Re: [Xen-devel] Windows Bug Check 0x101 issue

[Keir Fraser <keir.fraser@xxxxxxxxxxxxx>]

On 24/3/08 09:53, "Kouya Shimura" <kouya@xxxxxxxxxxxxxx> wrote:

> An attached patch fixes it. However I think the root cause is
> that a timer event can't interrupt an i/o emulation.
> How should we fix it?

The attached patch also makes the fsync asynchronous, which seems a bit
dangerous if we acknowledge the flush to the guest meanwhile. If whatever
happens on return from raw_flush() could be deferred to aio_fsync_cb() then
this might be a good fix.

In general timer events *cannot* interrupt i/o emulation, as we cannot
really take an interrupt in the middle of executing an instruction. That
would be a mess!

However in this case I guess the IDE/SCSI device model could allow
asynchronous reporting of flush completion to the guest, and this new aspect
of the device model would obviously tie into your patch, being done on
aio_fsync_cb().

 -- Keir



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