Andrej,
Thanks for your information. My setup includes glibc 2.20 rather than the 2.16
referenced in your patch file, so I don't think it'll work directly (the source
of 2.20 and 2.16 look pretty different - actually, I'm not even seeing the
files referenced in the 2.16 source I pulled from the glibc archive today), but
perhaps it'll help me find the answers I need.
My results from readelf on libc.so.6 look a bit different:
> 14499: 00058f21 0 FUNC GLOBAL DEFAULT 11 memcpy
> 12773: 00056f11 740 FUNC GLOBAL DEFAULT 11 strcmp
> 14536: 00057701 220 FUNC GLOBAL DEFAULT 11 strlen
there are no "FUNC LOCAL DEFAULT" for the base functions - though memcpy
does have a size of 0 here.
My ld-2.20.so (which is the file valgrind referenced) yeilds similar results to
yours:
> 916: 00012921 0 FUNC LOCAL DEFAULT 10 memcpy
> 1033: 00011b31 740 FUNC LOCAL DEFAULT 10 strcmp
> 1083: 00012041 220 FUNC LOCAL DEFAULT 10 strlen
with the exception of strcmp and strlen. Which makes sense given the valgrind
output in my last message. Replacing ld-2.20.so moved the error from strcmp to
memcpy.
I'm actually seeing similar entries ('0 FUNC LOCAL DEFAULT') for memcpy,
memchr, index, memset, and strchr in ld-2.20.so.
- Jason
-----Original Message-----
From: ptxdist [mailto:[email protected]] On Behalf Of
[email protected]
Sent: Wednesday, March 28, 2018 6:02 AM
To: [email protected]
Subject: Re: [ptxdist] Including valgrind in the configuration doesn't trigger
an un-optimized build of glibc.
Hi Jason,
we had experienced a similar problem with the OSELAS 2012.12.1 toolchain.
In our case the problem was the missing .size directive in assembler
reimplementations of strlen and memcpy from the linaro cortex-strings patch for
glibc.
Does your "readelf -s /path/to/your/libc.so.6" output look like this?
...
828: 00019491 0 FUNC LOCAL DEFAULT 10 memcpy ...
949: 00018690 164 FUNC LOCAL DEFAULT 10 strcmp ...
996: 000189d1 0 FUNC LOCAL DEFAULT 10 strlen ...
Note the zeroes in the 3rd column. If so, you may have the same problem.
You may find our workaround in the attachment. The patch has to be applied
after the Linaro NEON patch. No idea why strcmp also had to be touched though.
Andrej
-----Ursprüngliche Nachricht-----
Von: ptxdist [mailto:[email protected]] Im Auftrag von Crim, Jason
Gesendet: Dienstag, 27. März 2018 1:54
An: [email protected]
Betreff: Re: [ptxdist] Including valgrind in the configuration doesn't trigger
an un-optimized build of glibc.
Michael,
It looks like the toolchain is configured as OSELAS.Toolchain-2014.12. I've
recently been added to a team which had this existing ptxdist setup, so I've
been trying to learn my way around ptxdist while trying to get valgrind enabled
to test an issue.
The initial run of valgrind results in the message shown in my first message:
>
> ==5596== Memcheck, a memory error detector
> ==5596== Copyright (C) 2002-2015, and GNU GPL'd, by Julian Seward et al.
> ==5596== Using Valgrind-3.11.0 and LibVEX; rerun with -h for copyright
> info
> ==5596== Command: /usr/bin/device
> ==5596==
>
> valgrind: Fatal error at startup: a function redirection
> valgrind: which is mandatory for this platform-tool combination
> valgrind: cannot be set up. Details of the redirection are:
> valgrind:
> valgrind: A must-be-redirected function
> valgrind: whose name matches the pattern: strcmp
> valgrind: in an object with soname matching: ld-linux-armhf.so.3
> valgrind: was not found whilst processing
> valgrind: symbols from the object with soname: ld-linux-armhf.so.3
> valgrind:
> valgrind: Possible fixes: (1, short term): install glibc's debuginfo
> valgrind: package on this machine. (2, longer term): ask the packagers
> valgrind: for your Linux distribution to please in future ship a non-
> valgrind: stripped ld.so (or whatever the dynamic linker .so is called)
> valgrind: that exports the above-named function using the standard
> valgrind: calling conventions for this platform. The package you need
> valgrind: to install for fix (1) is called
> valgrind:
> valgrind: On Debian, Ubuntu: libc6-dbg
> valgrind: On SuSE, openSuSE, Fedora, RHEL: glibc-debuginfo
> valgrind:
> valgrind: Cannot continue -- exiting now. Sorry.
Since ld-linux-armhf.so.3 is a symlink to ld-2.20.so, I copied ld-2.20.so from
the root-debug/lib location and re-ran valgrind. The referenced function
changed, but the file it expected it in did not (for brevity, I've only
retained the function and file lines).
>
> ...
> valgrind: A must-be-redirected function
> valgrind: whose name matches the pattern: memcpy
> valgrind: in an object with soname matching: ld-linux-armhf.so.3
> valgrind: was not found whilst processing
> valgrind: symbols from the object with soname: ld-linux-armhf.so.3
> ...
The fact that the reference was expected in the same file, along with
additional searching, led me to think that the optimization was the problem.
Thanks,
- Jason
-----Original Message-----
From: ptxdist [mailto:[email protected]] On Behalf Of Michael
Olbrich
Sent: Tuesday, March 27, 2018 3:34 AM
To: [email protected]
Subject: Re: [ptxdist] Including valgrind in the configuration doesn't trigger
an un-optimized build of glibc.
Hi,
On Mon, Mar 26, 2018 at 07:03:21AM -0500, Crim, Jason wrote:
> I'd tried moving over the root-debug versions of the referenced files,
> and was still hitting the error. That's when I'd found the post
> referencing the optimization levels. I'll take a look at getting the
> nfs-root running when I've got some time.
Hmm, strage, The only problems I've seen are with stripped binaries. Which
toolchain are you using?
Michael
--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |
_______________________________________________
ptxdist mailing list
[email protected]
_______________________________________________
ptxdist mailing list
[email protected]
_______________________________________________
ptxdist mailing list
[email protected]