[gentoo-user] Can't create valid btrfs on NVMe

[gentoo-user] Can't create valid btrfs on NVMe

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 759 bytes --]

I'm putting together a new desktop using a Samsung SM951 NVMe drive. I
booted sysrescd, partitioned the drive and ran

mkfs.btrfs /dev/nvme0n1p3
mount /dev/nvme0n1p3 /mnt/custom

df -T, mount and findmnt all show this is mounted as a btrfs filesystem,
e.g.

/dev/nvme0n1p3 on /mnt/custom type btrfs (rw,relatime,ssd,space_cache)

I can create files in here but cannot do anything btrfs-y

% btrfs filesystem show /mnt/custom
ERROR: not a valid btrfs filesystem: /mnt/custom

% btrfs subvolume create /mnt/custom/test
Create subvolume '/mnt/custom/test'
ERROR: cannot create subvolume: Inappropriate ioctl for device

Google has been no help at all.


-- 
Neil Bothwick

TEXAS VIRUS: Makes sure that it's bigger than any other file.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: [gentoo-user] Can't create valid btrfs on NVMe

[Adam Carter]

[-- Attachment #1: Type: text/plain, Size: 1088 bytes --]

On Thu, Aug 18, 2016 at 8:26 AM, Neil Bothwick <neil@digimed.co.uk> wrote:

> I'm putting together a new desktop using a Samsung SM951 NVMe drive. I
> booted sysrescd, partitioned the drive and ran
>
> mkfs.btrfs /dev/nvme0n1p3
> mount /dev/nvme0n1p3 /mnt/custom
>
> df -T, mount and findmnt all show this is mounted as a btrfs filesystem,
> e.g.
>
> /dev/nvme0n1p3 on /mnt/custom type btrfs (rw,relatime,ssd,space_cache)
>
> I can create files in here but cannot do anything btrfs-y
>
> % btrfs filesystem show /mnt/custom
> ERROR: not a valid btrfs filesystem: /mnt/custom
>
> % btrfs subvolume create /mnt/custom/test
> Create subvolume '/mnt/custom/test'
> ERROR: cannot create subvolume: Inappropriate ioctl for device
>
> Google has been no help at all.
>
>
I'm assuming tools are expecting SATA (or SCSI) and need an update or
alternative for nvm;

# hdparm -i /dev/nvme0n1

/dev/nvme0n1:
 HDIO_DRIVE_CMD(identify) failed: Inappropriate ioctl for device
 HDIO_GET_IDENTITY failed: Inappropriate ioctl for device

Maybe SCSI emulation could work around it?
CONFIG_BLK_DEV_NVME_SCSI

[-- Attachment #2: Type: text/html, Size: 1592 bytes --]

Re: [gentoo-user] Can't create valid btrfs on NVMe

[Peter Humphrey]

On Thursday 18 Aug 2016 15:56:50 Adam Carter wrote:
> On Thu, Aug 18, 2016 at 8:26 AM, Neil Bothwick <neil@digimed.co.uk> wrote:
> > I'm putting together a new desktop using a Samsung SM951 NVMe drive. I
> > booted sysrescd, partitioned the drive and ran
> > 
> > mkfs.btrfs /dev/nvme0n1p3
> > mount /dev/nvme0n1p3 /mnt/custom
> > 
> > df -T, mount and findmnt all show this is mounted as a btrfs filesystem,
> > e.g.
> > 
> > /dev/nvme0n1p3 on /mnt/custom type btrfs (rw,relatime,ssd,space_cache)
> > 
> > I can create files in here but cannot do anything btrfs-y
> > 
> > % btrfs filesystem show /mnt/custom
> > ERROR: not a valid btrfs filesystem: /mnt/custom
> > 
> > % btrfs subvolume create /mnt/custom/test
> > Create subvolume '/mnt/custom/test'
> > ERROR: cannot create subvolume: Inappropriate ioctl for device
> > 
> > Google has been no help at all.
> 
> I'm assuming tools are expecting SATA (or SCSI) and need an update or
> alternative for nvm;
> 
> # hdparm -i /dev/nvme0n1
> 
> /dev/nvme0n1:
>  HDIO_DRIVE_CMD(identify) failed: Inappropriate ioctl for device
>  HDIO_GET_IDENTITY failed: Inappropriate ioctl for device
> 
> Maybe SCSI emulation could work around it?
> CONFIG_BLK_DEV_NVME_SCSI

I don't have Neil's e-mail here - another KMail problem?

