• Welcome to OS2World OLD-STATIC-BACKUP Forum.
 

News:

This is an old OS2World backup forum for reference only. IT IS READ ONLY!!!

If you need help with OS/2 - eComStation visit http://www.os2world.com/forum

Main Menu
Menu

Show posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.

Show posts Menu

Messages - rwklein

#31
Quote from: scrutch2001 on 2011.05.04, 01:10:43
I am having exactly the same problem - Firefox 4 causes a trap and VM crash in Virtual Box (Mac Snow Leopard host).  It does NOT cause a trap in either VMWare or Parallels Workstation on the Mac.

Honestly I don't have hard facts to back it up. But sometimes I really wonder how mature VM are.

Some the things I have seen:

Running MPTS with Virtual PC 2004 dragged down just about the complete host OS Windows XP (system responded almost
in no way to keyboard or mouse input anymore). Could be reproduced every time. Seems some 16 bit code in MPTS screwed things.

A developer I know had weird output with debugging an application with GCC. On real CPU the same
code give different results then in a VM. (Virtual box).

We tested Virtualbox 3.16 with eCS 2.0 at the time and Virtualbox 3.20. Later people reported problems with Virtualbox.
Especially traps and hangs.

One thing we still need to look at. If you load the SMP kernel in Virtual PC it will always trap the system in JFS.IFS (happens during system startup!).
Do the same on real hardware never seen...

For me virtual machines are not the most stable thing in town and with Virtualbox I do have my doubts looking at the varying results people have with installing eCS with Virtualbox.
From the postings I see I get the impression to much stuff is still happening with Virtualbox underneath the boned.
I just have not had the time yet to make a complete inventory. And again I see to little feedback in the ecomstation bugtracker.

From my point of view it should not be the OS that should adopt to the virtual machine but the virtual machine should be a stable environment.  An exception is of course if we clearly have a defect in eCS!

Roderick Klein
Mensys
#32
Quote from: miturbide on 2011.05.03, 18:35:12
Hi

I had read that newers version of Firefox cause traps on eComStation 2.0 when running on VirtualBox.

Now I had tried my self with Firefox 4.0.1 and eCS 2.0 installed on VirtualBox 4.0.4 under  Windows 7 (64bits) Intel i3 Machine.

The first time I run Firefox 4.0.1, when it started to update the profile it gave me:

Exception in Module: JFS
Trap 0003

(screenshot added)

On the next reboot I started Firefox 4.0.1 and worked for a file.
Later it gave me this trap.

Exception in Module: OS2Krnl
Trap 000e


I have both firefox installed (the default one 3.5.3) and the 4.0.1 in different folder using the same Mozilla home path on that VM.
Firefox 3.5.3 works fine on the VM without trapping.

Any ideas how to get newer Firefox running more stable on a VirtualBox machine?

Regards

Well first of all don't modify anything on that VM. With Theseus we need to see where JFS.IFS is loaded when the system comes up. Open a ticket with this trap screen from JFS. Since it is a TRAP 0003 (breakpoint) we should have luck in finding this problem.

Roderick
#33
Quote from: djcaetano on 2011.05.03, 17:37:38
Quote from: rwklein on 2011.05.03, 17:13:57
Paul has/had 2 problems.
1. His video is slow because of an improper setup of the MTRR registers.
2. His hard speed is fixed. He ran an early test build of the OS2AHCI driver and his compile speeds went up quite a lot...

  This is very good news, indeed. :)

Quote from: rwklein on 2011.05.03, 17:13:57
Try and boot ACPI.PSD with /EIS.
Its either the kernel that goes BOOM or its screen01.sys, thats my guess...

What you can also try todo is boot so you get the kernel trap and when it hangs press CTRL ALT F10 twice.
Will it ask for a floppy is so I will give you more instructions. We need to see if we can get a memory dump
to hard disc. Then we can analyse this.

   Well... I believe I've tested EIS before, but tried again... and no changes
on the output screen.
   Probably is the kernel going boom, yes... but maybe because it expects
some special configuration that is not being made. ACPI snooper is doing
a better job than traditional PCI Snooper, since when booting without ACPI
the kernel (or the OS2LDR) simply TRAPs with no further information.

