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

[xen stable-4.13] x86/ept: flush cache when modifying PTEs and sharing page tables



commit 572e349de17e001b385b6eba1a5104d1b38b3960
Author:     Roger Pau Monné <roger.pau@xxxxxxxxxx>
AuthorDate: Tue Jul 7 15:04:57 2020 +0200
Commit:     Jan Beulich <jbeulich@xxxxxxxx>
CommitDate: Tue Jul 7 15:04:57 2020 +0200

    x86/ept: flush cache when modifying PTEs and sharing page tables
    
    Modifications made to the page tables by EPT code need to be written
    to memory when the page tables are shared with the IOMMU, as Intel
    IOMMUs can be non-coherent and thus require changes to be written to
    memory in order to be visible to the IOMMU.
    
    In order to achieve this make sure data is written back to memory
    after writing an EPT entry when the recalc bit is not set in
    atomic_write_ept_entry. If such bit is set, the entry will be
    adjusted and atomic_write_ept_entry will be called a second time
    without the recalc bit set. Note that when splitting a super page the
    new tables resulting of the split should also be written back.
    
    Failure to do so can allow devices behind the IOMMU access to the
    stale super page, or cause coherency issues as changes made by the
    processor to the page tables are not visible to the IOMMU.
    
    This allows to remove the VT-d specific iommu_pte_flush helper, since
    the cache write back is now performed by atomic_write_ept_entry, and
    hence iommu_iotlb_flush can be used to flush the IOMMU TLB. The newly
    used method (iommu_iotlb_flush) can result in less flushes, since it
    might sometimes be called rightly with 0 flags, in which case it
    becomes a no-op.
    
    This is part of XSA-321.
    
    Signed-off-by: Roger Pau Monné <roger.pau@xxxxxxxxxx>
    Reviewed-by: Jan Beulich <jbeulich@xxxxxxxx>
    master commit: c23274fd0412381bd75068ebc9f8f8c90a4be748
    master date: 2020-07-07 14:40:11 +0200
---
 xen/arch/x86/mm/p2m-ept.c           | 24 ++++++++++++++++++-
 xen/drivers/passthrough/vtd/iommu.c | 47 -------------------------------------
 xen/include/asm-x86/iommu.h         |  4 ----
 3 files changed, 23 insertions(+), 52 deletions(-)

diff --git a/xen/arch/x86/mm/p2m-ept.c b/xen/arch/x86/mm/p2m-ept.c
index 07236e963d..9c1627dbb3 100644
--- a/xen/arch/x86/mm/p2m-ept.c
+++ b/xen/arch/x86/mm/p2m-ept.c
@@ -58,6 +58,19 @@ static int atomic_write_ept_entry(struct p2m_domain *p2m,
 
     write_atomic(&entryptr->epte, new.epte);
 
+    /*
+     * The recalc field on the EPT is used to signal either that a
+     * recalculation of the EMT field is required (which doesn't effect the
+     * IOMMU), or a type change. Type changes can only be between ram_rw,
+     * logdirty and ioreq_server: changes to/from logdirty won't work well with
+     * an IOMMU anyway, as IOMMU #PFs are not synchronous and will lead to
+     * aborts, and changes to/from ioreq_server are already fully flushed
+     * before returning to guest context (see
+     * XEN_DMOP_map_mem_type_to_ioreq_server).
+     */
+    if ( !new.recalc && iommu_use_hap_pt(p2m->domain) )
+        iommu_sync_cache(entryptr, sizeof(*entryptr));
+
     return 0;
 }
 
@@ -278,6 +291,9 @@ static bool_t ept_split_super_page(struct p2m_domain *p2m,
             break;
     }
 
+    if ( iommu_use_hap_pt(p2m->domain) )
+        iommu_sync_cache(table, EPT_PAGETABLE_ENTRIES * sizeof(ept_entry_t));
+
     unmap_domain_page(table);
 
     /* Even failed we should install the newly allocated ept page. */
@@ -337,6 +353,9 @@ static int ept_next_level(struct p2m_domain *p2m, bool_t 
read_only,
         if ( !next )
             return GUEST_TABLE_MAP_FAILED;
 
