Skip to contentSkip to conversation

Add PGR response PASID requirement check in Intel IOMMU.

PATCH Superseded [email protected] 4 participants 7 messages Feb 11, 2019 – Feb 13, 2019

Subsystems (2)

v2 of 3 Superseded Latest is v3 →

lore
Share & tools
lore
#0 lore raw

v2 cover Add PGR response PASID requirement check in Intel IOMMU.

From: Kuppuswamy Sathyanarayanan <[email protected]>

Intel IOMMU responds automatically when receiving page-requests from a PCIe endpoint and the page-request queue is full and it cannot accept any more page-requests. When it auto-responds to page-requests with a success to the endpoint, it automatically responds with the PASID if the page-request had a PASID in the incoming request. IOMMU doesn't actually have any place to check device capabilities (like whether the device expects PASID in PGR response or not) before sending the response message. Due to this restriction Intel IOMMU driver only enables PASID, if the endpoint is compliant to Intel IOMMU's.

Changes since v1:
* Changed interface name to pci_prg_resp_pasid_required().
* Update comment header format.

Kuppuswamy Sathyanarayanan (2):
PCI/ATS: Add pci_prg_resp_pasid_required() interface. iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

 drivers/iommu/intel-iommu.c   |  3 ++-
 drivers/pci/ats.c             | 31 +++++++++++++++++++++++++++++++
 include/linux/pci-ats.h       |  5 +++++
 include/uapi/linux/pci_regs.h |  1 +
 4 files changed, 39 insertions(+), 1 deletion(-)
Signature · 1 line
2.20.1
#1 lore raw

PATCH v2 1/2 PCI/ATS: Add pci_prg_resp_pasid_required() interface.

From: Kuppuswamy Sathyanarayanan <[email protected]>

Return the PRG Response PASID Required bit in the Page Request Status Register.

As per PCIe spec r4.0, sec 10.5.2.3, if this bit is Set then the device expects a PASID TLP Prefix on PRG Response Messages when the corresponding Page Requests had a PASID TLP Prefix. If Clear, the device does not expect PASID TLP Prefixes on any PRG Response Message, and the device behavior is undefined if this bit is Clear and the device receives a PRG Response Message with a PASID TLP Prefix. Also the device behavior is undefined in the this bit is Set and the device receives a PRG Response Message with no PASID TLP Prefix when the corresponding Page Requests had a PASID TLP Prefix.

This function will be used by drivers like IOMMU, if it is required to check the status of the PRG Response PASID Required bit before enabling the PASID support of the device.

---
 drivers/pci/ats.c             | 31 +++++++++++++++++++++++++++++++
 include/linux/pci-ats.h       |  5 +++++
 include/uapi/linux/pci_regs.h |  1 +
 3 files changed, 37 insertions(+)
drivers/pci/ats.c +31 -0
index 5b78f3b1b918..f843cd846dff 100644
@@ -368,6 +368,37 @@ int pci_pasid_features(struct pci_dev *pdev)
 }
 EXPORT_SYMBOL_GPL(pci_pasid_features);
 
+/**
+ * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit
+ *				 status.
+ * @pdev: PCI device structure
+ *
+ * Returns 1 if PASID is required in PRG Response message, 0 otherwise.
+ *
+ * Even though the PRG response PASID status is read from PRI status
+ * register, since this API will mainly be used by PASID users, this
+ * function is defined within #ifdef CONFIG_PCI_PASID instead of
+ * CONFIG_PCI_PRI.
+ *
+ */
+int pci_prg_resp_pasid_required(struct pci_dev *pdev)
+{
+	u16 status;
+	int pos;
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+	if (!pos)
+		return 0;
+
+	pci_read_config_word(pdev, pos + PCI_PRI_STATUS, &status);
+
+	if (status & PCI_PRI_STATUS_PASID)
+		return 1;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);
+
 #define PASID_NUMBER_SHIFT	8
 #define PASID_NUMBER_MASK	(0x1f << PASID_NUMBER_SHIFT)
 /**
include/linux/pci-ats.h +5 -0
index 7c4b8e27268c..facfd6a18fe1 100644
@@ -40,6 +40,7 @@ void pci_disable_pasid(struct pci_dev *pdev);
 void pci_restore_pasid_state(struct pci_dev *pdev);
 int pci_pasid_features(struct pci_dev *pdev);
 int pci_max_pasids(struct pci_dev *pdev);
+int pci_prg_resp_pasid_required(struct pci_dev *pdev);
 
 #else  /* CONFIG_PCI_PASID */
 
@@ -66,6 +67,10 @@ static inline int pci_max_pasids(struct pci_dev *pdev)
 	return -EINVAL;
 }
 
