known iwlwifi bugs? (and side note about an AHCI controller)

[Chris Torek]

I just brought up a new AMD 7950X based system on an ASRock Taichi
motherboard. Side note here: I had to use the SATA_Ax ports, the SATA_[1-4]
ports give channel timeouts, I have yet to look into those (other than
verifying that it's not a matter of waiting longer).

The "Killer Ethernet" isn't up yet so I'm using the iwl wlan interface at
the moment. Things were going pretty good (32 CPUs for building!
buildworld takes under 16 minutes and buildkernel took 15 seconds, using 32
jobs and the -jobs targets) but while building various ports I had a crash.

The relevant part of the stack trace is:

#6 0xffffffff83019ed2 at iwl_mvm_mac_itxq_xmit+0xc2
#7 0xffffffff83028a6f at iwl_mvm_queue_state_change+0x1ef
#8 0xffffffff8306b085 at iwl_txq_reclaim+0x7f5
#9 0xffffffff8304848e at iwl_mvm_rx_tx_cmd+0x14e
#10 0xffffffff83025d4e at iwl_mvm_rx_common+0x1ee
#11 0xffffffff8305a4c3 at iwl_pcie_rx_handle+0x483
#12 0xffffffff83059f20 at iwl_pcie_napi_poll_msix+0x30
#13 0xffffffff80dafcef at lkpi_napi_task+0xf
#14 0xffffffff80ba5922 at taskqueue_run_locked+0x182
#15 0xffffffff80ba6bb2 at taskqueue_thread_loop+0xc2

I didn't get a crash dump (not sure why) and am still running the release
kernel at the moment, but this looks like it's in one of the linuxkpi
adapters, perhaps linuxkpi_ieee80211_tx_dequeue.

(Back to the side note: the AHCI interface that works, ahci1, differs from
the one that doesn't, ahci, like this:

***@pci0:76:0:0: class=0x010601 rev=0x02 hdr=0x00 vendor=0x1b21
device=0x
0612 subvendor=0x1849 subdevice=0x0613
***@pci0:88:0:0: class=0x010601 rev=0x01 hdr=0x00 vendor=0x1022
device=0x
43f6 subvendor=0x1b21 subdevice=0x1062

if that triggers anything obvious in someone's memory.)

[Bjoern A. Zeeb]

On Sun, 14 Jan 2024, Chris Torek wrote:

Hi Chris,

Post by Chris Torek
I just brought up a new AMD 7950X based system on an ASRock Taichi
motherboard. Side note here: I had to use the SATA_Ax ports, the SATA_[1-4]
ports give channel timeouts, I have yet to look into those (other than
verifying that it's not a matter of waiting longer).
The "Killer Ethernet" isn't up yet so I'm using the iwl wlan interface at
the moment. Things were going pretty good (32 CPUs for building!
buildworld takes under 16 minutes and buildkernel took 15 seconds, using 32
jobs and the -jobs targets) but while building various ports I had a crash.
#6 0xffffffff83019ed2 at iwl_mvm_mac_itxq_xmit+0xc2
#7 0xffffffff83028a6f at iwl_mvm_queue_state_change+0x1ef
#8 0xffffffff8306b085 at iwl_txq_reclaim+0x7f5
#9 0xffffffff8304848e at iwl_mvm_rx_tx_cmd+0x14e
#10 0xffffffff83025d4e at iwl_mvm_rx_common+0x1ee
#11 0xffffffff8305a4c3 at iwl_pcie_rx_handle+0x483
#12 0xffffffff83059f20 at iwl_pcie_napi_poll_msix+0x30
#13 0xffffffff80dafcef at lkpi_napi_task+0xf
#14 0xffffffff80ba5922 at taskqueue_run_locked+0x182
#15 0xffffffff80ba6bb2 at taskqueue_thread_loop+0xc2
I didn't get a crash dump (not sure why) and am still running the release
kernel at the moment, but this looks like it's in one of the linuxkpi
adapters, perhaps linuxkpi_ieee80211_tx_dequeue.

Likely because you hit a firmware crash or some other problem.
Hard to say without whatever was before that.

Give you are mentioning linuxkpi_ieee80211_tx_dequeue, could have been this one?
https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=275710

When you say you are still running release, that is 14.0?

The freebsd-wireless list is generally a better place to to let us know
about wireless problems.

/bz