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.


Messages - Valery Sedletski

Pages: 1 ... 16 17 [18] 19 20 ... 24
256
Applications / Re: VirtualBox problems
« on: April 09, 2018, 09:33:27 pm »
2Mark Szkolnicki:

> First in regard to one of your previous questions the 1018 error could be "-1018" (See attached JPG). When I saw a "dash" after "Unknown error", I normally assume in english it was a pause between two phrases, not a "minus" mark.

So yes, it is -1018, which means VERR_INVALID_CPU_ID, that is what I suggested. Probably, it will be some workaround for that in the future, but now only this way, until your LAPIC ID's are consecutive.

> VIRTUALADDRESSLIMIT=2560

It's too little! I have VIRTUALADDRESSLIMIT=3072 on all my three machines, and everything is working. Please try to change VIRTUALADDRESSLIMIT to 3072, and THREADS to 511, PROCESSES to 128 explicitly. Also, it is desirable to change JFS cache to some lesser explicit value, like:

VIRTUALADDRESSLIMIT=3072
THREADS=511
PROCESSES=156
ifs=d:\os2\jfs.ifs /autocheck:* /cache:128000 /lazy:8,60,5 /p1

-- Here, JFS cache is limited to 128 MB, threads is limited to 511 threads in the system, PROCESSES are limited to 156 processes max. THREADS and PROCESSES eat up some structures per each thread/process, and they consume some hundreds of MB in the system arena, which could be too much. JFS cache is limited to 20% of available RAM, which could be too much too.  (20% from 3 GB is 614 MB, which is a large % from 1 GB of kernel memory, if VIRTUALADDRESSLIMIT=3072, so JFS cache should be decreased to say, 128 MB. The exact number should be determined in Theseus). THREADS/PROCESSES/JFS cache consume memory above VIRTUALADDRESS limit (kernel memory), while VBox consumes memory below it. So, if you have too little VIRTUALADDRESSLIMIT, it can be not sufficient for VBOX. If you have VIRTUALADDRESSLIMIT near to 3072, then only 1024 MB above it is available for kernel and drivers. So, if THREADS/PROCESSES/JFS cache are not limited explicitly, you can easily run out of kernel memory.

PS: How much memory you assigned to your VM? 330 MB?

257
Applications / Re: VirtualBox problems
« on: April 03, 2018, 08:17:34 pm »
2 Mark Szkolnicki:

Quote
As mentioned last night, I was able to get a VBOX VM to open consistently after a 15 second delay after the program initially opens, or after closing the 1018 error windows. However, trying it this morning, I was having the same difficulties, even after a 15 second delay, until I realized I had Seamonkey already open in the background (Seamonkey 2.35).

After closing it, I was able to start a VM, and, keeping VBOX running, to download some Windows XP update files, and then reopen and use Seamonkey normally.

However, closing Seamonkey and restarting a second time with VBOX running and reopen Seamonkey it (Seamonkey) would not open until i shut down VBOX

So yes, these are some memory-related problems. Could you, please, show me your config.sys? Also, how many physical RAM is available to OS/2 on your machine? What does XCenter "sentinel memory monitor" show?

258
Applications / Re: VirtualBox problems
« on: April 03, 2018, 07:26:30 pm »
2Dave Yeo:

Thanks for the log! So, it works with SET LIBCX_HIGHMEM=2 for you? That's what I talked about. Leave it permanently in your config.sys then (or only in VBox environment). Unfortunately, it works fine on my machine, so I don't experience this, so I cannot play with it and see potential workarounds. VirtualBox.exe starts fine on my machine without SET LIBCX_HIGHMEM=1. Only VBoxSDL.exe crashes without it. So, I'd recommend for now to leave SET LIBCX_HIGHMEM=2 in CONFIG.SYS permanently.

Also, there can be problems with insufficient address space on your machine. How large is VIRTUALADDRESSLIMIT on your machine? Did you tried to load VBox DLL's high? There is highmem.cmd for that purpose (you need to have highmem.exe on PATH).

