Re: [RFT] binutils 2.27

[Khem Raj <raj.khem@gmail.com>]

[-- Attachment #1.1: Type: text/plain, Size: 1544 bytes --]


> On Aug 8, 2016, at 4:16 PM, Burton, Ross <ross.burton@intel.com> wrote:
> 
> 
> On 8 August 2016 at 16:59, Richard Purdie <richard.purdie@linuxfoundation.org <mailto:richard.purdie@linuxfoundation.org>> wrote:
> FWIW, I think the connmand segfault on mips and powerpc is hinted at
> with this bug:
> 
> https://bugs.launchpad.net/ubuntu/+source/lua5.3/+bug/1570055 <https://bugs.launchpad.net/ubuntu/+source/lua5.3/+bug/1570055>
> 
> connmand is a binary linked with a linker version-script. If I build
> connmand without that option, it doesn't segfault, at least on mips.
> 
> The binutils manuals say you shouldn't build binaries with version
> scripts:
> 
> http://ftp.gnu.org/old-gnu/Manuals/ld-2.9.1/html_node/ld_25.html <http://ftp.gnu.org/old-gnu/Manuals/ld-2.9.1/html_node/ld_25.html>
> 
> Its suspected connman might do this to limit the function exposure to
> its plugins.
> 
> Now sure exactly what needs to happen here without more research but it
> certainly hits at where the problem is. I suspect its the same issue on
> ppc.
> 
> I've verified that with binutils 2.26 on x86-64, a minimal test case for an executable using a version script to limit exported symbols to dlopen()d modules does in fact work.  A toolchain is building now to see what happens in different environments.

I could also see it on ppc. backtrace, shows the segfault is in exit path and happens in libc
at this point, I think the problem is how libc is compiled with binutils 2.27, connman itself
is ok.

> 
> Ross


[-- Attachment #1.2: Type: text/html, Size: 2920 bytes --]

[-- Attachment #2: Message signed with OpenPGP using GPGMail --]
[-- Type: application/pgp-signature, Size: 211 bytes --]