Skip to contentSkip to conversation

ACPI: scan: Honor certain device identification rules

PATCH In review Rafael J. Wysocki 4 participants 10 messages Oct 26, 2021 – Oct 27, 2021 ACPI

lore
Share & tools
lore
#0 lore raw

cover ACPI: scan: Honor certain device identification rules

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

Thanks!

#1 lore raw

PATCH 1/2 ACPI: scan: Do not add device IDs from _CID if _HID is not valid

From: Rafael J. Wysocki <[email protected]>

Section 6.1.2 of ACPI 6.4 explicitly requires _HID to be present for _CID to be defined, so don't add device IDs from _CID to the device IDs list of a device if _HID is not valid.

---
 drivers/acpi/scan.c |   10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)
drivers/acpi/scan.c +5 -5
Index: linux-pm/drivers/acpi/scan.c
===================================================================
@@ -1338,11 +1338,11 @@ static void acpi_set_pnp_ids(acpi_handle
 		if (info->valid & ACPI_VALID_HID) {
 			acpi_add_id(pnp, info->hardware_id.string);
 			pnp->type.platform_id = 1;
-		}
-		if (info->valid & ACPI_VALID_CID) {
-			cid_list = &info->compatible_id_list;
-			for (i = 0; i < cid_list->count; i++)
-				acpi_add_id(pnp, cid_list->ids[i].string);
+			if (info->valid & ACPI_VALID_CID) {
+				cid_list = &info->compatible_id_list;
+				for (i = 0; i < cid_list->count; i++)
+					acpi_add_id(pnp, cid_list->ids[i].string);
+			}
 		}
 		if (info->valid & ACPI_VALID_ADR) {
 			pnp->bus_address = info->address;
#2 lore raw

PATCH 2/2 ACPI: scan: Do not set type.bus_address if _HID is valid

In reply to Rafael J. Wysocki

From: Rafael J. Wysocki <[email protected]>

According to Section 6.1 of ACPI 6.4, it is invalid to provide both _HID and _ADR for one device at the same time, so modify the code to set pnp.type.bus_address and pnp.bus_address for a device only if it has _ADR, but it doesn't have a valid _HID.

---
 drivers/acpi/scan.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)
drivers/acpi/scan.c +2 -2
Index: linux-pm/drivers/acpi/scan.c
===================================================================
@@ -1343,11 +1343,11 @@ static void acpi_set_pnp_ids(acpi_handle
 				for (i = 0; i < cid_list->count; i++)
 					acpi_add_id(pnp, cid_list->ids[i].string);
 			}
-		}
-		if (info->valid & ACPI_VALID_ADR) {
+		} else if (info->valid & ACPI_VALID_ADR) {
 			pnp->bus_address = info->address;
 			pnp->type.bus_address = 1;
 		}
+
 		if (info->valid & ACPI_VALID_UID)
 			pnp->unique_id = kstrdup(info->unique_id.string,
 							GFP_KERNEL);
Reviewed-by on the v1 series
#3 lore raw
In reply to Rafael J. Wysocki
Quoted 8 lines · Rafael J. Wysocki

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

I understand the motivation, but afraid about consequences on the OEM cheap devices that are not always follow letter of the specification.

As per Intel platforms I would look into Baytrail / Cherrytrail devices for the past (I think Hans may help here a lot) and into Elkhart Lake in the present (for the letter I mostly refer to CSRT + DSDT cooperation to get GP DMA devices enumerated, so I _hope_ DSDT shouldn't have _ADR and _HID together).

Hence, from the code perspective

From the practice I would wait for some tests. I will try to find any new information about latest firmware tables on Elkhart Lake machines.

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#4 lore raw
Quoted 19 lines · Andy Shevchenko
Quoted 8 lines · Rafael J. Wysocki

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

I understand the motivation, but afraid about consequences on the OEM cheap devices that are not always follow letter of the specification.

