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

[PATCH v3 21/23] AMD/IOMMU: replace all-contiguous page tables by superpage mappings


  • To: "xen-devel@xxxxxxxxxxxxxxxxxxxx" <xen-devel@xxxxxxxxxxxxxxxxxxxx>
  • From: Jan Beulich <jbeulich@xxxxxxxx>
  • Date: Mon, 10 Jan 2022 17:37:35 +0100
  • Arc-authentication-results: i=1; mx.microsoft.com 1; spf=pass smtp.mailfrom=suse.com; dmarc=pass action=none header.from=suse.com; dkim=pass header.d=suse.com; arc=none
  • Arc-message-signature: i=1; a=rsa-sha256; c=relaxed/relaxed; d=microsoft.com; s=arcselector9901; h=From:Date:Subject:Message-ID:Content-Type:MIME-Version:X-MS-Exchange-AntiSpam-MessageData-ChunkCount:X-MS-Exchange-AntiSpam-MessageData-0:X-MS-Exchange-AntiSpam-MessageData-1; bh=q21k2QRVQ9OpX2KFqNv7CXJl+zA4mrlUPekTbpVEMuw=; b=SJ1M0RER6D2D7JwcvFnzeexCvnh7pb7sWf1QxShwk5tuBjW/MSaHPrHnd2rkQSdWd48oJc+q29+5fNMGq84JqLhqtc1sO1wUllzBBggdB4iILEoa5sd8eXoMmPb2n2lzISy01QcFvsEo+ss/hlNM43g+O4Z9bwJLJZUJIhorvJCy7E2AkDPe7EyeV+uw1KkY1eJXvusl0SyPv/k3Su9g4byuVV4gvOXdNGiU9CStV8WWF57TW3Fjwz85BXEjxLYb+CUXk4ybdpoxNczrBgJfwAckJYnLW02bsvptkh8P12rViF+M5xMpXC3dWQGJxsRL6BWXLejbE3+k42wmHqem7A==
  • Arc-seal: i=1; a=rsa-sha256; s=arcselector9901; d=microsoft.com; cv=none; b=HNTYsTqPl0MOLUG5oL5WKuSnPzPUsOcULY9y25cUKFqwkL2uDZudg8NkI4NDcR6d0nlcsAlOjCN26KCkfkb5HF5X330Ix0hERhq6WZGNjK/uvTmibh85+0v6XWKQ8hIOpg9fljo4bKqMKXqssnqsjq+FqVNoMWypOw5O20x45zn15bRSiPPJhxR8ZsYkl0utAtrCvlWTflhIe01Rrer9gK5K+VH5iHbwHHJaKcg9bvdlI7JfZM9NZevImLlb0Vx/anBJSkZ/VoSjewkNrelZr2SmEKL+yHPhmtSlHthO+4yv50BJ0zTgfS6Q9IfOqncd3SBTBDgUTeeYGP2BK7WfMg==
  • Authentication-results: dkim=none (message not signed) header.d=none;dmarc=none action=none header.from=suse.com;
  • Cc: Andrew Cooper <andrew.cooper3@xxxxxxxxxx>, Paul Durrant <paul@xxxxxxx>, Roger Pau Monné <roger.pau@xxxxxxxxxx>
  • Delivery-date: Mon, 10 Jan 2022 16:37:48 +0000
  • List-id: Xen developer discussion <xen-devel.lists.xenproject.org>

When a page table ends up with all contiguous entries (including all
identical attributes), it can be replaced by a superpage entry at the
next higher level. The page table itself can then be scheduled for
freeing.

Signed-off-by: Jan Beulich <jbeulich@xxxxxxxx>
---
Unlike the freeing of all-empty page tables, this causes quite a bit of
back and forth for PV domains, due to their mapping/unmapping of pages
when they get converted to/from being page tables. It may therefore be
worth considering to delay re-coalescing a little, to avoid doing so
when the superpage would otherwise get split again pretty soon. But I
think this would better be the subject of a separate change anyway.

Of course this could also be helped by more "aware" kernel side
behavior: They could avoid immediately mapping freed page tables
writable again, in anticipation of re-using that same page for another
page table elsewhere.
---
v3: New.

--- a/xen/drivers/passthrough/amd/iommu_map.c
+++ b/xen/drivers/passthrough/amd/iommu_map.c
@@ -81,7 +81,8 @@ static union amd_iommu_pte set_iommu_pte
                                                  unsigned long dfn,
                                                  unsigned long next_mfn,
                                                  unsigned int level,
-                                                 bool iw, bool ir)
+                                                 bool iw, bool ir,
+                                                 bool *contig)
 {
     union amd_iommu_pte *table, *pde, old;
 
@@ -94,11 +95,15 @@ static union amd_iommu_pte set_iommu_pte
          old.iw != iw || old.ir != ir )
     {
         set_iommu_pde_present(pde, next_mfn, 0, iw, ir);
-        pt_update_contig_markers(&table->raw, pfn_to_pde_idx(dfn, level),
-                                 level, PTE_kind_leaf);
+        *contig = pt_update_contig_markers(&table->raw,
+                                           pfn_to_pde_idx(dfn, level),
+                                           level, PTE_kind_leaf);
     }
     else
+    {
         old.pr = false; /* signal "no change" to the caller */
+        *contig = false;
+    }
 
     unmap_domain_page(table);
 
@@ -346,6 +351,7 @@ int amd_iommu_map_page(struct domain *d,
 {
     struct domain_iommu *hd = dom_iommu(d);
     unsigned int level = (IOMMUF_order(flags) / PTE_PER_TABLE_SHIFT) + 1;
+    bool contig;
     int rc;
     unsigned long pt_mfn = 0;
     union amd_iommu_pte old;
@@ -386,8 +392,26 @@ int amd_iommu_map_page(struct domain *d,
 
     /* Install mapping */
     old = set_iommu_pte_present(pt_mfn, dfn_x(dfn), mfn_x(mfn), level,
-                                (flags & IOMMUF_writable),
-                                (flags & IOMMUF_readable));
+                                flags & IOMMUF_writable,
+                                flags & IOMMUF_readable, &contig);
+
+    while ( unlikely(contig) && ++level < hd->arch.amd.paging_mode )
+    {
+        struct page_info *pg = mfn_to_page(_mfn(pt_mfn));
+        unsigned long next_mfn;
+
+        if ( iommu_pde_from_dfn(d, dfn_x(dfn), level, &pt_mfn, flush_flags,
+                                false) )
+            BUG();
+        BUG_ON(!pt_mfn);
+
+        next_mfn = mfn_x(mfn) & (~0UL << (PTE_PER_TABLE_SHIFT * (level - 1)));
+        set_iommu_pte_present(pt_mfn, dfn_x(dfn), next_mfn, level,
+                              flags & IOMMUF_writable,
+                              flags & IOMMUF_readable, &contig);
+        *flush_flags |= IOMMU_FLUSHF_modified | IOMMU_FLUSHF_all;
+        iommu_queue_free_pgtable(d, pg);
+    }
 
     spin_unlock(&hd->arch.mapping_lock);
 




 


Rackspace

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