+static int pci_prg_resp_pasid_required(struct pci_dev *pdev)
+{
+	return 0;
+}
 #endif /* CONFIG_PCI_PASID */
 
 
include/uapi/linux/pci_regs.h +1 -0
index e1e9888c85e6..898be572b010 100644
@@ -880,6 +880,7 @@
 #define  PCI_PRI_STATUS_RF	0x001	/* Response Failure */
 #define  PCI_PRI_STATUS_UPRGI	0x002	/* Unexpected PRG index */
 #define  PCI_PRI_STATUS_STOPPED	0x100	/* PRI Stopped */
+#define  PCI_PRI_STATUS_PASID	0x8000	/* PRG Response PASID Required */
 #define PCI_PRI_MAX_REQ		0x08	/* PRI max reqs supported */
 #define PCI_PRI_ALLOC_REQ	0x0c	/* PRI max reqs allowed */
 #define PCI_EXT_CAP_PRI_SIZEOF	16
Signature · 1 line
2.20.1
#3 lore raw

PATCH v2 2/2 iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

In reply to [email protected]

From: Kuppuswamy Sathyanarayanan <[email protected]>

In Intel IOMMU, if the Page Request Queue (PRQ) is full, it will automatically respond to the device with a success message as a keep alive. And when sending the success message, IOMMU will include PASID in the Response Message when the Page Request has a PASID in Request Message and It does not check against the PRG Response PASID requirement of the device before sending the response. Also, If the device receives the PRG response with PASID when its not expecting it then the device behavior is undefined. So enable PASID support only if device expects PASID in PRG response message.

---
 drivers/iommu/intel-iommu.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/iommu/intel-iommu.c +2 -1
index 1457f931218e..af2e4a011787 100644
@@ -1399,7 +1399,8 @@ static void iommu_enable_dev_iotlb(struct device_domain_info *info)
 	   undefined. So always enable PASID support on devices which
 	   have it, even if we can't yet know if we're ever going to
 	   use it. */
-	if (info->pasid_supported && !pci_enable_pasid(pdev, info->pasid_supported & ~1))
+	if (info->pasid_supported && pci_prg_resp_pasid_required(pdev) &&
+	    !pci_enable_pasid(pdev, info->pasid_supported & ~1))
 		info->pasid_enabled = 1;
 
 	if (info->pri_supported && !pci_reset_pri(pdev) && !pci_enable_pri(pdev, 32))
Signature · 1 line
2.20.1
#4 lore raw

RE: [PATCH v2 2/2] iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

Quoted 49 lines

From: [email protected] [mailto:iommu- [email protected]] On Behalf Of [email protected]
Sent: Tuesday, February 12, 2019 5:51 AM
To: [email protected]; [email protected]; [email protected]
Cc: Raj, Ashok <[email protected]>; [email protected]; linux- [email protected]; Busch, Keith <[email protected]>; [email protected]; Pan, Jacob jun <[email protected]>
Subject: [PATCH v2 2/2] iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

From: Kuppuswamy Sathyanarayanan
<[email protected]>

In Intel IOMMU, if the Page Request Queue (PRQ) is full, it will automatically respond to the device with a success message as a keep alive. And when sending the success message, IOMMU will include PASID in the Response Message when the Page Request has a PASID in Request Message and It does not check against the PRG Response PASID requirement
of the device before sending the response. Also, If the device receives the PRG response with PASID when its not expecting it then the device behavior is undefined. So enable PASID support only if device expects PASID in PRG response message.

Cc Ashok Raj <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Keith Busch <[email protected]>
Signed-off-by Kuppuswamy Sathyanarayanan

<[email protected]>

---
 drivers/iommu/intel-iommu.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/iommu/intel-iommu.c +1 -1
index 1457f931218e..af2e4a011787 100644
@@ -1399,7 +1399,8 @@ static void iommu_enable_dev_iotlb(struct
 device_domain_info *info)
 	   undefined. So always enable PASID support on devices which
 	   have it, even if we can't yet know if we're ever going to
 	   use it. */
-	if (info->pasid_supported && !pci_enable_pasid(pdev, info-
 >pasid_supported & ~1))
+	if (info->pasid_supported && pci_prg_resp_pasid_required(pdev)

&&

drivers/iommu/intel-iommu.c static void iommu_enable_dev_iotlb(struct device_domain_info *info) · line 1404

Quoted 2 lines
drivers/iommu/intel-iommu.c +1 -0
+	    !pci_enable_pasid(pdev, info->pasid_supported & ~1))
 		info->pasid_enabled = 1;

Above logic looks problematic. As Dave commented in another thread, PRI and PASID are orthogonal capabilities. Especially with introduction of VT-d scalable mode, PASID will be used alone even w/o PRI...

Why not doing the check when PRI is actually enabled? At that point you can fail the request if above condition is false.