As per Intel platforms I would look into Baytrail / Cherrytrail devices for the past (I think Hans may help here a lot) and into Elkhart Lake in the present (for the letter I mostly refer to CSRT + DSDT cooperation to get GP DMA devices enumerated, so I _hope_ DSDT shouldn't have _ADR and _HID together).

Quoted 5 lines

Hence, from the code perspective

From the practice I would wait for some tests. I will try to find any new information about latest firmware tables on Elkhart Lake machines.

So, what I see in Elkhart Lake

Case 1 - Sound Wire devices (2 times):

    Name (_ADR, 0x40000000)  // _ADR: Address
    Name (_CID, Package (0x02)  // _CID: Compatible ID
    {
        "PRP00001",
        "PNP0A05" /* Generic Container Device */
    })

Case 2 - GP DMA devices (3 times):

    Name (_ADR, 0x001D0003)  // _ADR: Address
    Name (_HID, "80864BB4")  // _HID: Hardware ID

Case 3 - Camera PMIC devices (5 x 2 (CLPn/DSCn) + 1 (PMIC) times = 11x):

    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "INT3472")  // _HID: Hardware ID
    Name (_CID, "INT3472")  // _CID: Compatible ID

Case 4 - LNK devices (6 times):

Name (_ADR, Zero) // _ADR: Address
...

Name (_UID, One) // _UID: Unique ID
Method (_HID, 0, NotSerialized) // _HID: Hardware ID {
Return (HCID (One))
}

Case 5 - Camera sensors (2 times):

    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "INT34xx")  // _HID: Hardware ID
    Name (_CID, "INT34xx")  // _CID: Compatible ID

I have no idea about cameras or audio devices, but what I'm worrying about is GP DMA. This kind of devices are PCI, but due to Microsoft hack, called CSRT, we have to have a possibility to match DSDT with CSRT ot retrieve the crucial information from the latter while being enumerated by the former.

While it may be against the specification, there is no other way to achieve that as far as I understand (without either breaking things in Linux or getting yellow bang in Windows).

Can you confirm that your change won't modify behaviour for these devices?

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#5 lore raw
Quoted 71 lines · Andy Shevchenko
Quoted 19 lines · Andy Shevchenko
Quoted 8 lines · Rafael J. Wysocki

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

I understand the motivation, but afraid about consequences on the OEM cheap devices that are not always follow letter of the specification.

As per Intel platforms I would look into Baytrail / Cherrytrail devices for the past (I think Hans may help here a lot) and into Elkhart Lake in the present (for the letter I mostly refer to CSRT + DSDT cooperation to get GP DMA devices enumerated, so I _hope_ DSDT shouldn't have _ADR and _HID together).

Quoted 5 lines

Hence, from the code perspective

From the practice I would wait for some tests. I will try to find any new information about latest firmware tables on Elkhart Lake machines.

So, what I see in Elkhart Lake

Case 1 - Sound Wire devices (2 times):

    Name (_ADR, 0x40000000)  // _ADR: Address
    Name (_CID, Package (0x02)  // _CID: Compatible ID
    {
        "PRP00001",
        "PNP0A05" /* Generic Container Device */
    })

Case 2 - GP DMA devices (3 times):

    Name (_ADR, 0x001D0003)  // _ADR: Address
    Name (_HID, "80864BB4")  // _HID: Hardware ID

Case 3 - Camera PMIC devices (5 x 2 (CLPn/DSCn) + 1 (PMIC) times = 11x):

    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "INT3472")  // _HID: Hardware ID
    Name (_CID, "INT3472")  // _CID: Compatible ID

Case 4 - LNK devices (6 times):

Name (_ADR, Zero) // _ADR: Address
...

