Created attachment 239361 [details]
On a specific network, my IWL dies every time within 20 or so seconds...
Everywhere else, it seems to work fine...
The hardware is a "Network controller: Intel Corporation Wireless 8260 (rev 3a)" on a Lenovo X1 Yoga.
[ 70.457881] wlp4s0: associated
[ 88.896444] iwlwifi 0000:04:00.0: Timeout exiting D0i3 before hcmd
[ 88.896458] wlp4s0: HW problem - can not stop rx aggregation for 90:f6:52:ff:b5:55 tid 0
Full dmesg is attached...
Did you enable runtime power-management (CPTCFG_IWLWIFI_PCIE_RTPM) in your kernel? You may have problems with that in certain platform that's why it's disabled by default.
No, but CONFIG_IWLWIFI_PCIE_RTPM is enabled (I have no CPTCFG). I will attach full config...
Created attachment 239401 [details]
Sorry, that's what I meant (the CPTCFG part is from the backports stuff).
You should disable PCIE_RTPM, since this needs special integration with some platforms.
Tobias, I tried to contact you on #archlinux @freenode, but you were not online. So I'm adding you here. ;)
I noticed that you enabled runtime PM in our driver (CONFIG_IWLWIFI_PCIE_RTPM), but unfortunately this feature still requires some integration with specific platforms. Is there any reason why you enabled this feature?
There was no specific reason for enabling. Probably cause fedora uses it too.
I can revert this again.
It's safer to keep it disabled for now. We'll let you know when setting this option actually becomes beneficial for general platforms.
Ok fixed on trunk, next kernel will not have it enabled anymore.
I suggested to the Fedora kernel team to make the same change, they responded with:
"There's no indication in the code that it's broken or unsafe to
enable. Seems like a patch making it depend on config BROKEN or
config EXPERT or at least some kind of warning in the help text is
That seems a valid point to me, can someone from the iwl wifi team please submit a patch upstreaming making it clear in the Kconfig that enabling this is a bad idea atm ?
Yes, this indeed makes sense. Fedora and derivatives are not the only ones I saw enabling RTPM and having trouble. We should fix those problems, but we're not actively working on this at the moment, so marking it as EXPERT would be a good idea. I don't want to mark it as BROKEN, because it does work, but needs some integration with other parts of the kernel.
Added an internal bug for tracking.
(In reply to Luca Coelho from comment #10)
> Yes, this indeed makes sense. Fedora and derivatives are not the only ones
> I saw enabling RTPM and having trouble. We should fix those problems, but
> we're not actively working on this at the moment, so marking it as EXPERT
> would be a good idea. I don't want to mark it as BROKEN, because it does
> work, but needs some integration with other parts of the kernel.
Thank you, please also update the help text to reflect this.
Created attachment 250731 [details]
Update of the Kconfig
The patch above was submitted internally.
Closing the bug. You can re-open it at will if you see the need.