ownership

Downside of this approach is that it causes wakeups whenever the PMIC
bus is accessed. Unfortunately we cannot simply wait for the PMIC bus
to go idle when we hit a race, as forcewakes may be done from interrupt
handlers where we cannot sleep to wait for the i2c PMIC bus access to
finish.

Note that the notifications and thus the wakeups will only happen on
baytrail / cherrytrail devices using PMICs with a shared i2c bus for
P-Unit and host PMIC access (i2c busses with a _SEM method in their
APCI node), e.g. an axp288 PMIC.

I plan to write some patches for drivers accessing the PMIC bus to
limit their bus accesses to a bare minimum (e.g. cache registers, do not
update battery level more often then 4 times a minute), to limit the
amount of wakeups.

BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=155241
Signed-off-by: Hans de Goede <hdegoede@redhat.com>
Tested-by: tagorereddy <tagore.chandan@gmail.com>
---
Changes in v2:
-Spelling: P-Unit, PMIC
Changes in v3:
-Improve comment explaining why we call intel_uncore_forcewake_get(ALL) on
 MBI_PMIC_BUS_ACCESS_BEGIN notification
---
 drivers/gpu/drm/i915/Kconfig        |  1 +
 drivers/gpu/drm/i915/i915_drv.h     |  1 +
 drivers/gpu/drm/i915/intel_uncore.c | 39 +++++++++++++++++++++++++++++++++++++
 3 files changed, 41 insertions(+)

Comments

Ville Syrjälä Feb. 16, 2017, 6:44 p.m. UTC | #1
On Fri, Feb 10, 2017 at 11:28:02AM +0100, Hans de Goede wrote:
> Listen for PMIC bus access notifications and get FORCEWAKE_ALL while
> the bus is accessed to avoid needing to do any forcewakes, which need
> PMIC bus access, while the PMIC bus is busy:
> 
> This fixes errors like these showing up in dmesg, usually followed
> by a gfx or system freeze:
> 
> [drm:fw_domains_get [i915]] *ERROR* render: timed out waiting for forcewake ack request.
> [drm:fw_domains_get [i915]] *MEDIA* render: timed out waiting for forcewake ack request.
> i2c_designware 808622C1:06: punit semaphore timed out, resetting
> i2c_designware 808622C1:06: PUNIT SEM: 2
> i2c_designware 808622C1:06: couldn't acquire bus ownership
> 
> Downside of this approach is that it causes wakeups whenever the PMIC
> bus is accessed. Unfortunately we cannot simply wait for the PMIC bus
> to go idle when we hit a race, as forcewakes may be done from interrupt
> handlers where we cannot sleep to wait for the i2c PMIC bus access to
> finish.
> 
> Note that the notifications and thus the wakeups will only happen on
> baytrail / cherrytrail devices using PMICs with a shared i2c bus for
> P-Unit and host PMIC access (i2c busses with a _SEM method in their
> APCI node), e.g. an axp288 PMIC.
> 
> I plan to write some patches for drivers accessing the PMIC bus to
> limit their bus accesses to a bare minimum (e.g. cache registers, do not
> update battery level more often then 4 times a minute), to limit the
> amount of wakeups.
> 
> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=155241
> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
> Tested-by: tagorereddy <tagore.chandan@gmail.com>

I gave the previous versions a quick whirl on a few machines here, but
them not being CR versions I guess this stuff doesn't kick in at all.
And I don't see any _SEM stuff in the DSDT/SSDT, so I guess that
confirms it. Which is fine since I've not seem any stability issues
on those machines. So at least nothing seemed to break :)

Anyways the changes look all right to me, so for both i915 patches
Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

