http://www.os2world.com/content/view/19256/1/ (http://www.os2world.com/content/view/19256/1/)
QuotePosted by Martin Iturbide - Sunday, 30 August 2009
Seren. .. .
I am very pleased the development is reaching "gold" stage, but I still had no success on booting eCS 2.0 Silver
on my desktop machine (legacy, Athlon XP 2600+ running on an ASUS A7N8E-DeLuxe). Even setting
eCS to install in "legacy" mode, it hangs just after loading the boot logo. ALT-F2 told me the last thing
loaded was PCIBUS.SNP.
I have been using OS/2 Warp 4.52 with the latest fixes and everything on this machine, never had
a problem...
Any clues?
Anyway, I was able to have eCS (RC6) running it on my Positivo Modo Netbook (similar to MSI Wind).
Had not tried RC7 yet.
I'm using RC7 and i can say to you that it's very nice! ;)
DJ, did you try any of the other profiles? I got mine to install on my positively ancient eMachines 633ids by selecting ACPI (not safe, not legacy). Any other option would fail.
Quote from: Saijin_Naib on 2009.09.29, 22:58:50
DJ, did you try any of the other profiles? I got mine to install on my positively ancient eMachines 633ids by selecting ACPI (not safe, not legacy). Any other option would fail.
I tried legacy and normal (not safe) ACPI. On the first it hangs on PCIBUS.SNP. On the second it gave me the terrible "Operating system cannot operate your hard disk". I'll try the "safe" version. Just as a remark, the driver used was DANIS506 (tried both modes), since my machine is using a SATA harddisk, which is not recognized by IBM1S506.
OS/2 Warp 4.52 is running fine with DANIS506.
I'll do some more tests. Thanks for this suggestion.
Quote from: Saijin_Naib on 2009.09.29, 22:58:50
DJ, did you try any of the other profiles? I got mine to install on my positively ancient eMachines 633ids by selecting ACPI (not safe, not legacy). Any other option would fail.
Ok.... tried again. Legacy mode hangs on PCIBUS.SNP and both Modern modes gave me "OS/2 cannot operate your hard disk". Messages are all gone when this message is displayed, but turning on "extra messages" on boot menu I could see DANIS506.ADD seems to be loading ok (at least it has identified everything).
Now it seems the last thing the system tried to load was the OS2LVM thing. Did Serenity modified anything there?
PS: My system has an IDE HD side by side with a SATA HD. IDE is primary and SATA is secondary (althought my operating systems are installed on the secondary disk... my motherboard BIOS always place SATA disks after IDE disks.
Regards!
Is it a nvidia shipset ?
If yes, add nvidia specific flags on the APIC like /MTR ...
Remy
What switches do you have on the DANIS driver? The /!BIOS one?
Try running DANIS with NO switches what so ever. Also, remove one of the two drives. See if you can narrow down the issue.
Quote from: Saijin_Naib on 2009.10.02, 00:50:30
What switches do you have on the DANIS driver? The /!BIOS one?
Try running DANIS with NO switches what so ever. Also, remove one of the two drives. See if you can narrow down the issue.
I tested the system only with SATA HD and had some success (more on that later). Then
I tested it only with IDE HD and... it seems the problem is really with OS2LVM.DMD, it refuses
to work with my IDE harddisk (!?!), the "OS/2 Cannot operate your harddisk" error shows
up using IBMS506 or DANIS506, whatever switches are used on the adapter/device
initialization. This is bizarre, some kind of ugly regression. The problem happens with and
without ACPI.
About the "success" using SATA HD, it was limited. First, SNAP gave a weird result:
as soon as the graphical shell is started, my monitor (22" LCD that supports up to
1400x1050) goes out of sync ("No Signal"). This is weird, since I use SNAP for years
on my OS/2 4.52 without any glitches. My video board is a Matrox MGA-550.
This problem happens with and without ACPI.
I tried Panorama and then the system booted (with the inherent performance
problems) but then I noticed something: the partition where I would like to install
eCS 2.0 silver is located at my IDE HD (my test partition). Since I cannot boot to
that harddisk, no donuts for me.
