Linux-Hardware Digest #208, Volume #9 Mon, 18 Jan 99 08:13:38 EST
Contents:
Re: Need suggestion on Minimum platform to run Linux (Michael Meissner)
Re: My partition choice (Michael Meissner)
Re: My partition choice (Michael Meissner)
Re: Want Linux bogomips numbers for Intel PII-450 (Terje Mathisen)
Re: My partition choice (Radovan Brako)
Re: Iomega DITTO MAX woes (Bruce Barnett)
Sound on Toshiba 4000 CDT (Philipp =?iso-8859-1?Q?R=FCfenacht?=)
Re: linux support for expanded memory? (Justin The Cynical)
Re: Modem trouble (Justin The Cynical)
Re: 3DLabs Permedia2 and Redhat5.2, X problems (#1) (Rick Moen)
Re: Linux only sees 16 megs of RAM! (Erwin de Beus)
disabling on-board video card? ([EMAIL PROTECTED])
----------------------------------------------------------------------------
From: Michael Meissner <[EMAIL PROTECTED]>
Subject: Re: Need suggestion on Minimum platform to run Linux
Date: 18 Jan 1999 01:24:41 -0500
Blaine Lupulack <[EMAIL PROTECTED]> writes:
> For MINIMUM, you're looking at a 386-16 with 4MB ram and 40-50MB disk
> (don't even bother
> thinking about using a GUI or any software developement )
Note, RedHat and presumably other distributions won't boot in less than 8 MB,
but you might be able to load a disk on a friends system, or possibly go with
Slackware.
--
Michael Meissner, Cygnus Solutions (Massachusetts office)
4th floor, 955 Massachusetts Avenue, Cambridge, MA 02139, USA
[EMAIL PROTECTED], 617-354-5416 (office), 617-354-7161 (fax)
------------------------------
From: Michael Meissner <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc,comp.unix.questions,comp.os.linux
Subject: Re: My partition choice
Date: 18 Jan 1999 01:14:03 -0500
DaZZa <[EMAIL PROTECTED]> writes:
> On 17 Jan 1999, Ilya wrote:
>
> > Thanks for clarifying - the machine will have 256MB of RAM eventually, and
> > 128MB when I get it. I guess I will have to have several swap partitions.
> > I am interested in a server-type machine. I don't know exactly what I will
> > be doing with it, but it is conceivable I might end up doing RAM-intensive
> > work.
>
> As a general rule of thumb, swap space should be approximately twice your
> physical memory.
True if you are using BSD derrived systems that require all physical memory
have reserved pages in the swap section, not true if you are using Linux. It
basically fairly simple, you need as much swap space as you will EVER need at
once minus the amount of physical memory (don't forget memory hungry compilers
and gimp apps -- I have a program that I cannot compile at full optimization
level and inlining everything without needing more than 512 meg of swap -- at
least that is how much swap space we had the last time we tried to compile it
two years ago on a Sun).
I'm a GCC developer, continually building the compiler and debugging it. On my
home machine I originally had 48, then 64 meg of memory and definately needed ~
60-80 meg of swap at times, now that I have 128 meg of memory, I don't need as
much swap, though I have 2 128 meg partitions on this machine, and 3 128 meg
partitions on my work machine, just in case.
> So, if you're having 256 Mb of RAM, then 512 Mb of swap is OK - however,
> as others have stated, you'll need to do it over several partitions - 127
> Mb is the largest swap partition size Linux will allow.
Actually this is old information. I believe the latest 2.2.0-prex kernels and
latest e2fs tools will now allow you to create swap partitions > 128 megabytes.
