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

[PATCH v3 1/4] xen/arm: Reduce redundant clear root pages when teardown p2m



Currently, p2m for a domain will be teardown from two paths:
(1) The normal path when a domain is destroyed.
(2) The arch_domain_destroy() in the failure path of domain creation.

When tearing down p2m from (1), the part to clear and clean the root
is only needed to do once rather than for every call of p2m_teardown().
If the p2m teardown is from (2), the clear and clean of the root
is unnecessary because the domain is not scheduled.

Therefore, this patch introduces a helper `p2m_clear_root_pages()` to
do the clear and clean of the root, and move this logic outside of
p2m_teardown(). With this movement, the `page_list_empty(&p2m->pages)`
check can be dropped.

Signed-off-by: Henry Wang <Henry.Wang@xxxxxxx>
Reviewed-by: Michal Orzel <michal.orzel@xxxxxxx>
Acked-by: Julien Grall <jgrall@xxxxxxxxxx>
---
v2 -> v3:
1. No change.
v1 -> v2:
1. Move in-code comment for p2m_force_tlb_flush_sync() on top of
   p2m_clear_root_pages().
2. Add Reviewed-by tag from Michal and Acked-by tag from Julien.
---
 xen/arch/arm/domain.c          |  8 +++++++
 xen/arch/arm/include/asm/p2m.h |  1 +
 xen/arch/arm/p2m.c             | 40 +++++++++++++++++-----------------
 3 files changed, 29 insertions(+), 20 deletions(-)

diff --git a/xen/arch/arm/domain.c b/xen/arch/arm/domain.c
index 99577adb6c..b8a4a60570 100644
--- a/xen/arch/arm/domain.c
+++ b/xen/arch/arm/domain.c
@@ -959,6 +959,7 @@ enum {
     PROG_xen,
     PROG_page,
     PROG_mapping,
+    PROG_p2m_root,
     PROG_p2m,
     PROG_p2m_pool,
     PROG_done,
@@ -1021,6 +1022,13 @@ int domain_relinquish_resources(struct domain *d)
         if ( ret )
             return ret;
 
+    PROGRESS(p2m_root):
+        /*
+         * We are about to free the intermediate page-tables, so clear the
+         * root to prevent any walk to use them.
+         */
+        p2m_clear_root_pages(&d->arch.p2m);
+
     PROGRESS(p2m):
         ret = p2m_teardown(d, true);
         if ( ret )
diff --git a/xen/arch/arm/include/asm/p2m.h b/xen/arch/arm/include/asm/p2m.h
index 91df922e1c..bf5183e53a 100644
--- a/xen/arch/arm/include/asm/p2m.h
+++ b/xen/arch/arm/include/asm/p2m.h
@@ -281,6 +281,7 @@ int p2m_set_entry(struct p2m_domain *p2m,
 
 bool p2m_resolve_translation_fault(struct domain *d, gfn_t gfn);
 
+void p2m_clear_root_pages(struct p2m_domain *p2m);
 void p2m_invalidate_root(struct p2m_domain *p2m);
 
 /*
diff --git a/xen/arch/arm/p2m.c b/xen/arch/arm/p2m.c
index 948f199d84..f1ccd7efbd 100644
--- a/xen/arch/arm/p2m.c
+++ b/xen/arch/arm/p2m.c
@@ -1314,6 +1314,26 @@ static void p2m_invalidate_table(struct p2m_domain *p2m, 
mfn_t mfn)
     p2m->need_flush = true;
 }
 
+/*
+ * The domain will not be scheduled anymore, so in theory we should
+ * not need to flush the TLBs. Do it for safety purpose.
+ * Note that all the devices have already been de-assigned. So we don't
+ * need to flush the IOMMU TLB here.
+ */
+void p2m_clear_root_pages(struct p2m_domain *p2m)
+{
+    unsigned int i;
+
+    p2m_write_lock(p2m);
+
+    for ( i = 0; i < P2M_ROOT_PAGES; i++ )
+        clear_and_clean_page(p2m->root + i);
+
+    p2m_force_tlb_flush_sync(p2m);
+
+    p2m_write_unlock(p2m);
+}
+
 /*
  * Invalidate all entries in the root page-tables. This is
  * useful to get fault on entry and do an action.
@@ -1698,30 +1718,10 @@ int p2m_teardown(struct domain *d, bool 
allow_preemption)
     struct p2m_domain *p2m = p2m_get_hostp2m(d);
     unsigned long count = 0;
     struct page_info *pg;
-    unsigned int i;
     int rc = 0;
 
-    if ( page_list_empty(&p2m->pages) )
-        return 0;
-
     p2m_write_lock(p2m);
 
-    /*
-     * We are about to free the intermediate page-tables, so clear the
-     * root to prevent any walk to use them.
-     */
-    for ( i = 0; i < P2M_ROOT_PAGES; i++ )
-        clear_and_clean_page(p2m->root + i);
-
-    /*
-     * The domain will not be scheduled anymore, so in theory we should
-     * not need to flush the TLBs. Do it for safety purpose.
-     *
-     * Note that all the devices have already been de-assigned. So we don't
-     * need to flush the IOMMU TLB here.
-     */
-    p2m_force_tlb_flush_sync(p2m);
-
     while ( (pg = page_list_remove_head(&p2m->pages)) )
     {
         p2m_free_page(p2m->domain, pg);
-- 
2.25.1




 


Rackspace

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