Date:	Thu, 22 May 2014 17:08:13 -0600
From:	Alex Williamson <alex.williamson@...hat.com>
To:	linux-pci@...r.kernel.org, iommu@...ts.linux-foundation.org
Cc:	bhelgaas@...gle.com, acooks@...il.com,
	linux-kernel@...r.kernel.org, eddy0596@...il.com, linux@...izon.com
Subject: [PATCH v4 08/16] iommu: Create central IOMMU group lookup/creation
 interface

Currently each IOMMU driver that supports IOMMU groups has its own
code for discovering the base device used in grouping.  This code
is generally not specific to the IOMMU hardware, but to the bus of
the devices managed by the IOMMU.  We can therefore create a common
interface for supporting devices on different buses.

Signed-off-by: Alex Williamson <alex.williamson@...hat.com>
---
 drivers/iommu/iommu.c |  181 +++++++++++++++++++++++++++++++++++++++++++++++++
 include/linux/iommu.h |    1 
 2 files changed, 182 insertions(+)

diff --git a/drivers/iommu/iommu.c b/drivers/iommu/iommu.c
index e5555fc..db55fe6 100644
--- a/drivers/iommu/iommu.c
+++ b/drivers/iommu/iommu.c
@@ -29,6 +29,7 @@
 #include <linux/idr.h>
 #include <linux/notifier.h>
 #include <linux/err.h>
+#include <linux/pci.h>
 #include <trace/events/iommu.h>
 
 static struct kset *iommu_group_kset;
@@ -514,6 +515,186 @@ int iommu_group_id(struct iommu_group *group)
 }
 EXPORT_SYMBOL_GPL(iommu_group_id);
 
