linux-kernel - Re: [RFC PATCH 1/3] vfio: ccw: introduce schib region

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20180111151659.2d997abf.cohuck@redhat.com>
Date:   Thu, 11 Jan 2018 15:16:59 +0100
From:   Cornelia Huck <cohuck@...hat.com>
To:     Dong Jia Shi <bjsdjshi@...ux.vnet.ibm.com>
Cc:     linux-kernel@...r.kernel.org, linux-s390@...r.kernel.org,
        kvm@...r.kernel.org, qemu-devel@...gnu.org, qemu-s390x@...gnu.org,
        borntraeger@...ibm.com, pasic@...ux.vnet.ibm.com,
        pmorel@...ux.vnet.ibm.com
Subject: Re: [RFC PATCH 1/3] vfio: ccw: introduce schib region

On Thu, 11 Jan 2018 04:04:19 +0100
Dong Jia Shi <bjsdjshi@...ux.vnet.ibm.com> wrote:

> This introduces a new region for vfio-ccw to provide subchannel
> information for user space.
> 
> Signed-off-by: Dong Jia Shi <bjsdjshi@...ux.vnet.ibm.com>
> ---
>  drivers/s390/cio/vfio_ccw_fsm.c     | 21 ++++++++++
>  drivers/s390/cio/vfio_ccw_ops.c     | 79 +++++++++++++++++++++++++++----------
>  drivers/s390/cio/vfio_ccw_private.h |  3 ++
>  include/uapi/linux/vfio.h           |  1 +
>  include/uapi/linux/vfio_ccw.h       |  6 +++
>  5 files changed, 90 insertions(+), 20 deletions(-)
> 
> diff --git a/drivers/s390/cio/vfio_ccw_fsm.c b/drivers/s390/cio/vfio_ccw_fsm.c
> index c30420c517b1..be081ccabea3 100644
> --- a/drivers/s390/cio/vfio_ccw_fsm.c
> +++ b/drivers/s390/cio/vfio_ccw_fsm.c
> @@ -172,6 +172,22 @@ static void fsm_irq(struct vfio_ccw_private *private,
>  		complete(private->completion);
>  }
>  
> +static void fsm_update_subch(struct vfio_ccw_private *private,
> +			     enum vfio_ccw_event event)
> +{
> +	struct subchannel *sch;
> +
> +	sch = private->sch;
> +	if (cio_update_schib(sch)) {

This implies device gone. Do we also want to trigger some event, or
just wait until a machine check comes around and we're notified in the
normal way? (Probably the latter.)

> +		private->schib_region.cc = 3;
> +		return;
> +	}
> +
> +	private->schib_region.cc = 0;
> +	memcpy(private->schib_region.schib_area, &sch->schib,
> +	       sizeof(sch->schib));

We might want to add documentation that schib_area contains the schib
from the last successful invocation of stsch (if any). That makes sense
as the schib remains unchanged for cc=3 after stsch anyway, but it
can't hurt to spell it out.

> +}
> +
>  /*
>   * Device statemachine
>   */
> @@ -180,25 +196,30 @@ fsm_func_t *vfio_ccw_jumptable[NR_VFIO_CCW_STATES][NR_VFIO_CCW_EVENTS] = {
>  		[VFIO_CCW_EVENT_NOT_OPER]	= fsm_nop,
>  		[VFIO_CCW_EVENT_IO_REQ]		= fsm_io_error,
>  		[VFIO_CCW_EVENT_INTERRUPT]	= fsm_disabled_irq,
> +		[VFIO_CCW_EVENT_UPDATE_SUBCH]	= fsm_update_subch,
>  	},
>  	[VFIO_CCW_STATE_STANDBY] = {
>  		[VFIO_CCW_EVENT_NOT_OPER]	= fsm_notoper,
>  		[VFIO_CCW_EVENT_IO_REQ]		= fsm_io_error,
>  		[VFIO_CCW_EVENT_INTERRUPT]	= fsm_irq,
> +		[VFIO_CCW_EVENT_UPDATE_SUBCH]	= fsm_update_subch,
>  	},
>  	[VFIO_CCW_STATE_IDLE] = {
>  		[VFIO_CCW_EVENT_NOT_OPER]	= fsm_notoper,
>  		[VFIO_CCW_EVENT_IO_REQ]		= fsm_io_request,
>  		[VFIO_CCW_EVENT_INTERRUPT]	= fsm_irq,
> +		[VFIO_CCW_EVENT_UPDATE_SUBCH]	= fsm_update_subch,
>  	},
>  	[VFIO_CCW_STATE_BOXED] = {
>  		[VFIO_CCW_EVENT_NOT_OPER]	= fsm_notoper,
>  		[VFIO_CCW_EVENT_IO_REQ]		= fsm_io_busy,
>  		[VFIO_CCW_EVENT_INTERRUPT]	= fsm_irq,
> +		[VFIO_CCW_EVENT_UPDATE_SUBCH]	= fsm_update_subch,
>  	},
>  	[VFIO_CCW_STATE_BUSY] = {
>  		[VFIO_CCW_EVENT_NOT_OPER]	= fsm_notoper,
>  		[VFIO_CCW_EVENT_IO_REQ]		= fsm_io_busy,
>  		[VFIO_CCW_EVENT_INTERRUPT]	= fsm_irq,
> +		[VFIO_CCW_EVENT_UPDATE_SUBCH]	= fsm_update_subch,

Does it makes to trigger this through the state machine if we always do
the same action and never change state?

>  	},
>  };

Else, looks good.