Unable to get INQUIRY vpd 1 page 0x0

Unable to get INQUIRY vpd 1 page 0x0

[li nux]

On doing a '/etc/init.d/boot.udev start' or 'multipath
-v2 -d' i get following error for all the devices:

creating device nodes 0:0:0:0: sg_io failed status 0x0
0x1 0x0 0x0
0:0:0:0: Unable to get INQUIRY vpd 1 page 0x0.
0:0:0:1: sg_io failed status 0x0 0x1 0x0 0x0

Can somebody have an idea, what i need to correct.

-lnxluv

__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

Re: Unable to get INQUIRY vpd 1 page 0x0

[Patrick Mansfield]

On Mon, Mar 13, 2006 at 03:00:46AM -0800, li nux wrote:
> On doing a '/etc/init.d/boot.udev start' or 'multipath
> -v2 -d' i get following error for all the devices:
> 
> creating device nodes 0:0:0:0: sg_io failed status 0x0
> 0x1 0x0 0x0
> 0:0:0:0: Unable to get INQUIRY vpd 1 page 0x0.
> 0:0:0:1: sg_io failed status 0x0 0x1 0x0 0x0
> 
> Can somebody have an idea, what i need to correct.

You can run scsi_id manually like:

	scsi_id -g -s /block/sda
	scsi_id -g -s /block/sdb

Or for whatever devices are at 0:0:0:0 and 0:0:0:1

Also try normal read/write to the devices.

And you should have posted the full dmesg output (that includes all
SCSI and perhaps PCI messages).

The devices are getting a host status of 1, that is a DID_NO_CONNECT and
generally means the initiator could not talk to the target (the linux
driver or hardware can't communicate with the disk), it is also used by
linux scsi core when it kills IO (like when a device is hot-unplugged
while IO is in progress).

If you get this error all the time, you won't be able to send any commands
to the device. Since the device actually showed up (and linux scsi core
device discovery sent at least an INQUIRY to it OK), something else is now
causing it to fail, you need to figure that out.

Depending on results, discussion should probably be moved to
linux-scsi@vger.kernel.org 

-- Patrick Mansfield

Re: Unable to get INQUIRY vpd 1 page 0x0

[li nux]

--- Patrick Mansfield <patmans@us.ibm.com> wrote:

> On Mon, Mar 13, 2006 at 03:00:46AM -0800, li nux
> wrote:
> > On doing a '/etc/init.d/boot.udev start' or
> 'multipath
> > -v2 -d' i get following error for all the devices:
> > 
> > creating device nodes 0:0:0:0: sg_io failed status
> 0x0
> > 0x1 0x0 0x0
> > 0:0:0:0: Unable to get INQUIRY vpd 1 page 0x0.
> > 0:0:0:1: sg_io failed status 0x0 0x1 0x0 0x0
> > 
> > Can somebody have an idea, what i need to correct.
> 
> You can run scsi_id manually like:
> 
> 	scsi_id -g -s /block/sda
> 	scsi_id -g -s /block/sdb
> 
> Or for whatever devices are at 0:0:0:0 and 0:0:0:1
> 

This also gives the same error
0:0:0:1: sg_io failed status 0x0 0x1 0x0 0x0
0:0:0:1: Unable to get INQUIRY vpd 1 page 0x0.
error calling out /sbin/scsi_id -g -u -s /block/sda

Why it is enquiring page 0x0, vpd enquires from page
0x80 or 0x83, right ?

> Also try normal read/write to the devices.
> 

This works

> And you should have posted the full dmesg output
> (that includes all
> SCSI and perhaps PCI messages).
> 
> The devices are getting a host status of 1, that is
> a DID_NO_CONNECT and
> generally means the initiator could not talk to the
> target (the linux
> driver or hardware can't communicate with the disk),
> it is also used by
> linux scsi core when it kills IO (like when a device
> is hot-unplugged
> while IO is in progress).

You are right, The devices were hot-unplugged and then
new devices were hot-plugged. But udev should take
care of these events and should refresh its
device-name mappings accordingly ?

After a rebooted, new disks under /proc/partitions
looks like cciss/c0d0p1 etc.
how should i run scsi_id on these ?

> 
> If you get this error all the time, you won't be
> able to send any commands
> to the device. Since the device actually showed up
> (and linux scsi core
> device discovery sent at least an INQUIRY to it OK),
> something else is now
> causing it to fail, you need to figure that out.
> 
> Depending on results, discussion should probably be
> moved to
> linux-scsi@vger.kernel.org 
> 
> -- Patrick Mansfield


__________________________________________________
Do You Yahoo!?
Tired of spam?  Yahoo! Mail has the best spam protection around 
http://mail.yahoo.com 

Re: Unable to get INQUIRY vpd 1 page 0x0

[Patrick Mansfield]

On Mon, Mar 13, 2006 at 09:03:11PM -0800, li nux wrote:
> --- Patrick Mansfield <patmans@us.ibm.com> wrote:
> 
> > On Mon, Mar 13, 2006 at 03:00:46AM -0800, li nux
> > wrote:
> > > On doing a '/etc/init.d/boot.udev start' or
> > 'multipath
> > > -v2 -d' i get following error for all the devices:
> > > 
> > > creating device nodes 0:0:0:0: sg_io failed status
> > 0x0
> > > 0x1 0x0 0x0
> > > 0:0:0:0: Unable to get INQUIRY vpd 1 page 0x0.
> > > 0:0:0:1: sg_io failed status 0x0 0x1 0x0 0x0
> > > 
> > > Can somebody have an idea, what i need to correct.
> > 
> > You can run scsi_id manually like:
> > 
> > 	scsi_id -g -s /block/sda
> > 	scsi_id -g -s /block/sdb
> > 
> > Or for whatever devices are at 0:0:0:0 and 0:0:0:1
> > 
> 
> This also gives the same error
> 0:0:0:1: sg_io failed status 0x0 0x1 0x0 0x0
> 0:0:0:1: Unable to get INQUIRY vpd 1 page 0x0.
> error calling out /sbin/scsi_id -g -u -s /block/sda

Where did the "error calling out" come from if you are running scsi_id
from the command line?

> Why it is enquiring page 0x0, vpd enquires from page
> 0x80 or 0x83, right ?

No. See the SCSI SPC spec, vital product data parameters, or look at
scsi_id source code. Page 0 gives a list of the supported vpd pages.

You can override scsi_id to skip page 0, and use vpd page 0x80 or 0x83.

> > Also try normal read/write to the devices.
> > 
> 
> This works
> 
> > And you should have posted the full dmesg output
> > (that includes all
> > SCSI and perhaps PCI messages).

It might help if you supply more information ... also your configuration
and distro.

> > The devices are getting a host status of 1, that is
> > a DID_NO_CONNECT and
> > generally means the initiator could not talk to the
> > target (the linux
> > driver or hardware can't communicate with the disk),
> > it is also used by
> > linux scsi core when it kills IO (like when a device
> > is hot-unplugged
> > while IO is in progress).
> 
> You are right, The devices were hot-unplugged and then
> new devices were hot-plugged. But udev should take
> care of these events and should refresh its
> device-name mappings accordingly ?

Yes.

> After a rebooted, new disks under /proc/partitions
> looks like cciss/c0d0p1 etc.
> how should i run scsi_id on these ?

I thought that was what you did above ...  I don't know about cciss, and
have never used it.

If this is a problem with the cciss devices, you have to figure it out or
find someone that knows more about them.

I guess they don't support VPD page 0x0, but are failing in a very weird
way.

-- Patrick Mansfield