Re: [win-pv-devel] DRIVER_IRQL_NOT_LESS_OR_EQUAL in xenbus.sys

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: Michael Stern [mailto:mhstar89@xxxxxxxxx]
> Sent: 03 December 2016 16:00
> To: Paul Durrant <Paul.Durrant@xxxxxxxxxx>; win-pv-
> devel@xxxxxxxxxxxxxxxxxxxx
> Subject: Re: [win-pv-devel] DRIVER_IRQL_NOT_LESS_OR_EQUAL in
> xenbus.sys
> 
> On Fri, Nov 11, 2016 at 3:56 PM, Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> wrote:
> > De-htmling...
> >
> > -----
> > From: win-pv-devel [mailto:win-pv-devel-bounces@xxxxxxxxxxxxxxxxxxxx]
> On Behalf Of Michael Stern
> > Sent: 11 November 2016 14:52
> > To: win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> > Subject: [win-pv-devel] DRIVER_IRQL_NOT_LESS_OR_EQUAL in xenbus.sys
> >
> > I'm running a Windows 2016 VM under Xen, which works great - however,
> the moment I try to install Microsoft SQL Server 2016, Windows will crash
> with a BSOD due to DRIVER_IRQL_NOT_LESS_OR_EQUAL in xenbus.sys.
> >
> > Here's what I found out:
> > - The bug is perfectly reproducible, it'll happen every time, and it'll 
> > happen
> within seconds of the attempted install.
> > - If the SQL Server 2016 Setup is started while xenbus.sys is not installed,
> the install works fine, but obviously the performance of the server is very
> poor without the PV drivers.
> > - This happens with the signed drivers, and with the bleeding edge drivers.
> > - This is not limited to Windows 2016 and SQL Server 2016, installing the 
> > SP3
> of SQL Server 2012 on a Windows 2008 Server will also result in a BSOD in
> xenbus.sys.
> >
> > Do you have any pointers as to what I could try? Would you be interested
> in a memory dump, or some other information? I can setup a pristine
> Windows 2016 Server with all the files in place to cause the crash.
> >
> > Host system is a Xen 4.7.0 on Ubuntu 16.04 (however keep in mind that
> without the PV drivers, the install works without hiccups).
> > -----
> >
> > Yes, please send me a memory dump if you can. Also, the drivers log to
> QEMU so if you can grab the QEMU log too then that may yield more info.
> >
> > Thanks,
> >
> >   Paul
> 
> Paul,
> 
> I'm still trying to get a memory dump (currently the machine simply
> hangs without crashing), but the QEMU log has these lines at the end:
> 
> XENIFACE|SessionExecuteLog: USER: Guest agent lite main loop starting
> XENIFACE|SessionExecuteLog: USER: Set time to XenTime
> XENIFACE|SessionExecuteLog: USER: hosttimeIsUTC: Open Registry Key
> XENIFACE|SessionExecuteLog: USER: Time is now  2016.12.3 10:23:29.558
> XENIFACE|StartWatch: Start Watch FFFFDB0071914250
> XENIFACE|SessionExecuteLog: USER: Set time to 2016.12.3 10:23:29.630
> XENIFACE|SessionExecuteLog: USER: About to add feature shutdown
> XENIFACE|SessionExecuteLog: USER: Establish watch control/shutdown
> XENIFACE|StartWatch: Start Watch FFFFDB00718C6C20
> XENIFACE|SessionExecuteLog: USER: About to advertise features
> XENIFACE|SessionExecuteLog: USER: WmiSessionSetEntry:MethodExec
> Failed
> XENIFACE|SessionExecuteLog: USER: WmiSessionSetEntry:SessionStart
> Failed
> XENIFACE|SessionExecuteLog: USER: WmiSessionSetEntry:SetVValue Failed
> XENIFACE|SessionExecuteLog: USER: WmiSessionSetEntry:SetVPath Failed
> XENIFACE|SessionExecuteLog: USER: Failed to advertise shutdown
> XENIFACE|SessionExecuteLog: USER: About to kick xapi
> XENIFACE|SessionExecuteLog: USER: win agent going to sleep
> XENIFACE|SessionExecuteLog: USER: win agent woke up for 3
> XENIFACE|SessionExecuteLog: USER: Fire 0000027EDFABBEF0
> XENIFACE|SessionExecuteLog: USER: fire feature shutdown
> XENIFACE|SessionExecuteLog: USER: Check if we need to shutdown
> XENIFACE|SessionExecuteLog: USER: No need to shutdown
> XENIFACE|SessionExecuteLog: USER: fired feature shutdown
> XENIFACE|SessionExecuteLog: USER: win agent going to sleep
> XENBUS|RangeSetPop: fail1 (c000009a)
> GNTTAB: MAP XENMAPSPACE_grant_table[7] @ 00000000.f2008000
> XENBUS|GnttabExpand: added references [00000e00 - 00000fff]
> XENVBD|PdoCompleteResponse:Target[0] : DISCARD
> BLKIF_RSP_EOPNOTSUPP (Tag 18098)
> XENDISK|PdoSendTrimSynchronous:fail2
> XENDISK|PdoSendTrimSynchronous:fail1 (c0000185)
> 
> Storage is LVM2 on a MD RAID1.
> Once the system hangs, I find the bottom three lines (starting with
> DISCARD BLKIF_RSP_EOPNOTSUPP) in the log every time.
> 

I think these somewhat scary log lines are expected if the backend storage does 
not support discard. (The blkif protocol is a bit crazy so a frontend has to 
assume discard is supported until it gets an EOPNOTSUPP from the backend). I'll 
check the code to make sure the error is not incorrectly propagated... I'm 
suspicious of that 'fail1' message as it suggests that the error is being 
propagated. I also think that, since this is expected, the log messages should 
be much less scary.

  Paul

> Michael
_______________________________________________
win-pv-devel mailing list
win-pv-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/cgi-bin/mailman/listinfo/win-pv-devel