Stealing memory pages.

Stealing memory pages.

[amey d inamdar]

Hi,
    We are a group of four students, working on a project, "Implementation of Network RAM". We want to add the remote pages to the address space of a process ( by Modifying page fault handler). 
    For the same reason we need a fixed allocation of a pool of page frames on each machine, which will serve as source for NRAM pages. This pool on each machine will be handled by a "Server" which will be implemented as a kernel module.
    So at the initialization only, we have to allocate the page frames to the module. We did it successfully as follows:
1)  After setting up mem_map ( array of page *), while freeing individual page we didn't give last few pages to the buddy deallocator.
2)  We individually marked all those pages non-reserved and stored virtual address of start of the first page frame. (__va(page)). The virtual address is part of kernel address space.
3) Now our server module will use this address and total no of pages, to manage allocation of pages to a remote process.
   My question is that, whether blocking such virtual address space inside the kernel can cause harm to its functionality? The machine is still working fine, but are there any ill-effects of such page frame stealing?
   thank you in anticipation.
- Amey 

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/

Re: Stealing memory pages.

[Ravi K]

Hi,
  I guess you have to make sure that globals like
max_low_pfn, highstart_pfn, highend_pfn and max_mapnr
are consistent with your setup. Some of these are used
for validating page structures and in determining the
start address used by vmalloc(). Other than that, I
don't think there is any problem with this setup.
  Of course, someone will correct me if I am wrong
here..

-Regards,
 Ravi.

--- amey d inamdar <iamey@rediffmail.com> wrote:
> Hi,
>     We are a group of four students, working on a
> project, "Implementation of Network RAM". We want to
> add the remote pages to the address space of a
> process ( by Modifying page fault handler). 
>     For the same reason we need a fixed allocation
> of a pool of page frames on each machine, which will
> serve as source for NRAM pages. This pool on each
> machine will be handled by a "Server" which will be
> implemented as a kernel module.
>     So at the initialization only, we have to
> allocate the page frames to the module. We did it
> successfully as follows:
> 1)  After setting up mem_map ( array of page *),
> while freeing individual page we didn't give last
> few pages to the buddy deallocator.
> 2)  We individually marked all those pages
> non-reserved and stored virtual address of start of
> the first page frame. (__va(page)). The virtual
> address is part of kernel address space.
> 3) Now our server module will use this address and
> total no of pages, to manage allocation of pages to
> a remote process.
>    My question is that, whether blocking such
> virtual address space inside the kernel can cause
> harm to its functionality? The machine is still
> working fine, but are there any ill-effects of such
> page frame stealing?
>    thank you in anticipation.
> - Amey 


__________________________________________________
Do You Yahoo!?
Check out Yahoo! Shopping and Yahoo! Auctions for all of
your unique holiday gifts! Buy at http://shopping.yahoo.com
or bid at http://auctions.yahoo.com
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/