Re: [PULL 08/10] vfio-ccw: Introduce a new CRW region

[Cornelia Huck <cohuck@redhat.com>]

On Tue, 2 Jun 2020 10:46:25 -0400
Eric Farman <farman@linux.ibm.com> wrote:

> On 6/2/20 9:13 AM, Cornelia Huck wrote:
> > On Mon, 25 May 2020 11:41:13 +0200
> > Cornelia Huck <cohuck@redhat.com> wrote:
> >   
> >> From: Farhan Ali <alifm@linux.ibm.com>
> >>
> >> This region provides a mechanism to pass a Channel Report Word
> >> that affect vfio-ccw devices, and needs to be passed to the guest
> >> for its awareness and/or processing.
> >>
> >> The base driver (see crw_collect_info()) provides space for two
> >> CRWs, as a subchannel event may have two CRWs chained together
> >> (one for the ssid, one for the subchannel).  As vfio-ccw will
> >> deal with everything at the subchannel level, provide space
> >> for a single CRW to be transferred in one shot.
> >>
> >> Signed-off-by: Farhan Ali <alifm@linux.ibm.com>
> >> Signed-off-by: Eric Farman <farman@linux.ibm.com>
> >> Reviewed-by: Cornelia Huck <cohuck@redhat.com>
> >> Message-Id: <20200505122745.53208-7-farman@linux.ibm.com>
> >> Signed-off-by: Cornelia Huck <cohuck@redhat.com>
> >> ---
> >>  Documentation/s390/vfio-ccw.rst     | 19 ++++++++++
> >>  drivers/s390/cio/vfio_ccw_chp.c     | 55 +++++++++++++++++++++++++++++
> >>  drivers/s390/cio/vfio_ccw_drv.c     | 20 +++++++++++
> >>  drivers/s390/cio/vfio_ccw_ops.c     |  8 +++++
> >>  drivers/s390/cio/vfio_ccw_private.h |  4 +++
> >>  include/uapi/linux/vfio.h           |  2 ++
> >>  include/uapi/linux/vfio_ccw.h       |  8 +++++
> >>  7 files changed, 116 insertions(+)
> >>  
> > 
> > (...)
> >   
> >> @@ -413,6 +423,16 @@ static int __init vfio_ccw_sch_init(void)
> >>  		goto out_err;
> >>  	}
> >>  
> >> +	vfio_ccw_crw_region = kmem_cache_create_usercopy("vfio_ccw_crw_region",
> >> +					sizeof(struct ccw_crw_region), 0,
> >> +					SLAB_ACCOUNT, 0,
> >> +					sizeof(struct ccw_crw_region), NULL);  
> > 
> > Ugh, I just tested this rebased to the s390 features branch, and I must
> > have used some different options, because I now get
> > 
> >    kmem_cache_create(vfio_ccw_crw_region) integrity check failed
> > 
> > presumably due to the size of the ccw_crw_region.
> > 
> > We maybe need to pad it up (leave it unpacked)? Eric, what do you think?  
> 
> Certainly packing a single one-word struct is weird, and the message is
> coming out of the tiny struct itself:
> 
> mm/slab-common.c:88:
>         if (!name || in_interrupt() || size < sizeof(void *) ||
>                 size > KMALLOC_MAX_SIZE) {
>                 pr_err("kmem_cache_create(%s) integrity check failed\n",
> name);
> 
> That's protected by CONFIG_DEBUG_VM which wasn't enabled in my config.
> So playing around with things, we'd have to explicitly add a pad (or the
> second CRW, ha!) to get the struct back up to a doubleword. That'd be
> fine with me.

I think I'll just go with

struct ccw_crw_region {                                                         
        __u32 crw;                                                              
        __u32 pad;                                                              
} __packed;

here and in the doc. I'll do some tests and do another pull request
tomorrow.