As a remark, no matter the harddisk configuration, the "Legacy" mode of
eCS
never complete the boot. It always locks on the first thing loaded:
PCIBUS.SNP.
I really think it is very cool and I really expect eCS to boot and work fine with
modern hardware, but I find it very odd this regression, since it is not working
in a machine built to run OS/2 (every piece of hardware was selected and
bought due to its compatibility with OS/2). Even if newer eCS features do not
work on this old piece of hardware, I would expect at least to be able to install
it and run it in the same way I did using OS/2 4.52.
(I remember having some problems with OS/2 4.52 on SATA+IDE configuration,
but there was no problem when installing OS/2 on the IDE HD after disconnecting
the SATA HD).
I am not a noob OS/2 user, I am using it for almost 17 years now... but this
problem is really puzzling me.
I thank you for the help and I hope someone come up with some new
ideas so I can try to solve this problem.
Regards!
Can you give me a run-down of your BIOS Settings?
For me, I needed PnP Aware OS set to off (could have just needed to be toggled to do a PnP refresh), and I needed Legacy USB set to ON even though I only use PS/2 keyboard/mouse.
Please post all the settings you have.
Quote from: Saijin_Naib on 2009.10.02, 22:39:54
Can you give me a run-down of your BIOS Settings?
For me, I needed PnP Aware OS set to off (could have just needed to be toggled to do a PnP refresh), and I needed Legacy USB set to ON even though I only use PS/2 keyboard/mouse.
Please post all the settings you have.
I did some more tests yesterday.
My BIOS do not have "PnP OS" option, but does have APIC option and
Manual/Automatic hardware detection. The latter does nothing to OS/2.
I also messed with USB config, but nothing...
Disabling APIC, though, allowed be to pass PCIBUS.SNP on Legacy
mode, after some delay (it seems PCIBUS.SNP doesn't like APIC)...
but then, the same "Operating system cannot operate your harddisk"
message puzzled me.
I tried to disable SATA drive but it seems there is no way eComStation
understands my old IDE PATA harddrive, *even* running IBM1S506.ADD.
(The error message is displayed just after DANIS506.ADD is loaded).
The only thing I can suspect is OS2LVM.DMD doesn't like my PATA
IDE controller (the problem still happens even disabling SATA controller
on hardware!).
It seems the solution will be buy another SATA harddisk. :/
Weird.
Hrm. I had an issue on this computer once that was related to the drive needing to be formatted using DFSEE. It would pause during the boot process on the SDDHELP.SYS.
I ended up doing all types of deleting volumes, zeroing the drive, LVM refresh, Boot-area operations, etc to the drive. After I did everything possible and had the drive ready with LVM and CHS values eCS could understand, it loaded just fine.
This is a very wild guess, but concerning the PATA/SATA drives:
BIOS on my notebook has a AHCI support Enabled/Disabled.
Do you have the similar thing in you BIOS? If yes then please Disable AHCI - eCS does not support it.
Quote from: StefanZ on 2009.10.16, 18:28:15
This is a very wild guess, but concerning the PATA/SATA drives:
BIOS on my notebook has a AHCI support Enabled/Disabled.
Do you have the similar thing in you BIOS? If yes then please Disable AHCI - eCS does not support it.
I have eCS instaled with AHCI support and working. At least it is enabled in bios.
Update: ---------
IBM Lenovo x60 tablet. eCS silver release, installed with new hardware support.
Quote from: buca on 2009.10.17, 00:10:28
Quote from: StefanZ on 2009.10.16, 18:28:15
This is a very wild guess, but concerning the PATA/SATA drives:
BIOS on my notebook has a AHCI support Enabled/Disabled.
Do you have the similar thing in you BIOS? If yes then please Disable AHCI - eCS does not support it.
I have eCS instaled with AHCI support and working. At least it is enabled in bios.
Ah, that is interesting :)
On my notebook (MSI) the eCS will not install with AHCI enabled.
Well,
This week I bought a new SATA 250GB harddrive and was able to install eCS2.0 RC7
after removing the IDE harddrive. But the install process was not without some sort of pain.