PS Unfortunantely, these problems are likely to happen on each memory-hungry program like VBox, Mozilla, Qt or OpenOffice. OS/2 is a 32-bit OS. Additionally, it has special address space layout with Private/Shared/High Private/High Shared arenas, limiting each other. So, you can easily run out of address space, so no wonder.

259
Applications / Re: VirtualBox problems
« on: April 03, 2018, 10:20:59 am »
@Mark Szkolnicki:

> error 1018

%define VERR_INVALID_CPU_ID    (-1018)

Are you sure it is 1018, not -1018?

This means that current CPU ID is invalid. Unfortunately, there is no reliable way to determine the current CPU ID for ring0 on IBM's kernels. Currently, LAPIC ID's are used to emulate getting the CPU ID, but they are not guaranteed to be consecutive, which is assumed. If your current CPU number gets into the "hole" between valid CPU numbers, this happens. So, you're unlucky. The only way to avoid this error is currently to use OS/4 kernel. Unlike IBM kernels, it has a consistent way of getting CPU ID's which are guaranteed to be consecutive (which is assumed in VBox). So, you can try using the OS/4 kernel (if you're experienced enough)

2Dave Yeo: Do you have SET NSPR_OS2_NO_HIRES_TIMER=1 in your config.sys? If so, could you repeat the test with it disabled (just set it to an empty line in OS/2 window before running the test)?

260
Applications / Re: VirtualBox problems
« on: April 03, 2018, 10:09:02 am »
2Mark Szkolnicki:

> Result Code: NS_ERROR_FAILURE (0x80004005)
> Component: MachineWrap
> Interface: IMachine {f30138d4-e5ea-4b3a-8858-a059de4c93fd}

As I said above to Dave, you either have shared memory problems, or have some required DLL's missing. As you said that you sometimes started the VM sucvcessfully, I suspect that you have (shared) memory errors.

Quote
The log generated at the time in the VirtualBox \BIN directory is as Dave also describes (mine was slightly different, as sent to you previously):

Log created: 2018-04-03T00:38:20.588000000Z
Executable: K:\VPC\VIRTUALBOX.EXE
K:\VPC\VirtualBox.exe kLIBC á--comment ámint á--startvm á43cabf9e-2003-43ea-a9e0-db4fa1156a6c á--no-startvm-errormsgbox

This log is not relevant. I asked about the VM log.

> At least for me, I can get the VM's to start up more consistently with LIBCX 0.6.0 installed than under the experimental LIBCX 0.6.2, but that may just be my experience, not others.

Yes, so use libcx 0.6.0, or, better, add LIBCX_HIGHMEM=2 to your environment, if you use libcx 0.6.2

> In regards to your other question, when the Windows XP VM has opened, when the AMD-V was enabled, rendering of the desktop was exceptionally slow (sometimes minutes for the desktop icons to appear) and clicking on folders to open also took sometimes minutes. After disabling that feature, the desktop seemed to open normally, and opening folders or other icons seemed to be a lot faster (seconds not minutes).

This has nothing to do with performance. You should just disable AMD-V/VT-x on your machine.

> Just an observation and first impression, Sir. If we can get some way to consistently start VM sessions, I will try boots with the setting enabled and then disabled, to see if what I am observing is correct or not.

As I said above, fix the (shared) memory problems on your machine. Boot with which setting disabled/enabled? Did I asked you to disable/enable any options?

261
Applications / Re: VirtualBox problems
« on: April 03, 2018, 09:52:32 am »
What I get on the first try is
Quote
Failed to open a session for the virtual machine mint.

The virtual machine 'mint' has terminated unexpectedly during startup with exit code 1 (0x1).

Result Code: NS_ERROR_FAILURE (0x80004005)
Component: MachineWrap
Interface: IMachine {f30138d4-e5ea-4b3a-8858-a059de4c93fd}


This means that either you have some required DLL's missing, or incorrect version, or having high memory problems. This is not a problem with VirtusalBox itself. Did you tried LIBCX_HIGHMEM=2 before starting VBox? I don't need this on my machine, I only need this to start VBoxSDL successfully, otherwise SDL crashes. Yes, some high memory related problems, not VBox-related.

