USB3 3TB HDD boot

USB3 3TB HDD boot

[Bob Lemar]

Hi, guys

I got new SEAGATE Expansion USB 3.0 and want to boot linux from it.

I have successfully booted various linuxes before with different
loaders from USB thumb flashes. But I cannot do it with this external
drive.

There are three problems: USB3, 4K sectors and >2TB.

I try grub2 (fedora 17). Grub2 doesn't power on external USB drive.
The drive is spinning up only after some OS is starting to load after
grub2. So grub2 and BIOS don't see external HDD.


Can someone point me out:

1. either how to turn on external USB3 driver while computer is in grub2?

2. or how to load OS with kexec from external USB3 4K sector drive
(from OS which is located at internal HDD)?

--
Best regards, Bob

Re: USB3 3TB HDD boot

[Seth Goldberg]

Hi,

  Is this a 4k native drive? (4k physical AND logical sector size) or is it a 512e drive?

 In any case, if your system's BIOS does not present it as a boot option, you could try GRUB's USB support, but it is a bit flaky (and doesn't directly support usb3, so depending on how your USB controller is wired, you may not even be able to do that.)

  --S

On Feb 12, 2013, at 7:36 AM, Bob Lemar <forega17@gmail.com> wrote:

> Hi, guys
> 
> I got new SEAGATE Expansion USB 3.0 and want to boot linux from it.
> 
> I have successfully booted various linuxes before with different
> loaders from USB thumb flashes. But I cannot do it with this external
> drive.
> 
> There are three problems: USB3, 4K sectors and >2TB.
> 
> I try grub2 (fedora 17). Grub2 doesn't power on external USB drive.
> The drive is spinning up only after some OS is starting to load after
> grub2. So grub2 and BIOS don't see external HDD.
> 
> 
> Can someone point me out:
> 
> 1. either how to turn on external USB3 driver while computer is in grub2?
> 
> 2. or how to load OS with kexec from external USB3 4K sector drive
> (from OS which is located at internal HDD)?
> 
> --
> Best regards, Bob
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel

Re: USB3 3TB HDD boot

[Aleš Nesrsta]

Hi Bob,

it is probably not GRUB bug according to Your description - it looks to
be BIOS problem.

1.
Currently I have no idea about Your HW, but it looks like Your xHCI (USB
3.0 controller) is not "native" part of the PC handled by BIOS, i.e.
BIOS cannot boot from any device connected to USB 3.0 port.

Or, if You are able to boot some another USB 3.0 devices on the same USB
3.0 port, BIOS probably cannot handle such huge device (4K sectors and
>2TB) - or has some another problem with Your new device.
But, unfortunately, the final result is the same - BIOS cannot load GRUB
and any other boot loader from Your new SEAGATE drive => You cannot boot
anything directly from such device.

There is no (easy) solution for such situation - it looks to be similar
situation like e.g. if You want boot from CD on some very old machine
where BIOS is able to boot only from floppy and it doesn't support
booting from CD...

Possibly You can use some additional small USB flash disc connected to
USB port where it can be handled by BIOS. Install GRUB manually on this
additional device together with copy of linux and initrd images from
SEAGATE Expansion, properly modify grub.cfg and boot Linux kernel from
it. I am afraid there is no better way. :-(

2.
Maybe is better to use GRUB BugTracker for such things, this case looks
more like bug report than developer's talk.. :-)

BR,
Ales

Bob Lemar wrote:
> Hi, guys
> 
> I got new SEAGATE Expansion USB 3.0 and want to boot linux from it.
> 
> I have successfully booted various linuxes before with different
> loaders from USB thumb flashes. But I cannot do it with this external
> drive.
> 
> There are three problems: USB3, 4K sectors and >2TB.
> 
> I try grub2 (fedora 17). Grub2 doesn't power on external USB drive.
> The drive is spinning up only after some OS is starting to load after
> grub2. So grub2 and BIOS don't see external HDD.
> 
> 
> Can someone point me out:
> 
> 1. either how to turn on external USB3 driver while computer is in grub2?
> 
> 2. or how to load OS with kexec from external USB3 4K sector drive
> (from OS which is located at internal HDD)?
> 
> --
> Best regards, Bob
> 
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel

Re: USB3 3TB HDD boot

[Bob Lemar]

Thank you for your replies.


Yes, it is definitely BIOS problems as it doesn't allow to boot from
external drive directly (via BIOS boot order settings). It doesn't
show this drive at <F10> boot menu even after soft reboot. So BIOS
doesn't provide access to GRUB like to any other drives or USB
flashes.

And I suppose to think that it is not a bug, it is a feature. And
feature of external drive, not a BIOS. Vendors of spinning external
drive try to limit "uptime" of theirs HDD. External HDD could start as
soon as USB is powered but do not want to do it. It only starts when
OS tell it to start up (this commands are magic for me).

