[kerneltech] Fwd: a question about the virtual address space memory layout
- From: Jiaqiang Xu <xjqhust03@xxxxxxxxx>
- To: kerneltech@xxxxxxxxxxxxx
- Date: Fri, 30 Sep 2011 16:38:20 +0800
I searched out of some explanations pertaining to this issue. It seems like
a recommendation to avoid potential buffer overflows. We are recommended to
use the top 16M for library mappings. However, it comes with another
question. Why do we use the top 16M address space which seems to be the
most possible area to get null pointer dereference. If we place library
text and data segment there, is it possible for our pointer dereference to
coincidentally get some data without any warning instead of a formal kernel
oops in default settings?
citation from stackoverflow
"
By default, on linux/x86, you won't see low addresses below 0x08000000 used
for much; although the kernel may use it if requested in a mmap call, or if
it runs out of room for mmaps. Additionally, there have been proposals to
use addresses in the 0x00000000 - 0x01000000 range for library mappings, to
make buffer overflows more difficult (by embedding a NUL byte to terminate
strings)."
---------- Forwarded message ----------
From: Jiaqiang Xu <xjqhust03@xxxxxxxxx>
Date: Fri, Sep 30, 2011 at 4:22 PM
Subject: a question about the virtual address space memory layout
To: kerneltech@xxxxxxxxxxxxx
I just noticed that in my local i386 machine with kernel version 2.6.22 the
ld and libc.so files are memory mapped between heap and stack area.
However in some other lab machines, file&anonymous memory mapping regions
are mapped before 0x08048000(in the top 128M region normally reserved for
catching null pointer dereference). I am wondering the reason why they did
this.
jq
Other related posts:
- » [kerneltech] Fwd: a question about the virtual address space memory layout - Jiaqiang Xu