> ---
> Changes in v2:
> -Spelling: P-Unit, PMIC
> Changes in v3:
> -Improve comment explaining why we call intel_uncore_forcewake_get(ALL) on
>  MBI_PMIC_BUS_ACCESS_BEGIN notification
> ---
>  drivers/gpu/drm/i915/Kconfig        |  1 +
>  drivers/gpu/drm/i915/i915_drv.h     |  1 +
>  drivers/gpu/drm/i915/intel_uncore.c | 39 +++++++++++++++++++++++++++++++++++++
>  3 files changed, 41 insertions(+)
> 
> diff --git a/drivers/gpu/drm/i915/Kconfig b/drivers/gpu/drm/i915/Kconfig
> index 1ae0bb9..a5cd5da 100644
> --- a/drivers/gpu/drm/i915/Kconfig
> +++ b/drivers/gpu/drm/i915/Kconfig
> @@ -20,6 +20,7 @@ config DRM_I915
>  	select ACPI_VIDEO if ACPI
>  	select ACPI_BUTTON if ACPI
>  	select SYNC_FILE
> +	select IOSF_MBI
>  	help
>  	  Choose this option if you have a system that has "Intel Graphics
>  	  Media Accelerator" or "HD Graphics" integrated graphics,
> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
> index 998239f..4e405d0 100644
> --- a/drivers/gpu/drm/i915/i915_drv.h
> +++ b/drivers/gpu/drm/i915/i915_drv.h
> @@ -725,6 +725,7 @@ struct intel_uncore {
>  	const struct intel_forcewake_range *fw_domains_table;
>  	unsigned int fw_domains_table_entries;
>  
> +	struct notifier_block pmic_bus_access_nb;
>  	struct intel_uncore_funcs funcs;
>  
>  	unsigned fifo_count;
> diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c
> index b6d726b..1be3cdc 100644
> --- a/drivers/gpu/drm/i915/intel_uncore.c
> +++ b/drivers/gpu/drm/i915/intel_uncore.c
> @@ -25,6 +25,7 @@
>  #include "intel_drv.h"
>  #include "i915_vgpu.h"
>  
> +#include <asm/iosf_mbi.h>
>  #include <linux/pm_runtime.h>
>  
>  #define FORCEWAKE_ACK_TIMEOUT_MS 50
> @@ -436,12 +437,16 @@ static void __intel_uncore_early_sanitize(struct drm_i915_private *dev_priv,
>  
>  void intel_uncore_suspend(struct drm_i915_private *dev_priv)
>  {
> +	iosf_mbi_unregister_pmic_bus_access_notifier(
> +		&dev_priv->uncore.pmic_bus_access_nb);
>  	intel_uncore_forcewake_reset(dev_priv, false);
>  }
>  
>  void intel_uncore_resume_early(struct drm_i915_private *dev_priv)
>  {
>  	__intel_uncore_early_sanitize(dev_priv, true);
> +	iosf_mbi_register_pmic_bus_access_notifier(
> +		&dev_priv->uncore.pmic_bus_access_nb);
>  	i915_check_and_clear_faults(dev_priv);
>  }
>  
> @@ -1329,6 +1334,32 @@ static void intel_uncore_fw_domains_init(struct drm_i915_private *dev_priv)
>  	dev_priv->uncore.fw_domains_table_entries = ARRAY_SIZE((d)); \
>  }
>  
> +static int i915_pmic_bus_access_notifier(struct notifier_block *nb,
> +					 unsigned long action, void *data)
> +{
> +	struct drm_i915_private *dev_priv = container_of(nb,
> +			struct drm_i915_private, uncore.pmic_bus_access_nb);
> +
> +	switch (action) {
> +	case MBI_PMIC_BUS_ACCESS_BEGIN:
> +		/*
> +		 * forcewake all now to make sure that we don't need to do a
> +		 * forcewake later which on systems where this notifier gets
> +		 * called requires the punit to access to the shared pmic i2c
> +		 * bus, which will be busy after this notification, leading to:
> +		 * "render: timed out waiting for forcewake ack request."
> +		 * errors.
> +		 */
> +		intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
> +		break;
> +	case MBI_PMIC_BUS_ACCESS_END:
> +		intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
> +		break;
> +	}
> +
> +	return NOTIFY_OK;
> +}
> +
>  void intel_uncore_init(struct drm_i915_private *dev_priv)
>  {
>  	i915_check_vgpu(dev_priv);
> @@ -1338,6 +1369,8 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>  	__intel_uncore_early_sanitize(dev_priv, false);
>  
>  	dev_priv->uncore.unclaimed_mmio_check = 1;
> +	dev_priv->uncore.pmic_bus_access_nb.notifier_call =
> +		i915_pmic_bus_access_notifier;
>  
>  	switch (INTEL_INFO(dev_priv)->gen) {
>  	default:
> @@ -1392,6 +1425,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>  	if (INTEL_GEN(dev_priv) >= 8)
>  		intel_shadow_table_check();
>  
> +	iosf_mbi_register_pmic_bus_access_notifier(
> +		&dev_priv->uncore.pmic_bus_access_nb);
> +
>  	i915_check_and_clear_faults(dev_priv);
>  }
>  #undef ASSIGN_WRITE_MMIO_VFUNCS
> @@ -1399,6 +1435,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>  
>  void intel_uncore_fini(struct drm_i915_private *dev_priv)
>  {
> +	iosf_mbi_unregister_pmic_bus_access_notifier(
> +		&dev_priv->uncore.pmic_bus_access_nb);
> +
>  	/* Paranoia: make sure we have disabled everything before we exit. */
>  	intel_uncore_sanitize(dev_priv);
>  	intel_uncore_forcewake_reset(dev_priv, false);
> -- 
> 2.9.3
Jani Nikula Feb. 16, 2017, 7:02 p.m. UTC | #2
On Thu, 16 Feb 2017, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote:
> On Fri, Feb 10, 2017 at 11:28:02AM +0100, Hans de Goede wrote:
>> Listen for PMIC bus access notifications and get FORCEWAKE_ALL while
>> the bus is accessed to avoid needing to do any forcewakes, which need
>> PMIC bus access, while the PMIC bus is busy:
>> 
>> This fixes errors like these showing up in dmesg, usually followed
>> by a gfx or system freeze:
>> 
>> [drm:fw_domains_get [i915]] *ERROR* render: timed out waiting for forcewake ack request.
>> [drm:fw_domains_get [i915]] *MEDIA* render: timed out waiting for forcewake ack request.
>> i2c_designware 808622C1:06: punit semaphore timed out, resetting
>> i2c_designware 808622C1:06: PUNIT SEM: 2
>> i2c_designware 808622C1:06: couldn't acquire bus ownership
>> 
>> Downside of this approach is that it causes wakeups whenever the PMIC
>> bus is accessed. Unfortunately we cannot simply wait for the PMIC bus
>> to go idle when we hit a race, as forcewakes may be done from interrupt
>> handlers where we cannot sleep to wait for the i2c PMIC bus access to
>> finish.
>> 
>> Note that the notifications and thus the wakeups will only happen on
>> baytrail / cherrytrail devices using PMICs with a shared i2c bus for
>> P-Unit and host PMIC access (i2c busses with a _SEM method in their
>> APCI node), e.g. an axp288 PMIC.
>> 
>> I plan to write some patches for drivers accessing the PMIC bus to
>> limit their bus accesses to a bare minimum (e.g. cache registers, do not
>> update battery level more often then 4 times a minute), to limit the
>> amount of wakeups.
>> 
>> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=155241
>> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
>> Tested-by: tagorereddy <tagore.chandan@gmail.com>
>
> I gave the previous versions a quick whirl on a few machines here, but
> them not being CR versions I guess this stuff doesn't kick in at all.
> And I don't see any _SEM stuff in the DSDT/SSDT, so I guess that
> confirms it. Which is fine since I've not seem any stability issues
> on those machines. So at least nothing seemed to break :)
>
> Anyways the changes look all right to me, so for both i915 patches
> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>

