[PATCH 1/2 V2] PCI: introduce device assignment interface and refactory related code

[PATCH 1/2 V2] PCI: introduce device assignment interface and refactory related code

[Ethan Zhao]

This patch introduces two new device assignment functions

pci_sriov_assign_device(),
pci_sriov_deassign_device()

along with the existed one

pci_vfs_assigned()

They construct the VFs assignment management interface, used to assign/
deassign device to VM and query the VFs reference counter. instead of
direct manipulation of device flag.

This patch refashioned the related code and make them atomic.

v2: reorder the patchset and make it bisectable and atomic, steps clear
between interface defination and implemenation according to the
suggestion from alex.williamson@redhat.com

Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com>
---
 drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c |   17 ++---------------
 drivers/pci/iov.c                                  |   20 ++++++++++++++++++++
 drivers/xen/xen-pciback/pci_stub.c                 |    4 ++--
 include/linux/pci.h                                |    4 ++++
 virt/kvm/assigned-dev.c                            |    2 +-
 virt/kvm/iommu.c                                   |    4 ++--
 6 files changed, 31 insertions(+), 20 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
index 02c11a7..781040e 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_virtchnl_pf.c
@@ -693,22 +693,9 @@ complete_reset:
 static bool i40e_vfs_are_assigned(struct i40e_pf *pf)
 {
 	struct pci_dev *pdev = pf->pdev;
-	struct pci_dev *vfdev;
-
-	/* loop through all the VFs to see if we own any that are assigned */
-	vfdev = pci_get_device(PCI_VENDOR_ID_INTEL, I40E_DEV_ID_VF , NULL);
-	while (vfdev) {
-		/* if we don't own it we don't care */
-		if (vfdev->is_virtfn && pci_physfn(vfdev) == pdev) {
-			/* if it is assigned we cannot release it */
-			if (vfdev->dev_flags & PCI_DEV_FLAGS_ASSIGNED)
-				return true;
-		}
 
-		vfdev = pci_get_device(PCI_VENDOR_ID_INTEL,
-				       I40E_DEV_ID_VF,
-				       vfdev);
-	}
+	if (pci_vfs_assigned(pdev))
+		return true;
 
 	return false;
 }
diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
index de7a747..c082523 100644
--- a/drivers/pci/iov.c
+++ b/drivers/pci/iov.c
@@ -644,6 +644,26 @@ int pci_vfs_assigned(struct pci_dev *dev)
 EXPORT_SYMBOL_GPL(pci_vfs_assigned);
 
 /**
+ * pci_sriov_assign_device - assign device to VM
+ * @pdev: the device to be assigned.
+ */
+void pci_sriov_assign_device(struct pci_dev *pdev)
+{
+	pdev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
+}
+EXPORT_SYMBOL_GPL(pci_sriov_assign_device);
+
+/**
+ * pci_sriov_deassign_device - deasign device from VM
+ * @pdev: the device to be deassigned.
+ */
+void pci_sriov_deassign_device(struct pci_dev *pdev)
+{
+	pdev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
+}
+EXPORT_SYMBOL_GPL(pci_sriov_deassign_device);
+
+/**
  * pci_sriov_set_totalvfs -- reduce the TotalVFs available
  * @dev: the PCI PF device
  * @numvfs: number that should be used for TotalVFs supported
diff --git a/drivers/xen/xen-pciback/pci_stub.c b/drivers/xen/xen-pciback/pci_stub.c
index 62fcd48..1d7b747 100644
--- a/drivers/xen/xen-pciback/pci_stub.c
+++ b/drivers/xen/xen-pciback/pci_stub.c
@@ -133,7 +133,7 @@ static void pcistub_device_release(struct kref *kref)
 	xen_pcibk_config_free_dyn_fields(dev);
 	xen_pcibk_config_free_dev(dev);
 
-	dev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
+	pci_sriov_deassign_device(dev);
 	pci_dev_put(dev);
 
 	kfree(psdev);
@@ -404,7 +404,7 @@ static int pcistub_init_device(struct pci_dev *dev)
 	dev_dbg(&dev->dev, "reset device\n");
 	xen_pcibk_reset_device(dev);
 
