NotebookForums.com › Forums › General Notebook Discussions › Linux & Other OS's › 5670 Notebook with RedHat 9..Conclusions
New Posts  All Forums:Forum Nav:

5670 Notebook with RedHat 9..Conclusions

post #1 of 18
7/10/03 at 9:41pm
Thread Starter 
Yes, well how, no I must do my Hackus Dance....drum roll please....



I just got my 5670 Notebook, and well, it is doing well thank you...

Some notes and issues regarding the notebook I wish to confer, and perhaps resolve.

First of all, RedHat 9 works, about 85%.

Things out of the box that worked for me:

The Radeon controller worked fine, including full DRM/DRI support with 4.3.

Love the LCD plate on this laptop, it is twice as bright as my old IBM ThinkPAD A21p.

1600x1200 looks gorgeous.

I purchased the laptop with about 1GIG of ram. That helps alot, since besides doing Linux Kernel development, I also do alot of Java programming for my website: www.aesgi.com.

Sound works fine, the modem, is a winmode, and well couldn't care less about it working as modems are so PAS SE.

:-)

Considering the past machinations I usually had to go through to get Linux, specifically X windows working on laptops I owned, I am very pleased.

Now..for the unpleasing parts...:-(

The BIOS on this laptop is, shall I say, meager. Beyond meager...almost criminal witht he lack of options, and more importantly:

How do you update the firmware?

This is a valid question, because I think the APIC controller in the laptop, by default has bad routing information in it when HyperThreading is enabled, or APIC support for UNI Processors is enabled in the laptop.

However, I can't tell for sure....what follows is a complete mystery to me and has to do with PCMCIA support.

When APIC support is enabled the yenta_socket manager is unable to route a IRQ pin through the APIC routing table.

Obviously, this causes problems.

I get a lot of PCMCIA failures from yenta_socket when APIC features are enabled in any of the kernels I build.

APIC support is big problem with these notebooks, at the moment so I will have to do some more reading and get back to everyone if I find a cure all.

The other problem is DRM support for 2.4.21 kernels. I have tried a great deal many kernel drm updates, patches for kernel (linux) and kernel (XFree86) and I can't get DRM to work on my 5670 with a 2.4.21 kernel.

2.4.20 works just fine, but not 2.4.21.

Now with regards to USB support...

I purchased a external USB hard disk with both FireWire and USB support. Works just fine with RedHat 9.

The only glitch is that I lost USB mouse support whenever the hard disk is in use.

Nothing show stopping, and for the most part everything works.

We need:

1) BIOS updates
2) Better APIC support in Linux or a BIOS update.
3) Better DRM/DRI support for the Radeon 9000. (Comming soon once XFree 4.4 and kernel 2.4.22 comes out.)

-Hack
Reply
Reply
post #2 of 18
7/10/03 at 11:23pm
what bios do you have? is it an 'insyde' bios, with drop down menus in the config? because that's the one i have in my 8887, and i hate it too. you can't even set the agp aperture. talk about a castrated bios...

update the firmware... of what? do you mean update the bios? the cpu microcode? i'm not quite sure what you mean. i assume you mean bios update, since that's what you mention at the bottom. what version is your bios? i updated my bios, and it was the same as any other computer - just download and make a boot floppy.

i don't use pcmcia or usb-storage devices (other than the 8887's mp3 player) so i can't really help you with those, sorry. are there any error messages? those are always useful.

i don't really have problems with apic (or ioapic, for that matter), presumably because i don't use pcmcia. is your comp hyperthreading? (i.e. are you using an smp kernel with apic/ioapic automatically enabled, or did you manually enable it for a uniprocessor kernel?)

on the pcmcia note, are you using the pcmcia support in the kernel? i believe the pcmcia stuff in the kernel is outdated; supposedly it's much better and more reliable to disable kernel pcmcia support, and use the latest from the pcmcia-cs project.

i don't have any problems with 2.4.21 and drm. are you building the module into the kernel? i normally just enable only drm support in the kernel with no drivers, and use an external package to build drm modules. again, does it give error messages when it fails?

i think the drm support is not bad. of course, it could always use improvement, but i get 90 fps in quake3, which is pretty good for a reverse-engineered driver.

or you could try ati's linux drivers - that's what i do at the moment, and i get 160 fps in quake 3.
Reply
Reply
post #3 of 18
7/10/03 at 11:55pm
Thread Starter 
My BIOS is a Pheonix BIOS.