Quoted 10 lines

if (info->pri_supported && !pci_reset_pri(pdev)
&& !pci_enable_pri(pdev, 32))
--
2.20.1

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

#5 lore raw

Re: [PATCH v2 2/2] iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

Quoted 21 lines · Kevin
Quoted 14 lines
drivers/iommu/intel-iommu.c +1 -1
index 1457f931218e..af2e4a011787 100644
@@ -1399,7 +1399,8 @@ static void iommu_enable_dev_iotlb(struct
 device_domain_info *info)
 	   undefined. So always enable PASID support on devices which
 	   have it, even if we can't yet know if we're ever going to
 	   use it. */
-	if (info->pasid_supported && !pci_enable_pasid(pdev, info-
 >pasid_supported & ~1))
+	if (info->pasid_supported && pci_prg_resp_pasid_required(pdev)

&&

drivers/iommu/intel-iommu.c static void iommu_enable_dev_iotlb(struct device_domain_info *info) · line 1404

Quoted 2 lines
drivers/iommu/intel-iommu.c +1 -0
+	    !pci_enable_pasid(pdev, info->pasid_supported & ~1))
 		info->pasid_enabled = 1;

Above logic looks problematic. As Dave commented in another thread, PRI and PASID are orthogonal capabilities. Especially with introduction of VT-d scalable mode, PASID will be used alone even w/o PRI...

Quoted 3 lines

Why not doing the check when PRI is actually enabled? At that point you can fail the request if above condition is false.

That makes sense.

#6 lore raw

Re: [PATCH v2 2/2] iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

In reply to Tian, Kevin
Quoted 55 lines · Kevin
Quoted 49 lines

From: [email protected] [mailto:iommu- [email protected]] On Behalf Of [email protected]
Sent: Tuesday, February 12, 2019 5:51 AM
To: [email protected]; [email protected]; [email protected]
Cc: Raj, Ashok <[email protected]>; [email protected]; linux- [email protected]; Busch, Keith <[email protected]>; [email protected]; Pan, Jacob jun <[email protected]>
Subject: [PATCH v2 2/2] iommu/vt-d: Enable PASID only if device expects PASID in PRG Response.

From: Kuppuswamy Sathyanarayanan
<[email protected]>

In Intel IOMMU, if the Page Request Queue (PRQ) is full, it will automatically respond to the device with a success message as a keep alive. And when sending the success message, IOMMU will include PASID in the Response Message when the Page Request has a PASID in Request Message and It does not check against the PRG Response PASID requirement
of the device before sending the response. Also, If the device receives the PRG response with PASID when its not expecting it then the device behavior is undefined. So enable PASID support only if device expects PASID in PRG response message.

Cc Ashok Raj <[email protected]>
Cc Jacob Pan <[email protected]>
Cc Keith Busch <[email protected]>
Signed-off-by Kuppuswamy Sathyanarayanan

<[email protected]>

---
  drivers/iommu/intel-iommu.c | 3 ++-
  1 file changed, 2 insertions(+), 1 deletion(-)
drivers/iommu/intel-iommu.c +1 -1
index 1457f931218e..af2e4a011787 100644
@@ -1399,7 +1399,8 @@ static void iommu_enable_dev_iotlb(struct
 device_domain_info *info)
  	   undefined. So always enable PASID support on devices which
  	   have it, even if we can't yet know if we're ever going to
  	   use it. */
-	if (info->pasid_supported && !pci_enable_pasid(pdev, info-
 > pasid_supported & ~1))
+	if (info->pasid_supported && pci_prg_resp_pasid_required(pdev)

&&

drivers/iommu/intel-iommu.c static void iommu_enable_dev_iotlb(struct device_domain_info *info) · line 1404

Quoted 2 lines
drivers/iommu/intel-iommu.c +1 -0
+	    !pci_enable_pasid(pdev, info->pasid_supported & ~1))
  		info->pasid_enabled = 1;

Above logic looks problematic. As Dave commented in another thread, PRI and PASID are orthogonal capabilities. Especially with introduction of VT-d scalable mode, PASID will be used alone even w/o PRI...

Quoted 2 lines

Why not doing the check when PRI is actually enabled? At that point you can fail the request if above condition is false.

yes, makes sense. I will fix it in next version.

Quoted 10 lines
Quoted 10 lines

if (info->pri_supported && !pci_reset_pri(pdev)
&& !pci_enable_pri(pdev, 32))
--
2.20.1

_______________________________________________
iommu mailing list
[email protected]
https://lists.linuxfoundation.org/mailman/listinfo/iommu

Signature · 2 lines
Sathyanarayanan Kuppuswamy
Linux kernel developer
Acked-by on v2 1/2
#2 lore raw

Re: [PATCH v2 1/2] PCI/ATS: Add pci_prg_resp_pasid_required() interface.

