Re: vmx status report against changeset 15521 - 2 new issues

[Mats Petersson <mats@planetcatfish.com>]

At 08:31 13/07/2007, Steven Hand wrote:

> >>>1) Data corrupted after copied into a 32bit win2k3 guest on 64bit
> >>>hypervisor with rtl8139
> >>>http://bugzilla.xensource.com/bugzilla/show_bug.cgi?id=3D1025
> >>
> >>Have you tried this with:
> >>   - 32bit windows guest on 32bit or PAE hypervisor?
> >>   - 64bit windows guest on 64bit hypervisor?
> >[Yunfeng] I tried 32bit/pae and 64bit/64bit, and I cannot reproduce this
> >issue on them.
>
>Ok, that's interesting.
>
> >>It'd also be good to know what the extent of the corruption is, e.g.
> >>if it affects just certain pages, etc.
> >[Yunfeng] Do you have any good way to get this kind of info?
>
>Just comparing the source and destination file page by page (or byte
>by byte) and working out the ranges of differences?


Have you tried filling the buffer in QEMU with a known pattern - from 
the description Steven gave, it looks like it may be using an "old" 
buffer, so after the buffer has been "emptied", could you fill it 
with something, and see if that's what occurs in the "bad packet"? 
[Admittedly, I don't really know if there's any easy way to know when 
the packet has actually been "consumed" by the driver in Windows!).

It also looked like the data in the file was code or something like 
that - have you tried preparing a file where it's easy to see that 
the bytes are in the correct order (e.g. fill a file with longwords 
0, 1, 2, 3, 4, and so on). Then you should be able to see if it's a 
"newer" or "older" packet that is incorrectly inserted (I presume).

--
Mats


>(To make this even easier, start with a known, homogenous file (e.g.
>all 0xa5 or something) - on receipt check md5sum of each page, and
>note any differences in pages which fail the check).
>
>cheers,
>
>S.
>
>
>_______________________________________________
>Xen-devel mailing list
>Xen-devel@lists.xensource.com
>http://lists.xensource.com/xen-devel