Nvme SSD M2 not seen by OMLx 4.0

alpha
omlx-4
Tags: #<Tag:0x00007f0e47d1e3e8> #<Tag:0x00007f0e47d1e050>

(D27) #1

Hi,

As I said earlier, I have a big problem with this 2412 , but what happens is a bit long to explain, so I will try to make myself as clear as possible …if capable of it…

A fortnight ago I received a brand new equipment (LENOVO AIO with APU - graphic incorporated into processor : amdgpu – and recent technologies).

At first, I tried to install Rosa (as I wanted to keep KDE4 -I don’t like Plasma5-) and noted the SSD M2 (card-disk) was absent from the instalator.

My second HD was correctly detected.

I decided then, to test what would happen with previous OMLx4 -version 2280 - I still had in my documents.

Note well => that the SSD M2 appeared in the KDE partition program.

So, I decided to see further (launching an installation) what Calamares would say.

“NO” SSD M2 appearing in Calamares.

And this was the end of my try, I didn’t install anything.

Later on, I installed “Mint Cinnamon” without the faintest problem => on my recently formatted SSD M2 ! Start now in 6 seconds !

…But Cinnamon, is not made for me, I installed Fedora 29 and again, my SSD M2 was perfectly recognized.

To day, I launch OMLx4 version 2412 as recommended by Rugy and => No SSD M2 appearing => nor in KDE partition program, neither in Calamares.

A look into dmseg says :

[lun. févr. 4 09:45:16 2019] nvme nvme0: controller is down; will reset: CSTS=0xffffffff, PCI_STATUS=0xffff

[lun. févr. 4 09:45:16 2019] print_req_error: I/O error, dev nvme0n1, sector 250069504

[lun. févr. 4 09:45:16 2019] nvme 0000:03:00.0: Refused to change power state, currently in D3

[lun. févr. 4 09:45:16 2019] nvme nvme0: Removing after probe failure status: -19

[lun. févr. 4 09:45:16 2019] Buffer I/O error on dev nvme0n1, logical block 31258688, async page read

[lun. févr. 4 09:45:16 2019] Buffer I/O error on dev nvme0n1p1, logical block 10239984, async page read

[lun. févr. 4 09:45:16 2019] nvme nvme0: failed to set APST feature (-19)

I cannot guarantee my disk is perfect, I can guarantee nor Mint neither fedora dmesg have similar lines.

So I, of course, didn’t install the system; it is useless to have half of the equipment working and especially, If you cannot install things where you want.

The kernel of this 2412 is the same (4.20.4…) I have on the Fedora, why do they actuate differently ?

Why KDE partition was seeing the SSD M2 in 2280 version, and not in this 2412 ?

Why Calamares remains blind to this SSD in both versions ?

The rest of dmesg message has (afaik) no problem; everything -including IOMMU or recent technologies- is normally detected.

It has come to my mind that, maybe, the detect-system is looking for a uefi or boot partition, or again a GPT partition, while I have formatted this SSD with “bios boot partion” and start directly from MBR… I don’t know if this will help…

Sorry to bother but I think it is important for the devs, to have knowledge of this kind of annoyance with some equipments.

Regards


OMLx 4.0 Post Alpha1 improvements
Issue/bug tracking system
(Abucodonosor) #2

Hello @D27

This is an kernel bug. Fedora probably patched a quirck for your SSD model.
I would need to know your exact SSD model , would need to see a full dmesg
and a calamares log.

That can mean APST is not supported at all on your SSD or it is supported but has
some unusual firmware expecting different latency settings.

You can try to boot with it disabled by adding this to kernel cmd line:

nvme_core.default_ps_max_latency_us=0

or you can tr playing with that value need be somehting bigger than 200.

However if if your SSD vendor is WD you may need a value of 500 to 5000


(D27) #3

Hi Abucodonosor,

Despite all my searches I will not be very helpful, I’m afraid.

This SSD M2 only appears as : HFM128GDHTN-8310B

no maker, no nothing else nowhere.

Possibly Hynix (??)

Here is the dmesg obtained when booting with the USB pendrive.
dmesg-OMV.txt (93,5 Ko)

Concerning Calamares I could not locate any log file. Remember I stopped the install process when I didn’t see the SSD in the list of HD, at the partitioning page.

This SSD M2 being a PCIe connected drive (and not a SATA one -I think-), maybe Calamares :

  1. Didn’t search the slots PCI…

  2. Or merely, didn’t find the path in some archive (I ignore completely how the process works), since the device was not recognized at all.

…But this last hypothesis (archive) doesn’t explain why calamares did not list the SSD with the previous version 2280 …when KDE partition program could see it.

