This is a follow up to https://bugs.gentoo.org/show_bug.cgi?id=589154 because version r3 is now already in tree. sys-kernel/hardened-sources-4.6.4-r3 is still affected by the issue that network related actions can cause a panic. Spotted trigger on x86, every unplug of ethernet cable of DSL modem causes a panic (same behavior as r2). Revert back to 4.5 series kernel, no issue. Please be aware that testing and reporting back of kernel related issues is not always possible in short time. I have to consider and plan downtime on systems that have vital functions for the network. Not every issue is reproducible on a dedicated testbox... Thanks for your understanding, patience and work!
(In reply to Alexander Stoll from comment #0) > This is a follow up to https://bugs.gentoo.org/show_bug.cgi?id=589154 > because version r3 is now already in tree. > sys-kernel/hardened-sources-4.6.4-r3 is still affected by the issue that > network related actions can cause a panic. Spotted trigger on x86, every > unplug of ethernet cable of DSL modem causes a panic (same behavior as r2). > Revert back to 4.5 series kernel, no issue. > Please be aware that testing and reporting back of kernel related issues is > not always possible in short time. I have to consider and plan downtime on > systems that have vital functions for the network. Not every issue is > reproducible on a dedicated testbox... > Thanks for your understanding, patience and work! I closed that bug because a tester said that -r3 was fixed. I'll pass this upstream again. At the very least, an oops message would be helpful.
we already reverted the size overflow hash table changes, so i guess you're back to your original problem that you had with all 4.6 kernels, not just the ones with the regenerated hash table. i'd need dmesg again, if it's still size overflow related then boot with pax_size_overflow_report_only.
Thanks to all for your support, I´m not able to trigger the 2 remaining observed issues on the testbox, I will have to get the ooopses directly from the affected systems. Will report back ASAP.
Created attachment 443156 [details] Panic on x32 box after pulling cable on DSL modem, current 4.6.5-r1 sorry for this ridiculous amount of time, can´t remember when a rather trivial thing like getting this dump caused such and effort.... now get x64 panic at shutdown ASAP ;-)
can you reproduce this with frame pointers enabled? also what does addr2line decode the faulting EIP to?
In the meantime 4.7 has superseeded the problematic 4.6 series, tested 4.7.2 on all affected systems, issue is gone. Thanks for your efforts!
Resolved by 4.7 kernel series