Name (_UID, One) // _UID: Unique ID
Method (_HID, 0, NotSerialized) // _HID: Hardware ID {
Return (HCID (One))
}

Case 5 - Camera sensors (2 times):

    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "INT34xx")  // _HID: Hardware ID
    Name (_CID, "INT34xx")  // _CID: Compatible ID

I have no idea about cameras or audio devices, but what I'm worrying about is GP DMA. This kind of devices are PCI, but due to Microsoft hack, called CSRT, we have to have a possibility to match DSDT with CSRT ot retrieve the crucial information from the latter while being enumerated by the former.

Quoted 5 lines

While it may be against the specification, there is no other way to achieve that as far as I understand (without either breaking things in Linux or getting yellow bang in Windows).

Can you confirm that your change won't modify behaviour for these devices?

Okay, I have looked into acpi_dma_parse_resource_group() and I don't see that we actually use _HID there. We definitely use _CRS. However, _HID is used in case when device is ACPI-enumerated (drivers/dma/dw/platform.c). Seems like firmware should provide this part runtime (either _HID or _ADR, but not both).

Signature · 2 lines
With Best Regards,
Andy Shevchenko
#6 lore raw
Quoted 77 lines · Andy Shevchenko
Quoted 75 lines · Andy Shevchenko
Quoted 19 lines · Andy Shevchenko
Quoted 8 lines · Rafael J. Wysocki

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

I understand the motivation, but afraid about consequences on the OEM cheap devices that are not always follow letter of the specification.

As per Intel platforms I would look into Baytrail / Cherrytrail devices for the past (I think Hans may help here a lot) and into Elkhart Lake in the present (for the letter I mostly refer to CSRT + DSDT cooperation to get GP DMA devices enumerated, so I _hope_ DSDT shouldn't have _ADR and _HID together).

Quoted 5 lines

Hence, from the code perspective

From the practice I would wait for some tests. I will try to find any new information about latest firmware tables on Elkhart Lake machines.

So, what I see in Elkhart Lake

Case 1 - Sound Wire devices (2 times):

    Name (_ADR, 0x40000000)  // _ADR: Address
    Name (_CID, Package (0x02)  // _CID: Compatible ID
    {
        "PRP00001",
        "PNP0A05" /* Generic Container Device */
    })

Case 2 - GP DMA devices (3 times):

    Name (_ADR, 0x001D0003)  // _ADR: Address
    Name (_HID, "80864BB4")  // _HID: Hardware ID

Case 3 - Camera PMIC devices (5 x 2 (CLPn/DSCn) + 1 (PMIC) times = 11x):

    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "INT3472")  // _HID: Hardware ID
    Name (_CID, "INT3472")  // _CID: Compatible ID

Case 4 - LNK devices (6 times):

Name (_ADR, Zero) // _ADR: Address
...

Name (_UID, One) // _UID: Unique ID
Method (_HID, 0, NotSerialized) // _HID: Hardware ID {
Return (HCID (One))
}

Case 5 - Camera sensors (2 times):

    Name (_ADR, Zero)  // _ADR: Address
    Name (_HID, "INT34xx")  // _HID: Hardware ID
    Name (_CID, "INT34xx")  // _CID: Compatible ID

I have no idea about cameras or audio devices, but what I'm worrying about is GP DMA. This kind of devices are PCI, but due to Microsoft hack, called CSRT, we have to have a possibility to match DSDT with CSRT ot retrieve the crucial information from the latter while being enumerated by the former.

While it may be against the specification, there is no other way to achieve that as far as I understand (without either breaking things in Linux or getting yellow bang in Windows).

Quoted 6 lines
Quoted 1 line

Can you confirm that your change won't modify behaviour for these devices?

Okay, I have looked into acpi_dma_parse_resource_group() and I don't see that we actually use _HID there. We definitely use _CRS. However, _HID is used in case when device is ACPI-enumerated (drivers/dma/dw/platform.c). Seems like firmware should provide this part runtime (either _HID or _ADR, but not both).

Right.

