Linux-Hardware Digest #335, Volume #10 Wed, 26 May 99 21:13:34 EDT
Contents:
Keyboard problems with XF_86FBDev frame buffer server. (Ross)
Re: IBM 22 GByte hard disk ("marco viola")
Re: Solved SCSI problem (aha152x) ("Curt")
BUSLOGIC BT542B rev H ("Curt")
Re: Lucent WinModem-Will it work? (R G ELLIOTT)
Re: AMD 5x86/133Mhz (Warren B. Hapke)
Any way to print out IRQ's being used in linux? ([EMAIL PROTECTED])
CRC errors problems =( (Jonas Palsson)
Re: Help a hard drive problem ([EMAIL PROTECTED])
Re: removing cooling fans--how dangerous? (David C.)
Re: IBM 22 GByte hard disk ("marco viola")
Re: IBM 22 GByte hard disk ("marco viola")
Re: Adaptec 2940UW PCI problem (Chris Mauritz)
How FAST am I going? (waddle waddle) (Walt Shekrota)
Re: Terabite Plus Filesystems ("Ross Vandegrift")
Problem with Linux and Compaq deskpro (Rachid Himmi)
AMD 5x86/133Mhz (John Hong)
Attempted RedHat install kills HDD ("Robert C McDougle")
Run AMD k2 366 in 66*5.5 or 100*3.5 (Y Chen)
Re: IDE faster than SCSI UW? (David C.)
Re: AMD 5x86/133Mhz (John Hong)
Re: IDE faster than SCSI UW? (David C.)
Re: Backup solution for a single linux box with about 10Gig drive. ("Gen. Sisyphus")
----------------------------------------------------------------------------
From: Ross <[EMAIL PROTECTED]>
Subject: Keyboard problems with XF_86FBDev frame buffer server.
Date: Wed, 26 May 1999 23:17:01 +0000
I've had to install the XF86_FBDev Frame Buffer server (to support my
new ATI Rage Pro 32MB) and pasted my old bits over for keyboard section.
Now my up / down /left /right / delete / etc. keys are either non
functioning or throwing wobblies.
Here is /etc/XF86Config and keyboard section is unchanged from
configuration that worked OK when set up with SVGA server.
Section "Files"
RgbPath "/usr/X11R6/lib/X11/rgb"
FontPath "/usr/X11R6/lib/X11/fonts/misc:unscaled"
FontPath "/usr/X11R6/lib/X11/fonts/75dpi:unscaled"
FontPath "/usr/X11R6/lib/X11/fonts/100dpi:unscaled"
FontPath "/usr/X11R6/lib/X11/fonts/Type1"
FontPath "/usr/X11R6/lib/X11/fonts/Speedo"
FontPath "/usr/X11R6/lib/X11/fonts/misc"
FontPath "/usr/X11R6/lib/X11/fonts/75dpi"
EndSection
Section "ServerFlags"
AllowMouseOpenFail
EndSection
Section "Module"
EndSection
Section "XInput"
EndSection
Section "Keyboard"
Protocol "Standard"
XkbRules "xfree86"
XkbModel "pc104"
XkbLayout "gb"
EndSection
Section "Pointer"
Protocol "PS/2"
Device "/dev/psaux"
SampleRate 60
Emulate3Buttons
Emulate3Timeout 65
BaudRate 1200
EndSection
Section "Monitor"
Identifier "Primary Monitor"
VendorName "Unknown"
ModelName "Unknown"
HorizSync 31-69
VertRefresh 50-85
Modeline "1024x768" 65.00 1024 1048 1192 1340 768 771 777 806
-hsync -vsync
EndSection
Section "Device"
Identifier "Primary Card"
EndSection
Section "Screen"
Driver "FBDev"
Device "Primary Card"
Monitor "Primary Monitor"
SubSection "Display"
Modes "default"
EndSubSection
EndSection
Any ideas what might be different or causing the prob. Can't find a set
of values on the internet / help files etc. that could be of assistance.
Gratefully
--
From: [EMAIL PROTECTED]
http://www.ross75.freeserve.co.uk
Using SuSE Linux 6.0 / NS Mozilla
------------------------------
From: "marco viola" <[EMAIL PROTECTED]>
Subject: Re: IBM 22 GByte hard disk
Date: Wed, 26 May 1999 10:08:11 +0200
Nick Birkett ha scritto nel messaggio
<[EMAIL PROTECTED]>...
