[Xen-devel] [PATCH v5 2/8] mm: Extract allocation loop from alloc_heap_pages()

[Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>]

This will make code a bit more readable, especially with changes that
will be introduced in subsequent patches.

Signed-off-by: Boris Ostrovsky <boris.ostrovsky@xxxxxxxxxx>
---
Changes in v5:
* Constified get_free_buddy()'s struct domain argument
* Dropped request local variable in get_free_buddy().

Because of rebasing there were few more changes in this patch so I decided not
to keep Jan's ACK.


 xen/common/page_alloc.c | 143 ++++++++++++++++++++++++++----------------------
 1 file changed, 79 insertions(+), 64 deletions(-)

diff --git a/xen/common/page_alloc.c b/xen/common/page_alloc.c
index 570d1f7..89fe3ce 100644
--- a/xen/common/page_alloc.c
+++ b/xen/common/page_alloc.c
@@ -694,22 +694,15 @@ static void page_list_add_scrub(struct page_info *pg, 
unsigned int node,
         page_list_add(pg, &heap(node, zone, order));
 }
 
-/* Allocate 2^@order contiguous pages. */
-static struct page_info *alloc_heap_pages(
-    unsigned int zone_lo, unsigned int zone_hi,
-    unsigned int order, unsigned int memflags,
-    struct domain *d)
+static struct page_info *get_free_buddy(unsigned int zone_lo,
+                                        unsigned int zone_hi,
+                                        unsigned int order, unsigned int 
memflags,
+                                        const struct domain *d)
 {
-    unsigned int i, j, zone = 0, nodemask_retry = 0;
     nodeid_t first_node, node = MEMF_get_node(memflags), req_node = node;
-    unsigned long request = 1UL << order;
-    struct page_info *pg, *first_dirty_pg = NULL;
-    nodemask_t nodemask = (d != NULL ) ? d->node_affinity : node_online_map;
-    bool_t need_tlbflush = 0;
-    uint32_t tlbflush_timestamp = 0;
-
-    /* Make sure there are enough bits in memflags for nodeID. */
-    BUILD_BUG_ON((_MEMF_bits - _MEMF_node) < (8 * sizeof(nodeid_t)));
+    nodemask_t nodemask = d ? d->node_affinity : node_online_map;
+    unsigned int j, zone, nodemask_retry = 0;
+    struct page_info *pg;
 
     if ( node == NUMA_NO_NODE )
     {
@@ -725,34 +718,6 @@ static struct page_info *alloc_heap_pages(
     first_node = node;
 
     ASSERT(node < MAX_NUMNODES);
-    ASSERT(zone_lo <= zone_hi);
-    ASSERT(zone_hi < NR_ZONES);
-
-    if ( unlikely(order > MAX_ORDER) )
-        return NULL;
-
-    spin_lock(&heap_lock);
-
-    /*
-     * Claimed memory is considered unavailable unless the request
-     * is made by a domain with sufficient unclaimed pages.
-     */
-    if ( (outstanding_claims + request >
-          total_avail_pages + tmem_freeable_pages()) &&
-          ((memflags & MEMF_no_refcount) ||
-           !d || d->outstanding_pages < request) )
-        goto not_found;
-
-    /*
-     * TMEM: When available memory is scarce due to tmem absorbing it, allow
-     * only mid-size allocations to avoid worst of fragmentation issues.
-     * Others try tmem pools then fail.  This is a workaround until all
-     * post-dom0-creation-multi-page allocations can be eliminated.
-     */
-    if ( ((order == 0) || (order >= 9)) &&
-         (total_avail_pages <= midsize_alloc_zone_pages) &&
-         tmem_freeable_pages() )
-        goto try_tmem;
 
     /*
      * Start with requested node, but exhaust all node memory in requested 
@@ -764,17 +729,17 @@ static struct page_info *alloc_heap_pages(
         zone = zone_hi;
         do {
             /* Check if target node can support the allocation. */
-            if ( !avail[node] || (avail[node][zone] < request) )
+            if ( !avail[node] || (avail[node][zone] < (1UL << order)) )
                 continue;
 
             /* Find smallest order which can satisfy the request. */
             for ( j = order; j <= MAX_ORDER; j++ )
                 if ( (pg = page_list_remove_head(&heap(node, zone, j))) )
-                    goto found;
+                    return pg;
         } while ( zone-- > zone_lo ); /* careful: unsigned zone may wrap */
 
         if ( (memflags & MEMF_exact_node) && req_node != NUMA_NO_NODE )
-            goto not_found;
+            return NULL;
 