The reason for the pain was the fact eCS is set up to boot at 800x600 and since SNAP
seems to not recognize my monitor on the first time it runs, 800x600 with some sort of
unsupported refresh rate is selected... causing an annoying "No Signal" message ("No
Sync").
It took me some time to discover what was happening, because everything worked
just fine when boot with Panorama. I had to revert back to VGA (after "out-of-sync,
GenGradd support didn't work also) and start over.
I did it several times until I discovered that: reverting to VGA, installing Panorama,
setting 640x480x4 bits (sixty colors) and THEN installing SNAP will boot correctly
in 640x480 mode. THEN I go to Screen object, set the monitor (now correctly
detected as "LG TV" and THEN configure the full screen resolution (1680x1050)...
Not SNAP works. Cool. But I believe these steps should be hidden from user, and
not done by him.
As a suggestion, it would be nice one could set the next boot resolution using
some command line application. So, when desktop boot out-of-sync, CAD
Handler (which works ok, presenting the text-mode screen without problems)
could show a menu option to "Change Desktop Resolution" and allow one
to select resolution, color depth and refresh rate (it could even do it changing
directly on desktop Screen object), so this problem could be solved without
the need of many reboots. Even a "Reset to 640x480x4bit" option (without driver
replacement) would be nice.
Well, I installed the system with ACPI, which forced SMP (even if my computer
is not SMP) with /APIC switch. Everything was working somewhat ok, but when I
booted without turning my scsi scanner on, the network failed. Turning the scanner
on, the network was detected ok.
Messing around with PCI I noticed they were sharing IRQs and I had this
idea: "since my mouse and keyboard are PS/2, I'll disable USB mouse and
keyboard, just as saib by Saijin Naib, and see what happens".
Well, this simple change reordered IRQ allocation and now network runs
fine with or without scanner enabled.
This is just weird, since I expected these IRQ problems would be
extinct with APIC. Whatever.. the system seems to be working now.
The only weird thing is the system sometime "hangs" on boot.
Sometimes pressing a key allow the loading to continue, sometimes not.
I created ALTF2ON.$$$ on /OS2/BOOT directory to see where the problem
happens and noticed it is not always the same time: sometimes when
loading IBM1FLPY, sometimes when loading OS2LVM, sometimes
when loading DANIATAPI, sometimes when loading OS2KBD... always
when loading base device drivers (and, AFAIK, *before* loading ACPI).
My first gess is SMP kernel not being very happy about my nForce2
chipset (running an Athlon XP 2600+). First solution: exchange SMP
kernel using ACPI configuration utility on "Post-Installation Tasks".
Bad news: it only allows me to exchange between SMP and W4...
and warns me if I exchange SMP by W4, ACPI will not be used.
Cool. I am almost doing that... but there is not another way?
UNI kernel would not be enough for ACPI?
Added options /CD and /ALS on ACPI, since it is an nVidia chipset, but
it seems this have not helped at all. Anyone knows anything I can do about this?
I am experiencing some minor problems also: WinOS2 seamless session
simply hangs (not tested full screen one, but DOS session works, even
when using ACPI!)... and ACPI nor APM is able to shutdown my system.
Original IBM APM turned it off just as expected.
Yesterday I installed Windows XP on another partition. Before that, I disabled
the second SATA HD (160GB). After Windows was working, I re-enabled the
old HD, made some backups and everything was working fine.
When I tried to boot from eCS CD again to re-enable BootManager
(which is always disabled by WinXP install), I received the old "Operating
System cannot operate your harddisk" message. I had no doubt: disabled
the second harddisk, boot eCS CD and re-enabled the Boot Manager.
After the shutdown, plugged my secondary HD SATA HD again and
eCS from harddisk did not complained about my harddisks again.
I am almost sure eCS boot CD does something nasty thing which
is preventing it from operating the secondary harddisk, be it IDE
or SATA.
I think I will fill some tickets to Serenity. I hope these problems can
be addressed before GA (although I probably will not try to install
GA over RC7 for a long time... hehehe).
Hrm, something funky indeed. I can no longer test this on my desktop, but I previously had no issue installing/booting eCS from a secondary IDE hard-drive. It was the Primary Slave, boot manager/XP were on the Primary Master. Secondary Master/Slave were DVD Read and DVD Write drives.
