Linux-Development-Sys Digest #410, Volume #6 Thu, 18 Feb 99 12:14:21 EST
Contents:
syscall()? (Jim Seymour)
Re: 2.2.1: strange SMP (dual celeron) startup msgs (BL)
Re: Linux programming jobs? (Marco Anglesio)
Should serial device interrupts appear in /proc/interrupts? (Joe Pfeiffer)
Re: Bug in select? Behavior changes midstream (Felix Morley Finch)
Unresolved symbol __ucmpdi2 in driver ("Christoph Goos")
Re: Problem with autofs and local /home (Karl Heyes)
Re: Long UID/GID and POSIX compliance (Martin Recktenwald)
Re: Clock Skew (Julian Robert Yon)
Re: Long UID/GID and POSIX compliance (Mark Tranchant)
Re: 2.2.1: strange SMP (dual celeron) startup msgs (M Sweger)
Re: need help with kernel OOPS message (David Kirkpatrick)
My first Y2K problem (Diego Betancor)
Serious NFS bug in Linux 2.2.x (Solaris server) (Oliver Stahlhut)
Re: 2.2.1: strange SMP (dual celeron) startup msgs (BL)
Re: Serious NFS bug in Linux 2.2.x (Solaris server) (PFENNIGER Daniel)
----------------------------------------------------------------------------
From: [EMAIL PROTECTED] (Jim Seymour)
Subject: syscall()?
Date: 17 Feb 1999 19:46:24 GMT
Hi,
I'm trying to resurrect XView on Linux. Amongst other things, there's
this little tidbit:
In a thing called "linux_select.c":
/* Function for calling the select(2) system call in linux.
* Linux doesn't have a syscall() function, this replaces it
* for select(), fcntl() and read().
* Kudos to Rick Sladkey ([EMAIL PROTECTED]) for suggesting
* this method. */
This file doesn't compile. Before I hare-off trying to repair it,
I'm wondering if the comments above are still correct.
I note that "man syscall" says there's no such thing, but grepping
for "syscall" in /usr/include seems to indicate that there is. So,
does it exist? Is it correct? And where might I find the missing
docs?
Red Hat Linux 5.0, kernel 2.0.32, GCC 2.7.2.3.
TIA,
Jim
--
Jim Seymour | PGP Public Key available at:
[EMAIL PROTECTED] | http://www.uk.pgp.net/pgpnet/pks-commands.html
http://home.msen.com/~jimsun | http://www.trustcenter.de/cgi-bin/SearchCert.cgi
------------------------------
From: BL <[EMAIL PROTECTED]>
Subject: Re: 2.2.1: strange SMP (dual celeron) startup msgs
Date: 18 Feb 1999 05:44:56 GMT
Reply-To: no.spambots.please
Paul Rubin <[EMAIL PROTECTED]> wrote:
: So what's the outcome? Disregarding the confusing console log, does
: the system work?
yes!
: How long does it take to do a kernel compile
: compared to a single cpu? Enquiring minds...
man, you're going to be green with envy ;-) I beat the 2 minute build ;-)
and on the 2.2.1 kernel, that's impressive (I think). it was 1:50 to build
"make -j2" on the dual 450 celerys. The malaysia chips (boxed 300a) really DO
go to 450 (add some cooling and do the 2.2volt mod, first).
The chips costed about $99 each (they raised the price just yesterday since
they knew I was coming over to buy some) ;-) and they're in very short supply
(the malaysia chips, that is). with a pair of costa rica chips (the only
other place that manufactures them) I was only able to clock them at 337mhz.
still, at that rate, I could the 2.2.1 at 2:30 or so.
and from 1cpu to 2cpus, it was very close to 100% (like 90% or so).
(worth mentioning that this was on a fast disk array, running hardware raid).
--
AntiSpam: For email, change all 'zero' chars to letter 'o' chars.
------------------------------
From: [EMAIL PROTECTED] (Marco Anglesio)
Crossposted-To: comp.os.linux.development.apps,comp.os.linux.networking
Subject: Re: Linux programming jobs?
Reply-To: [EMAIL PROTECTED]
Date: Wed, 17 Feb 1999 20:04:42 GMT
I'd also mention that you're in the capital of a province just coming out
of recession, with a government about to hand down a budget in the red and
no other industries, really, around.
