x86 / iommu: set up a scratch page in the quarantine domain

This patch introduces a new iommu_op to facilitate a per-implementation
quarantine set up, and then further code for x86 implementations
(amd and vtd) to set up a read-only scratch page to serve as the source
for DMA reads whilst a device is assigned to dom_io. DMA writes will
continue to fault as before.

The reason for doing this is that some hardware may continue to re-try
DMA (despite FLR) in the event of an error, or even BME being cleared, and
will fail to deal with DMA read faults gracefully. Having a scratch page
mapped will allow pending DMA reads to complete and thus such buggy
hardware will eventually be quiesced.

NOTE: These modifications are restricted to x86 implementations only as
      the buggy h/w I am aware of is only used with Xen in an x86
      environment. ARM may require similar code but, since I am not
      aware of the need, this patch does not modify any ARM implementation.

Signed-off-by: Paul Durrant <pdurrant@amazon.com>
Reviewed-by: Jan Beulich <jbeulich@suse.com>
master commit: ea38867831da67eed0e9c61672c8941016b63dd9
master date: 2019-11-29 18:27:54 +0000

diff --git a/xen/drivers/passthrough/amd/iommu_map.c b/xen/drivers/passthrough/amd/iommu_map.c
index bcd90508a77a9dd4db904a2ca6d5135ce8398dfb..21fbea04674747eaee5345fed43b97cdbc248eef 100644 (file)
--- a/xen/drivers/passthrough/amd/iommu_map.c
+++ b/xen/drivers/passthrough/amd/iommu_map.c
@@ -741,3 +741,65 @@ void amd_iommu_share_p2m(struct domain *d)
                         mfn_x(pgd_mfn));
     }
 }
+
+int __init amd_iommu_quarantine_init(struct domain *d)
+{
+    struct domain_iommu *hd = dom_iommu(d);
+    unsigned long max_gfn =
+        PFN_DOWN((1ul << DEFAULT_DOMAIN_ADDRESS_WIDTH) - 1);
+    unsigned int level = amd_iommu_get_paging_mode(max_gfn);
+    uint64_t *table;
+
+    if ( hd->arch.root_table )
+    {
+        ASSERT_UNREACHABLE();
+        return 0;
+    }
+
+    spin_lock(&hd->arch.mapping_lock);
+
+    hd->arch.root_table = alloc_amd_iommu_pgtable();
+    if ( !hd->arch.root_table )
+        goto out;
+
+    table = __map_domain_page(hd->arch.root_table);
+    while ( level )
+    {
+        struct page_info *pg;
+        unsigned int i;
+
+        /*
+         * The pgtable allocator is fine for the leaf page, as well as
+         * page table pages, and the resulting allocations are always
+         * zeroed.
+         */
+        pg = alloc_amd_iommu_pgtable();
+        if ( !pg )
+            break;
+
+        for ( i = 0; i < PTE_PER_TABLE_SIZE; i++ )
+        {
+            uint32_t *pde = (uint32_t *)&table[i];
+
+            /*
+             * PDEs are essentially a subset of PTEs, so this function
+             * is fine to use even at the leaf.
+             */
+            set_iommu_pde_present(pde, mfn_x(page_to_mfn(pg)), level - 1,
+                                  false, true);
+        }
+
+        unmap_domain_page(table);
+        table = __map_domain_page(pg);
+        level--;
+    }
+    unmap_domain_page(table);
+
+ out:
+    spin_unlock(&hd->arch.mapping_lock);
+
+    amd_iommu_flush_all_pages(d);
+
+    /* Pages leaked in failure case */
+    return level ? -ENOMEM : 0;
+}

diff --git a/xen/drivers/passthrough/amd/pci_amd_iommu.c b/xen/drivers/passthrough/amd/pci_amd_iommu.c
index 7ae559d8c8e39388f61fb45ca7b1cfc4a8b70e18..0b641ff75ccc91fe55f62ee5f38a6897340a6f40 100644 (file)
--- a/xen/drivers/passthrough/amd/pci_amd_iommu.c
+++ b/xen/drivers/passthrough/amd/pci_amd_iommu.c
@@ -118,10 +118,6 @@ static void amd_iommu_setup_domain_device(
     u8 bus = pdev->bus;
     const struct domain_iommu *hd = dom_iommu(domain);
 
-    /* dom_io is used as a sentinel for quarantined devices */
-    if ( domain == dom_io )
-        return;
-
     BUG_ON( !hd->arch.root_table || !hd->arch.paging_mode ||
             !iommu->dev_table.buffer );
 
@@ -222,7 +218,7 @@ static int __must_check allocate_domain_resources(struct domain_iommu *hd)
     return rc;
 }
 