Facts:

- External drive is logically 4K (not a 512e thing). It is not a problem yet.

- USB 3 is not a problem yet. The drive doesn't work with USB 2 just
the same way. So it is not "nativeness" BIOS-USB hub communication
problem.

- BIOS behavior is correct (it power up USB);

- external drive behavior is correct (it is powered but starts only on
demand and it is sleeping in many cases including PC start);

- GRUB behavior is correct (it boots from drives that BIOS told him
about; only OS can detect this device as mass storage because it was
designed that way).

- Ales, you were absolutely right with floppy for booting CD trick
example. While MB, HDD firmwares and bootloader have predictable and
correct behavior some users would like to have bootloader feature
(GRUB command) that wake up external HDDs and re-detect drives.

Resolution:

- Boot USB thumb and use kexec to boot external HDD.
I did: (1) clone linux to external drive; (2) correct fstab at
external; (3) boot into internal drive OS; (4) mount external HDD
which became visible; (5) kexec images and boot external drive. // I
am not sure that only external drive files were used at second boot --
need to do clear experiment with minimal os at internal device or
thumb flash. kboot and petitboot are possible solve this problem.

- It is good point to create feature request at BugTracker. I wish
GRUB has command to power up external USB HDD and do manual redetect.
But I don't know what is happening technically when OS do it. I guess
I should find out it to be more specific while filling bug.

On Tue, Feb 12, 2013 at 11:19 PM, Aleš Nesrsta <starous@volny.cz> wrote:
> Hi Bob,
>
> it is probably not GRUB bug according to Your description - it looks to
> be BIOS problem.
>
> 1.
> Currently I have no idea about Your HW, but it looks like Your xHCI (USB
> 3.0 controller) is not "native" part of the PC handled by BIOS, i.e.
> BIOS cannot boot from any device connected to USB 3.0 port.
>
> Or, if You are able to boot some another USB 3.0 devices on the same USB
> 3.0 port, BIOS probably cannot handle such huge device (4K sectors and
>>2TB) - or has some another problem with Your new device.
> But, unfortunately, the final result is the same - BIOS cannot load GRUB
> and any other boot loader from Your new SEAGATE drive => You cannot boot
> anything directly from such device.
>
> There is no (easy) solution for such situation - it looks to be similar
> situation like e.g. if You want boot from CD on some very old machine
> where BIOS is able to boot only from floppy and it doesn't support
> booting from CD...
>
> Possibly You can use some additional small USB flash disc connected to
> USB port where it can be handled by BIOS. Install GRUB manually on this
> additional device together with copy of linux and initrd images from
> SEAGATE Expansion, properly modify grub.cfg and boot Linux kernel from
> it. I am afraid there is no better way. :-(
>
> 2.
> Maybe is better to use GRUB BugTracker for such things, this case looks
> more like bug report than developer's talk.. :-)
>
> BR,
> Ales
>
> Bob Lemar wrote:
>> Hi, guys
>>
>> I got new SEAGATE Expansion USB 3.0 and want to boot linux from it.
>>
>> I have successfully booted various linuxes before with different
>> loaders from USB thumb flashes. But I cannot do it with this external
>> drive.
>>
>> There are three problems: USB3, 4K sectors and >2TB.
>>
>> I try grub2 (fedora 17). Grub2 doesn't power on external USB drive.
>> The drive is spinning up only after some OS is starting to load after
>> grub2. So grub2 and BIOS don't see external HDD.
>>
>>
>> Can someone point me out:
>>
>> 1. either how to turn on external USB3 driver while computer is in grub2?
>>
>> 2. or how to load OS with kexec from external USB3 4K sector drive
>> (from OS which is located at internal HDD)?
>>
>> --
>> Best regards, Bob
>>
>> _______________________________________________
>> Grub-devel mailing list
>> Grub-devel@gnu.org
>> https://lists.gnu.org/mailman/listinfo/grub-devel
>
>
>
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel



-- 
Best regards, Bob

Re: USB3 3TB HDD boot

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 1353 bytes --]

On 04.03.2013 17:27, Bob Lemar wrote:

> Thank you for your replies.
> 
> 
> Yes, it is definitely BIOS problems as it doesn't allow to boot from
> external drive directly (via BIOS boot order settings). It doesn't
> show this drive at <F10> boot menu even after soft reboot. So BIOS
> doesn't provide access to GRUB like to any other drives or USB
> flashes.
> 
> And I suppose to think that it is not a bug, it is a feature. And
> feature of external drive, not a BIOS.


Wrong. Not supporting 4K disks is simply a lack of support, nothing
else. As for manufacturers producing such drives reasons are several and
partially in an attempt to mitigate msdos partition limitation.

