When iommu is enabled, kernel constantly flooding with the following errors: [ 3.943678] AMD-Vi: Completion-Wait loop timed out [ 4.069905] AMD-Vi: Completion-Wait loop timed out [ 4.193441] AMD-Vi: Completion-Wait loop timed out [ 4.314378] AMD-Vi: Completion-Wait loop timed out [ 4.435285] AMD-Vi: Completion-Wait loop timed out [ 4.556225] AMD-Vi: Completion-Wait loop timed out [ 4.710102] AMD-Vi: Completion-Wait loop timed out [ 4.836121] AMD-Vi: Completion-Wait loop timed out [ 4.962122] AMD-Vi: Completion-Wait loop timed out [ 5.088122] AMD-Vi: Completion-Wait loop timed out [ 5.214121] AMD-Vi: Completion-Wait loop timed out [ 5.340118] AMD-Vi: Completion-Wait loop timed out [ 5.466123] AMD-Vi: Completion-Wait loop timed out [ 5.603982] AMD-Vi: Completion-Wait loop timed out [ 5.730996] AMD-Vi: Completion-Wait loop timed out [ 5.857953] AMD-Vi: Completion-Wait loop timed out [ 5.984904] AMD-Vi: Completion-Wait loop timed out [ 6.111936] AMD-Vi: Completion-Wait loop timed out [ 6.269186] AMD-Vi: Completion-Wait loop timed out [ 6.398025] AMD-Vi: Completion-Wait loop timed out [ 6.524039] AMD-Vi: Completion-Wait loop timed out [ 6.650039] AMD-Vi: Completion-Wait loop timed out System boots with a lot of service start timeouts. Network, display, etc doesn't work. However I can see login prompt via serial console. %) Mainline kernel (vanilla-sources-4.8.8) with the same config works fine. Disabling IOMMU (amd_iommu=off on the kernel cmdline) workarounds this issue. The CPU in this system is Piledriver-based Opteron (4300 series).
Created attachment 453750 [details] kernel log (hardened-sources-4.8.8)
Created attachment 453752 [details] kernel log (vanilla-sources-4.8.8) Ignore IO_PAGE_FAULT errors caused by amdgpu driver. It's a known unrelated issue.
The problem persists with hardened-sources-4.8.9. I also tested gentoo-sources-4.8.9 and It works fine when IOMMU is enabled.
It's due to GRKERNSEC_KSTACKOVERFLOW
this is an issue that upstream already fixed as it got exposed by their new VMAP_STACK feature (similar to grsec's KSTACKOVERFLOW): http://git.kernel.org/cgit/linux/kernel/git/torvalds/linux.git/commit/?id=4bf5beef578e46393f11eb69dda7d17a065e05ff . i guess we'll just backport it into grsec if you can confirm that it fixes your 4.8 too.
(In reply to PaX Team from comment #5) It fixed my 4.8. Thanks!
Fixed in hardened-sources-4.8.10 (grsecurity-3.1-4.8.10-201611232213.patch)
(In reply to Alexander Tsoy from comment #7) > Fixed in hardened-sources-4.8.10 (grsecurity-3.1-4.8.10-201611232213.patch) thanks for testing i was just about to ask :)
