diff mbox

[RFC,2/2] ata: acpi: rework the ata acpi bind support

Message ID 1374731223-28685-3-git-send-email-aaron.lu@intel.com
State Not Applicable
Delegated to: David Miller
Headers show

Commit Message

Aaron Lu July 25, 2013, 5:47 a.m. UTC
Binding ACPI handle to SCSI device has several drawbacks, namely:
1 During ATA device initialization time, ACPI handle will be needed
  while SCSI devices are not created yet. So each time ACPI handle is
  needed, instead of retrieving the handle by ACPI_HANDLE macro,
  a namespace scan is performed to find the handle for the corresponding
  ATA device. This is inefficient, and also expose a restriction on
  calling path not holding any lock.
2 The binding to SCSI device tree makes code complex, while at the same
  time doesn't bring us any benefit. All ACPI handlings are still done
  in ATA module, not in SCSI.

Rework the ATA ACPI binding code to bind ACPI handle to ATA transport
devices(ATA port and ATA device). The binding needs to be done only once,
since the ATA transport devices do not go away with hotplug. And due to
this, the flush_work call in hotplug handler for ATA bay is no longer
needed.

Signed-off-by: Aaron Lu <aaron.lu@intel.com>
---
 drivers/ata/libata-acpi.c      | 278 ++++++++++++-----------------------------
 drivers/ata/libata-core.c      |   5 -
 drivers/ata/libata-scsi.c      |  13 +-
 drivers/ata/libata-transport.c |   2 +
 drivers/ata/libata-zpodd.c     |  12 +-
 drivers/ata/libata.h           |  16 ++-
 drivers/ata/pata_acpi.c        |   4 +-
 drivers/scsi/scsi_lib.c        |  22 ----
 include/linux/libata.h         |   2 -
 9 files changed, 106 insertions(+), 248 deletions(-)

Comments

Tejun Heo July 25, 2013, 2:52 p.m. UTC | #1
On Thu, Jul 25, 2013 at 01:47:03PM +0800, Aaron Lu wrote:
> Binding ACPI handle to SCSI device has several drawbacks, namely:
> 1 During ATA device initialization time, ACPI handle will be needed
>   while SCSI devices are not created yet. So each time ACPI handle is
>   needed, instead of retrieving the handle by ACPI_HANDLE macro,
>   a namespace scan is performed to find the handle for the corresponding
>   ATA device. This is inefficient, and also expose a restriction on
>   calling path not holding any lock.
> 2 The binding to SCSI device tree makes code complex, while at the same
>   time doesn't bring us any benefit. All ACPI handlings are still done
>   in ATA module, not in SCSI.
> 
> Rework the ATA ACPI binding code to bind ACPI handle to ATA transport
> devices(ATA port and ATA device). The binding needs to be done only once,
> since the ATA transport devices do not go away with hotplug. And due to
> this, the flush_work call in hotplug handler for ATA bay is no longer
> needed.

I like it but am wondering why we weren't doing this before.  Was the
acpi support added before we made ata objects proper devices?

Thanks.
Aaron Lu July 26, 2013, 1:37 a.m. UTC | #2
On 07/25/2013 10:52 PM, Tejun Heo wrote:
> On Thu, Jul 25, 2013 at 01:47:03PM +0800, Aaron Lu wrote:
>> Binding ACPI handle to SCSI device has several drawbacks, namely:
>> 1 During ATA device initialization time, ACPI handle will be needed
>>   while SCSI devices are not created yet. So each time ACPI handle is
>>   needed, instead of retrieving the handle by ACPI_HANDLE macro,
>>   a namespace scan is performed to find the handle for the corresponding
>>   ATA device. This is inefficient, and also expose a restriction on
>>   calling path not holding any lock.
>> 2 The binding to SCSI device tree makes code complex, while at the same
>>   time doesn't bring us any benefit. All ACPI handlings are still done
>>   in ATA module, not in SCSI.
>>
>> Rework the ATA ACPI binding code to bind ACPI handle to ATA transport
>> devices(ATA port and ATA device). The binding needs to be done only once,
>> since the ATA transport devices do not go away with hotplug. And due to
>> this, the flush_work call in hotplug handler for ATA bay is no longer
>> needed.
> 
> I like it but am wondering why we weren't doing this before.  Was the
> acpi support added before we made ata objects proper devices?

I think so. But since I didn't do the original binding, I can only guess
:-)

At the time of the original binding, ACPI binding logic requires the
binding device has a bus type, which these ATA transport devices don't
have.

Later, the ACPI binding code evolves and no such limitation exists
anymore. As you can see, we can simply set the ACPI handle before this
device is added in driver core, and the binding will be done.


