Re: nvme controller reset failures on recent -CURRENT

From: Warner Losh <imp_at_bsdimp.com>
Date: Tue, 27 Oct 2026 22:11:47 UTC
On Mon, Oct 26, 2026 at 10:48 PM Don Lewis <truckman@freebsd.org> wrote:

> On 25 Oct, Warner Losh wrote:
> > On Sun, Oct 25, 2026 at 10:38 PM Don Lewis <truckman@freebsd.org> wrote:
> >
> >> On 25 Oct, Abdelkader Boudih wrote:
> >> > You are running 002C firmware. Please update to latest .
> >> >
> >> > You should have nda0: <INTEL SSDPEKNW512G8 **005C** BTNH940617WE512A>
> >> not nda0: <INTEL SSDPEKNW512G8 *002C* BTNH940617WE512A>.
> >> >
> >> > This looks like a nvme firmware problem.
> >> >
> >> > Abdelkader
> >>
> >> Weird ... my browser time out if it tries to use an IPv6 connection to
> >> https://www.solidigm.com/
> >>
> >> I booted a Linux Live .iso.  Then I was able to download the Linux CLI
> >> version of the Solidigm Storage Tool and upgrade the device firmware.
> >>
> >>
> https://www.solidigm.com/us/en/support-page/drivers-downloads/ka-00085.html
> >>
> >> nda0: <INTEL SSDPEKNW512G8 005C BTNH940617WE512A>
> >>
> >
> > Next time, you may be able to just run that tool from FreeBSD. We have
> good
> > linux compat ioctl which allows passthru commands to work which should
> allow
> > the program to run (assuming you couldn't extract the image and just use
> > nvmecontrol fwupgrade).
>
> The tool is a dynamicly linked executable that relies on some system
> libraries.  Maybe a Linux jail would work, but then how to access
> /dev/nvme?
>
> There are a bunch of firmware files, but it's not obvious what is what:
> FirmwareModules 129%ls
> firmware_module_adp.so  firmware_module_hh.so   firmware_module_nh.so
> firmware_module_cd.so   firmware_module_hhdc.so firmware_module_oem.so
> firmware_module_cpc.so  firmware_module_kh.so   firmware_module_yv.so
> firmware_module_eh.so   firmware_module_lh.so
>
> and:
> %file firmware_module_adp.so
> firmware_module_adp.so: ELF 64-bit LSB shared object, x86-64, version 1
> (SYSV), dynamically linked, stripped
>

In theory, one could put that all into a OCI container, but I've not done
that with binaries that need to access /dev nodes yet. Maybe I will.

Please let me know if the new firmware still causes problems.

Warner