Re: [Xen-devel] [RFC Patch v3 01/18] copy the correct page to memory

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 08/09/14 12:27, Ian Campbell wrote:
> On Fri, 2014-09-05 at 17:10 +0800, Wen Congyang wrote:
>> From: Hong Tao <bobby.hong@xxxxxxxxxx>
>>
>> apply_batch() only handles MAX_BATCH_SIZE pages at one time. If
>> there is some bogus/unmapped/allocate-only/broken page, we will
>> skip it. So when we call apply_batch() again, the first page's
>> index is curbatch - invalid_pages. invalid_pages stores the number
>> of bogus/unmapped/allocate-only/broken pages we have found.
>>
>> In many cases, invalid_pages is 0, so we don't catch this error.
>>
>> Signed-off-by: Hong Tao <bobby.hong@xxxxxxxxxx>
>> Signed-off-by: Wen Congyang <wency@xxxxxxxxxxxxxx>
> Acked-by: Ian Campbell <ian.campbell@xxxxxxxxxx>
>
> Hopefully all of this stuff will soon be replaced by migration v2 but we
> may as well apply this fix in the meantime. CCing Andy C on the unlikely
> chance that he carried this misbehaviour over into v2.

I rewrote all of this from scratch.

Amongst other things, the v2 code doesn't batch explicitly batch on the
restoring side.  It will deal with an individual PAGE_DATA record at
once, which will generally contain 1024 pages of data, but will not
cache some parts of this to batch better with the following PAGE_DATA
record, which appears to be the source of this bug.

As for deletion.  I am planning to delete it as part of the v2 series,
although that might get complicated if remusv2 is not ready by that point.

~Andrew


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