Linux-Development-Sys Digest #208, Volume #7 Sat, 18 Sep 99 21:14:00 EDT
Contents:
Download Ia.n.i.!!! It's free! (madQ)
ACK storm and resulting amnesia (Wisquatuk)
Re: Figure Out The MS Source Code Yourself ("[EMAIL PROTECTED]")
Re: ACK storm and resulting amnesia (Andi Kleen)
What dist do kernel hackers use? ([EMAIL PROTECTED])
Re: Booting Linux on Sparc Classic using Null-Modem Connection for Console. (Pete
Zaitcev)
Re: What dist do kernel hackers use? ("Selious")
Re: Looking for a project ("Selious")
Re: Linux without PC BIOS ("Selious")
Re: where are process-table, region-table, inode-table? (Juergen Heinzl)
Re: What dist do kernel hackers use? (Juergen Heinzl)
Re: incorrect kernal version error? (Peter Samuelson)
Re: Adding swap space to drive with data... easy? ("Glen Parker")
Re: cross compile under linux (Peter Samuelson)
Re: Statically linked binaries - why so big? (Peter Samuelson)
Re: kernel_lock (Peter Samuelson)
Re: Looking for a project (Peter Samuelson)
Re: write / writev guaranteed autonomous? (David Wragg)
mkraid ("D. Stimits")
Re: Kernel Install Applet (Peter Samuelson)
Re: You can now use Winmodems in Linux!!!!!!! (Peter Samuelson)
----------------------------------------------------------------------------
From: madQ <[EMAIL PROTECTED]>
Subject: Download Ia.n.i.!!! It's free!
Date: 18 Sep 1999 13:16:59 GMT
Download Ia.n.i. RemoteControlSystem 1.2 beta. It's free!!!
New site: http://jump.to/IaniProject
------------------------------
From: Wisquatuk <[EMAIL PROTECTED]>
Subject: ACK storm and resulting amnesia
Crossposted-To: comp.os.linux.networking
Date: Sat, 18 Sep 1999 17:20:20 GMT
=====BEGIN PGP SIGNED MESSAGE=====
Hash: SHA1
(I don't generally crosspost, but I'm not sure which group this fits
into, and I *think* it's on-topic for both, so...)
Okay, so it's happened again. Ugh. Basically, my system seems to be
losing memory... suddenly, the RSS values just don't add up. Here,
take a look at this last bout, which I managed to record before I
barely got the machine to shut down:
PID USER PRI NI SIZE SWAP RSS SHARE STAT %CPU %MEM TIME COMMAND
9911 wisq 18 0 520 52 468 340 R 7.9 2.5 0:04 top
443 root 0 0 492 152 340 260 S 0.0 1.8 3:59 rpc.nfsd
488 rc5des 19 19 328 124 204 144 S N 0.0 1.0 0:43 rc5des
505 rc5des 19 19 328 124 204 144 S N 0.0 1.0 0:06 rc5des
508 rc5des 20 19 328 124 204 144 R N 91.5 1.0 2007m rc5des
580 root 0 0 296 120 176 164 S 0.3 0.9 0:34 nmbd
432 root 0 0 172 108 64 44 S 0.0 0.3 0:10 rpc.mountd
530 root 0 0 100 44 56 40 S 0.1 0.2 0:00 gpm
546 root 0 0 380 328 52 44 S 0.0 0.2 0:14 httpd
1 root 0 0 68 68 0 0 SW 0.0 0.0 0:04 init
2 root 0 0 0 0 0 0 SW 0.0 0.0 0:36 kflushd
3 root 0 0 0 0 0 0 SW 0.0 0.0 0:08 kupdate
4 root 0 0 0 0 0 0 SW 0.0 0.0 0:00 kpiod
5 root 0 0 0 0 0 0 SW 0.0 0.0 1:58 kswapd
<and so on -- this is sorted by RSS, so nothing below here is resident
at *all*>
total used free shared buffers cached
Mem: 18692 18108 584 372 408 1716
-/+ buffers/cache: 15984 2708
Swap: 34028 15680 18348
Stuff swaps tons (which is the main reason the system becomes
near-unusable), and if it goes too far, I've even had the entire
system hang. (Accompanied by a ton of 'out of memory for ...'
messages, and a few init panics. Not that I knew init could panic
until now.)
Oh, and incidentally, I'm not sure if this is related, but it's weird
in any case.. my magic SysRq key has stopped working. I know it's
configured in the kernel. I know it worked not to long ago (in fact,
I used it to safely reboot during one of the near-system-failures). I
know that the keyboard isn't broken, because showkey -s gives me 0x54
codes, and SysRq still gets stuck like it sometimes does, but it
doesn't actually *do* anything anymore. (Even after a cold reboot.)
