BUILD: 1.8.1: b74ea963cefffad9fbd91d4eb9b240f6a8c86cd0 ENVIRONMENT: genericx86-64 on NUC IMAGE: core-image-lsb-sdk STEPS TO REPRODUCE: Parse dmesg log and all log files in /var/log for keywords like: "error" , "failed" , "cannot", "can't" etc. Filter out the known and invalid errors or errors that appear in certain situations and need to be ignored. EXPECTED OUTCOME: No errors or fails should be present in logs Actual error: This error appears during LSB image testing on NUC in /var/log//auth.log : Central error: Oct 1 00:10:01 genericx86-64 getty[1028]: can't open '/dev/tty1': Input/output error^M
NUC model used for this test was D54250WYK
Hello Cristina, thanks for filing the bug and for the description. I wanted to ask you whether this message is seen every time that the test runs or if perhaps you had checked the reproducibility (% of occurrences)
Cristina, can you please let me know step by step how to copy the image to the USB stick? I'm using jethro and copying the image with wic and I'm getting kernel panics when booting the image on a D54250WYK nuc.
I was able to reproduce the error using the fido branch, I tried again in jethro and the error was not there. It seems the error is for pam. In both cases tty1 refers to the console default login, I logged into the device without any problem even when the error appears. Is this a critical bug that the solution has to be backported to fido?
This is a intermittent issue and it happens about 1/8 of the time. This happens at boot and I haven't never seen it after. This also affects jethro and current master. I have to investigate more on this.
After doing more test, I was able to reproduce it only with grub as the bootloader. When using syslinux I booted the devices 25 times without any error.
Today I tried to get the differences between using syslinux and grub. When I booted using syslinux, the error was there. So this happens with grub and syslinux.
Moving to 2.1 M2. I'm still not able to reproduce the error at will.
I tried to reproduce the error with "core-image-full-cmdline" using "poky-lsb" distro, but I couldn't reproduce it in 40 tries.
Today I tried to recreate the issue with core-image-lsb-sdk, the same used to open the bug. I couldn't trigger the error in 30 reboots. This time I did based on commit b3d6872bf169a13e572638e3809c50a868246616.
Checking the commits, I found f1c034be02b1e8cf13d580de0de634794f7623e8 added a couple of patches for libpam but only when using musl lib. I need to investigate more on this.
This bug is triggered when trying to run getty and the console doesn't exists yet in /dev. Commit ccb2a57ad25c98e77f81f8e3bf82e6a1da2d9cd2 adds a mechanism to verify if the console exists before trying to run getty on any tty, so getty will run when the virtual console is ready. So this solves this bug also. I verified this running over 100 times the test using jethro and cherry-picking this particular commit. Because this is not a critical bug (getty will be run over and over again and the console gets ready after a couple of seconds), I won't backport this commit to jethro.