Firmware I mean BIOS of course, for the motherboard.

I will try building a seperate package for the kernel pcmcia system tonight and see what happens.

Good suggestion I didn't think of that.

I have tried just about ever DRM combination there is and cannot get 2.4.21 to enable DRI.

I never had that sort of a problem before with DRM between kernel versions.

Normally I would just fetch a new drm distro, rename the old one in the char/drm tree and rebuild the kernel that way, and normally it would work.

Sometimes I would fetch new drivers from gatos as well for the XFree86 kernel.

Just haven't had any luck with 2.4.21

HyperThreading works fine, but like I said, APIC support for devices on the PCI bus is sub standard, and PCMCIA devices stop working when I am in Hyperthreaded mode due to the fault yenta_socket errors.

I will try the PCMCIA rebuild tonight and get back to ya.

-Hack
Reply
Reply
post #4 of 18
7/11/03 at 12:17am
well, how far do you get with drm? like does the module load or not, compile or not, does x not start up with the radeon module, are there errors in XFree86.0.log, etc? like, how does it 'not work?' i'm just trying to figure out what's going wrong with it.

there are also times when stupid little things cause big problems. i remember one time when i fought with the alsa driver for a while, before finally realizing that /usr/src/linux pointed to the wrong kernel sources. is your symlink correct? i'm just suggesting whatever i can think of at this point.

gatos is incompatible with drm, but you probably knew that already.

dunno... i'd like to help you get dri(drm) working, because i remember fighting with it to get it running. and i never successfully got it working with 2.5, so i know what the frustration is like...
Reply
Reply
post #5 of 18
7/11/03 at 4:03pm
Thread Starter 
Everything works, except that the XFree86 log shows DRM is not enabled.

It isn't too terribly clear as to why that is under 2.4.21. The error log isn't too terribly helpful.

I was going to do a trace, but have decided to wait for the DRM/DRI issues between the X 4.3 branch and 2.4.21 to be resolved under 2.4.22 for my particular machine.

The pcmcia package install was a bust.

pcmcia 3.2.4 package I got is pretty much a no go either. (pcmcia-cs.sourceforge.net)

From what I can tell, there are a number of RedHat instigated issues with 2.4.21 with PCMCIA as well:

1) Attempts to update the kernel with pcmcia package 3.2.4 is unpractical. First, 2.4.21 will not build yenta_sockets on my machine without it enabled.

If you don't enable pcmcia services, you don't get a yenta_socket.o module. Without it, the modules built with the pcmcia package will not work and no cardbus card I have will run without yenta_sockets.

2) I am afraid it is even worse than that. If you reenable pcmcia in the 2.4.21 config, yet do not build the pcmcia modules for specific cards in the kernel (I turned them individually off hoping I could just copy the new ones there.), the pcmcia package will not build the latest core pcmcia services module, and simply builds the bin utilities.

No core utils modules, no latest card drivers, no nothing.

No help at all there.

3) Furthermore I got a real nasty surprise. The init.d script pcmcia thatr RedHat packages with Linux is not pcmcia-cs friendly. That is, the script has a entry in it that incorrectly executes a depmod -a in /lib/kernel/modules/... and not /lib/kernel/pcmcia, like it is suppose to do.

That is where pcmcia-cs 3.2.4 puts the new pcmcia core, services management, and new drivers for the cards.

So when you startup, depmod -a puts the wrong entriesin the database, and the system can't find the module to insert during a hotplug event when you put the card in.

That sucked. RedHat 9 gets a big
for that one.

--

What I have found, is that if I force the pcmcia utils package to build the latest pcmcia_core services, and over write the core pcmcia, ds, etc modules, yenta_sockets complains of undefined symbols and simply exits.

