[PATCH v3 03/15] mm: when adding a region, merge with region after as well as before

[Daniel Axtens]

On x86_64-efi (at least) regions seem to be added from top down. The mm
code will merge a new region with an existing region that comes
immediately before the new region. This allows larger allocations to be
satisfied that would otherwise be the case.

On powerpc-ieee1275, however, regions are added from bottom up. So if
we add 3x 32MB regions, we can still only satisfy a 32MB allocation,
rather than the 96MB allocation we might otherwise be able to satisfy.

 * Define 'post_size' as being bytes lost to the end of an allocation
   due to being given weird sizes from firmware that are not multiples
   of GRUB_MM_ALIGN.

 * Allow merging of regions immediately _after_ existing regions, not
   just before. As with the other approach, we create an allocated
   block to represent the new space and the pass it to grub_free() to
   get the metadata right.

Signed-off-by: Daniel Axtens <dja@axtens.net>
Tested-by: Stefan Berger <stefanb@linux.ibm.com>
Reviewed-by: Daniel Kiper <daniel.kiper@oracle.com>
---
v2: Thanks Daniel K for feedback.
v3: Fix alignment of comments (8 spaces vs tab)
---
 grub-core/kern/mm.c       | 123 +++++++++++++++++++++++---------------
 include/grub/mm_private.h |   9 +++
 2 files changed, 85 insertions(+), 47 deletions(-)

diff --git a/grub-core/kern/mm.c b/grub-core/kern/mm.c
index 079c28da7cdf..6e4e8f325a05 100644
--- a/grub-core/kern/mm.c
+++ b/grub-core/kern/mm.c
@@ -130,53 +130,81 @@ grub_mm_init_region (void *addr, grub_size_t size)
 
