Files
2024-02-19 00:25:02 -05:00

582 lines
22 KiB
Plaintext

From: Digestifier <Linux-Activists-Request@news-digests.mit.edu>
To: Linux-Activists@news-digests.mit.edu
Reply-To: Linux-Activists@news-digests.mit.edu
Date: Thu, 26 Mar 92 23:15:09 EST
Subject: Linux-Activists Digest #164
Linux-Activists Digest #164, Volume #1 Thu, 26 Mar 92 23:15:09 EST
Contents:
Re: going faster than 38400... (Torbj|rn Lindgren)
Exclusive use of devices on Linux? (Joel M. Hoffman)
ps and the kernel (Dan Miner)
Re: Linux-Activists Digest #150 (Erik Fichtner)
Re: permissions (Drew Eckhardt)
Re: ESDI disk with 1629 cyl (Drew Eckhardt)
The Complete Includes (Steve Robbins)
Re: How to use shared lib with GCC2.0? (Drew Eckhardt)
Re: X386 - a unix platform xmindows (Drew Eckhardt)
Re: ps and the kernel (Dennis Flaherty)
Re: Questions - mtools, 0.95a Alpha Patches, Extended Partitions (Drew Eckhardt)
suid shell scripts? and mail programs (Brian Mcbee)
Re: SCSI problem w/seagate driver (Drew Eckhardt)
----------------------------------------------------------------------------
From: tl@etek.chalmers.se (Torbj|rn Lindgren)
Subject: Re: going faster than 38400...
Date: 26 Mar 92 15:31:54 GMT
In article <11752@ns-mx.uiowa.edu> williams@herky.cs.uiowa.edu (Kent Williams) writes:
>From article <64259@apple.Apple.COM>, by erc@Apple.COM (Ed Carp):
>> BTW, the formula for converting baud rate divisors is 115,200 / baud_rate.
>> Works both ways, too...
>
>You can't run an asynchronous port reliably with a divisor of 1. It has
>to do with how the chip samples the analog signal -- you need divide by 2
>at least. So 57600 is about the upper limit on PC hardware. And with kernel
>and interrupt latency, that's pushing it.
If you read National Semiconductors (or any elses) databooks this is
correct.. BUT it does work without problem at least between to 8250
compatible UART's. But to use it you need to have 16550A (or
compatible) with buffers so that the computer won't be overrun. I
don't know if Linux 0.95a uses the buffers, but it shouldn't be
impossible to add if it isn't there already.
------------------------------
From: joel@wam.umd.edu (Joel M. Hoffman)
Subject: Exclusive use of devices on Linux?
Date: 27 Mar 92 00:05:50 GMT
What happens on Linux if two processes both try to use the same device
at the same time? E.g., two processes, each tar'ing from the same
floppy? Does the driver take care of this? Is it a problem?
Likewise, what about serial ports?
-Joel
------------------------------
From: dminer@mcs213e.cs.umr.edu (Dan Miner)
Subject: ps and the kernel
Date: 27 Mar 92 01:29:24 GMT
Well, I'm trying to compile ps095 (on my roommate's computer).
We applied the patches that came in the ps095.tar. That went
ok. When we try to rebuild the kernel, gcc gets a fatal signal
and aborts. The file it vomits on is fork.c and the signal
number keeps changing (increasing). Can someone give us a
clue?
Thanks,
Dan
--
=========================================================================
Dan Miner email:
dminer@cs.umr.edu (prefered)
Computer Science S101908@UMRVMA.BITNET
------------------------------
From: techs@triton.unm.edu (Erik Fichtner)
Subject: Re: Linux-Activists Digest #150
Date: 27 Mar 92 01:41:42 GMT
In article <1992Mar26.152744@hammer.Prime.COM> cummings@hammer.Prime.COM (Kevin Cummings) writes:
>In article <1992Mar25.202614.7530@athena.mit.edu>, harvard!ames!amdcad!yarc!scott@EDDIE.MIT.EDU (Scott Beckstead) writes:
>> Well I see a trend concerning DTK motherboards. We at Yarc have also had
>> our shar of problems with DTK mother boards. I beleive this also concerned
>> running I/O at high speeds. You should be able to draw your own conclusions
>> at this point. DON'T RUN LINUX (or anything else for that matter) ON DTK.
>> We did do a little investigation into the problem (we thought it was our fault)
>> and discovered that the only motherboard in the world we had the problem with
>> was (low and behold) DTK.