Meanwhile, the last two times my system's been fried or near-fried
(including the one from which the data above comes from), I've noticed
heavy activity on my ADSL link. This last time, I hastily did a
tcpdump to grab some of it before denying it all at the firewall:
10:44:27.885107 pppaddress.64581 > mail.pop-3: . ack 4069160461 win 0
10:44:27.895107 mail.pop-3 > pppaddress.64581: . ack 1 win 8760 (DF)
10:44:27.895107 pppaddress.64581 > mail.pop-3: . ack 4069160461 win 0
10:44:27.915107 mail.pop-3 > pppaddress.64581: . ack 1 win 8760 (DF)
10:44:27.925107 mail.pop-3 > pppaddress.64581: . ack 1 win 8760 (DF)
10:44:27.925107 mail.pop-3 > pppaddress.64581: . ack 1 win 8760 (DF)
10:44:27.925107 mail.pop-3 > pppaddress.64581: . ack 1 win 8760 (DF)
10:44:27.935107 pppaddress.64581 > mail.pop-3: . ack 4069160461 win 0
10:44:27.935107 pppaddress.64581 > mail.pop-3: . ack 4069160461 win 0
10:44:27.935107 pppaddress.64581 > mail.pop-3: . ack 4069160461 win 0
And so on, seemingly randomly, in a strange sort of ACK loop.
Is this an ACK storm used to attack a link? Or a kernel bug, maybe?
At first I thought the loss of available memory was some sort of
cracker infiltration using a kernel module that hid
processes/files/whatnot, but I disabled kernel module support, and the
problem continued. And then I noticed the ACK traffic. (I've killed
my fetchmail daemons in the meantime, as damage control.)
Further information, to potentially trace this: The host that keeps
dying is furball (192.168.1.2 on my localnet, Linux 2.2.11), whose
packets are forwarded and masqueraded via firewall (192.168.1.1, Linux
2.2.10). Firewall uses a PPPoE ADSL link to the net, via
Sympatico.ca, and the 192.168.1.x machines are connected via 10Mbps
Ethernet.
Oh, and also: I'm not collecting my mail from Sympatico's mailserver;
the ACK traffic above was to/from my *old* ISP's mailserver, which
still gives me an e-mail address for reasons unknown. Hence, I'm
wondering if someone along that route is trying to desynchronise my
link. (I don't see *why*, if they could just sniff it, but..)
Okay, so I'll be quite frank.. I'm totally at a loss to figure out
what the heck is going on. :) Any help would be appreciated.
- --
- Wisq ([EMAIL PROTECTED] to email)
-----BEGIN GEEK CODE BLOCK-----
Version: 3.12
GCS/CC/M d-(--) a--- C++(+++)>$ UL++++>$ P+++ L+++ E>++ W-(+>++) N+++
!o>++ K- w--- O- M- !V PS++(+++) PE- Y+ PGP+++@ t+@ 5 X+++@ R+ tv b+
DI+@ D+ G>+++ e- h!(++) !r z
------END GEEK CODE BLOCK------
=====BEGIN PGP SIGNATURE=====
Version: PGP 6.5.1
iQA/AwUBN+PIhP8wZ8HVeo8aEQKXrgCgi8MPHQmZePB7twwl7yigKcRHNkUAnjD+
xe9CFQIP06/fCC122ymZlE8Z
=HIw2
=====END PGP SIGNATURE=====
------------------------------
From: "[EMAIL PROTECTED]"
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.misc
Subject: Re: Figure Out The MS Source Code Yourself
Date: Sat, 18 Sep 1999 18:29:06 GMT
On 17 Sep 1999 01:41:25 -0500, [EMAIL PROTECTED]
(Peter Samuelson) wrote:
>What offense? Gates has done many things in his career but I seriously
>doubt he has reverse engineered anything.
He pays other people to do it. Find out how he screwed Wang
out of OLE. It's now the very foundation of MicroScrew.
>Why reverse engineer when you can buy?
Or steal it. Like he did with his first paper tape basic
interpreter. Whole thing started with a theft.
>Multipart/* is of the devil. Please exorcise.
Sadly, that's why it will win, just like Gates. Because
good is dumb. /:-!
Luke, come over to the Dark Side....
Behind every great fortune there is a crime.
- Honore de Balzac (1799-1850)
------------------------------
From: Andi Kleen <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.networking
Subject: Re: ACK storm and resulting amnesia
Date: 18 Sep 1999 21:21:47 +0200
Update to 2.2.12. That'll fix the ACK storms.