Thanks,
Aaron
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Rafael J. Wysocki July 26, 2013, 12:48 p.m. UTC | #3
On Friday, July 26, 2013 09:37:09 AM Aaron Lu wrote:
> On 07/25/2013 10:52 PM, Tejun Heo wrote:
> > On Thu, Jul 25, 2013 at 01:47:03PM +0800, Aaron Lu wrote:
> >> Binding ACPI handle to SCSI device has several drawbacks, namely:
> >> 1 During ATA device initialization time, ACPI handle will be needed
> >>   while SCSI devices are not created yet. So each time ACPI handle is
> >>   needed, instead of retrieving the handle by ACPI_HANDLE macro,
> >>   a namespace scan is performed to find the handle for the corresponding
> >>   ATA device. This is inefficient, and also expose a restriction on
> >>   calling path not holding any lock.
> >> 2 The binding to SCSI device tree makes code complex, while at the same
> >>   time doesn't bring us any benefit. All ACPI handlings are still done
> >>   in ATA module, not in SCSI.
> >>
> >> Rework the ATA ACPI binding code to bind ACPI handle to ATA transport
> >> devices(ATA port and ATA device). The binding needs to be done only once,
> >> since the ATA transport devices do not go away with hotplug. And due to
> >> this, the flush_work call in hotplug handler for ATA bay is no longer
> >> needed.
> > 
> > I like it but am wondering why we weren't doing this before.  Was the
> > acpi support added before we made ata objects proper devices?
> 
> I think so. But since I didn't do the original binding, I can only guess
> :-)
> 
> At the time of the original binding, ACPI binding logic requires the
> binding device has a bus type, which these ATA transport devices don't
> have.
> 
> Later, the ACPI binding code evolves and no such limitation exists
> anymore. As you can see, we can simply set the ACPI handle before this
> device is added in driver core, and the binding will be done.

Yeah, we made that change in the ACPI core around 3.8 IIRC.

Thanks,
Rafael
Matthew Garrett July 27, 2013, 6:34 a.m. UTC | #4
On Thu, Jul 25, 2013 at 10:52:29AM -0400, Tejun Heo wrote:
> 
> I like it but am wondering why we weren't doing this before.  Was the
> acpi support added before we made ata objects proper devices?

Yes.
Aaron Lu Aug. 15, 2013, 1:33 a.m. UTC | #5
On 07/25/2013 10:52 PM, Tejun Heo wrote:
> On Thu, Jul 25, 2013 at 01:47:03PM +0800, Aaron Lu wrote:
>> Binding ACPI handle to SCSI device has several drawbacks, namely:
>> 1 During ATA device initialization time, ACPI handle will be needed
>>   while SCSI devices are not created yet. So each time ACPI handle is
>>   needed, instead of retrieving the handle by ACPI_HANDLE macro,
>>   a namespace scan is performed to find the handle for the corresponding
>>   ATA device. This is inefficient, and also expose a restriction on
>>   calling path not holding any lock.
>> 2 The binding to SCSI device tree makes code complex, while at the same
>>   time doesn't bring us any benefit. All ACPI handlings are still done
>>   in ATA module, not in SCSI.
>>
>> Rework the ATA ACPI binding code to bind ACPI handle to ATA transport
>> devices(ATA port and ATA device). The binding needs to be done only once,
>> since the ATA transport devices do not go away with hotplug. And due to
>> this, the flush_work call in hotplug handler for ATA bay is no longer
>> needed.
> 
> I like it but am wondering why we weren't doing this before.  Was the
> acpi support added before we made ata objects proper devices?
 
Hi Tejun,

This has been quiet for some time, may I know your opinion on this?

Thanks,
Aaron
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Tejun Heo Aug. 15, 2013, 3:19 a.m. UTC | #6
Hello, Aaron.

On Thu, Aug 15, 2013 at 09:33:28AM +0800, Aaron Lu wrote:
> This has been quiet for some time, may I know your opinion on this?

Ooh, I like it.  I was waiting for proper patch series w/o RFC tag and
hopefully some test results.

Thanks.
Aaron Lu Aug. 22, 2013, 7:15 a.m. UTC | #7
On 08/15/2013 11:19 AM, Tejun Heo wrote:
> Hello, Aaron.
> 
> On Thu, Aug 15, 2013 at 09:33:28AM +0800, Aaron Lu wrote:
>> This has been quiet for some time, may I know your opinion on this?
> 
> Ooh, I like it. I was waiting for proper patch series w/o RFC tag and
> hopefully some test results.

The patch series has two patches and the first has been taken, shall I
send the 2nd one again without the RFC tag in the patch subject? There
should be no other changes at the moment.

As for the test, I've tested runtime power management related code, but
I don't have access to PMP and ATA bay, so I'll ask some people to help.

Hi Drik,
Can you please test the two patches regarding your ATA bay? Thanks.

Hi Dimitris,
Can you please test the two patches regarding ACPI per device disable
functionality? Thanks.

Hi Dan,
Can you please test the two patches here regarding PMP? Thanks.

Hi Jeff,
It would be good if you can test AMD's ZPODD implementation with the
patches here, thanks.