   The really bad thing is this computer doesn't even have a serial port...
hindering the debug kernel useless on this machine.

#34
Quote from: djcaetano on 2011.05.03, 16:49:17

  Hi again, Roderick!

  First of all, thanks for your patience. I know these problems are
awful to spot. On the other hand, there is almost not mention of
new generation hardware on eComStation.ru compatibility list...
Paul Smedley has bought one (Gigabyte) and now I have another
(Intel). Interestingly, Paul's board (based on P67 chipset) is booting
when selecting IDE mode for SATA (instead of AHCI), although he
experienced performance problems (already known, since most
updated drivers are still in progress). I thought H67 would boot
ok, since the basic change from P67 to H67 is the addition of
a video port to use internal video (inside CPU processor)... but
the reality is proving me awfully wrong. :/

Quote from: rwklein on 2011.05.03, 12:22:15
Well lets analyze. Because I don't think this has todo with EFI.
What happens if you switch on ACPI debug in the preboot menu and turn off os2 boot logo and then instead of hitting F10 press ALT F2 what does it hang on ? Screen01.sys ?

  Well... the result is a blank screen with the following text:

--:--:--.-- ACPI_LOCAL_NMI: ProcID FFh LINT1 -

  And nothing else. When I disable all possible features* of my mainboard, it
seems to print a little more information, but hangs none the less... before
trying to load anything. The output screen is this:

----------------------------------
--:--:--.-- ACPI: MCFG @ 0xDAC4C1B0/0x3C (v1 INTEL  DH67CL   0x1072009 MSFT 0x97)
--:--:--.-- ACPI: HPET @ 0xDAC4C1F0/0x38 (v1 INTEL  DH67CL   0x1072009 AMI. 0x4)
--:--:--.-- AcpiOsTableOverride Ex:FEA2C140 New:FFC1FFA4 SIG:[DSDT]
--:--:--.-- Table [DSDT]
--:--:--.-- Table [FACS]
--:--:--.-- Table [FACP]
--:--:--.-- Table [APIC]
--:--:--.-- Local APIC address: FEE00000h
--:--:--.-- System also has 8259
--:--:--.-- APIC_PROCESSOR: ProcID 1h LocalApicID 0h - usable
--:--:--.-- APIC_IO: IO ApicID 0h addr FEC00000h INTI 0
--:--:--.-- APIC_XRUPT_OVERRIDE: Bus: 0h Source 0 Global SysInt 2 - Polarity active: Bus spec Trigger by: Bus spec
--:--:--.-- APIC_XRUPT_OVERRIDE: Bus: 0h Source 9 Global SysInt 9 - Polarity active: High 0 Trigger by: Level
--:--:--.-- APIC_LOCAL_NMI: ProcID FFh LINT1 -
----------------------------------

   Note: I am adding no parameter on ACPI driver. I tried it with several
of them earlier (/CD, /!NOD, /SMP, /APIC)... and nothing changed. The
APIC messages suggested a test with /PIC switch... but again, not a single
change.

   I am using eCS 2.0 GA (which means ACPI 3.18, AFAIK). Maybe testing the
3.19 beta could help? Just putting 3.19 files in the root of a pendrive would
do the work? (this computer doesn't even have FDC ports anymore!)

(*) I said before that I was disabling everything possible because I cannot
disable onboard video (my only video) nor SATA (CD/DVD is SATA) and
cannot disable all USB/USB Legacy, since this motherboard has no PS/2
keyboard port... and eCS CD boot menu (at least the first one) will not work
with the USB mouse with USB Legacy disabled.

   Thanks again for the help!

Paul has/had 2 problems.

1. His video is slow because of an improper setup of the MTRR registers.

2. His hard speed is fixed. He ran an early test build of the OS2AHCI driver and his compile speeds went up quite a lot...

Try and boot ACPI.PSD with /EIS.

Its either the kernel that goes BOOM or its screen01.sys, thats my guess...

Roderick
#35
Quote from: rwklein on 2011.05.03, 12:22:15
Well lets analyze. Because I don't think this has todo with EFI.