So it would appear, 2.4.21 and the latest 3.2.4 packages are not compatible with the yenta_socket services with 2.4.21 on my particular machine.

If you had one of the standard iNTEL controllers, that don't require yenta_sockets services, it might work. (Or if you only have 16 bit cards)

In any case, 2.4.21 is a pretty problematic kernel for my laptop and I am going to wait for 2.4.22 to resolve some of these issues.

-Hack
Reply
Reply
post #6 of 18
7/11/03 at 7:16pm
hmm... i've had absolutely no problems with 2.4.21. (the kernel itself, extra patches are a different story) i'm not a redhat user - not since 6 or 7 - so i can't really say much about the stuff that comes with it.

one thing... you're building these yourself, right? there were occasions i remember where installing the rpm of the package from rpmfind.net or something worked when building it myself didn't. just a suggestion.

i can't give you exact help with pcmcia since i have absolutely no pcmcia cards. i looked it up on the pcmcia-cs page, and yenta_sockets seems to be the main cardbus module for 2.4, unfortunately. i stumbled across a couple interesting things, though. i don't know if you've read the stuff on the pcmcia-cs page... but the documentation has a lot of info - there is a troubleshooting problem->explanation page, with a bunch entirely about irq's, which you mentioned was the problem. also, there is the possibility that you may have to pass specific irq options to the driver; the documentation has a list of a bunch of specific laptops and the corresponding irq options. while i didn't see sager on there, there was mention of a prostar that you may want to try. also, you could even try the others if you felt like it. maybe check out the bug list, too - i looked and didn't see your problem, but you would have a better understanding, so you might see something. maybe do a lspci -v and check if your cardbus bridge is supported by pcmcia-cs. also, the memory and irq info from lspci could help with passing irq options. ... that's all i can think of at the moment.

about drm:
what version of DRM are you using? (or datestamp, if that's what it is)

is the radeon module loaded in lsmod?

i assume that glxinfo also says no direct rendering, right? (i.e. it matches the x log or not)

here's something you might try - are you loading agpgart as a module? because if so, you might want to try loading it with the option (parameter) agp_try_unsupported=1, if you're not already doing so. or you could try an 'options agpgart agp_try_unsupported=1' below the line in the module aliases file associating agpgart with char-major-10-175 - that's not the way i do it, though, so i can't tell you if that'll work correctly; i just put it as a parameter in the module autoload file. i think that's why i've seen mention of the fact that agpgart _has_ to be a module for drm to work.

could you post your x log, if you don't mind? i've seen weird stuff, where the drmOpenDevice lines say they opened /dev/dri/card0 ok, but then something stupid happens in the middle with the [agp] lines. i have also seen cases where everything looked fine but it didn't work - in 2.5, x would start up fine and a log showed everything perfect, and it would only get a black screen, without ever getting to open fluxbox or show a mouse cursor, even.

also, you might want to try the most verbose x log possible, with startx -- -logverbose 8, if you haven't done so already.

i dunno... my laptop has no problems with 2.4.21 and i didn't think our laptops were _that_ different... somehow, i think this stuff should work... but i suppose it's different for everyone.
Reply
Reply
post #7 of 18
7/12/03 at 12:26am
Thread Starter 
Well,

I am much happier now....time for the Hackus Hack dance of kernel hacking...drum roll please....

(It helps if you have Sammy Hagar "Best of Both Worlds" playing in the background while watching the dance....)



With that said, I have I believe perfected my RedHat 9 install, with no problems. If you want a perfectly running system, at the fastest possible speed for your 5670, follow the directions below:

First you need to do some fairly heavy lifting though....as in rebuild a custom kernel.

The lame kernel that comes with the RedHat installation is compiled for 386.

It needs to be rebuilt with Pentium 4 extensions for the processor options.

You also need to turn on HotPlugging. (But DO NOT include the IBM HotPlug Extensions or the Compaq stuff, turn everything off except the HotPlug service itself.)

Also, turn off APIC/MultiProcessor support. I know, it sucks, but if you want your system to be rock solid, you have to do it.