In reply to [email protected]
Quoted 5 lines · On Mon, Feb 11, 2019 at 01:50:31PM -0800…

From: Kuppuswamy Sathyanarayanan <[email protected]>

Return the PRG Response PASID Required bit in the Page Request Status Register.

Quoted 9 lines

As per PCIe spec r4.0, sec 10.5.2.3, if this bit is Set then the device expects a PASID TLP Prefix on PRG Response Messages when the corresponding Page Requests had a PASID TLP Prefix. If Clear, the device does not expect PASID TLP Prefixes on any PRG Response Message, and the device behavior is undefined if this bit is Clear and the device receives a PRG Response Message with a PASID TLP Prefix. Also the device behavior is undefined in the this bit is Set and the device receives a PRG Response Message with no PASID TLP Prefix when the corresponding Page Requests had a PASID TLP Prefix.

s/Set then the device/Set, the device/
s/undefined if this bit is Clear and the device/undefined if the device/ s/is undefined in the this/is undefined if this/

Quoted 9 lines

This function will be used by drivers like IOMMU, if it is required to check the status of the PRG Response PASID Required bit before enabling the PASID support of the device.

With typos (also below) addressed,

Quoted 16 lines
---
 drivers/pci/ats.c             | 31 +++++++++++++++++++++++++++++++
 include/linux/pci-ats.h       |  5 +++++
 include/uapi/linux/pci_regs.h |  1 +
 3 files changed, 37 insertions(+)
drivers/pci/ats.c +2 -0
index 5b78f3b1b918..f843cd846dff 100644
@@ -368,6 +368,37 @@ int pci_pasid_features(struct pci_dev *pdev)
 }
 EXPORT_SYMBOL_GPL(pci_pasid_features);
 
+/**
+ * pci_prg_resp_pasid_required - Return PRG Response PASID Required bit

drivers/pci/ats.c int pci_pasid_features(struct pci_dev *pdev) · line 382

Quoted 10 lines
drivers/pci/ats.c +10 -0
+ *				 status.
+ * @pdev: PCI device structure
+ *
+ * Returns 1 if PASID is required in PRG Response message, 0 otherwise.
+ *
+ * Even though the PRG response PASID status is read from PRI status
+ * register, since this API will mainly be used by PASID users, this
+ * function is defined within #ifdef CONFIG_PCI_PASID instead of
+ * CONFIG_PCI_PRI.
+ *

Remove blank comment line.

Quoted 60 lines
drivers/pci/ats.c +19 -0
+ */
+int pci_prg_resp_pasid_required(struct pci_dev *pdev)
+{
+	u16 status;
+	int pos;
+
+	pos = pci_find_ext_capability(pdev, PCI_EXT_CAP_ID_PRI);
+	if (!pos)
+		return 0;
+
+	pci_read_config_word(pdev, pos + PCI_PRI_STATUS, &status);
+
+	if (status & PCI_PRI_STATUS_PASID)
+		return 1;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(pci_prg_resp_pasid_required);
+
 #define PASID_NUMBER_SHIFT	8
 #define PASID_NUMBER_MASK	(0x1f << PASID_NUMBER_SHIFT)
 /**
include/linux/pci-ats.h +5 -0
index 7c4b8e27268c..facfd6a18fe1 100644
@@ -40,6 +40,7 @@ void pci_disable_pasid(struct pci_dev *pdev);
 void pci_restore_pasid_state(struct pci_dev *pdev);
 int pci_pasid_features(struct pci_dev *pdev);
 int pci_max_pasids(struct pci_dev *pdev);
+int pci_prg_resp_pasid_required(struct pci_dev *pdev);
 
 #else  /* CONFIG_PCI_PASID */
 
@@ -66,6 +67,10 @@ static inline int pci_max_pasids(struct pci_dev *pdev)
 	return -EINVAL;
 }
 
+static int pci_prg_resp_pasid_required(struct pci_dev *pdev)
+{
+	return 0;
+}
 #endif /* CONFIG_PCI_PASID */
 
 
include/uapi/linux/pci_regs.h +1 -0
index e1e9888c85e6..898be572b010 100644
@@ -880,6 +880,7 @@
 #define  PCI_PRI_STATUS_RF	0x001	/* Response Failure */
 #define  PCI_PRI_STATUS_UPRGI	0x002	/* Unexpected PRG index */
 #define  PCI_PRI_STATUS_STOPPED	0x100	/* PRI Stopped */
+#define  PCI_PRI_STATUS_PASID	0x8000	/* PRG Response PASID Required */
 #define PCI_PRI_MAX_REQ		0x08	/* PRI max reqs supported */
 #define PCI_PRI_ALLOC_REQ	0x0c	/* PRI max reqs allowed */
 #define PCI_EXT_CAP_PRI_SIZEOF	16
Signature · 1 line
2.20.1