You shouldn't have problems with multiple SATA drives, I don't get what is going on there. By chance, do you have multiple volumes with the same LVM Letter or name? That could cause a conflict/confusion for the boot manager.
Hi djcaetano
Looks like you had a bit of fun with this install :-)
Some things you mention about ACPI may be worth a discussion.
"Well, I installed the system with ACPI, which forced SMP (even if my computer is not SMP) with /APIC switch."
Using /APIC does not force SMP mode - using /SMP forces SMP mode.
Are you sure you need either /CD or /ALS? - the 2 do not get used together - and have you run tmrtest to see if the /TMR switch is required?
As ACPI - especially versions later than 3.14 - is far from ready for use on most systems you can expect problems.
I have seen mentioned that it should be possible to use the uni kernel with ACPI but have no idea how well that would work. As I understand it the SMP kernel should work for single or multiple CPUs (or cores).
As you are using a single CPU on an older system you probably do not need to use ACPI; may reduce your problem count a bit if you sack it :-)
A thought on DOS/Win16 support.
Do you really need it?
If, like me, all you do with that is run a few DOS games then it is well worth *Not* installing DOS/Win16 support and instead use DOSBox to run your games. If you have a pressing need to run ancient win16 apps then DOSBox is a bit slow loading Windows3.1/3.11 but may still be worth consideration.
Regards
Pete
The APIC switch is supposed to be for any system that has an Advanced Programmable Interrupt Controller, and is for support of it's extended IRQ functions as well as (though not required) some things that interrelate with SMP. Newer machines, or older ones that are single CPU but SMP capable and support Intel's MP1.4 spec or a similar newer spec probably all use these.
So, as Pete pointed out, quite standard and not necessarily having to do with SMP unless some portion of the board/chipset was actually designed for SMP regardless of the number of CPUs.
Rob
Hi Pete!
Quote from: Pete on 2009.10.23, 00:29:29
Looks like you had a bit of fun with this install :-)
Yeah, you bet it. :P
Quote from: Pete on 2009.10.23, 00:29:29
Some things you mention about ACPI may be worth a discussion.
"Well, I installed the system with ACPI, which forced SMP (even if my computer is not SMP) with /APIC switch."
Using /APIC does not force SMP mode - using /SMP forces SMP mode.
I meant SMP kernel. I know /SMP switch is needed to enable SMP mode.
But the problem is SMP kernel seems to not like my motherboard. :P
It gave me problems even on OS/2 Warp 4.52... but I had been using
UNI kernel without a glitch for many months.
I don't know the difference between them (SMP x UNI), but UNI works
and SMP somewhat works. :P
Quote from: Pete on 2009.10.23, 00:29:29
Are you sure you need either /CD or /ALS? - the 2 do not get used together - and have you run tmrtest to see if the /TMR switch is required?
Well, since I was having trouble, I decided to do some tests... :)
I will run tmrtest and see what happens. Anyway, this utility comes with eCS?
Quote from: Pete on 2009.10.23, 00:29:29
As ACPI - especially versions later than 3.14 - is far from ready for use on most systems you can expect problems.
I have seen mentioned that it should be possible to use the uni kernel with ACPI but have no idea how well that would work. As I understand it the SMP kernel should work for single or multiple CPUs (or cores).
As you are using a single CPU on an older system you probably do not need to use ACPI; may reduce your problem count a bit if you sack it :-)
Well, my motherboard supports ACPI and APIC, even though it is not SMP. I just
want to make it run with "all features". :)
But maybe you are right and I should just revert to W4 kernel and strip ACPI / APIC.
But hey, can ACPI cause troubles while basic device drivers are being loaded?
I thought it was loaded a little later.
Quote from: Pete on 2009.10.23, 00:29:29
A thought on DOS/Win16 support.
Do you really need it?
If, like me, all you do with that is run a few DOS games then it is well worth *Not* installing DOS/Win16 support and instead use DOSBox to run your games. If you have a pressing need to run ancient win16 apps then DOSBox is a bit slow loading Windows3.1/3.11 but may still be worth consideration.