> - BIOS behavior is correct (it power up USB);

It has also to propose the disk to boot from and if chosen, boot from it.
> - Ales, you were absolutely right with floppy for booting CD trick

> example. While MB, HDD firmwares and bootloader have predictable and
> correct behavior some users would like to have bootloader feature
> (GRUB command) that wake up external HDDs and re-detect drives.

BIOS accessing disks is already wakeup enough.> I did: (1) clone linux
to external drive; (2) correct fstab at.
You can access the disk in GRUB using own GRUB modules but you can't
chainload on such drives as it would require BIOS support.


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: USB3 3TB HDD boot

[Lennart Sorensen]

On Mon, Mar 04, 2013 at 08:27:41PM +0400, Bob Lemar wrote:
> Thank you for your replies.
> 
> 
> Yes, it is definitely BIOS problems as it doesn't allow to boot from
> external drive directly (via BIOS boot order settings). It doesn't
> show this drive at <F10> boot menu even after soft reboot. So BIOS
> doesn't provide access to GRUB like to any other drives or USB
> flashes.
> 
> And I suppose to think that it is not a bug, it is a feature. And
> feature of external drive, not a BIOS. Vendors of spinning external
> drive try to limit "uptime" of theirs HDD. External HDD could start as
> soon as USB is powered but do not want to do it. It only starts when
> OS tell it to start up (this commands are magic for me).
> 
> Facts:
> 
> - External drive is logically 4K (not a 512e thing). It is not a problem yet.
> 
> - USB 3 is not a problem yet. The drive doesn't work with USB 2 just
> the same way. So it is not "nativeness" BIOS-USB hub communication
> problem.
> 
> - BIOS behavior is correct (it power up USB);
> 
> - external drive behavior is correct (it is powered but starts only on
> demand and it is sleeping in many cases including PC start);
> 
> - GRUB behavior is correct (it boots from drives that BIOS told him
> about; only OS can detect this device as mass storage because it was
> designed that way).
> 
> - Ales, you were absolutely right with floppy for booting CD trick
> example. While MB, HDD firmwares and bootloader have predictable and
> correct behavior some users would like to have bootloader feature
> (GRUB command) that wake up external HDDs and re-detect drives.
> 
> Resolution:
> 
> - Boot USB thumb and use kexec to boot external HDD.
> I did: (1) clone linux to external drive; (2) correct fstab at
> external; (3) boot into internal drive OS; (4) mount external HDD
> which became visible; (5) kexec images and boot external drive. // I
> am not sure that only external drive files were used at second boot --
> need to do clear experiment with minimal os at internal device or
> thumb flash. kboot and petitboot are possible solve this problem.
> 
> - It is good point to create feature request at BugTracker. I wish
> GRUB has command to power up external USB HDD and do manual redetect.
> But I don't know what is happening technically when OS do it. I guess
> I should find out it to be more specific while filling bug.

The fact it is 3TB would mean it uses GPT, not old style DOS partition
table.  Maybe the BIOS doesn't support booting from a GPT partitioned
drive.  If it had a hybrid partition table, then it might work assuming
other things are not also causing problems.

After all some bioses only list a drive as bootable if they see a
partition marked bootable in a dos style partition table.

-- 
Len Sorensen

Re: USB3 3TB HDD boot

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 3046 bytes --]

On 04.03.2013 20:19, Lennart Sorensen wrote:

> On Mon, Mar 04, 2013 at 08:27:41PM +0400, Bob Lemar wrote:
>> Thank you for your replies.
>>
>>
>> Yes, it is definitely BIOS problems as it doesn't allow to boot from
>> external drive directly (via BIOS boot order settings). It doesn't
>> show this drive at <F10> boot menu even after soft reboot. So BIOS
>> doesn't provide access to GRUB like to any other drives or USB
>> flashes.
>>
>> And I suppose to think that it is not a bug, it is a feature. And
>> feature of external drive, not a BIOS. Vendors of spinning external
>> drive try to limit "uptime" of theirs HDD. External HDD could start as
>> soon as USB is powered but do not want to do it. It only starts when
>> OS tell it to start up (this commands are magic for me).
>>
>> Facts:
>>
>> - External drive is logically 4K (not a 512e thing). It is not a problem yet.
>>
>> - USB 3 is not a problem yet. The drive doesn't work with USB 2 just
>> the same way. So it is not "nativeness" BIOS-USB hub communication
>> problem.
>>
>> - BIOS behavior is correct (it power up USB);
>>
>> - external drive behavior is correct (it is powered but starts only on
>> demand and it is sleeping in many cases including PC start);
>>
>> - GRUB behavior is correct (it boots from drives that BIOS told him
>> about; only OS can detect this device as mass storage because it was
>> designed that way).
>>
>> - Ales, you were absolutely right with floppy for booting CD trick
>> example. While MB, HDD firmwares and bootloader have predictable and
>> correct behavior some users would like to have bootloader feature
>> (GRUB command) that wake up external HDDs and re-detect drives.
>>
>> Resolution:
>>
>> - Boot USB thumb and use kexec to boot external HDD.
>> I did: (1) clone linux to external drive; (2) correct fstab at
>> external; (3) boot into internal drive OS; (4) mount external HDD
>> which became visible; (5) kexec images and boot external drive. // I
>> am not sure that only external drive files were used at second boot --
>> need to do clear experiment with minimal os at internal device or
>> thumb flash. kboot and petitboot are possible solve this problem.
>>
>> - It is good point to create feature request at BugTracker. I wish
>> GRUB has command to power up external USB HDD and do manual redetect.
>> But I don't know what is happening technically when OS do it. I guess
>> I should find out it to be more specific while filling bug.
> 
> The fact it is 3TB would mean it uses GPT, not old style DOS partition
> table.

