Re: Hyper-V SSD passthrough fix backport

[Stephen Hemminger <stephen@networkplumber.org>]

On Thu, 7 Dec 2017 07:10:45 +0000
Sitsofe Wheeler <sitsofe@gmail.com> wrote:

> (Forwarding to all Hyper-V maintainers due to lack of response to
> Greg's question.
> 
> There's something about this patch which makes it keep falling through
> the cracks: in March when someone asked for this patch to be
> backported (http://lkml.iu.edu/hypermail/linux/kernel/1703.3/03056.html
> ) no one confirmed it should so Greg wound up dropping it
> (http://lkml.iu.edu/hypermail/linux/kernel/1703.3/04774.html ). Given
> this, can I urge people not to overlook this latest query...)
> 
> ---------- Forwarded message ----------
> From: Greg KH <greg@kroah.com>
> Date: 4 December 2017 at 12:10
> Subject: Re: Hyper-V SSD passthrough fix backport
> To: Stephen Hemminger <stephen@networkplumber.org>
> Cc: Sitsofe Wheeler <sitsofe@gmail.com>, stable
> <stable@vger.kernel.org>, "Martin K. Petersen"
> <martin.petersen@oracle.com>
> 
> 
> On Wed, Nov 29, 2017 at 03:21:08PM -0800, Stephen Hemminger wrote:
> > On Wed, 29 Nov 2017 12:09:03 +0000
> > Sitsofe Wheeler <sitsofe@gmail.com> wrote:
> >  
> > > On 29 November 2017 at 10:40, Sitsofe Wheeler <sitsofe@gmail.com> wrote:  
> > > > On 29 November 2017 at 08:11, Sitsofe Wheeler <sitsofe@gmail.com> wrote:  
> > > >> On 29 November 2017 at 08:06, Greg KH <greg@kroah.com> wrote:  
> > > >>> On Tue, Nov 28, 2017 at 10:58:25PM +0000, Sitsofe Wheeler wrote:  
> > > >>>> I'd like to nominate f1c635b439a5c01776fe3a25b1e2dc546ea82e6f
> > > >>>> (https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/drivers/scsi/storvsc_drv.c?id=f1c635b439a5c01776fe3a25b1e2dc546ea82e6f
> > > >>>> ) to be backported to 4.4 stable because it looks like without it
> > > >>>> passthrough SSD disks won't work and will generate faux devices. It is
> > > >>>> also being carried by Ubuntu:
> > > >>>> http://kernel.ubuntu.com/git/ubuntu/ubuntu-xenial.git/log/drivers/scsi/storvsc_drv.c?h=Ubuntu-4.4.0-98.121
> > > >>>> along with other Hyper-V patches in their 4.4 tree.
> > > >>>>
> > > >>>> I'm not 100% sure that this one will fix what is being seen but I'm
> > > >>>> coming to this conclusion based on
> > > >>>> https://bugs.launchpad.net/ubuntu/xenial/+source/linux/+bug/1679898/comments/144  
> > > >>>
> > > >>> Can you test it to determine if this really does what you think it does
> > > >>> to solve the problem?  The changelog text makes it seem not necessary at
> > > >>> all.  
> > > >>
> > > >> I agree the changelog makes it seem totally unnecessary but the
> > > >> changelog is wrong in this case :-) I have checked out 4.4.102 and
> > > >> built it, run it and hit the error. When applying the patch attached
> > > >> (which is basically f1c635b439a5c01776fe3a25b1e2dc546ea82e6f with the
> > > >> lines removing static int msft_blist_flags = BLIST_TRY_VPD_PAGES;
> > > >> moved about) and the problem was resolved.
> > > >>
> > > >> I'm happy for others to weigh on what I'm seeing.  
> > > >
> > > > I will also note that this patch was deemed necessary/suitable for
> > > > backporting to 4.9 and 4.10 back in May -
> > > > (https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/log/drivers/scsi/storvsc_drv.c?h=linux-4.9.y
> > > > and https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/log/drivers/scsi/storvsc_drv.c?h=linux-4.10.y
> > > > ). Looking at https://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git/log/drivers/scsi/storvsc_drv.c?h=linux-4.1.y
> > > > perhaps this is also suitable for 4.1 stable too (but I haven't tested
> > > > 4.1, only 4.4.102)...  
> > >
> > > One more data point - back in May
> > > (https://www.spinics.net/lists/stable/msg170046.html ) you noted this
> > > in relation to one of the stable patch requests for this:
> > > "it's interesting how you deleted the changelog
> > > comments here saying this was only needed for 4.11 and no older kernels
> > > :)"
> > >
> > > and there's a follow up reply
> > > (https://www.spinics.net/lists/stable/msg170125.html ) where Stephen
> > > says:
> > >
> > > "At the time I only had reports of problem with 4.11.
> > > But others ran into issue with 4.10 and 4.9"  
> >
> > The changelog was about where the problem was seen. It makes sense
> > that this also occurs with other classes of storage devices;
> > hadn't tried all the possible configurations.  
> 
> So, what does the maintainer of the driver recommend here?  Should I
> backport this patch, or not?
> 
> thanks,
> 
> greg k-h
> 

I was waiting for some feedback from Dexuan who has done most of the recent work on
Hyper-V storage. His feedback is.

> Hi Stephen,
> My understanding is that the patch was made to work around a host emulation
> defect of SCSI DVD device, so old kernels should also need this patch, but I didn't
> know it could fix SSD passthrough, and actually I was not aware of any bug 
> related to SSD passthrough.
> 
> Thanks,
> -- Dexuan

So yes, I think the backport is appropriate. The SSD fix is an added beneficial
side effect.