Re: [Qemu-devel] [PATCH] Flush pending AIO on reboot and shutdown.

[Gleb Natapov]

On Wed, Aug 13, 2008 at 03:29:04PM +0100, Samuel Thibault wrote:
> Gleb Natapov, le Wed 13 Aug 2008 17:25:58 +0300, a écrit :
> > On Wed, Aug 13, 2008 at 03:06:41PM +0100, Samuel Thibault wrote:
> > > Gleb Natapov, le Wed 13 Aug 2008 16:59:41 +0300, a écrit :
> > > > On Wed, Aug 13, 2008 at 02:46:08PM +0100, Samuel Thibault wrote:
> > > > > Gleb Natapov, le Wed 13 Aug 2008 16:23:58 +0300, a écrit :
> > > > > > If there is outstanding IDE IO when BIOS starts execution then IDE
> > > > > > commands sent by BIOS will interfere with it and will leave IDE
> > > > > > subsystem in unpredictable state. This can happen when system 
> > > > > > reboots
> > > > > > unexpectedly without waiting for IO completion. Flushing IO before 
> > > > > > exit
> > > > > > prevents data lose.
> > > > > 
> > > > > I'm wondering: isn't that what happens with real machines?
> > > > > 
> > > > With shutdown yes, but why not try harder. With reboot I don't think
> > > > that in real HW you can hang IDE interface after reboot ;)
> > > 
> > > Mmm, I couldn't understand.
> > > 
> > > With real hardware, if you reboot into the bios the board is not
> > > resetted either, and thus the interference is the same.
> > > 
> > On real hardware if you press reset button
> 
> That's when you press the reset button.  When you have a triple fault
> the IDE board won't be resetted.
> 
As far as I see qemu doesn't handle triple faults yet and when it will
handle it we will flush aio there too.

> > Currently BOCHS BIOS doesn't check that IDE is busy (it doesn't even
> > know that HW is present) and sends overlapping command.
> 
> Then fix it.
> 
Wow you found one more reason to do nothing in qemu. Congratulations.
The real HW are much more complex then what qemu is implementing. The
real HW can even save you buffers after you unplugged your computer. And
for real HW it takes much less time to do it too.

--
                        Gleb.