>We recently bought a 22Gbyte hard disk for use under Linux.
>I can partition this and format it fine. It is used as scartch storage
>hence only one huge parition (we have some enourmous files).
>
>The disk : IBM 22Gbyte, 7200 rpm, 2Mbyte cache UDMA.
if chipset ali aladdin, http://www.dyer.vanderbilt.edu/
ciao
marco
------------------------------
Reply-To: "Curt" <[EMAIL PROTECTED]>
From: "Curt" <[EMAIL PROTECTED]>
Subject: Re: Solved SCSI problem (aha152x)
Date: Wed, 26 May 1999 17:16:57 -0500
IRQ 9 is not usually available, because it is used to redirect interrupts
from the 2nd
interrupt controller to the first via IRQ2. I have seen in some cases where
the driver is
smart that it is possible to use IRQ 9 and or 2, but not very often.
Kjell Petersen <[EMAIL PROTECTED]> wrote in message
news:7ibksj$pu5$[EMAIL PROTECTED]...
> Earlier on I posted a message describing my problems with using the
aha152x
> support together with an AVA1502AE Adaptec scsi card to control a Mustek
> scanner.
>
> I received several hints, thanks a lot to everyone who answered my
posting.
> None of the hints were head on, but together the hints drove me in the
right
> direction towards a solution ;^)
>
> I've seen several people post regarding similar problems to mine with the
> aha152x, so I'll try to make a summary of the hints I got.
>
>
> But first, the symptoms :
>
> When compiling the 152x support into the kernel, and giving parameters at
> boot time :
>
> aha152x=0x140,9,7 ( I/O, IRQ, SCSI_ID *for controller card, not
device*)
> or
> append="aha152x=0x140,9,7" in lilo.conf
>
> the following messages appeared during bootup :
>
> aha152x: processing commandline: ok
> aha152x: BIOS test: passed, detected 1 controller(s)
> aha152x0: vital data: PORTBASE=0x140, IRQ=9, SCSI ID=7, reconnect=enabled,
parity=enabled, synchronous=disabled, delay=500, extended
translation=disabled
> aha152x: trying software interrupt, lost.
> aha152x: IRQ 9 possibly wrong. Please verify.
> scsi : 0 hosts.
> scsi : detected total.
>
>
> Trying modules instead yielded:
>
> >insmod scsi_mod # needed by aha152x
> >insmod aha152x aha152x=0x140,9,7
> aha152x.o: Device or resource busy
> >
>
>
> According to my /proc/interrupts both irq 11 and 12 were taken, so I tried
the
> two last options for my card, irq 9 and 10. Both gave the same results as
> above.
>
> One of the first hints I got, was to also check the /proc/pci, and
> there I discovered that irq 10 was taken by the USB controller. But it
didn't
> mention anything about the irq 9, so I still had that one as an option, or
> at least that's what I thought. I did get the following warning in one
mail,
> but it didn't seem to be relevant for my system :
>
> "you should re-re-check with cat /proc/pci | grep IRQ, I have never see a
> PCI system not using the IRQ 9 for it's own or for some PCI card until now
!!!"
>
> In several postings I read something about a possibility to reserve an IRQ
> number in the bios. In my bios I found a table where I could chose which
IRQ
> numbers and DMA channels that was possible for the PnP system to use. Both
IRQ
> 9 and 10 were in this pool, so I disabled then for PnP use, but this
didn't
> solve any problems. The USB controller still used IRQ 10, and the aha152x
would
> not work on IRQ 9.
>
> Getting tired of failures, I din't want to mess too much my BIOS. But I
did
> register some options regarding the USB port.
>
> Then I found the following in a posting :
>
> "These cards are a bit of a pain. If your willing to build a kernel with
the
> adaptec support built in you can go in and edit the Makefile in
> /usr/src/linux/drivers/scsi and change the line that pertains to the
> aha152x (it will be obvious). Remove all the stuff that says -DAUTOCONF
(or
> whatever it says) and replace it with -DSKIP_BIOSTEST
> -DSETUP0="{0x140,11,7,1}"."