You can either use the branch here:
https://github.com/aaronlu/linux ata_acpi_rework

Or apply the following two patches on top of Linus' tree or a stable tree:
https://github.com/aaronlu/linux/commit/88ba5478837098b71573bb7d6002be33156ea7e4
https://github.com/aaronlu/linux/commit/98b61e1f52fe5582eb7f18c6d27773bf1da70096

Thanks,
Aaron
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Tejun Heo Aug. 22, 2013, 6:36 p.m. UTC | #8
Hello, Aaron.

On Thu, Aug 22, 2013 at 03:15:16PM +0800, Aaron Lu wrote:
> The patch series has two patches and the first has been taken, shall I
> send the 2nd one again without the RFC tag in the patch subject? There
> should be no other changes at the moment.

Yes, that was a request for you to send an updated patch.  I'm okay
with taking some risk with ACPI support for PMPs.  It most likely
won't matter as long as the code itself doesn't oops anyway, but can
you please describe what's tested and what's not but should be okay in
the patch description?

Thanks.
Dirk Griesbach Aug. 23, 2013, 7:02 a.m. UTC | #9
Hello Aaron,

On Do, Aug 22, 2013 at 03:15:16 +0800, Aaron Lu wrote:
> Hi Drik,
> Can you please test the two patches regarding your ATA bay? Thanks.

No problems with the bay. I applied the patches on top of 3.11-rc6.

Regards,
Dirk
--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Aaron Lu Aug. 23, 2013, 7:08 a.m. UTC | #10
On 08/23/2013 03:02 PM, Dirk Griesbach wrote:
> Hello Aaron,
> 
> On Do, Aug 22, 2013 at 03:15:16 +0800, Aaron Lu wrote:
>> Hi Drik,
>> Can you please test the two patches regarding your ATA bay? Thanks.
> 
> No problems with the bay. I applied the patches on top of 3.11-rc6.

Thanks a lot for the test!

-Aaron

> 
> Regards,
> Dirk
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-ide" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/drivers/ata/libata-acpi.c b/drivers/ata/libata-acpi.c
index cf4e702..2377a32 100644
--- a/drivers/ata/libata-acpi.c
+++ b/drivers/ata/libata-acpi.c
@@ -34,14 +34,6 @@  struct ata_acpi_gtf {
 	u8	tf[REGS_PER_GTF];	/* regs. 0x1f1 - 0x1f7 */
 } __packed;
 
-/*
- *	Helper - belongs in the PCI layer somewhere eventually
- */
-static int is_pci_dev(struct device *dev)
-{
-	return (dev->bus == &pci_bus_type);
-}
-
 static void ata_acpi_clear_gtf(struct ata_device *dev)
 {
 	kfree(dev->gtf_cache);
@@ -49,47 +41,18 @@  static void ata_acpi_clear_gtf(struct ata_device *dev)
 }
 
 /**
- * ata_ap_acpi_handle - provide the acpi_handle for an ata_port
- * @ap: the acpi_handle returned will correspond to this port
- *
- * Returns the acpi_handle for the ACPI namespace object corresponding to
- * the ata_port passed into the function, or NULL if no such object exists
- */
-acpi_handle ata_ap_acpi_handle(struct ata_port *ap)
-{
-	if (ap->flags & ATA_FLAG_ACPI_SATA)
-		return NULL;
-
-	return ap->scsi_host ?
-		DEVICE_ACPI_HANDLE(&ap->scsi_host->shost_gendev) : NULL;
-}
-EXPORT_SYMBOL(ata_ap_acpi_handle);
-
-/**
  * ata_dev_acpi_handle - provide the acpi_handle for an ata_device
- * @dev: the acpi_device returned will correspond to this port
+ * @dev: the acpi_handle returned will correspond to this device
  *
  * Returns the acpi_handle for the ACPI namespace object corresponding to
  * the ata_device passed into the function, or NULL if no such object exists
+ * or ACPI is disabled for this device due to consecutive errors.
  */
 acpi_handle ata_dev_acpi_handle(struct ata_device *dev)
 {
-	acpi_integer adr;
-	struct ata_port *ap = dev->link->ap;
-
-	if (libata_noacpi || dev->flags & ATA_DFLAG_ACPI_DISABLED)
-		return NULL;
-
-	if (ap->flags & ATA_FLAG_ACPI_SATA) {
-		if (!sata_pmp_attached(ap))
-			adr = SATA_ADR(ap->port_no, NO_PORT_MULT);
-		else
-			adr = SATA_ADR(ap->port_no, dev->link->pmp);
-		return acpi_get_child(DEVICE_ACPI_HANDLE(ap->host->dev), adr);
-	} else
-		return acpi_get_child(ata_ap_acpi_handle(ap), dev->devno);
+	return dev->flags & ATA_DFLAG_ACPI_DISABLED ?
+			NULL : ACPI_HANDLE(&dev->tdev);
 }
-EXPORT_SYMBOL(ata_dev_acpi_handle);
 