>> .
>
>You mean I've been lucky running LINUX on my DTK-PEM2500 with NO MAJOR PROBLEMS,
>386-DX running at 25MHz (IO bus at 10MHz)???? It's news to me. Perhaps you
>can be a little more specific about your problems (CPU type, CPU speed,
>IO bus speed, whether or not problems go away when not in TURBO mode,
>whether or not you are using peripherals not designed for a 10MHz IO
>bus, ... etc, etc, etc.).
>
I too have a DTK-PEM-2530 80386DX 25MHz with a 12.5MHz bus speed. Linux
works fine in Turbo mode. The troubles I've been having with "HD Timeout"
are all caused by a very cheap IDE hard disk controller card. The system works
just fine with a newer, faster card.
(In case you're curious, the HD Controller is talking to two Seagate 1144A
drives and occasionally the system will "miss" an interrupt and things
go haywire.)
--
Techs techs@triton.unm.edu
Disclaimer: any inaccuracies are caused by the phase of the moon
"...whaddya mean I ain't kind... Just not *your* kind!" - Megadeth
------------------------------
From: drew@cs.colorado.edu (Drew Eckhardt)
Subject: Re: permissions
Date: 27 Mar 92 02:06:30 GMT
In article <1992Mar24.105719.5321@monu6.cc.monash.edu.au> pca306c@nella7.cc.monash.edu.au (E Elu) writes:
>After reading the various news items on the recommended directory layout
>and contents for Linux, I have gotten myself quite confused trying to work
>out what the correct permissions, owners and groups should be for the
>various directories and standard distribution files. It took me quite some
>time to get everything to a state where some other user than root could login
>and do anything !!
>
>Has any one got a document that suggests how the permissions of a Unix systems
>should be set for starters?
>
>The /etc /dev/ and /bin directories and files are the ones I am having the
>most trouble with.
>
MOST Directory permissions :
should be
755 - read / write / execute owner, read / execute everyone else
/tmp and /usr/tmp should be 1777 - sticky, read / write / execute all.
Sticky keeps people from deleting files they don't own.
Executables should probably be 555, read / execute all, unless they
are setuid. Many people like to make mount / umount setuid to
root so that "normal" people can run them. The best thing to
do there is to make the permissions on mount / umount
u+s 750 , owner root, group wheel, and put yourself in group wheel.
/dev
terminals, etc should probably be rw all - ie 666, hard disks 600,
floppies 666.
------------------------------
From: drew@cs.colorado.edu (Drew Eckhardt)
Subject: Re: ESDI disk with 1629 cyl
Date: Fri, 27 Mar 1992 02:24:47 GMT
In article <1992Mar24.223607.27070@afterlife.ncsc.mil> aegoldm@afterlife.ncsc.mil (Arthur Goldman) writes:
>Dilemma: I have a maxtor 600 MB drive with Ultrastor 12F esdi ctrlr.
>The controller offers me track mapping, which, when invoked, fools
>(DOS) into thinking it only has 1024 cyl (increases sectors/trk
>accordingly).
>
>BUT, Linux, like all other good Unix systems, won't run with the
>controller in mapped mode. Thus, I can only use 300+/- MB of
>my disk.
>
>Is there anything I/we can do to take advantage of the track
>mapping? I assume one would have to use the controller's
>BIOS in some way. right?/wrong?
This depends if the mapping is done at the BIOS or hardware level. My guess
would be at the BIOS level, since it doesn't work under Linux.
>I'm willing to do some work myself - give me a place to start looking.
You can hack hd.c to perform appropriate translations before issuing
the commands to the controller.
------------------------------
From: Steve@nyongwa.cam.org (Steve Robbins)
Subject: The Complete Includes
Date: 26 Mar 92 10:09:44 EST
Where are they?
I'm having problems compiling stuff with gcc. I started out with gcc-1.40,
stuck it in the right places and made the right links, but couldn't even
#include <stdio.h> in my programs.
I read that I needed the compiler sources to get the headers, so I ftp'd
'em, and gcc2.0 at the same time. I installed the newer compiler, which
came with a stdio header, but now compiles fail when the program wants to
include other stuff, particularly <dirent.h>. This file exists in gcc2.0's
private includes directory, but then dirent.h includes sys/dirent.h, which
I can't find anywhere, not even with the linux 0.95a sources.