This next option is very important, turn OFF the default large memory (4GB) extension. It creates havoc with AGP aperture and memory areas with the AGP module kernel.
(I eliminated all of my Radeon context errors, memory free (allocation) errors.)

I haven't done a trace yet, but I believe this is yet another firmware issue with the BIOS.

WE NEED A FIRMWARE UPDATE FOR 5670 LAPTOPS....

Here is what I get with the AGP driver, with NO ERRORS. You should too.

[drm] AGP 0.99 on Intel i845 @ 0xec000000 64MB
[drm] Initialized radeon 1.7.0 20020828 on minor 0
[drm] Loading R200 Microcode

To get clean RADEON performance on your 5670, you have to have the proper config to build your kernel.

In your /usr/src/linux-2.4 directory you will see a "configs" directory.

This config directory covers all of the RedHat options you need.

In particular, you want this file:

"kernel-2.4.20-i686.config"

First, do a make mrproper;make clean before you load the config file to set it up! This is very important to do, so make sure you do it!

Do the standard make xconfig in windows, or make menuconfig, and turn on/off the options like I directed above by doing a load Configuration, using this file. Save it and exit.

Note: you want to insure that you DELETE any pcmcia or misc directories under your /lib/modules/<kernel>/ tree otherwise it could break the make install. So make sure you don't have any of those hanging around before you start your kernel build.

I am a VMware user, so the first time I did this I forgot vmware makes a misc directory in the kernel. This resulted in linker errors.

Any linkage errors will prevent grub.conf from being updated.

Note: You also must have at least ONE template in your grub.conf file. Otherwise, grubby won't be able to update the file automatically for you. So never delete your entire grub.conf file unless you know how to make a grub.conf kernel entry from scratch!

If not, then make install will fail as well.

So, I have PCMCIA working great and DRM is working great, and I have a kernel compiled with Pentium 4 extensions, which runs faster under gcc 3.3.

if you are interested, I can send you a .config for you to use on your 5670, so you can have a nicely running, behaving RedHat 9 installation.

Now, for something I learned that is not so pleasant. Appearently, the standard Pentium 4 processor doesn't cache anything above half a gig of memory.

Of course, you can turn on APIC (Enable SMP Support) processing if you WANT TO, with this config. Be aware though, I wouldn't run PCMCIA services with it on and your setup won't be as reliable.

Kinda sucks as Linux just got Posix Threads.

Cheers, Hack.
Reply
Reply
post #8 of 18
7/12/03 at 12:21pm
Thread Starter 

Another Update RedHat 9 and USB 2.0

My 5670 laptop, after building a custom kernel, exhibits more undesirable operation with APIC enabled either locally or through SMP.

Kernels compiled with APIC support for either hyperthreading or local APIC support do not mount USB devices.

Appearently the kernel can't assign a device ID to the ACOM 60 gigabyte USB2.0/Firewire drive I recently purchased to do my backups. (Got the drive at BestBuy)

On the other hand, the non APIC or non SMP kernel I built works fine with the drive, with a bit of a quirk.

That quirk is that while the drive is mounted and recognized OK by RedHat 9, large data transfers to the drive freezes X-Windows for about 3 minutes at a time.

During those 3 minutes you can't do anything, and it looks like the 5670 is locked up.

It isn't though, it is just really unresponsive while copying large blocks of data.
Reply
Reply
post #9 of 18
7/14/03 at 1:35pm
2.4.x is dead baby - grab a copy of the NEW 2.6.0 kernel (brank spanky new today) and tweak that. I have been using 2.5.74 for a while now and it is far superior to 2.4.21.


Just my 2c worth
Reply
Reply
post #10 of 18
7/14/03 at 2:58pm
Thread Starter 
Yeah I seen that.

2.6.0 TESTING branch is out.

I am a bit leary of testing a 2.6.0 kernel out on my laptop, perhaps on a TEST server, to screw around with.

RIght now I am wrestling with a Oracle 9i install in my Laptop under RedHat 9.

Talk about learning everything there is to know about ld and trace.

:-)

Not for the faint hearted I am afraid....

I just got XWindows, DRM/DRI, PCMCIA and all that stuff worked out under 2.4.20 on the 5670, I think I need a break!