Acked-by: Jani Nikula <jani.nikula@intel.com>

for merging the i915 patches through some other tree if that makes
managing the pile easier.


>
>> ---
>> Changes in v2:
>> -Spelling: P-Unit, PMIC
>> Changes in v3:
>> -Improve comment explaining why we call intel_uncore_forcewake_get(ALL) on
>>  MBI_PMIC_BUS_ACCESS_BEGIN notification
>> ---
>>  drivers/gpu/drm/i915/Kconfig        |  1 +
>>  drivers/gpu/drm/i915/i915_drv.h     |  1 +
>>  drivers/gpu/drm/i915/intel_uncore.c | 39 +++++++++++++++++++++++++++++++++++++
>>  3 files changed, 41 insertions(+)
>> 
>> diff --git a/drivers/gpu/drm/i915/Kconfig b/drivers/gpu/drm/i915/Kconfig
>> index 1ae0bb9..a5cd5da 100644
>> --- a/drivers/gpu/drm/i915/Kconfig
>> +++ b/drivers/gpu/drm/i915/Kconfig
>> @@ -20,6 +20,7 @@ config DRM_I915
>>  	select ACPI_VIDEO if ACPI
>>  	select ACPI_BUTTON if ACPI
>>  	select SYNC_FILE
>> +	select IOSF_MBI
>>  	help
>>  	  Choose this option if you have a system that has "Intel Graphics
>>  	  Media Accelerator" or "HD Graphics" integrated graphics,
>> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
>> index 998239f..4e405d0 100644
>> --- a/drivers/gpu/drm/i915/i915_drv.h
>> +++ b/drivers/gpu/drm/i915/i915_drv.h
>> @@ -725,6 +725,7 @@ struct intel_uncore {
>>  	const struct intel_forcewake_range *fw_domains_table;
>>  	unsigned int fw_domains_table_entries;
>>  
>> +	struct notifier_block pmic_bus_access_nb;
>>  	struct intel_uncore_funcs funcs;
>>  
>>  	unsigned fifo_count;
>> diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c
>> index b6d726b..1be3cdc 100644
>> --- a/drivers/gpu/drm/i915/intel_uncore.c
>> +++ b/drivers/gpu/drm/i915/intel_uncore.c
>> @@ -25,6 +25,7 @@
>>  #include "intel_drv.h"
>>  #include "i915_vgpu.h"
>>  
>> +#include <asm/iosf_mbi.h>
>>  #include <linux/pm_runtime.h>
>>  
>>  #define FORCEWAKE_ACK_TIMEOUT_MS 50
>> @@ -436,12 +437,16 @@ static void __intel_uncore_early_sanitize(struct drm_i915_private *dev_priv,
>>  
>>  void intel_uncore_suspend(struct drm_i915_private *dev_priv)
>>  {
>> +	iosf_mbi_unregister_pmic_bus_access_notifier(
>> +		&dev_priv->uncore.pmic_bus_access_nb);
>>  	intel_uncore_forcewake_reset(dev_priv, false);
>>  }
>>  
>>  void intel_uncore_resume_early(struct drm_i915_private *dev_priv)
>>  {
>>  	__intel_uncore_early_sanitize(dev_priv, true);
>> +	iosf_mbi_register_pmic_bus_access_notifier(
>> +		&dev_priv->uncore.pmic_bus_access_nb);
>>  	i915_check_and_clear_faults(dev_priv);
>>  }
>>  
>> @@ -1329,6 +1334,32 @@ static void intel_uncore_fw_domains_init(struct drm_i915_private *dev_priv)
>>  	dev_priv->uncore.fw_domains_table_entries = ARRAY_SIZE((d)); \
>>  }
>>  
>> +static int i915_pmic_bus_access_notifier(struct notifier_block *nb,
>> +					 unsigned long action, void *data)
>> +{
>> +	struct drm_i915_private *dev_priv = container_of(nb,
>> +			struct drm_i915_private, uncore.pmic_bus_access_nb);
>> +
>> +	switch (action) {
>> +	case MBI_PMIC_BUS_ACCESS_BEGIN:
>> +		/*
>> +		 * forcewake all now to make sure that we don't need to do a
>> +		 * forcewake later which on systems where this notifier gets
>> +		 * called requires the punit to access to the shared pmic i2c
>> +		 * bus, which will be busy after this notification, leading to:
>> +		 * "render: timed out waiting for forcewake ack request."
>> +		 * errors.
>> +		 */
>> +		intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
>> +		break;
>> +	case MBI_PMIC_BUS_ACCESS_END:
>> +		intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
>> +		break;
>> +	}
>> +
>> +	return NOTIFY_OK;
>> +}
>> +
>>  void intel_uncore_init(struct drm_i915_private *dev_priv)
>>  {
>>  	i915_check_vgpu(dev_priv);
>> @@ -1338,6 +1369,8 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>>  	__intel_uncore_early_sanitize(dev_priv, false);
>>  
>>  	dev_priv->uncore.unclaimed_mmio_check = 1;
>> +	dev_priv->uncore.pmic_bus_access_nb.notifier_call =
>> +		i915_pmic_bus_access_notifier;
>>  
>>  	switch (INTEL_INFO(dev_priv)->gen) {
>>  	default:
>> @@ -1392,6 +1425,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>>  	if (INTEL_GEN(dev_priv) >= 8)
>>  		intel_shadow_table_check();
>>  
>> +	iosf_mbi_register_pmic_bus_access_notifier(
>> +		&dev_priv->uncore.pmic_bus_access_nb);
>> +
>>  	i915_check_and_clear_faults(dev_priv);
>>  }
>>  #undef ASSIGN_WRITE_MMIO_VFUNCS
>> @@ -1399,6 +1435,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>>  
>>  void intel_uncore_fini(struct drm_i915_private *dev_priv)
>>  {
>> +	iosf_mbi_unregister_pmic_bus_access_notifier(
>> +		&dev_priv->uncore.pmic_bus_access_nb);
>> +
>>  	/* Paranoia: make sure we have disabled everything before we exit. */
>>  	intel_uncore_sanitize(dev_priv);
>>  	intel_uncore_forcewake_reset(dev_priv, false);
>> -- 
>> 2.9.3
Hans de Goede Feb. 20, 2017, 8:29 a.m. UTC | #3
Hi,