Waiting a bit after this allows the virtual machine to start.
The log seems to use a different charset,

Quote
Log created: 2018-04-03T00:38:20.588000000Z
Executable: K:\VPC\VIRTUALBOX.EXE
K:\VPC\VirtualBox.exe kLIBC á--comment ámint á--startvm á43cabf9e-2003-43ea-a9e0-db4fa1156a6c á--no-startvm-errormsgbox

Yes, there are some signs (probably, quotes, probably, in unicode) in command lines which VBox uses to start its processes. That's the way it works.

262
Applications / Re: VirtualBox problems
« on: April 03, 2018, 02:05:15 am »
2Dave Yeo: Thanks!

2Mark Szkolnicki:

> The Logs sent were before I switched off the AMD-VT. Now done. I did not see another log generated in the \LOG subdirectory but have seen a marked improvement in speed when I can get the WindowsXP VM to start.

Improvement in speed? Compared to what? AMD-V is disabled, so what improvement can be?

> So far, the start-up is still sporadic, but I have a couple of items I am looking into.

Sporadic? What do you mean? It starts only sometimes? What error? Any logs?

> Running a test start from an ArcaOS window session the VBOX window came up, but I noted no panic screen. Perhaps I'm misunderstanding what was being requested Valery if you could provide an example of what a panic screen should look like i will see if one comes up

I don't expect any panic screen. No panic screen, according to your log.

So, what do you mean by "sporadic" startup?

263
Applications / Re: VirtualBox problems
« on: April 02, 2018, 11:44:30 pm »
2Mark Szkolnicki:

Did you switched AMD-V/VT-x off? According to your log, it is enabled and it failed to initialize. (I see no Guru meditation, though, so this is not that case). That is what I suggested you above: switch off AMD-V/VT-x. Why did you enabled it? It was never supported by OS/2 port of VBox. Only last builds contain some support for that, and only on OS/4 kernels!

> In regard to "Guru Meditation" - from your description I assume that it would be seen if you attempt to start VirtualBox.exe from an OS/2 window, and the program will not start at all. As that has not been the problem I have been experiencing, I'm not sure if running that test would give you any added information, but could try it if you wish.

The program will wtart, but itt will show a panic screen. The test has no relation with "Guru meditation" error, I only need to know if they fixed a hi-res timer error in your version of ACPI.PSD. As you're the first user I know having this version.

> There are 4 VBox.log files in the Windows XP \LOG subdirectory which have been generated and which I've attached below.

Thanks, but I only need the last one.

> * timerTest.html (6.44 kB - downloaded 6 times.)

This is some trash with javascript code (look into the file you attached). Please attach the text file with the test output

264
Applications / Re: VirtualBox problems
« on: March 31, 2018, 01:37:06 pm »
2Mark Szkolnicki:

> Understood that 5.0.51 is a developmental release Valery.

Yes, it's a development version, but not less stable than 5.0.6. I planned to update the OS/2 VBox port to a newer version. 5.0.51 is not a final one, but some intermediate version. I'll probably try to update it to more newer version.

> I actually saw the notice about v5.0.51 that Martin posted here. (He should be a reporter, as he has a nose for news <smile>)

Yes, it's possibly, Martin is guilty for a leakage ;) But anyway it's good as it could gain more attention to the project, so it is ok.

> I am testing with ArcaOS v5.0.1 installed on this machine, with the kernel updated to 14.201_SMP. eCS v2.2 is also installed on this machine on a separate partition, but has not been used for the testing, as of yet. ACPI is at the latest released level (3.23.08) on both partitions.

AMD-V/VT-x is not supported on IBM kernels so far. You need an OS/4 kernel to use HW virtualization in VBox. PSD can be any, but note that newer ACPI.PSD are incompatible with OS/4 kernels (AN just disables all CPU cores. Any hacking with it won't help, so OS/4 devs gave up on patching ACPI.PSD for working on OS/4 kernel). So, you'll need os2apic.psd or os4apic.psd, or earlier versions of ACPI.PSD for use with OS/4 kernel.

