Re: [win-pv-devel] Next signed release - support for readonly disks?

[Paul Durrant <Paul.Durrant@xxxxxxxxxx>]

> -----Original Message-----
> From: win-pv-devel [mailto:win-pv-devel-bounces@xxxxxxxxxxxxxxxxxxxx] On
> Behalf Of Chris McClymont
> Sent: 18 October 2017 00:40
> To: win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> Subject: Re: [win-pv-devel] Next signed release - support for readonly disks?
> 
> > In my experience an NTFS filesystem requires a disk to be read/write
> otherwise it will not function... certainly this is true of a system disk. 
> Are you
> using a different filesystem? TBH I think it's quite reasonable for Windows to
> assume that a non-removable medium calling itself is a disk is read/write.
> 
> Readonly disks appear to behave as expected for me using Windows
> Server 2012 with the GPLPV drivers signed by EJB Digital, tested using
> both VFAT and NTFS. In the Disk Management UI, the disk appears as
> Read Only where it normally says Online. For some reason, the option
> to add a drive letter is not there, but I can mount them using
> diskpart. Files can be read but trying to paste a file into the
> mounted readonly disk results in an error dialog:
> "The disk is write-protected. ... Remove the write-protection or use
> another disk."  Try Again / Skip / Cancel
> 
> I believe this is achieved by setting the Write Protected flag in the
> SCSI parameters in response to the Sense command.
> It looks like it is set here in the GPLPV code:
> http://xenbits.xensource.com/ext/win-
> pvdrivers/file/2d392ecdd366/xenvbd_common/common_miniport.h#l497

Interesting, thanks for the detail.

> 
> These disks have "EJECTSUPPORTED" and also "SURPRISEREMOVALOK" as
> capabilities (In both GPLPV and this project) so I think they are
> classified as removable.
> 

It's a bit of a strange distinction but I believe that's different from 
removable media and I was wondering whether Windows will cope with 
non-removable media being read-only. From what you say above though, it appears 
it does.

> > This fix was unintended - and it does seem possible to boot with read-only
> non-boot disks attached. It's not a fully tested setup, and the frontend 
> offers
> no protection against writes to read only disks, so the results are undefined
> Yeah, my testing shows that writes appear to work, but after
> refreshing the disk or folder the written files disappear.
> 
> Thanks Paul and Owen, I will either keep using the GPLPV drivers or
> look into feasibility of patching in the SCSI level write protection.
> 

I'm sure one of us can look at patching xenvbd accordingly but, as I said, it 
may be a little while before we do a signed release of the 9.0 drivers.

Cheers,

  Paul

> Regards,
> Chris McClymont
> 
> 
> On Tue, Oct 17, 2017 at 1:46 AM, Paul Durrant <Paul.Durrant@xxxxxxxxxx>
> wrote:
> >> -----Original Message-----
> >> From: win-pv-devel [mailto:win-pv-devel-bounces@xxxxxxxxxxxxxxxxxxxx]
> On
> >> Behalf Of Chris McClymont
> >> Sent: 14 October 2017 02:19
> >> To: win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> >> Subject: [win-pv-devel] Next signed release - support for readonly disks?
> >>
> >> Hi everyone,
> >>
> >> Support for readonly disks was fixed in xenvbd commit
> >> 3bb41b49ff66433ab5ebb5d36d24946f40390d7c. When is this likely to make
> >> it into a signed release?
> >>
> >>
> >> I'm working on Windows templates for Xen and would love to use the
> >> newer Xen Project drivers signed by the Linux Foundation, rather than
> >> the EJB Digital or others floating around. But I have a requirement to
> >> support readonly disks, which don't work in the 8.2.0 release.
> >> Readonly disks appear in device manage only when emulated, and
> >> disappear (even with Show Hidden Devices) after the PV drivers are
> >> installed.
> >>
> >> The commit I mentioned is somewhere between 8.2.0 and master.
> >> Previously, pdo.c would set ScsiStatus to SCSI_ABORT if
> >> (DiskInfo->DiskInfo & VDISK_READONLY). Now, pdo.c is called target.c
> >> and the commit removes the readonly limitation.
> >>
> >
> > Hi Chris,
> >
> > Owen may have more to say, but I assume this was an unintended
> consequence of the patch.
> >
> > In my experience an NTFS filesystem requires a disk to be read/write
> otherwise it will not function... certainly this is true of a system disk. 
> Are you
> using a different filesystem? TBH I think it's quite reasonable for Windows to
> assume that a non-removable medium calling itself is a disk is read/write.
> >
> >> I'm wondering a few things:
> >>   1) When is this commit likely to make it into a signed release
> >
> > It's not currently a back-port candidate and there is no firm timeline for
> releasing 9.0 drivers.
> >
> >>   2) Is there a roadmap? I.e. How is it decided when to make releases
> >> and which features/fixes to include
> >
> > Because the Xen Project does not have dedicated infrastructure for testing
> the PV drivers, we generally rely on the testing done by Citrix XenServer and
> therefore releases tend to be aligned with updates made to the Citrix
> branded variant of the drivers.
> >
> >>   3) How can I search this mailing list? It doesn't seem to be
> >> included on http://xen.markmail.org/search/
> >>
> >
> > You can see the mailing list archives at
> https://lists.xenproject.org/archives/html/win-pv-devel. As for searching via
> markmail... I have requested that the list be added in the past, but I'll have
> another go.
> >
> > Cheers,
> >
> >   Paul
> >
> >> Thanks,
> >> Chris McClymont
> >>
> >> _______________________________________________
> >> win-pv-devel mailing list
> >> win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> >> https://lists.xenproject.org/cgi-bin/mailman/listinfo/win-pv-devel
> 
> _______________________________________________
> win-pv-devel mailing list
> win-pv-devel@xxxxxxxxxxxxxxxxxxxx
> https://lists.xenproject.org/cgi-bin/mailman/listinfo/win-pv-devel
_______________________________________________
win-pv-devel mailing list
win-pv-devel@xxxxxxxxxxxxxxxxxxxx
https://lists.xenproject.org/cgi-bin/mailman/listinfo/win-pv-devel