2.0-testing.bk

2.0-testing.bk

[Ian Pratt]

Folks,

It would be good if you could give 2.0-testing a try out as we're
close to releasing it as 2.0.4.

The main differences since 2.0.3:
  * Disabling the debugging that was accidentaly enabled in 2.0.2
that caused a drastic performance slow down
  * Move contents of /var/xen/ to /var/lib/xen as per LSB-FHS
  * fix network scripts not to be confused by ipv6
  * misc bug fixes (e.g. to VM restore code, confusion with >4GB
mem)


Thanks,
Ian




-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Nicholas Berry]

>>> Ian Pratt <Ian.Pratt@cl.cam.ac.uk> 01/25/05 08:17AM >>>

> Folks,
>
> It would be good if you could give 2.0-testing a try out as we're
> close to releasing it as 2.0.4.
> 
> The main differences since 2.0.3:
>   * Disabling the debugging that was accidentaly enabled in 2.0.2
> that caused a drastic performance slow down
>   * Move contents of /var/xen/ to /var/lib/xen as per LSB-FHS
>   * fix network scripts not to be confused by ipv6
>   * misc bug fixes (e.g. to VM restore code, confusion with >4GB
> mem)
>
>
> Thanks,
> Ian

Is it just me, or is the INSTALL_DIR definition in the Makefile
completely bogus?

make install-kernels

fails on:

cp -a install -d -m0755/boot/* /boot/

Changing INSTALL_DIR to the earlier definition fixed it for me.

Nik



**********************************************************
Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues.


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Ian Pratt]

> 
> Is it just me, or is the INSTALL_DIR definition in the Makefile
> completely bogus?

Grrr. It's totally bogus. Hopefully now fixed in 2.0-testing.

Ian



-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Keir Fraser]

> Is it just me, or is the INSTALL_DIR definition in the Makefile
> completely bogus?
> 
> make install-kernels
> 
> fails on:
> 
> cp -a install -d -m0755/boot/* /boot/
> 
> Changing INSTALL_DIR to the earlier definition fixed it for me.
> 
> Nik

Some of the recent build system 'cleanups' were broken. I've just done
some testing and sanitisation.

 -- Keir


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Paul Larson]

[-- Attachment #1: Type: text/plain, Size: 784 bytes --]

Am I mistaken, or didn't this used to build, even if you don't have
latex and such things that are needed to build the documentation?  It's
not a big deal to install it, but was this change intentional, or is it
preferred that things like this not cause a build failure?

*************************************************
*************************************************
* WARNING: Package 'latex' is required
*          to build Xen documentation
*************************************************
*************************************************
make[1]: *** [docs] Error 1
make[1]: Leaving directory
`/home/plars/cvs/xen/xentest/xen-2.0-testing.bk'
make: *** [world] Error 2

-- 
Thanks,
Paul Larson
plars@linuxtestproject.org
http://www.linuxtestproject.org

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.0-testing.bk

[Keir Fraser]

Yes, indeed it did. I actually have now fixed it again as part of my
re-sanitisation of the build system earlier today. I hope the build
system will now be left alone. ;-)

Let me know if you have the same problem even with the latest patch.

 -- Keir

> Am I mistaken, or didn't this used to build, even if you don't have
> latex and such things that are needed to build the documentation?  It's
> not a big deal to install it, but was this change intentional, or is it
> preferred that things like this not cause a build failure?
> 
> *************************************************
> *************************************************
> * WARNING: Package 'latex' is required
> *          to build Xen documentation
> *************************************************
> *************************************************
> make[1]: *** [docs] Error 1
> make[1]: Leaving directory
> `/home/plars/cvs/xen/xentest/xen-2.0-testing.bk'
> make: *** [world] Error 2
> 
> -- 
> Thanks,
> Paul Larson
> plars@linuxtestproject.org
> http://www.linuxtestproject.org
>    This is a digitally signed message part


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Paul Larson]

[-- Attachment #1: Type: text/plain, Size: 736 bytes --]

The previous build failure when latex not installed seems to have been
fixed.  However the patch from yesterday to create symlinks to the
kernels in /boot is circumvented now that the files aren't copied from
dist/install/* as before.

