OSDev.org

Well, this is more a sign-of-life message than anything else.

The os has got some debug sessions. I've found some bugs in the shared memory management (oh, there are still some which I need to squash). The GUI environment runs much stabler now. No more crashing after opening and closing a bunch of windows in succession.

In the login screen, just click on a user and then on login. root is preferred. Then you can go on and perform happy playing around. There have been more changes under the hood, so you won't see anything new.

If you want to see BlueIllusion crap out and puke the hell out on the screen, just start the clock application, and then either do naught or move around the mouse. It will go nasty. Well. I'm gonna work this out.

BlueIllusion is known to work in vmware server. I don't know about real machines. It might work. It might as well die. in a cruel way.

For the near future I'm intending to implement a rpc/message feature which runs over tcp/ip sockets and feases multicast of messages by the means of an observer pattern implementation. It's gonna work like a bus system.

stay safe.

Testing in QEMU, it takes quite a while to get over "vm86" and it found ata drive with 64k cylinders/heads/sectors (I configure none) then hang for a while, and couldn't see what it said after that, but I got a red screen, with black box in the middle, and when clicked on the window, the black box disappeared.

Then it seems to take 100% CPU (well at least QEMU is taking all CPU) but it doesn't seem to be progressing much... and then I tried clicking around and hitting keys and moving mouse and wheel, and now I notice it doesn't use any CPU anymore, but it isn't doing anything either.. Just a red screen.

Ok I tried again with an (all zero) HD image of 20MB, and now it gets over the ata detect properly.. still sits at vm86 for a while, and again gives me red screen with a black box, and the black box disappears as soon as I touch keyboard or mouse.

Oh ok, switching QEMU to use the Bochs VBE BIOS thingie standard VGA instead of cirrus gives me a proper login box now, with what looks like two mouse cursors (assuming mouse cursors are the crosshair thingies) and neither of which seems to move or anything... hmmh.. maybe I'm just not patient enough, as it seems to run with 100% CPU... I'll wait a bit, let's see what happens.. Okey, after having it run for 6 minutes worth CPU time (in QEMU on an AMD64 3k+ box, in 32-bit mode though) it still doesn't seem to do anything beyond running at full CPU...

I see a login box, and two of what looks like a cursor of some sort, but that's it. Nothing moves on the screen. Nothing seems to happen whether I hit keys or move/click mouse.

Unfortunately I can't test on real machine right now, because I don't have any empty CDs..

He better not have just made one big boot loader.

@mystran: thanks for testing!

The thing is, that my drawing primitives only work for 32 bpp and that I lack the time to incorporate 15 resp. 16 bpp drawing stuff again. I've had it working ere I've overhauled the whole gui stuff to work with shared memory.

That's why it dies in that horrible way. The Two Login Boxes should be one - that's the drawing. Have to fix it, I see. This ain't be anywhere in the near future. Studying is eating all the time. *gg* Ah, but it's fun.

Well ... I have to admit, that I have ceased testing with qemu and bochs for these two don't give the performance which vmware shells out with ease. So, I only test with vmware (the free server edition, which is also available for LInux)

@kevin McGuire: You can bet, that there is no single big bootloader. My os is microkernel based, and you can easily derive from the screenshots at my website, that it works well for that little of time which I can muster to do os deving and debugging. If you still haven't got the knack, here is a Hint: grub has it's knobby little fingers in the game.

It's not even booting, I've set it as image file to boot with in VMware Server but it says "Operating System not found"? I've checked the image with WinImage and it has a bootsector etc. in the floppy emulation image.

EDIT:
Yes it works fine now

EDIT2: Oh no For me it hangs when I click LOGIN after selecting the user 'root'.

@bughunter: Thank's a bunch! I've checked the issue by downloading the image - for me it's continuing happily after clicking login.

Things you can try: reboot the whole shebang with the vmware-reset button. Increasing the Memory assigned to the virtual machine. I'm used to give it 256 mb.- it has already 256 mb of framebuffer memory by default due to my big big graphics adapter. I know that it sometimes - very seldom - has issues with apparently not zeroed out memory. Should do this properly.

@mystran: how's the going with your networking stack?

I don't mean to be rude distantvoices, but perhaps you should try testing your OS in "Other" emulators and real systems?

It appears your OS has a lot of kinks to work out, and from the looks of it.. You have some sort of Vendor lock-in thing going on with VMware.

You've probably been using it so much you've become reliant on it's quicks and defects ..

Still I think it's a great project, With a extended "testing" range more people might get interested in the development & testing of your OS.

As you like saying.. Stay safe

@brynet: Sure. The lack of taking care of it shows up on occasion, I can't deny that. But vendor-lock ... aaach, vmware-server being free I'd be haunted by plague if I didn't use that for it is way more performant than f. ex. Bochs.

as for the real hardware -> you 're perfectly right, I haven't tested it on real hardware for long time now. Should do that to check out the weird stuff it is doing to a real cpu. *fg*

Just be patient with me. I'm more busy with studying than with osdeving these days, and this will last for 2 years - until I've got that masters degree.

distantvoices wrote:@brynet: Sure. The lack of taking care of it shows up on occasion, I can't deny that. But vendor-lock ... aaach, vmware-server being free I'd be haunted by plague if I didn't use that for it is way more performant than f. ex. Bochs.

It only works under one emulator, How isn't that Vendor lock-in?

distantvoices wrote:as for the real hardware -> you 're perfectly right, I haven't tested it on real hardware for long time now. Should do that to check out the weird stuff it is doing to a real cpu. *fg*

Cool

distantvoices wrote:Just be patient with me. I'm more busy with studying than with osdeving these days, and this will last for 2 years - until I've got that masters degree.

I was unaware of your education status, Apologies..

To quote Einstein though: "The only thing that interferes with my learning is my education."

distantvoices wrote: @mystran: how's the going with your networking stack?

Ehm.. well.. it answers ICMP echo.

Seriously, I've not been working on it, as I've been modifying my VFS to work asynchronously, one layer at a time, and having it run at all time time, I end up with lots of temporary code that's just there for the compatibility, but it's progressing, and I've been able to keep it stable. I've also been working on general improvements to kernel:

- my malloc now performs well enough to run GVim with LD_PRELOAD, without any perceivable slowdowns, which is cool

- interrupt management is pretty much worked again, with all kinds of cleanup still going on.. first I added proper handlers (with sharing and all), then I added generalized bottom halves (which is what the new asynchronous VFS relies on, and I got rid of a killer task by running the thread-teardown in a deferred procedure after reschedule to some other thread)

- after interrupt rework, I also made all threads fault to interrupts enabled, which required some fixing here and there, and some changes to rules on what context (normal vs. deferred vs. interrupt) you can call what functions, and should you have interrupts enabled or disable while doing so (typically either interrupts required to be enabled, or don't care, though some parts rely on they being disabled)

I guess I'll get back to networking after I've rewritten the top-VFS layer and the FAT driver to work asynchronously.


As for VMWare... hmmh... QEMU does perform pretty okayish.. even without KQEMU..

distantvoices wrote:@bughunter: Thank's a bunch! I've checked the issue by downloading the image - for me it's continuing happily after clicking login.

Things you can try: reboot the whole shebang with the vmware-reset button. Increasing the Memory assigned to the virtual machine. I'm used to give it 256 mb.- it has already 256 mb of framebuffer memory by default due to my big big graphics adapter. I know that it sometimes - very seldom - has issues with apparently not zeroed out memory. Should do this properly.

Yep it works with 256 MB. However when logged in, VMware consumes about 99% of the CPU time.