Re: [Xen-devel] [PATCH 2 of 4] get_page_type: Print out extra information when failing to get page_type

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 22.08.12 at 16:24, Konrad Rzeszutek Wilk <konrad@xxxxxxxxxx> wrote:
> On Wed, Aug 22, 2012 at 03:14:41PM +0100, Jan Beulich wrote:
>> >>> On 21.08.12 at 22:08, Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx> 
>> >>> wrote:
>> > # HG changeset patch
>> > # User Konrad Rzeszutek Wilk <konrad.wilk@xxxxxxxxxx>
>> > # Date 1345579709 14400
>> > # Node ID 8ed3eef706710c9c476a8d984bfb2861d92bedfb
>> > # Parent  635917c6dac4ab8748572fcbeb3e745428684e15
>> > get_page_type: Print out extra information when failing to get page_type.
>> > 
>> > When any reference to __get_page_type is called and it fails, we get:
>> > 
>> > (XEN) mm.c:2429:d0 Bad type (saw 1400000000000002 != exp 7000000000000000) 
>> > for mfn 10e392 (pfn 1bf6c)
>> > 
>> > with this patch we get some extra details such as:
>> > (XEN) debug.c:127:d0 cr3: 10d80b000, searching for 10e392
>> > (XEN) debug.c:111:d0 cr3(10d80b000) has mfn(10e392) in level PUD/L3: 
> [258][272]
>> > (XEN) debug.c:111:d0 cr3(10d80b000) has mfn(10e392) in level PMD/L2: 
> [258][511][511]
>> > (XEN) debug.c:111:d0 cr3(10d80b000) has mfn(10e392) in level PGD/L4: [272]
>> > (XEN) debug.c:111:d0 cr3(10d80b000) has mfn(10e392) in level PUD/L3: 
> [511][511]
>> > 
>> > where it actually is in the pagetable of the guest. This is useful
>> > b/c we can figure out where it is, and use that to figure out where
>> > the OS thinks it is.
>> 
>> In addition to my earlier reply, I also think that the printing
>> should be done at info level, so that nothing would get
>> additionally printed without special command line options.
> 
> I will be more than happy to make those changes. However I think
> you are feeling that this is not really that useful? Perhaps if I spiced
> it up by also dumping what those 'types' are and added some
> rudimentary logic troubleshooting (its an L2 type, and you are trying to
> add it as an L1 entry as writeable, so on..)

No, it doesn't need to go that far I think. My main concern is that
if something like that should get added, it should reduce the need
to consult the sources to understand the messages. Currently I'm
rather suspecting the inverse.

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel