[V6,09/11] ACPI: introduce flag .is_master_device
diff mbox series

Message ID 1400136256-2218-10-git-send-email-rui.zhang@intel.com
State New, archived
Headers show
Series
  • ACPI: ACPI enumeration rework
Related show

Commit Message

Zhang, Rui May 15, 2014, 6:44 a.m. UTC
For some ACPI device objects, they represent master devices,
and their children devices are enumerated by bus controller drivers
for the buses they are on.

In this case, we do not want to enumerate their children devices to
platform bus explicitly in acpi scan code.

Thus a new flag .is_master_device is introduced in this patch.

For devices with this flag set, we will not do default enumeration
for their children.

Signed-off-by: Zhang Rui <rui.zhang@intel.com>
---
 drivers/acpi/acpi_lpss.c | 62 +++++++++++++++++++++++++++++++++---------------
 include/acpi/acpi_bus.h  |  3 ++-
 2 files changed, 45 insertions(+), 20 deletions(-)

Comments

Mika Westerberg May 21, 2014, 8:52 a.m. UTC | #1
On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> For some ACPI device objects, they represent master devices,
> and their children devices are enumerated by bus controller drivers
> for the buses they are on.
> 
> In this case, we do not want to enumerate their children devices to
> platform bus explicitly in acpi scan code.
> 
> Thus a new flag .is_master_device is introduced in this patch.
> 
> For devices with this flag set, we will not do default enumeration
> for their children.

Is there any particular reason we would like to enumerate everything
below the first device by default?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Mika Westerberg May 21, 2014, 11:04 a.m. UTC | #2
On Wed, May 21, 2014 at 01:10:33PM +0200, Rafael J. Wysocki wrote:
> On Wednesday, May 21, 2014 11:52:07 AM Mika Westerberg wrote:
> > On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > > For some ACPI device objects, they represent master devices,
> > > and their children devices are enumerated by bus controller drivers
> > > for the buses they are on.
> > > 
> > > In this case, we do not want to enumerate their children devices to
> > > platform bus explicitly in acpi scan code.
> > > 
> > > Thus a new flag .is_master_device is introduced in this patch.
> > > 
> > > For devices with this flag set, we will not do default enumeration
> > > for their children.
> > 
> > Is there any particular reason we would like to enumerate everything
> > below the first device by default?
> 
> Yes, there is.  Device objects without _ADR under the PCI host bridge.

OK.

> Or we can skip the children under every *platform* device created by this by
> default and mark the ones where we want the children to be enumerated as
> platform devices too in a special way if needed.
> 
> I guess we could try that (that was the Rui's original idea IIRC).

That sounds better to me.

I wonder if we can do this analogous to of_platform_bus_probe() and
friends?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Rafael J. Wysocki May 21, 2014, 11:10 a.m. UTC | #3
On Wednesday, May 21, 2014 11:52:07 AM Mika Westerberg wrote:
> On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > For some ACPI device objects, they represent master devices,
> > and their children devices are enumerated by bus controller drivers
> > for the buses they are on.
> > 
> > In this case, we do not want to enumerate their children devices to
> > platform bus explicitly in acpi scan code.
> > 
> > Thus a new flag .is_master_device is introduced in this patch.
> > 
> > For devices with this flag set, we will not do default enumeration
> > for their children.
> 
> Is there any particular reason we would like to enumerate everything
> below the first device by default?

Yes, there is.  Device objects without _ADR under the PCI host bridge.

Or we can skip the children under every *platform* device created by this by
default and mark the ones where we want the children to be enumerated as
platform devices too in a special way if needed.

I guess we could try that (that was the Rui's original idea IIRC).

Rafael

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Rafael J. Wysocki May 21, 2014, 11:59 a.m. UTC | #4
On 5/21/2014 1:04 PM, Mika Westerberg wrote:
> On Wed, May 21, 2014 at 01:10:33PM +0200, Rafael J. Wysocki wrote:
>> On Wednesday, May 21, 2014 11:52:07 AM Mika Westerberg wrote:
>>> On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
>>>> For some ACPI device objects, they represent master devices,
>>>> and their children devices are enumerated by bus controller drivers
>>>> for the buses they are on.
>>>>
>>>> In this case, we do not want to enumerate their children devices to
>>>> platform bus explicitly in acpi scan code.
>>>>
>>>> Thus a new flag .is_master_device is introduced in this patch.
>>>>
>>>> For devices with this flag set, we will not do default enumeration
>>>> for their children.
>>> Is there any particular reason we would like to enumerate everything
>>> below the first device by default?
>> Yes, there is.  Device objects without _ADR under the PCI host bridge.
> OK.
>
>> Or we can skip the children under every *platform* device created by this by
>> default and mark the ones where we want the children to be enumerated as
>> platform devices too in a special way if needed.
>>
>> I guess we could try that (that was the Rui's original idea IIRC).
> That sounds better to me.
>
> I wonder if we can do this analogous to of_platform_bus_probe() and
> friends?

