Discussion:
Help With Custom Hyper Calls
John Backes
2011-08-12 16:19:53 UTC
Permalink
Hello,

So I'm new to Xen development and want to play around with Xen
Hypercalls, but havn't been able to find a whole lot of documentation
related to creating custom Hypercalls. I've found that to create a
custom Hypercall I need to define it with an unused Hypercall number in
"xen/include/public/xen.h", but then I am unsure of the next steps
(e.g., where to place the function to call with the Hypercall).

Can anyone point more towards some documentation or at least towards the
correct locations for registering new Hypercalls? Thanks in advance.

- John
Keir Fraser
2011-08-12 17:02:01 UTC
Permalink
Post by John Backes
Hello,
So I'm new to Xen development and want to play around with Xen
Hypercalls, but havn't been able to find a whole lot of documentation
related to creating custom Hypercalls. I've found that to create a
custom Hypercall I need to define it with an unused Hypercall number in
"xen/include/public/xen.h", but then I am unsure of the next steps
(e.g., where to place the function to call with the Hypercall).
Can anyone point more towards some documentation or at least towards the
correct locations for registering new Hypercalls? Thanks in advance.
Pick the name of a hypercall, e.g., update_va_mapping has a nice distinctive
name. Recursive grep for that name in the Xen and Linux source trees. Should
give you a pretty good start on the few places you need to register in the
hypervisor and in the guest kernel.

-- Keir
Post by John Backes
- John
_______________________________________________
Xen-devel mailing list
http://lists.xensource.com/xen-devel
John Backes
2011-08-15 14:45:06 UTC
Permalink
So I greped through the s

I've altered the hypercall_table and hypercall_args_table to have an
additional entry in xen/arch/x86/x86_32/entry.S and in
xen/arch/x86/x86_64/entry.S:

to the hypercall_table:

...........................
.long do_sysctl /* 35 */
.long do_domctl
.long do_kexec_op
.long do_tmem_op
.long do_new_hyper /* 39 */
...........................

to the hypercall_args_table:

...........................
.byte 1 /* do_sysctl */ /* 35 */
.byte 1 /* do_domctl */
.byte 2 /* do_kexec_op */
.byte 1 /* do_tmem_op */
.byte 0 /* do_new_hyper */ /* 39 */
...........................


I've also registered the name in xen/include/public/xen.h

...........................
#define __HYPERVISOR_new_hyper 39
...........................

and I've added the following function to xen/arch/x86/mm.c

...........................
void do_new_hyper ( void )
{
printk("NEW HYPERCALL RECEIVED\n");
}
...........................

To test this, I am running a fedora 14 dom0 and I wrote the following
kernel module (so the code runs in ring 1):

...........................
#include <linux/init.h>
#include <linux/module.h>
#include <linux/fs.h>

MODULE_LICENSE("GPL");

#define SUCCESS 0

static int hyper_init(void){

int output;

printk(KERN_ALERT "Testing Hypercall\n");

__asm__ ( "movl $39, %%eax;"
"int $0x82;"
: "=a" (output)
);

return SUCCESS;
}

static void hyper_exit(void){
printk(KERN_ALERT "Removing Hypercall Module");


}

module_init(hyper_init);
module_exit(hyper_exit);
...........................

I then run "xm dmesg" to see if I can see the "NEW HYPERCALL RECEIVED"
message, but nothing appears. Any thoughts?

- John
Post by Keir Fraser
Post by John Backes
Hello,
So I'm new to Xen development and want to play around with Xen
Hypercalls, but havn't been able to find a whole lot of documentation
related to creating custom Hypercalls. I've found that to create a
custom Hypercall I need to define it with an unused Hypercall number in
"xen/include/public/xen.h", but then I am unsure of the next steps
(e.g., where to place the function to call with the Hypercall).
Can anyone point more towards some documentation or at least towards the
correct locations for registering new Hypercalls? Thanks in advance.
Pick the name of a hypercall, e.g., update_va_mapping has a nice distinctive
name. Recursive grep for that name in the Xen and Linux source trees. Should
give you a pretty good start on the few places you need to register in the
hypervisor and in the guest kernel.
-- Keir
Post by John Backes
- John
_______________________________________________
Xen-devel mailing list
http://lists.xensource.com/xen-devel
Tim Deegan
2011-08-15 14:56:26 UTC
Permalink
Post by John Backes
So I greped through the s
I've altered the hypercall_table and hypercall_args_table to have an
additional entry in xen/arch/x86/x86_32/entry.S and in
You also need to edit xen/arch/x86/x86_64/compat/entry.S, for the case
where Xen is 64-bit and dom0 kernel is 32-bit.
Post by John Backes
static int hyper_init(void){
int output;
printk(KERN_ALERT "Testing Hypercall\n");
__asm__ ( "movl $39, %%eax;"
"int $0x82;"
: "=a" (output)
);
While this should work, you probably ought to be using the hypercall
page (and the existing kernel mechanisms) to make hypercalls.
Post by John Backes
return SUCCESS;
}
static void hyper_exit(void){
printk(KERN_ALERT "Removing Hypercall Module");
}
module_init(hyper_init);
module_exit(hyper_exit);
...........................
I then run "xm dmesg" to see if I can see the "NEW HYPERCALL RECEIVED"
message, but nothing appears. Any thoughts?
You could print the return value from the hypercall in your module?