-install-kernels:
-	cp -a $(DESTDIR)/boot/* /boot/
-	cp -a $(DESTDIR)/lib/modules/* /lib/modules/
-	cp -dR $(DESTDIR)/boot/*$(LINUX_VER)* $(prefix)/boot/
-	cp -dR $(DESTDIR)/lib/modules/* $(prefix)/lib/modules/

I can easily work around this by doing make dist and calling install.sh
manually, but if anyone wanted the symlinks to be created when doing a
make install, that won't happen anymore.

-- 
Thanks,
Paul Larson
plars@linuxtestproject.org
http://www.linuxtestproject.org

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: 2.0-testing.bk

[Kip Macy]

What changeset should we be up to? I'm currently running 1.1666 which
you committed yesterday evening.

Frustrated that xm save for freebsd would give me:
(XEN) DOM0: (file=memory.c, line=425) Non-privileged attempt to map I/O
space 00000000

I tried 'xm save 0 foo' - which I'm not sure should be allowed - and
saw the following:

Unable to handle kernel NULL pointer dereference at virtual address
00000000
 printing eip:
c0298b6a
*pde = ma 00000000 pa 55555000
 [<c01069f3>] __do_suspend+0x43/0x1e0
 [<c0117906>] __wake_up+0x46/0xa0
 [<c012b96f>] worker_thread+0x22f/0x340
 [<c0106c60>] __shutdown_handler+0x0/0x50
 [<c0117840>] default_wake_function+0x0/0x20
 [<c0117840>] default_wake_function+0x0/0x20
 [<c012b740>] worker_thread+0x0/0x340
 [<c013009a>] kthread+0xaa/0xb0
 [<c012fff0>] kthread+0x0/0xb0
 [<c01076b5>] kernel_thread_helper+0x5/0x10
Oops: 0000 [#1]
PREEMPT DEBUG_PAGEALLOC
Modules linked in:
CPU:    0
EIP:    0061:[<c0298b6a>]    Not tainted VLI
EFLAGS: 00011296   (2.6.10-xen0)
EIP is at netif_suspend+0xa/0x40
eax: fbffc000   ebx: 00000000   ecx: 00000001   edx: ce57e000
esi: c0101000   edi: c05e8000   ebp: 00000000   esp: c05e9f0c
ds: 007b   es: 007b   ss: 0069
Process events/0 (pid: 3, threadinfo=c05e8000 task=c05cbb20)
Stack: c013c69f c043ef40 c01069f3 c05e9f40 c0117906 c05caf10 00002a8a
2e981339
       00004a2e ce57e000 c043ef40 00000000 c05e8000 00000000 c012b96f
00000000
       c05e9f74 00000000 c05caf10 c05e8000 c05e8000 c0106c60 c05e8000
c05caf00
Call Trace:
 [<c013c69f>] __get_free_pages+0x1f/0x40
 [<c01069f3>] __do_suspend+0x43/0x1e0
 [<c0117906>] __wake_up+0x46/0xa0
 [<c012b96f>] worker_thread+0x22f/0x340
 [<c0106c60>] __shutdown_handler+0x0/0x50
 [<c0117840>] default_wake_function+0x0/0x20
 [<c0117840>] default_wake_function+0x0/0x20
 [<c012b740>] worker_thread+0x0/0x340
 [<c013009a>] kthread+0xaa/0xb0
 [<c012fff0>] kthread+0x0/0xb0
 [<c01076b5>] kernel_thread_helper+0x5/0x10
Code: 89 54 24 10 8b 5c 24 04 8b 7c 24 08 83 c4 0c e9 3d f7 ff ff 90 8d
b6 00
00 00 8d bf 00 00 00 00 53 83 ec 04 8b 1d 40 bc 53 c0 <8b> 03 0f 18 00
90 81 f
0 bc 53 c0 74 20 90 8d b4 26 00 00 00



Why does xen think that an unprivileged guest has a reference to address
0 with the valid bit set?


