Patches for the next release
Steve M. Robbins
steve at sumost.ca
Sat Aug 2 18:47:34 CEST 2008
On Sat, Aug 02, 2008 at 09:33:44AM +0200, Torbjorn Granlund wrote:
> > Why does only assmbly code suffer from this problem, why isn't C code
> > also affected?
> Good question. I hadn't considered this, but google led me to the
> When you compile source code normally, gcc takes care of adding
> the GNU_STACK markings so that the final object code is not marked
> with an executable stack unless it actually needs it. However, if
> you compile assembly code, gcc will not automatically add
> GNU_STACK markings. So, the most common source of executable
> stacks in ELF binaries are packages which include raw assembly
> code. Note that we're not talking about inline assembly code, but
> rather files like .S which are written in pure assembler.
> We can either patch each source file written in assembler and send
> the fixes upstream, or we can force the package build system to
> assemble the source files with the GNU as option --noexecstack
> Your quotation stopped abruptly just before "(but this is highly
> discouraged)". :-)
Yes, because the writer is adopting a Gentoo build system point of
view, which is not relevant here. The annotation he's advocating is
of the form
#if defined(__linux__) && defined(__ELF__)
which works only for linux. The patch I passed on works for *any*
platform using the GNU assembler.
The reason why the Gentoo guy discourages using --noexecstack is that
his solution is to put --noexecstack in the gentoo "ebuild" file
(whatever that is) meaning that it is a Gentoo-only solution. In this
light, he's right: better to have upstream patch it for all linux
users rather than just Gentoo users.
However, the patch I provided *is* for upstream so it benefits every
GNU user (not just linux user).
> I suppose we should really implement their suggested solution, i.e.,
> annotate the source. For for 4.2.3, the --noexecstack fix will have
> to do.
If the source annotation is for linux only, it would still meet my
needs. However, I think it would be a regression since the patch
provided works for all GNU assembler systems.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Size: 189 bytes
Desc: Digital signature
Url : http://gmplib.org/list-archives/gmp-bugs/attachments/20080802/6953de54/attachment.bin
More information about the gmp-bugs