Regression (latency spike) between 5.16.2-rt19 and 5.17.1-rt17

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 




Dear rt-users,

I was happily using kernel version 5.4.84-rt47 on this Celeron j4125
system (all power management and cpu frequency scaling disabled in bios)
and it was working great. Worst case cyclictest latency was around 30 to
40 us depending on system load.

Due to some USB fixes landing in 5.14 I went and tried upgrading my
kernel. I started with 5.19-rt10 but something weird showed up:
cyclictest reported _huge_ latencies on one of the four cores (thread 1
of cyclictest) of around 700 us. So I went back to my old kernel to make
sure it really was the kernel and sure thing. The old kernel still had
the nice 30 to 40 us worst case latencies I determined previously. So I
went through some kernel versions:

5.19-rt10    has the problem
5.17.1-rt17  has the problem
5.16.2-rt19  does not have the problem

So I would guess somewhere between 5.16.2-rt19 and 5.17.1-rt17 something
horrible happened ;)

Maybe interesting fact: It is always thread 1 of cyclictest that shows
the large latency. The other threads are seemingly completely in-line
with the expectation of 30 - 40 us.

Here's my cyclictest command:

sudo ./cyclictest -m -p 95 -S -M

cyclictest at stable/v1.0

I would possibly be willing to git bisect the issue. But I am a little
bit confused about how to go about that. What repo/tree would I
precisely bisect (I am using the the patch tarballs from
https://cdn.kernel.org/pub/linux/kernel/projects/rt/)?

Kind regards,
FPS

--
https://fps.io




[Index of Archives]     [RT Stable]     [Kernel Newbies]     [IDE]     [Security]     [Git]     [Netfilter]     [Bugtraq]     [Yosemite]     [Yosemite News]     [MIPS Linux]     [ARM Linux]     [Linux Security]     [Linux RAID]     [Linux ATA RAID]     [Samba]     [Video 4 Linux]     [Device Mapper]

  Powered by Linux