-	dev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
+	pci_sriov_assign_device(dev);
 	return 0;
 
 config_release:
diff --git a/include/linux/pci.h b/include/linux/pci.h
index aab57b4..ddfcceb 100644
--- a/include/linux/pci.h
+++ b/include/linux/pci.h
@@ -1603,6 +1603,8 @@ int pci_num_vf(struct pci_dev *dev);
 int pci_vfs_assigned(struct pci_dev *dev);
 int pci_sriov_set_totalvfs(struct pci_dev *dev, u16 numvfs);
 int pci_sriov_get_totalvfs(struct pci_dev *dev);
+void pci_sriov_assign_device(struct pci_dev *dev);
+void pci_sriov_deassign_device(struct pci_dev *dev);
 #else
 static inline int pci_enable_sriov(struct pci_dev *dev, int nr_virtfn)
 { return -ENODEV; }
@@ -1614,6 +1616,8 @@ static inline int pci_sriov_set_totalvfs(struct pci_dev *dev, u16 numvfs)
 { return 0; }
 static inline int pci_sriov_get_totalvfs(struct pci_dev *dev)
 { return 0; }
+static inline void pci_sriov_assign_device(struct pci_dev *dev) { }
+static inline void pci_sriov_deassign_device(struct pci_dev *dev) { }
 #endif
 
 #if defined(CONFIG_HOTPLUG_PCI) || defined(CONFIG_HOTPLUG_PCI_MODULE)
diff --git a/virt/kvm/assigned-dev.c b/virt/kvm/assigned-dev.c
index bf06577..80b477f 100644
--- a/virt/kvm/assigned-dev.c
+++ b/virt/kvm/assigned-dev.c
@@ -302,7 +302,7 @@ static void kvm_free_assigned_device(struct kvm *kvm,
 	else
 		pci_restore_state(assigned_dev->dev);
 
-	assigned_dev->dev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
+	pci_sriov_deassign_device(assigned_dev->dev);
 
 	pci_release_regions(assigned_dev->dev);
 	pci_disable_device(assigned_dev->dev);
diff --git a/virt/kvm/iommu.c b/virt/kvm/iommu.c
index 0df7d4b..6d3f5ef 100644
--- a/virt/kvm/iommu.c
+++ b/virt/kvm/iommu.c
@@ -194,7 +194,7 @@ int kvm_assign_device(struct kvm *kvm,
 			goto out_unmap;
 	}
 
-	pdev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
+	pci_sriov_assign_device(pdev);
 
 	dev_info(&pdev->dev, "kvm assign device\n");
 
@@ -220,7 +220,7 @@ int kvm_deassign_device(struct kvm *kvm,
 
 	iommu_detach_device(domain, &pdev->dev);
 
-	pdev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
+	pci_sriov_deassign_device(pdev);
 
 	dev_info(&pdev->dev, "kvm deassign device\n");
 
-- 
1.7.1


------------------------------------------------------------------------------
Open source business process management suite built on Java and Eclipse
Turn processes into business applications with Bonita BPM Community Edition
Quickly connect people, data, and systems into organized workflows
Winner of BOSSIE, CODIE, OW2 and Gartner awards
http://p.sf.net/sfu/Bonitasoft
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel&#174; Ethernet, visit http://communities.intel.com/community/wired

[PATCH 2/2 V2] PCI: implement VFs assignment reference counter

[Ethan Zhao]

Current implementation of helper function pci_vfs_assigned() is a
little complex, to get sum of VFs that assigned to VM, access low
level configuration space register and then loop in traversing
device tree.