         /* Pick next node. */
         if ( !node_isset(node, nodemask) )
@@ -791,39 +756,89 @@ static struct page_info *alloc_heap_pages(
         {
             /* When we have tried all in nodemask, we fall back to others. */
             if ( (memflags & MEMF_exact_node) || nodemask_retry++ )
-                goto not_found;
+                return NULL;
             nodes_andnot(nodemask, node_online_map, nodemask);
             first_node = node = first_node(nodemask);
             if ( node >= MAX_NUMNODES )
-                goto not_found;
+                return NULL;
         }
     }
+}
 
- try_tmem:
-    /* Try to free memory from tmem */
-    if ( (pg = tmem_relinquish_pages(order, memflags)) != NULL )
+/* Allocate 2^@order contiguous pages. */
+static struct page_info *alloc_heap_pages(
+    unsigned int zone_lo, unsigned int zone_hi,
+    unsigned int order, unsigned int memflags,
+    struct domain *d)
+{
+    nodeid_t node;
+    unsigned int i, buddy_order, zone;
+    unsigned long request = 1UL << order;
+    struct page_info *pg, *first_dirty_pg = NULL;
+    bool_t need_tlbflush = 0;
+    uint32_t tlbflush_timestamp = 0;
+
+    /* Make sure there are enough bits in memflags for nodeID. */
+    BUILD_BUG_ON((_MEMF_bits - _MEMF_node) < (8 * sizeof(nodeid_t)));
+
+    ASSERT(zone_lo <= zone_hi);
+    ASSERT(zone_hi < NR_ZONES);
+
+    if ( unlikely(order > MAX_ORDER) )
+        return NULL;
+
+    spin_lock(&heap_lock);
+
+    /*
+     * Claimed memory is considered unavailable unless the request
+     * is made by a domain with sufficient unclaimed pages.
+     */
+    if ( (outstanding_claims + request >
+          total_avail_pages + tmem_freeable_pages()) &&
+          ((memflags & MEMF_no_refcount) ||
+           !d || d->outstanding_pages < request) )
     {
-        /* reassigning an already allocated anonymous heap page */
         spin_unlock(&heap_lock);
-        return pg;
+        return NULL;
+    }
+ 
+    /*
+     * TMEM: When available memory is scarce due to tmem absorbing it, allow
+     * only mid-size allocations to avoid worst of fragmentation issues.
+     * Others try tmem pools then fail.  This is a workaround until all
+     * post-dom0-creation-multi-page allocations can be eliminated.
+     */
+    if ( ((order == 0) || (order >= 9)) &&
+         (total_avail_pages <= midsize_alloc_zone_pages) &&
+         tmem_freeable_pages() )
+     {
+        /* Try to free memory from tmem. */
+         pg = tmem_relinquish_pages(order, memflags);
+         spin_unlock(&heap_lock);
+         return pg;
+     }
+ 
+    pg = get_free_buddy(zone_lo, zone_hi, order, memflags, d);
+    if ( !pg )
+    {
+        /* No suitable memory blocks. Fail the request. */
+        spin_unlock(&heap_lock);
+        return NULL;
     }
 
- not_found:
-    /* No suitable memory blocks. Fail the request. */
-    spin_unlock(&heap_lock);
-    return NULL;
-
- found: 
+    node = phys_to_nid(page_to_maddr(pg));
+    zone = page_to_zone(pg);
+    buddy_order = PFN_ORDER(pg);
 
     if ( pg->u.free.first_dirty != INVALID_DIRTY_IDX )
         first_dirty_pg = pg + pg->u.free.first_dirty;
-
-    /* We may have to halve the chunk a number of times. */
-    while ( j != order )
+ 
+     /* We may have to halve the chunk a number of times. */
+    while ( buddy_order != order )
     {
         unsigned int first_dirty;
 
-        if ( first_dirty_pg && ((pg + (1 << j)) > first_dirty_pg) )
+        if ( first_dirty_pg && ((pg + (1 << buddy_order)) > first_dirty_pg) )
         {
             if ( pg < first_dirty_pg )
                 first_dirty = (first_dirty_pg - pg) / sizeof(*pg);
@@ -833,8 +848,8 @@ static struct page_info *alloc_heap_pages(
         else
             first_dirty = INVALID_DIRTY_IDX;
 
-        page_list_add_scrub(pg, node, zone, --j, first_dirty);
-        pg += 1 << j;
+        page_list_add_scrub(pg, node, zone, --buddy_order, first_dirty);
+        pg += 1 << buddy_order;
     }
 
     ASSERT(avail[node][zone] >= request);
-- 
1.8.3.1


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel