Re: [apostle-developers] design

[Guido de J. <gu...@us...>]

On Saturday 06 October 2001 04:32, you wrote:
> hi guido,
> i compiled the nucleus with MOREDEBUG defined. then i
> saw the following message. (i typed it, so please dont
> mind any tipos ;-)

OK

> ..
> normal stuff
> ..
> Setting CPU upcalls to (0x10000018,0x10000024) for
> process 512
> returning current process information
> creating capablity
> 1:(512,0x00007fdd,0x00000002,0x00000007)
> creating capablity

This creates a capability for the page directory, needed to change anything
in the address space. What you see is:
cap#:(PID,page#,cap-type,flags)

> 2:(512,0x00000000,0x00000000,0x00000000)
> creating capablity

Acquire an empty page, the page number is not shown

> 3:(512,0x000000b8,0x00000000,0x00000007)
> creating capablity

Create a capability for crt output, so it can write to the screen.

> 2:(512,0x00007fda,0x00000002,0x00000007)
> setting entry 400 in page directory 0x07fdd000 to
> 0x07fd0007

This sets the empty page to entry 400 in the page dir. The empty page is now
a page table.

> setting entry 0 in page table 0x07fda000 to 0x07fd0008

Maybe you mistyped something here, 'cause the last address should read
0x000b8007. Please verify.

> creating capablity
> 4:(512,0x000003d4,0x00000004,0x00000001)
> creating capablity
> 5:(512,0x000003d5,0x00000002,0x00000001)

This just creates capabilities to ports 0x3d4 and 0x3d5, needed to move the
cursor and stuff like that.

> page fault 0x08000002
> eip 0x00100806
> address 0x0bff8000

The next thing that should happen is that the ports above are being enabled.
This is done setting 2 bits in the IO bitmap to 0. Apparently this generates
a page fault. Don't now why at the moment.

> my query is weather this address not lying in the page
> table and directory entries u have made?

This has nothing to do with the page that are created. You can use nm to find
the function that is executed when the PF is generated.

> i assume that the address is faulty (please correct me
> here if iam wrong... i still dont have a hang on the
> address space used ;-)...) and iam currently trying to
> analyse what code is present in 0x00100806 and what
> gave it a faulty page.

Type 'nm -Cgn gemini-nucleus'. The output will help.

> if iam taking a wrong approach, or if u have already
> saw something of this kind tell me ;-)

Your approach is fine, please keep us informed on anything you find out. I
will look into it too in the meantime.

Guido