[PATCH 2/2] [SCSI] ibmvscsi: Fix SCSI scanning

[PATCH 2/2] [SCSI] ibmvscsi: Fix SCSI scanning

[Bart Van Assche]

Avoid that SCSI scanning triggers creation of targets with a non-zero
channel number.

Signed-off-by: Bart Van Assche <bvanassche@acm.org>
Cc: Brian King <brking@linux.vnet.ibm.com>
Cc: FUJITA Tomonori <fujita.tomonori@lab.ntt.co.jp>
Cc: <stable@kernel.org>
---
 drivers/scsi/ibmvscsi/ibmvscsi.c |    6 ++++++
 1 files changed, 6 insertions(+), 0 deletions(-)

Note: this patch hasn't been tested yet.

diff --git a/drivers/scsi/ibmvscsi/ibmvscsi.c b/drivers/scsi/ibmvscsi/ibmvscsi.c
index 3d391dc..c31adf5 100644
--- a/drivers/scsi/ibmvscsi/ibmvscsi.c
+++ b/drivers/scsi/ibmvscsi/ibmvscsi.c
@@ -498,6 +498,11 @@ static void purge_requests(struct ibmvscsi_host_data *hostdata, int error_code)
 	spin_unlock_irqrestore(hostdata->host->host_lock, flags);
 }
 
+static int ibmvscsi_target_alloc(struct scsi_target *starget)
+{
+	return starget->channel == 0 ? 0 : -ENODEV;
+}
+
 /**
  * ibmvscsi_reset_host - Reset the connection to the server
  * @hostdata:	struct ibmvscsi_host_data to reset
@@ -1803,6 +1808,7 @@ static struct scsi_host_template driver_template = {
 	.module = THIS_MODULE,
 	.name = "IBM POWER Virtual SCSI Adapter " IBMVSCSI_VERSION,
 	.proc_name = "ibmvscsi",
+	.target_alloc = ibmvscsi_target_alloc,
 	.queuecommand = ibmvscsi_queuecommand,
 	.eh_abort_handler = ibmvscsi_eh_abort_handler,
 	.eh_device_reset_handler = ibmvscsi_eh_device_reset_handler,
-- 
1.7.3.4

Re: [PATCH 2/2] [SCSI] ibmvscsi: Fix SCSI scanning

[Christoph Hellwig]

On Sun, Jun 19, 2011 at 01:48:40PM +0200, Bart Van Assche wrote:
> Avoid that SCSI scanning triggers creation of targets with a non-zero
> channel number.

Same comment as for the SRP one applies here as well.

Re: [PATCH 2/2] [SCSI] ibmvscsi: Fix SCSI scanning

[Bart Van Assche]

On Mon, Jun 20, 2011 at 1:40 PM, Christoph Hellwig <hch@infradead.org> wrote:
> On Sun, Jun 19, 2011 at 01:48:40PM +0200, Bart Van Assche wrote:
>> Avoid that SCSI scanning triggers creation of targets with a non-zero
>> channel number.
>
> Same comment as for the SRP one applies here as well.

If no kernel module parameters have been specified, max_channel is set
to three in the ibmvscsi driver. It's not clear to me why.

Bart.

Re: [PATCH 2/2] [SCSI] ibmvscsi: Fix SCSI scanning

[Brian King]

On 06/20/2011 12:58 PM, Bart Van Assche wrote:
> On Mon, Jun 20, 2011 at 1:40 PM, Christoph Hellwig <hch@infradead.org> wrote:
>> On Sun, Jun 19, 2011 at 01:48:40PM +0200, Bart Van Assche wrote:
>>> Avoid that SCSI scanning triggers creation of targets with a non-zero
>>> channel number.
>>
>> Same comment as for the SRP one applies here as well.
> 
> If no kernel module parameters have been specified, max_channel is set
> to three in the ibmvscsi driver. It's not clear to me why.

It could be set as high as 16 in order to support the full 3 bits worth
of bus values as defined in Logical Unit Addressing model in SAM, but
3 is generally sufficient for most users as it allows for 64 targets *
3 buses = 192 virtual scsi targets.

Thanks,

Brian

-- 
Brian King
Linux on Power Virtualization
IBM Linux Technology Center

Re: [PATCH 2/2] [SCSI] ibmvscsi: Fix SCSI scanning

[Brian King]

On 06/19/2011 06:48 AM, Bart Van Assche wrote:
> diff --git a/drivers/scsi/ibmvscsi/ibmvscsi.c b/drivers/scsi/ibmvscsi/ibmvscsi.c
> index 3d391dc..c31adf5 100644
> --- a/drivers/scsi/ibmvscsi/ibmvscsi.c
> +++ b/drivers/scsi/ibmvscsi/ibmvscsi.c
> @@ -498,6 +498,11 @@ static void purge_requests(struct ibmvscsi_host_data *hostdata, int error_code)
>  	spin_unlock_irqrestore(hostdata->host->host_lock, flags);
>  }
> 
> +static int ibmvscsi_target_alloc(struct scsi_target *starget)
> +{
> +	return starget->channel == 0 ? 0 : -ENODEV;
> +}
> +

This will break existing IBM Power users who have more than 64 virtual disks assigned
to a single virtual scsi adapter. The IBM VSCSI target uses Logical Unit Addressing (10b)
and does bus/id/lun -> vscsi LUN mapping as defined in SAM. The 65'th virtual scsi
disk would show up on bus 1 as target 0.

Thanks,

Brian

-- 
Brian King
Linux on Power Virtualization
IBM Linux Technology Center