i have one intel machine, when i reboot with kexex enabled, the display is completely distorted and i can't see why the kernel doesn't boot. restarting the pc without kexec works as expected the machine uses a 3.7.5 vanilla kernel. i also tried 3.6.11 and all versions of 3.7 - allways the same behavior as kexec was working before i'm not quit shure if its a kernel problem or if kexec or any other userland program kexec works fine on two other machines - but this machines use a nvidia card Reproducible: Always Steps to Reproduce: 1.start kexec 2.reboot 3.display distorted
Created attachment 337338 [details] picture showing the distortion after a reboot with kexec
Can you please attach the following information: 1. The part of /var/log/messages (or your logger's equivalent) after you did the kexec, this might reveal something if the kernel reports something. 2. Your current `dmesg` output, can be handy for comparison purposes. 3. /etc/conf.d/kexec, to see how kexec is configured 4. /usr/src/linux/.config, to be able to lookup and verify the settings. Also, at what point exactly does it distort? Do you see anything prior to the distortion?
my /etc/conf.d/kexec file # Load kexec kernel image into memory during shutdown instead of bootup # (default: yes) LOAD_DURING_SHUTDOWN="yes" # Additional arguments passed to kexec (8) #KEXEC_OPT_ARGS="" # Kernel image partition. Mounted automatically if not. # (default: /boot) #BOOTPART="/boot" # Root partition (should be autodetected) ROOTPART="/dev/sda2" # Kernel image pathname, relative from BOOTPART. # If it's one of # {kernel-genkernel,bzImage,vmlinuz,kernel}-<currently running kernel version>, # or bzImage, vmlinuz (without suffix), # then it's automaticaly detected. # Setting it to "-" will disable kexec. KNAME="efi/boot/bootx64.efi" # Initrd # Same automatic detection restriction as for KNAME apply. # initramfs-genkernel-<currently running kernel version>, # initrd{,.img}-<currently running kernel version>{,.img} # will be detected. #INITRD="/boot/fbsplash-emergence-1024x768" # Kernel parameters (should be autodetected) #KPARAM="splash=silent,theme:emergence"
Created attachment 337444 [details] kernel config
Created attachment 337446 [details] kernel messages logged with syslog-ng after kexec was started i rebooted and the log shows the system boot after i manually turn off the system and turned it on again, the kexec reboot just showed the distorted screen - as it seems nothing was logged
Right after you start your kexec you already get this: > Jan 31 13:01:08 onion /etc/init.d/kexec[5072]: Cannot find kernel image. Please make sure a valid kernel image is present before reboot. Maybe efi/boot has to be EFI/BOOT? > Jan 31 13:02:26 onion kernel: [ 0.000000] Kernel command line: root=/dev/sda2 pcie_aspm=force i915.i915_enable_rc6=1 i915.i915_enable_fbc=1 i915.lvds_downclock=1 What setting KPARAM="root=/dev/sda2" and then add parameters as you try again and again, maybe one of the parameters doesn't work well when ran twice on a single cold boot. > Jan 31 13:02:26 onion kernel: [ 0.508301] efifb: probing for efifb > Jan 31 13:02:26 onion kernel: [ 0.508754] efifb: framebuffer at 0xe0000000, > mapped to 0xffffc90006280000, using 1200k, total 1200k > Jan 31 13:02:26 onion kernel: [ 0.508763] efifb: mode is 640x480x32, linelength=2560, pages=1 > Jan 31 13:02:26 onion kernel: [ 0.508769] efifb: scrolling: redraw > Jan 31 13:02:26 onion kernel: [ 0.508775] efifb: Truecolor: size=8:8:8:8, shift=24:16:8:0 I'm not sure whether rebooting through kexec would still keep the system in EFI mode, using the EFI framebuffer when not in EFI could maybe cause distortion. You could try again with a kernel that doesn't have an EFI framebuffer. > Jan 31 13:02:26 onion kernel: [ 0.513675] ACPI: Deprecated procfs I/F for AC is loaded, please retry with CONFIG_ACPI_PROCFS_POWER cleared Just a warning, but you might want to fix this as well. > Jan 31 13:02:26 onion kernel: [ 0.587427] fb: conflicting fb hw usage inteldrmfb vs EFI VGA - removing generic driver > Jan 31 13:02:26 onion kernel: [ 0.590706] Console: switching to colour dummy device 80x25 There you go, you're trying to load two framebuffers simultaneously, choose one and get rid of the other; you probably will want to get rid of the EFI framebuffer. > Jan 31 13:02:26 onion kernel: [ 0.511227] Console: switching to colour frame buffer device 80x30 > ... > Jan 31 13:02:26 onion kernel: [ 0.590706] Console: switching to colour dummy device 80x25 > ... > Jan 31 13:02:26 onion kernel: [ 1.920764] Console: switching to colour frame buffer device 170x48 Your computer first does the low resolution EFI framebuffer, then switched to a low resolution Intel framebuffer after which it switches to a higher resolution Intel framebuffer; this can't be healthy... Conclusion: This is most likely caused by running two framebuffers simultaneously, therefore disabling the EFI framebuffer has a high chance of fixing this; although I've included other suggestions in case it doesn't.
thanks for your help the kernel is at /boot/efi/boot/bootx64.efi it also seems to find it as the kernel that is loaded by kexec is doing this as the last readable line is - "starting new kernel" i also removed the rootpart line from kexec config - but as the kernel never comes up it doesn't mount the root partition, so this was not the problem efi framebuffer and CONFIG_ACPI_PROCFS_POWER is now disabled - but i still end up with a distorted screen
i also tried to remove all options from the kernel command line except root=/dev/sda2 but this also didn't help
Can you provide a new dmesg with these changes to see what it says now?
Created attachment 338006 [details] currrent dmesg output with CONFIG_ACPI_PROCFS_POWER enabled - as i need it for a script
still the same problem with 3.8.1 but i still use 3.7.9 as >=3.8 has a bug in the e1000e driver https://lkml.org/lkml/2013/1/18/147 - should i report the error to the lkml?
(In reply to comment #11) > still the same problem with 3.8.1 but i still use 3.7.9 as >=3.8 has a bug > in the e1000e driver https://lkml.org/lkml/2013/1/18/147 - should i report > the error to the lkml? Upstream bugzilla (http://bugzilla.kernel.org) is an option to try; please also leave a link to the upstream bug here so we can follow along, thank you.
https://bugzilla.kernel.org/show_bug.cgi?id=54671