The reason that you're having trouble finding a job in Victoria is because
there probably aren't many, and the ones that you are fighting for, you
are fighting for against an oversupply of UVic graduates.
I would try to take a job as a tech and work your way up. One does not
become a programmer in a day. Open source experience (putting together
projects, learning internals) will help. So will, sad to say, getting at
least a couple of years in university. Even if it has nothing to do with
programming at all, the existence of some university makes a difference.
marco
--
,--------------------------------------------------------------------------.
> | We've been doing anal probing <
> Marco Anglesio | for years now, and all we've <
> [EMAIL PROTECTED] | discovered is that one in ten <
> http://www.the-wire.com/~mpa | doesn't seem to mind. <
> | --Dave Foley, of Kids in the Hall <
`--------------------------------------------------------------------------'
------------------------------
From: Joe Pfeiffer <[EMAIL PROTECTED]>
Subject: Should serial device interrupts appear in /proc/interrupts?
Date: 17 Feb 1999 23:08:30 -0700
(almost forgot -- Linux 2.2.1)
It seems as if the answer should be yes. However, my rc.serial file
contains
${SETSERIAL} /dev/ttyS2 uart 16550 port 0x3e8 irq 12
but my /proc/interrupts looks like
CPU0
0: 27692369 XT-PIC timer
1: 79980 XT-PIC keyboard
2: 0 XT-PIC cascade
3: 2189587 XT-PIC serial
4: 267087 XT-PIC serial
8: 1 XT-PIC rtc
9: 0 XT-PIC serial
11: 48797 XT-PIC NE2000
13: 1 XT-PIC fpu
14: 172726 XT-PIC ide0
15: 4 XT-PIC ide1
NMI: 0
--
Joseph J. Pfeiffer, Jr., Ph.D. Phone -- (505) 646-1605
Department of Computer Science FAX -- (505) 646-1002
New Mexico State University http://www.cs.nmsu.edu/~pfeiffer
------------------------------
From: [EMAIL PROTECTED] (Felix Morley Finch)
Subject: Re: Bug in select? Behavior changes midstream
Date: 18 Feb 1999 07:32:15 GMT
On Thu, 18 Feb 1999 01:52:22 GMT, Alan Curry <[EMAIL PROTECTED]> intoned:
>Now there are once again no writers on the FIFO. And here's the rule you
>apparently weren't aware of: when a FIFO has no writers, an EOF condition is
>placed on it. From now on, the FIFO will always be considered readable,
>because its EOF has been reached. That's why you are reading 0 bytes. A 0
>byte read is an EOF. The only way to reset it is to close all readers and do
>your RDONLY open() again.
>
>But you shouldn't do that, because there is a brief period of time between
>when you detect the EOF and when you close the FIFO that some other writer
>could show up and write something to the pipe. Then you'll close the reader
>and that data will be lost.
Ahhh.... it was the change from before reading anything to after that
baffled me. I don't quite understand why there's no EOF condition
placed from the beginning, but I'll go find the FAQ and dig in. I
wonder if I will remember this a year from now when I do it
again... :-)
Thanks very much.
--
... _._. ._ ._. . _._. ._. ___ .__ ._. . .__. ._ .. ._.
Felix Finch: scarecrow repairman & rocket surgeon / [EMAIL PROTECTED]
PGP = 91 B3 94 7C E9 E8 76 2D E1 63 51 AA A0 48 89 2F ITAR license #4933
I've found a solution to Fermat's Last Theorem but I see I've run out of room o
------------------------------
From: "Christoph Goos" <[EMAIL PROTECTED]>
Subject: Unresolved symbol __ucmpdi2 in driver
Date: Thu, 18 Feb 1999 07:49:08 GMT
Hello,
in a network driver I developed, I get an unresolved symbol __ucmpdi2
while doing something like that:
unsigned long long a,b;
a=20;
b=10;
if((a-b) == 0) {
do something
}
I can work around this by changigng the if-statement to:
if(!(a-b)) ...
but for some reasons I would like to keep it in the first form.
Any advice what to do?
