Linux, Hunting error logs during boot

I have this obession to know about kernel errors/warnings, even if my PC works fine. At least you learn some interesting stuff.

Huge pages

pmd_set_huge: Cannot satisfy [mem 0x383ff0000000-0x383ff0200000] with a huge-page mapping due to MTRR override.

MTRR memory type range register, used by processor to determine type of memory (uncacheable, write-back…) behind an address range.
- MTRR are mostly deprecated to PAT (page table attributes) which offers higher granularity
Memory range overlaps between huge pages and MTRR caused the huge page to be split.
However in this case, MTRR are set by the BIOS, they cannot be changed by the OS (even with kernel param enable_mtrr_cleanup)
- Changing 32bit or 64bit PCI mapping on the BIOS influences /proc/mtrr

# 32bit PCI mapping
reg05: base=0x0f0000000 ( 3840MB), size=   32MB, count=1: write-through
# lspci
03:00.0 VGA compatible controller: NVIDIA Corporation GM204 [GeForce GTX 970] (rev a1) (prog-if 00 [VGA controller])
  Memory at f0000000 (64-bit, prefetchable) [size=32M]

# 64bit PCI mapping
reg05: base=0x383ff0000000 (58982144MB), size=   32MB, count=1: write-through

Huge page conclusion

Not sure what was linux attempting to map with a 1GB page.
Not a big deal, I can live with a few extra 4K pages.

UDEV

systemd-udevd[412]: /usr/lib/udev/rules.d/51-android.rules:549 The line takes no effect, ignoring.
systemd-udevd[453]: sdb: /etc/udev/rules.d/60-schedulers.rules:2 Failed to write ATTR{/sys/devices/pci0000:00/0000:00:1f.2/ata5/host4/target4:0:0/4:0:0:0/block/sdb/queue/scheduler}, ignoring: Invalid argument
cfg80211: Process '/usr/bin/set-wireless-regdom' failed with exit code 1.

UDEV conclusion

Bad formed android rules are caused by adb, not a big dealio
Looks like the block device scheduler changed name from noop to none => FIXED
‘regdom’ refers to regulatory domain limitations on powerbands and tx power. Wifi works and has expected speeds.

Firmware bugs

TSC ADJUST: CPU0: -8605093541 force to 0
TSC ADJUST differs within socket(s), fixing all errors
  #2  #3  #4  #5  #6
MDS CPU bug present and SMT on, data leak possible. See https://www.kernel.org/doc/html/latest/admin-guide/hw-vuln/mds.html for more details.
  #7  #8  #9 #10 #11

TSC : timestamp counter, a x86 register.
- MDS : Microarchitectural Data Sampling, a side channel attack.

Firmware conclusion

Boot param libata.noacpi to avoid TSC adjust warnings => PC fails to boot
MDS mitigation flushes some buffers, it may impact perf, maybe disable ?

EDAC : Error Detection And Correction

EDAC sbridge: CPU SrcID #0, Ha #0, Channel #0 has DIMMs, but ECC is disabled
EDAC sbridge: Couldn't find mci handler
EDAC sbridge: Failed to register device with error -19.

EDAC are a set of linux modules to interface with hardware with error correcting capabilities and report error rates.
- For example PCI bus or ECC memory
- EDAC module may be statically linked (aka not displayed by lsmod), check /lib/modules/$(uname -r)/modules.builtin
- Most desktops do not have ECC (check sudo dmidecode -t 17)
sbridge refers to sandy bridge (intel arch). Turns out that edac_sbridge is the module config also for Haswell procs.
mci refers to memory controller information a linux struct.