Another program I was trying to build wanted to include termio.h, which
didn't exist, but termios.h did, so I changed the name, and it was happy I
think.
Is there a full set of standard include files somewhere, or at least a
list of names, so I know what I'm missing (and hopefully where to find them)?
P.S. If you post a reply to this, would you please mail a copy directly
to me as well? My news feed misses about 98% of the articles posted here.
--
Steve Robbins -- iconoclast at large
"If voting could change the system, it would be illegal."
steve@nyongwa.cam.org - or - uunet!nyongwa.cam.org!steve
------------------------------
From: drew@cs.colorado.edu (Drew Eckhardt)
Subject: Re: How to use shared lib with GCC2.0?
Date: 27 Mar 92 02:43:24 GMT
In article <1992Mar25.005149.183@athena.mit.edu> dhl@wet.net.netcom.com (Dahai Li) writes:
>Hello, every one,
>
>I know someone has asked this question before, and I read the replies. But
>I still make it work.
>
>Then I compiled a a simple "main(){puts("Hello, world!")}" program, by
>calling "gcc hello.c" command. (This gcc is from gcc2 package, not newgcc).
>The size of file "a.out" is 30,722.
>
>Then I copied the gcc lib files to the directory "/lib", and compiled the
>program again, (and removed the "a.out" file) and compiled the program
>again. Its size is the same as before, not "~5k" as mentioned in one of the
>replies to the previous help request to shared lib.
>
>I am terribly confused about people's comments on the contents in directory
>"/lib", one reply says:
> "shared library itself must be in /lib"
>another one says:
> ".. copy /usr/gcc2/shared/lib/lib92.03.15 (or whatever the date
> is)"
>and another one says:
>
> "/lib/what_this_version_of_the_shared_libraray_is_as_specified_in_crt0"
>
>Are shared libary (libaries) in different files than the static ones, I saw
>the "inst2.0" script linking "/usr/lib/shared/*" and "/usr/lib/static/*"
>to the same files?
>
This has to do with how shared libraries are currently implemented in
Linux.
The top 4MB of a process's address space is reserved for shared
libraries. Any references to this address space will result in the
shared library being demmand paged into this area, just like a
normal executable.
When you link with a library under /usr/lib/shared, the addresses in that
library point to areas in that top 4MB. The libraries here are
"stubs" so to speak, and have no actual code.
The /lib/lib92.03.16 or whatever file (the version of the shared
library is maintained in crt0.{s,o} is the actual code
which gets page faulted in.
Both the actual binary, and shread library stubs must be present
for shared libraries to work.
address space as a normal executable is.
------------------------------
From: drew@cs.colorado.edu (Drew Eckhardt)
Subject: Re: X386 - a unix platform xmindows
Date: 27 Mar 92 02:49:53 GMT
In article <4606@umriscc.isc.umr.edu> dminer@mcs213d.cs.umr.edu (Dan Miner) writes:
>
> I've heard about this xmindows (called X386?) that works on
>many of the other unix(s) out there. Has anyone heard of this too?
>I've been wanting something like xmindows, actually anything that
>will let be display graphics. (I don't know enough about any unix
>to figure out how to access and modify the IO prevl. map for the
>386)
X windows is an interaction between "client" processes (ie
xterm, xbiff, etc) and a server. X386 is the xserver for 386 unixes.
Also, you can do graphics if you have used obz's mmap(2) patches,
by mapping the frame buffer into user memory. Bank switching, etc
can be done through writing to /dev/port.
> I've found just about very thing on linux that my DOS
>can do except graphics... (DOS has better interfaces for word
>processing about I can live without it for now... :)
>
>PS. I think it is written entirely in assembly.
>
hardly. A VERY small portion of the device dependant code
(/src/mit/server/ddx/x386/drivers from the stock X11R5 distribution)
is written in assembler, and then only the bank switching code.
------------------------------
From: dflahert@davinci.helios.nd.edu (Dennis Flaherty)
Subject: Re: ps and the kernel
Date: 27 Mar 92 03:14:19 GMT
In article <4642@umriscc.isc.umr.edu> dminer@mcs213e.cs.umr.edu (Dan Miner) writes:
>
> Well, I'm trying to compile ps095 (on my roommate's computer).
