Package: src:linux
Version: 6.17.11-1
Severity: critical
Tags: d-i ipv6 upstream
Justification: breaks the whole system
X-Debbugs-Cc: debian-amd64@lists.debian.org, radostin.m.dimitrov@gmail.com, debian-boot@lists.debian.org
User: debian-amd64@lists.debian.org
Usertags: amd64
User: debian-boot@lists.debian.org
Usertags: amd64

Hi, Bastian !

Thanks for your swift reply, appreciated.
Well, I have been using this Intel-card for several years now (I think at
least 2 years),
going through Deb 12 / 13 / even Sid and never got any issues. I think it
broke yesterday,
when I updated my Forky and kernel 6.17.11 got in. At first, I wondered if
it was my router or switch that needed a reboot :)
To confirm, I don't find this being a hardware fault, since the card
continues working under Win11 (from which I am now writing to you).

Debian's boot goes normal, the ixgbe-driver load succeeds and the desktop
gets fully loaded (Plasma 6).
I was at something, when I suddenly got network issues - browser stopped
loading, or was it Steam ?
Anyway, it was exactly then, when I fired up dmesg and saw these lines :

"Adapter removed
Warning firmware error detected FWSM: some hex"

I need to reboot my machine, boot Deb with this NIC again, so that I
provide you with the exact dmesg-messages.
Please, tell me what log(s) you need to collect, since I generally do not
configure logging and rarely look at logs, thanks !

Cheers,
Radi

Please provide some information.  What is broken.  How is it broken.
Since when is it broken.  And the complete kernel log.

ipv6.disable=1 is unsupported.

You have proprietary modules loaded.  Does it also fail without?

Okay, the error counters are pretty high.

Bastian

Control: severity -1 important
Control: tags -1 + moreinfo
Control: tags -1 - d-i ipv6

There are absolutely information in this bugreport what is believed to
be broken: What does not function? Is it a regression from the
previous running kernel? Can you provide boot logs? (Do boot without
proprietary nvidia drivers so we might properly forward it to
upstream, as the kernel otherwise is tainted).

In case it is a regression from the previously running kernel, can you
do please a bisect? Do you need instructions on how to do it?

Regards,
Salvatore

Hey, Guys !

Using a single-chain to write to you both with Salvatore (and the entire
team), so that you don't get cluttered with mails
(and you have tons of those, I am sure). As instructed, I booted with the
IGD, waited for the desktop to load and gathered
both boot's log and dmesg. Please, find those attached - thank you very
much for your help !

Cheers,
Radi


On Thu, Dec 18, 2025 at 6:57 PM Radostin Dimitrov < radostin.m.dimitrov@gmail.com> wrote:

I still see this,

And nvidia loaded.

Anyway, please update the firmware.
https://de.msi.com/Motherboard/MAG-X870E-TOMAHAWK-WIFI/support shows
2.A91, while your system reports 1.A61.

You've run into broken firmware, I would say.  The whole network card
does not get any memory for the queues assigned:

Bastian

Hello, Bastian, Guys,

I followed your instructions to the letter and did the test without the
nvidia driver (results attached, as requested).
Same unpredictable behaviour - sometimes the NIC would work for like 5
mins, other times for 20 mins (before detaching itself with "Adapter
firmware error").
It works on Windows, hence I isolated this from being a hardware fault. As
mentioned, it's been working for several years without issues.
On the UEFI topic - I am monitoring MSI's mobo page for upcoming updates.
But you're wrong about the model - I am on "Edge Ti WiFi", not "Tomahawk",
if that would make a difference, of course. Let me know if you need further
details and if I need to do other tests.
Thanks & Cheers !

Hello Radostin,

This mail doesn't has an attachment and the previously attached kernel
log had the nvidia module loaded. So there is a misunderstanding
somewhere (or you just forgot to attach a recent kernel log?).

Can you install older kernels and check if this gives you a working
setup again? Please tell if you need instructions to do so. It would be
interesting to know the newest kernel that still worked and/or the
oldest broken kernel.

Best regards
Uwe

Hey, Guys,

Starting a "fresh" / cleaner mail loop, since it gets far longer and harder
for everyone to track, hope you don't mind. Dunno what happened - but I did
send requested logs, on the very same day they were requested. Attaching
these once again.

Thanks & Cheers !

Control: tag -1 + moreinfo

The dmesg still has

	[    7.412563] nvidia: loading out-of-tree module taints kernel.

. Can you please blacklist this driver for further tests?

Looking at the code, there is:

        /* read fwsm.ext_err_ind register and log errors */
        fwsm = IXGBE_READ_REG(hw, IXGBE_FWSM(hw));

        /* skip if E610's FW is reloading, warning in that case may be misleading */
        if (fwsm & IXGBE_FWSM_EXT_ERR_IND_MASK ||
            (!(fwsm & IXGBE_FWSM_FW_VAL_BIT) && !(hw->mac.type == ixgbe_mac_e610)))
                e_dev_warn("Warning firmware error detected FWSM: 0x%08X\n",
                           fwsm);

I think the diagnosis is wrong, the value read is 0xffffffff which in
other code locations is interpreted as IXGBE_FAILED_READ_REG.

Can you please:

 a) check the issue also occurs without the nvidia (and other tainting
    modules) loaded. Try adding `module_blacklist=nvidia` to the kernel
    commandline. (If the problem doesn't reproduce without the nvidia
    module, stop here. Then it seems the nvidia code is the culprit and
    we cannot do anything for you.)

 b) You wrote Debian 13 wasn't affected. Can you please install kernels
    from https://snapshot.debian.org/ between 6.12 and 6.17 and check
    which is the newest kernel still working and the oldest being
    broken? Do the working kernels also report things like:

	[    0.265544] pci 0000:01:00.0: VF BAR 2 [mem size 0x10000000 64bit pref]: can't assign; no space

    ?

Best regards
Uwe