Yes, we should actually.

Rafael

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Zhang, Rui May 21, 2014, 2:43 p.m. UTC | #5
On 三, 2014-05-21 at 11:52 +0300, Mika Westerberg wrote:
> On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > For some ACPI device objects, they represent master devices,
> > and their children devices are enumerated by bus controller drivers
> > for the buses they are on.
> > 
> > In this case, we do not want to enumerate their children devices to
> > platform bus explicitly in acpi scan code.
> > 
> > Thus a new flag .is_master_device is introduced in this patch.
> > 
> > For devices with this flag set, we will not do default enumeration
> > for their children.
> 
> Is there any particular reason we would like to enumerate everything
> below the first device by default?

we do not enumerate everything below the first device by default, we
just enumerate all the devices with _HID.
But if a device has _HID and it is enumerated by its parents to a
separate bus, we need this flag set for its parent.

thanks,
rui

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Zhang, Rui May 21, 2014, 3:09 p.m. UTC | #6
On 三, 2014-05-21 at 13:10 +0200, Rafael J. Wysocki wrote:
> On Wednesday, May 21, 2014 11:52:07 AM Mika Westerberg wrote:
> > On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > > For some ACPI device objects, they represent master devices,
> > > and their children devices are enumerated by bus controller drivers
> > > for the buses they are on.
> > > 
> > > In this case, we do not want to enumerate their children devices to
> > > platform bus explicitly in acpi scan code.
> > > 
> > > Thus a new flag .is_master_device is introduced in this patch.
> > > 
> > > For devices with this flag set, we will not do default enumeration
> > > for their children.
> > 
> > Is there any particular reason we would like to enumerate everything
> > below the first device by default?
> 
> Yes, there is.  Device objects without _ADR under the PCI host bridge.
> 
I'm confused. Mika, what do you mean by saying "first device"?

thanks,
rui
> Or we can skip the children under every *platform* device created by this by
> default and mark the ones where we want the children to be enumerated as
> platform devices too in a special way if needed.
> 
> I guess we could try that (that was the Rui's original idea IIRC).
> 
> Rafael
> 


--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Mika Westerberg May 22, 2014, 8:08 a.m. UTC | #7
On Wed, May 21, 2014 at 11:09:42PM +0800, Zhang Rui wrote:
> On 三, 2014-05-21 at 13:10 +0200, Rafael J. Wysocki wrote:
> > On Wednesday, May 21, 2014 11:52:07 AM Mika Westerberg wrote:
> > > On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > > > For some ACPI device objects, they represent master devices,
> > > > and their children devices are enumerated by bus controller drivers
> > > > for the buses they are on.
> > > > 
> > > > In this case, we do not want to enumerate their children devices to
> > > > platform bus explicitly in acpi scan code.
> > > > 
> > > > Thus a new flag .is_master_device is introduced in this patch.
> > > > 
> > > > For devices with this flag set, we will not do default enumeration
> > > > for their children.
> > > 
> > > Is there any particular reason we would like to enumerate everything
> > > below the first device by default?
> > 
> > Yes, there is.  Device objects without _ADR under the PCI host bridge.
> > 
> I'm confused. Mika, what do you mean by saying "first device"?

I mean the first level devices right below the host bridge, such as
\_SB.PCI0.* but not \_SB.PCI0.*.FOO.
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Mika Westerberg May 22, 2014, 8:51 a.m. UTC | #8
On Wed, May 21, 2014 at 10:43:07PM +0800, Zhang Rui wrote:
> On 三, 2014-05-21 at 11:52 +0300, Mika Westerberg wrote:
> > On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > > For some ACPI device objects, they represent master devices,
> > > and their children devices are enumerated by bus controller drivers
> > > for the buses they are on.
> > > 
> > > In this case, we do not want to enumerate their children devices to
> > > platform bus explicitly in acpi scan code.
> > > 
> > > Thus a new flag .is_master_device is introduced in this patch.
> > > 
> > > For devices with this flag set, we will not do default enumeration
> > > for their children.
> > 
> > Is there any particular reason we would like to enumerate everything
> > below the first device by default?
> 
> we do not enumerate everything below the first device by default, we
> just enumerate all the devices with _HID.

OK.

> But if a device has _HID and it is enumerated by its parents to a
> separate bus, we need this flag set for its parent.

