I just got a second hand laptop. It came with Windows 10 installed, and 1 hard disk and 2 SSD (NVMe) were detected. These two SSD are: - THNSN51T02DUK NVMe TOSHIBA Bus nbr 1, Target Id 0, LUN 0 - THNSN51T02DUK NVMe TOSHIBA Bus nbr 3, Target Id 0, LUN 0 Then I replaced Windows by FreeBSD -CURRENT, and these NVMe cannot be used: # grep -i nvme /var/run/dmesg.boot nvme0: <Generic NVMe Device> mem 0xdd600000-0xdd603fff irq 16 at device 0.0 on pci3 nvme1: <Generic NVMe Device> mem 0xdd400000-0xdd403fff irq 16 at device 0.0 on pci6 # cat /boot/loader.conf kern.geom.label.disk_ident.enable="0" kern.geom.label.gptid.enable="0" cryptodev_load="YES" zfs_load="YES" nvme_load="YES" nvd_load="YES" # nvmecontrol devlist nvme0: IDENTIFY (06) sqid:0 cid:0 nsid:0 cdw10:00000001 cdw11:00000000 nvme0: ABORTED - BY REQUEST (00/07) crd:0 m:0 dnr:0 sqid:0 cid:0 cdw0:0 nvme1: IDENTIFY (06) sqid:0 cid:0 nsid:0 cdw10:00000001 cdw11:00000000 nvme1: ABORTED - BY REQUEST (00/07) crd:0 m:0 dnr:0 sqid:0 cid:0 cdw0:0 # ls -l /dev/nvme* crw------- 1 root wheel 0x36 Mar 22 18:49 /dev/nvme0 crw------- 1 root wheel 0x37 Mar 22 18:49 /dev/nvme1 # ls -l /dev/nda* ls: /dev/nda*: No such file or directory Full dmesg and devinfo outputs are on https://wiki.freebsd.org/Laptops/Dell_Alienware_17R4. Any tip to make them usable?
I have tried many things: BIOS tweaking, with or without hw.nvme.use_nvd, with vmd_load, etc., but no satisfying result ATM. Note: initially, a device /dev/ntfs was shown and locked one NVMe. It was caused by KDE, and freed when I booted the machine without starting KDE. This is very strange: at some time /dev/nvd0 and /dev/nvd1 were there, nvmecontrol identify was OK for the two devices, gpart destroy of the Windows stuff, gpart create and gpart add -t freebsd-zfs were successfull, and I created two zpools on them. But they are not persistent! After a reboot, I see sometimes only one of them, and sometimes none… Remark 1: when I can see a zpool, this is always the same. And yes, when the machine ran Windows, both SSDs were always active. Remark 2: it seems that chances to get at least one NVMe are greater after a cold boot, and chances to see none of them are greater after a reboot.
Created attachment 241246 [details] Verbose dmesg after a successful boot After this boot the two NVMe were usable.
Created attachment 241247 [details] Verbose dmesg after an unsuccessful boot After this boot, none of the NVMe were shown. Same configuration than for the previous file.
(In reply to Thierry Thomas from comment #1) For completeness, here is the /boot/loader.conf giving the better results (as described in comment #1): kern.geom.label.disk_ident.enable="0" kern.geom.label.gptid.enable="0" cryptodev_load="YES" zfs_load="YES" nvme_load="YES" nvd_load="YES" #vmd_load="YES" hw.vga.textmode=1 nvidia-modeset="YES" # No success with nda (4) #hw.nvme.use_nvd=0 # See PR 264172 #hw.pci.enable_pcie_hp=1
(In reply to Thierry Thomas from comment #4) I'm not sure, but since you're not using `nda(4)` can you comment out `nvme_load="YES"` in your /boot/loader.conf ? I've Intel Optane M.2 SSD using nvd(4) driver and it works great (on 13.1).
(In reply to Zhenlei Huang from comment #5) You are right: this is not necessary, because nvme module is already in kernel. Then with or without it does not change the result.
Marcus Oliveira reported a similar problem in https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=262969#c15 But the question remains: with the same PSU, why are these NVMe usable on Windows and not on FreeBSD?
I might add that the problem also raises on Linux... I've tried all distributions and hypervisors you could think of. It only works on Windows. Marcus Oliveira
Ya'll need to be adding Warner on NVMe bugs. There's not much we can do I think if IDENTIFY fails. That status though is really odd. Normally you should only get that status if we send an explicit ABORT command to cancel the earlier IDENTIFY. I wonder if we had sent the IDENTIFY and then it timed out so we sent an ABORT?
It might be useful to include the output of `nvmecontrol identify nvme0` and `nvmecontrol identify -xv nvme0`
We get that too when the command times out due to interrup idrntify is a command that aleays works and os fast.t issues.
Created attachment 264085 [details] Output of nvmecontrol identify -xv nvme0 Adding the output of `nvmecontrol identify -xv nvme0' (imp@ is already Cc: )
Created attachment 264086 [details] Adding plain `nvmecontrol identify nvme0' Adding plain `nvmecontrol identify nvme0'.