Date: Sun, 12 Jul 1998 22:35:58 +0300 From: Alexander Litvin <archer@lucky.net> To: Terry Lambert <tlambert@primenet.com> Cc: current@FreeBSD.ORG Subject: Re: Arrgh ! resubscribing again again again.... Message-ID: <19980712223558.37658@carrier.kiev.ua> In-Reply-To: <199807120621.XAA29796@usr07.primenet.com>; from Terry Lambert on Sun, Jul 12, 1998 at 06:21:52AM %2B0000 References: <199807120536.IAA07205@grape.carrier.kiev.ua> <199807120621.XAA29796@usr07.primenet.com>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sun, Jul 12, 1998 at 06:21:52AM +0000, Terry Lambert wrote:
> > TL> You don't *have* to do that. You *could* shutdown your X server and
> > TL> restart it. You simply choose to have your demons die, instead of
> > TL> doing this and/or limiting the amount of VM available to the X server
> > TL> and Netscape, or disabling the X shm extensions.
> >
> > Terry, your explanation is ok. But what X-server should I shutdown?
> > For I don't have any X's. And yesterday I repeated 'make -j64 buildworld',
> > and as always: sendmail childs die with SIGSEGV, cron seems to be in
> > memory, but no cron jobs run. And all that after makeworld has finished,
> > system is _absolutely_ idle.
>
> Interesting. Clearly, you're not the person running Netscape to
> cause a similar problem. 8-).
>
> Do you have any zombie processes?
No.
> Have you looked at the vmstat output?
Noting interesting, it seems. Just ordinary idle system.
> If you kill -HUP cron, what happens?
It doesn't help, at least in the cases I tried it.
> If you kill -HUP sendmail, what happens?
Well, sendmail restarts. May be it is not the same always. Though today
it restarted just fine.
> Did you get any errors in the buildworld, or did it complete without
> damaging itself (only damaging sendmail and cron)?
Well, sometimes it worked, simetimes -- no, though I cannot say for
sure whether it was due to VM shortage, or because the world was broken
at the time (I tested it several times with sources of different dates).
> Are you overclocking your CPU?
No.
> How much memory do you have -- is it enough that if you missed a
> refresh cycle or two (say, because you did something silly, like
> a buildworld with -j64 and you have an Adaptec controller with a
> high bus-on time, or you are using a PIO IDE controller, etc...),
> that the upper memory contents might degrade and cause these problems?
I do not exactly understand what you're asking about here. Here's my boot
log:
Jul 10 23:35:06 grape /kernel: FreeBSD 3.0-CURRENT #96: Fri Jul 10 22:22:46 EEST
1998
Jul 10 23:35:06 grape /kernel: archer@grape.carrier.kiev.ua:/usr/src/sys/compile
/GRAPE
Jul 10 23:35:06 grape /kernel: Timecounter "i8254" frequency 1193182 Hz cost 4
239 ns
Jul 10 23:35:06 grape /kernel: Timecounter "TSC" frequency 120053411 Hz cost 1
96 ns
Jul 10 23:35:06 grape /kernel: CPU: Pentium/P54C (120.05-MHz 586-class CPU)
Jul 10 23:35:06 grape /kernel: Origin = "GenuineIntel" Id = 0x52c Stepping=12
Jul 10 23:35:06 grape /kernel: Features=0x1bf<FPU,VME,DE,PSE,TSC,MSR,MCE,CX8>
Jul 10 23:35:06 grape /kernel: real memory = 33554432 (32768K bytes)
Jul 10 23:35:06 grape /kernel: avail memory = 30408704 (29696K bytes)
Jul 10 23:35:06 grape /kernel: Probing for devices on PCI bus 0:
Jul 10 23:35:06 grape /kernel: chip0: <VIA 82C585 (Apollo VP1/VPX) system contro
ller> rev 0x23 on pci0.0.0
Jul 10 23:35:06 grape /kernel: chip1: <VIA 82C586 PCI-ISA bridge> rev 0x27 on pc
i0.7.0
Jul 10 23:35:06 grape /kernel: ide_pci0: <VIA 82C586x (Apollo) Bus-master IDE co
ntroller> rev 0x06 on pci0.7.1
Jul 10 23:35:06 grape /kernel: Probing for devices on the ISA bus:
Jul 10 23:35:06 grape /kernel: sc0 at 0x60-0x6f irq 1 on motherboard
Jul 10 23:35:06 grape /kernel: sc0: VGA color <16 virtual consoles, flags=0x0>
Jul 10 23:35:06 grape /kernel: sio0 at 0x3f8-0x3ff irq 4 on isa
Jul 10 23:35:06 grape /kernel: sio0: type 16550A
Jul 10 23:35:06 grape /kernel: sio1 at 0x2f8-0x2ff irq 3 on isa
Jul 10 23:35:06 grape /kernel: sio1: type 16550A
Jul 10 23:35:06 grape /kernel: fdc0 at 0x3f0-0x3f7 irq 6 drq 2 on isa
Jul 10 23:35:06 grape /kernel: wdc0 at 0x1f0-0x1f7 irq 14 flags 0xa0ffa0ff on is
a
Jul 10 23:35:06 grape /kernel: wdc0: unit 0 (wd0): <ST31277A>, DMA, 32-bit, mult
i-block-16
Jul 10 23:35:06 grape /kernel: wd0: 1221MB (2501856 sectors), 2482 cyls, 16 head
s, 63 S/T, 512 B/S
Jul 10 23:35:06 grape /kernel: wdc1 at 0x170-0x177 irq 15 on isa
Jul 10 23:35:06 grape /kernel: wdc1: unit 0 (wd2): <Maxtor 7270 AV>
Jul 10 23:35:06 grape /kernel: wd2: 257MB (527450 sectors), 959 cyls, 11 heads,
50 S/T, 512 B/S
Jul 10 23:35:06 grape /kernel: npx0 flags 0x1 on motherboard
Jul 10 23:35:07 grape /kernel: npx0: INT 16 interface
Jul 10 23:35:07 grape /kernel: Intel Pentium F00F detected, installing workaroun
d
> Does -j32 have the same problem?
No, there is enough memory to make world with -j32.
> -j16?
>
> etc.?
>
> It would be nice to know if this problem was the result of running
> out of VM, or if you can cause it by load without actually exhasting
> the VM...
>
> Makeworld takes a while; have you tried putting a crontab entry that
> spits the time out to the console once a minute, and bracketing the
> buildworld with "date"?
>
> Maybe it's a particular section of the tree, and not the load, that's
> getting you? Perhaps that region of the disk is corrupt?
I can repeat the effect without building world. In fact I may cause
it in some 1-2 minutes -- much faster than with makeworld. I just
encountered it for the first time trying to test softupdates, when
I ran 'make -j64 buildworld'.
Just the same happenes if I just exhaust VM with some other means, e.g.,
start some 10-15 progs, each mallocing 32M of memory and durtying it.
> Obviously, I don't expect you to answer all of these to the list, or
> even to me, if the answer to the questions is "no impact".
>
>
> Terry Lambert
> terry@lambert.org
> ---
> Any opinions in this posting are my own and not those of my present
> or previous employers.
---
Twenty Percent of Zero is Better than Nothing.
-- Walt Kelly
To Unsubscribe: send mail to majordomo@FreeBSD.org
with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?19980712223558.37658>