So after patch [2/2] _HID will always be used for the enumeration in such cases which may not be what happens now - and what happens now depends on the ordering in which the objects in question are seen during the namespace walk. If the DMA device is seen before the PCI host bridge, it will be enumerated using the _HID.

#7 lore raw
In reply to Andy Shevchenko
Quoted 24 lines · Andy Shevchenko
Quoted 22 lines · Andy Shevchenko
Quoted 8 lines · Rafael J. Wysocki

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

I understand the motivation, but afraid about consequences on the OEM cheap devices that are not always follow letter of the specification.

As per Intel platforms I would look into Baytrail / Cherrytrail devices for the past (I think Hans may help here a lot) and into Elkhart Lake in the present (for the letter I mostly refer to CSRT + DSDT cooperation to get GP DMA devices enumerated, so I _hope_ DSDT shouldn't have _ADR and _HID together).

Hence, from the code perspective

Quoted 8 lines
Quoted 2 lines

From the practice I would wait for some tests. I will try to find any new information about latest firmware tables on Elkhart Lake machines.

So, what I see in Elkhart Lake

Case 1 - Sound Wire devices (2 times):

    Name (_ADR, 0x40000000)  // _ADR: Address

No _HID, so the IDs returned by the _CID below won't be used.

Quoted 3 lines
    Name (_CID, Package (0x02)  // _CID: Compatible ID
    {
        "PRP00001",

The above device ID is invalid (one 0 too many).

Quoted 1 line

"PNP0A05" /* Generic Container Device */

Without the change this causes a container device to be created, but the only purpose of it may be offline/online (if the child devices support offline/online).

This change should not be functionally relevant.

Quoted 5 lines
    })

Case 2 - GP DMA devices (3 times):

    Name (_ADR, 0x001D0003)  // _ADR: Address

_ADR will be ignored which may not be expected. Is this a PCI device?

Quoted 5 lines
    Name (_HID, "80864BB4")  // _HID: Hardware ID

Case 3 - Camera PMIC devices (5 x 2 (CLPn/DSCn) + 1 (PMIC) times = 11x):

Name (_ADR, Zero) // _ADR: Address

_ADR will be ignored, which shouldn't matter.

Quoted 6 lines
    Name (_HID, "INT3472")  // _HID: Hardware ID
    Name (_CID, "INT3472")  // _CID: Compatible ID

Case 4 - LNK devices (6 times):

Name (_ADR, Zero) // _ADR: Address

Same here.

Quoted 11 lines

...

Name (_UID, One) // _UID: Unique ID
Method (_HID, 0, NotSerialized) // _HID: Hardware ID {
Return (HCID (One))
}

Case 5 - Camera sensors (2 times):

Name (_ADR, Zero) // _ADR: Address

And same here.

Quoted 4 lines
    Name (_HID, "INT34xx")  // _HID: Hardware ID
    Name (_CID, "INT34xx")  // _CID: Compatible ID
Quoted 8 lines

I have no idea about cameras or audio devices, but what I'm worrying about is GP DMA. This kind of devices are PCI, but due to Microsoft hack, called CSRT, we have to have a possibility to match DSDT with CSRT ot retrieve the crucial information from the latter while being enumerated by the former.

While it may be against the specification, there is no other way to achieve that as far as I understand (without either breaking things in Linux or getting yellow bang in Windows).

I'm not really sure why _HID is needed for this. The PCI device ID could be used for CRST matching just fine.

Quoted 1 line

Can you confirm that your change won't modify behaviour for these devices?

Well, the GP DMA thing may be broken by patch [2/2], but does Windows actually use _ADR if _HID is provided?

#8 lore raw
Quoted 4 lines · Rafael J. Wysocki
Quoted 2 lines · Andy Shevchenko
Quoted 1 line · Andy Shevchenko

On Tue, Oct 26, 2021 at 08:51:49PM +0200, Rafael J. Wysocki wrote:

...

