Hi Santosh :)

On Sat, Feb 4, 2012 at 13:22, SaNtosh kuLkarni
[off-list ref] wrote:

My doubt is when , in a, say 3:1 partition of kernel user space address
space, if a user space application tries to write/access the 1gb portion of
kernel space which it cant, is it because there is not page table entry in
the user space process page table entry or......... is it because of a PTE
?( page table entry) ?entry related to access control set to KWE and a
kernel trap occurs.

it's due to CPL (current privilege level) doesn't match the privilege
level of kernel address space (which is marked as 0 in their PTEs). In
user mode, your CPL is 3, whereas you're targetting privilege level 0.

If it is the other way around, kernel mode accessing user space, then
there won't be any problem.

What events take place when a user land process tries to
write to?kernel?space address.

i think it's page fault. I forgot the exact route, but I guess it
starts as hardware trap (due to the above explanation), kernel trap
handler kicks in, knowing that it is due privilege level mismatch and
simply throws instruction pointer back to user mode together with
error code (IMO it's EFAULT).

-- 
regards,

Mulyadi Santosa
Freelance Linux trainer and consultant

blog: the-hydra.blogspot.com
training: mulyaditraining.blogspot.com