Mount by WWN (FC, using qlogic 2200)

Mount by WWN (FC, using qlogic 2200)

[NH Support]

Hello,

Is it possible to mount a drive based on it's WWN when using the qla2xxx 
  drivers?

Thanks,
Jason
NH

Re: Mount by WWN (FC, using qlogic 2200)

[EXT / DEVOTEAM VAROQUI Christophe]

Hello,

Which path would you expect to mount by specifying the WWID ?

cvaroqui

> Hello,
> 
> Is it possible to mount a drive based on it's WWN when using the qla2xxx 
> drivers?
>

Re: Mount by WWN (FC, using qlogic 2200)

[Robert J. Adams (jason)]

EXT / DEVOTEAM VAROQUI Christophe wrote:

> Which path would you expect to mount by specifying the WWID ?

In Solaris it's possible to tie down a particular WWN to a device, 
otherwise the device my be detected at a different point at the next 
boot (thus get assigned a different /dev/sdX). Is this possible w/ Linux?

Thanks,
jason
NH

Re: Mount by WWN (FC, using qlogic 2200)

[Martin Peschke3]

Hi,

> > Which path would you expect to mount by specifying the WWID ?
>
> In Solaris it's possible to tie down a particular WWN to a device,
> otherwise the device my be detected at a different point at the next
> boot (thus get assigned a different /dev/sdX). Is this possible w/ Linux?

As there might be several paths to a device identified by a WWPN
in the form of 2 or more host bus adapters connecting to that WWPN,
don't you want to mount an FCP device by
<adapter identifier> + <WWPN> + <FCP_LUN> ?

There has been a discussion about how to represent such
device attributes in sysfs. If udev acknowledged these attributes
and created proper devices nodes, then you could safely
mount your FCP devices.

If multipathing support matters, then you want your favourite mp solution
to lay claim to the SCSI device node identified as above and
provide another device node, which might then have any convenient
name.

That's my theory...

Mit freundlichen Grüßen / with kind regards

Martin Peschke

IBM Deutschland Entwicklung GmbH
Linux for eServer Development
Phone: +49-(0)7031-16-2349


"Robert J. Adams (jason)" <radams@siscom.net>@vger.kernel.org on 04/02/2004
19:39:22

Sent by:    linux-scsi-owner@vger.kernel.org


To:    EXT / DEVOTEAM VAROQUI Christophe <ext.devoteam.varoqui@sncf.fr>
cc:    "'linux-scsi@vger.kernel.org'" <linux-scsi@vger.kernel.org>
Subject:    Re: Mount by WWN (FC, using qlogic 2200)


EXT / DEVOTEAM VAROQUI Christophe wrote:

> Which path would you expect to mount by specifying the WWID ?

In Solaris it's possible to tie down a particular WWN to a device,
otherwise the device my be detected at a different point at the next
boot (thus get assigned a different /dev/sdX). Is this possible w/ Linux?

Thanks,
jason
NH
-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html



-
To unsubscribe from this list: send the line "unsubscribe linux-scsi" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: Mount by WWN (FC, using qlogic 2200)

[James Bottomley]

On Wed, 2004-02-04 at 14:58, Martin Peschke3 wrote:
> As there might be several paths to a device identified by a WWPN
> in the form of 2 or more host bus adapters connecting to that WWPN,
> don't you want to mount an FCP device by
> <adapter identifier> + <WWPN> + <FCP_LUN> ?

Actually, I think the question was about WWN, i.e. the EVPD page 0x83
World Wide Name, which uniquely identifies the device independent of
transport.

The answer is that someone needs to think about doing it using udev.

James

Re: Mount by WWN (FC, using qlogic 2200)

[Patrick Mansfield]

On Wed, Feb 04, 2004 at 03:06:33PM -0500, James Bottomley wrote:
> On Wed, 2004-02-04 at 14:58, Martin Peschke3 wrote:
> > As there might be several paths to a device identified by a WWPN
> > in the form of 2 or more host bus adapters connecting to that WWPN,
> > don't you want to mount an FCP device by
> > <adapter identifier> + <WWPN> + <FCP_LUN> ?
> 
> Actually, I think the question was about WWN, i.e. the EVPD page 0x83
> World Wide Name, which uniquely identifies the device independent of
> transport.
> 
> The answer is that someone needs to think about doing it using udev.

[added hotplug list to the cc]

scsi_id is included with the udev source now, and can be used to solve
most of the problem.

We still need something that tells udev what to do with multiple matches -
for multi-pathed sd's showing up multiple times, we don't want to drop the
other sd's, and that is what will happen right now with udev.

-- Patrick Mansfield

Re: Mount by WWN (FC, using qlogic 2200)

[James Bottomley]

On Wed, 2004-02-04 at 18:12, Bryan Henderson wrote:
> But it doesn't identify the "device" as Linux defines a device.  Setting
> aside multiple paths for the moment, a Linux sd device corresponds to an
> FCP logical unit, and the WWN identifies an FCP target (aka physical unit).
> I.e. where a physical unit has multiple logical units, multiple linux sd
> devices can have the same WWN.

Actually, the SPC-3 (rev 16) section 7.6.4.1 does say the Device
Identification VPD Page ... [shall apply] to the logical unit.

Of course, not all devices obey this, as you say, but still.

James

Re: Mount by WWN (FC, using qlogic 2200)

[Bryan Henderson]

> On Wed, 2004-02-04 at 18:12, Bryan Henderson wrote:
> > But it doesn't identify the "device" as Linux defines a device. 
Setting
> > aside multiple paths for the moment, a Linux sd device corresponds to 
an
> > FCP logical unit, and the WWN identifies an FCP target (aka physical 
unit).
> > I.e. where a physical unit has multiple logical units, multiple linux 
sd
> > devices can have the same WWN.
> 
> Actually, the SPC-3 (rev 16) section 7.6.4.1 does say the Device
> Identification VPD Page ... [shall apply] to the logical unit.
> 
> Of course, not all devices obey this, as you say, but still.