Not necessarily. 3T drives often use msdos, just with 4K sectors which
increases the range of msdos to 16T.

>  Maybe the BIOS doesn't support booting from a GPT partitioned
> drive.  If it had a hybrid partition table, then it might work assuming
> other things are not also causing problems.
> 
> After all some bioses only list a drive as bootable if they see a
> partition marked bootable in a dos style partition table.
> 




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: USB3 3TB HDD boot

[Lennart Sorensen]

On Mon, Mar 04, 2013 at 08:33:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> Not necessarily. 3T drives often use msdos, just with 4K sectors which
> increases the range of msdos to 16T.

Hmm, I suppose that would work.  I have never seen one that did that yet,
but I don't have that many 3TB drives either.

-- 
Len Sorensen

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 4, 2013, at 9:27 AM, Bob Lemar <forega17@gmail.com> wrote:

> - External drive is logically 4K (not a 512e thing). It is not a problem yet.

They are now shipping? When you use 'parted -l' does it say Sector size is 4096B/4096B or does it say 512B/4096B ? What model?


Chris Murphy

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 4, 2013, at 12:43 PM, "Lennart Sorensen" <lsorense@csclub.uwaterloo.ca> wrote:

> On Mon, Mar 04, 2013 at 08:33:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>> Not necessarily. 3T drives often use msdos, just with 4K sectors which
>> increases the range of msdos to 16T.
> 
> Hmm, I suppose that would work.  I have never seen one that did that yet,
> but I don't have that many 3TB drives either.

I wonder what the BIOS does when it "asks" the drive for LBA 0, and receives 4096 bytes instead of 512 bytes. I'd think the BIOS would have to tolerate receiving 8x the data. I didn't know 4Kn drives were shipping yet. I'm not finding anything remotely recent about this when googling it.

Chris Murphy

Re: USB3 3TB HDD boot

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 1048 bytes --]

On 05.03.2013 00:16, Chris Murphy wrote:

> 
> On Mar 4, 2013, at 12:43 PM, "Lennart Sorensen" <lsorense@csclub.uwaterloo.ca> wrote:
> 
>> On Mon, Mar 04, 2013 at 08:33:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>>> Not necessarily. 3T drives often use msdos, just with 4K sectors which
>>> increases the range of msdos to 16T.
>>
>> Hmm, I suppose that would work.  I have never seen one that did that yet,
>> but I don't have that many 3TB drives either.
> 
> I wonder what the BIOS does when it "asks" the drive for LBA 0, and receives 4096 bytes instead of 512 bytes.

It doesn't get that far. Most likely it bails out after reading SCSI
descriptor.

> I'd think the BIOS would have to tolerate receiving 8x the data. I didn't know 4Kn drives were shipping yet. I'm not finding anything remotely recent about this when googling it.
> 
> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 5, 2013, at 12:22 AM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:

> On 05.03.2013 00:16, Chris Murphy wrote:
> 
>> 
>> On Mar 4, 2013, at 12:43 PM, "Lennart Sorensen" <lsorense@csclub.uwaterloo.ca> wrote:
>> 
>>> On Mon, Mar 04, 2013 at 08:33:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
>>>> Not necessarily. 3T drives often use msdos, just with 4K sectors which
>>>> increases the range of msdos to 16T.
>>> 
>>> Hmm, I suppose that would work.  I have never seen one that did that yet,
>>> but I don't have that many 3TB drives either.
>> 
>> I wonder what the BIOS does when it "asks" the drive for LBA 0, and receives 4096 bytes instead of 512 bytes.
> 
> It doesn't get that far. Most likely it bails out after reading SCSI
> descriptor.

I have anecdotal evidence these drives are now in the wild. Mac user with a new 3TB Seagate USB 3 external drive:
Device / Media Name:      Seagate Backup+ Desk Media

This is not very descriptive. But when I ask the user to display the GPT with gdisk, there's normal output, except for this line:
Logical sector size: 4096 bytes

Next, when I have him report the result from:
sudo dd if=/dev/disk3 count=2 2>/dev/null | hexdump -C

The 1st 512 bytes reported is the PMBR, but the 2nd 512 bytes is garbage, not the GPT header as expected.

