Created attachment 2863 [details] dmesg Image tested: meta-intel-corei7-64 Image Link: http://yocto-ab-master.jf.intel.com/pub/releases/yocto-meta-intel-4.0-jethro-2.0.rc1/machines/intel-corei7-64/core-image-sato-sdk-intel-corei7-64.hddimg HW Tested: Crystal Forest/Server CPU E5-2658 poky: fc45deac89ef63ca1c44e763c38ced7dfd72cbe1 meta-intel: 1ce13b3d39b2da79916e3a175cee6baa507290f1 Steps to reproduce: 1. Install image to device. 2. From terminal enter so that the HW will go to sleep: $ echo "mem" > /sys/power/state 3. After the system goes into stand by mode, move mouse or press any key to make it resume. Expected: 3. The system resumes from stand by. Actually: 3. The system starts but, the screen remains black (X fails to show up) If you boot the image from stick (live boot), put the system in stand by and try to resume it, the HW doesn't even start. Attached you can find dmesg and Xorg log files.
Created attachment 2864 [details] Xorg
It looks like very similar like this bug which can be fixed with 4.4 kernel in OE. https://bugzilla.yoctoproject.org/show_bug.cgi?id=8311 But we need to make sure it. I will send reporter an Image with 4.4 kernel to verify it.
Hi Alex, We suspect it could be the similar issue possibly fixed with 4.4 kernel. Our crystalforest target is hooked with KVM in lab, so it is not convenient for me to try this image for a suspend/resume issue. Could you try this image on impacted target to see if the issue can still be produced? Thanks lot! https://drive.google.com/open?id=0B5-srET1lZkiUUg1ZUNUUmk2blk
The image fails to install on my Cristalforest. When I try to install or boot it from USB I receive this message: "end Kernel panic - not syncing: Invalid configuration from end user prevents continuing"
Apologize for the first image's issue. I don't know what happened during my build but the test image shows same thing on my NUC 6. https://drive.google.com/open?id=0B5-srET1lZkiWnFyLWpiN1Fiems The above is an updated image which is actually a minimal build for genericx86-64. I still believe it is an issue n kernel so it shouldn't matter between a generic and a meta-intel build. I have tried this image on Nuc 6 and this issues seems fixed.
The core-image-minimal cannot be installed. It hangs with black screen after I choose the install or boot option through the USB stick.
Then it looks we have a more serious issue on crystalforest. I am checking people at my site to get another crystalforest because the one hooked in our test infrastructure is not easy to work with suspend/resume issues. I will update you once I find a way to try it. Thanks lot!
Hi Cristina, There is an update from my trial on shumway board in our lab. () I found the BIOS cannot boot UEFI by default. I have to create UEFI entries manually in Boot -> Add new boot option, and specify Bootx64.efi on a USB drive to live-boot image (a very recent intel-corei7-64 sat image) () After some tweaking in BIOS, I still cannot get display back in our lab infrastructure. But the behavior is slightly different when I enabled Advanced->ACPI Settings->S3 video repost. Display cannot come back but fans on system starts running in resume. Could you give this option a try? () To have a better understanding of this issue, could you also provide information of BIOS, video card and processor on your cystalforest system? Thanks
I enabled kenrel log from serial on shumway board but I cannot get any output after resume. Kernel configurations for PM debugging look enabled already. It means we still need to narrow down a little bit before concluding it is a video issue. In a talk with firmware team, the S3 should be supported, but they don't think yocto is a supported distro when shumway was released, so they don't know which kernel was verified at that time. I will ping Intel's PAE and do some debugging in parallel. Still need a talk with QA to see if we have same setup (e.g. video card model).