What happens if you switch on ACPI debug in the preboot menu and turn off os2 boot logo and then instead of hitting F10 press ALT F2 what does it hang on ? Screen01.sys ?

Roderick

What your basically seeing with these hangs is that the device drivers are getting old in OS/2. Thats why for example eCS 2.0 ships with an updated testcfg.sys, testcfg2.sys. PNP.SYS and ISAPNP.SNP are no longer loaded in ACPI mode.
Its to phase out legacy code. Thats why DANIS506.ADD uses the /!BIOS switch by default in eCS to make booting on modern hardware more successful.

There is being worked on a new build of screen01.sys which has a lot of legacy code taken out. Some calls it makes to I/o ports can be deadly.

So OS/2 is not really outdated. Its the device drivers and that is being worked on. Most of that code is the DDK from IBM.
OS2DASD.DMD and SCREEN01.SYS. As you have seen with USB host controller drivers, thats being worked on well.

P.S. on the EFI stuff people please inform yourself. Its possible, but never tried that when you create a with bootcamp a partition for Windows and then take the hard disc out the Mac and copy a JFS partition it will even boot. Again I never tried.
But most EFI implementations ship with legacy support. Like support for A20 gate, PCI BIOS and INT 13 support.

Roderick
#36
Well lets analyze. Because I don't think this has todo with EFI.

What happens if you switch on ACPI debug in the preboot menu and turn off os2 boot logo and then instead of hitting F10 press ALT F2 what does it hang on ? Screen01.sys ?

Roderick
#37
Quote from: djcaetano on 2011.05.02, 22:16:32
 Hello!

 Finally my new computer arrived. It's configuration is as follows:

- Motherboard DH67CL (LGA1155)
- Processor i7 2600 3.4GHz
- 8GB RAM
- 2TB SATA HardDisk
- SATA CD/DVD-RW
- 950W Corsair power supply

 For now I am (sort of) using Intel integrated video-board.

 The results: TRAP 000d at boot on "Legacy" and "Modern Hardware Safe"
modes. On "Modern Hardware, ACPI" it displays eCS2.0 boot logo, but hangs
right there. Disabling boot logo did not provided any useful information: it
hangs as soon as Kernel version is displayed.

 I have tried SATA legacy mode (IDE mode) as well as disabling
multi-threading. Nothing changed (not even a little bit). I can't remove
the extra memory, since it would void the warranty.
 Any ideas?

 New hardware generation... new instance of old problems.

 BTW, Windows 7 x64 and Ubuntu x64 are booting and working fine (I have
not fully installed them, but until the moment the hard disk must be partitioned,
both installers run just fine).

PS: The manual states that the motherboard has "minimal support to
Windows XP, from Home to Professional"... it should boot eCS 2.0! :/

*** UPDATE
My problem seems similar to this:
http://news.ecomstation.com/article.php?id=24493&group=ecomstation.support.install#24493
Unfortunately, it seems no one was able to provide a solution.


What happens if you switch the hard disc controller back to AHCI mode ?
Are the diskettes unpacked to the memdisc (the white bar).

If it gets passed unpacking the diskette images to the memory disc.
In the preboot menu switch of the DANIS506.ADD driver and see if you get the message
that you need to insert the CD ROM.

If this happens thats good news! Because the upcoming eCS 2.1 will contain the  new OS2AHCI driver!

Roderick Klein
Mensys
#38
Quote from: RobertM on 2011.05.02, 20:35:00
Hi Ned!

Thanks for the tip! Now, if you wouldn't mind, please let us know if in a week or so, the excessive network traffic the Windows boxes sends makes the OS/2 machines stop responding.

Best,
Rob

I think this only applies to servers/workstations that offer shares. If the requester is running to log onto a fileserver it might not get into trouble...

It seems its a messy adventure to get this to work. Congrats Nedcheese on getting this to work. We did try it a couple of times here at Mensys in our lab. But its certainly made a bit complicated from my point of view...

Do you mind if we put this information in the eComStation FAQ and our eComStation manual ?

For some companies it might not be an option to setup this up. In eCS we include a Samba client.
You can get a copy here:
http://svn.netlabs.org/samba
You need a license of Netdrive to make it work.