Tim.
--
Tim Deegan <***@xen.org>
Principal Software Engineer, Xen Platform Team
Citrix Systems UK Ltd. (Company #02937203, SL9 0BG)
John Backes
2011-08-15 15:00:41 UTC
Permalink
I changed the table entries in both xen/arch/x86/x86_32/entry.S and
xen/arch/x86/x86_64/entry.S. Could you point me towards these existing
kernel mechanisms for issuing hypercalls? Thanks.

- John
Post by Tim Deegan
Post by John Backes
So I greped through the s
I've altered the hypercall_table and hypercall_args_table to have an
additional entry in xen/arch/x86/x86_32/entry.S and in
You also need to edit xen/arch/x86/x86_64/compat/entry.S, for the case
where Xen is 64-bit and dom0 kernel is 32-bit.
Post by John Backes
static int hyper_init(void){
int output;
printk(KERN_ALERT "Testing Hypercall\n");
__asm__ ( "movl $39, %%eax;"
"int $0x82;"
: "=a" (output)
);
While this should work, you probably ought to be using the hypercall
page (and the existing kernel mechanisms) to make hypercalls.
Post by John Backes
return SUCCESS;
}
static void hyper_exit(void){
printk(KERN_ALERT "Removing Hypercall Module");
}
module_init(hyper_init);
module_exit(hyper_exit);
...........................
I then run "xm dmesg" to see if I can see the "NEW HYPERCALL RECEIVED"
message, but nothing appears. Any thoughts?
You could print the return value from the hypercall in your module?
Tim.
Tim Deegan
2011-08-15 15:10:09 UTC
Permalink
Hi,

Please don't top-post.
Post by John Backes
I changed the table entries in both xen/arch/x86/x86_32/entry.S and
xen/arch/x86/x86_64/entry.S.
Yes; you also need to edit xen/arch/x86/x86_64/compat/entry.S
Post by John Backes
Could you point me towards these existing
kernel mechanisms for issuing hypercalls? Thanks.
In linux: arch/x86/include/asm/xen/hypercall.h

Cheers,

Tim.
Post by John Backes
Post by Tim Deegan
Post by John Backes
So I greped through the s
I've altered the hypercall_table and hypercall_args_table to have an
additional entry in xen/arch/x86/x86_32/entry.S and in
You also need to edit xen/arch/x86/x86_64/compat/entry.S, for the case
where Xen is 64-bit and dom0 kernel is 32-bit.
Post by John Backes
static int hyper_init(void){
int output;
printk(KERN_ALERT "Testing Hypercall\n");
__asm__ ( "movl $39, %%eax;"
"int $0x82;"
: "=a" (output)
);
While this should work, you probably ought to be using the hypercall
page (and the existing kernel mechanisms) to make hypercalls.
Post by John Backes
return SUCCESS;
}
static void hyper_exit(void){
printk(KERN_ALERT "Removing Hypercall Module");
}
module_init(hyper_init);
module_exit(hyper_exit);
...........................
I then run "xm dmesg" to see if I can see the "NEW HYPERCALL RECEIVED"
message, but nothing appears. Any thoughts?
You could print the return value from the hypercall in your module?
Tim.
--
Tim Deegan <***@xen.org>
Principal Software Engineer, Xen Platform Team
Citrix Systems UK Ltd. (Company #02937203, SL9 0BG)
Continue reading on narkive:
Loading...