| Summary: | sys-kernel/gentoo-sources-4.12.5 causing random hangs on my AMD64 system in UEFI mode. | ||
|---|---|---|---|
| Product: | Gentoo Linux | Reporter: | JMW <jessemichaelwilson> |
| Component: | Current packages | Assignee: | Gentoo Kernel Bug Wranglers and Kernel Maintainers <kernel> |
| Status: | RESOLVED FIXED | ||
| Severity: | normal | CC: | tsmksubc |
| Priority: | Normal | ||
| Version: | unspecified | ||
| Hardware: | AMD64 | ||
| OS: | Linux | ||
| Whiteboard: | |||
| Package list: | Runtime testing required: | --- | |
| Attachments: |
emerge --info from new system
Dmesg taken from SSH immediately before a hang .config of the affected kernel .config of the unaffected vanilla-sources-4.13.3 kernel |
||
Created attachment 495802 [details]
Dmesg taken from SSH immediately before a hang
Attached is the output of dmesg -w performed over SSH immediately before a hang.
Created attachment 495804 [details]
.config of the affected kernel
Attached is the .config of the affected kernel.
Created attachment 495806 [details]
.config of the unaffected vanilla-sources-4.13.3 kernel
Attached is the .config from the new, apparently unaffected kernel.
Update: My system did eventually hang again after watching several youtube videos, but this hang seemed to be of a very different character. Instead of a sudden lockup, the video stopped and audio started looping, but I still had control over the mouse for ~15-30 more seconds. I'm not sure if this hang is related to the other one, and I've still yet to have a hang while emerging with the new kernel. Update 2: This hang occurred again after emerging the latest nvidia-drivers on vanilla-sources-4.13.3. I attempted to fix this by passing iommu=off to the kernel commandline at boot, which resulted in my mouse and keyboard unable to function after booting. I then went into my kernel configuration and found that IOMMUs for Calgary architecture had been enabled and default-enabled. Changing this setting didn't seem to help the problem. Passing IOMMU=soft to the kernel at bootup has resulted in several days uptime and has likely solved the problem. This seems to be an IOMMU issue, perhaps specific to my hardware, that just happens to be sensitive to small changes in kernel and drivers. Appears you have found a workaround. Let us know if this is not the case, still. |
Created attachment 495800 [details] emerge --info from new system After setting up my new system in legacy/UEFI mode and then switching it over to UEFI mode without CSM, I found my system had become subject to seemingly random hangs. They were initially aggravated by heavy system use (i.e. emerging packages) and playing video, but could not be induced predictably to occur. The hangs were total system lockups, preventing all activity including logging in from SSH. Enabling all options under "Lockup Detection" in the kernel did not allow the lockups to become kernel panics. Believing this to be some sort of graphics issue, perhaps having to do with nvidia binary drivers, I initially started booting with 'nomodeset', which did not help, and then I unmerged nvidia-drivers and eliminated every framebuffer in the kernel with the exception of efifb. This had the opposite effect - lockups went from being largely random to being highly predictable, occurring during any emerge lasting longer than 5 minutes and almost immediately upon watching a video in chrome. This behavior was only observed using gentoo-sources-4.12.5. After getting some help on IRC, I emerged vanilla-sources-4.13.3, copied over my previous .config, followed by 'make oldconfig', 'make', 'make modules_install', and 'make install'. Since switching to the vanilla-sources, my computer has been up for about 12 hours without a problem. I've both emerged packages and watched videos in chrome. Unfortunately Gentoo no longer has stable vanilla-sources, so I consider this a serious problem, at least on my hardware. Attached is my emerge --info from my system with the new kernel.