Roderick

Roderick
#39
That patched testcfg.sys is loaded in eCS 2.0 if ACPI is detected in your system.

eCS 2.0 has 4 operational hardware mode.

1. Modern hardware safe mode (all ACPI settings including the patched testcfg2.sys are loaded), except load ACPI itself.
2. Modern hardware ACPI mode (acpi is loaded).
3. Legacy system (no ACPI is detected). That means old testcfg.sys is loaded for legacy systems and isasnp.snp etc.
4. Virtual machine mode.


Roderick
#40
64 bit main boards to my understanding is just not 64 bit only. It can run 32 bit and 16 bit software as well.
I think I briefly read your previous thread. But I did not see you open a ticket at ecomstation.com.

If your machine has more then 2 GB of ram, in general switch to eCS 2.0 and any older version of OS/2. Depending on the hardware eCS 1.2R might just crash and reboot on loading the kernel.

64 bit claims with the current state of technology should not prevent eCS from booting. Some mainboards have problems. But always deal with problem you find first instead of writing long postings. Because clearing up 1 issue you encounter first might fix other issue's.

Roderick Klein
Mensys
#41
Hello Sigurd,

Let me put it like this. You make it sound as if ACPI is the only reason eCS 2.0 its release was delayed again and again. Thats your point of view. That point of view is far from the truth.

In does beta versions a lot of effort has gone into making JFS work better and making it a bootable file system.
I think you will see some performance difference between a HPFS file system and JFS.

A lot of effort also went into getting UNIAUD to work with HDA chipsets. In the end getting the people to work on it and coordination was done by me.

A lot of people complain about Panorama and its performance. Mensys tried a few years ago to buy the source code of SNAP but we could not reach a deal with the company that took over the source code of SNAP from Scitech. Infact we had systems in the test lab here where Scitech SNAP will just blunty crash on when PM comes up.
We would at the have liked to support more graphical chipsets but its to labour intensive.
So a dission was made...

One of the many items we also had todo was to make certain that all NIC drivers included in Genmac
could legaly be distributed with eCS. The license of eCS 2.0 was also updated so NIC drivers could be included.
If you want to try and get project management done with lawyers with some large companies.
Well its a big job... Thats why certain Atheros and Cisco NIC drivers had been removed from the Genmac package.

As to your opinion that airboot is feature creap is also far from true. Already 4 years we looked
at removing. Its true you found a method to add eCS to the Windows boot manager. And the disc checker
is not needed if you can use DFsee. But both of these scenario's are not an option for all users.
So again a kind of black and white picture view. Already around februari last year I was looking with Steve
at the issue how the boot manager could boot Windows 7. A lot of research was done for sollitions but it was
technically not possible to keep the boot manager. The estimate was made on the information around how long it
would take. But it did not make for the eCS 2.0.
We then decided to work on Airboot. If you think its feature this ticket is proof of its not "feature creap":
http://bugs.ecomstation.nl/view.php?id=2869 It was created July 2010 last year! Sadly Martin Kiewitz
it seems duo to private reasons never resolved the bug.

Somewhere around September 2009 there was a lot of comment eCS 2.0 would not easly install in virtual machines.
Based on the research of Ed Durrant from Australia Joachim worked 6 weeks to update the preboot menu
and the underlying code so it would install as *easy* as possible.

If you check the OS2WORLD archive its thanks to my research and the coding of Alex Taylor that the Windows Disc
Signature of Windows 7 and Vista does not get destroyed. People did post to the os2world forums about this problem
that it happened to them. Other OS'es also

When it comes to ACPI you will not accept that ACPI is easy. In the yahoogroups list I posted you a link to a forum
of guy thats get a kernel panic with Linux on Thinkpad T510. ACPI is far from an easy topic. 

As for the disc checker it went into a lot of testing. And just to point that Linux can also destroy eCS look at
this fresh bug just in.  http://bugs.ecomstation.nl/view.php?id=2914
For the people who don't have an eComStation.com account... Ubuntu basicly writes data to a portion of the JFS
partition and trashes LVM information. It also pooks around in the Windows NTFS according to the user.
Details can be found in the bug.

In the end we did our best. But in the end it seems to happen on more platforms data corruption.