 /* @ap and @dev are the same as ata_acpi_handle_hotplug() */
 static void ata_acpi_detach_device(struct ata_port *ap, struct ata_device *dev)
@@ -156,10 +119,8 @@  static void ata_acpi_handle_hotplug(struct ata_port *ap, struct ata_device *dev,
 
 	spin_unlock_irqrestore(ap->lock, flags);
 
-	if (wait) {
+	if (wait)
 		ata_port_wait_eh(ap);
-		flush_work(&ap->hotplug_task.work);
-	}
 }
 
 static void ata_acpi_dev_notify_dock(acpi_handle handle, u32 event, void *data)
@@ -216,37 +177,55 @@  static const struct acpi_dock_ops ata_acpi_ap_dock_ops = {
 	.uevent = ata_acpi_ap_uevent,
 };
 
-void ata_acpi_hotplug_init(struct ata_host *host)
+/* bind acpi handle to pata port */
+void ata_acpi_bind_port(struct ata_port *ap)
 {
-	int i;
+	acpi_handle host_handle = ACPI_HANDLE(ap->host->dev);
 
-	for (i = 0; i < host->n_ports; i++) {
-		struct ata_port *ap = host->ports[i];
-		acpi_handle handle;
-		struct ata_device *dev;
+	if (libata_noacpi || ap->flags & ATA_FLAG_ACPI_SATA || !host_handle)
+		return;
 
-		if (!ap)
-			continue;
+	ACPI_HANDLE_SET(&ap->tdev, acpi_get_child(host_handle, ap->port_no));
 
-		handle = ata_ap_acpi_handle(ap);
-		if (handle) {
-			/* we might be on a docking station */
-			register_hotplug_dock_device(handle,
-						     &ata_acpi_ap_dock_ops, ap,
-						     NULL, NULL);
-		}
+	if (ata_acpi_gtm(ap, &ap->__acpi_init_gtm) == 0)
+		ap->pflags |= ATA_PFLAG_INIT_GTM_VALID;
 
-		ata_for_each_dev(dev, &ap->link, ALL) {
-			handle = ata_dev_acpi_handle(dev);
-			if (!handle)
-				continue;
+	/* we might be on a docking station */
+	register_hotplug_dock_device(ACPI_HANDLE(&ap->tdev),
+				     &ata_acpi_ap_dock_ops, ap, NULL, NULL);
+}
 
-			/* we might be on a docking station */
-			register_hotplug_dock_device(handle,
-						     &ata_acpi_dev_dock_ops,
-						     dev, NULL, NULL);
-		}
+void ata_acpi_bind_dev(struct ata_device *dev)
+{
+	struct ata_port *ap = dev->link->ap;
+	acpi_handle port_handle = ACPI_HANDLE(&ap->tdev);
+	acpi_handle host_handle = ACPI_HANDLE(ap->host->dev);
+	acpi_handle parent_handle;
+	u64 adr;
+
+	/*
+	 * For both sata/pata devices, host handle is required.
+	 * For pata device, port handle is also required.
+	 */
+	if (libata_noacpi || !host_handle ||
+			(!(ap->flags & ATA_FLAG_ACPI_SATA) && !port_handle))
+		return;
+
+	if (ap->flags & ATA_FLAG_ACPI_SATA) {
+		if (!sata_pmp_attached(ap))
+			adr = SATA_ADR(ap->port_no, NO_PORT_MULT);
+		else
+			adr = SATA_ADR(ap->port_no, dev->link->pmp);
+		parent_handle = host_handle;
+	} else {
+		adr = dev->devno;
+		parent_handle = port_handle;
 	}
+
+	ACPI_HANDLE_SET(&dev->tdev, acpi_get_child(parent_handle, adr));
+
+	register_hotplug_dock_device(ata_dev_acpi_handle(dev),
+				     &ata_acpi_dev_dock_ops, dev, NULL, NULL);
 }
 
 /**
@@ -270,18 +249,34 @@  void ata_acpi_dissociate(struct ata_host *host)
 		struct ata_port *ap = host->ports[i];
 		const struct ata_acpi_gtm *gtm = ata_acpi_init_gtm(ap);
 
-		if (ata_ap_acpi_handle(ap) && gtm)
+		if (ACPI_HANDLE(&ap->tdev) && gtm)
 			ata_acpi_stm(ap, gtm);
 	}
 }
 
-static int __ata_acpi_gtm(struct ata_port *ap, acpi_handle handle,
-			  struct ata_acpi_gtm *gtm)
+/**
+ * ata_acpi_gtm - execute _GTM
+ * @ap: target ATA port
+ * @gtm: out parameter for _GTM result
+ *
+ * Evaluate _GTM and store the result in @gtm.
+ *
+ * LOCKING:
+ * EH context.
+ *
+ * RETURNS:
+ * 0 on success, -ENOENT if _GTM doesn't exist, -errno on failure.
+ */
+int ata_acpi_gtm(struct ata_port *ap, struct ata_acpi_gtm *gtm)
 {
 	struct acpi_buffer output = { .length = ACPI_ALLOCATE_BUFFER };
 	union acpi_object *out_obj;
 	acpi_status status;
 	int rc = 0;
+	acpi_handle handle = ACPI_HANDLE(&ap->tdev);
+
+	if (!handle)
+		return -EINVAL;
 
 	status = acpi_evaluate_object(handle, "_GTM", NULL, &output);
 