-static int get_paging_mode(unsigned long entries)
+int amd_iommu_get_paging_mode(unsigned long entries)
 {
     int level = 1;
 
@@ -251,7 +247,8 @@ static int amd_iommu_domain_init(struct domain *d)
      *   unilaterally.
      */
     hd->arch.paging_mode = is_hvm_domain(d)
-        ? IOMMU_PAGING_MODE_LEVEL_4 : get_paging_mode(get_upper_mfn_bound());
+        ? IOMMU_PAGING_MODE_LEVEL_4
+        : amd_iommu_get_paging_mode(get_upper_mfn_bound());
 
     return 0;
 }
@@ -315,10 +312,6 @@ void amd_iommu_disable_domain_device(struct domain *domain,
     int req_id;
     u8 bus = pdev->bus;
 
-    /* dom_io is used as a sentinel for quarantined devices */
-    if ( domain == dom_io )
-        return;
-
     BUG_ON ( iommu->dev_table.buffer == NULL );
     req_id = get_dma_requestor_id(iommu->seg, PCI_BDF2(bus, devfn));
     dte = iommu->dev_table.buffer + (req_id * IOMMU_DEV_TABLE_ENTRY_SIZE);
@@ -615,6 +608,7 @@ static void amd_dump_p2m_table(struct domain *d)
 const struct iommu_ops amd_iommu_ops = {
     .init = amd_iommu_domain_init,
     .hwdom_init = amd_iommu_hwdom_init,
+    .quarantine_init = amd_iommu_quarantine_init,
     .add_device = amd_iommu_add_device,
     .remove_device = amd_iommu_remove_device,
     .assign_device  = amd_iommu_assign_device,

diff --git a/xen/drivers/passthrough/iommu.c b/xen/drivers/passthrough/iommu.c
index 132bbb37629b9a41efb144a7a5558aa670e72aed..2eff0c11c7df2f2c11438e53e302ab2b7796d621 100644 (file)
--- a/xen/drivers/passthrough/iommu.c
+++ b/xen/drivers/passthrough/iommu.c
@@ -380,6 +380,21 @@ int iommu_iotlb_flush_all(struct domain *d)
     return rc;
 }
 
+static int __init iommu_quarantine_init(void)
+{
+    const struct domain_iommu *hd = dom_iommu(dom_io);
+    int rc;
+
+    rc = iommu_domain_init(dom_io);
+    if ( rc )
+        return rc;
+
+    if ( !hd->platform_ops->quarantine_init )
+        return 0;
+
+    return hd->platform_ops->quarantine_init(dom_io);
+}
+
 int __init iommu_setup(void)
 {
     int rc = -ENODEV;
@@ -413,7 +428,7 @@ int __init iommu_setup(void)
     printk("I/O virtualisation %sabled\n", iommu_enabled ? "en" : "dis");
     if ( iommu_enabled )
     {
-        if ( iommu_domain_init(dom_io) )
+        if ( iommu_quarantine_init() )
             panic("Could not set up quarantine\n");
 
         printk(" - Dom0 mode: %s\n",

diff --git a/xen/drivers/passthrough/vtd/iommu.c b/xen/drivers/passthrough/vtd/iommu.c
index 8712d3b4dc1057ba1ecf92aa15f4ec926f1f056e..42ad24adb86acfe1139224928853cff81d96c3bf 100644 (file)
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -192,7 +192,7 @@ u64 alloc_pgtable_maddr(struct acpi_drhd_unit *drhd, unsigned long npages)
     nodeid_t node = NUMA_NO_NODE;
     unsigned int i;
 
-    rhsa = drhd_to_rhsa(drhd);
+    rhsa = drhd ? drhd_to_rhsa(drhd) : NULL;
     if ( rhsa )
         node =  pxm_to_node(rhsa->proximity_domain);
 
@@ -1340,10 +1340,6 @@ int domain_context_mapping_one(
     int agaw, rc, ret;
     bool_t flush_dev_iotlb;
 
-    /* dom_io is used as a sentinel for quarantined devices */
-    if ( domain == dom_io )
-        return 0;
-
     ASSERT(pcidevs_locked());
     spin_lock(&iommu->lock);
     maddr = bus_to_context_maddr(iommu, bus);
@@ -1579,10 +1575,6 @@ int domain_context_unmap_one(
     int iommu_domid, rc, ret;
     bool_t flush_dev_iotlb;
 
-    /* dom_io is used as a sentinel for quarantined devices */
-    if ( domain == dom_io )
-        return 0;
-
     ASSERT(pcidevs_locked());
     spin_lock(&iommu->lock);
 
