|
From: <sv...@va...> - 2005-10-15 00:41:54
|
Author: sewardj Date: 2005-10-15 01:41:50 +0100 (Sat, 15 Oct 2005) New Revision: 4925 Log: Add comments re loss of pointercheck. Modified: trunk/coregrind/m_aspacemgr/aspacemgr.c Modified: trunk/coregrind/m_aspacemgr/aspacemgr.c =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D --- trunk/coregrind/m_aspacemgr/aspacemgr.c 2005-10-14 19:05:45 UTC (rev = 4924) +++ trunk/coregrind/m_aspacemgr/aspacemgr.c 2005-10-15 00:41:50 UTC (rev = 4925) @@ -253,6 +253,16 @@ complains if there is a difference. --sanity-level=3D3 runs it before and after each syscall, which is a powerful, if slow way of finding buggy syscall wrappers. + + Loss of pointercheck + ~~~~~~~~~~~~~~~~~~~~ + Up to and including Valgrind 2.4.1, x86 segmentation was used to + enforce seperation of V and C, so that wild writes by C could not + trash V. This got called "pointercheck". Unfortunately, the new + more flexible memory layout, plus the need to be portable across + different architectures, means doing this in hardware is no longer + viable, and doing it in software is expensive. So at the moment we + don't do it at all. */ =20 =20 |