This patch introduces an atomic reference counter for VFs those
were assigned to VM in struct pci_sriov. and simplify the code in
pci_vfs_assigned().

v2: reorder the patchset according to the suggestion from
alex.williamson@redhat.com

Signed-off-by: Ethan Zhao <ethan.zhao@oracle.com>
---
 drivers/pci/iov.c |   45 +++++++++++++++++----------------------------
 drivers/pci/pci.h |    1 +
 2 files changed, 18 insertions(+), 28 deletions(-)

diff --git a/drivers/pci/iov.c b/drivers/pci/iov.c
index c082523..5478a0c 100644
--- a/drivers/pci/iov.c
+++ b/drivers/pci/iov.c
@@ -382,6 +382,7 @@ found:
 	iov->nres = nres;
 	iov->ctrl = ctrl;
 	iov->total_VFs = total;
+	atomic_set(&iov->VFs_assigned_cnt, 0);
 	iov->offset = offset;
 	iov->stride = stride;
 	iov->pgsz = pgsz;
@@ -603,43 +604,21 @@ int pci_num_vf(struct pci_dev *dev)
 EXPORT_SYMBOL_GPL(pci_num_vf);
 
 /**
- * pci_vfs_assigned - returns number of VFs are assigned to a guest
- * @dev: the PCI device
+ * pci_vfs_assigned - returns number of VFs are assigned to VM
+ * @dev: the physical PCI device that contains the VFs.
  *
- * Returns number of VFs belonging to this device that are assigned to a guest.
+ * Returns number of VFs belonging to this device that are assigned to VM.
  * If device is not a physical function returns 0.
  */
 int pci_vfs_assigned(struct pci_dev *dev)
 {
-	struct pci_dev *vfdev;
-	unsigned int vfs_assigned = 0;
-	unsigned short dev_id;
-
 	/* only search if we are a PF */
 	if (!dev->is_physfn)
 		return 0;
 
-	/*
-	 * determine the device ID for the VFs, the vendor ID will be the
-	 * same as the PF so there is no need to check for that one
-	 */
-	pci_read_config_word(dev, dev->sriov->pos + PCI_SRIOV_VF_DID, &dev_id);
-
-	/* loop through all the VFs to see if we own any that are assigned */
-	vfdev = pci_get_device(dev->vendor, dev_id, NULL);
-	while (vfdev) {
-		/*
-		 * It is considered assigned if it is a virtual function with
-		 * our dev as the physical function and the assigned bit is set
-		 */
-		if (vfdev->is_virtfn && (vfdev->physfn == dev) &&
-		    (vfdev->dev_flags & PCI_DEV_FLAGS_ASSIGNED))
-			vfs_assigned++;
-
-		vfdev = pci_get_device(dev->vendor, dev_id, vfdev);
-	}
-
-	return vfs_assigned;
+	if (dev->sriov)
+		return atomic_read(&dev->sriov->VFs_assigned_cnt);
+	return 0;
 }
 EXPORT_SYMBOL_GPL(pci_vfs_assigned);
 
@@ -650,6 +629,11 @@ EXPORT_SYMBOL_GPL(pci_vfs_assigned);
 void pci_sriov_assign_device(struct pci_dev *pdev)
 {
 	pdev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
+	if (pdev->is_virtfn && !pdev->is_physfn)
+		if (pdev->physfn)
+			if (pdev->physfn->sriov)
+				atomic_inc(&pdev->physfn->sriov->
+					VFs_assigned_cnt);
 }
 EXPORT_SYMBOL_GPL(pci_sriov_assign_device);
 
@@ -660,6 +644,11 @@ EXPORT_SYMBOL_GPL(pci_sriov_assign_device);
 void pci_sriov_deassign_device(struct pci_dev *pdev)
 {
 	pdev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
+	if (pdev->is_virtfn && !pdev->is_physfn)
+		if (pdev->physfn)
+			if (pdev->physfn->sriov)
+				atomic_dec(&pdev->physfn->sriov->
+					VFs_assigned_cnt);
 }
 EXPORT_SYMBOL_GPL(pci_sriov_deassign_device);      
 
diff --git a/drivers/pci/pci.h b/drivers/pci/pci.h
index 6bd0822..d17bda2 100644
--- a/drivers/pci/pci.h
+++ b/drivers/pci/pci.h
@@ -235,6 +235,7 @@ struct pci_sriov {
 	u32 pgsz;		/* page size for BAR alignment */
 	u8 link;		/* Function Dependency Link */
 	u16 driver_max_VFs;	/* max num VFs driver supports */
+	atomic_t VFs_assigned_cnt; /* counter of VFs assigned to VM */
 	struct pci_dev *dev;	/* lowest numbered PF */
 	struct pci_dev *self;	/* this PF */
 	struct mutex lock;	/* lock for VF bus */
-- 
1.7.1


