RPL,DPL,CPL (x86 assembler i deskriptori/slektori)
Naso sam jedan veom zanimljivo post na comp.lang.asm.x86 news grupi :
> I'm reading a section in a book about Protected Mode
> and I've been having difficulties distinguishing
> between the functionality and purpose of the RPL
> located on the segment selector and the DPL located
> on the segment descriptor.
> In addition, there is the EPL which I understand by
> definition but no further than that.
No there isn't an EPL - i'll assume you mean CPL.
Well here goes. (First a little background)
Privledge levels have 4 possible values (binary)
00 highest privledge - operating system kernel
01 rest of the operating system software
10 trusted applications ie. database engines
11 normal applications (and v86 code) - you will
normally run most stuff in this mode
- lowest privledge
(The backward nature [00=highest->11=lowest] is confusing at first but you get $
course when you are coding your own OS you may elect
to only use a limited number of these modes (ie. 00
and 11 for system/user) or run every thing in ring
00 (no protection).
The CPL is the current privledge level - this is the
privledge the currently executing instruction/code is
running at - essentially the reference point against
which the other values are tested.
The DPL is the descriptor privledge level - this is
the privledge required to access the selector. It
is stored in the flags of the selector in the GDT or
LDT (depending on the location of the selector). The
selector can only be accessed if CPL<=DPL otherwise
fault 13h / general protection fault (#GF) occurs.
RPL is the requestor privledge level - this is used
when a selector is passed from a less privledged
procedure to a more privledged procedure to stop
possible accesses of the selector passed if the
application was unable to access the selector. The
RPL is stored in bits 1 and 0 of the selector. The
instruction ARPL is designed to alter the RPL before
it is loaded into a selector. If a selector is
accessed #GF occurs if CPL>DPL ior RPL>DPL. Setting
RPL to 0 allows it to be ignored, and if protection
is not an issue RPL should always be set to 0.
(nb: if you are passing a selector along a chain of
increasingly privledged procedures instead of setting
RPL = CPL of caller, set RPL as the higher of old RPL
and the CPL of the caller.
Ie. If CPL = 10,
SS = DPL10, RPL00
DS = DPL10, RPL10
ES = DPL00, RPL00
FS = DPL11, RPL11
GS = DPL10, RPL11
Then SS, DS and FS could be accessed but any attempt
to access ES or GS would cause #GF
(nb: I think even loading ES would cause a #GF however
you'll have to check - I'm writing this from memory)
That should answer you question - but I'll also mention
IOPL (stored in flags) this is the minimum privledge
IN or OUT instruction can occur at. If an IN or OUT is executed and CPL>IOPL t$
(in the current TSS) is checked. Then if the I/O is
accessing a port forbidden bu this bitmap #GF occurs.
A number of instructions are forbidden in certain
privledge levels.
If CPL>00 LIDT,LLDT,LGDT,LTR,LMSW,CLTS and HLT are
forbidden.
If CPL>IOPL STI and CLI are forbidden.
If CPL>IOPL IN,OUT,INS and OUTS are subject to the
I/O permission bitmap.
Also IRET and POPF are modified according to CPL
(IOPL field is unchanged if CPL>00 and IF field is
unchanged if CPL>IOPL)
[in v86 mode things are slightly different - but
that can be left for another day]
 |  |
 |
