[PATCH] scsi: storvsc: make INQUIRY response SPC-compliant

[PATCH] scsi: storvsc: make INQUIRY response SPC-compliant

[Vitaly Kuznetsov]

SPC-2/3/4 specs state that "The standard INQUIRY data (see table ...)
shall contain at least 36 bytes". Hyper-V host doesn't always honor this
requirement, e.g. when there is no physical device present at a particular
LUN host sets Peripheral qualifier to 011b and Additional length to 0
(thus making the reply 5-bytes long). Upper level SCSI stack complains
with 'INQUIRY result too short (5), using 36'. Fix the issue by mangling
Additional length field in host's reply at the driver level.

Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
---
This is a hack, the proper fix should probably be done in Hyper-V.
---
 drivers/scsi/storvsc_drv.c | 19 +++++++++++++++++++
 1 file changed, 19 insertions(+)

diff --git a/drivers/scsi/storvsc_drv.c b/drivers/scsi/storvsc_drv.c
index 3c6584f..bca31af 100644
--- a/drivers/scsi/storvsc_drv.c
+++ b/drivers/scsi/storvsc_drv.c
@@ -1153,6 +1153,7 @@ static void storvsc_on_io_completion(struct hv_device *device,
 {
 	struct storvsc_device *stor_device;
 	struct vstor_packet *stor_pkt;
+	struct vmbus_packet_mpb_array *payload = request->payload;
 
 	stor_device = hv_get_drvdata(device);
 	stor_pkt = &request->vstor_packet;
@@ -1174,6 +1175,24 @@ static void storvsc_on_io_completion(struct hv_device *device,
 		vstor_packet->vm_srb.srb_status = SRB_STATUS_SUCCESS;
 	}
 
+	/*
+	 * When there is no physical device attached to a LUN Hyper-V host
+	 * sets Peripheral qualifier to 011b and Additional length to 0. SPC
+	 * spec, however, says that "The standard INQUIRY data ... shall
+	 * contain at least 36 bytes". Upper level SCSI stack complains with
+	 * 'INQUIRY result too short (5), using 36'. Mangle host's reply here.
+	 */
+	if (stor_pkt->vm_srb.cdb[0] == INQUIRY && payload) {
+		u8 *buf, per_qual, data_len;
+		int range_len = payload->range.len;
+
+		buf = phys_to_virt(payload->range.pfn_array[0] << PAGE_SHIFT) +
+			payload->range.offset;
+		per_qual = (buf[0] >> 5) & 7;
+		data_len = buf[4] + 5;
+		if (per_qual == 3 && data_len < min(range_len, 36))
+			buf[4] = min(range_len, 36) - 5;
+	}
 
 	/* Copy over the status...etc */
 	stor_pkt->vm_srb.scsi_status = vstor_packet->vm_srb.scsi_status;
-- 
2.4.3

Re: [PATCH] scsi: storvsc: make INQUIRY response SPC-compliant

[Christoph Hellwig]

On Wed, Jul 01, 2015 at 11:04:08AM +0200, Vitaly Kuznetsov wrote:
> SPC-2/3/4 specs state that "The standard INQUIRY data (see table ...)
> shall contain at least 36 bytes". Hyper-V host doesn't always honor this
> requirement, e.g. when there is no physical device present at a particular
> LUN host sets Peripheral qualifier to 011b and Additional length to 0
> (thus making the reply 5-bytes long). Upper level SCSI stack complains
> with 'INQUIRY result too short (5), using 36'. Fix the issue by mangling
> Additional length field in host's reply at the driver level.

This looks like a big mess, and usage of phys_to_virt is not generally
safe to start with.

If HyperV really is that broken the warning seems correct, but if you
really have to get rid of it we could add a blist flag to not issue
the warning in the core code instead of hacking around it in the driver.

RE: [PATCH] scsi: storvsc: make INQUIRY response SPC-compliant

[KY Srinivasan]

> -----Original Message-----
> From: Christoph Hellwig [mailto:hch@infradead.org]
> Sent: Friday, July 3, 2015 9:19 AM
> To: Vitaly Kuznetsov
> Cc: linux-scsi@vger.kernel.org; Long Li; KY Srinivasan; Haiyang Zhang; James
> E.J. Bottomley; devel@linuxdriverproject.org; linux-kernel@vger.kernel.org
> Subject: Re: [PATCH] scsi: storvsc: make INQUIRY response SPC-compliant
> 
> On Wed, Jul 01, 2015 at 11:04:08AM +0200, Vitaly Kuznetsov wrote:
> > SPC-2/3/4 specs state that "The standard INQUIRY data (see table ...)
> > shall contain at least 36 bytes". Hyper-V host doesn't always honor this
> > requirement, e.g. when there is no physical device present at a particular
> > LUN host sets Peripheral qualifier to 011b and Additional length to 0
> > (thus making the reply 5-bytes long). Upper level SCSI stack complains
> > with 'INQUIRY result too short (5), using 36'. Fix the issue by mangling
> > Additional length field in host's reply at the driver level.
> 
> This looks like a big mess, and usage of phys_to_virt is not generally
> safe to start with.
> 
> If HyperV really is that broken the warning seems correct, but if you
> really have to get rid of it we could add a blist flag to not issue
> the warning in the core code instead of hacking around it in the driver.

Agreed. We have fixed this issue in win10 and I am trying to get the fix backported.

Regards,

K. Y

Re: [PATCH] scsi: storvsc: make INQUIRY response SPC-compliant

[Vitaly Kuznetsov]

KY Srinivasan <kys@microsoft.com> writes:

>> -----Original Message-----
>> From: Christoph Hellwig [mailto:hch@infradead.org]
>> Sent: Friday, July 3, 2015 9:19 AM
>> To: Vitaly Kuznetsov
>> Cc: linux-scsi@vger.kernel.org; Long Li; KY Srinivasan; Haiyang Zhang; James
>> E.J. Bottomley; devel@linuxdriverproject.org; linux-kernel@vger.kernel.org
>> Subject: Re: [PATCH] scsi: storvsc: make INQUIRY response SPC-compliant
>> 
>> On Wed, Jul 01, 2015 at 11:04:08AM +0200, Vitaly Kuznetsov wrote:
>> > SPC-2/3/4 specs state that "The standard INQUIRY data (see table ...)
>> > shall contain at least 36 bytes". Hyper-V host doesn't always honor this
>> > requirement, e.g. when there is no physical device present at a particular
>> > LUN host sets Peripheral qualifier to 011b and Additional length to 0
>> > (thus making the reply 5-bytes long). Upper level SCSI stack complains
>> > with 'INQUIRY result too short (5), using 36'. Fix the issue by mangling
>> > Additional length field in host's reply at the driver level.
>> 
>> This looks like a big mess, and usage of phys_to_virt is not generally
>> safe to start with.
>> 
>> If HyperV really is that broken the warning seems correct, but if you
>> really have to get rid of it we could add a blist flag to not issue
>> the warning in the core code instead of hacking around it in the driver.
>
> Agreed. We have fixed this issue in win10 and I am trying to get the fix backported.

In case this is fixed in future Hyper-V versions introducing new blist
flags looks like an overkill, let's leave things as they are.

Thanks,

-- 
  Vitaly