Linux Virtualization - Sep 2020 - [PATCH] Rescan the entire target on transport reset when LUN is 0

On 28/08/20 14:21, Matej Genci wrote:
> VirtIO 1.0 spec says
>     The removed and rescan events ... when sent for LUN 0, they MAY
>     apply to the entire target so the driver can ask the initiator
>     to rescan the target to detect this.
> 
> This change introduces the behaviour described above by scanning the
> entire scsi target when LUN is set to 0. This is both a functional and a
> performance fix. It aligns the driver with the spec and allows control
> planes to hotplug targets with large numbers of LUNs without having to
> request a RESCAN for each one of them.
> 
> Signed-off-by: Matej Genci <matej at nutanix.com>
> Suggested-by: Felipe Franciosi <felipe at nutanix.com>
> ---
>  drivers/scsi/virtio_scsi.c | 7 ++++++-
>  1 file changed, 6 insertions(+), 1 deletion(-)
> 
> diff --git a/drivers/scsi/virtio_scsi.c b/drivers/scsi/virtio_scsi.c
> index bfec84aacd90..a4b9bc7b4b4a 100644
> --- a/drivers/scsi/virtio_scsi.c
> +++ b/drivers/scsi/virtio_scsi.c
> @@ -284,7 +284,12 @@ static void virtscsi_handle_transport_reset(struct
virtio_scsi *vscsi,
>  
>  	switch (virtio32_to_cpu(vscsi->vdev, event->reason)) {
>  	case VIRTIO_SCSI_EVT_RESET_RESCAN:
> -		scsi_add_device(shost, 0, target, lun);
> +		if (lun == 0) {
> +			scsi_scan_target(&shost->shost_gendev, 0, target,
> +					 SCAN_WILD_CARD, SCSI_SCAN_INITIAL);
> +		} else {
> +			scsi_add_device(shost, 0, target, lun);
> +		}
>  		break;
>  	case VIRTIO_SCSI_EVT_RESET_REMOVED:
>  		sdev = scsi_device_lookup(shost, 0, target, lun);
> 

Acked-by: Paolo Bonzini <pbonzini at redhat.com>

> On Sep 8, 2020, at 3:22 PM, Paolo Bonzini <pbonzini at redhat.com>
wrote:
> 
> On 28/08/20 14:21, Matej Genci wrote:
>> VirtIO 1.0 spec says
>>    The removed and rescan events ... when sent for LUN 0, they MAY
>>    apply to the entire target so the driver can ask the initiator
>>    to rescan the target to detect this.
>> 
>> This change introduces the behaviour described above by scanning the
>> entire scsi target when LUN is set to 0. This is both a functional and
a
>> performance fix. It aligns the driver with the spec and allows control
>> planes to hotplug targets with large numbers of LUNs without having to
>> request a RESCAN for each one of them.
>> 
>> Signed-off-by: Matej Genci <matej at nutanix.com>
>> Suggested-by: Felipe Franciosi <felipe at nutanix.com>
>> ---
>> drivers/scsi/virtio_scsi.c | 7 ++++++-
>> 1 file changed, 6 insertions(+), 1 deletion(-)
>> 
>> diff --git a/drivers/scsi/virtio_scsi.c b/drivers/scsi/virtio_scsi.c
>> index bfec84aacd90..a4b9bc7b4b4a 100644
>> --- a/drivers/scsi/virtio_scsi.c
>> +++ b/drivers/scsi/virtio_scsi.c
>> @@ -284,7 +284,12 @@ static void virtscsi_handle_transport_reset(struct
virtio_scsi *vscsi,
>> 
>> 	switch (virtio32_to_cpu(vscsi->vdev, event->reason)) {
>> 	case VIRTIO_SCSI_EVT_RESET_RESCAN:
>> -		scsi_add_device(shost, 0, target, lun);
>> +		if (lun == 0) {
>> +			scsi_scan_target(&shost->shost_gendev, 0, target,
>> +					 SCAN_WILD_CARD, SCSI_SCAN_INITIAL);
>> +		} else {
>> +			scsi_add_device(shost, 0, target, lun);
>> +		}
>> 		break;
>> 	case VIRTIO_SCSI_EVT_RESET_REMOVED:
>> 		sdev = scsi_device_lookup(shost, 0, target, lun);
>> 
> 
> 
> Acked-by: Paolo Bonzini <pbonzini at redhat.com>
Cc: stable at vger.kernel.org

Thanks, Paolo.

