CONFIG_GENTOO_LINUX_INIT_SYSTEMD forces CONFIG_IPV6=y which in turn causes the (huge) ipv6 module to be linked in. For HPPA, this causes problems as the kernel image is then too large and the linker fails. CHK include/generated/uapi/linux/version.h CHK include/generated/utsrelease.h CC scripts/mod/devicetable-offsets.s GEN scripts/mod/devicetable-offsets.h HOSTCC scripts/mod/file2alias.o HOSTLD scripts/mod/modpost CALL scripts/checksyscalls.sh CHK include/generated/compile.h CHK kernel/config_data.h LINK vmlinux LD vmlinux.o MODPOST vmlinux.o GEN .version CHK include/generated/compile.h UPD include/generated/compile.h CC init/version.o LD init/built-in.o KSYM .tmp_kallsyms1.o hppa64-unknown-linux-gnu-ld: scripts/link-vmlinux.sh: line 52: 7349 Segmentation fault (core dumped) ${LD} ${LDFLAGS} ${LDFLAGS_vmlinux} -o ${2} -T ${lds} ${KBUILD_VMLINUX_INIT} --start-group ${KBUILD_VMLINUX_MAIN} --end-group ${1} make: *** [vmlinux] Error 139
The systemd ebuild expects them to be enabled; so, I did a mimic of that in the kernel. But okay, clearly not everyone uses or wants to use IPv6 at this point in time yet; so, we probably need to revise this setting, at least in the kernel. http://cgit.freedesktop.org/systemd/systemd/tree/README Upstream lists this as "Optional but strongly recommended"; so, I think it would be sufficient to drop "select CONFIG_IPV6" from CONFIG_GENTOO_LINUX_INIT_SYSTEMD in 3.10.9 and then update the GENPATCHES number in the current ebuild. That way, we could exclusively stabilize 3.10.9 on HPPA... Are you okay with that? We will then later as stabilization is done and time has passed probably want to unstable 3.10.7 for HPPA and then ask systemd@g.o later to revise the setting in the ebuild; because with the kernel option selected, it's going to look odd that the ebuild then still mentions something as missing. Seeing the upstream README some time ago; I do have plans in the future to add these optional options in some kind of sub menu such that users can more carefully choose these themselves, that's something I am going to work on by the next stable (which I aim for 3.10.14+ [= 2+ months from now] and/or somewhere late in the 3.11 branch). Thank you for your report.
Seeing you have assigned this to hppa@g.o and not kernel@g.o; the other option is that we future proof this now as we're getting closer and closer to IPv6 becoming mainstream, which I think might be your intention. If you solely want to see this fixed by the kernel team, and have no idea on how to patch this matter; then feel free to reassign to us such that we do not lose track of this... Thank you.
Perhaps CONFIG_GENTOO_INIT_SYSTEMD could be changed from bool to tristate to allow its dependencies to be modularized as needed.
Even with CONFIG_IPV6=m the linker segfaults, so I am looking into the other options that are automatically linking in modules.
With CONFIG_AUTOFS4_FS=y and after significantly stripping the kernel images of drivers not necessary before init, I got it to compile. But that's a pretty small module at 65 kB: IPV6 is the only really problematic one at 651 kB.
(In reply to Tom Wijsman (TomWij) from comment #2) > Seeing you have assigned this to hppa@g.o and not kernel@g.o Assignment wasn't intended to convey some hidden message.
------------------------------------------------------------------------ r2557 | tomwij | 2013-10-14 19:36:10 +0200 (Mon, 14 Oct 2013) | 1 line Select FHANDLE (bug #485154) and do not select IPV6 (bug #481998) for systemd. ------------------------------------------------------------------------