On 16-02-17 20:02, Jani Nikula wrote:
> On Thu, 16 Feb 2017, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote:
>> On Fri, Feb 10, 2017 at 11:28:02AM +0100, Hans de Goede wrote:
>>> Listen for PMIC bus access notifications and get FORCEWAKE_ALL while
>>> the bus is accessed to avoid needing to do any forcewakes, which need
>>> PMIC bus access, while the PMIC bus is busy:
>>>
>>> This fixes errors like these showing up in dmesg, usually followed
>>> by a gfx or system freeze:
>>>
>>> [drm:fw_domains_get [i915]] *ERROR* render: timed out waiting for forcewake ack request.
>>> [drm:fw_domains_get [i915]] *MEDIA* render: timed out waiting for forcewake ack request.
>>> i2c_designware 808622C1:06: punit semaphore timed out, resetting
>>> i2c_designware 808622C1:06: PUNIT SEM: 2
>>> i2c_designware 808622C1:06: couldn't acquire bus ownership
>>>
>>> Downside of this approach is that it causes wakeups whenever the PMIC
>>> bus is accessed. Unfortunately we cannot simply wait for the PMIC bus
>>> to go idle when we hit a race, as forcewakes may be done from interrupt
>>> handlers where we cannot sleep to wait for the i2c PMIC bus access to
>>> finish.
>>>
>>> Note that the notifications and thus the wakeups will only happen on
>>> baytrail / cherrytrail devices using PMICs with a shared i2c bus for
>>> P-Unit and host PMIC access (i2c busses with a _SEM method in their
>>> APCI node), e.g. an axp288 PMIC.
>>>
>>> I plan to write some patches for drivers accessing the PMIC bus to
>>> limit their bus accesses to a bare minimum (e.g. cache registers, do not
>>> update battery level more often then 4 times a minute), to limit the
>>> amount of wakeups.
>>>
>>> BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=155241
>>> Signed-off-by: Hans de Goede <hdegoede@redhat.com>
>>> Tested-by: tagorereddy <tagore.chandan@gmail.com>
>>
>> I gave the previous versions a quick whirl on a few machines here, but
>> them not being CR versions I guess this stuff doesn't kick in at all.
>> And I don't see any _SEM stuff in the DSDT/SSDT, so I guess that
>> confirms it. Which is fine since I've not seem any stability issues
>> on those machines. So at least nothing seemed to break :)
>>
>> Anyways the changes look all right to me, so for both i915 patches
>> Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
>
> Acked-by: Jani Nikula <jani.nikula@intel.com>
>
> for merging the i915 patches through some other tree if that makes
> managing the pile easier.

Actually the idea was for the entire pile to go through the drm-intel
tree. Daniel can you pick these up please (they seem to be ready) ?