Also, as you have "brand new" 3.23.08 ACPI.PSD, could you please run this test: http://gus.biysk.ru/os4/test/timerTest.html (Please, redirect the output to file, we need about 200 iterations of this test. This is a high resolution timer test. It is interesting to know if they fixed a hi-res timer bug in 3.23.08. It is still present in 3.23.07.)

> In regards to "Guru Mediation" screen, I have no idea as to what you are referring to. I have included the screen that comes up with the Windows XP information presented. If I have some idea as to what your referring to, and how to get it, I will include that as well.

"Guru meditation" is a VBox panic error. It should write it to the console (if you start VBox from OS/2 window), and in VBox VM log. VBox stores the VM log in Logs\VBox.log in each VM subdirectory. Yes, as far as I can see, you have AMD-V/VT-x enabled in your VM. Why did you so? It is disabled by default, and displays a warning in case you enabled it and your system lacks required API's and features. So, disable it, and it should work.

> BTW AMD-V/VT-X is indicated as "enabled" by default (see attached). However, the "invalid setting detected" message at the bottom when moused over, indicates that the current kernel doesn't allow enablement of hardware virtualization.

Yes, that's what I meant above. VBox warns you that your configuration is incorrect. So, disable AMD-V/VT-x checkbox and it should work.

2Pete:

>>... The link just "leaked" from the bug tracker when I gave it to Pete for testing.

> Guilty as charged, my Apologies.

Yes, I did not planned to distribute this version, but no problem ;)

265
Web applications / Re: Chrome...
« on: March 30, 2018, 04:47:15 pm »
This is not QEMU :) Though, there is QEMU 0.11.0 ported to OS/2 by Sawataishi Jun from Japan in 2009. It works fine on my machine, there are i386 and x86_64 versions available. I have it on my ftp: ftp://osfree.org/upload/bauxite.sakura.ne.jp/tmp/os2/. Unfortunately, original site is down.

266
Applications / Re: VirtualBox problems
« on: March 30, 2018, 07:49:37 am »
2Mark Szkolnicki:

> Both VM's are recognized by VirtualBox, but when I attempt to boot them, more times than not I get a message "Unknown error 1018 - during initialization of the runtime" (JPG attached). The detail screen comes up with a message that "The Virtual Machine WindowsXP has terminated unexpectedly during start-up with an exit code 1 (0x1) in Component "MachineWrap" (JPG attached).

Do you have AMD-V/VT-x enabled? What kernel are you using? IBM's or OS/4? Which PSD? (acpi.psd/os2apic.psd/os4apic.psd)? Did you saw a "Guru Meditation" screen? Can you show me VM log and the VM config?

PS: No special additions version is required for 5.0.51. 5.0.51 is just a bit newer version from 5.0.6 -- no big difference, so the same 5.0.x additions should work well. Additions don't enhance performance significantly. They are for user experience with the VM to be more comfortable, e.g., shared folders, clipboard sharing, mouse pointer integration etc.

PPS: 5.0.51 is a development version, not a release. That's because it did not have special additions version. Also, I did not "released" it officially. The link just "leaked" from the bug tracker when I gave it to Pete for testing.

267
Applications / Re: VirtualBox problems
« on: March 24, 2018, 06:58:08 am »
OK. LIBCX_HIGHMEM=2 is not needed for VirtualBox.exe. VirtualBox.exe should run fine without that. It only required for VBoxSDL, as there is some high memory-related problem with SDL.

268
Applications / Re: VirtualBox problems
« on: March 23, 2018, 08:54:21 am »
2Pete:

Quote
I have not really used either of my eCS installations since installing ArcaOS but decided to try booting eCS2.1 with various ArcaOS Driver Pack updates. Boot stops with a SYS1201: vboxdrv.sys not installed.

The same problem with eCS2.2b2 - which surprised me as ecS2.2b2 is the system that was in use when I first installed and used VirtualBox successfully.

