diff --git a/string/Makefile b/string/Makefile
index aaad743..f70dc7a 100644
--- a/string/Makefile
+++ b/string/Makefile
@@ -57,6 +57,8 @@ tests		:= tester inl-tester noinl-tester testcopy test-ffs	\
 tests-ifunc := $(strop-tests:%=test-%-ifunc)
 tests += $(tests-ifunc)
 
+xtests = tst-strcoll-overflow
+
 include ../Rules
 
 tester-ENV = LANGUAGE=C
diff --git a/string/strcoll_l.c b/string/strcoll_l.c
index 50ed84d..4ee101a 100644
--- a/string/strcoll_l.c
+++ b/string/strcoll_l.c
@@ -45,7 +45,7 @@
 typedef struct
 {
   int len;			/* Length of the current sequence.  */
-  int val;			/* Position of the sequence relative to the
+  size_t val;			/* Position of the sequence relative to the
 				   previous non-ignored sequence.  */
   size_t idxnow;		/* Current index in sequences.  */
   size_t idxmax;		/* Maximum index in sequences.  */
@@ -55,6 +55,12 @@ typedef struct
   const USTRING_TYPE *us;	/* The string.  */
   int32_t *idxarr;		/* Array to cache weight indices.  */
   unsigned char *rulearr;	/* Array to cache rules.  */
+  unsigned char rule;		/* Saved rule for the first sequence.  */
+  int32_t idx;			/* Index to weight of the current sequence.  */
+  int32_t save_idx;		/* Save looked up index of a forward
+				   sequence after the last backward
+				   sequence.  */
+  const USTRING_TYPE *back_us;	/* Beginning of the backward sequence.  */
 } coll_seq;
 
 /* Get next sequence.  The weight indices are cached, so we don't need to
@@ -64,7 +70,7 @@ get_next_seq_cached (coll_seq *seq, int nrules, int pass,
 		     const unsigned char *rulesets,
 		     const USTRING_TYPE *weights)
 {
-  int val = seq->val = 0;
+  size_t val = seq->val = 0;
   int len = seq->len;
   size_t backw_stop = seq->backw_stop;
   size_t backw = seq->backw;
@@ -146,7 +152,7 @@ get_next_seq (coll_seq *seq, int nrules, const unsigned char *rulesets,
 	      const USTRING_TYPE *extra, const int32_t *indirect)
 {
 #include WEIGHT_H
-  int val = seq->val = 0;
+  size_t val = seq->val = 0;
   int len = seq->len;
   size_t backw_stop = seq->backw_stop;
   size_t backw = seq->backw;
@@ -162,7 +168,7 @@ get_next_seq (coll_seq *seq, int nrules, const unsigned char *rulesets,
       ++val;
       if (backw_stop != ~0ul)
 	{
-	  /* The is something pushed.  */
+	  /* There is something pushed.  */
 	  if (backw == backw_stop)
 	    {
 	      /* The last pushed character was handled.  Continue
@@ -227,15 +233,199 @@ get_next_seq (coll_seq *seq, int nrules, const unsigned char *rulesets,
   seq->us = us;
 }
 
