On 24/08/18 18:41, Guenter Roeck wrote: > On Fri, Aug 24, 2018 at 09:30:22AM -0700, Guenter Roeck wrote: >> Hi Guillaume, >> >> On Fri, Aug 24, 2018 at 08:32:07AM +0100, Guillaume Tucker wrote: >>> On 22/08/18 14:51, kernelci.org bot wrote: >>>> stable/linux-4.4.y boot: 124 boots: 3 failed, 117 passed with 2 offline, 1 untried/unknown, 1 conflict (v4.4.151) >>>> >>>> Full Boot Summary: https://kernelci.org/boot/all/job/stable/branch/linux-4.4.y/kernel/v4.4.151/ >>>> Full Build Summary: https://kernelci.org/build/stable/branch/linux-4.4.y/kernel/v4.4.151/ >>>> >>>> Tree: stable >>>> Branch: linux-4.4.y >>>> Git Describe: v4.4.151 >>>> Git Commit: 78f654f6cce3442937b8c7eb4b640357871363c1 >>>> Git URL: http://git.kernel.org/pub/scm/linux/kernel/git/stable/linux-stable.git >>>> Tested: 43 unique boards, 20 SoC families, 22 builds out of 191 >>>> >>>> Boot Regressions Detected: >>> [...] >>>> x86: >>>> >>>> defconfig+kvm_guest: >>>> x86-x5-z8350: >>>> lab-mhart: failing since 6 days (last pass: v4.4.147 - first fail: v4.4.148) >>>> >>>> x86_64_defconfig: >>>> x86-x5-z8350: >>>> lab-mhart: failing since 6 days (last pass: v4.4.147 - first fail: v4.4.148) >>> >>> We've had a couple of automated boot bisections pointing at the >>> same commit on an x86 Atom x5-Z8350 platform (AAEON UP-CHT01), >>> please see details below. The issue seems to have started with >>> v4.4.148. >>> >> >> Can you test if a backport of upstream commit d367cef0a7f0 ("x86/mm/pat: >> Fix boot crash when 1GB pages are not supported by the CPU") fixes >> the problem ? It doesn't apply cleanly, but the conflict is easy to >> resolve. >> > > We also may have to reapply commit 87e2bd898d3a ("x86/mm/pat: Ensure cpa->pfn > only contains page frame numbers") and fix whatever problems it had. I've applied these 2 patches on top of v4.4.152: https://gitlab.collabora.com/gtucker/linux/commits/linux-4.4.152-bisect-atom-x5 Then I've run the tests using these 2 commits, but they both still failed to boot on that platform: http://lava.streamtester.net/scheduler/job/145549 http://lava.streamtester.net/scheduler/job/145548 For the record, here's the same test with v4.4.147 that works: http://lava.streamtester.net/scheduler/job/145547 Let me know if you have other fixes to apply, I can take another look at some point next week. Guillaume