------------------------------------------------------------------------------
Open source business process management suite built on Java and Eclipse
Turn processes into business applications with Bonita BPM Community Edition
Quickly connect people, data, and systems into organized workflows
Winner of BOSSIE, CODIE, OW2 and Gartner awards
http://p.sf.net/sfu/Bonitasoft
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel&#174; Ethernet, visit http://communities.intel.com/community/wired

Re: [PATCH 1/2 V2] PCI: introduce device assignment interface and refactory related code

[David Vrabel]

On 11/07/14 06:52, Ethan Zhao wrote:
> 
> --- a/drivers/xen/xen-pciback/pci_stub.c
> +++ b/drivers/xen/xen-pciback/pci_stub.c
> @@ -133,7 +133,7 @@ static void pcistub_device_release(struct kref *kref)
>  	xen_pcibk_config_free_dyn_fields(dev);
>  	xen_pcibk_config_free_dev(dev);
>  
> -	dev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
> +	pci_sriov_deassign_device(dev);
>  	pci_dev_put(dev);
>  
>  	kfree(psdev);
> @@ -404,7 +404,7 @@ static int pcistub_init_device(struct pci_dev *dev)
>  	dev_dbg(&dev->dev, "reset device\n");
>  	xen_pcibk_reset_device(dev);
>  
> -	dev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
> +	pci_sriov_assign_device(dev);
>  	return 0;

Xen's PCI passthrough works with all PCI devices not just SR-IOV ones,
so the naming of the helpers isn't correct.

David

Re: [PATCH 1/2 V2] PCI: introduce device assignment interface and refactory related code

[Ethan Zhao]

> 在 2014年7月11日，下午5:41，David Vrabel <david.vrabel@citrix.com> 写道：
> 
>> On 11/07/14 06:52, Ethan Zhao wrote:
>> 
>> --- a/drivers/xen/xen-pciback/pci_stub.c
>> +++ b/drivers/xen/xen-pciback/pci_stub.c
>> @@ -133,7 +133,7 @@ static void pcistub_device_release(struct kref *kref)
>>    xen_pcibk_config_free_dyn_fields(dev);
>>    xen_pcibk_config_free_dev(dev);
>> 
>> -    dev->dev_flags &= ~PCI_DEV_FLAGS_ASSIGNED;
>> +    pci_sriov_deassign_device(dev);
>>    pci_dev_put(dev);
>> 
>>    kfree(psdev);
>> @@ -404,7 +404,7 @@ static int pcistub_init_device(struct pci_dev *dev)
>>    dev_dbg(&dev->dev, "reset device\n");
>>    xen_pcibk_reset_device(dev);
>> 
>> -    dev->dev_flags |= PCI_DEV_FLAGS_ASSIGNED;
>> +    pci_sriov_assign_device(dev);
>>    return 0;
> 
> Xen's PCI passthrough works with all PCI devices not just SR-IOV ones,
> so the naming of the helpers isn't correct.
yes, but so far, no better name for it , though the helpers work for all PCI devices.
do you have any good suggestion ?

Thanks,
Ethan
> 
> David

------------------------------------------------------------------------------
Open source business process management suite built on Java and Eclipse
Turn processes into business applications with Bonita BPM Community Edition
Quickly connect people, data, and systems into organized workflows
Winner of BOSSIE, CODIE, OW2 and Gartner awards
http://p.sf.net/sfu/Bonitasoft
_______________________________________________
E1000-devel mailing list
E1000-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/e1000-devel
To learn more about Intel&#174; Ethernet, visit http://communities.intel.com/community/wired