Both eCS installations are using acpi 3.23.05 and os2krnl 14_106 SMP

Hm, VBoxDrv.sys doesn't install? I can't imagine why this can happen. Tried many kernels and PSD drivers -- never seen this. Could
you try to help me to track why this happen? So, this happens with 14.106_SMP and ACPI.PSD 3.23.05? I tested this combination, and everything booted well. I tested it in MCP2 and eCS 2.2 beta II, on a single-core Athlon 64 machine and two Core2Duo machines -- all worked well. Can you give me more details? Did you installed any other updates except the kernel and ACPI.PSD? Are you sure it was ACPI 3.23.05? Maybe, it was some other version? (also, it's possible that AN has different builds having the same version number, which behave differently. We can exchange more bldlevel details. For example, the copy of ACPI.PSD 3.23.05 I have is built 1 Feb 2017 12:52:11. Does your copy differ?)

Quote
Failed to create the VirtualBoxClient COM object.
 
   The application will now terminate.


   Callee RC: NS_ERROR_ABORT (0x80004004)

This happens when VBox cannot load VBoxRT.dll or some other DLL. This is usually because some dependent DLL's installed via rpm packages are missing, or are different version. You could try to update, or better, remove the DLL's listed here: http://trac.netlabs.org/vbox/ and install them over. This is basically the same problem as with newer versions of Mozilla, when some DLL's XUL.DLL depends on, are missing or incorrect version.

2Doug Bissett:

Quote
I run the old VBox 1.6.1, and the beta 5.0.6, under ArcaOS. No unexpected problems. The won't run OS/2, because you need VT-x/AMD-V enabled, and the OS/2 versions of VBox can't do that (be sure that the option is turned off). You can run a 64 bit OS, as a guest, but it will be dead slow, so don't bother even trying (like 12 hours to boot win 7 64 bit).

OS/2 versions of VBox since about r147 support hardware assisted virtualization. I was able to run OS/2 guest on OS/2 host successfully. 64-bit OS'es currently don't run ok. Most probably, you will get a hang. But it is still being worked on. Though, because of limitations of IBM kernels, this is currently supported on OS/4 kernels, only. They have some required mechanisms like API's for running some code on all CPU's in parallel, getting the current CPU number etc.

How did you managed to run a 64-bit OS? On which VBox version? I was unable to run a 64 bit OS neither in RAW mode, nor with HW virtualization.

2Dave Yeo: This error does nothing with NSPR. It is just a failed COM object initialization because it cannot load a DLL. NS_ERROR_ABORT seems to be an XPCOM error, not NSPR. Also, VBox doesn't use NSPR from yum repo. It carries its own version of NSPR based on VBox Runtime (IPRT, Innotek Portable RunTime) with itself.

Quote
BTW, about your ifx error. I saw dmt (the eCS version of ifx) hangs in the past with eCS but none with ArcaOS. Doug might have a point in the ordering in config.sys.

I never seen this in my eCS with DMT. Also, I have no idea how IFX/DMT can hang, or not, depending on VBoxDrv position in config.sys. I suspect this is some SMP-related problem, though.

2Pete*2:

Quote
The resolution to the "timing problem" with the vboxdrv.sys was to move the line higher up the config.sys file - no problem booting now, no need to use Shift as IFX is starting.

Probably, this is indeed some timing problem, so before VBoxDrv "settles a bit", i.e., finishes some activity, and activity done by IFX, should pass some time. An idea may be that VBoxDrv and IFX do something at the same time, but on different CPU's. This somehow leads to a hang. Currently SMP support in VBoxDrv is limited on IBM kernels. IBM kernels have no API returning the current CPU number. So, this is currently emulated with finding the current local APIC number. But LAPIC numbers are not guaraneteed to be consecutive. The may be "holes" between valid LAPIC numbers. So, if an attempt to use an invalid CPU number from the hole in the range 0..MaxCPUNumber is made, an unexpected result can happen, e. g., a hang. So, this currently works reliably only on OS/4 kernels, where an API exist returning a correct CPU number from a contiguous range. VBox assumes that the range 0..MaxCPUNumber should be contiguous.

