Created attachment 505276 [details] my kernel config Booting gentoo-sources-4.9.63 (see attached config) compiled by gcc-6.4.0 (using genkernel) in an ESXi x86 guest resulted in a hung guest system, caps lock and scroll lock lights flashing on the keyboard. The guest system got stuck in different places: one time during sysinit/udev-settle, another time a few minutes after booting while serving lots of read accesses to it's nfs server. After compiling the same sources (freshly installed, same config) using sys-devel/gcc-5.4.0-r3 the guest system runs just fine.
I think I've a similar problem with my gentoo/x86 running under ESXi4 as an x86 guest, I'm trying to migrate from hardened-sources-4.8.17-r2 to gentoo-source-4.9.72, because of "hardened-sources" masking. If I compile both kernels with gcc-6.4.0 the system become "unstable", ranging from sudden halts to "bad swap file entry". Compiling both kernels with gcc-5.4.0 seems instead to lead to a stable kernel. I've to use the word "seems" because I tested the kernels for short periods of times (in the order of one day). Last time I compiled hardened-sources-4.8.17-r2, with gcc-5.4.0 (before installing gcc-6.4.0, in October 2017) the system hasn't any apparent problem and was happily up and running for months. If requested I may provide my kernels config, but at a very first look, it seems quite similar to the already posted config, which have, by the way, HIGHMEM/PAE enabled.
please try with sys-kernel/gentoo-sources-4.9.76
This configuration works for me now more that 24 hours under stress: VMware ESXi 5.5.0 build-6480324 sys-kernel/gentoo-sources-4.9.76 sys-devel/gcc-6.4.0-r1 uname -m i686 uname -p Intel(R) Xeon(R) CPU E5-2650 v2 @ 2.60GHz
I tested 4.9.72/x86, compiled with gcc-6.4.0 under ESXi 4.1, VMware Workstation 12.5.8 build-7098237 and KVM/qemu, using a "fresh" gentoo installation, created with latest installers/stage3 (again using 4.9.72 compiled with gcc-6.4.0) and all of them are "unstable", they at least hangs under high filesystem (ext3/ext4) activity, especially when the kernel is compiled for CONFIG_MPENTIUM4. They looks more stable when compiled for CONFIG_M486 (like the gentoo installer) for x86 with gcc-6.4.0, but under the right conditions, all of them, only in a virtual machine (VMWare or KVM), after a while, hang. I didn't test 4.9.76, as suggested, but I rather jumped to 4.14.12, compiled for CONFIG_MPENTIUM4, with gcc-6.4.0. Both my hardened system, which I'm trying to "convert" to a gentoo-source kernel ,and my test fresh gentoo/x86 virtual machines seems stable with this gentoo-sources-4.14.12 kernel. What make me still worrying is the gcc version dependency. It looks gcc-6.4.0 is generating wrong machine code for the kernel, at least in a virtual machine, especially for aged x86 Intel Pentium architectures. May we be sure will not generate wrong code even for standard gentoo application ebuilds? G. Vitillaro.
It looks like you found a workable configuration, please let us know if this issue is still outstanding.
Yep, I found a solution, new kernel versions, actually running now 4.4.95 without any problem. This specific issue may be closed, but the problem about gcc-6.4.0 compiling wrong kernel code for x86/32-bit is still there, I guess. Regards, G. Vitillaro.