So you paint a picture that is far from the truth Sigurd. I have never denied that communication from Mensys about the
product status is not good. But to say that we uphelt the release for only problems with ACPI is only far from true.
We do far more research and bug fixing then just trying to fix ACPI.

Everybody takes it for granted (which is there right of course) but thanks to research at Mensys done with STece
Levine eCS 2.0 and 1.2R are capable of running RSPINST from IBM and install on AMD 64 based CPU's.

I guess we should have documented that does type if features where not supported ?
Come up with fixes is extremely time consuming where work late into night.

I just had Doug Bisset on the phone yesterday and thanks to our new Multimac driver he can run on a new Thinkpad
510 (or was it 500). So it seems not all new hardware is dead with eCS.

Again there is a lot of room for improvement but call the product a failure in every retrospect is an opinion.

Regards,

Roderick Klein
Mensys
#42
Just as an extra foot note IBM did release up until 2004 the so called DDK which has device driver samples in it.

Overall we have enough sample code to update so it can all work better, device drivers. Thats what is currently being done for example on the USB stack. We hope to release a new set host controller drivers, USB mouse and keyboard drivers in the next few months. But its a lot of work. After that USBMSD.ADD will be worked on.

Roderick Klein
Mensys
#43
I think if you start looking at the history of IBM and how they dealt with requests to open source OS/2 or hand over the source code they have not done. Esther Schindler (a known OS/2 author) wrote an article in I think 2006 about open sourcing OS/2.
Basically all infrastructure within IBM to support OS/2 is gone. Large customers that still have a maintenance contract, a TCO,

To get hardware support you need to make smart choices in your development environment.

One project I came up with that is being funded by Mensys and developed by Softtechnics from the Ukraine
and David Azarewicz is the project Multimac. http://svn.ecomstation.nl/multimac.
Its an as generic as possible C NIC driver skeleton. The developer then needs to take the LInux source code
and put the code in so you can get a driver that can communicate with the hardware.

The same is true for audio now that some bugs in uniaud16.sys have been taken out we just need to stay current with ALSA sources. That is what David Azarewicz is doing.

Thats how, with minimal human resources you can keep an OS on track.

Roderick Klein
Mensys
#44
At a glance the suggestion of Paul would make sense. But you know what most hardware is so identical these days which results in the following examples.

You only know this is you have been involved in device driver development/testing.

For example porting the ALSA driver to OS/2
with UNIAUD once you have one chipset working the rest comes as part of the package. Most of the problems UNIAUD had in the past on OS/2 where because of problems in the interface driver UNIAUD16.

The same goes for USB drivers, most of the drivers we have on OS/2 support basically all USB host controllers and base equipment because the interface to the hardware is pretty standard. Problem is that the old released drivers from IBM are currently needin an update. This is being  done by Lars Erdmann and his drivers are on hobbes.

The same goes for SATA support. Most chipsets have a pretty common interface. The same is true for the AHCI driver that is under development.

Take a look at multimac project (funded by Mensys) http://svn.ecomstation.nl/multimac Browse the source code.
Its split up in Realtek, Intel and Nvidia sources. Once one chipset works a whole truckload of chipsets works.

One last other problem is that the hardware its alive on the market is shorter and shorter. And not all equipment is just as good available in the world. We have had this with one large company customer that needed a *recent* hp workstation 1 year later we had to certify a new workstation for this customer because it was no longer for sale! So go figure!

The basic ball game is that when one chipset for modern hardware is supported and your code skeleton is oke you get a basicly all hardware supported without any trouble. Multimac, Uniaud and updated USB drivers and the Daniela driver are to a large degree proof of this.

The only hardware that is kind of tricky are video chipsets. Thats why eCS uses Panorama, this driver uses the VESA interface. Which won't go away that quickly from video cards BIOS'es.

Roderick Klein
Mensys
#45
A new Realtek gigabit driver is in the pipe line it will be released on svn.ecomstation.nl/multimac.


Realtek screwed up! They released new chipsets revisions under the same vendor and chipset ID. Some of the drivers we use don't work! Windows users have the same problem.

I should have the driver in my mailbox on monday! Then it needs internal testing which can take a week or so. Then we can release it to the world.