OSDev.org

how does the cpu know where to find the tss?

When setting up your system, you place an entry into the GDT. This entry is flagged as a TSS entry and includes the TSS base address. Please consult the manuals, they make this quite clear.

I - SS&ESP is current stack pointer (when you are in kernel it points to top of kernel stack, and when you are in user land it points to top of user (task's) stack.

II - Ther is only one GDT at the time, there is no such thing as user GDT, you can insert descriptor with DPL=3.

III:
1. allocate memory for TSS, write ss0&esp0 fields of TSS
2. add tss descriptor to your gdt
3. execute:
mov ax,YOUR_TSS_SELECTOR
ltr ax
(now cpu knows where is your current tss)

IV - If you want to go to user level then push desired selector (for ring3 code descriptor) & address on stack, execute retf instruction.

V - http://my.execpc.com/_geezer/os/pm.htm

ok i think im starting to get it: i make an entery in my gdt that is user level and when i set up my task i need to set it's ds, ss, ds, gs, ... to my user level gdt entery? but how do i make it a user level gdt entery? (manuals somewere) and what should the base adderess of the entery be? the the start of the tasks space, or 0x00000000???, thx

Your task need not be user-level. It will probably have a kernel- and user-level component.

but how do i make it a user level gdt entery?

Set the DPL field of the descriptor to your desired priveleged level (3).

what should the base adderess of the entery be? the the start of the tasks space, or 0x00000000?

The "start of the task's space", which is 0 in most cases.

Go here: http://www.intel.com/design/pentium4/manuals/index_new.htm. Click on the link under "IA-32 Intel? Architecture Software Developer's Manual, Volume 3: System Programming Guide". Find section 3.4.5 (Segment Descriptors). Another section you might want is 6.2.2 (TSS Descriptor). Someone can probably point you to the proper locations in AMD manuals, if you prefer those.

so what is my kernel is at 0x0 cant the program mess with it??? how does the cpu know that memory is protected when the user level ds includes all memory

Most (all?) people use paging, which allows you to decide for each page whether it is accessible in user mode or not. If you don't use paging, you may want to use segmentation with non-zero base addresses, but be sure you keep things compatible with the assumptions your compiler makes. It's a design decision for you to make.

so paging will protect space, and GDT will protect from execution of non privlaged code, right?

Hi,

GLneo wrote: so paging will protect space, and GDT will protect from execution of non privlaged code, right?

There's many ways to combine segment level protection with page level protection, but most OSs use page level protection for everything.

For example, to prevent user-level code from executing kernel code, you'd mark the kernel's pages as "supervisor level". That way user level code can only use the kernel through special entry points (the kernel API), and can't execute kernel code directly (it'd generate a page fault if it tried).

BTW have you read the Intel manual yet?


Cheers,

Brendan

BTW have you read the Intel manual yet?

yes, read the intel manual (or AMD if you prefer) cover to cover 2 times, then refer back to it frequently!

if you expect to port your OS to x86-64 at anytime, it would be a good idea to use only 'flat mode' segments (as discribed in section 3.2.1)

paging can provide complete protection, user/supervisor, read/only, and execute/noexecute

seriously, read sections 3&4 at least before you continue, and it will give you a better overview of what it is your doing (for task handleing, section 6 is also very important)

whats the danger of having everything in ring 0 but using paging for protection???

User programs is still in Ring 3.
In order to have protection you must have:
1x system code segment (ring 0)
1x user code segment (ring 3)
1x system data segment (ring 0)
1x user data segment (ring 3)
The with paging you can select level (only two: supervisor or user) for a single page.

GLneo wrote: whats the danger of having everything in ring 0 but using paging for protection???

They can perform priviledged instructions.

so what i should do is make a tss that points to the kernel task and ldt its address and make a GDT pointing to it also???, help!

p.s. what keeps a task from doing this: ???

what keeps a task from doing this:

Hardware protection... right before the processor executes that call, it checks the CPL, and if it is greater than zero, it will generate some fault or other (GPF probably). Try it in Windows and see what happens.

Why don't you read the manual carefully?

well that i could of looked up but the first part i cant find, or at lest im having troble understanding