Regards,

Hans



>
>
>>
>>> ---
>>> Changes in v2:
>>> -Spelling: P-Unit, PMIC
>>> Changes in v3:
>>> -Improve comment explaining why we call intel_uncore_forcewake_get(ALL) on
>>>  MBI_PMIC_BUS_ACCESS_BEGIN notification
>>> ---
>>>  drivers/gpu/drm/i915/Kconfig        |  1 +
>>>  drivers/gpu/drm/i915/i915_drv.h     |  1 +
>>>  drivers/gpu/drm/i915/intel_uncore.c | 39 +++++++++++++++++++++++++++++++++++++
>>>  3 files changed, 41 insertions(+)
>>>
>>> diff --git a/drivers/gpu/drm/i915/Kconfig b/drivers/gpu/drm/i915/Kconfig
>>> index 1ae0bb9..a5cd5da 100644
>>> --- a/drivers/gpu/drm/i915/Kconfig
>>> +++ b/drivers/gpu/drm/i915/Kconfig
>>> @@ -20,6 +20,7 @@ config DRM_I915
>>>  	select ACPI_VIDEO if ACPI
>>>  	select ACPI_BUTTON if ACPI
>>>  	select SYNC_FILE
>>> +	select IOSF_MBI
>>>  	help
>>>  	  Choose this option if you have a system that has "Intel Graphics
>>>  	  Media Accelerator" or "HD Graphics" integrated graphics,
>>> diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
>>> index 998239f..4e405d0 100644
>>> --- a/drivers/gpu/drm/i915/i915_drv.h
>>> +++ b/drivers/gpu/drm/i915/i915_drv.h
>>> @@ -725,6 +725,7 @@ struct intel_uncore {
>>>  	const struct intel_forcewake_range *fw_domains_table;
>>>  	unsigned int fw_domains_table_entries;
>>>
>>> +	struct notifier_block pmic_bus_access_nb;
>>>  	struct intel_uncore_funcs funcs;
>>>
>>>  	unsigned fifo_count;
>>> diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c
>>> index b6d726b..1be3cdc 100644
>>> --- a/drivers/gpu/drm/i915/intel_uncore.c
>>> +++ b/drivers/gpu/drm/i915/intel_uncore.c
>>> @@ -25,6 +25,7 @@
>>>  #include "intel_drv.h"
>>>  #include "i915_vgpu.h"
>>>
>>> +#include <asm/iosf_mbi.h>
>>>  #include <linux/pm_runtime.h>
>>>
>>>  #define FORCEWAKE_ACK_TIMEOUT_MS 50
>>> @@ -436,12 +437,16 @@ static void __intel_uncore_early_sanitize(struct drm_i915_private *dev_priv,
>>>
>>>  void intel_uncore_suspend(struct drm_i915_private *dev_priv)
>>>  {
>>> +	iosf_mbi_unregister_pmic_bus_access_notifier(
>>> +		&dev_priv->uncore.pmic_bus_access_nb);
>>>  	intel_uncore_forcewake_reset(dev_priv, false);
>>>  }
>>>
>>>  void intel_uncore_resume_early(struct drm_i915_private *dev_priv)
>>>  {
>>>  	__intel_uncore_early_sanitize(dev_priv, true);
>>> +	iosf_mbi_register_pmic_bus_access_notifier(
>>> +		&dev_priv->uncore.pmic_bus_access_nb);
>>>  	i915_check_and_clear_faults(dev_priv);
>>>  }
>>>
>>> @@ -1329,6 +1334,32 @@ static void intel_uncore_fw_domains_init(struct drm_i915_private *dev_priv)
>>>  	dev_priv->uncore.fw_domains_table_entries = ARRAY_SIZE((d)); \
>>>  }
>>>
>>> +static int i915_pmic_bus_access_notifier(struct notifier_block *nb,
>>> +					 unsigned long action, void *data)
>>> +{
>>> +	struct drm_i915_private *dev_priv = container_of(nb,
>>> +			struct drm_i915_private, uncore.pmic_bus_access_nb);
>>> +
>>> +	switch (action) {
>>> +	case MBI_PMIC_BUS_ACCESS_BEGIN:
>>> +		/*
>>> +		 * forcewake all now to make sure that we don't need to do a
>>> +		 * forcewake later which on systems where this notifier gets
>>> +		 * called requires the punit to access to the shared pmic i2c
>>> +		 * bus, which will be busy after this notification, leading to:
>>> +		 * "render: timed out waiting for forcewake ack request."
>>> +		 * errors.
>>> +		 */
>>> +		intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
>>> +		break;
>>> +	case MBI_PMIC_BUS_ACCESS_END:
>>> +		intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
>>> +		break;
>>> +	}
>>> +
>>> +	return NOTIFY_OK;
>>> +}
>>> +
>>>  void intel_uncore_init(struct drm_i915_private *dev_priv)
>>>  {
>>>  	i915_check_vgpu(dev_priv);
>>> @@ -1338,6 +1369,8 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>>>  	__intel_uncore_early_sanitize(dev_priv, false);
>>>
>>>  	dev_priv->uncore.unclaimed_mmio_check = 1;
>>> +	dev_priv->uncore.pmic_bus_access_nb.notifier_call =