> We applied the patches that came in the ps095.tar. That went
> ok. When we try to rebuild the kernel, gcc gets a fatal signal
> and aborts. The file it vomits on is fork.c and the signal
> number keeps changing (increasing). Can someone give us a
> clue?
>
> Thanks,
> Dan
You can't compile it with -finline-functions, for some reason. Or
at least it compiled when I took out that option. In fact there are
quite a few files in the kernel for which I had to switch compiler
options. Also, there's a lingering d_entry in swap.c that should
be removed, I think.
And I'm not even sure I did it all right, because ps doesn't work
for me-- it only displays the header line, and sometimes it complains
about /dev/kmem.
--
Dennis T. Flaherty dflahert@chekov.helios.nd.edu
U of Notre Dame Dept. of Electrical Engineering
Notre Dame *tells* me what opinions to have.
------------------------------
From: drew@cs.colorado.edu (Drew Eckhardt)
Subject: Re: Questions - mtools, 0.95a Alpha Patches, Extended Partitions
Date: 27 Mar 92 03:01:52 GMT
In article <1992Mar25.160630.27178@athena.mit.edu> apr%dec001.ziap.wtza-berlin.de@noc.belwue.de (Andreas Priebe) writes:
>
>Hi Linuxers,
>
>again some questions/remarks:
>
>
>1.
>MTOOLS:
>I made in devices.c several entries for A and B (to read 360/720/1.2 in 1.2M
>and 720/1.44 in 1.44MB drive).
>A 1.2MB disk in a 1.2MD drive and 1.44 in 1.44 work very fast and without
>any error messages.
>Reading a 720k in an 1.44MB drive gives:
>
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>floppy I/O error
>dev 021c, block 9
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>Reset floppy called
>floppy I/O error
>dev 021c, block 9
>Reset floppy called
>
>After that the normal mtools header (anyway I don't like that header,
>to much information for a simple copy, but he tells me that 9 sectors
>and all the other things are correctly recognized) appears and
>the data (ie dir or copy) are correct.
>Similiar results are obtained with 360k/720k in an 1.2MB drive, although the
>number of "Reset floppy called" may differ.
>The data don't seem to be corrupted.
>
>Is this the normal behaviour (I had this with 0.12 already!) - can't
>I use several formats on one drive type?
YES. This is normal behavior with Mtools. It tries every entry for
A: or whatever in the devices.c? file (in the source), and when
it gets an error it tries the next one. Set the devices up so
that the first entry for A: or whatever is the most commonly
used one.
>
>2.
>Maybe I'm wrong, but should'nt compile at least the kernel _without_ any
>warnings?
Those are WARNINGS. Warnings mean "Hey, you may have done something stupid,
but syntactically, it's correct. Just thought I'd let you know". gcc
includes the functionality of lint with its warnings.
>I applied the Alpha patches from Linus to 0.95a and get several (not too much,
>but some warnings) (using gcc1.40) - ok its Alpha only, I know,
>but I think its a priciple question.
>BTW: It would have been nice, if somebody had mentioned to use patch
>with "-l" (or is it assumed, that Alpha testers guess it themselves?).
>With me the patches work, but sometimes with very hard disk traffic
>(swapping during 2 parallel compilations and several background
>"ls -lR / &") I get HD timeout which don't seem to corrupt the results.
>
I used patch -p0 < patches from the appropriate directory and it worked
fine.
As far as having parallel compilations, etc you probably shouldn't do it.
With one build, you may swap a little on a typical 4M system,
, but mostly you'll still be in
real memory. With two, that won't be the case with limited
amounts of ram. The rule of thumb is that disk is 1000 times slower than
real memory.
>
>4.
>Shouldn't create Linux a core-file after dieing a process with
>segmentation fault ?
No. The die() routine prints out some information, but no
core dump is performed. Some one might want to
implement core dumps now that we have a debugger.
>5.
>A shutdown script would be very nice. Anybody working on it or another
>way to go to single user mode and back to multiuser?
You want to fork() then exec() not just exec when running the
init. That way kill -9 1 will take you into single user mode,
as with any other "Real Unix"
------------------------------
From: mcbeeb@nyssa.CS.ORST.EDU (Brian Mcbee)
Subject: suid shell scripts? and mail programs
Date: Fri, 27 Mar 1992 03:19:38 GMT
I've been messing around trying to get mail up on my machine. smail
compiles fairly cleanly, and same for elm after I got all the programs
necessary to run the configure script. I have a problem with local
mail delivery, though. Anyone have a suggestion as to shere to get lmail?