-/* Compare two sequences.  */
+/* Get next sequence.  Traverse the string as required.  This function does not
+   set or use any index or rule cache.  */
+static void
+get_next_seq_nocache (coll_seq *seq, int nrules, const unsigned char *rulesets,
+		      const USTRING_TYPE *weights, const int32_t *table,
+		      const USTRING_TYPE *extra, const int32_t *indirect,
+		      int pass)
+{
+#include WEIGHT_H
+  size_t val = seq->val = 0;
+  int len = seq->len;
+  size_t backw_stop = seq->backw_stop;
+  size_t backw = seq->backw;
+  size_t idxcnt = seq->idxcnt;
+  size_t idxmax = seq->idxmax;
+  int32_t idx = seq->idx;
+  const USTRING_TYPE *us = seq->us;
+
+  while (len == 0)
+    {
+      ++val;
+      if (backw_stop != ~0ul)
+	{
+	  /* There is something pushed.  */
+	  if (backw == backw_stop)
+	    {
+	      /* The last pushed character was handled.  Continue
+		 with forward characters.  */
+	      if (idxcnt < idxmax)
+		{
+		  idx = seq->save_idx;
+		  backw_stop = ~0ul;
+		}
+	      else
+		{
+		  /* Nothing anymore.  The backward sequence ended with
+		     the last sequence in the string.  Note that len is
+		     still zero.  */
+		  idx = 0;
+		  break;
+	        }
+	    }
+	  else
+	    {
+	      /* XXX Traverse BACKW sequences from the beginning of
+		 BACKW_STOP to get the next sequence.  Is ther a quicker way
+	         to do this?  */
+	      size_t i = backw_stop;
+	      us = seq->back_us;
+	      while (i < backw)
+		{
+		  int32_t tmp = findidx (&us, -1);
+		  idx = tmp & 0xffffff;
+		  i++;
+		}
+	      --backw;
+	      us = seq->us;
+	    }
+	}
+      else
+	{
+	  backw_stop = idxmax;
+	  int32_t prev_idx = idx;
+
+	  while (*us != L('\0'))
+	    {
+	      int32_t tmp = findidx (&us, -1);
+	      unsigned char rule = tmp >> 24;
+	      prev_idx = idx;
+	      idx = tmp & 0xffffff;
+	      idxcnt = idxmax++;
+
+	      /* Save the rule for the first sequence.  */
+	      if (__glibc_unlikely (idxcnt == 0))
+	        seq->rule = rule;
+
+	      if ((rulesets[rule * nrules + pass]
+		   & sort_backward) == 0)
+		/* No more backward characters to push.  */
+		break;
+	      ++idxcnt;
+	    }
+
+	  if (backw_stop >= idxcnt)
+	    {
+	      /* No sequence at all or just one.  */
+	      if (idxcnt == idxmax || backw_stop > idxcnt)
+		/* Note that len is still zero.  */
+		break;
+
+	      backw_stop = ~0ul;
+	    }
+	  else
+	    {
+	      /* We pushed backward sequences.  If the stream ended with the
+		 backward sequence, then we process the last sequence we
+		 found.  Otherwise we process the sequence before the last
+		 one since the last one was a forward sequence.  */
+	      seq->back_us = seq->us;
+	      seq->us = us;
+	      backw = idxcnt;
+	      if (idxmax > idxcnt)
+		{
+		  backw--;
+		  seq->save_idx = idx;
+		  idx = prev_idx;
+		}
+	      if (backw > backw_stop)
+		backw--;
+	    }
+	}
+
+      len = weights[idx++];
+      /* Skip over indices of previous levels.  */
+      for (int i = 0; i < pass; i++)
+	{
+	  idx += len;
+	  len = weights[idx];
+	  idx++;
+	}
+    }
+
+  /* Update the structure.  */
+  seq->val = val;
+  seq->len = len;
+  seq->backw_stop = backw_stop;
+  seq->backw = backw;
+  seq->idxcnt = idxcnt;
+  seq->idxmax = idxmax;
+  seq->us = us;
+  seq->idx = idx;
+}
+
+/* Compare two sequences.  This version does not use the index and rules
+   cache.  */
+static int
+do_compare_nocache (coll_seq *seq1, coll_seq *seq2, int position,
+		    const USTRING_TYPE *weights)
+{
+  int seq1len = seq1->len;
+  int seq2len = seq2->len;
+  size_t val1 = seq1->val;
+  size_t val2 = seq2->val;
+  int idx1 = seq1->idx;
+  int idx2 = seq2->idx;
+  int result = 0;
+
+  /* Test for position if necessary.  */
+  if (position && val1 != val2)
+    {
+      result = val1 > val2 ? 1 : -1;
+      goto out;
+    }
+
+  /* Compare the two sequences.  */
+  do
+    {
+      if (weights[idx1] != weights[idx2])
+	{
+	  /* The sequences differ.  */
+	  result = weights[idx1] - weights[idx2];
+	  goto out;
+	}
+
+      /* Increment the offsets.  */
+      ++idx1;
+      ++idx2;
+
+      --seq1len;
+      --seq2len;
+    }
+  while (seq1len > 0 && seq2len > 0);
+
+  if (position && seq1len != seq2len)
+    result = seq1len - seq2len;
+
+out:
+  seq1->len = seq1len;
+  seq2->len = seq2len;
+  seq1->idx = idx1;
+  seq2->idx = idx2;
+  return result;
+}
+
+/* Compare two sequences using the index cache.  */
 static int
 do_compare (coll_seq *seq1, coll_seq *seq2, int position,
 	    const USTRING_TYPE *weights)
 {
   int seq1len = seq1->len;
   int seq2len = seq2->len;
-  int val1 = seq1->val;
-  int val2 = seq2->val;
+  size_t val1 = seq1->val;
+  size_t val2 = seq2->val;
   int32_t *idx1arr = seq1->idxarr;
   int32_t *idx2arr = seq2->idxarr;
   int idx1now = seq1->idxnow;
@@ -245,7 +435,7 @@ do_compare (coll_seq *seq1, coll_seq *seq2, int position,
   /* Test for position if necessary.  */
   if (position && val1 != val2)
     {
-      result = val1 - val2;
+      result = val1 > val2 ? 1 : -1;
       goto out;
     }
 