Neil, have you tried inspecting your drive with sys-apps/nvme-cli? I don't 
have any experience to relate as I haven't worked out how to use it yet, but 
it may help you.

-- 
Rgds
Peter

Re: [gentoo-user] Can't create valid btrfs on NVMe

[james]

On 08/18/2016 05:06 AM, Peter Humphrey wrote:
> On Thursday 18 Aug 2016 15:56:50 Adam Carter wrote:
>> On Thu, Aug 18, 2016 at 8:26 AM, Neil Bothwick <neil@digimed.co.uk> wrote:
>>> I'm putting together a new desktop using a Samsung SM951 NVMe drive. I
>>> booted sysrescd, partitioned the drive and ran
>>>
>>> mkfs.btrfs /dev/nvme0n1p3
>>> mount /dev/nvme0n1p3 /mnt/custom
>>>
>>> df -T, mount and findmnt all show this is mounted as a btrfs filesystem,
>>> e.g.
>>>
>>> /dev/nvme0n1p3 on /mnt/custom type btrfs (rw,relatime,ssd,space_cache)
>>>
>>> I can create files in here but cannot do anything btrfs-y
>>>
>>> % btrfs filesystem show /mnt/custom
>>> ERROR: not a valid btrfs filesystem: /mnt/custom
>>>
>>> % btrfs subvolume create /mnt/custom/test
>>> Create subvolume '/mnt/custom/test'
>>> ERROR: cannot create subvolume: Inappropriate ioctl for device
>>>
>>> Google has been no help at all.
>>
>> I'm assuming tools are expecting SATA (or SCSI) and need an update or
>> alternative for nvm;
>>
>> # hdparm -i /dev/nvme0n1
>>
>> /dev/nvme0n1:
>>  HDIO_DRIVE_CMD(identify) failed: Inappropriate ioctl for device
>>  HDIO_GET_IDENTITY failed: Inappropriate ioctl for device
>>
>> Maybe SCSI emulation could work around it?
>> CONFIG_BLK_DEV_NVME_SCSI
>
> I don't have Neil's e-mail here - another KMail problem?
>
> Neil, have you tried inspecting your drive with sys-apps/nvme-cli? I don't
> have any experience to relate as I haven't worked out how to use it yet, but
> it may help you.
>



Did you see the doc::
NVMeSSD_User_Installation_Guide_whitepaper-0.pdf

That's some fancy piece of hardware, that says it requires special drivers::

Chipset Intel 5520 or later generation chipset
Slots
• Require PCIe Gen3 x4 link width slot (For Max Performance)
• Support PCIe Gen1/Gen2 slot


The following operating systems are supported, with some requiring 
additional drivers:
• Windows Server® 2008R2 (64bit)
• Windows Server 2012/2012 R2 (64bit)
• Windows 7, 8, 8.1 (32/64bit)
• Red Hat® Enterprise Linux (RHEL) 7.0 (64bit)
• Red Hat Enterprise Linux (RHEL) 6.4/6.5 (64bit)
• SUSE® Linux Enterprise (SLES) 11SP3 (64bit)
• Ubuntu® 12.04.02 LTS Server (64bit)
• Ubuntu® 12.04.04 LTS Server (64bit)
• Solaris 11 SP2 (64bit)


Installing the Driver on a Linux System
RHEL 6.4 / 6.5 64bit
1. Turn the power on and log on to the system with the admin account.
2. Copy the driver file package to the installation folder.
3. Install the rpm with the command below. After installing, the NVMe 
driver will automatically load when the system is booted.
# rpm --ivh nvme-kmp-default-1.17_3.0.76_0.11-0.x86.64.rpm
(The file name could change depending on the driver version.)
4. To upgrade from v1.9, execute the following command:
# rpm --Uvh nvme-kmp-default-1.17_3.0.76_0.11-0.x86.64.rpm
5. To uninstall the package, execute the following command:
# rpm --e nvme-kmp-default-1.17_3.0.76_0.11-0.x86.64.rpm


Verifying the Installation of the software package in the Linux System
1. Turn the system power on, and then start the terminal window.
2. Run “Modinfo mtip32xx.” If the NVMe driver has been installed 
successfully, the version and module information are shown.
3. Run “fdisk –l | grep rssd*”; the NVMe SSD is shown as 
“/dev/rssd<x>drive”.
Starting Re-Drive and Checking the NVMe SSD State
1. Click the “Identify” tab and Name Space Button.
2. The NVMe SSD information is displayed.