--
This is like TV. I don't like TV.
------------------------------
From: [EMAIL PROTECTED]
Subject: What dist do kernel hackers use?
Date: Sat, 18 Sep 1999 20:26:13 GMT
Hehe, i love the smell of flame war in the morning.
------------------------------
Subject: Re: Booting Linux on Sparc Classic using Null-Modem Connection for Console.
From: [EMAIL PROTECTED] (Pete Zaitcev)
Date: Sat, 18 Sep 1999 20:33:18 GMT
>: Actually, it seems that RH6.0 install image is broken, at least on
>: sparc64. "boot net", "boot net console=ttya" yeld this:
>: VFS: Mounted root (romfs filesystem).
>: Warning: unable to open an initial console.
tftp64.img from updates helps.
--Pete
------------------------------
From: "Selious" <[EMAIL PROTECTED]>
Subject: Re: What dist do kernel hackers use?
Date: Sat, 18 Sep 1999 23:12:51 +0200
Ehh, RedHat !!
Or Mandrake !!
Some even moved to windows !!
------------------------------
From: "Selious" <[EMAIL PROTECTED]>
Subject: Re: Looking for a project
Date: Sat, 18 Sep 1999 23:14:46 +0200
Make ZIP, gz, tar and other archive standards mountable like I can mount ISO
files !!
But ofcourse r/w !!
That way, I can open, compile and install downloaded tar.gz's without making
a unarchived copy of the contents !!
------------------------------
From: "Selious" <[EMAIL PROTECTED]>
Subject: Re: Linux without PC BIOS
Date: Sat, 18 Sep 1999 23:17:42 +0200
Well, make sure that all BIOS calls used by linux in the real mode (boot.S)
are replaced by your asm that loads the 'sectors' from the ROM's !!
Or does your system use a HD ?? I mean, how does it boot your 'bootsector'
?? And should linux be loaded into RAM, or run from ROM (then you will
discover how 'great' linux is, hehe)
------------------------------
From: [EMAIL PROTECTED] (Juergen Heinzl)
Subject: Re: where are process-table, region-table, inode-table?
Date: Sat, 18 Sep 1999 20:48:02 GMT
In article <7rv1f3$dm$[EMAIL PROTECTED]>, [EMAIL PROTECTED] wrote:
>Hi,
> Can anyone tell me where I can change the text table (the max # of
>executable that a system can run at any time) , region table , inode tables
>and file table? I want to tune my kernel a bit, since it's running slow
>w/ my 486 8M.. somehow the kernel is taking more than 10% of my total memory
>(624k for kernel code, 384k reserved, 376k data).... leaving me enough to
>barely run a X... I read a book called "System Tuning" (base on System V)
>which mentioned something about changing the text table , region table,
>inode tables anf file table to tune the system. Anyone know where I should
>change these defaults? Anyone think that the kernel configuration script
>should allow us to estimate the # of users our system will be serving..and
>then calculate(estimate) the size of these tables? Does these tables
>exist in Linux?
See the kernel documentation for the /proc interface. Linux does things
"slightly" different than the rest of the world. You might have to poke
around in the sources a bit too, it depends.
See http://kernelnotes.org/ for more, at the bottom of the page there
are some links regarding tuning the kernel.
Ta',
Juergen
--
\ Real name : J�rgen Heinzl \ no flames /
\ EMail Private : [EMAIL PROTECTED] \ send money instead /
------------------------------
From: [EMAIL PROTECTED] (Juergen Heinzl)
Subject: Re: What dist do kernel hackers use?
Date: Sat, 18 Sep 1999 21:32:00 GMT
In article <[EMAIL PROTECTED]>, [EMAIL PROTECTED] wrote:
>Hehe, i love the smell of flame war in the morning.
distribution, what's a distribution ?
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: incorrect kernal version error?
Date: 18 Sep 1999 16:45:25 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
> > [root@sleepy hello]# gcc -c hello.c
[Karlo Szabo <[EMAIL PROTECTED]>]
> gcc -D__KERNEL__ -I/usr/src/linux-2.2.12/include -Wall
> -Wstrict-prototypes -O2 -fomit-frame-pointer -pipe -fno-strength-reduce
> -m386 -DCPU=386 -DMODULE -DMODVERSIONS -include
> /usr/src/linux-2.2.12/include/linux/modversions.h -c -o hello.o
> hello.c
And here is the rationale for the important parts above, namely:
-O2 -D__KERNEL__ -DMODULE -I/usr/src/linux-2.2.xx/include
* You need -O2 mainly to make sure gcc inlines those inline functions in
some header files.
