"lshw" seems to get stuck at the "DMI" stage with 4.x

"lshw" seems to get stuck at the "DMI" stage with 4.x

[Andy Smith]

Hello,

I just noticed that the "lshw" command gets stuck at the "DMI" stage
for me on all (three of) my Xen 4.x installs. Meanwhile the Xen
dmesg is filling with:

(XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 8000000001592625 for l1e_owner=0, pg_owner=32753
(XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 8000000001592625 for l1e_owner=0, pg_owner=32753

until I ctrl-c the lshw.

Searching for this I found this thread:

http://lists.xensource.com/archives/html/xen-devel/2010-10/msg00257.html

although that is discussing a boot failure (crash) which I am not
experiencing. Alan does mention having to patch lshw to have it
complete the DMI probe.

lshw works fine on my Xen 3.x installs.

Does this mean the bug is in lshw, or in Xen?

Cheers,
Andy

Re: "lshw" seems to get stuck at the "DMI" stage with 4.x

[Jan Beulich]

>>> On 15.08.11 at 04:09, Andy Smith <andy@strugglers.net> wrote:
> I just noticed that the "lshw" command gets stuck at the "DMI" stage
> for me on all (three of) my Xen 4.x installs. Meanwhile the Xen
> dmesg is filling with:
> 
> (XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 
> 8000000001592625 for l1e_owner=0, pg_owner=32753
> (XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 
> 8000000001592625 for l1e_owner=0, pg_owner=32753

Hmm, 32753 is DOM_IO, and Dom0 is generally allowed to access DOM_IO
pages. But the MFN's value as well as it having a valid PFN suggests
that this is actually a RAM page (we'd know for sure only if you provided
the memory map of the machine(s)), which certainly isn't valid to be
accessed in the context of a DMI table access (or any mapping with the
owner set to DOM_IO).

Also you don't state what kernel you use in Dom0.

> until I ctrl-c the lshw.
> 
> lshw works fine on my Xen 3.x installs.

Quite possible that the tighter checking in newer Xen is responsible
for this, so you may want to add some further debugging code to
the code paths involved. But it's also possible that there was a bug
in 3.x accidentally letting the access go through.

Jan

Re: "lshw" seems to get stuck at the "DMI" stage with 4.x

[Ian Campbell]

On Mon, 2011-08-15 at 08:32 +0100, Jan Beulich wrote:
> >>> On 15.08.11 at 04:09, Andy Smith <andy@strugglers.net> wrote:
> > I just noticed that the "lshw" command gets stuck at the "DMI" stage
> > for me on all (three of) my Xen 4.x installs. Meanwhile the Xen
> > dmesg is filling with:
> > 
> > (XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 
> > 8000000001592625 for l1e_owner=0, pg_owner=32753
> > (XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 
> > 8000000001592625 for l1e_owner=0, pg_owner=32753
> 
> Hmm, 32753 is DOM_IO, and Dom0 is generally allowed to access DOM_IO
> pages. But the MFN's value as well as it having a valid PFN suggests
> that this is actually a RAM page (we'd know for sure only if you provided
> the memory map of the machine(s)), which certainly isn't valid to be
> accessed in the context of a DMI table access (or any mapping with the
> owner set to DOM_IO).
> 
> Also you don't state what kernel you use in Dom0.
> 
> > until I ctrl-c the lshw.
> > 
> > lshw works fine on my Xen 3.x installs.
> 
> Quite possible that the tighter checking in newer Xen is responsible
> for this, so you may want to add some further debugging code to
> the code paths involved. But it's also possible that there was a bug
> in 3.x accidentally letting the access go through.

It may also matter which kernel Andy is combining with his 3.x and 4.x
installs. Konrad did some work on improving the handling of identity
mapped regions of the p2m space which could be related to something like
this.

Ian.

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

Re: "lshw" seems to get stuck at the "DMI" stage with 4.x

[Konrad Rzeszutek Wilk]

On Mon, Aug 15, 2011 at 02:09:39AM +0000, Andy Smith wrote:
> Hello,
> 
> I just noticed that the "lshw" command gets stuck at the "DMI" stage
> for me on all (three of) my Xen 4.x installs. Meanwhile the Xen
> dmesg is filling with:
> 
> (XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 8000000001592625 for l1e_owner=0, pg_owner=32753
> (XEN) mm.c:860:d0 Error getting mfn 1592 (pfn 601f) from L1 entry 8000000001592625 for l1e_owner=0, pg_owner=32753
> 
> until I ctrl-c the lshw.

Hm, so I ran this on 3.0 (Ubuntu 11.04) with Xen 4.1.1 and it ran
fine for me. Can you try that kernel instead (just use the stock
3.0 kernel from ftp.kernel.org) and see if you get the same problem?

> 
> Searching for this I found this thread:
> 
> http://lists.xensource.com/archives/html/xen-devel/2010-10/msg00257.html
> 
> although that is discussing a boot failure (crash) which I am not
> experiencing. Alan does mention having to patch lshw to have it
> complete the DMI probe.
> 
> lshw works fine on my Xen 3.x installs.
> 
> Does this mean the bug is in lshw, or in Xen?

I would think Dom0, but we can narrow that down if you run
with 3.0 (which works for me) and see how that fares.
> 
> Cheers,
> Andy
> 
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xensource.com
> http://lists.xensource.com/xen-devel