Re: [PATCH v2 3/8] s390/cio: add basic protected virtualization support

[Halil Pasic <pasic@linux.ibm.com>]

On Mon, 27 May 2019 12:38:02 +0200
Cornelia Huck <cohuck@redhat.com> wrote:

> On Thu, 23 May 2019 18:22:04 +0200
> Michael Mueller <mimu@linux.ibm.com> wrote:
> 
> > From: Halil Pasic <pasic@linux.ibm.com>
> > 
> > As virtio-ccw devices are channel devices, we need to use the dma area
> > for any communication with the hypervisor.
> > 
> > It handles neither QDIO in the common code, nor any device type specific
> > stuff (like channel programs constructed by the DASD driver).
> > 
> > An interesting side effect is that virtio structures are now going to
> > get allocated in 31 bit addressable storage.
> > 
> > Signed-off-by: Halil Pasic <pasic@linux.ibm.com>
> 
> [Side note: you really should add your s-o-b if you send someone else's
> patches... if Halil ends up committing them, it's fine, though.]
> 
> > ---
> >  arch/s390/include/asm/ccwdev.h   |  4 +++
> >  drivers/s390/cio/ccwreq.c        |  9 +++---
> >  drivers/s390/cio/device.c        | 64 +++++++++++++++++++++++++++++++++-------
> >  drivers/s390/cio/device_fsm.c    | 53 ++++++++++++++++++++-------------
> >  drivers/s390/cio/device_id.c     | 20 +++++++------
> >  drivers/s390/cio/device_ops.c    | 21 +++++++++++--
> >  drivers/s390/cio/device_pgid.c   | 22 +++++++-------
> >  drivers/s390/cio/device_status.c | 24 +++++++--------
> >  drivers/s390/cio/io_sch.h        | 20 +++++++++----
> >  drivers/s390/virtio/virtio_ccw.c | 10 -------
> >  10 files changed, 164 insertions(+), 83 deletions(-)
> > 
> 
> (...)
> 
> > @@ -1593,20 +1622,31 @@ struct ccw_device * __init ccw_device_create_console(struct ccw_driver *drv)
> >  		return ERR_CAST(sch);
> >  
> >  	io_priv = kzalloc(sizeof(*io_priv), GFP_KERNEL | GFP_DMA);
> > -	if (!io_priv) {
> > -		put_device(&sch->dev);
> > -		return ERR_PTR(-ENOMEM);
> > -	}
> > +	if (!io_priv)
> > +		goto err_priv;
> > +	io_priv->dma_area = dma_alloc_coherent(&sch->dev,
> > +				sizeof(*io_priv->dma_area),
> > +				&io_priv->dma_area_dma, GFP_KERNEL);
> 
> Even though we'll only end up here for 3215 or 3270 consoles, this sent
> me looking.
> 
> This code is invoked via console_init(). A few lines down in
> start_kernel(), we have
> 
>         /*                                                                       
>          * This needs to be called before any devices perform DMA                
>          * operations that might use the SWIOTLB bounce buffers. It will         
>          * mark the bounce buffers as decrypted so that their usage will         
>          * not cause "plain-text" data to be decrypted when accessed.            
>          */
>         mem_encrypt_init();
> 
> So, I'm wondering if creating the console device interacts in any way
> with the memory encryption interface?

I do things a bit different than x86: the SWIOTLB stuff is set up in
mem_init(). So I think we should be fine. If there is a down-side to
calling swiotlb_update_mem_attributes() earlier, honestly I'm
not sure.

> 
> [Does basic recognition work if you start a protected virt guest with a
> 3270 console? I realize that the console is unlikely to work, but that
> should at least exercise this code path.]

I've already had some thoughts along these lines and slapped  
-device x-terminal3270,chardev=char_0,devno=fe.0.000a,id=terminal_0 \
on my qemu command line. The ccw device does show up in the guest...

Device   Subchan.  DevType CU Type Use  PIM PAM POM  CHPIDs
----------------------------------------------------------------------
0.0.0000 0.0.0000  0000/00 3832/01 yes  80  80  ff   00000000 00000000 
0.0.000a 0.0.0001  0000/00 3270/00      80  80  ff   01000000 00000000 
0.0.0002 0.0.0002  0000/00 3832/09 yes  80  80  ff   00000000 00000000 
0.0.0300 0.0.0003  0000/00 3832/02 yes  80  80  ff   00000000 00000000 
0.0.0301 0.0.0004  0000/00 3832/02 yes  80  80  ff   00000000 00000000 

But I would not call it a comprehensive test...

Mimu, do we have something more elaborate with regards to this?

Regards,
Halil