Re: [OxLUG] 4k sector size disks

Top Page
Message as email
+ (text/plain)
Delete this message
Reply to this message
Author: Simon Iremonger (oxlug)
To: Oxfordshire Linux User Group Discussion List
Subject: Re: [OxLUG] 4k sector size disks
> no, users are turning up with disks which present the 4k sectors to the OS:
>  kernel: usb 2-5: new high speed USB device using ehci_hcd and address 6
>  kernel: usb 2-5: configuration #1 chosen from 1 choice
>  kernel: scsi10 : SCSI emulation for USB Mass Storage devices
>  kernel:   Vendor: HitachiG  Model: ST                Rev: 0000
>  kernel:   Type:   Direct-Access                      ANSI SCSI
> revision: 06
>  kernel: SCSI device sdb: 976754431 4096-byte hdwr sectors (4000786 MB)
>  kernel: sdb: Write Protect is off
>  kernel: sdb: assuming drive cache: write through
>  kernel: SCSI device sdb: 976754431 4096-byte hdwr sectors (4000786 MB)
>  kernel: sdb: Write Protect is off
>  kernel: sdb: assuming drive cache: write through
>  kernel:  sdb: sdb1
>  kernel: sd 10:0:0:0: Attached scsi disk sdb
>  kernel: sd 10:0:0:0: Attached scsi generic sg1 type 0

Hrrm That is via USB? Does the same drive do the same thing
when presented over SATA bypassing the USB bridge controller?

But, fair-enough, it may be the case that the latest "very large"
disks really are no longer saying 512-byte-sectors...

I'm wondering if the USB mass storage protocol, has some LBA limit

> I think this is probably inevitable given that it is a 4TB disk...

I'm not convinced.
With LBA28, 512byte sectors, was a traditional 128GB limit.

However, now there is LBA48 ;-).
That looks like 128 Binary Petabytes according to my calculations,
with 512byte sectors...

2^48 * 512 / 1048576 /  1024  / 1024  / 1024
  (bytes)      (->mb)  (->gb)  (->tb)  (->pb)

>> This reminds me, there was the case, in DOS, dos 3.3 or so
>>    introduced FAT16 but couldn't cope with>64k sectors and hence
>>    was limited to 32mb/partition and weird drivers were invented
>>    to do weird translation to present the illusion of larger sectors
>>    to dos, in order that it could cope, or something ;-).
> are you talking about LBA? Logical Block Allocation was a replacement
> for Cylinders, Heads, Sectors (CHS).

No no no this predated all of that LBA stuffs ;-).

> so far I have ascertained that RHEL5 is incompatible with disks which
> have 4k sectors. RHEL6 is probably ok but I would like to have a disk
> that I can use for testing.

OK now all makes sense, I have a machine with a large 4k sector drive
but would need to reboot it to compare kernel messages and comment.
I do wonder if the "hdwr sectors" COULD refer to the fact that linux
''knows'' it is 4k-sectors but the interface still speaks 512...
Maybe not =).

> see above. I need the OS to be compatible, I do not want to right my own
> kernel space drivers!

Aaaaah! Now that you have explained you are actually getting drives
that, when connected over USB, appear as 4k sector drives, this
makes more sense. In that case, apologies, I don't know yet.

Seemingly there is a "Reduced Block Commands" USB mass storage
interface, but it isn't clear what this does/doesn't support
in terms of LBA size!!
Maybe there is a sector size translation going on in the
SATA->USB bridge insize the unit... Again, maybe-not!

Supposedly, some 4k physical sector drives were provided as an
'engineering sample' !! Code introduced in Linux 2.6.37

Presumably RHEL5 is <2.6.37 kernel...

It is also interesting to notice that Microsoft say
  "Most of the drives larger than 2 terabytes and with a USB
    connection are this kind of drive." [4k native]

I don't have a direct answer/solution, but hope the above notes/
links will be of interest to some, at least.