Note: This is a public test instance of Red Hat Bugzilla. The data contained within is a snapshot of the live data so any changes you make will not be reflected in the production Bugzilla. Email is disabled so feel free to test any aspect of the site that you want. File any problems you find or give feedback at bugzilla.redhat.com.
Bug 1186363 - kernel 3.18.3 breaks wifi
Summary: kernel 3.18.3 breaks wifi
Keywords:
Status: CLOSED UPSTREAM
Alias: None
Product: Fedora
Classification: Fedora
Component: kernel
Version: 21
Hardware: x86_64
OS: Linux
unspecified
urgent
Target Milestone: ---
Assignee: fedora-kernel-wireless-iwl
QA Contact: Fedora Extras Quality Assurance
URL:
Whiteboard:
Depends On: 1190933
Blocks:
TreeView+ depends on / blocked
 
Reported: 2015-01-27 14:36 UTC by Conrad S
Modified: 2015-02-25 01:40 UTC (History)
7 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2015-02-25 01:40:07 UTC
Type: Bug
Embargoed:

Attachments (Terms of Use)

Description Conrad S 2015-01-27 14:36:32 UTC 
Description of problem:
- Upgrading to kernel 3.18.3-201.fc21 breaks wifi connections 
- Previous good kernel is kernel-3.17.8-300.fc21

Version-Release number of selected component (if applicable):
3.18.3-201.fc21

How reproducible:
- enable wifi
- wifi appears as only "half connected": in the top-right panel in gnome shell, there is a question mark over the wifi symbol
- choosing "select network" from wifi menu shows that the correct wifi network has a tick next to it, but no internet connection is available

Steps to Reproduce:
1. assuming previous good kernel is running (kernel-3.17.8-300)
2. yum check-update
3. yum update kernel
4. reboot
5. running kernel is kernel-3.18.3-201
6. enable wifi from top-right menu in gnome-shell

Actual results:
wifi no longer works after booting with updated kernel

Expected results:
wifi connects and works

Additional info:
relevant lspci (from working kernel 3.17.8-300)
00:16.0 Communication controller: Intel Corporation 7 Series/C210 Series Chipset Family MEI Controller #1 (rev 04)
01:00.0 Network controller: Intel Corporation Centrino Advanced-N 6235 (rev 24)
02:00.0 Ethernet controller: Realtek Semiconductor Co., Ltd. RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller (rev 06)

$ lsmod | grep wifi
iwlwifi               129618  1 iwldvm
cfg80211              505401  3 iwlwifi,mac80211,iwldvm

(lsmod is from working kernel 3.17.8-300)
Comment 1 Conrad S 2015-01-27 15:11:55 UTC 
Related bug:
https://bugzilla.kernel.org/show_bug.cgi?id=92151
Comment 2 Stanislaw Gruszka 2015-02-20 10:23:38 UTC 
Please attach dmesg output from working and non-working kernel.
Comment 3 Conrad S 2015-02-24 05:17:15 UTC 
This turned out to be a problem with ext4 in kernel 3.18.

My /etc/fstab had journal_async_commit for the root partition. Apparently the journal_async_commit option cannot be enabled during a remount.

As the root partition is already mounted by the time it reads /etc/fstab, the ext4 driver in 3.18 sees journal_async_commit and "helpfully" remounts root as read-only.  This in turn causes wifi to break.

This is arguably boneheaded behaviour on the part of the ext4 driver.

The real fix is to have a more strict kernel patch review process, in order prevent such boneheaded behaviour ending up in the kernel.


https://bugzilla.redhat.com/show_bug.cgi?id=1190933
Comment 4 Stanislaw Gruszka 2015-02-24 08:27:33 UTC 
I don't know how wifi problem could possibly depend on ext4 problem. Can we get info requested in comment 2 ?
Comment 5 Conrad S 2015-02-25 01:40:07 UTC 
It's as simple as it sounds: ext4 has a bug which causes the root filesystem (/) to become read-only at boot. This in turn stops wifi connecting properly. I presume this is because NetworkManager (or a related subsystem) can't write to the root filesystem (maybe to the /etc directory?).

By having journal_async_commit in /etc/fstab, I can reliably put the root filesystem into read-only at boot, and hence cause problems for wifi.
 
Without journal_async_commit in /etc/fstab, wifi works properly.

Removing journal_async_commit solves the issue for me, but it does point to the actual bug being in the ext4 driver.
Note You need to log in before you can comment on or make changes to this bug.