intremap not working after kernel update?

[c64wolf]

Hello,

I have been having issues with my server's interrupt remapping in the past, related to Intel chipset errata (Intel 55x0 Chipset Errata - Interrupt Remapping Issue). I managed to fix the issues by adding following boot parameters to grub:
intremap=off pci=nomsi,noaer

I have confirmed via dmesg that the parameters are still passed to the kernel after latest kernel upgrade:
[ 0.000000] Linux version 4.4.44-1-pve (root@nora) (gcc version 4.9.2 (Debian 4.9.2-10) ) #1 SMP PVE 4.4.44-84 (Thu, 9 Mar 2017 12:06:34 +0100) ()
[ 0.000000] Command line: BOOT_IMAGE=/boot/vmlinuz-4.4.44-1-pve root=/dev/md2 ro rootdelay=15 quiet intremap=off pci=nomsi,noaer

The interrupt remapping issue caused one of my LXC containers to lose network connectivity with ssh daemon instances going <defuct>. Now after the kernel upgrade I saw this same issue today again, but could not see anything strange in system logs. Has those boot parameters changed since the earlier kernel version or is the new kernel missing a module or a compile time setting?

 

[fabian]

I don't see anything obvious that would have caused such a change. Maybe you can try narrowing down the kernel version where your issues start?

 

[c64wolf]

We had no issues with the earlier kernel version. Saw the issue again when upgrading to 4.4.44-1-pve (PVE 4.4.44-84), which was released last week.

 

[fabian]

what is "the earlier kernel version" exactly?

 

[c64wolf]

Thank you for your fast response. The earlier kernel version we were running was Linux 4.4.40-1-pve.

 

[fabian]

Thank you for your fast response. The earlier kernel version we were running was Linux 4.4.40-1-pve.

if you are up for the task, you could bisect the Ubuntu kernels those two were based on (4.4.0-63.84 for 4.4.44 , and 4.4.0.62-83 for 4.4.40).

if this is a remote machine, please only test such kernels if you have iKVM access and can easily revert to working kernels in the bootloader's menu. it is possible that a test kernel does not boot on your system (because of missing modules).

first, I'd test whether the Ubuntu-4.40-63.84 kernel package works: http://packages.ubuntu.com/xenial/amd64/linux-image-4.4.0-63-generic/download

if it does, please report back and skip the rest of this post (this would indicate that one of our patches is at fault).

otherwise, you could test the respective mainline kernels from 4.4.40 to 4.4.44 (from http://kernel.ubuntu.com/~kernel-ppa/mainline/), if those are affected as well please report back in which mainline stable kernel the problem starts to show and skip the rest (this would indicate it's an actual upstream mainline kernel issue).

if you get to this point, the only thing that remains to test is a full bisect of the Ubuntu kernel's git repo (from 4.4.0-62.83 to 4.4.0-63.84):

 git bisect start Ubuntu-4.4.0-63.84 Ubuntu-4.4.0-62.83
Bisecting: 181 revisions left to test after this (roughly 8 steps)
[f2097e3e4039edb56271b6db1c380be0bb210514] HID: hid-cypress: validate length of report

that would take around 8 test kernels which you'd need to build, which is also why it's the last option here