Re: support of hardware nested page table for vkernels

[Ankur Agrawal, Jitendra Pratap Singh <anagrawal@xxxxxxxxxxxxxxxxx>, jpsingh@xxxxxxxxxxxxxxxxx]

Samuel J. Greear <sjg <at> evilcode.net> writes:

> 
> 
> On Mon, Mar 19, 2012 at 7:29 AM, Jitendra Singh <jpsingh <at> 
cs.stonybrook.edu> wrote:
> 
> 
> 
> 
> 
> 
> 
> Hi,
> 
> 
> 
> 
> We are a group of two Grad students from Stony Brook University, NY. We are 
keenly interested and want to contribute in the project for supporting hardware 
nested page table for vkernels for mentioned in the GSoC 2012. In fact, we would 
even love to add support
>  for all hardware virtualization features provided by AMD-V architecture and 
Intel-VT architecture if it already has not been supported. 
> 
> 
> 
> 
> As a matter of fact, we are already working towards providing software and 
hardware nested paging support for JOSÂ (JOS is an Operating System developed at 
MIT for academic purposes)Â and also towards providing support for emulation of 
AMD-V virtualization features
>  in GEM5 emulator.
> 
> 
> 
> 
> Our past experience with Operating Systems and Virtualization includes design 
and implementation of various features of JOS form scratch (like the bootloader, 
Virtual Memory set-up, Multiprogramming and Interrupt Setup, IPC setup, 
filesystem and network driver
>  set up). We also design and implemented a paravirtualized Virtual Machine 
Manager for JOS for one our course projects. Apart from this, we are working 
towards design and implementation of first ever device driver and filesystem for 
Shingled Magnetic Recording
>  disks in LINUX.
> 
> 
> 
> 
> We strongly believe that our current work towards support of AMD-V emulation 
in GEM5 and support of nested paging for JOS will give us significant exposure 
to gain considerable insights into the hardware Virtualization and hence would 
help us in adding an efficient
>  support for hardware Virtualization features in DragonFly BSD.
> 
> 
> 
> 
> We are looking forward to hear from you.
> 
> 
> 
> 
> Thanks,
> 
> Ankur Agrawal, Jitendra Pratap Singh
> 
> MS, Computer Science
> Stony Brook University
> 
> 
> 
> 
> 
> 
> 
> 
> This is potentially a large (difficult) enough project that it could be split 
into two independent GSoC projects with no overlap, but the two of you will have 
to work out how to make that happen and it is possible that only one of the two 
proposals might be selected, or etc. Both proposals (there must be two, since 
there are two of you -- that is not to say that both of you can't work on it, 
just that you both need an independent proposal to be an official student and 
paid/recognized by Google) must be able to stand on their own and each must 
propose to result in something that will be potentially commit-able and use-able 
on its own.
> 
> These two articles:
> 
> http://lwn.net/Articles/228404/
> 
> 
> http://lwn.net/Articles/230658/
> 
> Give an excellent overview of the architecture of the vkernel architecture, 
how it manages page tables and how it all relates to the vm subsystem in 
DragonFly. I believe it should be possible to implemented (optional) hardware 
nested pagetable support within the existing vmspaces kernel API, which already 
implements software nested pagetables (but if it turns out slight tweaks need to 
happen, that's probably OK too).
> 
> There is nothing saying you can't take this further, and one of the two of you 
work on optional IO-MMU support or etc., the project ideas are simply that, 
ideas.
> 
> Best,
> Sam
> 


Hi,

Thanks for your encouraging mails. We eagerly look forward to this project

As suggested by you, we are already aware of the difficulty and the size of the 
project and hence we are working towards splitting this idea into two non-
overlapping but correlated modules. 

While we are already working at it, we will be glad to have any ideas or 
suggestion that you think might be useful for us in splitting the project. 

Thanks
Jitendra, Ankur