Xen devel - Mar 2008 - Is it possible to make hypercall from user application directly?

Hi,

I implemented my own Hypercall in xen.

Now I wish my application under dom0 could talk to Xen via my hypercall.
Currently, I make a stub in dom0''s kernel, which works as a proxy to
forward
the request. So, my application first makes a system call to dom0, which
then makes hypercall to xen.

Is this design necessary? May I call hypercalls from user application under
dom0 directly?

Thanks,
Weiming


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

On Tue, 2008-03-11 at 12:13 -0500, weiming wrote:
> Hi,
> 
> I implemented my own Hypercall in xen.
> 
> Now I wish my application under dom0 could talk to Xen via my
> hypercall.
> Currently, I make a stub in dom0''s kernel, which works as a proxy
to
> forward the request. So, my application first makes a system call to
> dom0, which then makes hypercall to xen. 
> 
> Is this design necessary? May I call hypercalls from user application
> under dom0 directly?
No, not directly (didn''t we have that syscall conversation recently?
;).

But for dom0-only privileged operations there''s /proc/xen/privcmd to
rule them all. See e.g. do_xen_hypercall() in tools/libxc/.

regards,
Daniel

-- 
Daniel Stodden
LRR     -      Lehrstuhl für Rechnertechnik und Rechnerorganisation
Institut für Informatik der TU München             D-85748 Garching
http://www.lrr.in.tum.de/~stodden         mailto:stodden@cs.tum.edu
PGP Fingerprint: F5A4 1575 4C56 E26A 0B33  3D80 457E 82AE B0D8 735B



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

After some study, I got the following understanding. Please correct me if
I''m wrong; otherwise, you can just say "correct".

1. In x86_64 mode, xen sets the LSTAR (the EIP of syscall target), and this
value keeps constant. This value is code of "syscall_enter" defined in
xen/arch/x86/x86_64/entry.S

2.
In ENTRY(syscall_enter)
...
...
testb $TF_kernel_mode, VCPU_thread_flag(%rbx)
jz switch_to_kernel

/*hypercall*/
...

so whenever a syscall is issued, "syscall_enter" checks the context.
If it
decides that the context is "kernel mode", it will treat the syscall
as a
plain "system call" from user applications. So it goes to "jz
switch_to_kernel". Otherwise, it will think it''s a hypercall
issued from
guest domains.

3. So an application can''t make "hypercalls" directly because
it''s context
is "kernel mode", so it will always be treated as a system call to the
guest
OS.

Am I right?

Thanks,
Weiming

On Tue, Mar 11, 2008 at 2:50 PM, Daniel Stodden <stodden@cs.tum.edu>
wrote:
> On Tue, 2008-03-11 at 12:13 -0500, weiming wrote:
> > Hi,
> >
> > I implemented my own Hypercall in xen.
> >
> > Now I wish my application under dom0 could talk to Xen via my
> > hypercall.
> > Currently, I make a stub in dom0''s kernel, which works as a
proxy to
> > forward the request. So, my application first makes a system call to
> > dom0, which then makes hypercall to xen.
> >
> > Is this design necessary? May I call hypercalls from user application
> > under dom0 directly?
>
> No, not directly (didn''t we have that syscall conversation
recently? ;).
>
> But for dom0-only privileged operations there''s /proc/xen/privcmd
to
> rule them all. See e.g. do_xen_hypercall() in tools/libxc/.
>
> regards,
> Daniel
>
> --
> Daniel Stodden
> LRR     -      Lehrstuhl für Rechnertechnik und Rechnerorganisation
> Institut für Informatik der TU München             D-85748 Garching
> http://www.lrr.in.tum.de/~stodden
<http://www.lrr.in.tum.de/%7Estodden>
>       mailto:stodden@cs.tum.edu
> PGP Fingerprint: F5A4 1575 4C56 E26A 0B33  3D80 457E 82AE B0D8 735B
>
>
>

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

On Tue, 2008-03-11 at 21:42 -0500, weiming wrote:
> After some study, I got the following understanding. Please correct me
> if I''m wrong; otherwise, you can just say "correct".
> 
> 1. In x86_64 mode, xen sets the LSTAR (the EIP of syscall target), and
> this value keeps constant. This value is code of "syscall_enter"
> defined in xen/arch/x86/x86_64/entry.S
correct.
> 2. 
> In ENTRY(syscall_enter)
> ...
> ...
> testb $TF_kernel_mode, VCPU_thread_flag(%rbx)
> jz switch_to_kernel
> 
> /*hypercall*/
> ...
> 
> so whenever a syscall is issued, "syscall_enter" checks the
context.
> If it decides that the context is "kernel mode", it will treat
the
> syscall as a  plain "system call" from user applications. So it
goes
> to "jz switch_to_kernel". Otherwise, it will think it''s
a hypercall
> issued from guest domains.
correct. PT switching is then done in
create_bounce_frame/toggle_guest_mode. 
> 3. So an application can''t make "hypercalls" directly
because it''s
> context is "kernel mode", so it will always be treated as a
system
> call to the guest OS.
believe you mean "user mode", but otherwise correct.

best,
daniel


-- 
Daniel Stodden
LRR     -      Lehrstuhl für Rechnertechnik und Rechnerorganisation
Institut für Informatik der TU München             D-85748 Garching
http://www.lrr.in.tum.de/~stodden         mailto:stodden@cs.tum.edu
PGP Fingerprint: F5A4 1575 4C56 E26A 0B33  3D80 457E 82AE B0D8 735B



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