Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 5 Jun 2019 20:24:31 -0400
From: Rich Felker <dalias@...c.org>
To: "tenspd137 ." <dcday137@...il.com>
Cc: musl@...ts.openwall.com
Subject: Re: Illegal instruction in __copy_tls()

On Wed, Jun 05, 2019 at 05:37:49PM -0600, tenspd137 . wrote:
> bummer....  Hmm - I wonder if an i686 cross compiler would work?

Yes, you can do that. It's just that gcc's multilib dir layout is not
compatible with musl -- it assumes library headers can be shared
between 32- and 64-bit versions of "the same" arch rather than
treating them as completely separate things.

Rich


> On Wed, Jun 5, 2019 at 5:34 PM Michael Everitt <m.j.everitt@....org> wrote:
> >
> > Musl doesn't support multilib ... :] fyi
> >
> > On 06/06/19 00:32, tenspd137 . wrote:
> > > Quick question - anyone know how to make a crossdev crosscompiler
> > > multilib so I can use -m32 when I compile?
> > >
> > > Thanks!
> > >
> > > On Wed, Jun 5, 2019 at 5:09 PM tenspd137 . <dcday137@...il.com> wrote:
> > >> I probably don't fully understand how it works.  In
> > >> /etc/x86_64-pc-linux-musl/etc.portage/make.conf, march is not
> > >> specified.  It is specified in /etc/portage/make.conf.  When I removed
> > >> it from there, things worked.  Then I just put it back after using
> > >> crossdev to make x86_64-pc-linux-musl-*.
> > >>
> > >> Either way, loads of fun :)
> > >>
> > >> Thanks!
> > >>
> > >> -C
> > >>
> > >> On Wed, Jun 5, 2019 at 3:56 PM Michael Everitt <m.j.everitt@....org> wrote:
> > >>> In your /usr/${CTARGET}/etc/portage/make.conf you can set appropriate
> > >>> target flags there. But using 'march=native' is going to catch you out a
> > >>> lot cross-compiling anything ... :) At least if you've got it set to
> > >>> something, you're going to see errors that point immediately to your CFLAGS
> > >>> when/if it goes wrong!
> > >>>
> > >>> Michael.
> > >>>
> > >>> On 05/06/19 22:27, tenspd137 . wrote:
> > >>>> Thanks!  That was indeed the problem.  I had march=native for all
> > >>>> portage related stuff set, and my guess is that crossdev read that to
> > >>>> start.  I removed that, reran the crossdev tool to buiild the cross
> > >>>> toolchain, and can now build static binaries that run on different
> > >>>> systems.
> > >>>>
> > >>>> Much appreciated!
> > >>>>
> > >>>> -C
> > >>>>
> > >>>> On Wed, Jun 5, 2019 at 2:07 PM Rich Felker <dalias@...c.org> wrote:
> > >>>>> On Wed, Jun 05, 2019 at 01:53:58PM -0600, tenspd137 . wrote:
> > >>>>>> Sure -
> > >>>>>>
> > >>>>>> strace -d -v ./a.out:
> > >>>>>> strace: ptrace_setoptions = 0x51
> > >>>>>> strace: new tcb for pid 26302, active tcbs:1
> > >>>>>> strace: [wait(0x80137f) = 26302] WIFSTOPPED,sig=SIGSTOP,EVENT_STOP (128)
> > >>>>>> strace: pid 26302 has TCB_STARTUP, initializing it
> > >>>>>> strace: [wait(0x80057f) = 26302] WIFSTOPPED,sig=SIGTRAP,EVENT_STOP (128)
> > >>>>>> strace: [wait(0x00127f) = 26302] WIFSTOPPED,sig=SIGCONT
> > >>>>>> strace: [wait(0x00857f) = 26302] WIFSTOPPED,sig=133
> > >>>>>> execve("./a.out", ["./a.out"], ["SSH_CONNECTION=15.86.136.27 5196"...,
> > >>>>>> "LANG=en_US.UTF-8", "XDG_SESSION_ID=41", "USER=dayd",
