i have in stalled redhat 8.0 on a proliant server 2500
i had to type in (linux mem=exactmap mem640@0 mem=256M@1M) that allowed me to install it but now its so slow you cant even use it please help ANY IDEAS???????????????

I have the same problem with RedHat 9, Compaq Proliant 2500 dual proc. I know what the problem is- same as install, it's only seeing 16MB memory, due to that particular machine having a Chipset-enforced "memory hole" at 16M. The "mem=exactmap" and defining the memory trick got it installed, but it didn't get passed on to the boot loader. Of course, RH9 uses GRUB, and I have yet to find how to pass mem arguments via GRUB. So, I had to reinstall, and chose LILO as my boot loader.

However, those tricks that worked so well on install don't do a doggone thing on an installed server. I tried using the exactmap at a LILO boot- my exact syntax was:

linux mem=exactmap mem=640k@0 mem=447m@1m

That only came back with "Invalid Number".

So, I tried method #2, specifying memory in hex:

linux mem=exactmap mem=0x9f000@0 mem=0x1bf00000@100000

Same result, Invalid Number.

If I try just "linux mem=448m" it boots, but only recognizes 16M, and 3 days later it's STILL swapping and hasn't brought up Gnome (I do get a login).

So, I think I have a pretty good handle on WHAT is happening, but not HOW to solve it. Why Linux can't seem to naturally jump over that 16MB memory hole when the evil NT didn't even notice it is beyond me- folks have been complaining about this problem for years. And I can't go back to RH8 and try it- the installer sets the Processor 2 failure flag on the server (seen quite a few messages about THAT too) and that's unacceptable.

I have a Proliant 5000 and on Redhat 9 I am getting the same issues.

Here are two websites that address the issue

how to tell compaqs how much ram you have

discussion on why lilo doesn't work with redhat 9 default install

I am not sure if the latest 2.4.20-20 kernel has that fixed or not. I will probably try it out soon and see if that will fix it.

However on a side note if I do a linux rescue disk and set the boot params to be:

linux mem=exactmap mem=0x9f000@0 mem=0xff00000@0x100000

then my compute sees the full 256mb of ram that i have.

helinut from what i can tell you should be able to use those links and make yours work in lilo just fine.

I was able to make it work by doing the following.
here is my lilo.conf file:

------------------------------------------------------------------------------
prompt
timeout=10
default=Redhat-smp
boot=/dev/ida/c0d0
map=/boot/map
install=/boot/boot.b
message=/boot/message
append= " mem=exactmap mem=0x9f000@0 mem=0xff00000@0x100000 root=LABEL=/"
linear

image=/boot/vmlinuz-2.4.20-8smp
label=Redhat-smp
initrd=/boot/initrd-2.4.20-8smp.img
read-only

image=/boot/vmlinuz-2.4.20-8
label=Redhat-up
initrd=/boot/initrd-2.4.20-8.img
read-only
------------------------------------------------------------------------------

this worked just fine for me and I now have all my RAM to use.

After a terrible time trying to get Redhat 7.2 configured to see all our Proliant 5000's 656M of RAM we could get the system to see the RAM when an argument was entered at boot, but we could never get lilo edited properly. We punted and tried to load Redhat ES. At the install boot prompt we entered mem=620M@16M. The install went very smoothly. When later given the opportunity, we opted for lilo and entered the same string (mem=620M@16M) in the screen form. The system now boots properly and sees all the memory. We may still experiment a bit to find out how much RAM can be used. When we tried 640 in our earlier install it failed, although 620 worked. Perhaps we need to try 639 or less.

I'm a newbie here so if it seems like I'm trying to teach my grandmother how to suck eggs then just say. Have you tried the following .....

during memory test take note of the memory that is displayed and divide that number by 1024.

Round this number down to be on the safe side and take away a meg for example my machine reports 196 blah / 1024 gives me 192 blah - 1 = 191MB

when installing type the following

linux mem=exactmap mem=640K@0 mem=191M@1M

This should be more or less spot on for your actual memory

I only just got my ProLiant 5000 running on Sunday and had been trying on and off for a week perusing web sites galore but the following worked for me

Insert SmartStart when you turn the machine on
Perform system erase
When it asks what OS you want to install pick SCO 2.xx
Let it do it's thing
Configure RAID controller
Let it do it's thing
When it comes to reboot time pop in your RH CD
type "linux mem=exactmap mem=640K@0 mem=191M@1M" at the prompt (for my 192Mb machine)
When you get to partitioning don't erase all partitions as this will scrub the 36 Meg CMOS partition if I remeber rightly
Pick LILO as the boot loader and specifiy addtional parameters as "mem=exactmap mem=640K@0 mem=191M@1M" (for my 192Mb machine)
Pick your packages
Restart
Hey presto I am Windowless

I found that if I used GRUB my machine was so slow it was unusable so I tried LILO and it worked fine. I think when you reboot after telling RH directly at install and then use GRUB GRUB cannot see past the 16Mb barrier. By using LILO and specifying these parameters at install it passes these through on first boot whereas with my machine GRUB couldn't see past the 16Mb barrier and I automatically thought my machine couldn't hack it.