Created attachment 106830 [details] dmesg output when error occurred. Linux 3.14.7/3.14.14 Baytrail CPU Boot into linux without X, use console output to HDMIA1. Sometimes ther is no display output after perform reboot[occurrance 1/10], power on seems okay. When there display not output, ttyS4 of HSUART is okay. Use intel_reg_dumper to dump the register value to comapre normal and abnormal situation: DPLL_A_MD: 0x60ff2cdf (have output) DPLL_A_MD: 0x70ff2ddf (no output) DPLL_B: 0x60ff2cdf (disabled, dvo, VGA, TV A clock, unknown mode, p1 = 1, p2 = 0) (have output) DPLL_B: 0x70ff2ddf (disabled, dvo, TV A clock, unknown mode, p1 = 1, p2 = 0, using FPx1!) (no output)
Please try current upstream v3.17-rc6 to see if it works. Sadly intel_reg_dumper does not properly support BYT, please try tools/quick_dump/quick_dump.py -a from the intel-gpu-tools package [1]. [1] http://cgit.freedesktop.org/drm-intel
test 3.17.0-rc7, 30 times pass, occurrence 0/30 How do I merge the change into 3.14 series? Thanks
(In reply to Rick from comment #2) > test 3.17.0-rc7, 30 times pass, occurrence 0/30 > > How do I merge the change into 3.14 series? The first step is *finding* the change that makes the difference. It may be it's not just one change, but a series of changes. To give you an idea what kind of haystack you're looking for the needle in, there have been 1300+ changesets merged into the i915 driver alone between v3.14 and v3.17, with a diffstat of: 67 files changed, 25925 insertions(+), 9362 deletions(-). That doesn't even include the drm core changes. You could try to do a reverse git bisect to find the commit that fixed the issue for you. It's roughly 14 build/test steps. For example https://wiki.ubuntu.com/Kernel/KernelBisection has this explained. Once you've found the commit(s), the next step is backporting the fixes back to v3.14, preferrably contributing them to the longterm stable v3.14. Unfortunately they may not be directly applicable to v3.14, and a lot of work may be required. Finally, this is upstream bugzilla, and we have fixed the issue upstream, thus I'm closing the bug. We may be able to help if you identify the working commit. HTH.
Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.