--
Michael Meissner, Cygnus Solutions (Massachusetts office)
4th floor, 955 Massachusetts Avenue, Cambridge, MA 02139, USA
[EMAIL PROTECTED], 617-354-5416 (office), 617-354-7161 (fax)
------------------------------
From: Michael Meissner <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.misc,comp.unix.questions,comp.os.linux
Subject: Re: My partition choice
Date: 18 Jan 1999 01:22:44 -0500
Ernst-Udo Wallenborn <[EMAIL PROTECTED]> writes:
> In my personal experience (SuSE 5.1->5.3 (currently) and 6.0 (planned))
> this was always the biggest problem after upgrades: personal setups,
> programs that ran within shell script wrappers, upgraded software,
> specialized libraries, and hacks like that. And based on this experience
> i'd say: Make a 128MB swap, put /home on a separate partition (~4GB),
> put /usr/local on a separate partition (~3GB), install every tar.gz
> emacs-update, python-beta or teTeX-990117 in /usr/local/ only, and
> give / all the rest. In general, imho, things that are likely to change
> simultaneously or not at all, or whose expiration date you determine on
> your own justify a partition of their own.
One technique that I use, particularly if you have the disk space, is alternate
root partitions the same size as the main root partition, moving user data out
to separate partitions, and a small /boot. That way you can play with a
different installation and/or upgrade in the alternate partition and then
switch lilo to use that as the root partition when you are satisfied, but until
then you don't wipe out your working system. It also helps when you blow away
your shared libraries and nothing runs (however many of the distributions now
support the concept of rescue disks for this case :-). The /boot partition is
to allow you to move your root partitions anywhere and not be limited by
needing to keep vmlinuz under 1024 sectors. I'm inching up to ~ 1 gig for root
+ /var + /usr partition size (currently 850 meg and 900 meg on my two
systems).
--
Michael Meissner, Cygnus Solutions (Massachusetts office)
4th floor, 955 Massachusetts Avenue, Cambridge, MA 02139, USA
[EMAIL PROTECTED], 617-354-5416 (office), 617-354-7161 (fax)
------------------------------
From: Terje Mathisen <[EMAIL PROTECTED]>
Crossposted-To: comp.sys.intel,comp.os.linux.advocacy,comp.arch,alt.os.linux
Subject: Re: Want Linux bogomips numbers for Intel PII-450
Date: Mon, 18 Jan 1999 11:55:15 +0100
Sander Vesik wrote:
>
> Bjorn Lindgren ?[EMAIL PROTECTED]? wrote:
> ? In comp.sys.intel Moshe Bar ?[EMAIL PROTECTED]? wrote:
> ? ? Hi
>
> ? ? My 400Mhz has a bogomips of 378.
>
> ? ? You may or may not know that Bogomips is NOT a measure of processor speed,
> ? ? it is just a calibration loop to measure how long it takes the processor to
> ? ? do NOTHING (no-op).
>
> ? ? So, careful with those Bogomips figures.
>
> ? What benchmark should i use to check that my PII-450 is a real PII-450 and
> ? not a remarked 350/400, i dont have access to other PII-450's so i need a
> ? open benchmark that there numbers for. any suggestions?
>
> There can be *NO* such benchmark. A PII-350 overclocked to 450 will get
> *EXACTLY* the same bogomips number. It will just probably not function
> 100% right every time and break down eventually.
There are actually some differences between the PII cpus, which are
software detectable, i.e. obvious things like the size/speed of the L2
cache, and the ability to use ECC on more/less of the memory hierarchy.
If the 450 is made on a later stepping, it should be possible to just
check the return values from the CPUID opcode.
...
I managed to find a PII-450 downstairs, and it returned exactly the same
stepping info as my PII-400, so this would not work. :-(
Terje
--
- <[EMAIL PROTECTED]>
Using self-discipline, see http://www.eiffel.com/discipline
"almost all programming can be viewed as an exercise in caching"
------------------------------
From: [EMAIL PROTECTED] (Radovan Brako)
Crossposted-To: comp.os.linux.misc,comp.unix.questions,comp.os.linux
Subject: Re: My partition choice
Date: 18 Jan 1999 12:39:48 +0100
In <77r94l$1o9$[EMAIL PROTECTED]> Ilya <[EMAIL PROTECTED]> writes:
>I'd like to know if this will work pretty well for a personal workstation
>and if I can improve this setup in any way. Redhat 5.2, 9.1GB hard drive.
