On Tue, 29 Sep 1998 12:04:45 Canul Podkopayeva wrote:

> Why does there exists a VPFN if a process is going to use only one PFN.
> (VPFN = Virtual Page Frame Number)
> (PFN = Page Frame Number)
> 
> The thing is I'm understanding the concept well.
> When a process is mapped to its Virtual Page where does the
> instructions go? What does the offset part (on an x86 it's 11:0 and
> below) do? And why does it need a VFPN?

The text file attached is a copy of an email from Linus which provides a 
description of i386 address mapping. It is the best information I have found on 
the topic.

Intel provides online access to alot of technical information at their web site. 
There is another web site, www.8086.org (or something like that). It is a good 
source of technical information and constructively critical commentary on a wide 
range of topics.

> The reason why I need to know these things is because I'm soon going
> to get into IPC programming and I believe that I most know these for
> it to be easier.

Shared memory is the only form of IPC that I can think of where addresses can 
get you into trouble. If data structures in a shared memory segment contain 
pointers, like a linked list, must map the segment into their address space at 
the same base. If they don't one or more of them will go psychotic posibly 
imploding and taking the rest of the system along for the ride.

"Holes" in user address space can be reserved with linker command line options ( 
I'm not sure about this) or linker scripts (I'm sure about this). There are 
other approaches like rolling your own PIDS (position independent data 
structure), but you be giving up some of the speed of share memory because of 
the extra integer instruction (base increment) for each pointer access. This may 
be not much of a hit in our brave new super scalar/pipelined world, but this 
doesn't help the cause much given the dearth of registers in the current X86 
architecture.

I hope this helps get you started.

Regards,


david
--
David Ross

[EMAIL PROTECTED]
Toad Technologies

"I'll be good! I will, I will !"

IO-mapping.txt.35pre5

  • :ER Canul Podkopayeva
    • David Ross

Reply via email to