@@ -1715,10 +1707,6 @@ static int domain_context_unmap(struct domain *domain, u8 devfn,
         goto out;
     }
 
-    /* dom_io is used as a sentinel for quarantined devices */
-    if ( domain == dom_io )
-        goto out;
-
     /*
      * if no other devices under the same iommu owned by this domain,
      * clear iommu in iommu_bitmap and clear domain_id in domid_bitmp
@@ -2684,9 +2672,69 @@ static void vtd_dump_p2m_table(struct domain *d)
     vtd_dump_p2m_table_level(hd->arch.pgd_maddr, agaw_to_level(hd->arch.agaw), 0, 0);
 }
 
+static int __init intel_iommu_quarantine_init(struct domain *d)
+{
+    struct domain_iommu *hd = dom_iommu(d);
+    struct dma_pte *parent;
+    unsigned int agaw = width_to_agaw(DEFAULT_DOMAIN_ADDRESS_WIDTH);
+    unsigned int level = agaw_to_level(agaw);
+    int rc;
+
+    if ( hd->arch.pgd_maddr )
+    {
+        ASSERT_UNREACHABLE();
+        return 0;
+    }
+
+    spin_lock(&hd->arch.mapping_lock);
+
+    hd->arch.pgd_maddr = alloc_pgtable_maddr(NULL, 1);
+    if ( !hd->arch.pgd_maddr )
+        goto out;
+
+    parent = map_vtd_domain_page(hd->arch.pgd_maddr);
+    while ( level )
+    {
+        uint64_t maddr;
+        unsigned int offset;
+
+        /*
+         * The pgtable allocator is fine for the leaf page, as well as
+         * page table pages, and the resulting allocations are always
+         * zeroed.
+         */
+        maddr = alloc_pgtable_maddr(NULL, 1);
+        if ( !maddr )
+            break;
+
+        for ( offset = 0; offset < PTE_NUM; offset++ )
+        {
+            struct dma_pte *pte = &parent[offset];
+
+            dma_set_pte_addr(*pte, maddr);
+            dma_set_pte_readable(*pte);
+        }
+        iommu_flush_cache_page(parent, 1);
+
+        unmap_vtd_domain_page(parent);
+        parent = map_vtd_domain_page(maddr);
+        level--;
+    }
+    unmap_vtd_domain_page(parent);
+
+ out:
+    spin_unlock(&hd->arch.mapping_lock);
+
+    rc = iommu_flush_iotlb_all(d);
+
+    /* Pages leaked in failure case */
+    return level ? -ENOMEM : rc;
+}
+
 const struct iommu_ops intel_iommu_ops = {
     .init = intel_iommu_domain_init,
     .hwdom_init = intel_iommu_hwdom_init,
+    .quarantine_init = intel_iommu_quarantine_init,
     .add_device = intel_iommu_add_device,
     .enable_device = intel_iommu_enable_device,
     .remove_device = intel_iommu_remove_device,

diff --git a/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h b/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h
index 99bc21c7b3b0eb8abe0136c5d94ac524cf241a59..c42688fe514f0fd549a093b9465c4a0fffce268a 100644 (file)
--- a/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h
+++ b/xen/include/asm-x86/hvm/svm/amd-iommu-proto.h
@@ -51,6 +51,9 @@ void get_iommu_features(struct amd_iommu *iommu);
 int amd_iommu_init(void);
 int amd_iommu_update_ivrs_mapping_acpi(void);
 
+int amd_iommu_get_paging_mode(unsigned long entries);
+int amd_iommu_quarantine_init(struct domain *d);
+
 /* mapping functions */
 int __must_check amd_iommu_map_page(struct domain *d, unsigned long gfn,
                                     unsigned long mfn, unsigned int flags);

diff --git a/xen/include/xen/iommu.h b/xen/include/xen/iommu.h
index 235d2a620b14f97774447c4fc06c7eaf56483397..a0573b659ccf15dbedcf5ba580fff6048f8c3fc4 100644 (file)
--- a/xen/include/xen/iommu.h
+++ b/xen/include/xen/iommu.h
@@ -139,6 +139,7 @@ typedef int iommu_grdm_t(xen_pfn_t start, xen_ulong_t nr, u32 id, void *ctxt);
 struct iommu_ops {
     int (*init)(struct domain *d);
     void (*hwdom_init)(struct domain *d);
+    int (*quarantine_init)(struct domain *d);
     int (*add_device)(u8 devfn, device_t *dev);
     int (*enable_device)(device_t *dev);
     int (*remove_device)(u8 devfn, device_t *dev);