On Apr 18, Jamie Bainbridge <jamie.bainbridge@gmail.com> wrote: I would like to have feedback from the kernel team about this proposed change. > I have a laptop with iwlwifi wireless card: > > $ sudo lspci -nn | grep Net > Network controller [0280]: Intel Corporation Wireless 7265 [8086:095a] (rev 59) > > This hardware requires the firmware-iwlwifi package, and the firmware > works fine. However, on nonfree install or nonfree Live, the wireless > interface repeatedly dies as soon as it's used with "Microcode SW error > detected. Restarting" in dmesg and long firmware dump from the driver. > it's not possible to even "apt update" because the interface dies. So does the firmware work or not? > Ubuntu ships a file in its kmod package with the following contents: > > # /etc/modprobe.d/iwlwifi.conf > # iwlwifi will dyamically load either iwldvm or iwlmvm depending on the > # microcode file installed on the system. When removing iwlwifi, first > # remove the iwl?vm module and then iwlwifi. > remove iwlwifi \ > (/sbin/lsmod | grep -o -e ^iwlmvm -e ^iwldvm -e ^iwlwifi | xargs > /sbin/rmmod) \ > && /sbin/modprobe -r mac80211 But why then also unload mac80211? > Adding this file to Debian resolves the problem. But why? At no point you described something or somebody requesting that iwlwifi is unloaded. > This is a request to include the above iwlwifi.conf file in Debian's > kmod package too. > > I can only assume from the massive Ubuntu install base that this file > doesn't cause any problems for devices which don't need the modules > loaded in this order, while also resolving whatever problem requires > that modules are loaded in the order which the above file forces. > > I could not find any previous bug with this request. I also couldn't > find Ubuntu's history of this package to find why it was included there > in the first place. It's been there for a very long time, at least since > kmod 9 from over 10 years ago. So I am not very comfortable with adding code whose purpose and origin is unclear. > Given that Ubuntu has shipped this file for so long, the risk of any > regression in Debian seems extremely low. Given also that Debian never shipped this, I would like to better understand why we should do it now. -- ciao, Marco
Attachment:
signature.asc
Description: PGP signature