How about checking if the device has *SerialBus() connector and in such
case skip the device (given that it is not listed in a special list,
like acpi_platform_device_ids)?
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/
Zhang, Rui May 22, 2014, 2:26 p.m. UTC | #9
On Thu, 2014-05-22 at 11:51 +0300, Mika Westerberg wrote:
> On Wed, May 21, 2014 at 10:43:07PM +0800, Zhang Rui wrote:
> > On 三, 2014-05-21 at 11:52 +0300, Mika Westerberg wrote:
> > > On Thu, May 15, 2014 at 02:44:14PM +0800, Zhang Rui wrote:
> > > > For some ACPI device objects, they represent master devices,
> > > > and their children devices are enumerated by bus controller drivers
> > > > for the buses they are on.
> > > > 
> > > > In this case, we do not want to enumerate their children devices to
> > > > platform bus explicitly in acpi scan code.
> > > > 
> > > > Thus a new flag .is_master_device is introduced in this patch.
> > > > 
> > > > For devices with this flag set, we will not do default enumeration
> > > > for their children.
> > > 
> > > Is there any particular reason we would like to enumerate everything
> > > below the first device by default?
> > 
> > we do not enumerate everything below the first device by default, we
> > just enumerate all the devices with _HID.
> 
> OK.
> 
> > But if a device has _HID and it is enumerated by its parents to a
> > separate bus, we need this flag set for its parent.
> 
> How about checking if the device has *SerialBus() connector and in such
> case skip the device (given that it is not listed in a special list,
> like acpi_platform_device_ids)?

This sounds like a good idea.
I think we can just ignore devices with ACPI_RESOURCE_TYPE_SERIAL_BUS
resources, and this can be done in drivers/acpi/scan.c for all _HID
devices w/o handler attached, right?

thanks,
rui

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Patch
diff mbox series

diff --git a/drivers/acpi/acpi_lpss.c b/drivers/acpi/acpi_lpss.c
index 965428f..4ab0915 100644
--- a/drivers/acpi/acpi_lpss.c
+++ b/drivers/acpi/acpi_lpss.c
@@ -54,6 +54,7 @@  struct lpss_shared_clock {
 struct lpss_private_data;
 
 struct lpss_device_desc {
+	bool is_master_device;
 	bool clk_required;
 	const char *clkdev_name;
 	bool ltr_required;
@@ -91,6 +92,7 @@  static void lpss_uart_setup(struct lpss_private_data *pdata)
 }
 
 static struct lpss_device_desc lpt_dev_desc = {
+	.is_master_device = true,
 	.clk_required = true,
 	.prv_offset = 0x800,
 	.ltr_required = true,
@@ -98,6 +100,7 @@  static struct lpss_device_desc lpt_dev_desc = {
 };
 
 static struct lpss_device_desc lpt_uart_dev_desc = {
+	.is_master_device = true,
 	.clk_required = true,
 	.prv_offset = 0x800,
 	.ltr_required = true,
@@ -106,6 +109,7 @@  static struct lpss_device_desc lpt_uart_dev_desc = {
 };
 
 static struct lpss_device_desc lpt_sdio_dev_desc = {
+	.is_master_device = true,
 	.prv_offset = 0x1000,
 	.prv_size_override = 0x1018,
 	.ltr_required = true,
@@ -127,6 +131,7 @@  static struct lpss_shared_clock uart_clock = {
 };
 
 static struct lpss_device_desc byt_uart_dev_desc = {
+	.is_master_device = true,
 	.clk_required = true,
 	.prv_offset = 0x800,
 	.clk_gate = true,
@@ -140,6 +145,7 @@  static struct lpss_shared_clock spi_clock = {
 };
 
 static struct lpss_device_desc byt_spi_dev_desc = {
+	.is_master_device = true,
 	.clk_required = true,
 	.prv_offset = 0x400,
 	.clk_gate = true,
@@ -147,6 +153,7 @@  static struct lpss_device_desc byt_spi_dev_desc = {
 };
 
 static struct lpss_device_desc byt_sdio_dev_desc = {
+	.is_master_device = true,
 	.clk_required = true,
 };
 
@@ -156,16 +163,24 @@  static struct lpss_shared_clock i2c_clock = {
 };
 
 static struct lpss_device_desc byt_i2c_dev_desc = {
+	.is_master_device = true,
 	.clk_required = true,
 	.prv_offset = 0x800,
 	.shared_clock = &i2c_clock,
 };
 
 #define LPSS_PTR(desc) ((unsigned long)&desc)
+#define LPSS_MASTER_PTR(desc) LPSS_PTR(desc)
 
 #else
 
-#define LPSS_PTR(desc) 0
+static struct lpss_device_desc lpss_dummy_desc;
+static struct lpss_device_desc lpss_dummy_master_desc = {
+	.is_master_device = true,
+};
+
+#define LPSS_PTR(desc) (&lpss_dummy_desc)
+#define LPSS_MASTER_PTR(desc) (&lpss_dummy_master_desc)
 
