[committed v2.37] realloc: Limit chunk reuse to only growing requests [BZ #30579]

[committed v2.37] realloc: Limit chunk reuse to only growing requests [BZ #30579]

[Siddhesh Poyarekar]

The trim_threshold is too aggressive a heuristic to decide if chunk
reuse is OK for reallocated memory; for repeated small, shrinking
allocations it leads to internal fragmentation and for repeated larger
allocations that fragmentation may blow up even worse due to the dynamic
nature of the threshold.

Limit reuse only when it is within the alignment padding, which is 2 *
size_t for heap allocations and a page size for mmapped allocations.
There's the added wrinkle of THP, but this fix ignores it for now,
pessimizing that case in favor of keeping fragmentation low.

This resolves BZ #30579.

Signed-off-by: Siddhesh Poyarekar <siddhesh@sourceware.org>
Reported-by: Nicolas Dusart <nicolas@freedelity.be>
Reported-by: Aurelien Jarno <aurelien@aurel32.net>
Reviewed-by: Aurelien Jarno <aurelien@aurel32.net>
Tested-by: Aurelien Jarno <aurelien@aurel32.net>
(cherry picked from commit 2fb12bbd092b0c10f1f2083216e723d2406e21c4)
---
 malloc/malloc.c | 23 +++++++++++++++--------
 1 file changed, 15 insertions(+), 8 deletions(-)

diff --git a/malloc/malloc.c b/malloc/malloc.c
index fd8b52bfac..67df9f8c51 100644
--- a/malloc/malloc.c
+++ b/malloc/malloc.c
@@ -3398,16 +3398,23 @@ __libc_realloc (void *oldmem, size_t bytes)
   if (__glibc_unlikely (mtag_enabled))
     *(volatile char*) oldmem;
 
-  /* Return the chunk as is whenever possible, i.e. there's enough usable space
-     but not so much that we end up fragmenting the block.  We use the trim
-     threshold as the heuristic to decide the latter.  */
-  size_t usable = musable (oldmem);
-  if (bytes <= usable
-      && (unsigned long) (usable - bytes) <= mp_.trim_threshold)
-    return oldmem;
-
   /* chunk corresponding to oldmem */
   const mchunkptr oldp = mem2chunk (oldmem);
+
+  /* Return the chunk as is if the request grows within usable bytes, typically
+     into the alignment padding.  We want to avoid reusing the block for
+     shrinkages because it ends up unnecessarily fragmenting the address space.
+     This is also why the heuristic misses alignment padding for THP for
+     now.  */
+  size_t usable = musable (oldmem);
+  if (bytes <= usable)
+    {
+      size_t difference = usable - bytes;
+      if ((unsigned long) difference < 2 * sizeof (INTERNAL_SIZE_T)
+	  || (chunk_is_mmapped (oldp) && difference <= GLRO (dl_pagesize)))
+	return oldmem;
+    }
+
   /* its size */
   const INTERNAL_SIZE_T oldsize = chunksize (oldp);
 
-- 
2.41.0