Possible regression with the FW643 chipset

[Jonathan W. <jw...@ju...>]

Hi all

The FFADO project has received several reports from people running recent
kernels which suggest the kernel has stopped detecting bus resets when a
PCIe FW643 host controller is in use, and (perhaps as a result) is failing
to enumerate devices on the bus.  This was first reported by a user running
Fedora 42.  When running Fedora 41 their system would detect firewire
devices without any issues, and /dev/fw1 was created.  When they upgraded to
Fedora 42, however, /dev/fw1 wasn't being created and the system logs didn't
contain any messages from the firefire subsystem to indicate that any
devices had been seen.

Another user running Endeavour OS also reported the same trouble.

The failure to detect devices applied when devices were connected at boot
and if they were plugged in sometime after booting.

Follow up discussions suggest that kernel 6.6 is okay while 6.12 and all
versions which follow it are not.  The problem only seems to affect users
with FW643 host controllers.  The precise kernel version that first
displayed the symptoms is not yet identified.

For those affected by this issue and running kernel 6.14.5, the following
command sequence usually results in a correctly functioning firewire system:

    sudo modprobe -r firewire-ohci
    sudo modprobe -r firewire-core
    sleep 2
    sudo modprobe firewire-ohci

It has been reported that this does not work in 6.14.3 or 5.14.4.

Since the reporters are not developers, the kernels they have access to are
limited to those supported out of the box by their distributions.  A git
bisect will only be possible if someone with the requisite knowledge can
also reproduce the problem.  The information to hand at present is therefore
somewhat incomplete.  It may take some time to obtain a clear picture of
what might be going on.

Unfortunately I don't personally have a FW643 host controller so I am not
able to investigate this myself.  For what it's worth, the 6.12.26 kernel is
fine for me when using a TI PCI host controller.

At this stage I was wondering whether there have been any changes between
the 6.6 and 6.12 kernels within the firewire subsystem (or maybe the PCIe
subsystem) which could have inadvertently caused this issue with FW643 host
controllers.

Alternatively, are there any FW643 users who have no problem with device
detection under a 6.12.x (or later) kernel?  If so this would conclusively
rule out a kernel issue.

Regards
  jonathan