EDAC conclusion

Nothing to worry since I do not have ECC.
- try kernel boot param edac_report=off to see if logs get less verbose

USB

usb 2-2.4: config 1 has an invalid interface number: 3 but max is 2
usb 2-2.4: config 1 has an invalid interface number: 3 but max is 2
usb 2-2.4: config 1 has an invalid interface number: 3 but max is 2
usb 2-2.4: config 1 has no interface number 2
hid-generic 0003:1852:7022.0005: No inputs registered, leaving
hid-generic 0003:1852:7022.0005: hidraw2: USB HID v1.00 Device [GFEC ASSP MyAMP] on usb-0000:00:14.0-2.4/input0

USB devices present themselves using configurations,interfaces, endpoints
- In this case my USB DAC present 1 configuration with 3 interfaces (audio control, audio stream, audio stream hidef)
- Although I paid good money for it, the firmware devs did not comform to the USB spec, interface ids should be sequential
HID human interface device : a type of USB interface
- Again this complains about my USB DAC
- Looks like the input declaration on the report descriptor are okish though
- cf hexdump -C "/sys/bus/usb/devices/<device>/<interface>/<id>/report_descriptor"

USB conclusion

USB dac works fine, even the HID (not sure why since no inputs got detected)
For the price I paid for the damn audio card, I would have expected usb2 support …

KDE

kscreen.kded: UPower not available, lid detection won't work
kded5[1037]: kscreen.kded: PowerDevil SuspendSession action not available!
powerdevil: Xrandr not supported, trying ddc, helper
backlighthelper[1229]: powerdevil: no kernel backlight interface found
powerdevil: org.kde.powerdevil.backlighthelper.brightness failed
powerdevil: The profile  "AC" tried to activate "DimDisplay" a non-existent action. This is usually due to an installation problem, a configuration problem, or because the action is not supported

kf5.kpackage: No metadata file in the package, expected it at: "/usr/share/wallpapers/Breath/contents/images/"
plasmashell[1093]: file:///usr/share/plasma/wallpapers/org.kde.image/contents/ui/main.qml:76:9: Unable to assign [undefined] to QStringList
plasmashell[1093]: file:///usr/share/plasma/wallpapers/org.kde.image/contents/ui/main.qml:75:9: Unable to assign [undefined] to int
plasmashell[1093]: file:///usr/share/plasma/wallpapers/org.kde.image/contents/ui/main.qml:75:9: Unable to assign [undefined] to int
plasmashell[1093]: file:///usr/share/plasma/wallpapers/org.kde.image/contents/ui/main.qml:76:9: Unable to assign [undefined] to QStringList
plasmashell[1093]: trying to show an empty dialog
plasmashell[1093]: file:///usr/share/plasma/shells/org.kde.plasma.desktop/contents/views/Desktop.qml:146:19: QML Loader: Binding loop detected for property "height"

plasma-nm: Unhandled active connection state change:  1

kdeinit5[1374]: kf5.kio.kio_tags: tag fetch failed: "Failed to open the database"
kdeinit5[1374]: kf5.kio.kio_tags: "tags:/" list() invalid url

xrandr (Resize and Rotate) is a configuration tool for xsrever screens.
- ddc (Display Data Channel) is a protocol to communicate with the monitor
There are too many power management options.
- To handle power events (like poweroff button) you can use powerdevil, acpid, sytemd-logind, TLP, upower (this one looks more like middleware)
kpackage is an API for installation and loading of additional content (ex: scripts, images…)
- no idea about this metadata deal …
QML is a declarative language that to build GUIs.
plasma-nm is the networkmanager applet frontend
KIO is an API for file access (local or via network)

KDE Conclusion

most warning related to power saving are not applicable to a desktop => dump powerdevil
Do not care about KDE bugs (KIO, QML or applets), the stuff I use works fine

Misc Cr*p

thermal thermal_zone0: failed to read out thermal zone (-61)
kauditd_printk_skb: 20 callbacks suppressed

thermal_zone0 looks to be related to wifi (cf /sys/class/thermal/thermal_zone0/type)
- Since this is pcie adapter into the mobo maybe it just cannot be read (even running sensors-detect again)
kauditd is the kernel part of the auditing framework. kauditd communicates with user space via a netlink socket.
- Not sure why this is firing since I disabled all rules in ‘/etc/auditd/rules.d’ (there is not even /var/log/audit/*.log)
- Maybe somehow systemd-journal is getting those audit events ? (I disabled audit.service unit)

Misc Cr*p conclusion

auditd is not needed
although temperature sensors are broken for wifi card, it works ok