I tried writing lmail as a sh script. To make it work,wit has to run suid.
However, even suid, I can't chown the mail file after appending a message
to it. Is this a bug, or am I doing something wrong? My lmail is just:
#!/bin/sh
cat - >>/usr/spool/mail/$1
chown $1 /usr/spool/mail/$1
Any help appreciated.
--
Brian McBee mcbeeb@jacobs.cs.orst.edu or brianop@opac.UUCP
"The strongest reason for the people to retain the right to keep and bear
arms is, as a last resort, to protect themselves against tyranny in
government." Thomas Jefferson
------------------------------
From: drew@cs.colorado.edu (Drew Eckhardt)
Subject: Re: SCSI problem w/seagate driver
Date: 27 Mar 92 03:28:55 GMT
In article <1992Mar26.015549.26806@src.umd.edu> chad@src.umd.edu (R Michael McMahon) writes:
>The seagate driver is not working on my system (386/7, Phoniex BIOS,
>C&T chipset). I can boot with 0.95a without SCSI support, and have
>no problems except the lack of a harddisk. Using the latest 0.95a
>SCSI bootimage (ftp'ed on 3/23), here are the diagnostics:
>
>Host 0 is detected as a(n) Seagate ST-01/ST-02.
>Seagate ST-0X SCSI dirver by Drew Ekhardt
>$Header .....blah.... $
>1 Host adapters detected
>Detected SCSI disk at host 0, ID 0, lun 0
>Detected SCSI disk at host 0, ID 1, lun 0
>Detected 2 disks, 0 tapes total
>Kernel panic: scsi_done() called with invalid host number
>
>(Note: I have a Quantum 105s and a Seagate 296N installed. The
> panic occurs with either or both drives attached.)
>
>I have not moved any jumpers on the ST-02, meaning on-board BIOS is
>at the default location, etc.
>
>Any help for this would be appreciated. In particular, is there a way
>for me to recompile (or a place to ftp) a SCSI bootimage with full
>debug options? I don't have a non-scsi HD in order to do this on
>my system very easily.
Every now and then, a few people have had problems with the bootable
images for SCSI. I think 3 people have had an identical problem with
the .95a images. The only "common" thing I've seen, and possible
difference from my system is that no one is using shoelace.
You might try the bootimage-0.12.seagate.3 Linux .12 bootimage,
and make a root file system with shoelace and the kernel on it.
As far as a debugging version of the kernel : I will upload
a version compiled with debug flags to headrest.colorado.edu
RSN.
>I have access to the Adaptec driver written by James da Silva of U.M.
>(he works upstairs!) for Minix. As soon as my Seagate is working, I
>will work on porting James' driver to my Adaptec. Any others working
>on this task, please send e-mail so we can collaborate.
>
Tommy Thorn tthorn@daimi.aau.dk is currently working on this,
and in a post of his I saw today, it was promised "RSN"
It looks like Wierdness has struck with Linux .95 - the larger
kernel (compared to .12) + larger SCSI patches (seagate + ultrastor)
move an area being stomped on to something noticeable. If you used
every patch available with .12 together, you also noticed
this problem.
In some cases, Linux (the stock kernel) stomps on Kernel memory,
and depending on who you are, something not noticeable / file_table[] / stack
, etc. gets corrupted. In my case, it was file_table[] that
was being stepped on, with SCSI drivers in place and "active",
or with the initialization routine replaced by a return to eliminate
any chance the SCSI drivers had for messing things up.
The distribution SCSI kernel, or the kludge.diff moved this and made
it go away.
Also, if you are booting off of floppy without shoelace, BSS will be
uninitialized, and is not guranteed to be all 0's. This
may or may not effect Linux. I don't know what happens to BSS
if you are using shoelace.
------------------------------
** 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: Linux-Activists-Request@NEWS-DIGESTS.MIT.EDU
You can send mail to the entire list (and alt.os.linux) via:
Internet: Linux-Activists@NEWS-DIGESTS.MIT.EDU
Linux may be obtained via one of these FTP sites:
nic.funet.fi pub/OS/Linux
tsx-11.mit.edu pub/linux
tupac-amaru.informatik.rwth-aachen.de pub/msdos/replace
The current version of Linux is 0.95a released on March 17, 1992
End of Linux-Activists Digest
******************************