>>> +		i915_pmic_bus_access_notifier;
>>>
>>>  	switch (INTEL_INFO(dev_priv)->gen) {
>>>  	default:
>>> @@ -1392,6 +1425,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>>>  	if (INTEL_GEN(dev_priv) >= 8)
>>>  		intel_shadow_table_check();
>>>
>>> +	iosf_mbi_register_pmic_bus_access_notifier(
>>> +		&dev_priv->uncore.pmic_bus_access_nb);
>>> +
>>>  	i915_check_and_clear_faults(dev_priv);
>>>  }
>>>  #undef ASSIGN_WRITE_MMIO_VFUNCS
>>> @@ -1399,6 +1435,9 @@ void intel_uncore_init(struct drm_i915_private *dev_priv)
>>>
>>>  void intel_uncore_fini(struct drm_i915_private *dev_priv)
>>>  {
>>> +	iosf_mbi_unregister_pmic_bus_access_notifier(
>>> +		&dev_priv->uncore.pmic_bus_access_nb);
>>> +
>>>  	/* Paranoia: make sure we have disabled everything before we exit. */
>>>  	intel_uncore_sanitize(dev_priv);
>>>  	intel_uncore_forcewake_reset(dev_priv, false);
>>> --
>>> 2.9.3
>
Daniel Vetter Feb. 26, 2017, 8:01 p.m. UTC | #4
On Mon, Feb 20, 2017 at 09:29:10AM +0100, Hans de Goede wrote:
> Hi,
> 
> On 16-02-17 20:02, Jani Nikula wrote:
> > On Thu, 16 Feb 2017, Ville Syrjälä <ville.syrjala@linux.intel.com> wrote:
> > > On Fri, Feb 10, 2017 at 11:28:02AM +0100, Hans de Goede wrote:
> > > > Listen for PMIC bus access notifications and get FORCEWAKE_ALL while
> > > > the bus is accessed to avoid needing to do any forcewakes, which need
> > > > PMIC bus access, while the PMIC bus is busy:
> > > > 
> > > > This fixes errors like these showing up in dmesg, usually followed
> > > > by a gfx or system freeze:
> > > > 
> > > > [drm:fw_domains_get [i915]] *ERROR* render: timed out waiting for forcewake ack request.
> > > > [drm:fw_domains_get [i915]] *MEDIA* render: timed out waiting for forcewake ack request.
> > > > i2c_designware 808622C1:06: punit semaphore timed out, resetting
> > > > i2c_designware 808622C1:06: PUNIT SEM: 2
> > > > i2c_designware 808622C1:06: couldn't acquire bus ownership
> > > > 
> > > > Downside of this approach is that it causes wakeups whenever the PMIC
> > > > bus is accessed. Unfortunately we cannot simply wait for the PMIC bus
> > > > to go idle when we hit a race, as forcewakes may be done from interrupt
> > > > handlers where we cannot sleep to wait for the i2c PMIC bus access to
> > > > finish.
> > > > 
> > > > Note that the notifications and thus the wakeups will only happen on
> > > > baytrail / cherrytrail devices using PMICs with a shared i2c bus for
> > > > P-Unit and host PMIC access (i2c busses with a _SEM method in their
> > > > APCI node), e.g. an axp288 PMIC.
> > > > 
> > > > I plan to write some patches for drivers accessing the PMIC bus to
> > > > limit their bus accesses to a bare minimum (e.g. cache registers, do not
> > > > update battery level more often then 4 times a minute), to limit the
> > > > amount of wakeups.
> > > > 
> > > > BugLink: https://bugzilla.kernel.org/show_bug.cgi?id=155241
> > > > Signed-off-by: Hans de Goede <hdegoede@redhat.com>
> > > > Tested-by: tagorereddy <tagore.chandan@gmail.com>
> > > 
> > > I gave the previous versions a quick whirl on a few machines here, but
> > > them not being CR versions I guess this stuff doesn't kick in at all.
> > > And I don't see any _SEM stuff in the DSDT/SSDT, so I guess that
> > > confirms it. Which is fine since I've not seem any stability issues
> > > on those machines. So at least nothing seemed to break :)
> > > 
> > > Anyways the changes look all right to me, so for both i915 patches
> > > Reviewed-by: Ville Syrjälä <ville.syrjala@linux.intel.com>
> > 
> > Acked-by: Jani Nikula <jani.nikula@intel.com>
> > 
> > for merging the i915 patches through some other tree if that makes
> > managing the pile easier.
> 
> Actually the idea was for the entire pile to go through the drm-intel
> tree. Daniel can you pick these up please (they seem to be ready) ?