   /* Attempt to merge this region with every existing region */
   for (p = &grub_mm_base, q = *p; q; p = &(q->next), q = *p)
-    /*
-     * Is the new region immediately below an existing region? That
-     * is, is the address of the memory we're adding now (addr) + size
-     * of the memory we're adding (size) + the bytes we couldn't use
-     * at the start of the region we're considering (q->pre_size)
-     * equal to the address of q? In other words, does the memory
-     * looks like this?
-     *
-     * addr                          q
-     *   |----size-----|-q->pre_size-|<q region>|
-     */
-    if ((grub_uint8_t *) addr + size + q->pre_size == (grub_uint8_t *) q)
-      {
-       /*
-        * Yes, we can merge the memory starting at addr into the
-        * existing region from below. Align up addr to GRUB_MM_ALIGN
-        * so that our new region has proper alignment.
-        */
-       r = (grub_mm_region_t) ALIGN_UP ((grub_addr_t) addr, GRUB_MM_ALIGN);
-       /* Copy the region data across */
-       *r = *q;
-       /* Consider all the new size as pre-size */
-       r->pre_size += size;
-
-       /*
-        * If we have enough pre-size to create a block, create a
-        * block with it. Mark it as allocated and pass it to
-        * grub_free (), which will sort out getting it into the free
-        * list.
-        */
-       if (r->pre_size >> GRUB_MM_ALIGN_LOG2)
-         {
-           h = (grub_mm_header_t) (r + 1);
-           /* block size is pre-size converted to cells */
-           h->size = (r->pre_size >> GRUB_MM_ALIGN_LOG2);
-           h->magic = GRUB_MM_ALLOC_MAGIC;
-           /* region size grows by block size converted back to bytes */
-           r->size += h->size << GRUB_MM_ALIGN_LOG2;
-           /* adjust pre_size to be accurate */
-           r->pre_size &= (GRUB_MM_ALIGN - 1);
-           *p = r;
-           grub_free (h + 1);
-         }
-       /* Replace the old region with the new region */
-       *p = r;
-       return;
-      }
+    {
+      /*
+       * Is the new region immediately below an existing region? That
+       * is, is the address of the memory we're adding now (addr) + size
+       * of the memory we're adding (size) + the bytes we couldn't use
+       * at the start of the region we're considering (q->pre_size)
+       * equal to the address of q? In other words, does the memory
+       * looks like this?
+       *
+       * addr                          q
+       *   |----size-----|-q->pre_size-|<q region>|
+       */
+      if ((grub_uint8_t *) addr + size + q->pre_size == (grub_uint8_t *) q)
+        {
+          /*
+           * Yes, we can merge the memory starting at addr into the
+           * existing region from below. Align up addr to GRUB_MM_ALIGN
+           * so that our new region has proper alignment.
+           */
+          r = (grub_mm_region_t) ALIGN_UP ((grub_addr_t) addr, GRUB_MM_ALIGN);
+          /* Copy the region data across */
+          *r = *q;
+          /* Consider all the new size as pre-size */
+          r->pre_size += size;
+
+          /*
+           * If we have enough pre-size to create a block, create a
+           * block with it. Mark it as allocated and pass it to
+           * grub_free (), which will sort out getting it into the free
+           * list.
+           */
+          if (r->pre_size >> GRUB_MM_ALIGN_LOG2)
+            {
+              h = (grub_mm_header_t) (r + 1);
+              /* block size is pre-size converted to cells */
+              h->size = (r->pre_size >> GRUB_MM_ALIGN_LOG2);
+              h->magic = GRUB_MM_ALLOC_MAGIC;
+              /* region size grows by block size converted back to bytes */
+              r->size += h->size << GRUB_MM_ALIGN_LOG2;
+              /* adjust pre_size to be accurate */
+              r->pre_size &= (GRUB_MM_ALIGN - 1);
+              *p = r;
+              grub_free (h + 1);
+            }
+          /* Replace the old region with the new region */
+          *p = r;
+          return;
+        }
+
+      /*
+       * Is the new region immediately above an existing region? That
+       * is:
+       *   q                       addr
+       *   |<q region>|-q->post_size-|----size-----|
+       */
+      if ((grub_uint8_t *)q + sizeof(*q) + q->size + q->post_size ==
+         (grub_uint8_t *) addr)
+       {
+         /*
+          * Yes! Follow a similar pattern to above, but simpler.
+          * Our header starts at address - post_size, which should align us
+          * to a cell boundary.
+          */
+         h = (grub_mm_header_t) ((grub_uint8_t *)addr - q->post_size);
+         /* our size is the allocated size plus post_size, in cells */
+         h->size = (size + q->post_size) >> GRUB_MM_ALIGN_LOG2;
+         h->magic = GRUB_MM_ALLOC_MAGIC;
+         /* region size grows by block size converted back to bytes */
+         q->size += h->size << GRUB_MM_ALIGN_LOG2;
+         /* adjust new post_size to be accurate */
+         q->post_size = (q->post_size + size) & (GRUB_MM_ALIGN - 1);
+         grub_free (h + 1);
+         return;
+       }
+    }
 
   /* Allocate a region from the head.  */
   r = (grub_mm_region_t) ALIGN_UP ((grub_addr_t) addr, GRUB_MM_ALIGN);
@@ -195,6 +223,7 @@ grub_mm_init_region (void *addr, grub_size_t size)
   r->first = h;
   r->pre_size = (grub_addr_t) r - (grub_addr_t) addr;
   r->size = (h->size << GRUB_MM_ALIGN_LOG2);
+  r->post_size = size - r->size;
 
   /* Find where to insert this region. Put a smaller one before bigger ones,
      to prevent fragmentation.  */
diff --git a/include/grub/mm_private.h b/include/grub/mm_private.h
index d3f2321e14fb..64d88ee86915 100644
--- a/include/grub/mm_private.h
+++ b/include/grub/mm_private.h
@@ -81,8 +81,17 @@ typedef struct grub_mm_region
    */
   grub_size_t pre_size;
 
+  /*
+   * Likewise, the post-size is the number of bytes we wasted at the end
+   * of the allocation because it wasn't a multiple of GRUB_MM_ALIGN
+   */
+  grub_size_t post_size;
+
   /* How many bytes are in this region? (free and allocated) */
   grub_size_t size;
+
+  /* pad to a multiple of cell size */
+  char padding[3 * GRUB_CPU_SIZEOF_VOID_P];
 }
 *grub_mm_region_t;
 
-- 
2.32.0