@@ -317,27 +312,6 @@  static int __ata_acpi_gtm(struct ata_port *ap, acpi_handle handle,
 	return rc;
 }
 
-/**
- * ata_acpi_gtm - execute _GTM
- * @ap: target ATA port
- * @gtm: out parameter for _GTM result
- *
- * Evaluate _GTM and store the result in @gtm.
- *
- * LOCKING:
- * EH context.
- *
- * RETURNS:
- * 0 on success, -ENOENT if _GTM doesn't exist, -errno on failure.
- */
-int ata_acpi_gtm(struct ata_port *ap, struct ata_acpi_gtm *gtm)
-{
-	if (ata_ap_acpi_handle(ap))
-		return __ata_acpi_gtm(ap, ata_ap_acpi_handle(ap), gtm);
-	else
-		return -EINVAL;
-}
-
 EXPORT_SYMBOL_GPL(ata_acpi_gtm);
 
 /**
@@ -374,8 +348,8 @@  int ata_acpi_stm(struct ata_port *ap, const struct ata_acpi_gtm *stm)
 	input.count = 3;
 	input.pointer = in_params;
 
-	status = acpi_evaluate_object(ata_ap_acpi_handle(ap), "_STM", &input,
-				      NULL);
+	status = acpi_evaluate_object(ACPI_HANDLE(&ap->tdev), "_STM",
+				      &input, NULL);
 
 	if (status == AE_NOT_FOUND)
 		return -ENOENT;
@@ -850,7 +824,7 @@  void ata_acpi_on_resume(struct ata_port *ap)
 	const struct ata_acpi_gtm *gtm = ata_acpi_init_gtm(ap);
 	struct ata_device *dev;
 
-	if (ata_ap_acpi_handle(ap) && gtm) {
+	if (ACPI_HANDLE(&ap->tdev) && gtm) {
 		/* _GTM valid */
 
 		/* restore timing parameters */
@@ -894,8 +868,7 @@  static int ata_acpi_choose_suspend_state(struct ata_device *dev, bool runtime)
 		d_max_in = ACPI_STATE_D3_HOT;
 
 out:
-	return acpi_pm_device_sleep_state(&dev->sdev->sdev_gendev,
-					  NULL, d_max_in);
+	return acpi_pm_device_sleep_state(&dev->tdev, NULL, d_max_in);
 }
 
 static void sata_acpi_set_state(struct ata_port *ap, pm_message_t state)
@@ -932,7 +905,7 @@  static void pata_acpi_set_state(struct ata_port *ap, pm_message_t state)
 	struct ata_device *dev;
 	acpi_handle port_handle;
 
-	port_handle = ata_ap_acpi_handle(ap);
+	port_handle = ACPI_HANDLE(&ap->tdev);
 	if (!port_handle)
 		return;
 
@@ -1063,109 +1036,16 @@  void ata_acpi_on_disable(struct ata_device *dev)
 	ata_acpi_clear_gtf(dev);
 }
 