Maybe around the end of August I will take another swipe at Kernel update.

Right now, I just got my Java, and my Kernel Development environment all nice and neat and I am starting to get some serious work done. (i.e. WORK=Pay Bills, Eat Food).

Next week I start hitting the Madison Wisconsin area with my Penguin act and DLP projector....

I wonder how many companies really know how to fit Linux into thier business beyond the server room or the techies?

Gonna find out!

-Hack
Reply
Reply
post #11 of 18
7/14/03 at 3:54pm
kernel 2.6.0 is _not_ out. today, linus announced on the lkml that 2.6 was to enter the _pretesting_ phase soon. it's essentially a beta test, where the codebase is mostly frozen and no more major feature changes will be made - a stress test, as opposed to the 2.5 series, which is basically an alpha since the features were still being added/removed. and while it has been announced, the first prerelease hasn't even come out yet. linus has always maintained a 'release when it's ready' mentality, so there's no real way to tell how long pretesting will take place until the real 2.6.0 comes out - and even then, the first few versions of 2.6 should still be considered testing until everything is ironed out.

pretesting generally lasts for quite a few months. there was a 7 month testing phase between the release of 2.4 pre1, and 2.4.0. don't expect to see 2.6.0 for some time. the only thing that happened today was an announcement; there was nothing released. got to kernel.org and look - no 2.6-related releases anywhere.

2.4 is far from dead. it will never be dead, considering the fact that 2.6 is geared for desktop usage - 2.4 will still be used in many places on servers.

2.5.74 is good, but it's not for everyone, because it's still alpha. i tried it out and it was nice, but it ate my LVM setup a couple times upon conversion to LVM2. so in some ways - namely, stability and compatibility - it is not really superior.
Reply
Reply
post #12 of 18
7/14/03 at 4:06pm
Ok coupel of things

1) 2.6. is out today - pretest yes its a test01 version but it is still labelled as 2.6.0 - replacing the 2.5.x tree. Look on kernel.org (its very slow atm because of people grabbing the 2.6.0-test1)

2) I was using linux in 1992/3 matey when slackware was all we had and mostly we just got it from friends who had the time to download the disks so I am well aware of the release protocol.

3) I meant 2.4.x is dead as a joke - hence the light terminology. Before you FLAME ON! like the torch perhaps you should dial your seriousness back a few notches

4) In many respects the 2.5.74 kernel is far superior to the 2.5.21 kernel. But as you said in regards to stability and compiling every time - its not, though I would also say I have had zero problems with it after the initial config issues. Indeed speed of IO and compile and general performance are much better.

Finally - Do not assume that everything posted is something that needs to be responded to - thats why I say just my 2c worth at the end - not "This is the word of Linus". Not everyone is a moron who needs to be tuned in. Some of us are veteran Linux (and UNIX) users who are just watching the boards and posting occasionally.

Again

Just my 2c worth

Quote:
Originally posted by xiphux
kernel 2.6.0 is _not_ out. today, linus announced on the lkml that 2.6 was to enter the _pretesting_ phase soon. it's essentially a beta test, where the codebase is mostly frozen and no more major feature changes will be made - a stress test, as opposed to the 2.5 series, which is basically an alpha since the features were still being added/removed. and while it has been announced, the first prerelease hasn't even come out yet. linus has always maintained a 'release when it's ready' mentality, so there's no real way to tell how long pretesting will take place until the real 2.6.0 comes out - and even then, the first few versions of 2.6 should still be considered testing until everything is ironed out.

pretesting generally lasts for quite a few months. there was a 7 month testing phase between the release of 2.4 pre1, and 2.4.0. don't expect to see 2.6.0 for some time. the only thing that happened today was an announcement; there was nothing released. got to kernel.org and look - no 2.6-related releases anywhere.

2.4 is far from dead. it will never be dead, considering the fact that 2.6 is geared for desktop usage - 2.4 will still be used in many places on servers.

