Re: Add 'info pg' command to monitor

[Don Porter <porter@cs.unc.edu>]

On 4/15/24 12:37, Peter Maydell wrote:
> On Mon, 15 Apr 2024 at 17:09, Don Porter <porter@cs.unc.edu> wrote:
>
> This patch set adds an 'info pg' command to the monitor, which prints
> a nicer view of the page tables.  A project in my graduate OS course
> involves implementing x86 page table support, and my students have
> found this helpful for debugging.
> So, my issue with this is that it's x86 specific, and it adds
> yet another monitor command that is doing "show some kind of debug
> info related to the guest page tables", along with "info mem"
> and "info tlb". Plus it is yet another lump of code that's
> walking the guest page tables and interpreting them.
>
> What I'd really like to see is some infrastructure that is
> at least somewhat guest-architecture-agnostic, so we can
> minimise what a guest architecture needs to implement (and
> then make providing that mandatory).
>
> The other thing I'd like to see is perhaps some investigation of
> whether there's any way to implement something useful by
> using/extending the existing get_phys_page_attrs_debug() and
> similar functions, so that you don't have to write one lot
> of page-table-walking code for QEMU to use when it's executing
> guest code and a separate lot (that's bound to get out of
> sync or not support new functionality/changes) that's only
> handling these monitor debug commands. There's a lot of
> complexity in figuring out things like permissions in a
> modern architecture...
>
> thanks
> -- PMM
>   

Thank you for the feedback!

There is still a lot I am learning about the code base, but it seems 
that qemu_get_guest_memory_mapping() does most of what one would need.  
It currently only returns the "leaves" of the page table tree in a list.

What if I extend this function with an optional argument to either
1) return the interior nodes of the page table in additional lists (and 
then parse+print in the monitor code), or
2) inline the monitor printing in the arch-specific hook, and pass a 
flag to get_guest_memory_mapping() that turns on/off the statements that 
pretty print the page tables?

It looks like most CPUs implement this function as part of checkpointing.

Thoughts?

Don