I indeed use Win16 sometimes, to run old software to read old documents.
A lot of PM5 and AutoCad 13 documents...
Also, I like to use DOS session to run some emulators with fast debugger;
running an emulator inside DOS Box is somewhat clumsy, I had alread
tried.
Anyway, thanks for your help!
Regards,
Caetano
Quote from: RobertM on 2009.10.23, 01:17:08
The APIC switch is supposed to be for any system that has an Advanced Programmable Interrupt Controller, and is for support of it's extended IRQ functions as well as (though not required) some things that interrelate with SMP. Newer machines, or older ones that are single CPU but SMP capable and support Intel's MP1.4 spec or a similar newer spec probably all use these.
Yes, the main reason I installed ACPI was the APIC support. My system has lots of devices, and
many potential IRQ conflicts.
Quote from: RobertM on 2009.10.23, 01:17:08
So, as Pete pointed out, quite standard and not necessarily having to do with SMP unless some portion of the board/chipset was actually designed for SMP regardless of the number of CPUs.
I dunno... But it is strange that SMP kernel give strange problems... but UNI and W4 don't.
It would be nice to be able to buy this without having a software subscription. :'( When will I - a non software subscription holder - be able to get 2.0? ???
Hi djcaetano
Starting towards the bottom of your post and working up
ACPI.PSD is the 1st driver loaded.
Your mainboard is an nForce2 chipset which may not have the best of ACPI implementations and should work fairly well - with minimal problems - using /PIC rather than /APIC if you insist on using ACPI; which is about the same as not using ACPI.
I have 1 mainboard with an nForce4 chipset and single core CPU here. It works much better without ACPI than with - no problems with USB devices, no missing features. Oh, Sorry: Herself has just reminded me that onboard sound does not work - either with or without ACPI. I suspect that is more a uniaud issue myself but as the SBLive soundcard I bunged in works fine...
tmrtest.exe is somewhere within either acpi\utils or ecs\bin - comes with the full acpi packages.
If I remember correctly the uni kernel works with the same doscall1.dll as the smp kernel - if you swap to the w4 kernel do remember to swap that dll file as well.
Regards
Pete
Quote from: djcaetano on 2009.10.23, 16:23:57
Quote from: RobertM on 2009.10.23, 01:17:08
The APIC switch is supposed to be for any system that has an Advanced Programmable Interrupt Controller, and is for support of it's extended IRQ functions as well as (though not required) some things that interrelate with SMP. Newer machines, or older ones that are single CPU but SMP capable and support Intel's MP1.4 spec or a similar newer spec probably all use these.
Yes, the main reason I installed ACPI was the APIC support. My system has lots of devices, and
many potential IRQ conflicts.
Quote from: RobertM on 2009.10.23, 01:17:08
So, as Pete pointed out, quite standard and not necessarily having to do with SMP unless some portion of the board/chipset was actually designed for SMP regardless of the number of CPUs.
I dunno... But it is strange that SMP kernel give strange problems... but UNI and W4 don't.
I am wondering if that is indicative of mis-implemented ACPI support on the board (ie: the board's chipsets/BIOS are not properly using it based off the combination of dual or single core CPUs). By using UNI or W4's UNI, it would probably be preventing/not using the SMP functions related to the /APIC mode. I suspect that if such is the case, then Microsoft and others may not be following the ACPI's /APIC utilization for certain boards with incorrect implementations - much like how Dani had to scan for certain chipsets in her driver and selectively not use certain functions that were incorrectly supported on various chipsets/motherboards.
Kinda like a bunch of "If this crappy implementation, then dont use these APIC functions even though chipset reports supporting them"
Quote from: RobertM on 2009.10.23, 23:51:59
I am wondering if that is indicative of mis-implemented ACPI support on the board (ie: the board's chipsets/BIOS are not properly using it based off the combination of dual or single core CPUs). By using UNI or W4's UNI, it would probably be preventing/not using the SMP functions related to the /APIC mode. I suspect that if such is the case, then Microsoft and others may not be following the ACPI's /APIC utilization for certain boards with incorrect implementations - much like how Dani had to scan for certain chipsets in her driver and selectively not use certain functions that were incorrectly supported on various chipsets/motherboards.
