Date: Wed, 22 Nov 2017 09:36:13 -0800 From: bob prohaska <fbsd@www.zefox.net> To: Mark Millard <markmi@dsl-only.net> Cc: Freebsd-arm <freebsd-arm@freebsd.org>, ports@freebsd.org Subject: Re: ssh sessions close spontaneously on rpi2 Message-ID: <20171122173613.GA63035@www.zefox.net> In-Reply-To: <FB9B37A4-CD5B-45F7-85A5-9AF1E6174F86@dsl-only.net> References: <20171118165638.GA47956@www.zefox.net> <20171120165832.GA55836@www.zefox.net> <FB9B37A4-CD5B-45F7-85A5-9AF1E6174F86@dsl-only.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, Nov 21, 2017 at 01:15:05PM -0800, Mark Millard wrote: > > But I've another oddity that you might try: > > A) I booted and ran top on the serial console ( top -CaePosize ) > B) I logged in 4 ssh sessions and had each do "openssl speed" > > top gets an unexpected result: CPU 2 shows as 100% idle and the > others 0.0% idle. > On FreeBSD 12.0-CURRENT (RPI2) #11 r326038: Wed Nov 22 01:04:56 PST 2017 the behavior seems reasonable: last pid: 743; load averages: 3.88, 2.12, 0.97 up 0+00:35:32 08:51:39 28 processes: 5 running, 23 sleeping CPU: 97.3% user, 0.0% nice, 0.6% system, 2.2% interrupt, 0.0% idle Mem: 22M Active, 4380K Inact, 65M Wired, 33M Buf, 824M Free Swap: 2048M Total, 2048M Free PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND 739 bob 1 102 0 7404K 4496K CPU3 3 3:56 103.68% openssl 737 bob 1 103 0 7404K 4496K RUN 2 4:16 99.53% openssl 741 bob 1 103 0 7404K 4496K CPU1 1 2:25 98.15% openssl 740 bob 1 102 0 7404K 4496K CPU0 0 3:47 92.67% openssl 694 bob 1 20 0 6508K 3060K CPU2 2 0:02 0.42% top 714 bob 1 20 0 11188K 6756K select 2 0:00 0.05% sshd 706 bob 1 20 0 11188K 6756K select 2 0:00 0.04% sshd 603 root 1 20 0 8156K 5096K select 1 0:00 0.02% sendmail Far as I can tell the problems with ssh disconnection are tied to running make -jN processes in /usr/ports. Make -j4 buildworld in /usr/src does not obviously interfere with subsequent ssh connections. Running make -j4 -DBATCH in /usr/ports/www/firefox also breaks ssh connections, but not quite so fast, it takes a few minutes for new ssh sessions to fail. Incidentally, the -DBATCH flag is ignored. That make stopped on a stale readline installation. After manually upgrading readline, the make was restarted without -j4 in /usr/ports/www/firefox and, some ten minutes later, the ssh connections are still working. Looks like the trouble is related to make -jN, but only in /usr/ports. It's understood the -j option is not a sure thing in ports, but having it interfere with ssh connections seems most strange. Thanks for reading, bob prohaska
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20171122173613.GA63035>