study_chunk: extract rck_elide_nothing

(CVE-2020-10878)

(cherry picked from commit a3a7598c8ec6efb0eb9c0b786d80c4d2a3751b70)

4 files changed, 48 insertions, 27 deletions

diff --git a/embed.fnc b/embed.fnc
index e762fe1eec..cf89277163 100644
--- a/embed.fnc
+++ b/embed.fnc
@@ -2477,6 +2477,7 @@ Es	|SSize_t|study_chunk	|NN RExC_state_t *pRExC_state \
                                 |I32 stopparen|U32 recursed_depth \
 				|NULLOK regnode_ssc *and_withp \
 				|U32 flags|U32 depth
+Es	|void	|rck_elide_nothing|NN regnode *node
 EsR	|SV *	|get_ANYOFM_contents|NN const regnode * n
 EsRn	|U32	|add_data	|NN RExC_state_t* const pRExC_state \
 				|NN const char* const s|const U32 n
diff --git a/embed.h b/embed.h
index a5416a1148..886551ce5c 100644
--- a/embed.h
+++ b/embed.h
@@ -1202,6 +1202,7 @@
 #define output_or_return_posix_warnings(a,b,c)	S_output_or_return_posix_warnings(aTHX_ a,b,c)
 #define parse_lparen_question_flags(a)	S_parse_lparen_question_flags(aTHX_ a)
 #define populate_ANYOF_from_invlist(a,b)	S_populate_ANYOF_from_invlist(aTHX_ a,b)
+#define rck_elide_nothing(a)	S_rck_elide_nothing(aTHX_ a)
 #define reg(a,b,c,d)		S_reg(aTHX_ a,b,c,d)
 #define reg2Lanode(a,b,c,d)	S_reg2Lanode(aTHX_ a,b,c,d)
 #define reg_node(a,b)		S_reg_node(aTHX_ a,b)
diff --git a/proto.h b/proto.h
index 66bb29b132..d3f8802c1d 100644
--- a/proto.h
+++ b/proto.h
@@ -5485,6 +5485,9 @@ STATIC void	S_parse_lparen_question_flags(pTHX_ RExC_state_t *pRExC_state);
 STATIC void	S_populate_ANYOF_from_invlist(pTHX_ regnode *node, SV** invlist_ptr);
 #define PERL_ARGS_ASSERT_POPULATE_ANYOF_FROM_INVLIST	\
 	assert(node); assert(invlist_ptr)
+STATIC void	S_rck_elide_nothing(pTHX_ regnode *node);
+#define PERL_ARGS_ASSERT_RCK_ELIDE_NOTHING	\
+	assert(node)
 PERL_STATIC_NO_RET void	S_re_croak2(pTHX_ bool utf8, const char* pat1, const char* pat2, ...)
 			__attribute__noreturn__;
 #define PERL_ARGS_ASSERT_RE_CROAK2	\
diff --git a/regcomp.c b/regcomp.c
index dd18add1db..0a9c6a8085 100644
--- a/regcomp.c
+++ b/regcomp.c
@@ -4178,6 +4178,44 @@ S_unwind_scan_frames(pTHX_ const void *p)
     } while (f);
 }
 
+/* Follow the next-chain of the current node and optimize away
+   all the NOTHINGs from it.
+ */
+STATIC void
+S_rck_elide_nothing(pTHX_ regnode *node)
+{
+    dVAR;
+
+    PERL_ARGS_ASSERT_RCK_ELIDE_NOTHING;
+
+    if (OP(node) != CURLYX) {
+        const int max = (reg_off_by_arg[OP(node)]
+                        ? I32_MAX
+                          /* I32 may be smaller than U16 on CRAYs! */
+                        : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
+        int off = (reg_off_by_arg[OP(node)] ? ARG(node) : NEXT_OFF(node));
+        int noff;
+        regnode *n = node;
+
+        /* Skip NOTHING and LONGJMP. */
+        while (
+            (n = regnext(n))
+            && (
+                (PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
+                || ((OP(n) == LONGJMP) && (noff = ARG(n)))
+            )
+            && off + noff < max
+        ) {
+            off += noff;
+        }
+        if (reg_off_by_arg[OP(node)])
+            ARG(node) = off;
+        else
+            NEXT_OFF(node) = off;
+    }
+    return;
+}
+
 /* the return from this sub is the minimum length that could possibly match */
 STATIC SSize_t
 S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
@@ -4277,28 +4315,10 @@ S_study_chunk(pTHX_ RExC_state_t *pRExC_state, regnode **scanp,
          */
         JOIN_EXACT(scan,&min_subtract, &unfolded_multi_char, 0);
 
-	/* Follow the next-chain of the current node and optimize
-	   away all the NOTHINGs from it.  */
-	if (OP(scan) != CURLYX) {
-	    const int max = (reg_off_by_arg[OP(scan)]
-		       ? I32_MAX
-		       /* I32 may be smaller than U16 on CRAYs! */
-		       : (I32_MAX < U16_MAX ? I32_MAX : U16_MAX));
-	    int off = (reg_off_by_arg[OP(scan)] ? ARG(scan) : NEXT_OFF(scan));
-	    int noff;
-	    regnode *n = scan;
-
-	    /* Skip NOTHING and LONGJMP. */
-	    while ((n = regnext(n))
-		   && ((PL_regkind[OP(n)] == NOTHING && (noff = NEXT_OFF(n)))
-		       || ((OP(n) == LONGJMP) && (noff = ARG(n))))
-		   && off + noff < max)
-		off += noff;
-	    if (reg_off_by_arg[OP(scan)])
-		ARG(scan) = off;
-	    else
-		NEXT_OFF(scan) = off;
-	}
+        /* Follow the next-chain of the current node and optimize
+           away all the NOTHINGs from it.
+         */
+        rck_elide_nothing(scan);
 
 	/* The principal pseudo-switch.  Cannot be a switch, since we
 	   look into several different things.  */
@@ -5425,11 +5445,7 @@ Perl_re_printf( aTHX_  "LHS=%" UVuf " RHS=%" UVuf "\n",
 		if (data && (fl & SF_HAS_EVAL))
 		    data->flags |= SF_HAS_EVAL;
 	      optimize_curly_tail:
-		if (OP(oscan) != CURLYX) {
-		    while (PL_regkind[OP(next = regnext(oscan))] == NOTHING
-			   && NEXT_OFF(next))
-			NEXT_OFF(oscan) += NEXT_OFF(next);
-		}
+		rck_elide_nothing(oscan);
 		continue;
 
 	    default: