[Xen-devel] FW: miniOS page table allocation

[ROSSIER Daniel <Daniel.Rossier@xxxxxxxxxx>]

 

Just for info, in-between, I got a flash ;-) and realized that the calculation is correct. Actually, we always consider 2-levels of page table, therefore

there is at least one L1 table, and it covers all the pages which are allocated.

 

But now, I've some other questions regarding miniOS; however, I'm note sure it's still relevant in this mailing list; miniOS is maybe not

of interest anymore (for me it is since I'm using it on ARM processor). Can somebody tell me if there is still some activities along miniOS or another mailing list devoted to it?

 

Thanks

 

Daniel

 

 

From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of ROSSIER Daniel
Sent: vendredi, 22. mai 2009 08:26
To: xen-devel@xxxxxxxxxxxxxxxxxxx
Subject: [Xen-devel] miniOS page table allocation

 

Hi Folks,

 

I'm sorry to come with a so technical question, but I'm working out a port of XEN on ARM in the context of EmbeddedXEN project.

 

I digged into miniOS and found something which seems strange to me: the build_pagetable() function in mm.c.

This function aims at building up the additional page tables which were not allocated by the hypervisor during the

domain construction (domain_build.c). I do not understand the following calculation:

 

pfn_to_map = (start_info.nr_pt_frames - NOT_L1_FRAMES) * L1_PAGETABLE_ENTRIES;

 

pfn_to_map is supposed to be the first page frame number to be mapped by the function. However, the guest OS gets

a list of pfns (start_pfn - max_pfn, including the phys_to_machine table) assigned by the hypervisor, and the first pfns are supposed to contain the startup code and

initial page tables (L2-L1). In this calculation, NOT_L1_FRAMES are typically L2 (and L3,L4 eventually) page table frames. Then, going through the code

pfn_to_map is directly converted to the virtual address and mapped in the corresponding page tables. So, my problem is we do not consider

the L2,L3,L4 frames which are supposed to be in the same virtual address space of the guest OS.

 

I would expect something like pfn_to_map = *start_pfn + (start_info.nr_pt_frames - NOT_L1_FRAMES)*L1_PAGETABLE_ENTRIES.

 

Many thanks for any inputs.

 

Daniel

 

 

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel