Dynamo Kernel 0.2.1 (alpha)

senaus · Post by **senaus** » Thu Nov 23, 2006 12:34 pm

Hi, I'm looking for people to test the latest version of my kernel. See this thread for more details:
http://www.osdev.org/phpBB2/viewtopic.php?t=12209

Thanks!
The Senaus

Brynet-Inc · Post by **Brynet-Inc** » Thu Nov 23, 2006 3:52 pm

Hello, I tested your OS in QEMU(Running on OpenBSD) and it works, It looks pretty neat.

Nice work..

Combuster · Post by **Combuster** » Thu Nov 23, 2006 5:38 pm

Works on bochs, breaks on MS VirtualPC 2004 with a pagefault

Screendump:

Code: Select all

[INFO]mm: Start of heap: 0x0
[INFO]mm: Pages in pool: 64
[INFO]mm: Initializing Physical Master Array...
[DEBUG]mm: Page Manager Initialized
[DEBUG]mm: Start of pool: 0x263080
[INFO]mm-arch: Initializing...
[DEBUG]mm: __kScanPhysicalMaster(16, 240, 1)
[DEBUG]mm: __kInitKernelPage(0x122500) pa=0x20102
[DEBUG]mm-arch: gKernelPageDirectory=0x20000
[DEBUG]mm: __kScanPhysicalMaster(16, 240, 1)
[DEBUG]mm: __kInitKernelPage(0x122500) pa=0x20102
[DEBUG]mm-arch: pt=0x20000
    - Paging Enabled
    - Virtual Master Array at 0x11f7c0
    - Virtual Master Array has 256 entries
    - Freed 260443 kernel pages
[DEBUG]mm: Kernel VM allocator initialized
[INFO]proc: Initializing...
TSS LOADED
[DEBUG]proc: Creating Threads

FATAL ERROR:
Page Fault

Press any key for debug console...

I'll be trying it on a real machine shortly to see wether it was some glitch in the ms stuff...

[edit]Real machines behave worse: it apparently prints some messages to the screen before resetting (dont ask me what it says, the screen goes blank before you can actually read anything)
Test box: Callisto[/edit]
[edit2]Io also errors, this time with a "Thread killed - Invalid memory address" message. since qemu and bochs work my first guess is that you didn't zero memory (consistently)[/edit]

senaus · Post by **senaus** » Mon Nov 27, 2006 7:18 am

Thanks for trying it out guys, I appreciate your help. I've now discovered a major bug in parsing the BIOS memory map, whereby all non-reserved sections are marked as free, regardless of the total RAM in the system; I'm pretty sure that this is causing the triple fault. As for the page fault in Virtual PC, it is most probably due to the incompleteness of the memory manager.

I've been baffled by the "Invalid Memory Address" bug for a while, it has cropped up occasionally on my systems. I'll try my best to get it rooted out for the next version.

Regards,
Senaus