Thanks for any help,
Christoph
Christoph Goos <[EMAIL PROTECTED]> http://www.syskonnect.de
SysKonnect - The Server Connectivity Company
------------------------------
From: Karl Heyes <[EMAIL PROTECTED]>
Subject: Re: Problem with autofs and local /home
Date: Thu, 18 Feb 1999 09:56:09 -0500
Craig J Copi wrote:
> In article <7a2oiu$lj2$[EMAIL PROTECTED]>,
> [EMAIL PROTECTED] (H. Peter Anvin) writes:
> > Followup to: <7a2iea$igs$[EMAIL PROTECTED]>
> > By author: [EMAIL PROTECTED] (Craig J Copi)
> > In newsgroup: comp.os.linux.development.system
> >>
> >> I am using the latest autofs with kernel 2.2.1. The users home directory
> >> actually lives in /export/home by I use autofs to "mount" this in /home
> >> (/export/home is then nfs exported and autofs on clients map them to /home
> >> also). On external clients it works great. On local clients it makes a
> >> symbolic as it should so
> >> /home/test -> /export/home/test
> >> Now I do the following
> >> root> cd /home/test
> >> root> pwd
> >> /home/test
> >>
> >> This is good.
> >>
> >> root> su - test
> >> test> pwd
> >> /export/home/test
> >>
> >> This is bad! This breaks scripts (or GNU queue) that do things like
> >> rsh remotehost "cd `pwd`; ./runbinary_in_this_dir"
> >> I don't know why the two cases above are giving different results.
> >>
> >> Any help would be appreciated.
> >>
> >
> > "pwd" is a bash built-in, so it knows how you got there. "su", or "rsh"
> > don't have that information.
> >
> > -hpa
>
> So does this mean I'm stuck? On the nfs server I have to keep the
> home directories in /home, export them from here, and not use autofs if I
> want to maintain similar looking directory structures across multiple
> machines? Is there a simple work around for this that I'm missing? This
> would seem like a big problem if I were using nis maps for autofs.
>
A feature was added to 2.2 to help with this facility. The dentry scheme and
loopback (something like that!!, not the network interface) module enabled you
to mount certain directories in other areas. I'll have to look at the details
but
this is what you want.
karl
------------------------------
From: Martin Recktenwald <[EMAIL PROTECTED]>
Subject: Re: Long UID/GID and POSIX compliance
Date: 18 Feb 1999 11:11:33 +0100
Mark Tranchant <[EMAIL PROTECTED]> writes:
> What's to be done, or am I doomed to run Windows forever on my x86 box
> here?
IIRC 32 bit uids/gids are likely to be added along the 2.3.x
development tree. Currently you have to live with 16 bit uids/gids.
Martin.
------------------------------
From: Julian Robert Yon <[EMAIL PROTECTED]>
Subject: Re: Clock Skew
Date: Thu, 18 Feb 1999 11:13:10 +0000
On Thu, 18 Feb 1999, Alex Rhomberg wrote:
> So here comes the command for the lazy. Just replace MY_TIMESERVER with
> the name
> of you central time source or NFS server and run :)
> echo '#\!/bin/sh\nping -q -c 1 MY_TIMESERVER > /dev/null && rdate -s
> MY_TIMESERVER' > /etc/cron.hourly/getnettime;chmod 744
> /etc/cron.hourly/getnettime
>
> I use this exactly to avoid clock skew warnings in make
Thank you kindly - will do!
Julian
~~~~~~~~~~~~~~~~~~~~
[EMAIL PROTECTED]
[EMAIL PROTECTED]
------------------------------
From: Mark Tranchant <[EMAIL PROTECTED]>
Subject: Re: Long UID/GID and POSIX compliance
Date: Wed, 17 Feb 1999 11:24:35 +0000
Reply-To: [EMAIL PROTECTED]
^a byte^a 16-bit integer
Sorry.
Mark Tranchant wrote:
>
> I know this one has come up before, but it's frustrating me. At work,
> the NIS servers are running Solaris, and our UIDs and GIDs may be
> greater than 65535. If I want to run Linux and use my network drives, I
> can't, because one of the groups I am in is numbered 120,000-ish.
>
> I *know* (or am lead to believe) that POSIX says t_uid and t_gid are of
> type short int, which is a byte on IA-32 systems, but this is
> restricting usage, which I am sure is not what the POSIX crew intended.
>
> What's to be done, or am I doomed to run Windows forever on my x86 box
> here?
>
> Cheers,
>
> Mark.