>
> At this point I did two things at the same go. Remembering the options
about
> the USB port in the BIOS, and the fact that I don't have any USB devices
yet
> (and the Linux support isn't around yet(?)), I disabled the use of the USB
> port and thereby freeing the IRQ 10. Changing the Makefile as described,
but
> with my parameters, compiling and booting, and it worked !!
>
> Then I thought, maybe only the IRQ was the problem ? So I tried with old
> Makefile again, compiling and booting, still working ! But I will never
(?)
> know if the card did need that initial direct assignment of parameters
> (anyone ?)
>
> I hope this can be of any help to anyone ;^)
>
>
> My guess is that my system somehow uses IRQ 9, but is not listing it in
neither
> /proc/interrupts or /proc/pci. Anyplace else one should look ?
>
>
> Kjell ;^)
>
------------------------------
Reply-To: "Curt" <[EMAIL PROTECTED]>
From: "Curt" <[EMAIL PROTECTED]>
Subject: BUSLOGIC BT542B rev H
Date: Wed, 26 May 1999 17:22:25 -0500
I recently saw some posts regarding this board, but never saw that the
problem was resolved.
The board it pretty old, but I know it works, I use it daily under another
PC UNIX OS (QNX).
I wanted to install RH5.2 linux on this system as well. I need the SCSI
interface to access my CDROM.
The board and CDROM are recognized, but when the install process tries to
install the driver it
constantly timeout and resets. Any suggestions?
------------------------------
From: R G ELLIOTT <[EMAIL PROTECTED]>
Subject: Re: Lucent WinModem-Will it work?
Date: Wed, 26 May 1999 18:43:03 +0100
In article <7ih4sr$u17$[EMAIL PROTECTED]>, [EMAIL PROTECTED] writes
>I have a Lucent WinModem. It currently works under Win98, but I am
>beginning to experiment with Linux (RH 5.2). Will this modem work? I
>suspect the answer is no. Can someone recommend a modem that will
>perform well under both Win98 (not ready to make the full leap just yet)
>and Red Hat?
>
>
>Bill
I too have a Lucent and it drove me nuts for a week until I found out it
was a Winmodem and does not work in Linux.
--
Bob
------------------------------
From: [EMAIL PROTECTED] (Warren B. Hapke)
Subject: Re: AMD 5x86/133Mhz
Date: Wed, 26 May 1999 22:53:50 GMT
John Hong ([EMAIL PROTECTED]) wrote:
: Anyone running with one of these? Anyhow, I o/c this to 160Mhz
: (4x40Mhz) and everything was going just fine until I tried recompiling
: the kernel (ended up with Sig11). Going back to 133Mhz (4x33Mhz)
: everything recompiled just fine. I was wondering had anyone tried o/c to
: 150Mhz (3x50Mhz) with this. Did it recompile fine for you? Or how about
: 120Mhz (3x40Mhz)?
I have one of these, but I've never run it at anything except the
standard 133 MHz configuration. It's been rock-solid at that level.
(PS. Sig 11 errors normally indicate a memory problem. What is your
memory speed; 70ns or 60ns? Maybe the overclocking will work w/
faster memory?)
The potential gains from the kinds of overclocking you're talking about
(150 vs 133) don't seem worth it.
Warren B. Hapke
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED]
Subject: Any way to print out IRQ's being used in linux?
Date: Wed, 26 May 1999 22:19:29 GMT
Any way to print out IRQ's being used in linux?
I have a linux system running and I want to
add a card, but I don't know what interrupt requests
are being used.
How do I know what to set the IRQ of the card being installed?
Is there a faq on this somewhere?
--== Sent via Deja.com http://www.deja.com/ ==--
---Share what you know. Learn what you don't.---
------------------------------
From: Jonas Palsson <[EMAIL PROTECTED]>
Subject: CRC errors problems =(
Date: Wed, 26 May 1999 21:41:30 +0200
I get LOTS of CRC errors when i download files (ethernet connection). I
know that the NIC not is broken so my last hope is that anyone here
knows what the problem could be. My system:
Harddisk: IBM Deskstar 16GP (udma33 disk)
CPU : AMD K6-2 350
MB: DFI P5BV3+ with the VIA MVP3 chipset
RAM: 64 MB PC100 SDRAM
NIC: 3com905B
Hope anyone can solve my problem
Jonas P
[EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED]
Subject: Re: Help a hard drive problem
Date: Wed, 26 May 1999 19:55:35 GMT
Just out of curiosity do you use chinese characters for your files?