> > >>>>>> "PWD=/home/dayd/projects/musl-tes"..., "HOME=/home/dayd",
> > >>>>>> "SSH_CLIENT=xx.xxx.xxx.xx 51960 22", "CCACHE_DIR=/ssd/dayd/.ccache",
> > >>>>>> "VMWARE_USE_SHIPPED_GTK=yes", "XDG_DATA_DIRS=/usr/local/share:/"...,
> > >>>>>> "USECCACHE=ON", "SSH_TTY=/dev/pts/2", "MAIL=/var/mail/dayd",
> > >>>>>> "TERM=rxvt-unicode-256color", "SHELL=/bin/bash", "SHLVL=1",
> > >>>>>> "PYTHONPATH=:/home/dayd/projects/"..., "LOGNAME=dayd",
> > >>>>>> "DBUS_SESSION_BUS_ADDRESS=unix:pa"...,
> > >>>>>> "XDG_RUNTIME_DIR=/run/user/6686",
> > >>>>>> "PATH=/usr/lib/ccache/bin/:/usr/l"..., "USEDISTCC=OFF",
> > >>>>>> "_=/usr/bin/strace", "OLDPWD=/home/dayd/projects"]strace:
> > >>>>>> [wait(0x04057f) = 26302] WIFSTOPPED,sig=SIGTRAP,EVENT_EXEC (4)
> > >>>>>> strace: [wait(0x00857f) = 26302] WIFSTOPPED,sig=133
> > >>>>>> ) = 0
> > >>>>>> strace: [wait(0x00047f) = 26302] WIFSTOPPED,sig=SIGILL
> > >>>>>> --- SIGILL {si_signo=SIGILL, si_code=ILL_ILLOPN, si_addr=0x40169e} ---
> > >>>>>> strace: [wait(0x06057f) = 26302] WIFSTOPPED,sig=SIGTRAP,EVENT_EXIT (6)
> > >>>>>> strace: [wait(0x000084) = 26302] WIFSIGNALED,core,sig=SIGILL
> > >>>>>> +++ killed by SIGILL (core dumped) +++
> > >>>>>> strace: dropped tcb for pid 26302, 0 remain
> > >>>>>> Illegal instruction (core dumped)
> > >>>>>> ******************************************************************************************
> > >>>>>> (gdb) bt
> > >>>>>> #0  0x00000000004016ae in __copy_tls ()
> > >>>>>> #1  0x0000000000000000 in ?? ()
> > >>>>>> *****************************************************************************************
> > >>>>>>
> > >>>>>> (gdb) disassemble
> > >>>>>> Dump of assembler code for function __copy_tls:
> > >>>>>>    0x0000000000401680 <+0>:    push   %r13
> > >>>>>>    0x0000000000401682 <+2>:    mov    %rdi,%r13
> > >>>>>>    0x0000000000401685 <+5>:    push   %r12
> > >>>>>>    0x0000000000401687 <+7>:    push   %rbp
> > >>>>>>    0x0000000000401688 <+8>:    push   %rbx
> > >>>>>>    0x0000000000401689 <+9>:    sub    $0x8,%rsp
> > >>>>>>    0x000000000040168d <+13>:    mov    0x410c(%rip),%rax        #
> > >>>>>> 0x4057a0 <__libc+32>
> > >>>>>>    0x0000000000401694 <+20>:    mov    0x410d(%rip),%rcx        #
> > >>>>>> 0x4057a8 <__libc+40>
> > >>>>>>    0x000000000040169b <+27>:    mov    0x40f6(%rip),%rbx        #
> > >>>>>> 0x405798 <__libc+24>
> > >>>>>>    0x00000000004016a2 <+34>:    lea    -0xe0(%rdi,%rax,1),%rax
> > >>>>>>    0x00000000004016aa <+42>:    lea    -0x1(%rcx),%r12
> > >>>>>> => 0x00000000004016ae <+46>:    andn   %rax,%r12,%r12
> > >>>>> This instruction, andn, is not in the base x86_64 ISA. It's part of
> > >>>>> the BMI1 extension:
> > >>>>>
> > >>>>> https://en.wikipedia.org/wiki/Bit_Manipulation_Instruction_Sets#BMI1_(Bit_Manipulation_Instruction_Set_1)
> > >>>>>
> > >>>>> Apparently Gentoo is providing cross compilers that default to
> > >>>>> producing "native ISA level" code for the host, rather than baseline
> > >>>>> ISA for the target. But:
> > >>>>>
> > >>>>>> as a bonus (I thought this might also help):
> > >>>>>> (%:~/projects/musl-test)┌- x86_64-pc-linux-musl-gcc -v -static -O0 -g
> > >>>>>> hello.c
> > >>>>>> Using built-in specs.
