Detection of Advanced Format drives

Detection of Advanced Format drives

[Arun]

Is there any way to detect the Advanced Format hard drives with 4K
sector size? I understand that above kernel v. 2.6.31 it is supported.
But is it possible detect automatically if the drive is an AF drive
and how?

In case this is not the correct place to ask please redirect me to
where I should look for an answer.

Thanks,

Arun

Re: Detection of Advanced Format drives

[Alan Cox]

On Mon, 26 Sep 2011 15:25:52 +0530
Arun <engineerarun@gmail.com> wrote:

> Is there any way to detect the Advanced Format hard drives with 4K
> sector size? I understand that above kernel v. 2.6.31 it is supported.
> But is it possible detect automatically if the drive is an AF drive
> and how?

It's in the drive identify data (see the ATA 8 specification documents)

Re: Detection of Advanced Format drives

[Joshua Scoggins]

Keep in mind that some AF drives will not report their actual sector
sizes. I have this issue with the WD AV-25 series. This is to provide
compatibility with older operating systems that don't support AF
drives. These are called generation 1 devices. Newer drives will
directly report their sector size as 4096 through the drive identify
data older ones will say 512. At that point you just need to do a
google search to see if it's AF or not.

-Josh


On Mon, Sep 26, 2011 at 3:15 AM, Alan Cox <alan@lxorguk.ukuu.org.uk> wrote:
> On Mon, 26 Sep 2011 15:25:52 +0530
> Arun <engineerarun@gmail.com> wrote:
>
>> Is there any way to detect the Advanced Format hard drives with 4K
>> sector size? I understand that above kernel v. 2.6.31 it is supported.
>> But is it possible detect automatically if the drive is an AF drive
>> and how?
>
> It's in the drive identify data (see the ATA 8 specification documents)
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

Re: Detection of Advanced Format drives

[Mike Snitzer]

On Mon, Sep 26, 2011 at 3:48 PM, Joshua Scoggins
<theoretically.x64@gmail.com> wrote:
> Keep in mind that some AF drives will not report their actual sector
> sizes. I have this issue with the WD AV-25 series. This is to provide
> compatibility with older operating systems that don't support AF
> drives. These are called generation 1 devices. Newer drives will
> directly report their sector size as 4096 through the drive identify
> data older ones will say 512. At that point you just need to do a
> google search to see if it's AF or not.

Linux should publish the discrepancy with these <disk> specific sysfs files:

/sys/block/<disk>/queue/logical_block_size=512
/sys/block/<disk>/queue/physical_block_size=4096

But are you saying you have an AF disk that doesn't publish this
discrepancy via the IDENTIFY DEVICE (ATA) or READ CAPACITY 16 (SCSI)
command?

Re: Detection of Advanced Format drives

[Joshua Scoggins]

Didn't know about the physical_block_size field. My AV-25 reports the
physical_block_size as 4096 and the logical block size as 512.

Thanks for the heads up and sorry about the misinformation. Not my
intention to spread FUD like that

On Mon, Sep 26, 2011 at 1:09 PM, Mike Snitzer <snitzer@gmail.com> wrote:
> On Mon, Sep 26, 2011 at 3:48 PM, Joshua Scoggins
> <theoretically.x64@gmail.com> wrote:
>> Keep in mind that some AF drives will not report their actual sector
>> sizes. I have this issue with the WD AV-25 series. This is to provide
>> compatibility with older operating systems that don't support AF
>> drives. These are called generation 1 devices. Newer drives will
>> directly report their sector size as 4096 through the drive identify
>> data older ones will say 512. At that point you just need to do a
>> google search to see if it's AF or not.
>
> Linux should publish the discrepancy with these <disk> specific sysfs files:
>
> /sys/block/<disk>/queue/logical_block_size=512
> /sys/block/<disk>/queue/physical_block_size=4096
>
> But are you saying you have an AF disk that doesn't publish this
> discrepancy via the IDENTIFY DEVICE (ATA) or READ CAPACITY 16 (SCSI)
> command?
>

Re: Detection of Advanced Format drives

[Arun]

Mike,

Regarding "Linux does, and will bubble this additional information up
the stack."...

Should I suppose that it is an already available piece of information
to the userland? Or do I need to modify the hdparm/fdisk code to get
this piece of info?

Once again, if this is not the place to ask about
userland/hdpark/fdisk etc. please redirect to the correct location.