>
>So far, I am thinking about this partition setup. I will assume that the
>real capacity is about 90% of the pre-formatted 9.1GB capacity.
>
>/swap 512 MB
>/tmp 350 MB
>/usr 4096 MB
>/var 1648 MB
>/home 1024 MB
>/ 512 MB
>/proc 48 MB
>
>I have pretty much made up my mind about /swap. What I'd like to know if
>1648MB is enough for /var and if I should decrease /usr to give /var more
>space or if this is OK. Also, is 48MB enough for /proc? 350MB for /tmp?
>
>Please post your replies, thank you.
Probably a lot of this has been already said in other replies,
but anyway:
1. Separate partitions are really necessary on multiuser or server
machines, where there is a serious risk that a vital service will
be starved of disk space while the machine is unattended. For
personal workstations, it is often best to leave everything as
a single partition, except, of course, swap.
2. I think that the most recent stable kernel (2.0.36) still has the
limit of 128 k swap partition. You can make two or more swap
partitions though.
3. /proc doesn't need disk space.
4. If you still decide to make several partitions, the /usr partition
you propose is huge, /var large, /tmp and /home smallish. Note
that user data for all applications will reside in /home, temporary
runtime data in /tmp or /usr/tmp.
RB
------------------------------
From: Bruce Barnett <[EMAIL PROTECTED]>
Crossposted-To: linux.dev.tape,linux.redhat.install
Subject: Re: Iomega DITTO MAX woes
Date: 18 Jan 1999 11:49:31 GMT
I'm not sure my last posting got out.Sorry for a duplicate.
>I am not sure, too. It looks like bpck driver can't find your Ditto.
>Here is a piece of my config:
>insmod ./bpck-fdc.o ft_fdc_threshold=16 ft_fdc_rate_limit=2000 \
>ft_fdc_base=0x378 ft_fdc_irq=7
How do I find the right value for ft_fcd_base?
I found the address of the parallel port, and of the floppy disk.
But the Iomega on the parallal port doesn't show up.
Is there some way I can test if I have the right address?
On other UNIX systems I used adb to probe the kernel memory.
Is there some set of values I can look for to make sure I have the
right values?
No matter what I try, I get
kernel: [006] bpck-fdc.c (bpck_fdc_query_proto) - Got 0xf8, expected either 0xc0
or 0x40.
------------------------------
Date: Mon, 18 Jan 1999 12:55:32 +0100
From: Philipp =?iso-8859-1?Q?R=FCfenacht?= <[EMAIL PROTECTED]>
Subject: Sound on Toshiba 4000 CDT
Has anyone succeded in configuring sound on a Toshiba Satellite 4000
CDT? I managed to
be able to play audio CDs (with the 'Toshiba Satellite Sound module' OSS
driver) - but
somehow couldn't find any good solution for other sound output (WAV,
MIDI etc.).
Thank you in advance,
kind regards,
Philipp
------------------------------
From: [EMAIL PROTECTED] (Justin The Cynical)
Subject: Re: linux support for expanded memory?
Date: 18 Jan 1999 12:01:12 GMT
Reply-To: [EMAIL PROTECTED]
On Sat, 16 Jan 1999 20:56:53 -0800, Paul Hovnanian <[EMAIL PROTECTED]> wrote:
->Dick Repasky wrote:
->>
->> Does linux support expanded memory? I can't find a definitive statement
->> one way or the other. The Hardware-HOWO says dram, sram, and edo, but
->> I do not know whether expanded memory qualifies as dram.
[snip]
->> If the memory isn't supported, with it foul up the kernel if the
->> board stays in the computer?
->
->When you say "the board", are you refering to system memory? That is,
<sigh> Kids today.. ;-)
[snip description of Extended Memory (XMS), IIRC. It's been awhile since I
thought about this stuff]
->On the other hand, there were some memory boards that plugged into ISA
->I/O slots (the ones you plug other interface cards into). These required
The memory boards that plugged into a slot are expanded memory (EMS).