For the future: Both Ville and Jani have commit rights for drm-intel,
Jani's even officially co-maintainer. No need to wait for me to get back
from travelling at all.

I'm applying them now, but only this time :-)

Cheers, Daniel
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/Kconfig b/drivers/gpu/drm/i915/Kconfig
index 1ae0bb9..a5cd5da 100644
--- a/drivers/gpu/drm/i915/Kconfig
+++ b/drivers/gpu/drm/i915/Kconfig
@@ -20,6 +20,7 @@  config DRM_I915
 	select ACPI_VIDEO if ACPI
 	select ACPI_BUTTON if ACPI
 	select SYNC_FILE
+	select IOSF_MBI
 	help
 	  Choose this option if you have a system that has "Intel Graphics
 	  Media Accelerator" or "HD Graphics" integrated graphics,
diff --git a/drivers/gpu/drm/i915/i915_drv.h b/drivers/gpu/drm/i915/i915_drv.h
index 998239f..4e405d0 100644
--- a/drivers/gpu/drm/i915/i915_drv.h
+++ b/drivers/gpu/drm/i915/i915_drv.h
@@ -725,6 +725,7 @@  struct intel_uncore {
 	const struct intel_forcewake_range *fw_domains_table;
 	unsigned int fw_domains_table_entries;
 