Regards,

Arun




On Tue, Sep 27, 2011 at 23:29, Joshua Scoggins
<theoretically.x64@gmail.com> wrote:
> Didn't know about the physical_block_size field. My AV-25 reports the
> physical_block_size as 4096 and the logical block size as 512.
>
> Thanks for the heads up and sorry about the misinformation. Not my
> intention to spread FUD like that
>
> On Mon, Sep 26, 2011 at 1:09 PM, Mike Snitzer <snitzer@gmail.com> wrote:
>> On Mon, Sep 26, 2011 at 3:48 PM, Joshua Scoggins
>> <theoretically.x64@gmail.com> wrote:
>>> Keep in mind that some AF drives will not report their actual sector
>>> sizes. I have this issue with the WD AV-25 series. This is to provide
>>> compatibility with older operating systems that don't support AF
>>> drives. These are called generation 1 devices. Newer drives will
>>> directly report their sector size as 4096 through the drive identify
>>> data older ones will say 512. At that point you just need to do a
>>> google search to see if it's AF or not.
>>
>> Linux should publish the discrepancy with these <disk> specific sysfs files:
>>
>> /sys/block/<disk>/queue/logical_block_size=512
>> /sys/block/<disk>/queue/physical_block_size=4096
>>
>> But are you saying you have an AF disk that doesn't publish this
>> discrepancy via the IDENTIFY DEVICE (ATA) or READ CAPACITY 16 (SCSI)
>> command?
>>
>

Re: Detection of Advanced Format drives

[Martin K. Petersen]

>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:

Arun> Should I suppose that it is an already available piece of
Arun> information to the userland? 

Yes, in the sysfs locations Mike mentioned:

  /sys/block/foo/queue/physical_block_size

Or you can use the BLKPBSZGET ioctl.


Arun> Or do I need to modify the hdparm/fdisk code to get this piece of
Arun> info?

Recent hdparm versions will also report the right thing:

# hdparm -I /dev/sdc | grep Sector
        Logical  Sector size:                   512 bytes
        Physical Sector size:                  4096 bytes
        Logical Sector-0 offset:                512 bytes

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: Detection of Advanced Format drives

[Arun]

Thanks for confirming "hdparm -I" Martin... I guessed it would resolve
my detection problem but was not quite sure as I'm still waiting to
get the drive and had some time to get my hands dirty :)

Thanks everyone for taking interest in my problem.

Regards,

Arun



On Wed, Sep 28, 2011 at 00:29, Martin K. Petersen
<martin.petersen@oracle.com> wrote:
>>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:
>
> Arun> Should I suppose that it is an already available piece of
> Arun> information to the userland?
>
> Yes, in the sysfs locations Mike mentioned:
>
>  /sys/block/foo/queue/physical_block_size
>
> Or you can use the BLKPBSZGET ioctl.
>
>
> Arun> Or do I need to modify the hdparm/fdisk code to get this piece of
> Arun> info?
>
> Recent hdparm versions will also report the right thing:
>
> # hdparm -I /dev/sdc | grep Sector
>        Logical  Sector size:                   512 bytes
>        Physical Sector size:                  4096 bytes
>        Logical Sector-0 offset:                512 bytes
>
> --
> Martin K. Petersen      Oracle Linux Engineering
>

Re: Detection of Advanced Format drives

[Mark Knecht]

On Tue, Sep 27, 2011 at 11:59 AM, Martin K. Petersen
<martin.petersen@oracle.com> wrote:
>>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:
>
> Arun> Should I suppose that it is an already available piece of
> Arun> information to the userland?
>
> Yes, in the sysfs locations Mike mentioned:
>
>  /sys/block/foo/queue/physical_block_size
>
> Or you can use the BLKPBSZGET ioctl.
>
>
> Arun> Or do I need to modify the hdparm/fdisk code to get this piece of
> Arun> info?
>
> Recent hdparm versions will also report the right thing:
>
> # hdparm -I /dev/sdc | grep Sector
>        Logical  Sector size:                   512 bytes
>        Physical Sector size:                  4096 bytes
>        Logical Sector-0 offset:                512 bytes
>
> --
> Martin K. Petersen      Oracle Linux Engineering

As someone else pointed out earlier, not every drive with 4K sectors
will tell you it has 4K sectors, at least not using hdparm. This is a
WD 1TB drive circa early last year for which I've not updated
firmware. (If there is an update)

gandalf ~ # hdparm -I /dev/sda | grep Sector
        Logical/Physical Sector size:           512 bytes
gandalf ~ #

gandalf ~ # cat /sys/block/sda/queue/physical_block_size
512
gandalf ~ #

- Mark

Re: Detection of Advanced Format drives

[Arun]

Mark,

I got your point.

A quick check of hdparm tells me that it is using the SG_IO ioctl for
the ATA_OP_PIDENTIFY command. Maybe I have some hopes of having a
correct result with the BLKPBSZGET ioctl. Will check that out.

