On Sat, 30 Dec 2023, Kevin Oberman wrote:
On Fri, Dec 22, 2023 at 9:33 PM Kevin Oberman <rkober...@gmail.com> wrote:
After the ssh vulnerability was announced, I updated my kernel. After
completing the upgrade, it crashed when the system was booted. All stack
dumps showed lots of drm points on hte stack. So, I tried removing the
i915kms from the kld_list in rc.conf. The system booted and started the
wifi without crashing. When I loaded i915kms manually, the system froze.
I still am wondering if there is some collision in the lkpi when the drm
driver on Alder Lake is present and the wifi starts. While I failed to get
much saved during the multiple reboots, all showed the crash after
wpa_supplcant started.
Is this remotely possible or just a coincidence?
I have rolled my system back to 06c083f79dd7 (Dec. 2) just to get it
running with both graphics and wifi.
I have now booted several times with no graphics kmod loaded (i915kms).
There have been no failures or panics while booting. One system freeze when
loading i915kms. No state messages at all.
iwlwifi0: WRT: Invalid buffer destination
iwlwifi0: WFPM_UMAC_PD_NOTIFICATION: 0x20
iwlwifi0: WFPM_LMAC2_PD_NOTIFICATION: 0x1f
iwlwifi0: WFPM_AUTH_KEY_0: 0x90
iwlwifi0: CNVI_SCU_SEQ_DATA_DW9: 0x0
iwlwifi0: RFIm is deactivated, reason = 4
wlan0: Ethernet address: 70:a8:d3:df:3c:de
Created wlan(4) interfaces: wlan0.
lo0: link state changed to UP
Starting wpa_supplicant.
Starting dhclient.
wlan0: no link ....
wlan0: link state changed to UP
got link
DHCPREQUEST on wlan0 to 255.255.255.255 port 67
DHCPACK from 192.168.1.1
unknown dhcp option value 0x7d
bound to 192.168.1.5 -- renewal in 7200 seconds.
Starting Network: lo0 em0 wlan0.
I should also mention that, when I manually load i915kms, I do see many
WITNESS reports. Do we still keep a list of known ones?
have you at any time tried the other way round:
load drm-kmod on boot and and NO wireless (possibly blocklist it) and
then try to manually load wireless?
While an interaction in LinuxKPI would mean some resources are not
properly save per device most likely (or we get the wrong one), loading
drm-kmod as well probably also changes timing which means you could hit
the net80211/LinuxKPI co-bugs which are triggered by two specifics in
net80211 and then cause assertions in LinuxKPI code (it seems I have
much improvement in a development branch but am still on it after the
New Year to clean things up and track secondary effects of the change).
/bz
--
Bjoern A. Zeeb r15:7