When this drive is connected to the computer and he tries to boot Windows, the system hangs at a black screen. That's kinda interesting/concerning, if even CSM-BIOS implementations on recent EFI firmware can have problems with such drives.


Chris Murphy

Re: USB3 3TB HDD boot

[Andrey Borzenkov]

В Thu, 14 Mar 2013 20:18:46 -0600
Chris Murphy <lists@colorremedies.com> пишет:

> 
> On Mar 5, 2013, at 12:22 AM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:
> 
> > On 05.03.2013 00:16, Chris Murphy wrote:
> > 
> >> 
> >> On Mar 4, 2013, at 12:43 PM, "Lennart Sorensen" <lsorense@csclub.uwaterloo.ca> wrote:
> >> 
> >>> On Mon, Mar 04, 2013 at 08:33:51PM +0100, Vladimir 'φ-coder/phcoder' Serbinenko wrote:
> >>>> Not necessarily. 3T drives often use msdos, just with 4K sectors which
> >>>> increases the range of msdos to 16T.
> >>> 
> >>> Hmm, I suppose that would work.  I have never seen one that did that yet,
> >>> but I don't have that many 3TB drives either.
> >> 
> >> I wonder what the BIOS does when it "asks" the drive for LBA 0, and receives 4096 bytes instead of 512 bytes.
> > 
> > It doesn't get that far. Most likely it bails out after reading SCSI
> > descriptor.
> 
> I have anecdotal evidence these drives are now in the wild. Mac user with a new 3TB Seagate USB 3 external drive:
> Device / Media Name:      Seagate Backup+ Desk Media
> 
> This is not very descriptive. But when I ask the user to display the GPT with gdisk, there's normal output, except for this line:
> Logical sector size: 4096 bytes
> 
> Next, when I have him report the result from:
> sudo dd if=/dev/disk3 count=2 2>/dev/null | hexdump -C
> 
> The 1st 512 bytes reported is the PMBR, but the 2nd 512 bytes is garbage, not the GPT header as expected.
> 

It is possible that you need to read in physical sector size; what if
you try bs=4k?

> When this drive is connected to the computer and he tries to boot Windows, the system hangs at a black screen. That's kinda interesting/concerning, if even CSM-BIOS implementations on recent EFI firmware can have problems with such drives.
> 
> 
> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 14, 2013, at 8:38 PM, Andrey Borzenkov <arvidjaar@gmail.com> wrote:
>> 
>> I have anecdotal evidence these drives are now in the wild. Mac user with a new 3TB Seagate USB 3 external drive:
>> Device / Media Name:      Seagate Backup+ Desk Media
>> 
>> This is not very descriptive. But when I ask the user to display the GPT with gdisk, there's normal output, except for this line:
>> Logical sector size: 4096 bytes
>> 
>> Next, when I have him report the result from:
>> sudo dd if=/dev/disk3 count=2 2>/dev/null | hexdump -C
>> 
>> The 1st 512 bytes reported is the PMBR, but the 2nd 512 bytes is garbage, not the GPT header as expected.
>> 
> 
> It is possible that you need to read in physical sector size; what if
> you try bs=4k?

Right, for a 4Kn drive, to read LBA 1 and get the GPT header I'd need:

sudo dd if=/dev/disk3 skip=8 count=1 2>/dev/null | hexdump -C


Chris Murphy

Re: USB3 3TB HDD boot

[Lennart Sorensen]

On Thu, Mar 14, 2013 at 08:56:18PM -0600, Chris Murphy wrote:
> Right, for a 4Kn drive, to read LBA 1 and get the GPT header I'd need:
> 
> sudo dd if=/dev/disk3 skip=8 count=1 2>/dev/null | hexdump -C

Wouldn't using 'bs=4096' make things simpler and more obvious?

-- 
Len Sorensen

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 15, 2013, at 8:51 AM, Lennart Sorensen <lsorense@csclub.uwaterloo.ca> wrote:

> On Thu, Mar 14, 2013 at 08:56:18PM -0600, Chris Murphy wrote:
>> Right, for a 4Kn drive, to read LBA 1 and get the GPT header I'd need:
>> 
>> sudo dd if=/dev/disk3 skip=8 count=1 2>/dev/null | hexdump -C
> 
> Wouldn't using 'bs=4096' make things simpler and more obvious?

How would it?

These structures are still predicated on a 512 byte block. The MBR is always 512 bytes, but on a 4096/4096 4Kn drive, LBA 0 is 4096 bytes. So if I don't want to see 3584 bytes of useless garbage, I can't set the dd block size to 4096, or I get extra information.

If I want to see just the GPT header, which while only ~92 bytes, by spec it gets its own sector, there's far less superfluous information using a bs of 512 bytes than 4096 bytes.