Using the NVMe SSD as a boot Drive (etc etc etc)

You might have to go digging around the kernel modules to find drivers, 
specs and such...


good_hunting::hth,
James

Re: [gentoo-user] Can't create valid btrfs on NVMe

[J. Roeleveld]

On Wednesday, August 17, 2016 11:26:13 PM Neil Bothwick wrote:
> I'm putting together a new desktop using a Samsung SM951 NVMe drive. I
> booted sysrescd, partitioned the drive and ran
> 
> mkfs.btrfs /dev/nvme0n1p3
> mount /dev/nvme0n1p3 /mnt/custom
> 
> df -T, mount and findmnt all show this is mounted as a btrfs filesystem,
> e.g.
> 
> /dev/nvme0n1p3 on /mnt/custom type btrfs (rw,relatime,ssd,space_cache)
> 
> I can create files in here but cannot do anything btrfs-y
> 
> % btrfs filesystem show /mnt/custom
> ERROR: not a valid btrfs filesystem: /mnt/custom
> 
> % btrfs subvolume create /mnt/custom/test
> Create subvolume '/mnt/custom/test'
> ERROR: cannot create subvolume: Inappropriate ioctl for device
> 
> Google has been no help at all.

Neil,

I've got a similar drive in my desktop and it actually works.
Difference: I am using Ext4.

Can you try Ext4 and see if it works? 
Next test: Does it work with a non-NVMe drive?

Also, which kernel version?

--
Joost

Re: [gentoo-user] Can't create valid btrfs on NVMe

[Rich Freeman]

On Thu, Aug 18, 2016 at 8:57 AM, J. Roeleveld <joost@antarean.org> wrote:
> On Wednesday, August 17, 2016 11:26:13 PM Neil Bothwick wrote:
>> I'm putting together a new desktop using a Samsung SM951 NVMe drive. I
>> booted sysrescd, partitioned the drive and ran
>>
>> mkfs.btrfs /dev/nvme0n1p3
>> mount /dev/nvme0n1p3 /mnt/custom
>>
>> df -T, mount and findmnt all show this is mounted as a btrfs filesystem,
>> e.g.
>>
>> /dev/nvme0n1p3 on /mnt/custom type btrfs (rw,relatime,ssd,space_cache)
>>
>> I can create files in here but cannot do anything btrfs-y
>>
>> % btrfs filesystem show /mnt/custom
>> ERROR: not a valid btrfs filesystem: /mnt/custom
>>
>> % btrfs subvolume create /mnt/custom/test
>> Create subvolume '/mnt/custom/test'
>> ERROR: cannot create subvolume: Inappropriate ioctl for device
>>
>> Google has been no help at all.
>
> I've got a similar drive in my desktop and it actually works.
> Difference: I am using Ext4.
>
> Can you try Ext4 and see if it works?
> Next test: Does it work with a non-NVMe drive?
>
> Also, which kernel version?

I think you guys are going down the wrong road.  The kernel drivers
are almost certainly working for the drive, otherwise the filesystem
wouldn't work at all, and I'm sure a million ext4 users would have
noticed a problem by now.

This is almost certainly a bug in btrfs-progs, or maybe the btrfs
filesystem driver in the kernel.

I'd suggest raising this on the btrfs mailing list, where it is going
to get a lot more attention from the people who develop btrfs.  There
are a few of us who use it around here, but I'd have to spend a day
tweaking the btrfs-progs source to have a guess at where this is
bailing out.  I suspect somebody over there would have an answer
almost immediately.

-- 
Rich

Re: [gentoo-user] Can't create valid btrfs on NVMe

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 810 bytes --]

On Thu, 18 Aug 2016 14:57:04 +0200, J. Roeleveld wrote:

> I've got a similar drive in my desktop and it actually works.
> Difference: I am using Ext4.

Yes, it works on ext4 here.
 
> Can you try Ext4 and see if it works? 
> Next test: Does it work with a non-NVMe drive?

That's my next test, but I've used sysrescd to set up btrfs before, so I
doubt that's the cause.
 
> Also, which kernel version?

4.1.30 - then I realised that sysrescd defaults to an older kernel (my
rescue version in /boot always boots to the alt kernel). I tried the alt
kernel, which is 4.4.17, and it worked!

Many thanks, Joost. It always helps to have someone point out the one
obvious point I've overlooked!


-- 
Neil Bothwick

I am Ken Dodd of the Borg... What a fine day to be assimilated missus!!