PS: I never sort CONFIG.SYS automatically, because I prefer to edit it manually. So, I have it being sorted out logically, grouping related statements together. This is easier to support, with a little performance impact. I always placed the VBoxDrv.sys loading string at the end of CONFIG.SYS.

269
General Discussion / Re: 2ine - OS/2 on Linux
« on: February 24, 2018, 10:50:29 am »
2Sergey Posokhov: Hmm, where did you took this from? Man tries to write some funny things for fun, and wrote about that in his blog. A masterpiece? Do you mean resume? I don't think he will write about that in his resume. He seems to be experient in game development, mostly based on SDL. So, he took yet another toy project just for fun, just like any retro game simulator. He achieved a good progress with it. He has some good progress, compared to osFree. But looking at his sources at a glance, I now see that he did not advanced significantly, compared to osFree. He has a number of DOSCALLS functions implemented. Some of them are missing in osFree, but some functions we implemented are missing in 2ine. Implemented TCPIP32 API's are just small wrappers around Linux socket functions. VIOCALLS are simply implemented based on NCURSES, so no proper VIO lattice is implemented, yet. PMWIN functions include almost the same set as implemented by FreePM. FreePM has only WinBeginPaint and WinEndPaint missing. Drawing is implemented directly via SDL. No GPI is implemented, yet. Also, on FreePM only a message loop is working, but it does not draw anything on screen. 2ine draws a rectangle on screen, instead of a window. The same simpliest application, creating HAB, message queue, a window and running a message loop. DosExecPgm is implemented via fork(). So, these are only simpliest API's implemented, to almost the same degree as on osFree. The only significant advancement is thunking and 16-bit API wrappers. Also, he successfully runs EMX runtime, and even EMX GCC. osFree can't do that, because we still have no 16-bit functions implemented. But looking how it's implemented in 2ine, we could try implementing the them similar way on L4. Some ideas are very useful to get from 2ine sources. So, ATM these sources can't be directly reused by project like osFree. It directly uses some system functions, and required abstraction layers are missing yet. We hope that it will change in the future. Also, looking at last paragraph of his blog post, I see that he paused the work for another project. It looks like implementing PM API is too complex task for quick progress. But I hope that he will continue the development, so maybe the project will advance much further, who knows. I hope him good luck.

PS: His LX loader looks much cleaner that our LX loader written by Viking. Maybe, we can use it, instead of Viking's one.

270
General Discussion / Re: 2ine - OS/2 on Linux
« on: February 23, 2018, 09:18:41 pm »
2 Sergey Posokhov: You seems to confuse something. https://rv8.io/ looks to be like a RISC-V software emulator. RISC-V is a new open source hardware architecture, not a bytecode. There are some development boards manufactured based on this architecture. It is intended as a future opensource replacement for Intel/AMD arch's. At least, I know it being supported by Genode, which is a project I follow too. Note that RISC-V is not yet complete. Genode supports compiling RISC-V binaries for this arch, and running them in another emulator I know, called Spike: https://github.com/riscv/riscv-isa-sim/. https://rv8.io/ is just another RISC-V emulator. I see nothing wrong with it. What is a symptome of illness of software industry?

Also, an author of https://rv8.io/ is Michael J. Clark, whereas 2ine is written by Ryan C. Gordon. I see nothing common between them.

I am a bit sceptical about the idea to run OS/2 binaries on Linux too. But, for example, ReactOS can benefit from WINE. ReactOS has its own kernel, but the user mode DLL's are mostly shared with WINE (though, they need to add some manual modifications to WINE code for porting them to ReactOS. Also, they have an alternative win32 subsystem implementation, called arwinss, which can reuse WINE libraries almost as is).

We also can benefit from 2ine, like ReactOS does from WINE. We can port some code to osFree, and back, so both projects can benefit. ReactOS does contributions back to WINE code, and we can. Also, some ideas exchange would not hurt.

Pages: 1 ... 16 17 [18] 19 20 ... 24