-static int compat_pci_ata(struct ata_port *ap)
-{
-	struct device *dev = ap->tdev.parent;
-	struct pci_dev *pdev;
-
-	if (!is_pci_dev(dev))
-		return 0;
-
-	pdev = to_pci_dev(dev);
-
-	if ((pdev->class >> 8) != PCI_CLASS_STORAGE_SATA &&
-	    (pdev->class >> 8) != PCI_CLASS_STORAGE_IDE)
-		return 0;
-
-	return 1;
-}
-
-static int ata_acpi_bind_host(struct ata_port *ap, acpi_handle *handle)
-{
-	if (libata_noacpi || ap->flags & ATA_FLAG_ACPI_SATA)
-		return -ENODEV;
-
-	*handle = acpi_get_child(DEVICE_ACPI_HANDLE(ap->tdev.parent),
-			ap->port_no);
-
-	if (!*handle)
-		return -ENODEV;
-
-	if (__ata_acpi_gtm(ap, *handle, &ap->__acpi_init_gtm) == 0)
-		ap->pflags |= ATA_PFLAG_INIT_GTM_VALID;
-
-	return 0;
-}
-
-static int ata_acpi_bind_device(struct ata_port *ap, struct scsi_device *sdev,
-				acpi_handle *handle)
-{
-	struct ata_device *ata_dev;
-
-	if (ap->flags & ATA_FLAG_ACPI_SATA) {
-		if (!sata_pmp_attached(ap))
-			ata_dev = &ap->link.device[sdev->id];
-		else
-			ata_dev = &ap->pmp_link[sdev->channel].device[sdev->id];
-	}
-	else {
-		ata_dev = &ap->link.device[sdev->id];
-	}
-
-	*handle = ata_dev_acpi_handle(ata_dev);
-
-	if (!*handle)
-		return -ENODEV;
-
-	return 0;
-}
-
-static int is_ata_port(const struct device *dev)
-{
-	return dev->type == &ata_port_type;
-}
-
-static struct ata_port *dev_to_ata_port(struct device *dev)
-{
-	while (!is_ata_port(dev)) {
-		if (!dev->parent)
-			return NULL;
-		dev = dev->parent;
-	}
-	return to_ata_port(dev);
-}
-
-static int ata_acpi_find_device(struct device *dev, acpi_handle *handle)
-{
-	struct ata_port *ap = dev_to_ata_port(dev);
-
-	if (!ap)
-		return -ENODEV;
-
-	if (!compat_pci_ata(ap))
-		return -ENODEV;
-
-	if (scsi_is_host_device(dev))
-		return ata_acpi_bind_host(ap, handle);
-	else if (scsi_is_sdev_device(dev)) {
-		struct scsi_device *sdev = to_scsi_device(dev);
-
-		return ata_acpi_bind_device(ap, sdev, handle);
-	} else
-		return -ENODEV;
-}
-
-static struct acpi_bus_type ata_acpi_bus = {
-	.name = "ATA",
-	.find_device = ata_acpi_find_device,
-};
-
-int ata_acpi_register(void)
+void ata_scsi_acpi_bind(struct ata_device *dev)
 {
-	return scsi_register_acpi_bus_type(&ata_acpi_bus);
+	acpi_handle handle = ata_dev_acpi_handle(dev);
+	if (handle)
+		acpi_dev_pm_add_dependent(handle, &dev->sdev->sdev_gendev);
 }
 
-void ata_acpi_unregister(void)
+void ata_scsi_acpi_unbind(struct ata_device *dev)
 {
-	scsi_unregister_acpi_bus_type(&ata_acpi_bus);
+	acpi_handle handle = ata_dev_acpi_handle(dev);
+	if (handle)
+		acpi_dev_pm_remove_dependent(handle, &dev->sdev->sdev_gendev);
 }
diff --git a/drivers/ata/libata-core.c b/drivers/ata/libata-core.c
index c24354d..39e78e4 100644
--- a/drivers/ata/libata-core.c
+++ b/drivers/ata/libata-core.c
@@ -6150,8 +6150,6 @@  int ata_host_register(struct ata_host *host, struct scsi_host_template *sht)
 	if (rc)
 		goto err_tadd;
 
-	ata_acpi_hotplug_init(host);
-
 	/* set cable, sata_spd_limit and report */
 	for (i = 0; i < host->n_ports; i++) {
 		struct ata_port *ap = host->ports[i];
@@ -6632,8 +6630,6 @@  static int __init ata_init(void)
 
 	ata_parse_force_param();
 
-	ata_acpi_register();
-
 	rc = ata_sff_init();
 	if (rc) {
 		kfree(ata_force_tbl);
@@ -6660,7 +6656,6 @@  static void __exit ata_exit(void)
 	ata_release_transport(ata_scsi_transport_template);
 	libata_transport_exit();
 	ata_sff_exit();
-	ata_acpi_unregister();
 	kfree(ata_force_tbl);
 }
 
diff --git a/drivers/ata/libata-scsi.c b/drivers/ata/libata-scsi.c
index b1e880a..f097d66 100644
--- a/drivers/ata/libata-scsi.c
+++ b/drivers/ata/libata-scsi.c
@@ -49,7 +49,6 @@ 
 #include <linux/hdreg.h>
 #include <linux/uaccess.h>
 #include <linux/suspend.h>
-#include <linux/pm_qos.h>
 #include <asm/unaligned.h>
 
 #include "libata.h"
@@ -3667,9 +3666,7 @@  void ata_scsi_scan_host(struct ata_port *ap, int sync)
 			if (!IS_ERR(sdev)) {
 				dev->sdev = sdev;
 				scsi_device_put(sdev);
-				if (zpodd_dev_enabled(dev))
-					dev_pm_qos_expose_flags(
-							&sdev->sdev_gendev, 0);
+				ata_scsi_acpi_bind(dev);
 			} else {
 				dev->sdev = NULL;
 			}
@@ -3757,6 +3754,8 @@  static void ata_scsi_remove_dev(struct ata_device *dev)
 	struct scsi_device *sdev;
 	unsigned long flags;
 
+	ata_scsi_acpi_unbind(dev);
+
 	/* Alas, we need to grab scan_mutex to ensure SCSI device
 	 * state doesn't change underneath us and thus
 	 * scsi_device_get() always succeeds.  The mutex locking can
@@ -3766,9 +3765,6 @@  static void ata_scsi_remove_dev(struct ata_device *dev)
 	mutex_lock(&ap->scsi_host->scan_mutex);
 	spin_lock_irqsave(ap->lock, flags);
 