Chris Murphy

Re: USB3 3TB HDD boot

[Bruce Dubbs]

Chris Murphy wrote:
>
> On Mar 15, 2013, at 8:51 AM, Lennart Sorensen
> <lsorense@csclub.uwaterloo.ca> wrote:
>
>> On Thu, Mar 14, 2013 at 08:56:18PM -0600, Chris Murphy wrote:
>>> Right, for a 4Kn drive, to read LBA 1 and get the GPT header I'd
>>> need:
>>>
>>> sudo dd if=/dev/disk3 skip=8 count=1 2>/dev/null | hexdump -C
>>
>> Wouldn't using 'bs=4096' make things simpler and more obvious?
>
> How would it?
>
> These structures are still predicated on a 512 byte block. The MBR is
> always 512 bytes, but on a 4096/4096 4Kn drive, LBA 0 is 4096 bytes.
> So if I don't want to see 3584 bytes of useless garbage, I can't set
> the dd block size to 4096, or I get extra information.
>
> If I want to see just the GPT header, which while only ~92 bytes, by
> spec it gets its own sector, there's far less superfluous information
> using a bs of 512 bytes than 4096 bytes.

There is more than one way to do things.

sudo dd if=/dev/disk3 bs=4096 skip=1 count=1 | cut -c 1-92 | hexdump -C

is one way that matches the physical drive.

The first mentioned way works too, but doesn't reflect as well what is 
really going on.

   -- Bruce

Re: USB3 3TB HDD boot

[Lennart Sorensen]

On Fri, Mar 15, 2013 at 09:59:22AM -0600, Chris Murphy wrote:
> How would it?
> 
> These structures are still predicated on a 512 byte block. The MBR is always 512 bytes, but on a 4096/4096 4Kn drive, LBA 0 is 4096 bytes. So if I don't want to see 3584 bytes of useless garbage, I can't set the dd block size to 4096, or I get extra information.

Well if you want to get LBA 1 on a 4096 byte sector device, using dd
bs=4096 skip=1 count=1 would get that sector and make it more obvious
which sector you are getting (to me at least).

So yeah you get extra stuff that way, but it is what is really there
after all.  If you only wanted 92 bytes you could have used a bs=128 as
well and gotten even less junk.

> If I want to see just the GPT header, which while only ~92 bytes, by spec it gets its own sector, there's far less superfluous information using a bs of 512 bytes than 4096 bytes.

True, although the drive has to read 4096 bytes no matter how much you
ask for.

-- 
Len Sorensen

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 15, 2013, at 11:53 AM, Lennart Sorensen <lsorense@csclub.uwaterloo.ca> wrote:

> On Fri, Mar 15, 2013 at 09:59:22AM -0600, Chris Murphy wrote:
> 
>> If I want to see just the GPT header, which while only ~92 bytes, by spec it gets its own sector, there's far less superfluous information using a bs of 512 bytes than 4096 bytes.
> 
> True, although the drive has to read 4096 bytes no matter how much you
> ask for.

OK.

In any case, these drives are floating around, and with at least one company's firmware (of unknown affected models) it causes a hang when trying to boot Windows. If the CSM-BIOS method is used to install Linux, I'd expect that to hang as well when such a drive is attached. I think it's reasonable to expect more strange behaviors as a result of these drives.

Chris Murphy

Re: USB3 3TB HDD boot

[Bob Lemar]

Guys, forget about 4K sectors. The problem is starting USB attached drive.

On Fri, Mar 15, 2013 at 10:38 PM, Chris Murphy <lists@colorremedies.com> wrote:
>
> On Mar 15, 2013, at 11:53 AM, Lennart Sorensen <lsorense@csclub.uwaterloo.ca> wrote:
>
>> On Fri, Mar 15, 2013 at 09:59:22AM -0600, Chris Murphy wrote:
>>
>>> If I want to see just the GPT header, which while only ~92 bytes, by spec it gets its own sector, there's far less superfluous information using a bs of 512 bytes than 4096 bytes.
>>
>> True, although the drive has to read 4096 bytes no matter how much you
>> ask for.
>
> OK.
>
> In any case, these drives are floating around, and with at least one company's firmware (of unknown affected models) it causes a hang when trying to boot Windows. If the CSM-BIOS method is used to install Linux, I'd expect that to hang as well when such a drive is attached. I think it's reasonable to expect more strange behaviors as a result of these drives.
>
> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel



--
Best regards, Bob

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 18, 2013, at 12:34 PM, Bob Lemar <forega17@gmail.com> wrote:

> Guys, forget about 4K sectors. The problem is starting USB attached drive.

Other USB drives aren't causing the reported problem, including other USB 3TB drives.

Maybe the issue is the computer is USB 2.0 and this particular 4Kn drive is in a USB 3.0 enclosure, even if in theory the USB 3.0 drive should just fallback to 2.0.