> > >>>>>> COLLECT_GCC=x86_64-pc-linux-musl-gcc
> > >>>>>> COLLECT_LTO_WRAPPER=/usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/lto-wrapper
> > >>>>>> Target: x86_64-pc-linux-musl
> > >>>>>> Configured with:
> > >>>>>> /var/tmp/portage/cross-x86_64-pc-linux-musl/gcc-9.1.0/work/gcc-9.1..0/configure
> > >>>>>> --host=x86_64-pc-linux-gnu --target=x86_64-pc-linux-musl
> > >>>>>> --build=x86_64-pc-linux-gnu --prefix=/usr
> > >>>>>> --bindir=/usr/x86_64-pc-linux-gnu/x86_64-pc-linux-musl/gcc-bin/9..1.0
> > >>>>>> --includedir=/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/include
> > >>>>>> --datadir=/usr/share/gcc-data/x86_64-pc-linux-musl/9.1.0
> > >>>>>> --mandir=/usr/share/gcc-data/x86_64-pc-linux-musl/9.1.0/man
> > >>>>>> --infodir=/usr/share/gcc-data/x86_64-pc-linux-musl/9.1.0/info
> > >>>>>> --with-gxx-include-dir=/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/include/g++-v9
> > >>>>>> --with-python-dir=/share/gcc-data/x86_64-pc-linux-musl/9.1.0/python
> > >>>>>> --enable-languages=c,c++,fortran --enable-obsolete --enable-secureplt
> > >>>>>> --disable-werror --with-system-zlib --enable-nls
> > >>>>>> --without-included-gettext --enable-checking=release
> > >>>>>> --with-bugurl=https://bugs.gentoo.org/ --with-pkgversion='Gentoo 9.1.0
> > >>>>>> p1.0' --disable-esp --enable-libstdcxx-time
> > >>>>>> --enable-poison-system-directories
> > >>>>>> --with-sysroot=/usr/x86_64-pc-linux-musl --disable-bootstrap
> > >>>>>> --enable-__cxa_atexit --disable-multilib --with-multilib-list=m64
> > >>>>>> --disable-altivec --disable-fixed-point --enable-targets=all
> > >>>>>> --enable-libgomp --disable-libmudflap --disable-libssp
> > >>>>>> --disable-systemtap --disable-vtable-verify --disable-libvtv
> > >>>>>> --enable-lto --without-isl --disable-libsanitizer --enable-default-pie
> > >>>>>> --enable-default-ssp
> > >>>>>> Thread model: posix
> > >>>>>> gcc version 9.1.0 (Gentoo 9.1.0 p1.0)
> > >>>>>> COLLECT_GCC_OPTIONS='-v' '-static' '-O0' '-g' '-mtune=generic' '-march=x86-64'
> > >>>>>>  /usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/cc1 -quiet -v hello.c
> > >>>>>> -quiet -dumpbase hello.c -mtune=generic -march=x86-64 -auxbase hello
> > >>>>>> -g -O0 -version -o /tmp/ccoTkh38.s
> > >>>>>> GNU C17 (Gentoo 9.1.0 p1.0) version 9.1.0 (x86_64-pc-linux-musl)
> > >>>>>>     compiled by GNU C version 8.3.0, GMP version 6.1.2, MPFR version
> > >>>>>> 4.0.2, MPC version 1.1.0, isl version none
> > >>>>>> GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
> > >>>>>> ignoring nonexistent directory "/usr/x86_64-pc-linux-musl/usr/local/include"
> > >>>>>> ignoring nonexistent directory
> > >>>>>> "/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/../../../../x86_64-pc-linux-musl/include"
> > >>>>>> #include "..." search starts here:
> > >>>>>> #include <...> search starts here:
> > >>>>>>  /usr/x86_64-pc-linux-musl/usr/include
> > >>>>>>  /usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/include
> > >>>>>> End of search list.