------------------------------
From: [EMAIL PROTECTED] (David C.)
Crossposted-To:
alt.comp.hardware,alt.comp.hardware.pc-homebuilt,alt.comp.hardware.overclocking
Subject: Re: removing cooling fans--how dangerous?
Date: 26 May 1999 19:09:07 -0400
[EMAIL PROTECTED] writes:
>
> Brilliant! Several people report good success in running their
> computers without cpu fans, so it seems to work. My questions:
>
> - If I have a large (huge) heat sink, do I also need a Peltier
> underneath (and then a thermostat for that, too)?
I think it would depend on your case and other environmental factors.
The Dell and Micron systems I use do not use CPU fans. Instead, they
have large heatsinks. The power-supply fan blows air over the
heat-sink, and a separate case-fan draws the warm air out again. This
definitely works - the air blowing out the back of the case is always
cool to the touch, and these computers are left on for months at a
time. (These are Pentium-II systems running at 300, 350, 400 and
450MHz)
Assuming your case's power-supply fan is blowing air over the CPU, a
large aluminum heat-sink will probably work fine by itself.
-- David
------------------------------
From: "marco viola" <[EMAIL PROTECTED]>
Subject: Re: IBM 22 GByte hard disk
Date: Wed, 26 May 1999 21:58:41 +0200
Nick Birkett ha scritto nel messaggio
<[EMAIL PROTECTED]>...
>We recently bought a 22Gbyte hard disk for use under Linux.
>I can partition this and format it fine. It is used as scartch storage
>hence only one huge parition (we have some enourmous files).
>
>The disk : IBM 22Gbyte, 7200 rpm, 2Mbyte cache UDMA.
if the chipset is an ali aladdin, look at
http://www.dyer.vanderbilt.edu/
ciao
marco
------------------------------
From: "marco viola" <[EMAIL PROTECTED]>
Subject: Re: IBM 22 GByte hard disk
Date: Wed, 26 May 1999 22:03:52 +0200
marco viola ha scritto nel messaggio
<7ihj9e$j36$[EMAIL PROTECTED]>...
>
sorry
ciao
marco
------------------------------
From: Chris Mauritz <[EMAIL PROTECTED]>
Subject: Re: Adaptec 2940UW PCI problem
Date: Wed, 26 May 1999 22:45:47 GMT
Is there a place to get the newest driver? I'd love to get the
aic7890 in my Asus P2B-S motherboard working. It works fine for
hard disks, but fails with DAT tapes. I simply get IO errors
when I try to read/write tapes, though I can do things like
rewind and eject them.
C
E.W. Veltman <[EMAIL PROTECTED]> wrote:
> Hello Nigel,
> I received a reply from SuSE support with a URL for updated SuSE6.1 boot
> disks.
> The boot disks on ftp://ftp.suse.com/pub/suse/i386/current/disks/ should
> contain 2.2.7.
> I did not yet try myself and probably won't have time for it until this
> weekend.
> If you try this disk, could you please let me know the result ?
> Then I can inform SuSE support whether it solved the problems.
> Thanks.
>>I do indeed have a CD drive and a tape drive on the SCSI bus.
>>
>>The version of the aic7xx driver seems to have changed between SuSE 6.0 and
>>6.1 (v. 5.1.12 -> v. 5.1.15), so I wonder if something went wrong there. I
>>haven't had a reply from SuSE yet.
>>
>>I'll be very interested in any further news you might have!
--
Christopher Mauritz
[EMAIL PROTECTED]
------------------------------
From: Walt Shekrota <[EMAIL PROTECTED]>
Subject: How FAST am I going? (waddle waddle)
Date: Wed, 26 May 1999 16:16:35 -0400
Anyone know of a neat widget to display or log serial line speed
negotiated? The log from PPP seems to just reflect what chat allows you
to request not necessarily the true line speed negotiated. Can't find
anything on this.
example
I told chat to connect at 57600 ..... thats whats in the log. I suspect
speed won't go higher than 52k in my area. I had a rock solid connection
of about 52 reported before I moved this modem from NT to my Linux
gateway box.
