Merge tag 'iommu-fixes-v5.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu

Pull iommu fixes from Joerg Roedel:

 - AMD IOMMU fix to make sure features are detected before they are
   queried.

 - Intel IOMMU address alignment check fix for an IOLTB flushing
   command.

 - Performance fix for Intel IOMMU to make sure the code does not do
   full IOTLB flushes all the time. Those flushes are very expensive
   on emulated IOMMUs.

* tag 'iommu-fixes-v5.11-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/joro/iommu:
  iommu/vt-d: Do not use flush-queue when caching-mode is on
  iommu/vt-d: Correctly check addr alignment in qi_flush_dev_iotlb_pasid()
  iommu/amd: Use IVHD EFR for early initialization of IOMMU features

drivers/iommu/amd/amd_iommu.h

@@ -84,12 +84,9 @@ static inline bool is_rd890_iommu(struct pci_dev *pdev)
 	       (pdev->device == PCI_DEVICE_ID_RD890_IOMMU);
 }
 
-static inline bool iommu_feature(struct amd_iommu *iommu, u64 f)
+static inline bool iommu_feature(struct amd_iommu *iommu, u64 mask)
 {
-	if (!(iommu->cap & (1 << IOMMU_CAP_EFR)))
-		return false;
-
-	return !!(iommu->features & f);
+	return !!(iommu->features & mask);
 }
 
 static inline u64 iommu_virt_to_phys(void *vaddr)

drivers/iommu/amd/amd_iommu_types.h

@@ -387,6 +387,10 @@
 #define IOMMU_CAP_NPCACHE 26
 #define IOMMU_CAP_EFR     27
 
+/* IOMMU IVINFO */
+#define IOMMU_IVINFO_OFFSET     36
+#define IOMMU_IVINFO_EFRSUP     BIT(0)
+
 /* IOMMU Feature Reporting Field (for IVHD type 10h */
 #define IOMMU_FEAT_GASUP_SHIFT	6
 

drivers/iommu/amd/init.c

@@ -257,6 +257,8 @@ static void init_device_table_dma(void);
 
 static bool amd_iommu_pre_enabled = true;
 
+static u32 amd_iommu_ivinfo __initdata;
+
 bool translation_pre_enabled(struct amd_iommu *iommu)
 {
 	return (iommu->flags & AMD_IOMMU_FLAG_TRANS_PRE_ENABLED);
@@ -296,6 +298,18 @@ int amd_iommu_get_num_iommus(void)
 	return amd_iommus_present;
 }
 
+/*
+ * For IVHD type 0x11/0x40, EFR is also available via IVHD.
+ * Default to IVHD EFR since it is available sooner
+ * (i.e. before PCI init).
+ */
+static void __init early_iommu_features_init(struct amd_iommu *iommu,
+					     struct ivhd_header *h)
+{
+	if (amd_iommu_ivinfo & IOMMU_IVINFO_EFRSUP)
+		iommu->features = h->efr_reg;
+}
+
 /* Access to l1 and l2 indexed register spaces */
 
 static u32 iommu_read_l1(struct amd_iommu *iommu, u16 l1, u8 address)
@@ -1577,6 +1591,9 @@ static int __init init_iommu_one(struct amd_iommu *iommu, struct ivhd_header *h)
 
 		if (h->efr_reg & BIT(IOMMU_EFR_XTSUP_SHIFT))
 			amd_iommu_xt_mode = IRQ_REMAP_X2APIC_MODE;
+
+		early_iommu_features_init(iommu, h);
+
 		break;
 	default:
 		return -EINVAL;
@@ -1770,6 +1787,35 @@ static const struct attribute_group *amd_iommu_groups[] = {
 	NULL,
 };
 