2.5.74 is good, but it's not for everyone, because it's still alpha. i tried it out and it was nice, but it ate my LVM setup a couple times upon conversion to LVM2. so in some ways - namely, stability and compatibility - it is not really superior.
Reply
Reply
post #13 of 18
7/14/03 at 4:28pm
ok, i only glanced over the page and missed the 2.6 prerelease; that's my mistake.

however, i did not mean to make you out as a moron. but as you've probably noticed, there are a number of newbies on this board; more than there are experienced users. that is not intended as a derogatory term, but rather, to indicate the simple fact that they don't know as much of linux as experienced users - i wanted to clarify the difference between a testing version and the real 2.6.0. it was to prevent the possibility of a lot of newbies going and unknowingly running the test version, which naturally has a very similar codebase to 2.5.75, and ending up with a broken system for some reason or another.

i'm an extremely serious person, which has to do with a family history that i don't like to talk about. i can't really help it if i don't easily catch onto other people's jokes.

i don't really consider my post a flame - flames generally have a lot more aggression or insults in them. i was merely making a couple corrections, ones that in retrospect, were not really necessary. but there wasn't really aggression towards you in the post.

but even if you say that it's only your two cents' worth, it is a post in a message board thread, which obviously would warrant the possibility of it being replied to.
Reply
Reply
post #14 of 18
7/14/03 at 5:04pm
I would think that most linux neophites would just grab a pre-packaged distro (a la Red Hat or Mandrake) rather than head over to kernel.org and roll their own.

By the time a user is ready to roll their own, they are most likely familiar with linux and it's ups and downs. I consider myself a "newbie" at linux - and I have for many years (even back in the early 90's and trying my first Slackware install).

I appreciate all the input in the Linux section here, and keep my eye on it regularly as I've never lost interest in linux (though I don't have it running on my 8887 - yet).

-myrkat
Reply
Reply
post #15 of 18
7/14/03 at 5:06pm
Forsake the dark side Luke - come back to the light - Linux will make you more powerful than you could ever imagine.

Just keep a windows partition for running those games that wont run under winex

Signed

Obi-wan-linuxobi


Quote:
Originally posted by myrkat
I would think that most linux neophites would just grab a pre-packaged distro (a la Red Hat or Mandrake) rather than head over to kernel.org and roll their own.

By the time a user is ready to roll their own, they are most likely familiar with linux and it's ups and downs. I consider myself a "newbie" at linux - and I have for many years (even back in the early 90's and trying my first Slackware install).

I appreciate all the input in the Linux section here, and keep my eye on it regularly as I've never lost interest in linux (though I don't have it running on my 8887 - yet).

-myrkat
Reply
Reply
post #16 of 18
7/14/03 at 5:19pm
Quote:
I would think that most linux neophites would just grab a pre-packaged distro (a la Red Hat or Mandrake) rather than head over to kernel.org and roll their own.
i guess... but i remember the first thing i did after installing my first distro and learning the basic command-line operations was to make my own kernel. maybe i'm just weird.


Quote:
Just keep a windows partition for running those games that wont run under winex
that's what i do, too. a necessary evil, i suppose.
Reply
Reply
post #17 of 18
7/14/03 at 10:25pm
Thread Starter 

Redhat 9/Oracle 9i

I did it.

Anyone here interested how it performs on a 5670?

If so I will post my conclusions here.

To summarize:

9i is a beast! It puts my 5670 through its paces with the work I am giving it to do.

However, my 5670 tames Oracle 9i very nicely.

Runs very well so far....

-Hack
Reply
Reply
post #18 of 18
7/15/03 at 12:41am

Re: Redhat 9/Oracle 9i

Nice one - out of interest got the 2.6.0-test1 kernel running - nothing really amazing from 2.5.74 but works well enough.




Quote:
Originally posted by Hackus
I did it.

Anyone here interested how it performs on a 5670?

If so I will post my conclusions here.

To summarize:

9i is a beast! It puts my 5670 through its paces with the work I am giving it to do.

However, my 5670 tames Oracle 9i very nicely.

Runs very well so far....

-Hack
Reply
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Linux & Other OS's
NotebookForums.com › Forums › General Notebook Discussions › Linux & Other OS's › 5670 Notebook with RedHat 9..Conclusions