* You need -D__KERNEL__ to get the kernel interfaces in many header
files which are designed to be also useful for some userspace
programs.
* You need -DMODULE and possibly -DMODVERSIONS so that <linux/module.h>
will define the necessary things for linking your object as a module as
opposed to linking into the kernel image proper.
* You need the -I flag because with glibc 2.x you can no longer assume
that /usr/include/linux is a symlink to /usr/src/linux/include/linux.
Besides, even with libc 5.x, if you have more than one kernel tree
installed it's hard to be sure /usr/src/linux points to the right
tree all the time. (This is the same problem that crops up if you
follow the bogus advice in the kernel README
cd /usr/src; gzip -dc patch-2.2.12.gz | patch -p0
rather than
cd /usr/src/linux-2.2.11; gzip -cd ../patch-2.2.12.gz | patch -Ep1
which is what I always recommend.)
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: "Glen Parker" <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc,alt.os.linux,comp.os.linux.setup
Subject: Re: Adding swap space to drive with data... easy?
Date: Sat, 18 Sep 1999 14:54:45 -0700
> did you consider using a swap file?
I second that. A *very* simple way to go, without so much as a reboot ;-)
I don't know how much of a performance hit you'll take, but I think if you
look into the priority number assignment, you can make any file-system based
page files inactive until the primary swap space is used up (as apposed to
the striping Linux does between peer page files).
There is (or was anyhow) a daemon available that will add/remove page files
as conditions warrant. I've never used it, but it sounds interesting. The
URL I found at freshmeat is now invalid, but you might have some luck if you
mail the author directly at mailto:[EMAIL PROTECTED]. The daemon is called
"swapd".
HTH
Glen Parker
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Crossposted-To: alt.unix.wizards,comp.os.linux.development.apps
Subject: Re: cross compile under linux
Date: 18 Sep 1999 17:57:48 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Vie^.t Nha^n <[EMAIL PROTECTED]>]
> yes, you can use gcc in linux, it must be on the architecture that
> you want to run your prog on (e.g alpha/linux; sparc/linux;
> intel/linux...).
While I was under the impression that OSF/1 (as it was called at the
time) offered the ability to run Linux/Alpha binaries, or maybe it was
the other way around, I don't know the details of that. (One question
mark is that DU 3.2 uses COFF and I don't know if it can also use ELF.)
What I do know is that you can configure gcc and binutils for a cross-
compilation target. Once you have your toolchain (as Cygnus calls it)
in place, you have to either copy or nfs-mount the include files and
libraries from the target environment onto your build machine and point
gcc at them. Details are in the gcc and binutils documentation.
Last I checked there were some issues with trying to do this between a
32-bit and a 64-bit environment. If this is still true, your build
machine may need to be Alpha/Linux rather than, say, i386/Linux.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Statically linked binaries - why so big?
Date: 18 Sep 1999 18:50:18 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Todd Knarr <[EMAIL PROTECTED]>]
> Repairs. Suppose, during the course of an upgrade, you hose libc.so
> thoroughly. All the dynamically-linked programs now will not run,
> that includes the shell and all the things like cp, mv, tar,
> rpm/dpkg, gzip, cpio and so on that you might use to get a working
> version of libc back. Statically-linked binaries, though, will run
> even when libc is hosed.
If static binaries get too big (so that you have to provide a dummy
exit() and so forth) the alternative would be to dynamically link your
essential tools against a duplicate copy of libc and ld-linux.so etc.
The duplicate libs would sit in /stand/lib or something and would be
just as impervious to screw-ups as the original statics would have
been.
The other thing to do is have one static binary swiss army knife. That
would be sash (the standalone shell), which is a bit primitive in the
UI dept but has builtins corresponding to ls, mv, cp, rm and much much
more.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: kernel_lock
Date: 18 Sep 1999 19:01:42 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[David Lee <[EMAIL PROTECTED]>]
> Does anyone know the purpose of kernel_lock(). It is used for
> pervent task switching in kernel or to prevent other processes to
> enter kernel.
Hmmmm, my 2.2 source tree does not have a kernel_lock(). Perhaps you
meant lock_kernel()? (Or is kernel_lock() something that 2.0 used?)
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Looking for a project
Date: 18 Sep 1999 19:28:45 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Rishabh Tandon <[EMAIL PROTECTED]>]
> guys I'm looking for an undergrad project in linux. If u have
> anything in mind please let me know.