> > >>>>>> GNU C17 (Gentoo 9.1.0 p1.0) version 9.1.0 (x86_64-pc-linux-musl)
> > >>>>>>     compiled by GNU C version 8.3.0, GMP version 6.1.2, MPFR version
> > >>>>>> 4.0.2, MPC version 1.1.0, isl version none
> > >>>>>> GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
> > >>>>>> Compiler executable checksum: a3183e5f435d1938198d432af6db159b
> > >>>>>> COLLECT_GCC_OPTIONS='-v' '-static' '-O0' '-g' '-mtune=generic' '-march=x86-64'
> > >>>>>>  /usr/libexec/gcc/x86_64-pc-linux-musl/as -v --64 -o /tmp/ccHq7CL6..o
> > >>>>>> /tmp/ccoTkh38.s
> > >>>>>> GNU assembler version 2.32.0 (x86_64-pc-linux-musl) using BFD version
> > >>>>>> (Gentoo 2.32 p2) 2.32.0
> > >>>>>> COMPILER_PATH=/usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/:/usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/:/usr/libexec/gcc/x86_64-pc-linux-musl/:/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/:/usr/lib/gcc/x86_64-pc-linux-musl/
> > >>>>>> LIBRARY_PATH=/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/:/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/../../../../x86_64-pc-linux-musl/lib/:/usr/x86_64-pc-linux-musl/lib/:/usr/x86_64-pc-linux-musl/usr/lib/
> > >>>>>> COLLECT_GCC_OPTIONS='-v' '-static' '-O0' '-g' '-mtune=generic' '-march=x86-64'
> > >>>>>>  /usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/collect2 -plugin
> > >>>>>> /usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/liblto_plugin.so
> > >>>>>> -plugin-opt=/usr/libexec/gcc/x86_64-pc-linux-musl/9.1.0/lto-wrapper
> > >>>>>> -plugin-opt=-fresolution=/tmp/ccWcifv4.res
> > >>>>>> -plugin-opt=-pass-through=-lgcc -plugin-opt=-pass-through=-lgcc_eh
> > >>>>>> -plugin-opt=-pass-through=-lc --sysroot=/usr/x86_64-pc-linux-musl -m
> > >>>>>> elf_x86_64 -static /usr/x86_64-pc-linux-musl/usr/lib/crt1.o
> > >>>>>> /usr/x86_64-pc-linux-musl/usr/lib/crti.o
> > >>>>>> /usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/crtbeginT.o
> > >>>>>> -L/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0
> > >>>>>> -L/usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/../../../../x86_64-pc-linux-musl/lib
> > >>>>>> -L/usr/x86_64-pc-linux-musl/lib -L/usr/x86_64-pc-linux-musl/usr/lib
> > >>>>>> /tmp/ccHq7CL6.o --start-group -lgcc -lgcc_eh -lc --end-group
> > >>>>>> /usr/lib/gcc/x86_64-pc-linux-musl/9.1.0/crtend.o
> > >>>>>> /usr/x86_64-pc-linux-musl/usr/lib/crtn.o
> > >>>>>> COLLECT_GCC_OPTIONS='-v' '-static' '-O0' '-g' '-mtune=generic' '-march=x86-64'
> > >>>>>>
> > >>>>>> mtune=generic and march=x86-64 suggests it isn't trying to do
> > >>>>>> something weird on the compile computer.
> > >>>>> Indeed, it looks like the default is right. So you probably need to
> > >>>>> check how musl was built. Is it also a package provided by Gentoo, and
> > >>>>> did they perhaps use -march=native even though it's being cross
> > >>>>> compiled?
> > >>>>>
> > >>>>> Rich
> > >>>
> >
> >

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.