+/*
+ * Note: IVHD 0x11 and 0x40 also contains exact copy
+ * of the IOMMU Extended Feature Register [MMIO Offset 0030h].
+ * Default to EFR in IVHD since it is available sooner (i.e. before PCI init).
+ */
+static void __init late_iommu_features_init(struct amd_iommu *iommu)
+{
+	u64 features;
+
+	if (!(iommu->cap & (1 << IOMMU_CAP_EFR)))
+		return;
+
+	/* read extended feature bits */
+	features = readq(iommu->mmio_base + MMIO_EXT_FEATURES);
+
+	if (!iommu->features) {
+		iommu->features = features;
+		return;
+	}
+
+	/*
+	 * Sanity check and warn if EFR values from
+	 * IVHD and MMIO conflict.
+	 */
+	if (features != iommu->features)
+		pr_warn(FW_WARN "EFR mismatch. Use IVHD EFR (%#llx : %#llx\n).",
+			features, iommu->features);
+}
+
 static int __init iommu_init_pci(struct amd_iommu *iommu)
 {
 	int cap_ptr = iommu->cap_ptr;
@@ -1789,8 +1835,7 @@ static int __init iommu_init_pci(struct amd_iommu *iommu)
 	if (!(iommu->cap & (1 << IOMMU_CAP_IOTLB)))
 		amd_iommu_iotlb_sup = false;
 
-	/* read extended feature bits */
-	iommu->features = readq(iommu->mmio_base + MMIO_EXT_FEATURES);
+	late_iommu_features_init(iommu);
 
 	if (iommu_feature(iommu, FEATURE_GT)) {
 		int glxval;
@@ -2607,6 +2652,11 @@ static void __init free_dma_resources(void)
 	free_unity_maps();
 }
 
+static void __init ivinfo_init(void *ivrs)
+{
+	amd_iommu_ivinfo = *((u32 *)(ivrs + IOMMU_IVINFO_OFFSET));
+}
+
 /*
  * This is the hardware init function for AMD IOMMU in the system.
  * This function is called either from amd_iommu_init or from the interrupt
@@ -2661,6 +2711,8 @@ static int __init early_amd_iommu_init(void)
 	if (ret)
 		goto out;
 
+	ivinfo_init(ivrs_base);
+
 	amd_iommu_target_ivhd_type = get_highest_supported_ivhd_type(ivrs_base);
 	DUMP_printk("Using IVHD type %#x\n", amd_iommu_target_ivhd_type);
 

drivers/iommu/intel/dmar.c

@@ -1496,7 +1496,7 @@ void qi_flush_dev_iotlb_pasid(struct intel_iommu *iommu, u16 sid, u16 pfsid,
 	 * Max Invs Pending (MIP) is set to 0 for now until we have DIT in
 	 * ECAP.
 	 */
-	if (addr & GENMASK_ULL(size_order + VTD_PAGE_SHIFT, 0))
+	if (!IS_ALIGNED(addr, VTD_PAGE_SIZE << size_order))
 		pr_warn_ratelimited("Invalidate non-aligned address %llx, order %d\n",
 				    addr, size_order);
 

drivers/iommu/intel/iommu.c

@@ -5440,6 +5440,36 @@ intel_iommu_domain_set_attr(struct iommu_domain *domain,
 	return ret;
 }
 
+static bool domain_use_flush_queue(void)
+{
+	struct dmar_drhd_unit *drhd;
+	struct intel_iommu *iommu;
+	bool r = true;
+
+	if (intel_iommu_strict)
+		return false;
+
+	/*
+	 * The flush queue implementation does not perform page-selective
+	 * invalidations that are required for efficient TLB flushes in virtual
+	 * environments. The benefit of batching is likely to be much lower than
+	 * the overhead of synchronizing the virtual and physical IOMMU
+	 * page-tables.
+	 */
+	rcu_read_lock();
+	for_each_active_iommu(iommu, drhd) {
+		if (!cap_caching_mode(iommu->cap))
+			continue;
+
+		pr_warn_once("IOMMU batching is disabled due to virtualization");
+		r = false;
+		break;
+	}
+	rcu_read_unlock();
+
+	return r;
+}
+
 static int
 intel_iommu_domain_get_attr(struct iommu_domain *domain,
 			    enum iommu_attr attr, void *data)
@@ -5450,7 +5480,7 @@ intel_iommu_domain_get_attr(struct iommu_domain *domain,
 	case IOMMU_DOMAIN_DMA:
 		switch (attr) {
 		case DOMAIN_ATTR_DMA_USE_FLUSH_QUEUE:
-			*(int *)data = !intel_iommu_strict;
+			*(int *)data = domain_use_flush_queue();
 			return 0;
 		default:
 			return -ENODEV;