Triple fault when enabling paging

[neon]

pgtbl[0] = 0x00000003 is correct however pgdir[0] = 0x0009B803 is not. Assuming the page table is at 0x00137000, pgdir[0] should be 0x137003. Without farther information, I cannot confirm where that 0x0009B803 is coming from - it points to a nonexistent page table at 0x9b000.

In fact, pgtbl[0] = 0x00000003, pgtbl[1] should be 0x00001003, pgtbl[2] should be 0x00002003 ... and pgtbl[1023] should be 0x003ff003. This maps frames 0 - 0x3ff, or addresses 0 - 0x400000 inclusive. If you got these as well, then we can focus on why the page directory is wrong.

[abcdef4bfd]

I'm puzzled. In your initial post you said the problem happened when you turned on paging. Now you are saying that it isn't in your code.

If it's crashing before you can do any debugging then from where are you getting the values that you tell us your page table entries have?

No, no. You misunderstanded me. The paging problem is in my code, my message about "not my code" is about Bochs crash on GRUB menu, not about paging.

[abcdef4bfd]

pgtbl[0] = 0x00000003 is correct however pgdir[0] = 0x0009B803 is not. Assuming the page table is at 0x00137000, pgdir[0] should be 0x137003. Without farther information, I cannot confirm where that 0x0009B803 is coming from - it points to a nonexistent page table at 0x9b000.

In fact, pgtbl[0] = 0x00000003, pgtbl[1] should be 0x00001003, pgtbl[2] should be 0x00002003 ... and pgtbl[1023] should be 0x003ff003. This maps frames 0 - 0x3ff, or addresses 0 - 0x400000 inclusive. If you got these as well, then we can focus on why the page directory is wrong.

I don't know why, but pgtbl is increases by 0x800, not by 0x1000.

[abcdef4bfd]

0x00000003, 0x00000803, 0x00001003, ...

[iansjack]

Your definition of Page Table entries is wrong. You're missing the PWT bit. And the address field should be 20 bits, not 21

[abcdef4bfd]

Your definition of Page Table entries is wrong. You're missing the PWT bit. And the address field should be 20 bits, not 21

Thanks, fixed. Now pgtbl is increasing by 0x1000. What next, @neon?

[neon]

Make sure that you made the same update to your page directory structure (it follows the same structure as the page table; the address field should also be 20 bits.) What is (uint32_t)pgdir[0] and the address of page directory now?

[abcdef4bfd]

pgdir[0] is changed, now it's 0x00137003. Address is the same - 0x00138000. My page directory structure is:

Code: Select all

struct page_directory_t
{
    unsigned int present       : 1;
    unsigned int rw_flag       : 1;
    unsigned int access_lvl    : 1; //0 is for only ring0. 1 is for anybody.
    unsigned int write_through : 1;
    unsigned int cache_off     : 1;
    unsigned int accessed      : 1;
    unsigned int zero          : 1;
    unsigned int page_size     : 1;
    unsigned int unused        : 4; //Now it's 4, because of 1 ignored bit (G in wiki article)
    unsigned int tbl_addr      : 20;
};

[neon]

Before you told us that the page table was at 0x00137000 not 0x00138000. Please double check which one is correct. pgdir[0] = 0x00137003 is only valid if the page table is at 0x00137000.

[abcdef4bfd]

Yes, that's a my typo The real address of pgdir is 0x00138000 and address of pgtbl is 0x00137000.

[neon]

Okay, great - this means pgdir[0] is now correct and so is your page table. Now, post the value of the cr3 register. It should be 0x00138000.

[abcdef4bfd]

CR3==0x00138000
YAAAY!

[neon]

Does it work without error or still triple fault?

[abcdef4bfd]

Does it work without error or still triple fault?

Works, I'll go and map the 32 MB of memory. Paging is enabling now, but I inserted the cli; hlt after it, so it cannot access non-accessible yet VBE framebuffer.

[abcdef4bfd]

Thank you, neon and iansjack!