Any help or tips appreciated.
Please remove 'nul' from my id in emails.
-Walt
Also might be neat to have a modem lights widget. Have used these on
other os's when internal card used. You just can't see activity!
------------------------------
From: "Ross Vandegrift" <[EMAIL PROTECTED]>
Crossposted-To:
comp.os.linux.development.system,comp.os.ms-windows.nt.admin.misc,comp.sys.sun.admin,comp.sys.hp.misc
Subject: Re: Terabite Plus Filesystems
Date: Wed, 26 May 1999 16:27:35 -0400
> We would
>want to pick either HPUX, linux, NT or Solaris x86. Any experience
>that could be passed would be great.
No experiance here, but I do know that SGI has announced to have the source
to XFS realeased sometime soon this summer. XFS is a 64-bit journaled
filesystem. As far as specific system advice, I can't help.
Ross
------------------------------
From: Rachid Himmi <[EMAIL PROTECTED]>
Subject: Problem with Linux and Compaq deskpro
Date: Wed, 26 May 1999 18:16:56 +0100
Can anybody help me on problems I met on Compaq deskpro computer.
I installed the Redhat 5.2 software and I got the following problems:
- the keyboard is not systematically working (sometimes after
rebooting the keyboard is not operational)
- after a long time, the system automatically go down and I am not
able to re-start it
(but the system is running because I can continue to work by
telnet)
- sometimes it is not possible to restart the computer because the
power switch does not work
- if I want to connect to the system by network (telnet or ftp) it
takes a long time before having a prompt
(about 1 minutes)
pleaser answer me by mail
thak you in advance for any help or suggestion to solve this blocking
problems
--
__________________________________________________
Rachid HIMMI
Alcatel Business Systems -France
1 Route Dr A. Schweitzer
67408 Illkirch Cedex FRANCE
tel: (+33) (0) 3 88 67 71 43
fax: (+33) (0) 3 88 67 71 33
e-mail: [EMAIL PROTECTED]
------------------------------
From: [EMAIL PROTECTED] (John Hong)
Subject: AMD 5x86/133Mhz
Date: 26 May 1999 20:23:42 GMT
Anyone running with one of these? Anyhow, I o/c this to 160Mhz
(4x40Mhz) and everything was going just fine until I tried recompiling
the kernel (ended up with Sig11). Going back to 133Mhz (4x33Mhz)
everything recompiled just fine. I was wondering had anyone tried o/c to
150Mhz (3x50Mhz) with this. Did it recompile fine for you? Or how about
120Mhz (3x40Mhz)?
------------------------------
From: "Robert C McDougle" <[EMAIL PROTECTED]>
Subject: Attempted RedHat install kills HDD
Date: Wed, 26 May 1999 20:14:24 -0400
I just tried to install RedHat 5.2 to my Tyan 1563D with 1 p133 processor
and a WD 2.5g HDD. The drive was recognized fine by the BIOS prior to the
install and could be hit by FDISK on a win98 boot disk. After attempting
the Linux install (which seemed to go fine), the boot fails with the message
"not found any [active partition] in HDD
DISK BOOT FAILURE, INSERT SYSTEM DISK AND PRESS ENTER"
pressing enter just gets me to a repeat of the message.
Now that this message has occurred, I can't even boot from a floppy. I
changed the boot sequence in the BIOS to A, C, SCSI but even with a
previously working boot disk only get to the same message.
Does anyone know what is going on???
--
R C McDougle
rcm Arch Design
------------------------------
From: [EMAIL PROTECTED] (Y Chen)
Subject: Run AMD k2 366 in 66*5.5 or 100*3.5
Date: Wed, 26 May 1999 16:30:34 -0400
Hello, everyone.
I have a AMD K6II 366 CPU, currently I set the CPU
bus speed to 100MHz and multiplied by 3.5.
My system board also support 66 MHz bus speed and
multiplied by 5.5.
I do not know which one is better for me?
Any comment?
Thank you ahead!
------------------------------
From: [EMAIL PROTECTED] (David C.)
Subject: Re: IDE faster than SCSI UW?