@@ -334,57 +524,70 @@ STRCOLL (const STRING_TYPE *s1, const STRING_TYPE *s2, __locale_t l)
   memset (&seq1, 0, sizeof (seq1));
   seq2 = seq1;
 
-  /* We need the elements of the strings as unsigned values since they
-     are used as indices.  */
-  seq1.us = (const USTRING_TYPE *) s1;
-  seq2.us = (const USTRING_TYPE *) s2;
+  size_t size_max = SIZE_MAX / (sizeof (int32_t) + 1);
 
-  if (! __libc_use_alloca ((s1len + s2len) * (sizeof (int32_t) + 1)))
+  if (MIN (s1len, s2len) > size_max
+      || MAX (s1len, s2len) > size_max - MIN (s1len, s2len))
+    {
+      /* If the strings are long enough to cause overflow in the size request,
+         then skip the allocation and proceed with the non-cached routines.  */
+    }
+  else if (! __libc_use_alloca ((s1len + s2len) * (sizeof (int32_t) + 1)))
     {
       seq1.idxarr = (int32_t *) malloc ((s1len + s2len) * (sizeof (int32_t) + 1));
-      seq2.idxarr = &seq1.idxarr[s1len];
-      seq1.rulearr = (unsigned char *) &seq2.idxarr[s2len];
-      seq2.rulearr = &seq1.rulearr[s1len];
-
-      if (seq1.idxarr == NULL)
-	/* No memory.  Well, go with the stack then.
-
-	   XXX Once this implementation is stable we will handle this
-	   differently.  Instead of precomputing the indices we will
-	   do this in time.  This means, though, that this happens for
-	   every pass again.  */
-	goto try_stack;
-      use_malloc = true;
+
+      /* If we failed to allocate memory, we leave everything as NULL so that
+	 we use the nocache version of traversal and comparison functions.  */
+      if (seq1.idxarr != NULL)
+	{
+	  seq2.idxarr = &seq1.idxarr[s1len];
+	  seq1.rulearr = (unsigned char *) &seq2.idxarr[s2len];
+	  seq2.rulearr = &seq1.rulearr[s1len];
+	  use_malloc = true;
+	}
     }
   else
     {
-    try_stack:
       seq1.idxarr = (int32_t *) alloca (s1len * sizeof (int32_t));
       seq2.idxarr = (int32_t *) alloca (s2len * sizeof (int32_t));
       seq1.rulearr = (unsigned char *) alloca (s1len);
       seq2.rulearr = (unsigned char *) alloca (s2len);
     }
 
-  seq1.rulearr[0] = 0;
+  int rule = 0;
 
   /* Cache values in the first pass and if needed, use them in subsequent
      passes.  */
   for (int pass = 0; pass < nrules; ++pass)
     {
       seq1.idxcnt = 0;
+      seq1.idx = 0;
+      seq2.idx = 0;
       seq1.backw_stop = ~0ul;
       seq1.backw = ~0ul;
       seq2.idxcnt = 0;
       seq2.backw_stop = ~0ul;
       seq2.backw = ~0ul;
 