-	if (zpodd_dev_enabled(dev))
-		zpodd_exit(dev);
-
 	/* clearing dev->sdev is protected by host lock */
 	sdev = dev->sdev;
 	dev->sdev = NULL;
@@ -3818,6 +3814,9 @@  static void ata_scsi_handle_link_detach(struct ata_link *link)
 		dev->flags &= ~ATA_DFLAG_DETACHED;
 		spin_unlock_irqrestore(ap->lock, flags);
 
+		if (zpodd_dev_enabled(dev))
+			zpodd_exit(dev);
+
 		ata_scsi_remove_dev(dev);
 	}
 }
diff --git a/drivers/ata/libata-transport.c b/drivers/ata/libata-transport.c
index 077a856..150a917 100644
--- a/drivers/ata/libata-transport.c
+++ b/drivers/ata/libata-transport.c
@@ -287,6 +287,7 @@  int ata_tport_add(struct device *parent,
 	dev->release = ata_tport_release;
 	dev_set_name(dev, "ata%d", ap->print_id);
 	transport_setup_device(dev);
+	ata_acpi_bind_port(ap);
 	error = device_add(dev);
 	if (error) {
 		goto tport_err;
@@ -644,6 +645,7 @@  static int ata_tdev_add(struct ata_device *ata_dev)
 		dev_set_name(dev, "dev%d.%d.0", ap->print_id, link->pmp);
 
 	transport_setup_device(dev);
+	ata_acpi_bind_dev(ata_dev);
 	error = device_add(dev);
 	if (error) {
 		ata_tdev_free(ata_dev);
diff --git a/drivers/ata/libata-zpodd.c b/drivers/ata/libata-zpodd.c
index cd8daf4..68f9e32 100644
--- a/drivers/ata/libata-zpodd.c
+++ b/drivers/ata/libata-zpodd.c
@@ -2,6 +2,7 @@ 
 #include <linux/cdrom.h>
 #include <linux/pm_runtime.h>
 #include <linux/module.h>
+#include <linux/pm_qos.h>
 #include <scsi/scsi_device.h>
 
 #include "libata.h"
@@ -190,8 +191,8 @@  void zpodd_enable_run_wake(struct ata_device *dev)
 	sdev_disable_disk_events(dev->sdev);
 
 	zpodd->powered_off = true;
-	device_set_run_wake(&dev->sdev->sdev_gendev, true);
-	acpi_pm_device_run_wake(&dev->sdev->sdev_gendev, true);
+	device_set_run_wake(&dev->tdev, true);
+	acpi_pm_device_run_wake(&dev->tdev, true);
 }
 
 /* Disable runtime wake capability if it is enabled */
@@ -200,8 +201,8 @@  void zpodd_disable_run_wake(struct ata_device *dev)
 	struct zpodd *zpodd = dev->zpodd;
 
 	if (zpodd->powered_off) {
-		acpi_pm_device_run_wake(&dev->sdev->sdev_gendev, false);
-		device_set_run_wake(&dev->sdev->sdev_gendev, false);
+		acpi_pm_device_run_wake(&dev->tdev, false);
+		device_set_run_wake(&dev->tdev, false);
 	}
 }
 
@@ -262,7 +263,7 @@  static void ata_acpi_add_pm_notifier(struct ata_device *dev)
 
 static void ata_acpi_remove_pm_notifier(struct ata_device *dev)
 {
-	acpi_handle handle = DEVICE_ACPI_HANDLE(&dev->sdev->sdev_gendev);
+	acpi_handle handle = ata_dev_acpi_handle(dev);
 	acpi_remove_notify_handler(handle, ACPI_SYSTEM_NOTIFY, zpodd_wake_dev);
 }
 
@@ -290,6 +291,7 @@  void zpodd_init(struct ata_device *dev)
 	ata_acpi_add_pm_notifier(dev);
 	zpodd->dev = dev;
 	dev->zpodd = zpodd;
+	dev_pm_qos_expose_flags(&dev->tdev, 0);
 }
 
