Timestamps in journalctl logs are not matching the dmesg's timestamps

[ctheegal]

Timestamps in journalctl logs are not matching the dmesg's timestamps at device boot. These timestamps somehow get aligned after 4th second or so.
Command used to collect journalctl logs: "journalctl --output=short-monotonic"

Device used: RB8 (IQ-9075) EVK ( Issue also seen on other targets like RB3-Gen2 etc..)

Build Tag: https://github.com/qualcomm-linux/meta-qcom-releases/releases/tag/qli-2.0-rc2
Logs are from above build-tag. But, this problem is there from long back. Not a recent regression as such

root@iq-9075-evk:/# journalctl --output=short-monotonic | grep -E "Booting Linux|init process|Freeing|System time"
[ 3.034623] iq-9075-evk kernel: Booting Linux on physical CPU 0x0000000000 [0x410fd4b2]
[ 3.050614] iq-9075-evk kernel: Freeing initrd memory: 9480K
[ 3.073432] iq-9075-evk kernel: Freeing unused kernel memory: 2048K
[ 3.073447] iq-9075-evk kernel: Run /init as init process
[ 3.073519] iq-9075-evk systemd[1]: System time advanced to timestamp on /var/lib/systemd/timesync/clock: Fri 2026-03-13 16:31:45 UTC
root@iq-9075-evk:/#

root@iq-9075-evk:/#
root@iq-9075-evk:/# dmesg | grep -E "Booting Linux|init process|Freeing|System time"
[ 0.000000] Booting Linux on physical CPU 0x0000000000 [0x410fd4b2]
[ 0.166559] Freeing initrd memory: 9480K
[ 1.226482] Freeing unused kernel memory: 2048K
[ 1.226545] Run /init as init proc
[ 1.975133] systemd[1]: System time advanced to timestamp on /var/lib/systemd/timesync/clock: Fri 2026-03-13 16:31:45 UTC

root@iq-9075-evk:/#
dmesg.txt
journalctl.log

[koenkooi]

Can you show the output of systemd-analyze? I suspect you'll see that the offset is a close match to the sum of firmware+loader+kernel times, here's a more pronounced example:

root@gertrude:~# journalctl -k -b --output=short-monotonic | grep -E "Booting Linux|init process|Freeing|System time"
[   11.640275] gertrude kernel: Booting Linux on physical CPU 0x0000000000 [0x410fd4b2]
[   11.653399] gertrude kernel: Freeing initrd memory: 22860K
[   11.655065] gertrude kernel: Freeing unused kernel memory: 3392K
[   11.655076] gertrude kernel: Run /init as init process
[   11.687063] gertrude systemd[1]: System time advanced to timestamp on /var/lib/systemd/timesync/clock: Tue 2026-04-14 08:48:33 CEST
root@gertrude:~# dmesg | grep -E "Booting Linux|init process|Freeing|System time"
[    0.299245] Freeing initrd memory: 22860K
[    5.660522] Freeing unused kernel memory: 3392K
[    5.665475] Run /init as init process
[    9.071334] systemd[1]: System time advanced to timestamp on /var/lib/systemd/timesync/clock: Tue 2026-04-14 08:48:33 CEST
root@gertrude:~# systemd-analyze 
Startup finished in 1.922s (firmware) + 775ms (loader) + 9.012s (kernel) + 14.163s (userspace) = 25.873s 
graphical.target reached after 14.158s in userspace.
root@gertrude:~# echo  "1.922 + 0.775 + 9.012" | bc
11.709

Systemd flushes the kmsg ring buffer into journal, but it seems to loose the timestamps when doing so.

[ctheegal]

Attaching new logs set. dmesg.txt

journalctl.log

systemd_analyze.txt

Logs-1.zip

[ricardosalveti]

Is this a regression comparing with QLI 1.8?

[ctheegal]

No. QLI-1.8 doesnt have this issue.