Chris Murphy

Re: USB3 3TB HDD boot

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 773 bytes --]

On 18.03.2013 19:56, Chris Murphy wrote:

> 
> On Mar 18, 2013, at 12:34 PM, Bob Lemar <forega17@gmail.com> wrote:
> 
>> Guys, forget about 4K sectors. The problem is starting USB attached drive.
> 
> Other USB drives aren't causing the reported problem, including other USB 3TB drives.
> 
> Maybe the issue is the computer is USB 2.0 and this particular 4Kn drive is in a USB 3.0 enclosure, even if in theory the USB 3.0 drive should just fallback to 2.0.
> 

It doesn't matter. I also tried with a 160GB 4Kn native disk. It wasn't
bootable on any BIOS-based system either

> 
> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel
> 




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 18, 2013, at 1:41 PM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:

> On 18.03.2013 19:56, Chris Murphy wrote:
> 
>> 
>> On Mar 18, 2013, at 12:34 PM, Bob Lemar <forega17@gmail.com> wrote:
>> 
>>> Guys, forget about 4K sectors. The problem is starting USB attached drive.
>> 
>> Other USB drives aren't causing the reported problem, including other USB 3TB drives.
>> 
>> Maybe the issue is the computer is USB 2.0 and this particular 4Kn drive is in a USB 3.0 enclosure, even if in theory the USB 3.0 drive should just fallback to 2.0.
>> 
> 
> It doesn't matter. I also tried with a 160GB 4Kn native disk. It wasn't
> bootable on any BIOS-based system either

What this user is reporting isn't an attempt to boot off the 4Kn disk. The EFI CSM seems to be hanging just by virtue of having the 4Kn disk attached (via USB). That disk has a GPT with PMBR, I had him zero first 5 sectors so they don't contain boot loader code or other garbage.

Nevertheless what you're reporting, if its a wide spread behavior, obviates any value of 4Kn drives breathing life into MBR/BIOS such that they could support 2.2+TB boot drives. At least, short of flash updates for the BIOS.

Chris Murphy

Re: USB3 3TB HDD boot

[Vladimir 'φ-coder/phcoder' Serbinenko]

[-- Attachment #1: Type: text/plain, Size: 1784 bytes --]

On 18.03.2013 22:05, Chris Murphy wrote:

> 
> On Mar 18, 2013, at 1:41 PM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:
> 
>> On 18.03.2013 19:56, Chris Murphy wrote:
>>
>>>
>>> On Mar 18, 2013, at 12:34 PM, Bob Lemar <forega17@gmail.com> wrote:
>>>
>>>> Guys, forget about 4K sectors. The problem is starting USB attached drive.
>>>
>>> Other USB drives aren't causing the reported problem, including other USB 3TB drives.
>>>
>>> Maybe the issue is the computer is USB 2.0 and this particular 4Kn drive is in a USB 3.0 enclosure, even if in theory the USB 3.0 drive should just fallback to 2.0.
>>>
>>
>> It doesn't matter. I also tried with a 160GB 4Kn native disk. It wasn't
>> bootable on any BIOS-based system either
> 
> What this user is reporting isn't an attempt to boot off the 4Kn disk. The EFI CSM seems to be hanging just by virtue of having the 4Kn disk attached (via USB). That disk has a GPT with PMBR, I had him zero first 5 sectors so they don't contain boot loader code or other garbage.
> 
> Nevertheless what you're reporting, if its a wide spread behavior, obviates any value of 4Kn drives breathing life into MBR/BIOS such that they could support 2.2+TB boot drives. At least, short of flash updates for the BIOS.
> 

BIOS interfaces support up-to 64-bit LBA adressing so >2T is just issue
of fixing bugs in BIOS. On, the other hand problem with 4Kn is deeper as
it doesn't seem to even be any implicit agreement of how such disks
should be handled for booting. Nor does it seem any impmentation on
either bootloader or BIOS side.

> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel




[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 294 bytes --]

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 18, 2013, at 3:46 PM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:
> 
> BIOS interfaces support up-to 64-bit LBA adressing so >2T is just issue
> of fixing bugs in BIOS.

Aha. OK good to know.

> On, the other hand problem with 4Kn is deeper as
> it doesn't seem to even be any implicit agreement of how such disks
> should be handled for booting. Nor does it seem any impmentation on
> either bootloader or BIOS side.

This is concerning. So at the moment the 4Kn drives are effectively data only, non-bootable. Is it the same situation for UEFI and BIOS? It seems the 4Kn release is premature. The 512e method has few downsides for any recent partitioning tool, which ensures alignment, and they are also bootable without concerns. I'm unaware of a downside.


Chris Murphy

Re: USB3 3TB HDD boot

[Chris Murphy]

On Mar 18, 2013, at 3:46 PM, Vladimir 'φ-coder/phcoder' Serbinenko <phcoder@gmail.com> wrote:

> BIOS interfaces support up-to 64-bit LBA adressing so >2T is just issue
> of fixing bugs in BIOS.

