[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Xen-devel] Re: [PATCH 2/6] Provide a mechanism to register domain owner of a PCI device.



On 12/09/09 12:51, Konrad Rzeszutek Wilk wrote:
. and also to find the domain owner based on the PCI device and
to unregister a domain owner of a PCI device.

Signed-off-by: Konrad Rzeszutek Wilk<konrad.wilk@xxxxxxxxxx>
---
  arch/x86/include/asm/xen/pci.h |   16 +++++++++
  arch/x86/xen/pci.c             |   73 ++++++++++++++++++++++++++++++++++++++++
  2 files changed, 89 insertions(+), 0 deletions(-)

diff --git a/arch/x86/include/asm/xen/pci.h b/arch/x86/include/asm/xen/pci.h
index cb84abe..f4312d5 100644
--- a/arch/x86/include/asm/xen/pci.h
+++ b/arch/x86/include/asm/xen/pci.h
@@ -7,6 +7,11 @@ int xen_create_msi_irq(struct pci_dev *dev,
                        struct msi_desc *msidesc,
                        int type);
  int xen_destroy_irq(int irq);
+
+int find_device_owner(struct pci_dev *dev);
+int register_device_owner(struct pci_dev *dev, domid_t domain);
+int unregister_device_owner(struct pci_dev *dev);

Could you give these less generic names?

+
  #else
  static inline int xen_register_gsi(u32 gsi, int triggering, int polarity)
  {
@@ -23,6 +28,17 @@ static inline int xen_destroy_irq(int irq)
  {
        return -1;
  }
+
+static inline int find_device_owner(struct pci_dev *dev) { return -1; }
+static inline int register_device_owner(struct pci_dev *dev, domid_t domain)
+{
+       return -1;
+}
+static inline int unregister_device_owner(struct pci_dev *dev)
+{
+       return -1;
+}

Make the formatting consistent here.

+
  #endif

  #if defined(CONFIG_PCI_MSI)&&  defined(CONFIG_XEN_DOM0_PCI)
diff --git a/arch/x86/xen/pci.c b/arch/x86/xen/pci.c
index 44d91ad..4fa734c 100644
--- a/arch/x86/xen/pci.c
+++ b/arch/x86/xen/pci.c
@@ -2,6 +2,7 @@
  #include<linux/acpi.h>
  #include<linux/pci.h>
  #include<linux/msi.h>
+#include<linux/slab.h>

  #include<asm/mpspec.h>
  #include<asm/io_apic.h>
@@ -109,3 +110,75 @@ error:
        return ret;
  }
  #endif
+
+struct device_owner {

Again, a bit generic (the name should at least convey we're talking about owner domains).


+       domid_t domain;
+       struct pci_dev *dev;
+       struct list_head list;
+};
+
+static DEFINE_SPINLOCK(dev_domain_list_spinlock);
+static struct list_head dev_domain_list = LIST_HEAD_INIT(dev_domain_list);
+
+struct device_owner *find_device(struct pci_dev *dev)
+{
+
+       struct device_owner *owner;
Stray blank line.
+
+       list_for_each_entry(owner,&dev_domain_list, list) {
+               if (owner->dev == dev)
+                       return owner;
+       }
+       return NULL;
+}
+
+int find_device_owner(struct pci_dev *dev)
+{
+
+       struct device_owner *owner = NULL;
Again.
+
+       owner = find_device(dev);
+       if (!owner)
+               return -ENODEV;
+
+       return owner->domain;
+}
+EXPORT_SYMBOL(find_device_owner);
+
+int register_device_owner(struct pci_dev *dev, domid_t domain)
+{
+       struct device_owner *owner;
+       unsigned long flags;
+
+       if (find_device(dev))
Should this be under the lock?
+               return -EEXIST;
+
+       owner = kzalloc(sizeof(struct device_owner), GFP_KERNEL);
Check for NULL.
+       owner->domain = domain;
+       owner->dev = dev;
+
+       spin_lock_irqsave(&dev_domain_list_spinlock, flags);
+       list_add_tail(&owner->list,&dev_domain_list);
+       spin_unlock_irqrestore(&dev_domain_list_spinlock, flags);
+
+       return 0;
+}
+EXPORT_SYMBOL(register_device_owner);
+
+int unregister_device_owner(struct pci_dev *dev)
+{
+       struct device_owner *owner = NULL;
+       unsigned long flags;
+
+       owner = find_device(dev);
Shouldn't this be under the lock too?
+       if (!owner)
+               return -ENODEV;
+
+       spin_lock_irqsave(&dev_domain_list_spinlock, flags);
+       list_del(&owner->list);
+       spin_unlock_irqrestore(&dev_domain_list_spinlock, flags);
+
+       kfree(owner);
+       return 0;
+}
+EXPORT_SYMBOL(unregister_device_owner);

Thanks,
    J

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel


 


Rackspace

Lists.xenproject.org is hosted with RackSpace, monitoring our
servers 24x7x365 and backed by RackSpace's Fanatical Support®.