RE: [Xen-devel] Informatio required on VMEXit settings and process.

["Petersson, Mats" <Mats.Petersson@xxxxxxx>]

> -----Original Message-----
> From: xen-devel-bounces@xxxxxxxxxxxxxxxxxxx 
> [mailto:xen-devel-bounces@xxxxxxxxxxxxxxxxxxx] On Behalf Of 
> aditya shevalkar
> Sent: 30 January 2007 12:47
> To: xen devel
> Subject: [Xen-devel] Informatio required on VMEXit settings 
> and process.
> 
> Hi all,
> Please help me in understanding who update the VMEXIT_REASON 
> and VMEXIT_QUALIFICATION fields in the VMCS field i.e. Xen or 
> CPU and how when the VMExit occurs.And how the VMexit 
> occurs(porcess of vmexit).

The VMEXIT_REASON is set by the CPU when it sees a condition that
generates a vmexit (i.e. the guest is doing something that the
hypervisor/host has set to exit in "EXEC_CONTROLS"). The
VMEXIT_QUALIFICATION is likewise filled in with the relevant information
by the CPU (such as the address of a page-fault or the interrupt vector
on an interrupt exit). 

[I'm not an Intel expert, but seeing as the overall concept is almost
identical for AMD and Intel, the question can be answered in a way that
applies to both, with the slight difference that the AMD uses a VMCB and
different names for the fields. For example, exit_reason in Intel is
called exitcode in AMD's world, and exit_qualification is called
exitinfo{1,2} and exitintinfo]

> Also explain this process in case of page fault if possible.

Processor sees a memory access. The processor checks the page-tables,
and if the access is not allowed, the processor gives a vmexit to the
hypervisor. The hypervisor checks the exit_reason to see that it's a
page-fault, and determines if it's a "hypervisor" related page-fault
(such as a page-table update or a MMIO access to a virtual device) or a
"guest fault". For the former, the hypervisor will then perform the
relevant work to either update the page-table or to send MMIO operation
to QEMU. If it's not a hypervisor fault, it is by process of elimination
a guest-fault (for example a page-access that hits a page that is stored
in the swap or a write to a page that is shared with "copy-on-write" in
a forked process). Guest faults are forwarded to the guest by
"injection". 

A more concrete question would probably give you a more precise answer -
there are MANY pages in the manual on the subject, so it's pretty much a
waste of everyones time to rewrite it here... 

> Give the code reference if possible.

.../xen/arch/x86/hvm/svm/svm.c for AMD code. 
.../xen/arch/x86/hvm/vmx/vmx.c for Intel code. 

--
Mats

> 
> Thanks and Regards,
> Aditya Shevalkar.
> 
> 
>               
> __________________________________________________________
> Yahoo! India Answers: Share what you know. Learn something new
> http://in.answers.yahoo.com/
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@xxxxxxxxxxxxxxxxxxx
> http://lists.xensource.com/xen-devel
> 
> 
> 



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