That is what the original poster is aasking about.
AFAIK, Linux doesn't support them. It would be interesting if it did.
It would be a good was to recycle all those old memory boards and SIMMs.
Granted, it would be slower than XMS/motherboard based memory, but it would be
nice to have around.
--
"I just went visual on this goofy looking Finn riding on a gnu,
wielding one pissed off penguin... gah"
- Bob The Sane
Justin The Cynical - [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (Justin The Cynical)
Subject: Re: Modem trouble
Date: 18 Jan 1999 12:01:14 GMT
Reply-To: [EMAIL PROTECTED]
On Sat, 16 Jan 1999 20:52:13 -0500, Mircea <[EMAIL PROTECTED]> wrote:
->Quote from your first post:
->"When I check on windows, I find the modem on IRQ11, with automatic
->setting checked. Also looks like other things are on IRQ11"
->
->That's a hallmark for winmodems. Your vendor either doesn't know his
->business, or is plainly lying.
Not always. PCI devices can share IRQ's.
Now, as for the IRQ 11 bit, it sounds like the motherboard has the
onboard serial ports activated, and Winblows put the modem on the first
open IRQ. It's not uncommon to find Plug and Pray modems on COM 3, IRQ 10+.
My suggestion would to be check the BIOS settings, and disable the
COM ports. Then either try changing the settings for the modem in Windows, or
remove it and let Win redetect it. Once it's at the I/O and IRQ that all good
modems should be (which in my case is COM2, IRQ3) ;-), try to use it under
Linux.
--
"I just went visual on this goofy looking Finn riding on a gnu,
wielding one pissed off penguin... gah"
- Bob The Sane
Justin The Cynical - [EMAIL PROTECTED]
------------------------------
From: Rick Moen <[EMAIL PROTECTED]>
Crossposted-To: comp.os.linux.x,comp.os.linux.setup
Subject: Re: 3DLabs Permedia2 and Redhat5.2, X problems (#1)
Date: 18 Jan 1999 07:44:31 GMT
[Followups snipped. Please pick _one_ newsgroup.]
In comp.os.linux.setup Martin Schultz <[EMAIL PROTECTED]> wrote:
[You upgraded to XFree86 3.3.3 via Red Hat's RPMs for your
3DLabs Permedia2-based video card.]
: After I searched www.xfree86.org for Permedia 2, I realized that I would
: have to download version 3.3.3-1 (Redhat 5.2 ships 3.3.2), including
: a 3DLabs driver.
I hope you downloaded all of the required RPMs. There should ideally
be about eight of them.
By the way, what you observed at Red Hat's ftp site is not version
3.3.3-1, but rather 3.3.3, inside _RPM package build version #1_.
: As I understand, I now have to run xf86config to configure the X server.
Yes.
: This seems to be straightforward; I figured if I specify everything as
: plain VGA, chances ar best to arrive at some result.
Yes and no. This would be at best sub-optimal, and there are some
video chipsets that, surprisingly enough, don't support VGA16 well
or at all. I have no idea if this is true of the Permedia2, but
I would in any event have specified to xf86config the 3DLabs server
(on the "accelerated servers" list), and allowed XFree86 to use all
of the possible colour depths -- 8/16/24/32 bits per pixel, according
to http://www.xfree86.org/3.3.3.1/3DLabs2.html.
You didn't mention the capabilities of your monitor. When in doubt,
I choose a _very_ high range of horizontal and vertical scanning
frequencies, allow xf86config to finish and write its XF86Config
file, and then manually edit that file's
HorizSync 30-82
VertRefresh 50-90
lines in the Monitor section to match what my monitor can do. That
way, you have the maximum possible set of Modelines in the configuration
file, to work with.
(Whoops. After writing the above, I see that you _did_ sort of
specify what your monitor is.)
