Re: [Xen-devel] Share memory between domains

[Facoltà di informatica <ferrucci@xxxxxxxxxxxxxxx>]

> You can share memory between the kernels of two domains by using the
> grant table "grant access" mechanism. This is described with example
> code in docs/misc/grant-tables.txt. The gist of it is:
>
> domA creates a grant reference, and then communicates it to domB somehow
> (for example, over a network connection would be one option).
> see: gnttab_grant_foreign_access
>
> domB uses the grant reference to map the foreign frame into its address
> space.
> see: HYPERVISOR_grant_table_op( GNTTABOP_map_grant_ref, ...)
>
> The block device front and back drivers in xen-unstable use this to
> communicate.
>
> To share memory between user space processes in different domains would
> require an extension to the grant table mapping mechanism to allow a
> domain to specify the PTE it wishes to put the foreign frame number
> into, rather than using the linear page table.
>
> There is an alternative grant table mode used to transfer frames from
> one domain to another; this wouldn't help you out with shared memory.
>
> The grant table code currently in libxc is disabled, of debugging
> interest only and unlikely to be of use to you.
>
> Event channels are a way for one domain to notify the other that an
> event of interest has occurred. This is an extremely low bandwidth
> channel and should be used to trigger inspection of shared memory for
> the actual communication. "ring.h" provides macros to assist with
> building a structured communication channel between domains using shared
> memory.
>
> Christopher
My really intention is to construct a communication channel between a daemon 
in a guest Os and  a daemon in Dom0 without use the network.I think it's the 
same of the block device driver under Xen, but I don't know how Xend 
communicate to the frontend driver the number of the port of the event 
channel and the frames of the shared pages of the ring. Can you halp me?
Thanks

Ferrucci Luca 

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