Kinda like a bunch of "If this crappy implementation, then dont use these APIC functions even though chipset reports supporting them"
Probably, Robert. Maybe even a bad programmed ACPI support... Without ACPI I had to
*disable* ACPI on BIOS, otherwise the system would hang during PCIBUS.SNP.
Anyway, I tried the following combination (never with /SMP on ACPI, for obvious
reasons):
ACPI | Kernel | Result |
YES | SMP | Random hangs at boot time |
YES | UNI | Random hangs at boot time |
NO | SMP | Random hangs at boot time |
NO | UNI | Boot always ok (sometimes it takes some more time, but always boot) |
Two remarks:
1) UNI doesn't work with DOSCALL1.DLL from SMP. It requires the
same DOSCALL1.DLL as W4 (I tested UNI kernel with SMP DOSCALL and
received a lot of errors during boot and a BSOD on the end)
2) GenMac is not very trustworthy. Sometimes it detected my
gigabit ethernet (marvell yukon 88E0001 or something like it), but
more usually it doesn't. I replaced GenMac by the native driver which
*seems* to be detecting the ethernet every boot.
Results: nForce2 chipset doesn't like ACPI, doesn't like APIC and
doesn't like SMP kernel.
Quote from: djcaetano on 2009.10.27, 01:15:02
Results: nForce2 chipset doesn't like ACPI, doesn't like APIC and
doesn't like SMP kernel.
Which is probably why it has it's own custom chipset drivers for Windows.
Blame Nvidia for that one :\
There is a reason why I, even as a gamer, will never use an Nvidia chipset-based board. They've gotten much more responsible now but I'll stick with VIA or Intel.
Quote from: djcaetano on 2009.10.27, 01:15:02
2) GenMac is not very trustworthy. Sometimes it detected my
gigabit ethernet (marvell yukon 88E0001 or something like it), but
more usually it doesn't. I replaced GenMac by the native driver which
*seems* to be detecting the ethernet every boot.
I have to apologize with GenMac developers. :)
Even the native driver is not working ok at all times. In fact,
I installed GenMac for nForce2 Ethernet and the result was also
bizarre: sometimes BOTH Ethernet are detected (Marvell and
nForce2) and sometimes NONE of them are detected... it
doesn't matter what drivers are used (both GenMac or one
with GenMac and another native).
And most bizarre: even when the driver cannot detect the
devices, they ARE detected by PCI.EXE.
This is the most annoying problem with this eCS 2.0RC7
and I am not being able to solve it. I am using UNI kernel
and no ACPI. This problem bugged me even with ACPI
and with SMP kernel... but I NEVER suffered from this
with the previous install (OS/2 4.52).
PS: WinOS2 *is* working. The bug is on eCS installer, which
doesn't add VMOUSE.SYS device line on Config.SYS. It
was just a matter of adding this line and WinOS2 is now
working, be it on Seamless of fullscreen mode.
Quote from: djcaetano on 2009.10.22, 19:52:08
This week I bought a new SATA 250GB harddrive and was able to install eCS2.0 RC7
after removing the IDE harddrive. But the install process was not without some sort of pain.
The reason for the pain was the fact eCS is set up to boot at 800x600 and since SNAP
seems to not recognize my monitor on the first time it runs, 800x600 with some sort of
unsupported refresh rate is selected... causing an annoying "No Signal" message ("No
Sync").
Oddly enough, this problem *still* happens when installing eCS 2.1.
Hi djcaetano
Yes, I saw that "unsupported mode" message on a monitor here when installing eCs2.1 using SNAP - the monitor wants 800 * 600 @ 75Hz for whatever reason and cannot use 800 * 600 @ 60Hz; maybe because it was designed primarily as a tv monitor, maybe because the computer to monitor connection is Analogue.
I worked around this by using Panorama during the installation but selected to install SNAP for use knowing that I could add the necessary resolution and refresh mode later using the SNAP gamode.exe utility - which worked out fine.
It would be nice if eCS updated the installer to check what the monitor is capable of rather than simply expecting all monitors to be 800 * 600 @ 60Hz capable.
Regards
Pete