+        if ( iommu_use_hap_pt(p2m->domain) )
+            iommu_sync_cache(next, EPT_PAGETABLE_ENTRIES * 
sizeof(ept_entry_t));
+
         rc = atomic_write_ept_entry(p2m, ept_entry, e, next_level);
         ASSERT(rc == 0);
     }
@@ -821,7 +840,10 @@ out:
          need_modify_vtd_table )
     {
         if ( iommu_use_hap_pt(d) )
-            rc = iommu_pte_flush(d, gfn, &ept_entry->epte, order, 
vtd_pte_present);
+            rc = iommu_iotlb_flush(d, _dfn(gfn), (1u << order),
+                                   (iommu_flags ? IOMMU_FLUSHF_added : 0) |
+                                   (vtd_pte_present ? IOMMU_FLUSHF_modified
+                                                    : 0));
         else if ( need_iommu_pt_sync(d) )
             rc = iommu_flags ?
                 iommu_legacy_map(d, _dfn(gfn), mfn, order, iommu_flags) :
diff --git a/xen/drivers/passthrough/vtd/iommu.c 
b/xen/drivers/passthrough/vtd/iommu.c
index 2919ada738..1e61e93d3f 100644
--- a/xen/drivers/passthrough/vtd/iommu.c
+++ b/xen/drivers/passthrough/vtd/iommu.c
@@ -1884,53 +1884,6 @@ static int intel_iommu_lookup_page(struct domain *d, 
dfn_t dfn, mfn_t *mfn,
     return 0;
 }
 
-int iommu_pte_flush(struct domain *d, uint64_t dfn, uint64_t *pte,
-                    int order, int present)
-{
-    struct acpi_drhd_unit *drhd;
-    struct vtd_iommu *iommu = NULL;
-    struct domain_iommu *hd = dom_iommu(d);
-    bool_t flush_dev_iotlb;
-    int iommu_domid;
-    int rc = 0;
-
-    iommu_sync_cache(pte, sizeof(struct dma_pte));
-
-    for_each_drhd_unit ( drhd )
-    {
-        iommu = drhd->iommu;
-        if ( !test_bit(iommu->index, &hd->arch.iommu_bitmap) )
-            continue;
-
-        flush_dev_iotlb = !!find_ats_dev_drhd(iommu);
-        iommu_domid= domain_iommu_domid(d, iommu);
-        if ( iommu_domid == -1 )
-            continue;
-
-        rc = iommu_flush_iotlb_psi(iommu, iommu_domid,
-                                   __dfn_to_daddr(dfn),
-                                   order, !present, flush_dev_iotlb);
-        if ( rc > 0 )
-        {
-            iommu_flush_write_buffer(iommu);
-            rc = 0;
-        }
-    }
-
-    if ( unlikely(rc) )
-    {
-        if ( !d->is_shutting_down && printk_ratelimit() )
-            printk(XENLOG_ERR VTDPREFIX
-                   " d%d: IOMMU pages flush failed: %d\n",
-                   d->domain_id, rc);
-
-        if ( !is_hardware_domain(d) )
-            domain_crash(d);
-    }
-
-    return rc;
-}
-
 static int __init vtd_ept_page_compatible(struct vtd_iommu *iommu)
 {
     u64 ept_cap, vtd_cap = iommu->cap;
diff --git a/xen/include/asm-x86/iommu.h b/xen/include/asm-x86/iommu.h
index 864e025078..6c9d5e5632 100644
--- a/xen/include/asm-x86/iommu.h
+++ b/xen/include/asm-x86/iommu.h
@@ -97,10 +97,6 @@ static inline int iommu_adjust_irq_affinities(void)
            : 0;
 }
 
-/* While VT-d specific, this must get declared in a generic header. */
-int __must_check iommu_pte_flush(struct domain *d, u64 gfn, u64 *pte,
-                                 int order, int present);
-
 static inline bool iommu_supports_x2apic(void)
 {
     return iommu_init_ops && iommu_init_ops->supports_x2apic
--
generated by git-patchbot for /home/xen/git/xen.git#stable-4.13



 


Rackspace

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