+/*
+ * To consider a PCI device isolated, we require ACS to support Source
+ * Validation, Request Redirection, Completer Redirection, and Upstream
+ * Forwarding.  This effectively means that devices cannot spoof their
+ * requester ID, requests and completions cannot be redirected, and all
+ * transactions are forwarded upstream, even as it passes through a
+ * bridge where the target device is downstream.
+ */
+#define REQ_ACS_FLAGS   (PCI_ACS_SV | PCI_ACS_RR | PCI_ACS_CR | PCI_ACS_UF)
+
+struct group_for_pci_data {
+	struct pci_dev *pdev;
+	struct iommu_group *group;
+};
+
+/*
+ * DMA alias iterator callback, return the last seen device.  Stop and return
+ * the IOMMU group if we find one along the way.
+ */
+static int get_pci_alias_or_group(struct pci_dev *pdev, u16 alias, void *opaque)
+{
+	struct group_for_pci_data *data = opaque;
+
+	data->pdev = pdev;
+	data->group = iommu_group_get(&pdev->dev);
+
+	return data->group != NULL;
+}
+
+/*
+ * Use standard PCI bus topology, isolation features, and DMA alias quirks
+ * to find or create an IOMMU group for a device.
+ */
+static struct iommu_group *iommu_group_get_for_pci_dev(struct pci_dev *pdev)
+{
+	struct group_for_pci_data data;
+	struct pci_bus *bus;
+	struct iommu_group *group = NULL;
+	struct pci_dev *tmp;
+
+	/*
+	 * Find the upstream DMA alias for the device.  A device must not
+	 * be aliased due to topology in order to have its own IOMMU group.
+	 * If we find an alias along the way that already belongs to a
+	 * group, use it.
+	 */
+	if (pci_for_each_dma_alias(pdev, get_pci_alias_or_group, &data))
+		return data.group;
+
+	pdev = data.pdev;
+
+	/*
+	 * Continue upstream from the point of minimum IOMMU granularity
+	 * due to aliases to the point where devices are protected from
+	 * peer-to-peer DMA by PCI ACS.  Again, if we find an existing
+	 * group, use it.
+	 */
+	for (bus = pdev->bus; !pci_is_root_bus(bus); bus = bus->parent) {
+		if (!bus->self)
+			continue;
+
+		if (pci_acs_path_enabled(bus->self, NULL, REQ_ACS_FLAGS))
+			break;
+
+		pdev = bus->self;
+
+		group = iommu_group_get(&pdev->dev);
+		if (group)
+			return group;
+	}
+
+	/*
+	 * Next we need to consider DMA alias quirks.  If one device aliases
+	 * to another, they should be grouped together.  It's theoretically
+	 * possible that aliases could create chains of devices where each
+	 * device aliases another device.  If we then factor in multifunction
+	 * ACS grouping requirements, each alias could incorporate a new slot
+	 * with multiple functions, each with aliases.  This is all extremely
+	 * unlikely as DMA alias quirks are typically only used for PCIe
+	 * devices where we usually have a single slot per bus.  Furthermore,
+	 * the alias quirk is usually to another function within the slot
+	 * (and ACS multifunction is not supported) or to a different slot
+	 * that doesn't physically exist.  The likely scenario is therefore
+	 * that everything on the bus gets grouped together.  To reduce the
+	 * problem space, share the IOMMU group for all devices on the bus
+	 * if a DMA alias quirk is present on the bus.
+	 */
+	tmp = NULL;
+	for_each_pci_dev(tmp) {
+		if (tmp->bus != pdev->bus ||
+		    !(tmp->dev_flags & PCI_DEV_FLAGS_DMA_ALIAS_DEVFN))
+			continue;
+
+		pci_dev_put(tmp);
+		tmp = NULL;
+
+		/* We have an alias quirk, search for an existing group */
+		for_each_pci_dev(tmp) {
+			struct iommu_group *group_tmp;
+			if (tmp->bus != pdev->bus)
+				continue;
+
+			group_tmp = iommu_group_get(&tmp->dev);
+			if (!group) {
+				group = group_tmp;
+				continue;
+			}
+
+			if (group_tmp) {
+				WARN_ON(group != group_tmp);
+				iommu_group_put(group_tmp);
+			}
+		}
+
+		return group ? group : iommu_group_alloc();
+	}
+
+	/*
+	 * Non-multifunction devices or multifunction devices supporting
+	 * ACS get their own group.
+	 */
+	if (!pdev->multifunction || pci_acs_enabled(pdev, REQ_ACS_FLAGS))
+		return iommu_group_alloc();
+
+	/*
+	 * Multifunction devices not supporting ACS share a group with other
+	 * similar devices in the same slot.
+	 */
+	tmp = NULL;
+	for_each_pci_dev(tmp) {
+		if (tmp == pdev || tmp->bus != pdev->bus ||
+		    PCI_SLOT(tmp->devfn) !=  PCI_SLOT(pdev->devfn) ||
+		    pci_acs_enabled(tmp, REQ_ACS_FLAGS))
+			continue;
+
+		group = iommu_group_get(&tmp->dev);
+		if (group) {
+			pci_dev_put(tmp);
+			return group;
+		}
+	}
+
+	/* No shared group found, allocate new */
+	return iommu_group_alloc();
+}
+
+/**
+ * iommu_group_get_for_dev - Find or create the IOMMU group for a device
+ * @dev: target device
+ *
+ * This function is intended to be called by IOMMU drivers and extended to
+ * support common, bus-defined algorithms when determining or creating the
+ * IOMMU group for a device.  On success, the caller will hold a reference
+ * to the returned IOMMU group, which will already include the provided
+ * device.  The reference should be released with iommu_group_put().
+ */
+struct iommu_group *iommu_group_get_for_dev(struct device *dev)
+{
+	struct iommu_group *group = ERR_PTR(-EIO);
+	int ret;
+
+	group = iommu_group_get(dev);
+	if (group)
+		return group;
+
+	if (dev_is_pci(dev))
+		group = iommu_group_get_for_pci_dev(to_pci_dev(dev));
+
+	if (IS_ERR(group))
+		return group;
+
+	ret = iommu_group_add_device(group, dev);
+	if (ret) {
+		iommu_group_put(group);
+		return ERR_PTR(ret);
+	}
+
+	return group;
+}
+
 static int add_iommu_group(struct device *dev, void *data)
 {
 	struct iommu_ops *ops = data;
diff --git a/include/linux/iommu.h b/include/linux/iommu.h
index b96a5b2..a2e5843 100644
--- a/include/linux/iommu.h
+++ b/include/linux/iommu.h
@@ -181,6 +181,7 @@ extern int iommu_group_register_notifier(struct iommu_group *group,
 extern int iommu_group_unregister_notifier(struct iommu_group *group,
 					   struct notifier_block *nb);
 extern int iommu_group_id(struct iommu_group *group);
+extern struct iommu_group *iommu_group_get_for_dev(struct device *dev);
 
 extern int iommu_domain_get_attr(struct iommu_domain *domain, enum iommu_attr,
 				 void *data);

--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/

Powered by blists - more mailing lists