Create another journalled filesystem to compete with the other five:
lfs, dtfs, xfs, ext3fs and reiserfs. (The first three are vaporware;
the last two exist but still have beta journalling support).
[That was a joke.]
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: David Wragg <[EMAIL PROTECTED]>
Subject: Re: write / writev guaranteed autonomous?
Date: 18 Sep 1999 13:14:28 +0000
"Glen Parker" <[EMAIL PROTECTED]> writes:
> man 2 write says this (under the errors section):
>
> EINTR The call was interrupted by a signal before any
> data was written.
>
> That's pretty specific, but I wonder how portable code that relies on
> that would be...
That is basically consistent with the POSIX.1 text for EINTR returned
by write(), which says that either no data was written, or the file
does not support partial writes (which doesn't apply to the case under
discussion).
David Wragg
------------------------------
Date: Sat, 18 Sep 1999 18:04:05 -0600
From: "D. Stimits" <[EMAIL PROTECTED]>
Reply-To: [EMAIL PROTECTED]
Subject: mkraid
Hi:
I had an older raid 0 setup, which I deleted in preparation for moving
to a newer kernel (was 2.1.130, 2.2.12). I intended to use the newer
raidtools, but I'm having a problem. There may be some precondition
the documents are not listing.
Anyway, a sample is:
mkraid /dev/md0 (when mdtab is properly set up).
Then of course I end up doing:
mkraid --really-force /dev/md0
It always fails, stating that it is aborting due to finding an ext2
filesystem on it. I even used fdisk to flag those partitions as type 0
(empty) as a last resort. I also deleted and recreated them.
What is necessary to get mkraid --really-force to continue to run? Do
I have to format with some non-ext2 system? Is newer raid broken?
According to /proc, I have raid support (and I specifically compiled
the kernel for this).
For reference, this is RH6.
Thanks!
D. Stimits, [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: Kernel Install Applet
Date: 18 Sep 1999 19:16:21 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
> ------------------ Posted via CNET Linux Help ------------------
> http://www.searchlinux.com
Oooh, looky, looky, it's CNET Linux Help again! This footer conveys
roughly the same information to a seasoned usenetter that "@aol.com"
did in 1993 (and to some extent still does).
[Cocheese <[EMAIL PROTECTED]>]
> After teaching a few crash courses to the other service techs I work
> with about Linux and the programs, I got to thinking about a few
> "Suggestions".
> Nobody I know personally has ever installed the kernel and lived to
> tell about it. In fact, a friend of mine and I spent an entire week
> trying and to no such luck.
There is absolutely nothing difficult about installing a new kernel.
Nothing. (Well, LILO can be a little arcane; the first time I needed
to use it I spent half an hour reading docs.) If you can't figure it
out from the README file and the Documentation directory that comes
with the kernel tarball, you really don't need to be in charge of a
Unix/Linux system. If you're not sure how to unpack the tarball in the
first place, you really really *really* don't need to be in charge of a
Unix/Linux system.
If your use of Linux is so casual as not to need an actual Unix person
around, you most likely don't need to recompile your kernel.
Prepackaged kernels distributed via .deb or .rpm should work.
> If anyone is up to the challenge I assure you many of the
> distribution's would eat that program right up (since they are really
> going all out to make Linux a little more user friendly- thus getting
> your name in the "Program Hall Of Fame" -LOL
Debian has scripts to compile a kernel from the source packages they
distribute. As far as I know, though (never having tried them), these
will pick the Debian default options, and if you want to do that you
aren't buying a whole lot over just installing a binary image.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
From: [EMAIL PROTECTED] (Peter Samuelson)
Subject: Re: You can now use Winmodems in Linux!!!!!!!
Date: 18 Sep 1999 19:20:16 -0500
Reply-To: Peter Samuelson <[EMAIL PROTECTED]>
[Fox <[EMAIL PROTECTED]>]
> I just a beginner of Linux,
[...]
> ------------------ Posted via CNET Linux Help ------------------
Redundant.
--
Peter Samuelson
<sampo.creighton.edu!psamuels>
------------------------------
** FOR YOUR REFERENCE **
The service address, to which questions about the list itself and requests
to be added to or deleted from it should be directed, is:
Internet: [EMAIL PROTECTED]
You can send mail to the entire list (and comp.os.linux.development.system) via:
Internet: [EMAIL PROTECTED]
Linux may be obtained via one of these FTP sites:
ftp.funet.fi pub/Linux
tsx-11.mit.edu pub/linux
sunsite.unc.edu pub/Linux
End of Linux-Development-System Digest
******************************