Date: 26 May 1999 19:59:54 -0400
[EMAIL PROTECTED] writes:
> Stuart R. Fuller <[EMAIL PROTECTED]> wrote:
>>
>> The SCSI driver will query each SCSI device to get various
>> information about them. One piece of information is some negotiation
>> of how fast the device can run. The driver will then configure each
>> device to run at the fastest common speed of the controller and the
>> device. Therefore, scanners will run at 5MB/sec, disks at 10, 20 or
>> 40MB/sec, etc.
>
> You sure about that? Is that only with certain scsi
> drivers/interfaces? I've always understood scsi to run at the speed
> of the slowest device on the bus (which is why you don't hook up your
> zip drive to the same card as your Ultra2 RAID). I'd be happy to find
> out I was wrong though...
This is more complicated than it seems.
With older drives and/or older host adapters, yes. The entire bus will
run at the speed of the slowest device.
With more modern hardware, however, the host adapter should determine
the speed of each device and run the bus at the best speed it can when
accessing each device.
But there are several gotchas.
1: If you're using Ultra2 devices and non-Ultra2 devices, you will not
get any of Ultra2's benefit. Ultra2 is a differential protocol and
is electrically incompatible with single-ended devices. If the host
adapter detects a single-ended device on the bus, it will disable all
differential features, running your U2 drive in Ultra-Wide mode
(meaning the top speed is halved and much shorter cable length
limits.)
2: Remember that slow devices use up more bus cycle time for same-sized
transfers. A 10M/s transfer at Fast SCSI speeds uses up all of the
bus, while a 10M/s transfer at Ultra-Wide SCSI speeds uses up only
1/4 of the bandwidth. So.... If you have an Ultra-wide drive and a
Fast-SCSI-2 drive on the bus together, and each are transferring data
at 5M/s, you have used up 5/8 of the bus's bandwidth, leaving behind
3.75M/s worth of Fast-SCSI bandwidth or 15M/s of Ultra-wide
bandwidth, not the 30M/s that you might have thought.
3: Bus cycle time and bit-width are two independant factors. If you've
got a Fast-Wide SCSI-2 drive and an Ultra (not wide) drive, the
fastest speed in common between them is 10M/s, even though each one
has a theoretical maximum of 20M/s. One can't handle the faster
cycle time and the other can't handle the wide data path.
4: If you're doing any device-device transfers (like duplicating a hard
drive onto tape), that transfer will proceed at whatever speed the
drives negotiate among each other, which may not be what your host
adapter and OS can negotiate with the drives.
So, the rule-of-thumb (the bus running at the speed of the slowest
device) may not be 100% accurate, but it is still a good idea to put
your high-speed devices (like Ultra-Wide hard drives) on a separate bus
from your slow-speed devices (like Zip and tape drives) if you want to
maximize your throughput.
-- David
------------------------------
From: [EMAIL PROTECTED] (John Hong)
Subject: Re: AMD 5x86/133Mhz
Date: 26 May 1999 23:03:45 GMT
Warren B. Hapke ([EMAIL PROTECTED]) wrote:
: I have one of these, but I've never run it at anything except the
: standard 133 MHz configuration. It's been rock-solid at that level.
: (PS. Sig 11 errors normally indicate a memory problem. What is your
: memory speed; 70ns or 60ns? Maybe the overclocking will work w/
: faster memory?)
After reading the Sig11 FAQ regarding kernel recompilation's, it
is normally due to o/c and memory (whether faulty memory or just bad
timing settings).
: The potential gains from the kinds of overclocking you're talking about
: (150 vs 133) don't seem worth it.
Not by that number surely, but regarding the bus speed there is.
------------------------------
From: [EMAIL PROTECTED] (David C.)
Subject: Re: IDE faster than SCSI UW?
Date: 26 May 1999 19:44:05 -0400
Andy Longton <[EMAIL PROTECTED]> writes:
>
> Since there is only so much CPU available, and the busmastering
> available under IDE seems to have about the same demands as SCSI, why
> not use two UDMA drives in an IDE system ...
Depending on your usage pattern, there may be no good reason to use
SCSI.
As you say, UDMA drives put just as little load on the CPU as SCSI
drives.
You may notice a difference, however, if multiple processes try to
access the disk at once, or if multiple processes try to access multiple
disks at once.
