Date: Wed, 7 Mar 2018 10:21:17 -0800 From: bob prohaska <fbsd@www.zefox.net> To: freebsd-arm@freebsd.org Subject: Garbled console output on Pi3 Message-ID: <20180307182117.GA55819@www.zefox.net>
next in thread | raw e-mail | index | archive | help
>From time to time I observe garbled console output on a Pi3 running -current.
The present revision is
FreeBSD 12.0-CURRENT (ZEFOX) #1 r330571: Wed Mar 7 00:58:23 PST 2018
but I don't think that's special; the same thing has shown up sporadically
for the past month or so.
The output looks like this:
Mar 7 09:53:53 www kernel: pid 44970 (llvm-tblgen), uid 0, was killed: out of swap space
sswawpa_spp_apgaegr:er iw:ndsaewap fii_ppnand_ieptge fiweaanrig:eti r:teb u wfiifannidt deefebr:i fnbuiiuffontife tberwajit:e : bu0f, bufeblfokb nj:r:o: 0,w2 b b22u3fo94aibj,tl: k b0nosui:,z fe b1siz87erlk6:n, o:sbi 4zue: : 202f901o669b604,
ej0sw: :
r:ag le4,0paap9 g6_epb
inrkd: inendefifnioini:te t w1eai35t bu w2fai6t0f b,er:uf fbufoerbj: :bufo bjs0:i , z0bl, blkeknno: 2: o16: 4260191,52
9393s7ize:, 1siz63e:84 4
096
swap_pager: indefinite wait buffer: bufobj: 0, blkno: 210573, size: 4096
The first and last lines look normal, which seems to rule out errors in
the serial link.
It seems to happen when the system is overloaded. In this case make -j8
buildworld was running as a stress test, but much lower j values have done
the same thing in the past. All the output seen in ssh sessions looks
normal, although sometimes very delayed. In this case top reported
(shortly after the fact):
last pid: 58112; load averages: 1.98, 1.45, 1.52 up 0+01:25:14 10:10:15
49 processes: 3 running, 46 sleeping
CPU: 56.5% user, 0.0% nice, 6.6% system, 8.0% interrupt, 29.0% idle
Mem: 586M Active, 63M Inact, 36M Laundry, 154M Wired, 97M Buf, 68M Free
Swap: 3072M Total, 1369M Used, 1703M Free, 44% Inuse, 22M In
PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND
58053 root 1 81 0 271M 45704K RUN 0 0:53 97.15% llvm-tblgen
58086 root 1 96 0 185M 26820K CPU2 2 3:24 93.95% llvm-tblgen
58049 root 1 29 0 265M 24924K swread 2 0:50 27.49% llvm-tblgen
58048 root 1 26 0 392M 63920K swread 2 0:37 16.72% llvm-tblgen
58046 root 1 22 0 254M 21816K swread 1 0:45 8.05% llvm-tblgen
58051 root 1 21 0 268M 26680K swread 1 0:49 3.14% llvm-tblgen
58044 root 1 20 0 67828K 11000K swread 2 0:35 1.51% llvm-tblgen
643 bob 1 20 0 13872K 876K CPU1 1 0:10 0.18% top
638 bob 1 20 0 21148K 600K select 2 0:01 0.03% sshd
58041 root 1 20 0 23776K 588K select 3 0:19 0.00% make
573 root 1 20 0 15468K 560K select 3 0:00 0.00% sendmail
45034 root 1 20 0 11024K 184K select 0 0:00 0.00% make
19864 bob 1 20 0 13788K 32K ttyin 3 0:00 0.00% tcsh
58016 root 1 20 0 10920K 184K select 0 0:00 0.00% make
30893 bob 1 20 0 13788K 32K ttyin 3 0:00 0.00% tcsh
570
It seems odd that the console should malfunction in this fashion,
perhaps there is some significance to more learned eyes.
In this case buildworld worked through the bottleneck and
is again running at 0% idle.
Thanks for reading,
bob prohaska
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20180307182117.GA55819>