I stand corrected.  Upon further investigation, I realize that I confused 
WWN with WWNN.  WWN (worldwide name) is a generic term referring to, well, 
world wide names.  There are at least 3 WWNs in FCP systems:  WWPN 
(worldwide port name) to identify a port in an FC network, WWNN (worldwide 
node name) to identify a node in an FC network, which corresponds to an 
FCP target (physical unit), and a logical unit WWN, which I have seen 
called LUSN (Logical Unit Serial Number), but now I see others calling it 
the LU WWN and logical unit name too.

I haven't see any relevant standards, but I'll take your word for it that 
SPC-3 calls the VPD field in question a 'WWN' and because it's a VPD page 
for a logical, that means it is the LU WWN.

RE: Mount by WWN (FC, using qlogic 2200)

[EXT / DEVOTEAM VAROQUI Christophe]

Note that the current behaviour of the multipath config tool (packaged with
udev) is to map a multipath DM target over a LUN if it has at least one
path. I.e. in a single path environment, you would get a DM target mapped
over every sd behind a FC HBA. Using this tool, one can just parse through
the configured DM to find the WWID to mount.

> On Wed, Feb 04, 2004 at 03:06:33PM -0500, James Bottomley wrote:
> > On Wed, 2004-02-04 at 14:58, Martin Peschke3 wrote:
> > > As there might be several paths to a device identified by a WWPN
> > > in the form of 2 or more host bus adapters connecting to that WWPN,
> > > don't you want to mount an FCP device by
> > > <adapter identifier> + <WWPN> + <FCP_LUN> ?
> > 
> > Actually, I think the question was about WWN, i.e. the EVPD page 0x83
> > World Wide Name, which uniquely identifies the device independent of
> > transport.
> > 
> > The answer is that someone needs to think about doing it using udev.
> 
> [added hotplug list to the cc]
> 
> scsi_id is included with the udev source now, and can be used to solve
> most of the problem.
> 
> We still need something that tells udev what to do with multiple matches -
> for multi-pathed sd's showing up multiple times, we don't want to drop the
> other sd's, and that is what will happen right now with udev.
> 
> -- Patrick Mansfield

Re: Mount by WWN (FC, using qlogic 2200)

[Tom Coughlan]

On Wed, 2004-02-04 at 19:33, Bryan Henderson wrote:

> I stand corrected.  Upon further investigation, I realize that I confused 
> WWN with WWNN.  WWN (worldwide name) is a generic term referring to, well, 
> world wide names.  There are at least 3 WWNs in FCP systems:  WWPN 
> (worldwide port name) to identify a port in an FC network, WWNN (worldwide 
> node name) to identify a node in an FC network, which corresponds to an 
> FCP target (physical unit), and a logical unit WWN, which I have seen 
> called LUSN (Logical Unit Serial Number), but now I see others calling it 
> the LU WWN and logical unit name too.
> 
> I haven't see any relevant standards, but I'll take your word for it that 
> SPC-3 calls the VPD field in question a 'WWN' and because it's a VPD page 
> for a logical, that means it is the LU WWN.

Each identifier on the SPC-3 Device Identification VPD page (page 83)
has an "association" field, with one of the following values:

- The IDENTIFIER field is associated with the addressed logical unit. 
- The IDENTIFIER field is associated with the SCSI target port that
received the request. 
- The IDENTIFIER field is associated with the SCSI target device that
contains the addressed logical unit. 

So, an identifier on page 83 may be either an LU WWN, WWPN, or a WWNN. 

We need to check the association field, and use the identifier
appropriately, and we need to insist that storage vendors implement it
correctly.

Re: Mount by WWN (FC, using qlogic 2200)

[James Bottomley]

On Thu, 2004-02-05 at 08:56, Tom Coughlan wrote:
> Each identifier on the SPC-3 Device Identification VPD page (page 83)
> has an "association" field, with one of the following values:
> 
> - The IDENTIFIER field is associated with the addressed logical unit. 
> - The IDENTIFIER field is associated with the SCSI target port that
> received the request. 
> - The IDENTIFIER field is associated with the SCSI target device that
> contains the addressed logical unit. 
> 
> So, an identifier on page 83 may be either an LU WWN, WWPN, or a WWNN. 

Actually, this isn't an either/or, the page may contain a list. 
However, the standard is very clear (7.6.4.11.3) that each LUN that is
not a well known LUN shall have at least one identifier of association 0
(Logical Unit Identifier).

> We need to check the association field, and use the identifier
> appropriately, and we need to insist that storage vendors implement it
> correctly.

I agree, but I see this as a userspace issue.

James

Re: Mount by WWN (FC, using qlogic 2200)

[Patrick Mansfield]

On Thu, Feb 05, 2004 at 10:10:29AM -0500, James Bottomley wrote:
> On Thu, 2004-02-05 at 08:56, Tom Coughlan wrote:

> > We need to check the association field, and use the identifier
> > appropriately, and we need to insist that storage vendors implement it
> > correctly.
> 
> I agree, but I see this as a userspace issue.

Yes, I have this code in scsi_id (scsi_serial.c), AFAIR the same as we
once had in the kernel. The return 1 means we won't use this id:

        /* 
         * ASSOCIATION must be with the device (value 0)
         */
        if ((page_83[1] & 0x30) != 0)
                return 1;

The source code is included with the udev release (under the extras
directory), or at:

http://www-124.ibm.com/storageio/scsi_id/scsi_id-0.3.tar.gz

-- Patrick Mansfield