Here's an example. Process A reads a large block from the disk. The OS
sets up the DMA transfer and goes on to service another process. Now,
before the first operation has completed, process B reads a different
large block from the disk. What happens?
On both systems, if DMA is actively happening (meaning the drive has
completed its operation and we're waiting for the host adapter to move
the data into RAM), process B's request has to wait. This is simply
because a single host adapter is only going to be using one DMA channel.
Suppose, however, that DMA is not happening. We're waiting the few
miliseconds for the drive to locate process A's data when B's request
happens.
With an IDE drive, process B must wait for A to locate the data, read
it, and transfer it into memory. Of course, the CPU can service other
processes while this is happening, since we're still talking about DMA
transfers.
With a SCSI drive, the OS can submit process B's request to the host
adapter immediately. Depending on where the data for the two requests
is located on the disk, it is possible for B's request to be completed
before A's. This is something IDE can not do.
Of course, the OS can re-order requests before submitting them to the
host adapter, which helps enormously. But once a request is submitted
to an IDE host adapter, no other request can be submitted until the
first one completes.
This is true both in the case where A and B are reading the same device
and in the case where they are reading different devices attached to the
same host adapter.
Whether this difference translates into a noticeable performance
improvement will depend on what programs you're running and their disk
access patterns. In your case (given your test results), the difference
(if any) may not be worth the extra cost of the drives.
-- David
------------------------------
From: "Gen. Sisyphus" <[EMAIL PROTECTED]>
Subject: Re: Backup solution for a single linux box with about 10Gig drive.
Date: Wed, 26 May 1999 16:52:52 -0700
Thanks for your complete description. How do I know what CD drives are
supported by Linux? The more you talk about this solution, the more I like
it. I am hoping that after the initial complete dump, I can fit all
incremental changes on a single CD. That would be wonderful. Also, is there
a way to create a boot CD? That would also be very nice. I am still having
problem creating my boot floppies!!
Si
killbill wrote:
> I researched my purchase about 9 months ago, and chose the Philips 2600
> (I think... I am going from memory) IDE 2x/6x CDRW. At the time, there
> were only about 4 units selling at the $200 level. The Philips unit had
> 1 MB cache, supported every format I could dream of, was widely sold and
> supported by Linux, and was in stock at buy.com. I was nervous about
> support at the time, so I went with the drive that everyone else had.
>
> It is a good drive, and I am happy with it, but I would probably pick
> another today. I would look for a drive with at least 2MB cache, IDE
> interface, at least 4x write, and specifically supported by cdrecord.
> It would also be nice if it supported UDMA, but I would not make that a
> requirement. You should be able to order one off the web for around
> $200 to $250.
>
> I would ask others for recomendations of drives that meet these criteria
> and that they actually have working. I have not heard anyone complain
> about the HP drives, and my experience has been that Philips equipment
> looks cool and is more innovative, but tends to be less reliable, while
> HP parts are un-inspiring, but perform great and are just about
> indestructable.
>
> As for media cost, it's great. If I am in a hurry, I can go down to my
> local CompUSA and get blank write-once (recordable) disks for $1 each,
> or rewritable disks for about $5 each. If I have more time, I can mail
> order the disks (in larger quantities) from the web, and probably pay
> around $0.50 each for recordables and $3 each for rewritables. Note
> that this works out to about $0.50 to $3 per gigabyte... decent by any
> standard.
>
> For the master system image, I probably generate less then ten or so a
> year, so I just burn them to recordable disks, and have a copy forever.
> The incremental data (/home and /mnt/dosc/data) backups go to a
> rewritable disk. Recordables can be read in just about any CDRom drive
> ever built, and only the newer drives can read the rewritables (which
> have a lower reflectivity), so if portability is an issue then go with
> the recordables.
>
> I love having a CDRW drive... definately one of the coolest peripherals
> I have gotten in a long time. The windows support (via adaptec tools)
> is outstanding as well.
>
> --
> Bil Kilgallon ([EMAIL PROTECTED])
> --"I believe, what I believe, has made me what I am. I did not make
> it, It is making me, it is the very truth of God, not the invention
> of any man". Rich Mullins, quoting G.K. Chesterton.
>
> --== Sent via Deja.com http://www.deja.com/ ==--
> ---Share what you know. Learn what you don't.---
------------------------------
** 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
******************************
