Hyper-V and IDE disks > 128GB

Hyper-V and IDE disks > 128GB

[David Shaw]

Hello,

I am seeing some rather odd behavior when using Linux (I've tried 2.6.30.10, 2.6.34.7, and 2.6.35.6, all with the same results) under the Hyper-V virtualization platform.  Specifically, the virtual IDE disks probe as "ATA-0", which seems to turn off 48-bit LBA, thus capping the disk size at 128GB.

This seems to have worked properly (at least in the sense that the full size of the disk was visible) on kernel 2.6.27.41.  See http://www.serverphorums.com/read.php?12,79638

In any event, this is how the virtual disk shows up at boot:

Oct 15 11:06:06 localhost kernel: ata1.01: ATA-0: Virtual HD, 1.1.0, max MWDMA2
Oct 15 11:06:06 localhost kernel: ata1.01: 267382800 sectors, multi 128: LBA

hdparm -I /dev/sdb says:

/dev/sdb:

ATA device, with non-removable media
	Model Number:       Virtual HD                              
	Serial Number:      
	Firmware Revision:  1.1.0   
Standards:
	Likely used: 2
Configuration:
	Logical		max	current
	cylinders	16383	65535
	heads		16	16
	sectors/track	63	255
	--
	bytes/track: 65024	bytes/sector: 512
	CHS current addressable sectors:  267382800
	LBA    user addressable sectors:  267382800
	Logical/Physical Sector size:           512 bytes
	device size with M = 1024*1024:      130558 MBytes
	device size with M = 1000*1000:      136899 MBytes (136 GB)
	cache/buffer size  = 64 KBytes (type=DualPortCache)
Capabilities:
	LBA, IORDY(can be disabled)
	Buffer size: 64.0kB
	Standby timer values: spec'd by Vendor
	R/W multiple sector transfer: Max = 128	Current = 128
	DMA: sdma0 sdma1 sdma2 mdma0 mdma1 *mdma2 
	     Cycle time: min=120ns recommended=120ns
	PIO: pio0 pio1 pio2 pio3 pio4 
	     Cycle time: no flow control=333ns  IORDY flow control=120ns

Can someone suggest a direction to go in next?

Thanks very much,

David

Re: Hyper-V and IDE disks > 128GB

[Alan Cox]

> with the same results) under the Hyper-V virtualization platform.  Specifically, the virtual IDE disks
> probe as "ATA-0", which seems to turn off 48-bit LBA, thus capping the
> disk size at 128GB.

Early ATA does not support LBA48 so that behaviour is correct from Linux

> 
> This seems to have worked properly (at least in the sense that the full size of the disk was visible) on kernel 2.6.27.41.  See http://www.serverphorums.com/read.php?12,79638
> 
> In any event, this is how the virtual disk shows up at boot:
> 
> Oct 15 11:06:06 localhost kernel: ata1.01: ATA-0: Virtual HD, 1.1.0, max MWDMA2
> Oct 15 11:06:06 localhost kernel: ata1.01: 267382800 sectors, multi 128: LBA

> Can someone suggest a direction to go in next?

Well it looks like a bug in the hypedvisor so I guess the first steps are
to check with Microsoft if there is an update and to open a bug report
with them. If the volume naming is consistent then I guess we can also in
theory add a hack for Linux to recognize the broken volumes (assuming
LBA48 actually works if you ignore the drive identify data)

Re: Hyper-V and IDE disks > 128GB

[David Shaw]

On Nov 2, 2010, at 3:27 PM, Alan Cox wrote:

>> with the same results) under the Hyper-V virtualization platform.  Specifically, the virtual IDE disks
>> probe as "ATA-0", which seems to turn off 48-bit LBA, thus capping the
>> disk size at 128GB.
> 
> Early ATA does not support LBA48 so that behaviour is correct from Linux
> 
>> 
>> This seems to have worked properly (at least in the sense that the full size of the disk was visible) on kernel 2.6.27.41.  See http://www.serverphorums.com/read.php?12,79638
>> 
>> In any event, this is how the virtual disk shows up at boot:
>> 
>> Oct 15 11:06:06 localhost kernel: ata1.01: ATA-0: Virtual HD, 1.1.0, max MWDMA2
>> Oct 15 11:06:06 localhost kernel: ata1.01: 267382800 sectors, multi 128: LBA
> 
>> Can someone suggest a direction to go in next?
> 
> Well it looks like a bug in the hypedvisor so I guess the first steps are
> to check with Microsoft if there is an update and to open a bug report
> with them. If the volume naming is consistent then I guess we can also in
> theory add a hack for Linux to recognize the broken volumes (assuming
> LBA48 actually works if you ignore the drive identify data)

I'll file a report with Microsoft and see what happens.  Thanks!

David

Re: Hyper-V and IDE disks > 128GB

[Tim Small]

On 02/11/10 18:52, David Shaw wrote:
> Can someone suggest a direction to go in next?
>   

I think I've seen this problem reported before, with the same diagnosis
that Alan gave.  i.e. it's a HyperV bug, albeit one which doesn't affect
Windows operating systems.

As I believe Microsoft and Redhat have some sort of co-support agreement
with respect to each other's virtualisation platforms, you may want to
report it on Redhat's bugzilla (after verifying that the fault is also
present in the RHEL6 beta).  I'd guess they'll either implement a
workaround or have some relatively straightforward mechanism to report
the fault to Microsoft - courtesy of their co-support agreement...

Cheers,

Tim.