Beside from the previous comment I made, we also have ACPI errors and suspicious logs from kernel from a new dmesg of recent build. (also can be seen in existing dmesg in attachment.) But I am not sure if this means the root cause... 364:[ 0.584222] ACPI: PCI Root Bridge [PCI0] (domain 0000 [bus 00-fe]) 393:[ 0.584939] pci 0000:00:01.0: System wakeup disabled by ACPI 396:[ 0.585101] pci 0000:00:02.0: System wakeup disabled by ACPI 399:[ 0.585262] pci 0000:00:03.0: System wakeup disabled by ACPI 426:[ 0.590609] pci 0000:00:1c.0: System wakeup disabled by ACPI 430:[ 0.590799] pci 0000:00:1d.0: System wakeup disabled by ACPI 489:[ 0.597346] pci 0000:04:00.0: System wakeup disabled by ACPI
What's inserted on the Shumway/crystalforest platform in my lab is a NVIDIA video card. I can reproduce the issue with very recent revisions, poky: 2f2f784727 and meta-intel: 3f637d34. It turns out its DRM driver is necessary to enable S3 power state on the card. CONFIG_DRM_NOUVEAU must be enabled in kernel configuration to fix this issue per my test. However, the original Shumway Development Kit should be shipped with a Matrox graphic card which could need another DRM driver. We need to know Alex or Cristina's video card models before we can close this issue at all sites. It is also possible the original issue has a different cause too. I still cannot reach Cristina in these days in email, IM or IRC. I assign this issue back to Cristina as a heads up. Please assign it back to me ASAP when you can provide video card model info on your Shumway board, so we can deliver test images for validation. I will talk to Saul in BSP team to find the best location to enable NVIDIA config for our board in parallel. Thanks
Adding Bejanmin and AlexG for QA
Created attachment 3066 [details] Image Hi, sorry for responding so late. I tested again for a core-image-sato-sdk image downloaded from http://yocto-ab-master.jf.intel.com/pub/releases/yocto-2.1_M2.rc1/machines/intel-corei7-64/ I installed the image, put the HW in sleep mode $ echo "mem" > /sys/power/state , tried to resume from stand by but only fans on system starts running. When I enable Advanced->ACPI Settings->S3 video repost I get the result from Image attachment.
Created attachment 3067 [details] dmesg-file Attached you will find the output of dmesg command. The output of lspci | grep vga command is: 03:00.0 VGA compatible controller: Advanced Micro Devices, Inc. [AMD/ATI] Park [Mobility Radeon HD 5430] The output of grep -i chipset /var/log/Xorg.0.log command is: [3287968.387] (II) VESA: driver for VESA chipsets: vesa Hope this is useful.
Information appears to have been provided. If you still NEEDINFO, please add a comment on what additional information is needed.
Alex, Thanks for the information. I found a Radeon card (R500/R520) in lab but don't have exact ATI model you mentioned. I did a cross test on this ATI video again: Without drivers, same issue as Nvidia card, black screen after resume. With two enabled arm drivers for ATI/AMD in build, mem power state works as expected. (BIOS ver CCFRCLC0.021) I have shared the test image for your test with ATI card: https://drive.google.com/drive/folders/0B5-srET1lZkiWnFyLWpiN1Fiems BTW, you probably see a low and scaled resolution in console with existing builds. But with proper driver enabled, you should see system switches to the high and default resolution for display properly after reboot. "S3 Repost video" option in bios should not be relevant. Anyway, I think we almost conclude the cause is lacking driver for video cards inserted onto shumway board. We may close this as "won't fix" unless you find new issues in this case. I haven't found anything wrong with Shumway board itself. Steven, I will change state accordingly based on Alex's next feedback.
Created attachment 3070 [details] screenshot I downloaded your image and when I want to boot it or install it on my device the process get stuck. See screenshot attachment.
Alex, Thanks to try the image. I noticed your ATI card does show some special symptoms in our cross-testing. It shows color blocks after resume at your side, but I always get black screen on my ATI and NV cards. Hope these suggestions can help you. () Check if the downloaded image's md5sum, I assume no problem at this step () Check BIOS version of your Shammy platform, update and retry if it is not the version I mentioned. () I found there is an ATI proprietary linux driver for Radeon Mobility HD 5000, you can try to run it after unzip on a released yocto image, to see if it can help. http://support.amd.com/en-us/download/desktop/previous/detail?os=Linux%20x86_64&rev=15.9 () Alternatively, you can try other video cards from ATI or Nvidia, I have uploaded another good image for my Nvidia card at same link: https://drive.google.com/drive/folders/0B5-srET1lZkiWnFyLWpiN1Fiems I still think all clues and different results clearly point to video cards instead of shumway board, assuming we have same version of firmware on our boards.
Hi Jianxun, I followed your steps, the BIOS version is CCFRCLC0.021 like you mentioned. I downloaded the driver, run it on a core-image-sato-sdk image installed on my CristalForest, but still the same result, the system doesn't resume from stand by. I tested you new image, I was able to install it, but when I try to resume the HW from stand by only fans starts running. I think you're right, the problem is from video card.
Alex, Appreciate your support on this issue. I just want to reminder you the 2nd image I sent is for Nvidia card only, so that if you need your shumway setup works, you can try that image with another Nvidia graphic card. We will close this bug as "won't fix' since the issue is from video card instead of Shumway board. Developers should enable expected vendor driver for their cards but this is out of our control. Thanks lot!
Talked with Saul. close as "won't fix". Thanks for Alex's support
refer to previous discussion. Concluded as a video card issue instead of Shumway's bug.
According to the discussion this bug is resolved wontfix. Thanks Jianxun! Put this on verified.