[vbox-dev] [RESOLVED] vbox bug causing Windows system cache problems?

Thu Sep 29 18:38:33 GMT 2011

Huihong,

unfortunately this isn't really a fix. There are two code paths in the HGCM
code to share data between guest and host. The VBGLR0_CAN_USE_PHYS_PAGE_LIST
flag is set if the guest can use lists of physical pages. If you remove that
condition, the old (slower) code is used. This is only a workaround but we
have to find the correct fix for the real bug. Though this workaround might
help you for the time being.

Thanks for your help anyway,

Frank

On Thursday 29 September 2011 19:38:05 Huihong Luo wrote:
> r36201 caused this issue, kind of weird.
>  
> VMMDev.cpp
>  
> this condition:
>  
> && requestHeader.requestType != VMMDevReq_GetHostVersion
>  
> After I removed this condition from latest svn code, the cache error no
> longer displays. 
> Is there any special call to make from guest addition to ensure the above
> condition?
> 
> --- On Thu, 9/29/11, Huihong Luo <huisinro at yahoo.com> wrote:
> 
> 
> From: Huihong Luo <huisinro at yahoo.com>
> Subject: Re: [vbox-dev] vbox bug causing Windows system cache problems?
> To: "Vitali Pelenjow" <Vitali.Pelenjow at oracle.com>
> Cc: vbox-dev at virtualbox.org
> Date: Thursday, September 29, 2011, 8:32 AM
> 
> 
> 
> 
> 
> 
> 
> Vitali,
>  
> No problem, I've been doing that. Now narrowed down to r36210 - r36190,
> where r36210 has the bug, but r36190 works. 
> I will go further. Most likely a bug in vmm.
>  
> - huihong
> 
> --- On Thu, 9/29/11, Vitali Pelenjow <Vitali.Pelenjow at oracle.com> wrote:
> 
> 
> From: Vitali Pelenjow <Vitali.Pelenjow at oracle.com>
> Subject: Re: [vbox-dev] vbox bug causing Windows system cache problems?
> To: "Huihong Luo" <huisinro at yahoo.com>
> Cc: vbox-dev at virtualbox.org
> Date: Thursday, September 29, 2011, 1:58 AM
> 
> 
> Hi Huihong,
> 
> we would appreciate, if you could narrow down the changeset, which
> introduced the regression.
> This would be a great help. The changeset range is still too wide to
> have a guess.
> 
> Thanks,
> Vitali.
> 
> Huihong Luo wrote:
> > After last few days' buiding different OSE revisions, the bug exists
> > between
> > r36100 - r36222 (Feb 28, 11 - Mar-09, 11)
> > if you have some wild guess on which code change may have caused that,
> > let me know, to save some time.
> > Otherwise I will continue the brutal force way of narrowing down the
> > code set that caused the bug.
> >
> > --- On *Tue, 9/27/11, Huihong Luo /<huisinro at yahoo.com>/* wrote:
> >     From: Huihong Luo <huisinro at yahoo.com>
> >     Subject: Re: [vbox-dev] vbox bug causing Windows system cache
> >     problems?
> >     To: vbox-dev at virtualbox.org
> >     Date: Tuesday, September 27, 2011, 12:33 PM
> >
> >     After some backing tracing, this error occurs from vbox 4.1.0
> >     r73009, and works well with 4.0.12 and before.
> >     The windows cache manager displays an error message as attached
> >     from our driver.
> >     We run exactly the same guest (same disk, same configure) using
> >     different versions of vbox downloaded from virtualbox.org.
> >     so very likely there is a bug in vbox between 4.0.12 and 4.1.0
> >
> >     --- On *Mon, 9/26/11, Huihong Luo /<huisinro at yahoo.com>/* wrote:
> >
> >
> >         From: Huihong Luo <huisinro at yahoo.com>
> >         Subject: [vbox-dev] vbox bug causing Windows system cache
> >         problems?
> >         To: vbox-dev at virtualbox.org
> >         Date: Monday, September 26, 2011, 6:47 PM
> >
> >         After one week of debugging, there might be a bug in vbox that
> >         causes Windows MmFlushSection() to behave weirdly.
> >         We have a Windows driver working very well on old versions of
> >         vbox, but not on v4. The problem was caused by MmFlushSection,
> >         which wrongly updates files that are opened for read. In other
> >         words, MmFlushSection() causes files to write even if they are
> >         opened for read, and thus get ACCESS denined error.
> >         It seems somehow the page dirty bit is not set correctly in
> >         some case, because MmFlushSection checks dirty pages.
> >         I first thought there might be some bugs in our driver's code,
> >         but strange thing is that it works fine before with old
> >         versions of vbox, so I post this message just in case
> >         something wrong with vbox code.
> >
> >
> >         -----Inline Attachment Follows-----
> >
> >         _______________________________________________
> >         vbox-dev mailing list
> >         vbox-dev at virtualbox.org
> >     
> >   <http://us.mc1603.mail.yahoo.com/mc/compose?to=vbox-dev@virtualbox.org
> >> http://vbox.innotek.de/mailman/listinfo/vbox-dev
> >
> >
> >     -----Inline Attachment Follows-----
> >
> >     _______________________________________________
> >     vbox-dev mailing list
> >     vbox-dev at www.virtualbox.org
> > 
> >   <http://us.mc1603.mail.yahoo.com/mc/compose?to=vbox-dev@www.virtualbox
> >.org> https://www.virtualbox.org/mailman/listinfo/vbox-dev
> >
> > _______________________________________________
> > vbox-dev mailing list
> > vbox-dev at virtualbox.org
> > https://www.virtualbox.org/mailman/listinfo/vbox-dev

-- 
Dr.-Ing. Frank Mehnert
Senior Manager Software Development Desktop Virtualization, VirtualBox
ORACLE Deutschland B.V. & Co. KG | Werkstr. 24 | 71384 Weinstadt, Germany

Hauptverwaltung: Riesstr. 25, D-80992 München
Registergericht: Amtsgericht München, HRA 95603

Komplementärin: ORACLE Deutschland Verwaltung B.V.
Hertogswetering 163/167, 3543 AS Utrecht, Niederlande
Handelsregister der Handelskammer Midden-Niederlande, Nr. 30143697
Geschäftsführer: Jürgen Kunz, Marcel van de Molen, Alexander van der Ven
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://www.virtualbox.org/pipermail/vbox-dev/attachments/20110929/8a918155/attachment.sig>