[PATCH rfc] vfio-pci: Allow write combining

[Keith Busch <kbusch@meta.com>]

From: Keith Busch <kbusch@kernel.org>

Write combining can be provide performance improvement for places that
can safely use this capability.

Previous discussions on the topic suggest a vfio user needs to
explicitly request such a mapping, and it sounds like a new vfio
specific ioctl to request this is one way recommended way to do that.
This patch implements a new ioctl to achieve that so a user can request
write combining on prefetchable memory. A new ioctl seems a bit much for
just this purpose, so the implementation here provides a "flags" field
with only the write combine option defined. The rest of the bits are
reserved for future use.

Link: https://lore.kernel.org/all/20171009025000.39435-1-aik@ozlabs.ru/
Link: https://lore.kernel.org/lkml/ZLFBnACjoTbDmKuU@nvidia.com/
Signed-off-by: Keith Busch <kbusch@kernel.org>
---
 drivers/vfio/pci/vfio_pci_core.c | 39 +++++++++++++++++++++++++++++++-
 include/linux/vfio_pci_core.h    |  1 +
 include/uapi/linux/vfio.h        | 17 ++++++++++++++
 3 files changed, 56 insertions(+), 1 deletion(-)

diff --git a/drivers/vfio/pci/vfio_pci_core.c b/drivers/vfio/pci/vfio_pci_core.c
index ba0ce0075b2fb..c275c95eafe32 100644
--- a/drivers/vfio/pci/vfio_pci_core.c
+++ b/drivers/vfio/pci/vfio_pci_core.c
@@ -1042,12 +1042,18 @@ static int vfio_pci_ioctl_get_region_info(struct vfio_pci_core_device *vdev,
 		info.flags = VFIO_REGION_INFO_FLAG_READ |
 			     VFIO_REGION_INFO_FLAG_WRITE;
 		if (vdev->bar_mmap_supported[info.index]) {
+			struct resource *res;
+
 			info.flags |= VFIO_REGION_INFO_FLAG_MMAP;
 			if (info.index == vdev->msix_bar) {
 				ret = msix_mmappable_cap(vdev, &caps);
 				if (ret)
 					return ret;
 			}
+
+			res = &vdev->pdev->resource[index];
+			if (res->flags & IORESOURCE_PREFETCH)
+				info.flags |= VFIO_REGION_INFO_FLAG_PREFETCH;
 		}
 
 		break;
@@ -1223,6 +1229,32 @@ static int vfio_pci_ioctl_set_irqs(struct vfio_pci_core_device *vdev,
 	return ret;
 }
 
+static int vfio_pci_ioctl_set_region_flags(struct vfio_pci_core_device *vdev,
+				   struct vfio_region_flags __user *arg)
+{
+	struct vfio_region_flags region_flags;
+	struct resource *res;
+	u32 index;
+
+	if (copy_from_user(&region_flags, arg, sizeof(region_flags)))
+		return -EFAULT;
+
+	index = region_flags.index;
+	if (index >= PCI_STD_NUM_BARS)
+		return -EINVAL;
+
+	if (region_flags.flags & VFIO_REGION_FLAG_WRITE_COMBINE) {
+		res = &vdev->pdev->resource[index];
+		if (!(res->flags & IORESOURCE_MEM) ||
+		    !(res->flags & IORESOURCE_PREFETCH))
+			return -EINVAL;
+		vdev->bar_write_combine[index] = true;
+	} else
+		vdev->bar_write_combine[index] = false;
+
+	return 0;
+}
+
 static int vfio_pci_ioctl_reset(struct vfio_pci_core_device *vdev,
 				void __user *arg)
 {
@@ -1484,6 +1516,8 @@ long vfio_pci_core_ioctl(struct vfio_device *core_vdev, unsigned int cmd,
 		return vfio_pci_ioctl_reset(vdev, uarg);
 	case VFIO_DEVICE_SET_IRQS:
 		return vfio_pci_ioctl_set_irqs(vdev, uarg);
+	case VFIO_DEVICE_SET_REGION_FLAGS:
+		return vfio_pci_ioctl_set_region_flags(vdev, uarg);
 	default:
 		return -ENOTTY;
 	}
@@ -1756,7 +1790,10 @@ int vfio_pci_core_mmap(struct vfio_device *core_vdev, struct vm_area_struct *vma
 	}
 
 	vma->vm_private_data = vdev;
-	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+	if (vdev->bar_write_combine[index])
+		vma->vm_page_prot = pgprot_writecombine(vma->vm_page_prot);
+	else
+		vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
 	vma->vm_page_prot = pgprot_decrypted(vma->vm_page_prot);
 
 	/*
diff --git a/include/linux/vfio_pci_core.h b/include/linux/vfio_pci_core.h
index fbb472dd99b36..0e0122ce4196a 100644
--- a/include/linux/vfio_pci_core.h
+++ b/include/linux/vfio_pci_core.h
@@ -54,6 +54,7 @@ struct vfio_pci_core_device {
 	struct pci_dev		*pdev;
 	void __iomem		*barmap[PCI_STD_NUM_BARS];
 	bool			bar_mmap_supported[PCI_STD_NUM_BARS];
+	bool			bar_write_combine[PCI_STD_NUM_BARS];
 	u8			*pci_config_map;
 	u8			*vconfig;
 	struct perm_bits	*msi_perm;
diff --git a/include/uapi/linux/vfio.h b/include/uapi/linux/vfio.h
index 2b68e6cdf1902..5537b20b23541 100644
--- a/include/uapi/linux/vfio.h
+++ b/include/uapi/linux/vfio.h
@@ -275,6 +275,7 @@ struct vfio_region_info {
 #define VFIO_REGION_INFO_FLAG_WRITE	(1 << 1) /* Region supports write */
 #define VFIO_REGION_INFO_FLAG_MMAP	(1 << 2) /* Region supports mmap */
 #define VFIO_REGION_INFO_FLAG_CAPS	(1 << 3) /* Info supports caps */
+#define VFIO_REGION_INFO_FLAG_PREFETCH	(1 << 4) /* Region is prefetchable */
 	__u32	index;		/* Region index */
 	__u32	cap_offset;	/* Offset within info struct of first cap */
 	__aligned_u64	size;	/* Region size (bytes) */
@@ -1821,6 +1822,22 @@ struct vfio_iommu_spapr_tce_remove {
 };
 #define VFIO_IOMMU_SPAPR_TCE_REMOVE	_IO(VFIO_TYPE, VFIO_BASE + 20)
 
+/**
+ * VFIO_DEVICE_SET_REGION_FLAGS	 - _IOW(VFIO_TYPE, VFIO_BASE + 21, struct vfio_region_flags)
+ *
+ * Set mapping options for the region
+ *
+ * Flags supported:
+ * - VFIO_REGION_FLAG_WRITE_COMBINE: use write-combine when requested to map
+ *   this region. Supported only if the region is prefetchable.
+ */
+struct vfio_region_flags {
+	__u32	index;		/* Region index */
+	__u32	flags;		/* Region flags */
+#define VFIO_REGION_FLAG_WRITE_COMBINE	(1 << 0)
+};
+#define VFIO_DEVICE_SET_REGION_FLAGS	_IO(VFIO_TYPE, VFIO_BASE + 21)
+
 /* ***************************************************************** */
 
 #endif /* _UAPIVFIO_H */
-- 
2.43.5