------------------------------
From: [EMAIL PROTECTED] (M Sweger)
Subject: Re: 2.2.1: strange SMP (dual celeron) startup msgs
Date: 18 Feb 1999 12:24:41 GMT
Reply-To: [EMAIL PROTECTED]
: Hmmm... 2:30 seems mysteriously fast. I get 3:28.89elapsed for
: a "make -j2 bzImage" with dual 333 Mhz PII and an ultrawide
: 40 Mbytes/sec SCSI disk interface.
Just so you know that the UW SCSI can go 80 Mbytes/sec. At least mine
is cable of doing it. The only reason that the scsi has been limited
by Linux is that the I/O buss between the controller and the memory
is still 33mhz. Thus, why make it 80 mb/s when the adaptor card will
just be put into wait mode due to the I/O bus being so slow. Hence,
the closest match between scsi and memory bus is 40 mb/s.
--
Mike,
[EMAIL PROTECTED]
------------------------------
From: David Kirkpatrick <[EMAIL PROTECTED]>
Subject: Re: need help with kernel OOPS message
Date: Thu, 18 Feb 1999 09:03:31 +0000
Reply-To: [EMAIL PROTECTED]
Do a find * -name oops* -print to find oops-tracing which will
give you some info.
Stuart Marshall wrote:
>
> Hi,
>
> I have a soft-real-time process which occasionally gets a kernel
> oops and is then killed by the kernel. Sometimes the machine locks
> up and gets rebooted by a watchdog card. I have enclosed the most
> recent oops message below. I would like to know how to proceed from
> here. Should I gather more info or should I send a possible bug
> report to someone? Unfortunately I don't understand what the
> meaning of "do_mmap+800/832" is. The process that is running, camserverd,
> is threaded, runs in SCHED_FIFO mode, and is locked into memory if that
> matters.
>
> Any suggestions on how to proceed are welcome.
>
> thanks,
> Stuart
> --------------------------------------------------------------------
>
> klogd startup message:
>
> klogd 1.3-3#31, log source = /proc/kmsg started.
> Loaded 3376 symbols from /boot/System.map-2.0.36.
> Symbols match kernel version 2.0.36.
> Loaded 70 symbols from 3 modules.
>
> kernel oops message:
>
> Unable to handle kernel paging request at virtual address 4111d000
> current->tss.cr3 = 00c09000, %cr3 = 00c09000
> *pde = 00810067
> *pte = 00000000
> Oops: 0000
> CPU: 0
> EIP: 0010:[do_mmap+800/832do_mmap+800/832]
> EFLAGS: 00010202
> eax: 00000200 ebx: 02392b18 ecx: 02ae4a18 edx: 4111d000
> esi: 00002070 edi: 00200000 ebp: 00000000 esp: 016bcf80
> ds: 0018 es: 0018 fs: 002b gs: 002b ss: 0018
> Process camserverd (pid: 21580, process nr: 25, stackpage=016bc000)
> Stack: bfffea9c 00000000 400c8640 bfffeac0 02ae4a18 001111d3 00000000 4111d000
> 00200000 00000000 00000022 00000000 02973414 400c8640 0010a8d9 bfffea9c
> 00000080 400cc8ec 400c8640 400c8640 bfffeac0 ffffffda 0000002b 0000002b
> Call Trace: [old_mmap+115/128] [system_call+85/124]
> Code: 64 8a 02 81 c7 00 f0 ff ff 81 c2 00 10 00 00 85 ff 75 ed 8b
--
[EMAIL PROTECTED]
[EMAIL PROTECTED]
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Diego Betancor)
Subject: My first Y2K problem
Date: Thu, 18 Feb 1999 15:34:31 GMT
Everyday at 3:05 am cron goes and runs
netdate -l0 times.cs.columbia.edu ntp0.cornell.edu jane/jpl.nasa.gov
What was my surprise that in monday I was in february 2036. After
some inspection columbia's clock gave me that date. cornell and nasa
didn't respond.
Seems to me that 2036 maybe at the end of time_t and that cornell
and nasa discovered and shut down their clocks but columbia did not.
As an after thought never use -l0 in netdate (so at least 2 clocks
have to give the date) and we will begin to see problems like this
more often from now on...
Diego
Diego Betancor @ Duo Business Communications
for email: dbetancor is my userid and my company's domain is duocom.net
** Do not send me unsolicited commercial e-mail spam of any kind **
------------------------------
From: Oliver Stahlhut <[EMAIL PROTECTED]>
Subject: Serious NFS bug in Linux 2.2.x (Solaris server)
Date: Thu, 18 Feb 1999 15:20:46 +0000
Hi there!