 #endif
 
@@ -174,30 +189,30 @@  static const struct acpi_device_id acpi_lpss_device_ids[] = {
 	{ "INTL9C60", LPSS_PTR(lpss_dma_desc) },
 
 	/* Lynxpoint LPSS devices */
-	{ "INT33C0", LPSS_PTR(lpt_dev_desc) },
-	{ "INT33C1", LPSS_PTR(lpt_dev_desc) },
-	{ "INT33C2", LPSS_PTR(lpt_dev_desc) },
-	{ "INT33C3", LPSS_PTR(lpt_dev_desc) },
-	{ "INT33C4", LPSS_PTR(lpt_uart_dev_desc) },
-	{ "INT33C5", LPSS_PTR(lpt_uart_dev_desc) },
-	{ "INT33C6", LPSS_PTR(lpt_sdio_dev_desc) },
+	{ "INT33C0", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT33C1", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT33C2", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT33C3", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT33C4", LPSS_MASTER_PTR(lpt_uart_dev_desc) },
+	{ "INT33C5", LPSS_MASTER_PTR(lpt_uart_dev_desc) },
+	{ "INT33C6", LPSS_MASTER_PTR(lpt_sdio_dev_desc) },
 	{ "INT33C7", },
 
 	/* BayTrail LPSS devices */
 	{ "80860F09", LPSS_PTR(byt_pwm_dev_desc) },
-	{ "80860F0A", LPSS_PTR(byt_uart_dev_desc) },
-	{ "80860F0E", LPSS_PTR(byt_spi_dev_desc) },
-	{ "80860F14", LPSS_PTR(byt_sdio_dev_desc) },
-	{ "80860F41", LPSS_PTR(byt_i2c_dev_desc) },
+	{ "80860F0A", LPSS_MASTER_PTR(byt_uart_dev_desc) },
+	{ "80860F0E", LPSS_MASTER_PTR(byt_spi_dev_desc) },
+	{ "80860F14", LPSS_MASTER_PTR(byt_sdio_dev_desc) },
+	{ "80860F41", LPSS_MASTER_PTR(byt_i2c_dev_desc) },
 	{ "INT33B2", },
 
-	{ "INT3430", LPSS_PTR(lpt_dev_desc) },
-	{ "INT3431", LPSS_PTR(lpt_dev_desc) },
-	{ "INT3432", LPSS_PTR(lpt_dev_desc) },
-	{ "INT3433", LPSS_PTR(lpt_dev_desc) },
-	{ "INT3434", LPSS_PTR(lpt_uart_dev_desc) },
-	{ "INT3435", LPSS_PTR(lpt_uart_dev_desc) },
-	{ "INT3436", LPSS_PTR(lpt_sdio_dev_desc) },
+	{ "INT3430", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT3431", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT3432", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT3433", LPSS_MASTER_PTR(lpt_dev_desc) },
+	{ "INT3434", LPSS_MASTER_PTR(lpt_uart_dev_desc) },
+	{ "INT3435", LPSS_MASTER_PTR(lpt_uart_dev_desc) },
+	{ "INT3436", LPSS_MASTER_PTR(lpt_sdio_dev_desc) },
 	{ "INT3437", },
 
 	{ }
@@ -285,6 +300,8 @@  static int acpi_lpss_create_device(struct acpi_device *adev,
 	if (!dev_desc)
 		return acpi_create_platform_device(adev, id);
 
+	adev->flags.is_master_device = dev_desc->is_master_device;
+
 	pdata = kzalloc(sizeof(*pdata), GFP_KERNEL);
 	if (!pdata)
 		return -ENOMEM;
@@ -528,6 +545,13 @@  static struct acpi_scan_handler lpss_handler = {
 static int acpi_lpss_dummy_attach(struct acpi_device *adev,
 				const struct acpi_device_id *id)
 {
+	struct lpss_device_desc *dev_desc;
+
+	dev_desc = (struct lpss_device_desc *)id->driver_data;
+	if (!dev_desc)
+		return 1;
+
+	adev->flags.is_master_device = dev_desc->is_master_device;
 	return 1;
 }
 
diff --git a/include/acpi/acpi_bus.h b/include/acpi/acpi_bus.h
index ec92ad3..13ac75f 100644
--- a/include/acpi/acpi_bus.h
+++ b/include/acpi/acpi_bus.h
@@ -207,7 +207,8 @@  struct acpi_device_flags {
 	u32 no_hotplug:1;
 	u32 hotplug_notify:1;
 	u32 is_dock_station:1;
-	u32 reserved:22;
+	u32 is_master_device:1;
+	u32 reserved:21;
 };
 
 /* File System */