FS#55665 - [linux] RTL8192EE PCIe Wireless Network Adapter can't connect with intel_iommu enabled kernel

Attached to Project: Arch Linux
Opened by Wind.Lee (zwindl) - Sunday, 17 September 2017, 23:48 GMT
Last edited by Doug Newgard (Scimmia) - Sunday, 01 October 2017, 15:10 GMT
Task Type Bug Report
Category Packages: Core
Status Closed
Assigned To Tobias Powalowski (tpowa)
Architecture x86_64
Severity Medium
Priority Normal
Reported Version
Due in Version Undecided
Due Date Undecided
Percent Complete 100%
Votes 1
Private No

Details

Description:
The RTL8192EE wifi card can't work properly with NetworkManager since kernel 4.13, it's gets stuck at 'configuring'. Here's the dmesg log https://ptpb.pw/dlJN and here's the journalctl log https://ptpb.pw/N9eT and I've tried compile myself with two different .config, one is from aur package linux-git(https://ptpb.pw/HF1g) which can run properly, wifi can connect on login, another is from official repo's linux package(https://ptpb.pw/7GuV) which can cause that issue.

Additional info:
* package version(s) linux 4.13.2-1
* config and/or log files etc.
dmesg log: https://ptpb.pw/dlJN
journalctl log: https://ptpb.pw/N9eT
.config from linux-git: https://ptpb.pw/HF1g
.config from archlinux's official linux package: https://ptpb.pw/7GuV

Steps to reproduce:
I reported to the kernel mailing-list fist, and tried to git bisect 4.13/4.12, but it's all pass when I was using linux-git's .config, and I shifted to official .config, it failed.
This task depends upon

Closed by  Doug Newgard (Scimmia)
Sunday, 01 October 2017, 15:10 GMT
Reason for closing:  Fixed
Comment by loqs (loqs) - Monday, 18 September 2017, 00:22 GMT
The bisection with the official config fromm 4.13 that failed when you built the package locally using the 4.12 tag?
Comment by Wind.Lee (zwindl) - Monday, 18 September 2017, 05:35 GMT
@loqs Thanks for the reply. No, it's totally OK for 4.12 tag, actually I'm using 4.12.10 now, it's from official repo. If I update package `linux` to 4.13, the WiFi card won't work properly.
Comment by loqs (loqs) - Monday, 18 September 2017, 11:21 GMT
"I reported to the kernel mailing-list fist, and tried to git bisect 4.13/4.12, but it's all pass when I was using linux-git's .config, and I shifted to official .config, it failed."
How exactily did you shift to the official config?
Going over the kernels you built:
4.12 with official 4.13 config good?
4.13 with official 4.13 config bad?
4.12 with linux-git config good?
4.13 with linux-git config good?
Comment by Wind.Lee (zwindl) - Tuesday, 19 September 2017, 02:49 GMT
@loqs
4.12 with official 4.13 config didn't test.
4.13 with official 4.13 config bad.
4.12 with linux-git config good.
4.13 with linux-git config good.
I'm testing the 4.12 with official 4.13 config. Report later.
Comment by Wind.Lee (zwindl) - Tuesday, 19 September 2017, 04:39 GMT
@loqs Now I'm sure that issue is caused by that .config file, 4.12 with that .config file can't work neither.
Comment by loqs (loqs) - Tuesday, 19 September 2017, 10:50 GMT
https://git.archlinux.org/svntogit/packages.git/commit/trunk?h=packages/linux&id=7648ebf0fa9b0190aa26a11acd46ac31f0593b69
enabled CONFIG_RTLWIFI_DEBUG which is not set in linux-git
If that is not the cause I would suggest diffing the two configs and merging the differences.
Make sure to diff the config after it has been processed by the 4.12 kernel so the third line references 4.12 not 4.13.
Comment by Wind.Lee (zwindl) - Tuesday, 19 September 2017, 12:11 GMT
@loqs No, not caused by the `CONFIG_RTLWIFI_DEBUG`, I'm diffing two config files now. But I'm not very pro on kernel building either configuring, I can't fully understand these flags mean, maybe I'll ask some foolish questions here, hope you wont' mind :(
Comment by Wind.Lee (zwindl) - Wednesday, 20 September 2017, 12:43 GMT
@loqs Finally, I figured out it's the `CONFIG_INTEL_IOMMU_DEFAULT_ON=y` flag caused that issue. Is that a kernel bug?
Comment by Eli Schwartz (eschwartz) - Wednesday, 20 September 2017, 13:00 GMT
So, does it work if you disable IOMMU via the kernel commandline intel_iommu=off using the <s>kernel in [core]</s> 4.13 kernel?

EDIT: Obviously, you were supposed to read my mind and know that I really meant the one in [testing]. :D
Comment by loqs (loqs) - Wednesday, 20 September 2017, 14:37 GMT
@eschwartz as linux in core 4.12.13-1 has '# CONFIG_INTEL_IOMMU_DEFAULT_ON is not set' according to the documentation https://cateee.net/lkddb/web-lkddb/INTEL_IOMMU_DEFAULT_ON.html intel_iommu=off will have no effect.
Comment by loqs (loqs) - Wednesday, 20 September 2017, 15:45 GMT
@zwindl it is certainly a possibility the issue is the result of a kernel bug but other possibilities exist
such as firmware supplying bad information to the kernel or bad interaction with another component that is using IOMMU.
You could investigate if upstream is aware of the issue and if upstream considers it a kernel bug.

Would hetfig consider reverting the change CONFIG_INTEL_IOMMU_DEFAULT_ON=y as it appears to have triggered this issue and  FS#55629 ?
Comment by Wind.Lee (zwindl) - Thursday, 21 September 2017, 12:20 GMT
@eschwartz Yes, it works! If I have to use that kernel, that is the best solution for me, thanks a lot.
@loqs I didn't meet any freeze problem. seems OK for me.
Comment by Wind.Lee (zwindl) - Sunday, 01 October 2017, 02:35 GMT
Seems fixed after yesterday's update, and now it works fine. Thanks devs and maintainers.

Loading...