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: [gentoo-user] Can't create valid btrfs on NVMe

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 839 bytes --]

On Thu, 18 Aug 2016 09:38:03 -0400, Rich Freeman wrote:

> This is almost certainly a bug in btrfs-progs, or maybe the btrfs
> filesystem driver in the kernel.

The latter, a later kernel appears to have done the trick.
 
> I'd suggest raising this on the btrfs mailing list, where it is going
> to get a lot more attention from the people who develop btrfs.  There
> are a few of us who use it around here, but I'd have to spend a day
> tweaking the btrfs-progs source to have a guess at where this is
> bailing out.  I suspect somebody over there would have an answer
> almost immediately.

As our resident btrfs expert, I was expecting you to come up with an
immediate answer ;-)


-- 
Neil Bothwick

The nice thing about Windows is - It does not just crash,
it displays a dialog box and lets you press OK first.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: [gentoo-user] Can't create valid btrfs on NVMe

[Rich Freeman]

On Thu, Aug 18, 2016 at 9:45 AM, Neil Bothwick <neil@digimed.co.uk> wrote:
>
> 4.1.30 - then I realised that sysrescd defaults to an older kernel (my
> rescue version in /boot always boots to the alt kernel). I tried the alt
> kernel, which is 4.4.17, and it worked!
>

Hmm, that longterm is starting to look somewhat mature.  I might
consider switching over myself.  Usually I try to give them a good six
months before I consider them ready for btrfs (they have a tendency to
introduce regressions in new kernel versions; I gave up on tracking
non-longterm ages ago).

-- 
Rich

[gentoo-user] Re: Can't create valid btrfs on NVMe

[Kai Krakow]

[-- Attachment #1: Type: text/plain, Size: 983 bytes --]

Am Thu, 18 Aug 2016 14:47:07 +0100
schrieb Neil Bothwick <neil@digimed.co.uk>:

> On Thu, 18 Aug 2016 09:38:03 -0400, Rich Freeman wrote:
> 
> > This is almost certainly a bug in btrfs-progs, or maybe the btrfs
> > filesystem driver in the kernel.  
> 
> The latter, a later kernel appears to have done the trick.
>  
> > I'd suggest raising this on the btrfs mailing list, where it is
> > going to get a lot more attention from the people who develop
> > btrfs.  There are a few of us who use it around here, but I'd have
> > to spend a day tweaking the btrfs-progs source to have a guess at
> > where this is bailing out.  I suspect somebody over there would
> > have an answer almost immediately.  
> 
> As our resident btrfs expert, I was expecting you to come up with an
> immediate answer ;-)

Have you tried an explicit "btrfs dev scan"? If that helps, problems
maybe arise from the udev rules...

-- 
Regards,
Kai

Replies to list-only preferred.

[-- Attachment #2: Digitale Signatur von OpenPGP --]
[-- Type: application/pgp-signature, Size: 181 bytes --]

Re: [gentoo-user] Re: Can't create valid btrfs on NVMe

[Neil Bothwick]

[-- Attachment #1: Type: text/plain, Size: 1272 bytes --]

On Fri, 2 Sep 2016 03:07:31 +0200, Kai Krakow wrote:

> Am Thu, 18 Aug 2016 14:47:07 +0100
> schrieb Neil Bothwick <neil@digimed.co.uk>:
> 
> > On Thu, 18 Aug 2016 09:38:03 -0400, Rich Freeman wrote:
> >   
> > > This is almost certainly a bug in btrfs-progs, or maybe the btrfs
> > > filesystem driver in the kernel.    
> > 
> > The latter, a later kernel appears to have done the trick.
> >    
> > > I'd suggest raising this on the btrfs mailing list, where it is
> > > going to get a lot more attention from the people who develop
> > > btrfs.  There are a few of us who use it around here, but I'd have
> > > to spend a day tweaking the btrfs-progs source to have a guess at
> > > where this is bailing out.  I suspect somebody over there would
> > > have an answer almost immediately.    
> > 
> > As our resident btrfs expert, I was expecting you to come up with an
> > immediate answer ;-)  
> 
> Have you tried an explicit "btrfs dev scan"? If that helps, problems
> maybe arise from the udev rules...

I tried that. It turned out the problem was that my kernel was too old.
Switching to the alt kernel on the CD fixed it.


-- 
Neil Bothwick

To be sure of hitting the target, shoot first and call whatever you hit
the target.

[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 163 bytes --]