			-Kip

On Tue, 25 Jan 2005, Keir Fraser wrote:

> > Is it just me, or is the INSTALL_DIR definition in the Makefile
> > completely bogus?
> >
> > make install-kernels
> >
> > fails on:
> >
> > cp -a install -d -m0755/boot/* /boot/
> >
> > Changing INSTALL_DIR to the earlier definition fixed it for me.
> >
> > Nik
>
> Some of the recent build system 'cleanups' were broken. I've just done
> some testing and sanitisation.
>
>  -- Keir
>
>
> -------------------------------------------------------
> This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
> Tool for open source databases. Create drag-&-drop reports. Save time
> by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
> Download a FREE copy at http://www.intelliview.com/go/osdn_nl
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/xen-devel
>


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Keir Fraser]

> What changeset should we be up to? I'm currently running 1.1666 which
> you committed yesterday evening.
> 
> Frustrated that xm save for freebsd would give me:
> (XEN) DOM0: (file=memory.c, line=425) Non-privileged attempt to map I/O
> space 00000000

Probably some structure not set up correctly -- my guess might be the
phys-to-machine mapping info that is supposed to get poked into the
shared_info page by the unprivileged guest. It is a list of the page
frames that make up the guest's p2m table. If you haven't init'ed it
then the xm_save tool will probably try to map NULL and hence the
error (Xen has low 1MB of memory marked as being an I/O page).

> I tried 'xm save 0 foo' - which I'm not sure should be allowed - and
> saw the following:

There should be a safety catch on this. DOM0 is definitely not
saveable! 

 -- Keir


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Kip Macy]

> Probably some structure not set up correctly -- my guess might be the
> phys-to-machine mapping info that is supposed to get poked into the
> shared_info page by the unprivileged guest. It is a list of the page
> frames that make up the guest's p2m table. If you haven't init'ed it
> then the xm_save tool will probably try to map NULL and hence the
> error (Xen has low 1MB of memory marked as being an I/O page).

Thanks. I'll look there. I assumed it was looking at the page tables
themselves (which shouldn't be a problem because xen shouldn't allow it
in the first place.


> There should be a safety catch on this. DOM0 is definitely not
> saveable!

I would think not - but it doesn't appear to realize that ;-).


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl

Re: 2.0-testing.bk

[Kip Macy]

> Probably some structure not set up correctly -- my guess might be the
> phys-to-machine mapping info that is supposed to get poked into the
> shared_info page by the unprivileged guest. It is a list of the page
> frames that make up the guest's p2m table. If you haven't init'ed it
> then the xm_save tool will probably try to map NULL and hence the
> error (Xen has low 1MB of memory marked as being an I/O page).

Thanks. I added the following:
    HYPERVISOR_shared_info->arch.pfn_to_mfn_frame_list = (unsigned long)xen_phys_machine;
    HYPERVISOR_shared_info->arch.mfn_to_pfn_start = (unsigned long)xen_machine_phys;

xen_phys_machine is start_info->mfn_list an xen_machine_phys is
(1008 << 22)

I'm now seeing:
(XEN) (file=/t/niners/users/xen/xen-2.0-testing.bk/xen/include/asm/mm.h,
line=157) Error pfn 00073000: ed=fc5f8040, sd=00000000, caf=00000000,
taf=0000000a

I know that comes from this:
        if ( unlikely((x & PGC_count_mask) == 0) ||  /* Not allocated?
*/
             unlikely((nx & PGC_count_mask) == 0) || /* Count overflow?
*/
             unlikely(d != domain) )                 /* Wrong owner? */
        {
            DPRINTK("Error pfn %08lx: ed=%p, sd=%p, caf=%08x,
taf=%08x\n",
                    page_to_pfn(page), domain, d,
                    x, page->u.inuse.type_info);
            return 0;
        }

But maybe you'll know off the top of your head what is wrong?


			-Kip

>
> > I tried 'xm save 0 foo' - which I'm not sure should be allowed - and
> > saw the following:
>
> There should be a safety catch on this. DOM0 is definitely not
> saveable!
>
>  -- Keir
>


-------------------------------------------------------
This SF.Net email is sponsored by: IntelliVIEW -- Interactive Reporting
Tool for open source databases. Create drag-&-drop reports. Save time
by over 75%! Publish reports on the web. Export to DOC, XLS, RTF, etc.
Download a FREE copy at http://www.intelliview.com/go/osdn_nl