Bug#995223: libffi: SIGILL on powerpc and ppc64 systems since libffi8
Hello! On 9/29/21 00:38, John Paul Adrian Glaubitz wrote: > There is actually a baseline violation on i386 as "-march=amdfam10" is > passed to the compiler, see the build log in [1], for example. Comparing the build logs, it seems that this an issue with the newer version of autoconf that was used to build version 3.4.2-2. Looking at the log for 3.4.2-2, a lot of autoconf warnings can be seen which are not present in the log for 3.4.1-1. > https://buildd.debian.org/status/fetch.php?pkg=libffi=powerpc=3.4.2-1=1626443428=0 > https://buildd.debian.org/status/fetch.php?pkg=libffi=powerpc=3.4.2-2=1632868252=0 So, upstream needs to fix compatibility issues with the newer autoconf version. Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `-GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
Bug#995223: libffi: SIGILL on powerpc and ppc64 systems since libffi8
Hi! On 9/28/21 22:40, John Paul Adrian Glaubitz wrote: > We should therefore pass "--enable-portable-binary" in debian/rules. There is actually a baseline violation on i386 as "-march=amdfam10" is passed to the compiler, see the build log in [1], for example. Adrian > [1] > https://buildd.debian.org/status/fetch.php?pkg=libffi=i386=3.4.2-2=1632469242=0 -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `-GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
Bug#995223: libffi: SIGILL on powerpc and ppc64 systems since libffi8
Control: severity -1 serious Hello! It turns out that m4/ax_gcc_archflag.m4 contains code to detect the baseline of the host system and sets the GCC architecture accordingly. Thus, a libffi compiled on a POWER8 machine will not work on a POWER5 machine as the compiler is emitting POWER8 instructions in this case. Since the m4 script contains such a host enviroment detection for aarch64 as well [1], this bug can potentially affect arm64 which is a release architecture. We should therefore pass "--enable-portable-binary" in debian/rules. Adrian > [1] https://github.com/libffi/libffi/blob/master/m4/ax_gcc_archflag.m4#L209 -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `-GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913
Bug#995223: libffi: SIGILL on powerpc and ppc64 systems since libffi8
Source: libffi Version: 3.3-6 Severity: important User: debian-powe...@lists.debian.org Usertags: powerpc ppc64 X-Debbugs-Cc: debian-powe...@lists.debian.org Hi! Multiple users on powerpc and ppc64 have reported SIGILL crashes after upgrading to libffi8 (3.4.x): # dmesg ... [ 16.257543] fail2ban-server[384]: illegal instruction (4) at 3fffb4283970 nip 3fffb4283970 lr 3fffb4282f90 code 1 in libffi.so.8.1.0[3fffb427b000+c000] ... Downgrading to libffi7 fixes the problem. It has been reported that building and using the upstream version does not reproduce this issue. But I have not yet independently verified that. It might be that libffi performs a runtime detection during build which causes the library to be built with a higher baseline on the POWER8 buildds. Adrian -- .''`. John Paul Adrian Glaubitz : :' : Debian Developer - glaub...@debian.org `. `' Freie Universitaet Berlin - glaub...@physik.fu-berlin.de `-GPG: 62FF 8A75 84E0 2956 9546 0006 7426 3B37 F5B5 F913