Date: Thu, 20 Feb 1997 14:52:56 -0800 (PST) From: Mark.Mayo To: freebsd-gnats-submit@freebsd.org Subject: misc/2784: userland PPP rises load to 1.00 Message-ID: <199702202252.OAA12677@freefall.freebsd.org> Resent-Message-ID: <199702202300.PAA13284@freefall.freebsd.org>
next in thread | raw e-mail | index | archive | help
>Number: 2784
>Category: misc
>Synopsis: userland PPP rises load to 1.00
>Confidential: no
>Severity: serious
>Priority: medium
>Responsible: freebsd-bugs
>State: open
>Class: sw-bug
>Submitter-Id: current-users
>Arrival-Date: Thu Feb 20 15:00:01 PST 1997
>Last-Modified:
>Originator: mark@quickweb.com
>Organization:
RingZero Computing
>Release: 2.2-GAMMA
>Environment:
FreeBSD celebris.quickweb.com 2.2-GAMMA FreeBSD 2.2-GAMMA #0: Thu Feb 13 12:36:53 EST 1997 mark@celebris.quickweb.com:/usr/src/sys/compile/RingZero i386
>Description:
Quite often (roughly 1/3 of the time) when I run userland PPP to connect
to my ISP, the load average on my machine rises to 1.00. Obviously, PPP
should not be using that much of my resources - and I don't think it is.
It seems to be "artificially" inflating the load average. When I monitor
the machine when it's in this state with top of systat, the CPU is usually
99% idle, as in this output from TOP:
last pid: 219; load averages: 1.22, 0.68, 0.38 17:39:46
27 processes: 1 running, 26 sleeping
CPU states: 0.0% user, 0.0% nice, 0.0% system, 0.0% interrupt, 100% idle
Mem: 11M Active, 3968K Inact, 6600K Wired, 3583K Buf, 8788K Free
If my machine load was .33 before I run ppp, afterwards it rises to roughly
1.3. It seems like it just adds +1 to the load! Stange. I'm not sure if
this is a bug in ppp, or just the way load average is calculated. In any case,
it looks really weird, and is troublesome to a new-comer to see the load
indicate that my machine is 'maxed out'.
I also see the load rising like this when I'm running the Afterstep (1.0pre2)
window manager - but the load only rises to 0.33 and levels off right there
for eternity. Even though I'm not doing a damn thing. SOmetime, I'll do
something (like startup a CPU/memory intensive app) and the load will
drop back down to 0.00 - where I think it should be. Again, I can't seem to
figure out how load is calculated, and I suspect the odd numbers come
from there...
The ppp problem is independant of whether or not there is data moving
through the connection.
>How-To-Repeat:
Run userland PPP 10 times. The load will almost certainly rise to above
1.00 one of these times... For me, it happend rought 1/3 of the time. Also,
it doesn't seem dependant on a particular ISP - it has happened on 3
different ISPs.
>Fix:
I'm not technically competent enough to figure this one out - I do have
the results of a ktrace -> kdump however, if they are useful:
With no data moving through the link:
524 ppp PSIG SIGALRM caught handler=0x11910 mask=0x0 code=0x0
524 ppp RET select -1 errno 4 Interrupted system call
524 ppp CALL sigreturn(0xefbfd2ac)
524 ppp RET sigreturn JUSTRETURN
524 ppp CALL select(0xe,0xefbfdb54,0xefbfdb34,0xefbfdb14,0)
524 ppp PSIG SIGALRM caught handler=0x11910 mask=0x0 code=0x0
524 ppp RET select -1 errno 4 Interrupted system call
524 ppp CALL sigreturn(0xefbfd2ac)
524 ppp RET sigreturn JUSTRETURN
524 ppp CALL select(0xe,0xefbfdb54,0xefbfdb34,0xefbfdb14,0)
524 ppp PSIG SIGALRM caught handler=0x11910 mask=0x0 code=0x0
524 ppp RET select -1 errno 4 Interrupted system call
524 ppp CALL sigreturn(0xefbfd2ac)
524 ppp RET sigreturn JUSTRETURN
524 ppp CALL selec
...
524 ppp CALL sigreturn(0xefbfd2ac)
524 ppp RET sigreturn JUSTRETURN
524 ppp CALL sigprocmask(0x1,0x2000)
524 ppp RET sigprocmask 0
524 ppp CALL sigprocmask(0x3,0)
524 ppp RET sigprocmask 8192/0x2000
524 ppp CALL sigprocmask(0x1,0x2000)
524 ppp RET sigprocmask 0
524 ppp CALL sigprocmask(0x3,0)
524 ppp RET sigprocmask 8192/0x2000
...
And here is with a ping through the ppp link
524 ppp CALL select(0xe,0xefbfdb54,0xefbfdb34,0xefbfdb14,0)
524 ppp RET select 1
524 ppp CALL read(0x4,0xefbfd334,0x7d0)
524 ppp GIO fd 4 read 64 bytes
"E\0\0@\v\0\0@\^Q\^E\^E\M^F1M\b\^D\^R\0005\0,@\^A\0\0\^A\0\0\0\
\0\0\0\^Evinyl\bquickweb\^Ccom\0\0\^A\0\^A"
524 ppp RET read 64/0x40
524 ppp CALL sigprocmask(0x1,0x2000)
524 ppp RET sigprocmask 0
524 ppp CALL sigprocmask(0x3,0)
524 ppp RET sigprocmask 8192/0x2000
524 ppp CALL select(0xe,0xefbfdb54,0xefbfdb34,0xefbfdb14,0)
524 ppp RET select 1
524 ppp CALL write(0x5,0x5f080,0x45)
524 ppp GIO fd 5 wrote 69 bytes
"~!E\0\0@\v\0\0@\^Q\^E\^E\M^F1M\b\^D\^R\0005\0,@\^A\0\0\^A\0\0\
\0\0\0\0\^Evinyl\bquickweb\^Ccom\0\0\^A\0\^A\M^N~"
524 ppp RET write 69/0x45
524 ppp CALL select(0xe,0xefbfdb54,0xefbfdb34,0xefbfdb14,0)
524 ppp PSIG SIGALRM caught handler=0x11910 mask=0x0 code=0x0
:
>Audit-Trail:
>Unformatted:
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199702202252.OAA12677>