: Afterwards I looked at my XF86Config file (which doesn't tell me too
: much, I must admit), and I made sure that the path where it resides
: /usr/X11R6/lib/X11 is included in PATH (after /usr/X11R6/bin).
We should stop at this point, and discuss the above. First, XF86Config
is not an executable file, so there's no point in ensuring it's in
PATH (not for _any_ user -- each user account has its own environment
variables).
Second, if there's really a copy of XF86Config in /usr/X11R6/lib/X11,
then that's probably not the correct one -- unless that happens to
be a symbolic link to the correct location, /etc/X11.
When you look at an XF86Config file created by xf86config, realise
that it's horribly, horribly verbose. Your best chance of understanding
it is to first make a backup copy of it, e.g.,
cd /etc/X11
cp XF86Config XF86Config-INITIAL
...and then edit the working copy (/etc/X11/XF86Config) with a text
editor to eliminate almost all of the comment lines that begin
with a "#" -- excepting only the comment line that accompanies each
"Modeline". Also eliminate blank lines other than those separating
the sections. Also, delete all "Device" and "Screen" sections other
than the ones (one apiece) concerned with your particular X server.
Having done that and saved your work, you'll find that the file is
radically simpler to read. Also, be aware that it's best read from
the _bottom up_. The Screen section at the bottom ties everything
together, referencing the Monitor and Device sections about it and
specifying what resolutions to try and which colour depths.
You may, at this point, wish to eliminate from the Momitor section
the pairs of Modeline and comment for all resolutions you definitely
will not want to use -- e.g., the 600x400 mode, all the low-resolution
doublescan modes at the bottom of the section, etc.
If you happen to mess up, and write erroneous changes to disk, you
still don't lose much, because you can always revert to your backup
copy of XF86Config.
I see you mention the Ctrl-Alt-Del shutdown sequence. You should
also be aware of the much less drastic Ctrl-Alt-Backspace combo,
whose effect is to kill the X server process (only).
: My monitor is a Korea Data 9500, i.e. a 19" thing that can display
: 1280x1024 with 16M colors and 80Hz under Win98.
I cannot find anything on this specific monitor in Hotbot, Infoseek,
etc. Therefore, we have to guess. It should be safe to use the
horizontal and vertical limits I quoted above:
HorizSync 30-82
VertRefresh 50-90
Those are from a ViewSonic 17" monitor, and so should be more than
adequately conservative. The two lines quoted above are safety
limits: The X server will ensure that it tells the video card to
send a signal outside those frequency limits. Sending a monitor a
frequency outside its scanning limits for more than some number of
seconds or minutes can damage it physically.
If you're ever attempting to adjust XF86Config, start X to test your
work, and observe the screen not quite synching on the video signal,
kill it using Ctrl-Alt-Bkspc, and adjust XF86Config to eliminate that
video mode.
: QUESTIONS:
: * What could cause this behaviour?
See answer to next question. ;->
: * What do I have to enter into the XF86Config file in order to have
: linux understand my graphics card?
This is the key question. Howard Mann's http://www.xmission.com/~howardm/
pages will help, especially the "xwoes" section.
: * Are there any other files that I have to modify in order to be
: able to run X?
You must have two symbolic links set correctly.
/etc/X11/X should be a symbolic link to /usr/X11R6/bin/XF86_3DLabs
(since that's your X server binary), and
/usr/X11R6/bin/X
should be a symbolic link pointing the same place.
The XF86_3DLabs binary should have "rwxr-xr-x" (755) permissions, and
be owned by user root, group root.
You may find the "Installing..." portion of the XFree86 Release Notes
of some use: http://www.xfree86.org/3.3.3.1/RELNOTES11.html#11
Also: http://www.xfree86.org/3.3.3.1/QuickStart.html
Please note that the XFree86 documents above are not Red Hat-specific,
but rather are generic. They therefore tell you how to get the XFree86
Project's tar-gzipped archives, rather than Red Hat's RPM binary
packages. You presumably don't need to do the former, having already
done the latter.
