lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Tomasz Chmielewski]

lspci implies that the virtio block device is a "SCSI storage 
controller", i.e.:

00:05.0 SCSI storage controller: Qumranet, Inc. Virtio block device


However, virtio block devide does not have much to do with SCSI (in 
sense: sdparm does not think it is a SCSI device; virtio_blk does not 
depend on any SCSI moduled like sd_mod).


Is "SCSI storage controller" a proper description for this device?


-- 
Tomasz Chmielewski
http://wpkg.org

Re: lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Luca Tettamanti]

On Mon, Oct 19, 2009 at 2:14 PM, Tomasz Chmielewski <mangoo@wpkg.org> wrote:
> lspci implies that the virtio block device is a "SCSI storage controller",
> i.e.:
>
> 00:05.0 SCSI storage controller: Qumranet, Inc. Virtio block device
>
>
> However, virtio block devide does not have much to do with SCSI (in sense:
> sdparm does not think it is a SCSI device; virtio_blk does not depend on any
> SCSI moduled like sd_mod).
>
> Is "SCSI storage controller" a proper description for this device?

It does not talk SCSI protocol if that's what you're asking. The
description you see comes from the PCI class (storage controller) and
subclass (SCSI controller); the meaning of the class/subclass is fixed
by the PCI standard.

Luca

Re: lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Tomasz Chmielewski]

Luca Tettamanti wrote:
> On Mon, Oct 19, 2009 at 2:14 PM, Tomasz Chmielewski <mangoo@wpkg.org> wrote:
>> lspci implies that the virtio block device is a "SCSI storage controller",
>> i.e.:
>>
>> 00:05.0 SCSI storage controller: Qumranet, Inc. Virtio block device
>>
>>
>> However, virtio block devide does not have much to do with SCSI (in sense:
>> sdparm does not think it is a SCSI device; virtio_blk does not depend on any
>> SCSI moduled like sd_mod).
>>
>> Is "SCSI storage controller" a proper description for this device?
> 
> It does not talk SCSI protocol if that's what you're asking. The
> description you see comes from the PCI class (storage controller) and
> subclass (SCSI controller); the meaning of the class/subclass is fixed
> by the PCI standard.

So why was "SCSI storage controller" any better than "IDE interface" or 
"SATA controller" for virtio block device, if it does not talk SCSI 
protocol (other than "SCSI storage controller" being the first on the 
list of subclasses)?


Doesn't "80  Mass storage controller" ("0x80  	0x00  	Other mass storage 
controller") fit better for virtio block device?

Generally, I see that 0x80 is reserved for "other/unspecified" types of 
devices from a given PCI class.


Let me know if I'm asking a stupid question ;)



C 01  Mass storage controller
         00  SCSI storage controller
         01  IDE interface
         02  Floppy disk controller
         03  IPI bus controller
         04  RAID bus controller
         05  ATA controller
                 20  ADMA single stepping
                 30  ADMA continuous operation
         06  SATA controller
                 00  Vendor specific
                 01  AHCI 1.0
         07  Serial Attached SCSI controller
         80  Mass storage controller


-- 
Tomasz Chmielewski
http://wpkg.org

Re: lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Luca Tettamanti]

On Mon, Oct 19, 2009 at 3:31 PM, Tomasz Chmielewski <mangoo@wpkg.org> wrote:
> Luca Tettamanti wrote:
>>
>> On Mon, Oct 19, 2009 at 2:14 PM, Tomasz Chmielewski <mangoo@wpkg.org>
>> wrote:
>>>
>>> lspci implies that the virtio block device is a "SCSI storage
>>> controller",
>>> i.e.:
>>>
>>> 00:05.0 SCSI storage controller: Qumranet, Inc. Virtio block device
>>>
>>>
>>> However, virtio block devide does not have much to do with SCSI (in
>>> sense:
>>> sdparm does not think it is a SCSI device; virtio_blk does not depend on
>>> any
>>> SCSI moduled like sd_mod).
>>>
>>> Is "SCSI storage controller" a proper description for this device?
>>
>> It does not talk SCSI protocol if that's what you're asking. The
>> description you see comes from the PCI class (storage controller) and
>> subclass (SCSI controller); the meaning of the class/subclass is fixed
>> by the PCI standard.
>
> So why was "SCSI storage controller" any better than "IDE interface" or
> "SATA controller" for virtio block device, if it does not talk SCSI protocol
> (other than "SCSI storage controller" being the first on the list of
> subclasses)?

Because both ATA and SATA classes have a generic driver that would try
to bind to that controller (and the whole point of virtio block device
is to avoid emulating a ATA/SATA controller).

> Doesn't "80  Mass storage controller" ("0x80    0x00    Other mass storage
> controller") fit better for virtio block device?

Maybe. I guess that are compatibility problem with "other" operating systems.

Luca

Re: lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Tomasz Chmielewski]

Luca Tettamanti wrote:

>> So why was "SCSI storage controller" any better than "IDE interface" or
>> "SATA controller" for virtio block device, if it does not talk SCSI protocol
>> (other than "SCSI storage controller" being the first on the list of
>> subclasses)?
> 
> Because both ATA and SATA classes have a generic driver that would try
> to bind to that controller (and the whole point of virtio block device
> is to avoid emulating a ATA/SATA controller).
> 
>> Doesn't "80  Mass storage controller" ("0x80    0x00    Other mass storage
>> controller") fit better for virtio block device?
> 
> Maybe. I guess that are compatibility problem with "other" operating systems.

Thanks for clarifications.

It makes sense in that case - I don't have any more questions ;)


-- 
Tomasz Chmielewski
http://wpkg.org

Re: lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Vadim Rozenfeld]

On 10/19/2009 03:42 PM, Tomasz Chmielewski wrote:
> Luca Tettamanti wrote:
>
>>> So why was "SCSI storage controller" any better than "IDE interface" or
>>> "SATA controller" for virtio block device, if it does not talk SCSI 
>>> protocol
>>> (other than "SCSI storage controller" being the first on the list of
>>> subclasses)?
>>
>> Because both ATA and SATA classes have a generic driver that would try
>> to bind to that controller (and the whole point of virtio block device
>> is to avoid emulating a ATA/SATA controller).
>>
>>> Doesn't "80  Mass storage controller" ("0x80    0x00    Other mass 
>>> storage
>>> controller") fit better for virtio block device?
>>
>> Maybe. I guess that are compatibility problem with "other" operating 
>> systems.
>
> Thanks for clarifications.
>
> It makes sense in that case - I don't have any more questions ;)
>
>
we need it for windows viostor driver to be WHQL'ed

Re: lspci says: "SCSI storage controller: Qumranet, Inc. Virtio block device". Is it really?

[Gerd Hoffmann]

> C 01 Mass storage controller
> 00 SCSI storage controller
> 01 IDE interface
> 02 Floppy disk controller
> 03 IPI bus controller
> 04 RAID bus controller
> 05 ATA controller
> 20 ADMA single stepping
> 30 ADMA continuous operation
> 06 SATA controller
> 00 Vendor specific
> 01 AHCI 1.0
> 07 Serial Attached SCSI controller
> 80 Mass storage controller

Virtio used to advertise itself as class 0x0180 aka "mass storage 
controller".  You can even switch the driver into that mode for backward 
compatibility reasons.  Problem with that is that you can't get windows 
driver certifications then ...

cheers,
   Gerd