[Troubleshoot] [ArchLinux] [X] GPU Hang

Preface: The following content has been in hang after about half a month, then want to write, has not come to write, write blog is also already spent time things.

have been on vacation recently, the use of computer is not much. Later or for the sake of life, reluctantly to start work, the first thing to work is what? Of course, after the network was configured, I started rolling the system (the meaning of updating the whole system to the newest package, because one of the features of our Archer is the rolling update, well, the updates are all rolled. , for a long time no serious use, of course, to enter the best up-to-date status to a slightly happy start to work, casually also for the new colleagues show me a system.

On the sauce, it took two days. I found something wrong. Use the X to hang up. Needless to say, the first is the suspicion of rolling, the system rolled into what unstable package.

Check the list of all the scroll packages, which are here:

[[email protected] ~]$ vim/var

View System log, two strokes: 1, Journalctl. 2. DMESG

I'm really sorry that I didn't leave log at that time (just half a month ago), and the main thing in log is that

GPU Hang, please go to "https://bugs.freedesktop.org/" to report the bug

To synthesize the above information, I first judged that Intel's driver was out of the question. Check out the list of roll-up packages and it's true that the kernel Pack (Linux) update was found. (Intel drivers are in the Linux kernel package). Then I downgrade the Linux kernel package.

Downgrade method: Because Arch is a rolling update and all packages are provided upstream, the official downgrade method is not available. In other words, Arch is only responsible for letting you roll up, not to help you roll down.

Based on the above, the general practice is to keep the old version of the update package, when the problem occurs, you can manually say that a single package fallback to the specified version.

Old bags are kept in this place./var/cache/pacman/pkg/system will be based on rules to delete the oldest, the path and rules are configurable, how to do not write, interested in self-inquiry manual.

Then find the Linux version of the package I need to drop back, use Pacman to drop back, and restart:

[Email protected] ~]$ pacman-u/var/cache/pacman/pkg/linux-4.7. 6-1-x86_64.pkg.Tar

The above, is to confirm that is indeed a package failure when the general solution.

However, it is conceivable that my question is not so simple. Downgrade the kernel, the problem still exists.

And then plunged into a state of no clue, so, ask Google to help. Also no useful information was found.

During this period I have been searching for clues that led to this problem, I was still looking at DPDK documents, while driving a win virtual machine and some other project colleagues through the QQ communication, yes, we rely on QQ. In this way, the memory will reach a critical value shortly after work, and both Chrome and QEMU are eating memory. So I started to wonder what was wrong with chrome. Chrome's hardware acceleration, for Firefox, did not solve the problem successfully.

Become helpless, I began to suspect KDE (unlucky KDE, not once again suspected), want to change to FVWM, this time again mentioned FVWM, reconfigure as soon as possible. So before reconfiguring the FVWM, I am desperate to report the bug, because this problem, it is difficult to provide useful information to the developer, from the report bug to solve, it is a long process.

Before reporting a bug, it is natural to look for a bug that has been reported. And then I found this:

https://bugs.freedesktop.org/show_bug.cgi?id=89360

which mentions:

Short version---'intel_iommu=igfx_off' helped

Yes, remember, in order to debug DPDK, I opened the vt-d function? is to allow the PCI device to be used directly in the user state. Join http://www.cnblogs.com/hugetong/p/5904024.html

So in retrospect, every time the problem started with QEMU.

Summarize:

1. Remove the kernel option Intel_iommu=on

2. Upgrade the kernel to the latest.

3. Restart

Problem solving:)

Solve the settlement, or there are two problems arising out of

1. FVWM, need to be reconfigured, in a usable state, ready to stand by.

2. Heard about the Zfs,brtfs two file systems? Can file system level snapshot, rollback??

[Troubleshoot] [ArchLinux] [X] GPU Hang