Quoted 9 lines
Quoted 8 lines
Quoted 2 lines

From the practice I would wait for some tests. I will try to find any new information about latest firmware tables on Elkhart Lake machines.

So, what I see in Elkhart Lake

Case 1 - Sound Wire devices (2 times):

    Name (_ADR, 0x40000000)  // _ADR: Address
Quoted 7 lines

No _HID, so the IDs returned by the _CID below won't be used.

Quoted 3 lines
    Name (_CID, Package (0x02)  // _CID: Compatible ID
    {
        "PRP00001",

The above device ID is invalid (one 0 too many).

Probably we have to communicate this to EHL program owners internally... I dunno what this means in case of Sound Wire.

Quoted 8 lines
Quoted 1 line

"PNP0A05" /* Generic Container Device */

Without the change this causes a container device to be created, but the only purpose of it may be offline/online (if the child devices support offline/online).

This change should not be functionally relevant.

Quoted 7 lines
Quoted 5 lines
    })

Case 2 - GP DMA devices (3 times):

    Name (_ADR, 0x001D0003)  // _ADR: Address

_ADR will be ignored which may not be expected. Is this a PCI device?

It depends on the BIOS decision at boot time. No idea if it's only one possibility (what I have heard is that device is PCI enumerated, that's why they chose PCI ID in the CSRT, to avoid allocating new IDs for truly ACPI-enumerated device).

But seems another point to discuss internally.

Quoted 43 lines
Quoted 5 lines
    Name (_HID, "80864BB4")  // _HID: Hardware ID

Case 3 - Camera PMIC devices (5 x 2 (CLPn/DSCn) + 1 (PMIC) times = 11x):

Name (_ADR, Zero) // _ADR: Address

_ADR will be ignored, which shouldn't matter.

Quoted 6 lines
    Name (_HID, "INT3472")  // _HID: Hardware ID
    Name (_CID, "INT3472")  // _CID: Compatible ID

Case 4 - LNK devices (6 times):

Name (_ADR, Zero) // _ADR: Address

Same here.

Quoted 11 lines

...

Name (_UID, One) // _UID: Unique ID
Method (_HID, 0, NotSerialized) // _HID: Hardware ID {
Return (HCID (One))
}

Case 5 - Camera sensors (2 times):

Name (_ADR, Zero) // _ADR: Address

And same here.

Quoted 11 lines
    Name (_HID, "INT34xx")  // _HID: Hardware ID
    Name (_CID, "INT34xx")  // _CID: Compatible ID

I have no idea about cameras or audio devices, but what I'm worrying about is GP DMA. This kind of devices are PCI, but due to Microsoft hack, called CSRT, we have to have a possibility to match DSDT with CSRT ot retrieve the crucial information from the latter while being enumerated by the former.

While it may be against the specification, there is no other way to achieve that as far as I understand (without either breaking things in Linux or getting yellow bang in Windows).

Quoted 7 lines

I'm not really sure why _HID is needed for this. The PCI device ID could be used for CRST matching just fine.

Quoted 1 line

Can you confirm that your change won't modify behaviour for these devices?

Well, the GP DMA thing may be broken by patch [2/2], but does Windows actually use _ADR if _HID is provided?

No idea. Let's discuss internally.

P.S. The issue here is that some BIOS versions are floating around and we never know who is using what... :-(

Signature · 2 lines
With Best Regards,
Andy Shevchenko
Reviewed-by on the v1 series
#9 lore raw
In reply to Rafael J. Wysocki

Hi,

Quoted 8 lines · Rafael J. Wysocki

Hi All,

There are some rules in the ACPI spec regarding which device identification objects can be used together etc., but they are not followed by the kernel code.

This series modifies the code to follow the spec more closely (see patch changelogs for details).

Both changes seem sensible to me; and since you make _HID take precedence over _ADR I don't expect this to cause any regressions on BYT / CHT hardware (the other way around will likely be an issue).

So for the series:

Regards,

Hans