Page Fault! on Multitasking James's Kernel code !!

[smainoo]

N.B: Take a look here Before reading the thread : http://wiki.osdev.org/James_Molloy%27s_ ... Known_Bugs
Hi guys,
To build an OS you should Know how you can catch Bugs , So i tried to catch some
but it seems like debug an OS is a tough task .
take a look at this Toaruos Code source (according to Kevin "it's not a refrence, There are some bugs").
https://github.com/klange/toaruos/tree/ ... a34e573d2a

if comment out the fork() function in main.c, everything goes fine.
did anyone have an idea or suggestions ??

image:
http://postimg.org/image/47gbsm0m1/

[klange]

The bugginess of the JamesM tutorial's approach to multitasking is well known. I wouldn't go digging through any of my old code (or much of my current code for that matter) as a good lot of it is broken.

[smainoo]

The bugginess of the JamesM tutorial's approach to multitasking is well known. I wouldn't go digging through any of my old code (or much of my current code for that matter) as a good lot of it is broken.

You are right kevin , & delving in such problems is a waste of time if you've something interesting , but for someone who want to learn some debugging skills, he should know why this code didn't work . (am i crazy ??)

[smainoo]

After some debugging, finally found some Bugs ... lol.

Code: Select all

  
   .............
   .............
   0x001037d7 <+159>:	mov    ds:0x105010,eax
   0x001037dc <+164>:	mov    eax,ds:0x105010
   0x001037e1 <+169>:	mov    ebx,DWORD PTR [eax+0x2000]
   0x001037e7 <+175>:	mov    eax,DWORD PTR [ebp-0x14]       <----- eax = eip
   0x001037ea <+178>:	mov    edx,DWORD PTR [ebp-0xc]
   0x001037ed <+181>:	mov    ecx,DWORD PTR [ebp-0x10]       <-----  ecx = ebp
    
   0x001037f0 <+184>:	cli                                                                   asm volatile("   cli;                 \
   0x001037f1 <+185>:	mov    ecx,eax  <---- ecx = eip (overwriting ebp)    mov %0, %%ecx;       \
   0x001037f3 <+187>:	mov    esp,edx                                           mov %1, %%esp;       \
   0x001037f5 <+189>:	mov    ebp,ecx  <---- ebp = ecx =eip             mov %2, %%ebp;       \
   0x001037f7 <+191>:	mov    cr3,ebx                                            mov %3, %%cr3;       \
   0x001037fa <+194>:	mov    eax,0x12345                                     mov $0x12345, %%eax; \         
   0x001037ff <+199>:	sti                                                            sti                 \
   0x00103800 <+200>:	jmp    ecx                                                  jmp *%%ecx    \
   0x00103802 <+202>:	add    esp,0x14                           : : "r"(eip), "r"(esp), "r"(ebp), "r"(current_directory->physicalAddr) ); 
   0x00103805 <+205>:	pop    ebx
   0x00103806 <+206>:	pop    ebp
   0x00103807 <+207>:	ret    
End of assembler dump.

You can see in the above code that ebp will receive the value of eip instead of our stack ebp.
Add clobbered registers to our inline asm, & use ebx instead of ecx solves the problem of "page fault" (on James'S Multitasking code).

Code: Select all

asm volatile("         \
      cli;                 \
      mov %0, %%ebx;       \
      mov %1, %%esp;       \
      mov %2, %%ebp;       \
      mov %3, %%cr3;       \
      mov $0x12345, %%eax; \
      sti;                 \
      jmp *%%ebx           "
                 : : "r"(eip), "r"(esp), "r"(ebp), "r"(current_directory->physicalAddr)// );
                 : "%ebx", "%esp", "%eax");

N.B: after viewing some newer commits i found that Kevin just resolve the problem with this solution, which help me to get the problem. https://github.com/klange/toaruos/commi ... df1c38be3a.

N.B : On ToaruOS commit (after switch_task() the code just leave the main & stuck in the infinite loop of start.s).