Skip to contentSkip to conversation

Memory management in Linux

DISCUSSION Josef E. Galea 4 participants 5 messages Dec 29, 2004 – Dec 30, 2004

lore
Share & tools
lore
#0 lore raw

Hi all,

Does the linux kernel allow a process to handle its own memory pages instead of using the kernel's virtual memory manager?

Thanks & Happy Holidays
Josef

#1 lore raw
Quoted 7 lines · Josef E. Galea

Hi all,

Does the linux kernel allow a process to handle its own memory pages instead of using the kernel's virtual memory manager?

Thanks & Happy Holidays
Josef

That's quite related to ``adaptive page replacement''. Linux doesn't support that at present imho.

--cqh

Signature · 2 lines
Coywolf Qi Hunt
Homepage http://sosdg.org/~coywolf/
#2 lore raw
Quoted 15 lines · Coywolf Qi Hunt
Quoted 9 lines · Josef E. Galea

Hi all,

Does the linux kernel allow a process to handle its own memory pages instead of using the kernel's virtual memory manager?

Thanks & Happy Holidays
Josef

Quoted 8 lines

That's quite related to ``adaptive page replacement''. Linux doesn't support that at present imho.

--cqh

Ok I may have got the name wrong :). What I am trying to do is to implement a package on linux similiar to the TreadMarks by Alan Cox et al. (ref. http://citeseer.ist.psu.edu/amza96treadmarks.html) that runs at kernel level instead of user level. Right now I think that inorder to achieve what I want to do, I have to change the code of the linux virtual memory manager. This is ok for academic purposes (which is my aim) however it severly reduces portability (it is much easier to just load a kernel module than to patch and recompile the kernel).

Thanks
Josef

#3 lore raw
Quoted 2 lines · Josef E. Galea

Ok I may have got the name wrong :). What I am trying to do is to implement a package on linux similiar to the TreadMarks by Alan Cox et al. (ref. http://citeseer.ist.psu.edu/amza96treadmarks.html) that runs at kernel level instead of user level. Right now I think that inorder to achieve what I want to do, I have to change the code of the linux virtual memory manager. This is ok for academic purposes (which is my aim) however it severly reduces portability (it is much easier to just load a kernel module than to patch and recompile the kernel).

Kernel-level thread or code can modify mm_struct,vma,pde,pte,page protection bits,etc. I think that it can modify everything in kernel.

Regards,
LWT

#4 lore raw
In reply to Josef E. Galea

* Josef E. Galea:

Quoted 2 lines

Does the linux kernel allow a process to handle its own memory pages instead of using the kernel's virtual memory manager?

You could play some tricks using mprotect(2), but I'm not sure if this is enough for your application.