+	struct notifier_block pmic_bus_access_nb;
 	struct intel_uncore_funcs funcs;
 
 	unsigned fifo_count;
diff --git a/drivers/gpu/drm/i915/intel_uncore.c b/drivers/gpu/drm/i915/intel_uncore.c
index b6d726b..1be3cdc 100644
--- a/drivers/gpu/drm/i915/intel_uncore.c
+++ b/drivers/gpu/drm/i915/intel_uncore.c
@@ -25,6 +25,7 @@ 
 #include "intel_drv.h"
 #include "i915_vgpu.h"
 
+#include <asm/iosf_mbi.h>
 #include <linux/pm_runtime.h>
 
 #define FORCEWAKE_ACK_TIMEOUT_MS 50
@@ -436,12 +437,16 @@  static void __intel_uncore_early_sanitize(struct drm_i915_private *dev_priv,
 
 void intel_uncore_suspend(struct drm_i915_private *dev_priv)
 {
+	iosf_mbi_unregister_pmic_bus_access_notifier(
+		&dev_priv->uncore.pmic_bus_access_nb);
 	intel_uncore_forcewake_reset(dev_priv, false);
 }
 
 void intel_uncore_resume_early(struct drm_i915_private *dev_priv)
 {
 	__intel_uncore_early_sanitize(dev_priv, true);
+	iosf_mbi_register_pmic_bus_access_notifier(
+		&dev_priv->uncore.pmic_bus_access_nb);
 	i915_check_and_clear_faults(dev_priv);
 }
 
@@ -1329,6 +1334,32 @@  static void intel_uncore_fw_domains_init(struct drm_i915_private *dev_priv)
 	dev_priv->uncore.fw_domains_table_entries = ARRAY_SIZE((d)); \
 }
 
+static int i915_pmic_bus_access_notifier(struct notifier_block *nb,
+					 unsigned long action, void *data)
+{
+	struct drm_i915_private *dev_priv = container_of(nb,
+			struct drm_i915_private, uncore.pmic_bus_access_nb);
+
+	switch (action) {
+	case MBI_PMIC_BUS_ACCESS_BEGIN:
+		/*
+		 * forcewake all now to make sure that we don't need to do a
+		 * forcewake later which on systems where this notifier gets
+		 * called requires the punit to access to the shared pmic i2c
+		 * bus, which will be busy after this notification, leading to:
+		 * "render: timed out waiting for forcewake ack request."
+		 * errors.
+		 */
+		intel_uncore_forcewake_get(dev_priv, FORCEWAKE_ALL);
+		break;
+	case MBI_PMIC_BUS_ACCESS_END:
+		intel_uncore_forcewake_put(dev_priv, FORCEWAKE_ALL);
+		break;
+	}
+
+	return NOTIFY_OK;
+}
+
 void intel_uncore_init(struct drm_i915_private *dev_priv)
 {
 	i915_check_vgpu(dev_priv);
@@ -1338,6 +1369,8 @@  void intel_uncore_init(struct drm_i915_private *dev_priv)
 	__intel_uncore_early_sanitize(dev_priv, false);
 
 	dev_priv->uncore.unclaimed_mmio_check = 1;
+	dev_priv->uncore.pmic_bus_access_nb.notifier_call =
+		i915_pmic_bus_access_notifier;
 
 	switch (INTEL_INFO(dev_priv)->gen) {
 	default:
@@ -1392,6 +1425,9 @@  void intel_uncore_init(struct drm_i915_private *dev_priv)
 	if (INTEL_GEN(dev_priv) >= 8)
 		intel_shadow_table_check();
 
+	iosf_mbi_register_pmic_bus_access_notifier(
+		&dev_priv->uncore.pmic_bus_access_nb);
+
 	i915_check_and_clear_faults(dev_priv);
 }
 #undef ASSIGN_WRITE_MMIO_VFUNCS
@@ -1399,6 +1435,9 @@  void intel_uncore_init(struct drm_i915_private *dev_priv)
 
 void intel_uncore_fini(struct drm_i915_private *dev_priv)
 {
+	iosf_mbi_unregister_pmic_bus_access_notifier(
+		&dev_priv->uncore.pmic_bus_access_nb);
+
 	/* Paranoia: make sure we have disabled everything before we exit. */
 	intel_uncore_sanitize(dev_priv);
 	intel_uncore_forcewake_reset(dev_priv, false);