 void zpodd_exit(struct ata_device *dev)
diff --git a/drivers/ata/libata.h b/drivers/ata/libata.h
index cc8a6e8..3e9cc95 100644
--- a/drivers/ata/libata.h
+++ b/drivers/ata/libata.h
@@ -118,9 +118,11 @@  extern void ata_acpi_on_resume(struct ata_port *ap);
 extern int ata_acpi_on_devcfg(struct ata_device *dev);
 extern void ata_acpi_on_disable(struct ata_device *dev);
 extern void ata_acpi_set_state(struct ata_port *ap, pm_message_t state);
-extern int ata_acpi_register(void);
-extern void ata_acpi_unregister(void);
-extern void ata_acpi_hotplug_init(struct ata_host *host);
+extern void ata_acpi_bind_port(struct ata_port *ap);
+extern void ata_acpi_bind_dev(struct ata_device *dev);
+extern acpi_handle ata_dev_acpi_handle(struct ata_device *dev);
+extern void ata_scsi_acpi_bind(struct ata_device *dev);
+extern void ata_scsi_acpi_unbind(struct ata_device *dev);
 #else
 static inline void ata_acpi_dissociate(struct ata_host *host) { }
 static inline int ata_acpi_on_suspend(struct ata_port *ap) { return 0; }
@@ -129,9 +131,11 @@  static inline int ata_acpi_on_devcfg(struct ata_device *dev) { return 0; }
 static inline void ata_acpi_on_disable(struct ata_device *dev) { }
 static inline void ata_acpi_set_state(struct ata_port *ap,
 				      pm_message_t state) { }
-static inline int ata_acpi_register(void) { return 0; }
-static inline void ata_acpi_unregister(void) { }
-static inline void ata_acpi_hotplug_init(struct ata_host *host) {}
+static inline void ata_acpi_bind_port(struct ata_port *ap) {}
+static inline void ata_acpi_bind_dev(struct ata_device *dev) {}
+static inline acpi_handle ata_dev_acpi_handle(struct ata_device *dev) { return NULL; }
+static inline void ata_scsi_acpi_bind(struct ata_device *dev) {}
+static inline void ata_scsi_acpi_unbind(struct ata_device *dev) {}
 #endif
 
 /* libata-scsi.c */
diff --git a/drivers/ata/pata_acpi.c b/drivers/ata/pata_acpi.c
index 09723b7..73212c9 100644
--- a/drivers/ata/pata_acpi.c
+++ b/drivers/ata/pata_acpi.c
@@ -39,7 +39,7 @@  static int pacpi_pre_reset(struct ata_link *link, unsigned long deadline)
 {
 	struct ata_port *ap = link->ap;
 	struct pata_acpi *acpi = ap->private_data;
-	if (ata_ap_acpi_handle(ap) == NULL || ata_acpi_gtm(ap, &acpi->gtm) < 0)
+	if (ACPI_HANDLE(&ap->tdev) == NULL || ata_acpi_gtm(ap, &acpi->gtm) < 0)
 		return -ENODEV;
 
 	return ata_sff_prereset(link, deadline);
@@ -195,7 +195,7 @@  static int pacpi_port_start(struct ata_port *ap)
 	struct pci_dev *pdev = to_pci_dev(ap->host->dev);
 	struct pata_acpi *acpi;
 
-	if (ata_ap_acpi_handle(ap) == NULL)
+	if (ACPI_HANDLE(&ap->tdev) == NULL)
 		return -ENODEV;
 
 	acpi = ap->private_data = devm_kzalloc(&pdev->dev, sizeof(struct pata_acpi), GFP_KERNEL);
diff --git a/drivers/scsi/scsi_lib.c b/drivers/scsi/scsi_lib.c
index 124392f..f0ddc7d 100644
--- a/drivers/scsi/scsi_lib.c
+++ b/drivers/scsi/scsi_lib.c
@@ -68,28 +68,6 @@  static struct scsi_host_sg_pool scsi_sg_pools[] = {
 
 struct kmem_cache *scsi_sdb_cache;
 
-#ifdef CONFIG_ACPI
-#include <acpi/acpi_bus.h>
-
-static bool acpi_scsi_bus_match(struct device *dev)
-{
-	return dev->bus == &scsi_bus_type;
-}
-
-int scsi_register_acpi_bus_type(struct acpi_bus_type *bus)
-{
-        bus->match = acpi_scsi_bus_match;
-        return register_acpi_bus_type(bus);
-}
-EXPORT_SYMBOL_GPL(scsi_register_acpi_bus_type);
-
-void scsi_unregister_acpi_bus_type(struct acpi_bus_type *bus)
-{
-	unregister_acpi_bus_type(bus);
-}
-EXPORT_SYMBOL_GPL(scsi_unregister_acpi_bus_type);
-#endif
-
 /*
  * When to reinvoke queueing after a resource shortage. It's 3 msecs to
  * not change behaviour from the previous unplug mechanism, experimentation
diff --git a/include/linux/libata.h b/include/linux/libata.h
index 283d66b..3198617 100644
--- a/include/linux/libata.h
+++ b/include/linux/libata.h
@@ -1244,8 +1244,6 @@  int ata_acpi_stm(struct ata_port *ap, const struct ata_acpi_gtm *stm);
 int ata_acpi_gtm(struct ata_port *ap, struct ata_acpi_gtm *stm);
 unsigned long ata_acpi_gtm_xfermask(struct ata_device *dev,
 				    const struct ata_acpi_gtm *gtm);
-acpi_handle ata_ap_acpi_handle(struct ata_port *ap);
-acpi_handle ata_dev_acpi_handle(struct ata_device *dev);
 int ata_acpi_cbl_80wire(struct ata_port *ap, const struct ata_acpi_gtm *gtm);
 #else
 static inline const struct ata_acpi_gtm *ata_acpi_init_gtm(struct ata_port *ap)