I did not make myself very clear : I am not trying to boot in any manner.

The USB pendrive boots perfectly !

I just don’t want to go further until my SSD is (first) detected by the system, and (second) by Calamares too.

If APST is or is not supported by this SSD, I think mention of this would appear in MINT or FEDORA dmesg, and that’s not the case.

Bye


(Abucodonosor) #4

Probably ADATA…

Well as long you SSD is not powered ON , see log it stuckt in D3 nothing can see it.
And even when some build saw it , like you see once you touched it with partition manager is gone.

From the dmesg you now posted you whole PCI bus times out.
Can you please attach lspci output too ?

I got here SSD’s connected to PCIE , M2 slot , M2 ones into PCIE card … and nothing
has any problem with these. However all my SSD’s/NVme’s are EVOs/EVOs PRO.

@bero got such strange issue on one of his boxes too , but these are kernel bugs.

There aare brunch quirck for particular disk in the kernel , just to fix such issues.

As example on this laptop:

[    2.853586] ata1.00: supports DRM functions and may not be fully accessible
[    2.854762] ata1.00: ATA-11: Samsung SSD 860 EVO 250GB, RVT01B6Q, max UDMA/133

Same way these quircks will disable or workaround other things are wrong implemented,
make for windows or simple bugs in the firmware.

That means just : these kernel have different versions , are patched , have different configuration.
But it still remains an kernel bug :slight_smile:

First need be detected or detected and fully functional by the kernel …

Also can you boot with nvme_core.default_ps_max_latency_us=0 please ?
( from the USB stick )


(rugyada) #5

Yes, card games have been included in ISO live mode only.

From D27’s suggestion :

:grin:


(Abucodonosor) #6

@D27

can you also post me a ‘sudo nvme list’ , output , please ?


(Abucodonosor) #7

Just to see you are not alone just see ubuntu generic bug about exactly that:

https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1678184


(Abucodonosor) #8

@D27

last thing I need is output of:

nvme id-ctrl /dev/nvme0


(Ben Bullard) #9

@abucodonosor @D27 Should we have a bug report for this?


(Abucodonosor) #10

@ben79
probably yes


(Ben Bullard) #11

@D27 could you file a bug report for this issue? It is always best if someone that has the issue files the bug report.


(Ben Bullard) #12

@rugyada @abucodonosor

Also Beta ISO has been made so maybe it’s time for a new thread? If so I wonder if issues would not be easier to track if each issue has it’s own thread. In other words a general thread for people to report their OMLx 4.0 Beta testing but any issues needing attention we start a new thread for that specific issue?

To me it got very confusing trying to keep up with specific issues in one long thread. But I’m a Linux Monkey.

:speak_no_evil: :hear_no_evil: :see_no_evil:


(Abucodonosor) #13

@ben79

you can start new beta thread but I need some infos from @D27
to be able to workaround this issue.


(Ben Bullard) #14

I just made this a separate thread as it seems important enough. Hope this makes this issue easier to follow.

Post-edit: Please feel free to edit title as you see fit.


(Abucodonosor) #15

@ben79

yes thx :slight_smile:


(Ben Bullard) #16

OT but need to say: This is more or less in response to discussion on IRC about devs and issues in forum. I’m hoping if we enforce the “One issue, one thread” concept this will make it easier both on devs to follow an issue and for users to find an issue.

Post-edit: And I have been guilty of doing it the other way myself but we/me need to do better.


(rugyada) #17

Right, just I was waiting for Beta release to be announced, or at least Beta candidate :wink:


(rugyada) #18

Yes, definitely.


(D27) #19

Hi Abucodonosor,

You did it!
Here is what you requested :
1 - lspci :
lspci-OMLx4.txt (2,5 Ko)

2 - nvme list
nvme-list-OMLx4.txt (163 Octets)
From french => No command available

3 - nvme id_ctrl /dev/nvme0
nvme-id-ctrl-OMLx4.txt (183 Octets)
From french => No command available

4 - Then I booted with : nvme_core_default_…us=0
and here is the dmesg (will be send later, lost it at the moment)

5 - Then here is KDE partition program with the SSD => OK with option just above inserted in grub2
KDE-partition

6 - And finally Calamares Where the SSD appears now
calamares

Congrat.


(Abucodonosor) #21

@D27

Nice … and not nice since it seems your SSD need an NVME_QUIRK_NO_APST.

It seems nvme-cli gone from our repos … but I badly need that output of
nvme id_ctrl /dev/nvme0 , to be able adding that to the kernel.

So if is possible for you can you run that command from any other Distribution
and post me the output ?

Thx

( PS: we’ll re-add nvmi-cli soon )