: * a Netgear Fast Ethernet adapter FA310TX
This uses the Digital (aka "DEC") "Tulip" driver. Extract the
card from your case and examine the largest chip on it. If it says
"Digital" on it, it's a genuine, good-old Tulip chip. If not, it's
one of the Bay Networks imitation Tulip chips that NetGear (Bay
Networks's budget line) was forced to manufacture, when Digital's
chip production line was swallowed up by Intel.
If it's a genuine-Digital chip, you should have zero problems
with any Linux kernel's Tulip driver. If it's an imitation, then
it's claimed that the very recent Tulip drivers have been patched
to deal adequately with it -- though I'd personally take it back
and shove it down the vendor's throat. ;->
: * a iomega SCSI adapter (came with the new 2GB jaz drive)
You didn't specify which SCSI chipset this uses. Advansys?
Again, extract the card and read what it says on the largest
chip. Typically, the cheap SCSI cards included with things like
Jaz drives and scanners do not include a SCSI BIOS ROM, which
means that Linux kernel drivers, even when present on one's
system, cannot auto-detect the SCSI host adapter. For those,
typically, you must supply the driver with the necesary
hardware-resource information (IRQ, I/O base address) manually.
As with funky NetGear cards, the alternative (feasible on an
all-SCSI system, but perhaps not otherwise) is to cram the card
down the vendor's throat and get something better.
--
Cheers, The cynics among us might say: "We laugh,
Rick Moen monkeyboys -- Linux IS the mainstream UNIX now!
rick (at) linuxmafia.com MuaHaHaHa!" but that would be rude. -- Jim Dennis
------------------------------
From: Erwin de Beus <[EMAIL PROTECTED]>
Subject: Re: Linux only sees 16 megs of RAM!
Date: Mon, 18 Jan 1999 13:25:11 +0100
"Steve Z." wrote:
>
> I've installed Linux on a new Compaq Presario 5710 (I think that's the model
> number... I'm not sitting by it right now). When I run "free", it only
> reports 16 megs of RAM, when there is 128 megs in the system! Does anyone
> know why?
>
> I remembered an option in the kernel to limit memory usage to the lower 16
> megs, so I re-compiled the kernel with this option turned off, to no avail.
>
> FYI, it's the latest Slackware (downloaded yesterday) installed with UMSDOS
> onto my C: drive. This is a stock Presario, nothing has been added...
>
> TIA for any advice!
>
> Steve
Check your bios settings. I don't know about Compaqs, but some HP Vectra
machines have an option that says 'Limit memory to 16 Mb y/n' and
default to 'yes'.............
Erwin de Beus
------------------------------
From: [EMAIL PROTECTED]
Subject: disabling on-board video card?
Date: Mon, 18 Jan 1999 10:38:02 GMT
Hello,
I've installed RedHat 5.2. The system runs great, no problems so far exept for
this one I'm writing about. The cause of trouble isn't Linux, but rather my
peculiar video configuration.
The original video card is ATI Mach64 PCI with 1M. It is built-in on the
motherboard, and, according to manufacturer, there's no way to disable it in
the hardware. I have installed a better video card, STB Velocity 128 ZX with
8M VRAM. It works well under Windows this way(advised by the manufacturer): I
disable the ATI chip in the device manager, and then Windows detects Velocity
and installs its drivers and works well. Now I installed RedHat with X. It
runs well in text mode, and it even runs X, but the old card is getting on
its way. When X server is starting, it reports only 1M VRAM, and it also
mentions ATI chip. So I'm limited to 1024*768*8 due to lack of VRAM. If I try
a better resolution, X server understandably fails to start at all.
So, how can I tell Linux to ignore the ATI chip, similar to Windows'
disabling?
TIA
============= Posted via Deja News, The Discussion Network ============
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own
------------------------------
** 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.hardware) 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-Hardware Digest
******************************