I'm Cc'ing stable as I believe this fixes a driver bug where it
doesn't follow the spec. Per commit message, today devices are
required to issue RESCAN events for each LUN behind a target when
hotplugging, or risking the driver not seeing the new LUNs.

Is this enough? Or should we resend after merge per below?
https://www.kernel.org/doc/Documentation/process/stable-kernel-rules.rst

F.

On Tue, Sep 08, 2020 at 05:53:16PM +0000, Felipe Franciosi
wrote:
> 
> 
> > On Sep 8, 2020, at 3:22 PM, Paolo Bonzini <pbonzini at
redhat.com> wrote:
> > 
> > On 28/08/20 14:21, Matej Genci wrote:
> >> VirtIO 1.0 spec says
> >>    The removed and rescan events ... when sent for LUN 0, they MAY
> >>    apply to the entire target so the driver can ask the initiator
> >>    to rescan the target to detect this.
> >> 
> >> This change introduces the behaviour described above by scanning
the
> >> entire scsi target when LUN is set to 0. This is both a functional
and a
> >> performance fix. It aligns the driver with the spec and allows
control
> >> planes to hotplug targets with large numbers of LUNs without
having to
> >> request a RESCAN for each one of them.
> >> 
> >> Signed-off-by: Matej Genci <matej at nutanix.com>
> >> Suggested-by: Felipe Franciosi <felipe at nutanix.com>
> >> ---
> >> drivers/scsi/virtio_scsi.c | 7 ++++++-
> >> 1 file changed, 6 insertions(+), 1 deletion(-)
> >> 
> >> diff --git a/drivers/scsi/virtio_scsi.c
b/drivers/scsi/virtio_scsi.c
> >> index bfec84aacd90..a4b9bc7b4b4a 100644
> >> --- a/drivers/scsi/virtio_scsi.c
> >> +++ b/drivers/scsi/virtio_scsi.c
> >> @@ -284,7 +284,12 @@ static void
virtscsi_handle_transport_reset(struct virtio_scsi *vscsi,
> >> 
> >> 	switch (virtio32_to_cpu(vscsi->vdev, event->reason)) {
> >> 	case VIRTIO_SCSI_EVT_RESET_RESCAN:
> >> -		scsi_add_device(shost, 0, target, lun);
> >> +		if (lun == 0) {
> >> +			scsi_scan_target(&shost->shost_gendev, 0, target,
> >> +					 SCAN_WILD_CARD, SCSI_SCAN_INITIAL);
> >> +		} else {
> >> +			scsi_add_device(shost, 0, target, lun);
> >> +		}
> >> 		break;
> >> 	case VIRTIO_SCSI_EVT_RESET_REMOVED:
> >> 		sdev = scsi_device_lookup(shost, 0, target, lun);
> >> 
> > 
> > 
> > Acked-by: Paolo Bonzini <pbonzini at redhat.com>
> 
> Cc: stable at vger.kernel.org
> 
> Thanks, Paolo.
> 
> I'm Cc'ing stable as I believe this fixes a driver bug where it
> doesn't follow the spec. Per commit message, today devices are
> required to issue RESCAN events for each LUN behind a target when
> hotplugging, or risking the driver not seeing the new LUNs.
> 
> Is this enough? Or should we resend after merge per below?
> https://www.kernel.org/doc/Documentation/process/stable-kernel-rules.rst
You need to let stable know the git commit id of the patch in Linus's
tree if the cc: stable is not on the final commit that gets merged.

thanks,

greg k-h