Pass sys_cacheflush along to VG_(discard_translations), so we have

transparent zero-cost self-modifying code support on ARM.

git-svn-id: svn://svn.valgrind.org/valgrind/trunk@11018

diff --git a/coregrind/m_syswrap/syswrap-arm-linux.c b/coregrind/m_syswrap/syswrap-arm-linux.c
index c566edcd1c420977d9fe7abb682f27a47d84e624..88ed10a76a54b8bb83468e60e0e0ae7931d4019e 100644 (file)
--- a/coregrind/m_syswrap/syswrap-arm-linux.c
+++ b/coregrind/m_syswrap/syswrap-arm-linux.c
@@ -51,6 +51,7 @@
 #include "pub_core_syswrap.h"
 #include "pub_core_tooliface.h"
 #include "pub_core_stacks.h"        // VG_(register_stack)
+#include "pub_core_transtab.h"      // VG_(discard_translations)
 
 #include "priv_types_n_macros.h"
 #include "priv_syswrap-generic.h"   /* for decls of generic wrappers */
@@ -1214,6 +1215,10 @@ PRE(sys_cacheflush)
 {
    PRINT("cacheflush (%lx, %#lx, %#lx)",ARG1,ARG2,ARG3);
    PRE_REG_READ3(long, "cacheflush", void*, addrlow,void*, addrhigh,int, flags);
+   VG_(discard_translations)( (Addr64)ARG1,
+                              ((ULong)ARG2) - ((ULong)ARG1) + 1ULL/*paranoia*/,
+                              "PRE(sys_cacheflush)" );
+   SET_STATUS_Success(0);
 }