Data written to a SUN Solaris NFS server by a Linux 2.2.1 NFS client can
be truncated.
I managed to find a reproducible test for this problem:
1.) Get jpegsrc-6b.tar.gz
2.) unpack it into an arbitrary directory on the SUN server
3.) issue "configure; make; make check" on the sources
What you will see is a) compilation errors of the jpeg utilities due to
a corrupted libjpeg.a, or b) segmentation faults at "make check". This
is only one example. It is not possible anymore to compile huge
source-packages on a Solaris NFS server. Usually generating libraries or
shared objects fails.
This problem only appears with the Linux 2.2.x kernel version. With
2.1.125 or 2.0.36 everything works fine. Neither it depends on the
system-release, or the compiler used for compiling the 2.2.x sources. I
tested it with glibc & libc5 systems, gcc 2.7.2.3 & egcs.
Linux & IRIX NFS servers work fine. I only have the problem when
accessing Solaris (2.5/.6) servers - unfortunately all the huge
fileservers in our institute.
Can anyone reproduce my problem ? ...
Oliver
--
/*
Oliver Stahlhut - Universitaet Hannover
Institut für Theoretische Nachrichtentechnik
und Informationsverarbeitung (TNT)
mailto:[EMAIL PROTECTED]
http://www.tnt.uni-hannover.de/~stahlhut/
*/
------------------------------
From: BL <[EMAIL PROTECTED]>
Subject: Re: 2.2.1: strange SMP (dual celeron) startup msgs
Date: 18 Feb 1999 16:16:24 GMT
Reply-To: no.spambots.please
Douglas Jerome <[EMAIL PROTECTED]> wrote:
: Hmmm... 2:30 seems mysteriously fast. I get 3:28.89elapsed for
: a "make -j2 bzImage" with dual 333 Mhz PII and an ultrawide
: 40 Mbytes/sec SCSI disk interface.
of course we'd have to use the exact same kernel config options. maybe I'm
not invoking stuff that you are?
I don't build ide cdrom or sound or linux-video. I do a lot of modules (have
to - kernel size is getting up there...). but I do include the mylex dac960
instead of standard scsi drivers. and other stuff.
I guess if you mailed me your .config I could better compare apples to apples.
--
AntiSpam: For email, change all 'zero' chars to letter 'o' chars.
------------------------------
From: PFENNIGER Daniel <[EMAIL PROTECTED]>
Subject: Re: Serious NFS bug in Linux 2.2.x (Solaris server)
Date: Thu, 18 Feb 1999 17:12:16 +0100
Oliver Stahlhut wrote:
>
> Hi there!
>
> Data written to a SUN Solaris NFS server by a Linux 2.2.1 NFS client can
> be truncated.
> I managed to find a reproducible test for this problem:
>
> 1.) Get jpegsrc-6b.tar.gz
> 2.) unpack it into an arbitrary directory on the SUN server
> 3.) issue "configure; make; make check" on the sources
>
> What you will see is a) compilation errors of the jpeg utilities due to
> a corrupted libjpeg.a, or b) segmentation faults at "make check". This
> is only one example. It is not possible anymore to compile huge
> source-packages on a Solaris NFS server. Usually generating libraries or
> shared objects fails.
>
> This problem only appears with the Linux 2.2.x kernel version. With
> 2.1.125 or 2.0.36 everything works fine. Neither it depends on the
> system-release, or the compiler used for compiling the 2.2.x sources. I
> tested it with glibc & libc5 systems, gcc 2.7.2.3 & egcs.
>
> Linux & IRIX NFS servers work fine. I only have the problem when
> accessing Solaris (2.5/.6) servers - unfortunately all the huge
> fileservers in our institute.
>
> Can anyone reproduce my problem ? ...
Yes, I had a probably related problem. When compiling (g77) with a
2.2.1 linux SMP PII on a disk controlled by a Sun Solaris sparc, through NFS
all the compilations failed with core generation. But when copying
the program on the Linux system and compiling it with the same compiler
worked fine.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Dr Daniel Pfenniger | [EMAIL PROTECTED]
Geneva Observatory, University of Geneva | tel: +41 (22) 755 2611
CH-1290 Sauverny, Switzerland | fax: +41 (22) 755 3983
__________________________________________________________________________
------------------------------
** 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
******************************