Regards,

Arun



On Wed, Sep 28, 2011 at 01:40, Mark Knecht <markknecht@gmail.com> wrote:
>
> On Tue, Sep 27, 2011 at 11:59 AM, Martin K. Petersen
> <martin.petersen@oracle.com> wrote:
> >>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:
> >
> > Arun> Should I suppose that it is an already available piece of
> > Arun> information to the userland?
> >
> > Yes, in the sysfs locations Mike mentioned:
> >
> >  /sys/block/foo/queue/physical_block_size
> >
> > Or you can use the BLKPBSZGET ioctl.
> >
> >
> > Arun> Or do I need to modify the hdparm/fdisk code to get this piece of
> > Arun> info?
> >
> > Recent hdparm versions will also report the right thing:
> >
> > # hdparm -I /dev/sdc | grep Sector
> >        Logical  Sector size:                   512 bytes
> >        Physical Sector size:                  4096 bytes
> >        Logical Sector-0 offset:                512 bytes
> >
> > --
> > Martin K. Petersen      Oracle Linux Engineering
>
> As someone else pointed out earlier, not every drive with 4K sectors
> will tell you it has 4K sectors, at least not using hdparm. This is a
> WD 1TB drive circa early last year for which I've not updated
> firmware. (If there is an update)
>
> gandalf ~ # hdparm -I /dev/sda | grep Sector
>        Logical/Physical Sector size:           512 bytes
> gandalf ~ #
>
> gandalf ~ # cat /sys/block/sda/queue/physical_block_size
> 512
> gandalf ~ #
>
> - Mark

Re: Detection of Advanced Format drives

[Martin K. Petersen]

>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:

Arun> A quick check of hdparm tells me that it is using the SG_IO ioctl
Arun> for the ATA_OP_PIDENTIFY command. Maybe I have some hopes of
Arun> having a correct result with the BLKPBSZGET ioctl. Will check that
Arun> out.

hdparm and the kernel are looking the same place for the physical block
size. There should be no discrepancy between what hdparm can see and
what's reported in physical_block_size by the kernel. The BLKPBSZGET
returns the same topology parameter as sysfs' physical_block_size.

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: Detection of Advanced Format drives

[NamJae Jeon]

2011/9/28 Martin K. Petersen <martin.petersen@oracle.com>:
>>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:
>
> Arun> A quick check of hdparm tells me that it is using the SG_IO ioctl
> Arun> for the ATA_OP_PIDENTIFY command. Maybe I have some hopes of
> Arun> having a correct result with the BLKPBSZGET ioctl. Will check that
> Arun> out.
>
> hdparm and the kernel are looking the same place for the physical block
> size. There should be no discrepancy between what hdparm can see and
> what's reported in physical_block_size by the kernel. The BLKPBSZGET
> returns the same topology parameter as sysfs' physical_block_size.
>
> --
> Martin K. Petersen      Oracle Linux Engineering
> --
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>
plz remember, There are native 4k and 512B emulate.
In case of 512B emulate, you can see logical block size is 512Byte and
physical block size is 4KB.
Although you can access 512Byte, It should be access per 4KB in order
to prevent the loss of performance by RMW penalty.

Re: Detection of Advanced Format drives

[Martin K. Petersen]

>>>>> "Arun" == Arun  <engineerarun@gmail.com> writes:

Arun> [...] couldn't find any direct way of detecting an SSD drive like
Arun> an ioctl.

Arun> Is there any such direct way of detecting SSD drives yet? I am
Arun> working on kernel version 2.6.32-19.

You can check /sys/block/foo/queue/rotational.

The matching BLKROTATIONAL ioctl didn't go in until very recently.

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: Detection of Advanced Format drives

[Martin K. Petersen]

>>>>> "Joshua" == Joshua Scoggins <theoretically.x64@gmail.com> writes:

Joshua> Keep in mind that some AF drives will not report their actual
Joshua> sector sizes. I have this issue with the WD AV-25 series. This
Joshua> is to provide compatibility with older operating systems that
Joshua> don't support AF drives. These are called generation 1
Joshua> devices. Newer drives will directly report their sector size as
Joshua> 4096 through the drive identify data older ones will say 512. 

AF drives continue to report a logical block size of 512. There's an
additional parameter in IDENTIFY DEVICE that describes the physical
block size. Legacy operating systems do not check the physical block
size parameter. Linux does, and will bubble this additional information
up the stack.

There have been a few attempts at switching ATA drives over to 4KB
logical blocks as well. We are ready. However, both boot and legacy OS
compatibility concerns make this a bit of an uphill battle.

As you mention there's one vendor that pretty consistently gets their
physical block size reporting wrong. We've pointed this out to them many
times. Apparently it's been fixed with the latest green drive
firmware...

-- 
Martin K. Petersen	Oracle Linux Engineering

Re: Detection of Advanced Format drives

[Martin K. Petersen]

>>>>> "Alan" == Alan Cox <alan@lxorguk.ukuu.org.uk> writes:

>> Is there any way to detect the Advanced Format hard drives with 4K
>> sector size? I understand that above kernel v. 2.6.31 it is
>> supported.  But is it possible detect automatically if the drive is
>> an AF drive and how?

Alan> It's in the drive identify data (see the ATA 8 specification
Alan> documents)

Or:

  http://oss.oracle.com/~mkp/docs/linux-advanced-storage.pdf

-- 
Martin K. Petersen	Oracle Linux Engineering