Xen devel - Nov 2005 - [PATCH][XM-TEST] Test that tries to create a 16MB domain

This patch adds a test that tries to create a domain with only 16MB of
RAM.  It verifies that the domain is created and that the console
works.

Right now this test fails for me: the domain just dies immediately
with no indication of why, or with any useful console data.  Should we
put a sanity check on xm''s create and mem-set commands that prevent
the user from doing something we know won''t work?  Perhaps we can add
a --force flag to allow overrides.

Thoughts?

Signed-off-by: Dan Smith <danms@us.ibm.com>




-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms@us.ibm.com

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

On Wed, Nov 16, 2005 at 10:19:35AM -0800, Dan Smith wrote:
> This patch adds a test that tries to create a domain with only 16MB of
> RAM.  It verifies that the domain is created and that the console
> works.
> 
> Right now this test fails for me: the domain just dies immediately
> with no indication of why, or with any useful console data.  Should we
> put a sanity check on xm''s create and mem-set commands that
prevent
> the user from doing something we know won''t work?  Perhaps we can
add
> a --force flag to allow overrides.
I''ve applied the patch, thanks.

Regarding the failure, perhaps a check or a warning would be a good idea, but
I''m wary of blocking out at certain memory limits.  If someone wrote a
tiny OS
(say the OS equivalent of Hello World) then surely it would work in less than
16MB, so why should we put such a limit?

Ewan.


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

EM> Regarding the failure, perhaps a check or a warning would be a
EM> good idea, but I''m wary of blocking out at certain memory
limits.
EM> If someone wrote a tiny OS (say the OS equivalent of Hello World)
EM> then surely it would work in less than 16MB, so why should we put
EM> such a limit?

I completely agree with you.  The --force flag would be one way around
this.  We could also add it to the xend configuration as an adjustable
limit.  Another option might be to have the limit determined by the
builder that is used, right?

  switch (type) {
    case linux:  limit = 24;
    case fatos:  limit = 128;
    default:     limit = 0;
}
    
Something like that?  I''m not sure off hand if this info is available
after build time for protecting mem-set, but we could make the limit a
property of the domain to solve that easily.

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms@us.ibm.com


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

On 11/16/05, Dan Smith <danms@us.ibm.com> wrote:
> This patch adds a test that tries to create a domain with only 16MB of
> RAM.  It verifies that the domain is created and that the console
> works.
>
> Right now this test fails for me: the domain just dies immediately
> with no indication of why, or with any useful console data.  Should we
> put a sanity check on xm''s create and mem-set commands that
prevent
> the user from doing something we know won''t work?  Perhaps we can
add
> a --force flag to allow overrides.
I routinely create 4M domains (and I would be able to create them even
smaller if the domain builder would allow it), there is event a recent
Xen paper that mentions running ttylinux in this amout of space.

I think this is a problem in XenLinux, it was discussed recently, and
that you can ''fix'' it by setting mem=16M on the linux command
line.

Jacob


--
Save time and bandwidth with EDelta: http://www.diku.dk/~jacobg/edelta/

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

JH> I routinely create 4M domains (and I would be able to create them
JH> even smaller if the domain builder would allow it), there is event
JH> a recent Xen paper that mentions running ttylinux in this amout of
JH> space.

But not with a 4M ramdisk, I''m guessing, which is what xm-test uses :)

JH> I think this is a problem in XenLinux, it was discussed recently,
JH> and that you can ''fix'' it by setting mem=16M on the
linux command
JH> line.

This may be true, however, I am unable to start a 16MB domain with
extra="mem=16M" on x86_64 (x86_32 works).  The domain hangs when
trying to unpack the ramdisk.  I''ll submit a patch that adds the
mem=16M argument to the kernel command line so that it will pass on 32
bit.

If there''s something different about x86_64 that limits the minimum
size of a domain, then I re-argue for some sort of arch-specific soft
limit to "help" the user realize that they are doing something that is
likely to fail.

-- 
Dan Smith
IBM Linux Technology Center
Open Hypervisor Team
email: danms@us.ibm.com


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