Created attachment 685792 [details] Patch to fix watchdog/NMI to resolve the problem Description of problem: Update to any 3.7.x kernel breaks suspend/hibernation in the freezing-process (reproduceable with a chance of 50%). The problem was bisect at https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1079534 and a fix was proposed at http://thread.gmane.org/gmane.linux.kernel/1397985 According to the author that fix is already in Linux 3.8 and should backported to 3.7, although it seems that it did not happen till now. Version-Release number of selected component (if applicable): 3.7.2-201.fc18 and 3.7.3-201.fc18 How reproducible: 50% Steps to Reproduce: 1. Boot affected kernel 2. Try to suspend to RAM or disk 3. With a 50% chance the system locks up while freezing. If not, just resume and try again. Actual results: System freezes. Expected results: System should suspend properly. Additional info: I attached the small patch which fixes the issue. I added the patch to the Fedora Kernel build 3.7.4-201.fc18 and can confirm that the problem is resolved. Since I suspect that Fedora 18 will stick with Linux 3.7 for a while I suggest to add the patch until it is properly backported upstream.
The patch you point to was included in 3.7 as commit 8d4516904b39. The thread was mistaken in that. However, there was a subsequent patch: commit 3935e89505a1c3ab3f3b0c7ef0eae54124f48905 Author: Bjørn Mork <bjorn> Date: Wed Dec 19 20:51:31 2012 +0100 watchdog: Fix disable/enable regression Commit 8d4516904b39 ("watchdog: Fix CPU hotplug regression") causes an oops or hard lockup when doing that fixes another issue. That was backported to 3.7.4, which has been built in F18. It would be good if you could test that and let us know if it solves your issue.
Thanks, indeed it seems that I overlooked that patch. However, I used 3.7.3-201.fc18 from koji since sunday evening and that build did not solve the issue. I just looked at the source and confirmed that this version did already include the patch above. I went straight to 3.7.4-201.fc18 with the additional return-statement in watchdog_disable and had no issues so far. What puzzles me is, that 3.7.2-201.fc18 also had the additional return-statement but did crash anyway. So maybe both patches are needed, but I will try with the original 3.7.4-203.fc18 from koji first and check if that kernel still crashes on suspend.
OK, I made some quick test with 3.7.4-203.fc18 from koji without any further patches and was not able to reproduce the issue. So for the moment it seems that the bug was indeed already fixed. Sorry for the noise. I will make some further test within the next days and will report should the issue raise again. Till then this bug can be closed.