Actually, as I think about it, >2TB with BIOS is a dead end for most of the market because Windows on BIOS requires MBR, and thus far there's no extension for MBR beyond 32-bit addressing. There's sort of a crude work around that can allow up to 4.4TB disks with MBR, requiring two partitions; since MBR specifies partitions with starting LBA and number of partitions rather than ending LBA. I don't know this yet, but I suspect this is how Apple has implemented 3TB support for CSM booting (totally ill advised in my opinion).

GRUB+Linux of course can use GPT in such a case, so long as the BIOS bugs are fixed to tolerate GPT (which some BIOS's notably Lenovo) the larger disk sizes are bootable.

Chris Murphy

RE: USB3 3TB HDD boot

[Elliott, Robert (Server Storage)]

> -----Original Message-----
> From: grub-devel-bounces+elliott=hp.com@gnu.org [mailto:grub-devel-
> bounces+elliott=hp.com@gnu.org] On Behalf Of Chris Murphy
> Sent: Monday, 18 March, 2013 5:37 PM
> To: The development of GNU GRUB
> Subject: Re: USB3 3TB HDD boot
> 
> 
> On Mar 18, 2013, at 3:46 PM, Vladimir 'φ-coder/phcoder' Serbinenko
> <phcoder@gmail.com> wrote:
> >
> > BIOS interfaces support up-to 64-bit LBA adressing so >2T is just issue
> > of fixing bugs in BIOS.
> 
> Aha. OK good to know.
> 
> > On, the other hand problem with 4Kn is deeper as
> > it doesn't seem to even be any implicit agreement of how such disks
> > should be handled for booting. Nor does it seem any impmentation on
> > either bootloader or BIOS side.
> 
> This is concerning. So at the moment the 4Kn drives are effectively data only,
> non-bootable. Is it the same situation for UEFI and BIOS? It seems the 4Kn
> release is premature. The 512e method has few downsides for any recent
> partitioning tool, which ensures alignment, and they are also bootable without
> concerns. I'm unaware of a downside.

One reason is that 4Kn forces all I/Os to be aligned, while 512e allows non-aligned I/Os.  512e drives are somewhat obligated to ensure that writing one 512 byte logical block won't affect the other seven 512-byte logical blocks sharing that physical block if a power loss occurs or an error happens; this is not simple to do. It is cheaper and easier just to ship the drive as 4Kn and force software to adapt.  That's why you are seeing this first in cheap large-capacity USB drives.

CD/DVDs use 2048 byte logical block sizes with the CDFS ISO 9660 filesystem, and most systems can boot from CDs nowadays, so booting from non-512 byte logical block sizes is not impossible.  There's just a lot of legacy BIOS, MBR based code that assumes 512 byte logical blocks on hard drives (and 32-bit LBAs).

UEFI systems using GPT are likely to support booting from 4Kn drives and use 64-bit LBAs throughout.  Some OSes already claim boot support from 4Kn drives in those systems.  The UEFI spec has included warnings about detecting the logical block size since 2006.

---
Rob Elliott    HP Server Storage


> 
> Chris Murphy
> _______________________________________________
> Grub-devel mailing list
> Grub-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/grub-devel

Re: USB3 3TB HDD boot

[Phillip Susi]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 3/18/2013 6:59 PM, Chris Murphy wrote:
> Actually, as I think about it, >2TB with BIOS is a dead end for
> most of the market because Windows on BIOS requires MBR, and thus
> far there's no extension for MBR beyond 32-bit addressing. There's
> sort

I wonder what happens when you throw "dynamic disks" into the mix?
Could you have the system partition only in the limited MBR, then ldm
takes over and can address the whole disk?

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2.0.17 (MingW32)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQEcBAEBAgAGBQJRScaZAAoJEJrBOlT6nu75eQMH/inJd62eUiuvPXNo78sDnd8j
50CGb+i7EgfA/k2jT51k5rSre+HV4QyTANuhRiHCcF8MZG2zwFZOOsdhpiETqtR3
DypbxyAZ7gHn7aufT+5W1KJzdKb2hHY/Sww+FZDamSRuWzmLtYuLdP0LgibFHPGV
TFgztX5RnFBDoqjdv9Ff5hLfHmY3Qo+zdkWxCd1zaOTtPYdq5yrw3hHeCB+qKyTn
lXaBC7gpJmEVlrcw+UJDbOZzPktuVvJ0WJruzV6SJjJz6uoPYGeINM7lRID4i9s7
RjbFMlNm6DhHJwYHz+eN2DBrgXAXhZFVa6lnSYUj4YSf8w+LM4FMnHu69lmLvZA=
=Os3d
-----END PGP SIGNATURE-----