NetBSD-Bugs archive
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]
port-mips/58072: mips: mis-counting of CPU states for accounting & monitoring
>Number: 58072
>Category: port-mips
>Synopsis: mips: mis-counting of CPU states for accounting & monitoring
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: port-mips-maintainer
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sun Mar 24 11:20:00 +0000 2024
>Originator: Havard Eidnes
>Release: NetBSD 10.99.10
>Organization:
I try...
>Environment:
System: NetBSD mipsel 10.99.10 NetBSD 10.99.10 (MIPSSIM) #8: Mon Dec 25 18:08:12 CET 2023 he%mt.urc.uninett.no@localhost:/u/build/HEAD/obj/evbmips-el/sys/arch/evbmips/compile/MIPSSIM evbmips
Architecture: mips
Machine: evbmips
>Description:
The evbmips port using the MIPSSIM kernel appears to
mis-characterize and mis-count the CPU states with respect
to whether it's "User", "Nice", "System" or "Idle".
While trying to test rust 1.76.0, I have two builds running
and they are makign progress, but despite this, e.g. "top"
reports 100% "System" CPU, which is clearly bogus:
load averages: 2.22, 2.19, 2.13; up 90+00:06:28 17:40:21
49 processes: 2 runnable, 46 sleeping, 1 on CPU
CPU states: 0.0% user, 0.0% nice, 100% system, 0.0% interrupt, 0.0% idle
Memory: 230M Act, 112M Inact, 16K Wired, 97M Exec, 148M File, 3028K Free
Swap: 2048M Total, 361M Used, 1687M Free
PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND
27998 he 21 4 289M 122M RUN 2:53 47.51% 47.51% rustc
19669 he 21 4 212M 60M RUN 317:27 47.02% 47.02% rustc
12376 he 85 0 18M 700K select 0:26 1.03% 1.03% sshd
13513 he 85 0 6036K 784K nanoslp 0:03 0.05% 0.05% vmstat
23354 he 43 0 9168K 1276K CPU 0:11 0.00% 0.00% top
and with vmstat:
procs memory page disks faults cpu
r b avm fre flt re pi po fr sr l0 m0 in sy cs us sy id
2 0 600584 2352 12 0 1 0 0 0 1 0 52 2074 22 0 100 0
2 0 600584 2228 10 0 1 0 0 0 12 0 56 1901 23 0 100 0
4 0 600836 1736 28 0 1 0 0 0 0 0 52 1467 21 0 100 0
2 0 601332 1576 11 0 0 0 0 0 0 0 52 1511 21 0 100 0
2 0 601836 1076 33 0 1 0 0 0 1 0 52 2295 21 0 100 0
2 0 602524 5152 216 0 13 223 223 223 25 0 70 1342 37 0 100 0
2 0 602872 4624 39 0 2 0 0 0 3 0 53 1665 22 0 100 0
2 0 603328 4452 13 0 1 0 0 0 5 0 58 1794 22 0 100 0
and with systat:
7 users Load 2.15 2.17 2.13 Sun Mar 24 17:41:09
Proc:r d s Csw Traps SysCal Intr Soft Fault PAGING SWAPPING
3 11 52 284 1858 56 24 277 in out in out
ops 2
100.0% Sy 0.0% Us 0.0% Ni 0.0% In 0.0% Id pages
| | | | | | | | | | |
================================================== 2 forks
2 fkppw
Anon 98924 20% zero 56 Interrupts 2 fksvm
Exec 100132 20% wired 16 6 int 1 (virtio) pwait
File 151196 30% inact 115140 int 2 (uart) 2 relck
Meta 39944 4% bufs 3767 50 int 5 (clock) 2 rlkok
(kB) real swaponly free int 5 (clock) m noram
Active 234616 363800 3176 42 ndcpy
Namei Sys-cache Proc-cache 19 fltcp
Calls hits % hits % 134 zfod
68 68 100 cow
256 fmin
Disks: ld0 md0 341 ftarg
seeks itarg
xfers 1 flnan
bytes 2600 pdfre
%busy pdscn
>How-To-Repeat:
Run a program which consumes significant amounts of CPU time.
Watch any of the tools above report all CPU usage as "System".
>Fix:
Sorry, don't know.
Home |
Main Index |
Thread Index |
Old Index