+      /* We need the elements of the strings as unsigned values since they
+	 are used as indices.  */
+      seq1.us = (const USTRING_TYPE *) s1;
+      seq2.us = (const USTRING_TYPE *) s2;
+
       /* We assume that if a rule has defined `position' in one section
 	 this is true for all of them.  */
-      int position = rulesets[seq1.rulearr[0] * nrules + pass] & sort_position;
+      int position = rulesets[rule * nrules + pass] & sort_position;
 
       while (1)
 	{
-	  if (pass == 0)
+	  if (__glibc_unlikely (seq1.idxarr == NULL))
+	    {
+	      get_next_seq_nocache (&seq1, nrules, rulesets, weights, table,
+				    extra, indirect, pass);
+	      get_next_seq_nocache (&seq2, nrules, rulesets, weights, table,
+				    extra, indirect, pass);
+	    }
+	  else if (pass == 0)
 	    {
 	      get_next_seq (&seq1, nrules, rulesets, weights, table, extra,
 			    indirect);
@@ -411,10 +614,18 @@ STRCOLL (const STRING_TYPE *s1, const STRING_TYPE *s2, __locale_t l)
 	      goto free_and_return;
 	    }
 
-	  result = do_compare (&seq1, &seq2, position, weights);
+	  if (__glibc_unlikely (seq1.idxarr == NULL))
+	    result = do_compare_nocache (&seq1, &seq2, position, weights);
+	  else
+	    result = do_compare (&seq1, &seq2, position, weights);
 	  if (result != 0)
 	    goto free_and_return;
 	}
+
+      if (__glibc_likely (seq1.rulearr != NULL))
+	rule = seq1.rulearr[0];
+      else
+	rule = seq1.rule;
     }
 
   /* Free the memory if needed.  */
diff --git a/string/tst-strcoll-overflow.c b/string/tst-strcoll-overflow.c
new file mode 100644
index 0000000..bb665ac
--- /dev/null
+++ b/string/tst-strcoll-overflow.c
@@ -0,0 +1,61 @@
+/* Copyright (C) 2013 Free Software Foundation, Inc.
+   This file is part of the GNU C Library.
+
+   The GNU C Library is free software; you can redistribute it and/or
+   modify it under the terms of the GNU Lesser General Public
+   License as published by the Free Software Foundation; either
+   version 2.1 of the License, or (at your option) any later version.
+
+   The GNU C Library is distributed in the hope that it will be useful,
+   but WITHOUT ANY WARRANTY; without even the implied warranty of
+   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+   Lesser General Public License for more details.
+
+   You should have received a copy of the GNU Lesser General Public
+   License along with the GNU C Library; if not, see
+   <http://www.gnu.org/licenses/>.  */
+
+#include <locale.h>
+#include <stdio.h>
+#include <stdint.h>
+#include <stdlib.h>
+#include <string.h>
+
+/* Verify that strcoll does not crash for large strings for which it cannot
+   cache weight lookup results.  The size is large enough to cause integer
+   overflows on 32-bit as well as buffer overflows on 64-bit.  The test should
+   work reasonably reliably when overcommit is disabled, but it obviously
+   depends on how much memory the system has.  There's a limitation to this
+   test in that it does not run to completion.  Actually collating such a
+   large string can take days and we can't have xcheck running that long.  For
+   that reason, we run the test for about 5 minutes and then assume that
+   everything is fine if there are no crashes.  */
+#define SIZE 0x40000000ul
+
+int
+do_test (void)
+{
+  if (setlocale (LC_COLLATE, "en_GB.UTF-8") == NULL)
+    {
+      puts ("setlocale failed, cannot test for overflow");
+      return 0;
+    }
+
+  char *p = malloc (SIZE);
+
+  if (p == NULL)
+    {
+      puts ("could not allocate memory");
+      return 1;
+    }
+
+  memset (p, 'x', SIZE - 1);
+  p[SIZE - 1] = 0;
+  printf ("%d\n", strcoll (p, p));
+  return 0;
+}
+
+#define TIMEOUT 300
+#define EXPECTED_SIGNAL SIGALRM
+#define TEST_FUNCTION do_test ()
+#include "../test-skeleton.c"