I've upgraded my 2.6.25-r9 gentoo-sources kernel to 2.6.29-r5. The kernel loads, but after mouting the root (readonly) Gives the message: Kernel panic - not syncing No init found. Try passing init = option to kernel It loads and works with the old kernel. I've runned fsck but found nothing. The same happens with 2.6.28-r6 If I try to boot with init=/sbin/init or /sbin/sh Then I get: cannot execute /sbin/init (or /sbin/sh) And then try to bhave as default, and get the same error Kernel panic - not syncing No init found. Try passing init = option to kernel Reproducible: Always Expected Results: After the mount the kernel should execute init
Created attachment 192769 [details] conifg for kernel 2.6.29-gentoo-r5
Created attachment 192771 [details] config for the (working) 2.6.25 kernel
1. The "working" .config you included is from 2.6.29-gentoo-r5, read the top of the file. I think you just copied the wrong one. Please attach the correct file and mark the other one as obsolete. 2. please capture your kernel boot log output (serial console recommended) for the failed boot, and your full dmesg right after boot from the working kernel.
Created attachment 192884 [details] config of the (working) kernel
Created attachment 192885 [details] working dmesg
Sorry, but I can't attach a serial console to my home desktop. But at the start, and at the end, there is no error message such the root partition cannot be mounted. It says it is mounted (readonly) kjournald starting (delay the write 5s). Then the above error: no init found. I've a new videocard so I've needed a new kernel. I've a new 2.6.25 kernel, and today if i start the 2.6.29 kernel after the startup with the new (2.6.25) kernel It didn't says anything, simply freez.
It's possible that sda and sdb are being swapped. can you check by looking carefully at the boot messages on the failing kernel? You can see which one is which in your working dmesg.. each disk has a different product number and size scsi 0:0:0:0: Direct-Access ATA HDT722520DLA380 V44O PQ: 0 ANSI: 5 sd 0:0:0:0: [sda] 390721968 512-byte hardware sectors (200050 MB) scsi 2:0:0:0: Direct-Access ATA Hitachi HDP72502 GM2O PQ: 0 ANSI: 5 sd 2:0:0:0: [sdb] 488395055 512-byte hardware sectors (250058 MB)
I'm in agreement with Daniel here. The devices have probably swapped order. You aren't seeing it fail to mount because the mount IS working, but to a device that doesn't have /bin/init, because it was your secondary drive. The async boot stuff in recent kernels DOES make it possible for the order to be different. If it does happen that you need to change the order, you'd have to change fstab. To avoid this, I suggest changing the driver for the secondary disk to be a module, so that /dev/sda is fixed at being a specific disk. Alternatively start using labels or uuids with an initramfs. Please test changing the root= argument, and reopen the bug as needed.
No there were no swap. I've tried to start with hda2, hdb2 sdb2 etc. I've get the message (with e.g. sdb2) that there are parititions which can be mount as root such: hda1, hda2, hda3, hda4, hdb - cannot be mounted because it is a cd/dvd drive (it is diagnosed by the kernel) sda1, sda2 (the size is match with the real sda2), sda3, sda4 So i cannot mount sdb2. And If i try hda2 the kernal says: I cannot mount (because it is a swap partition) There is no swap. The kernel mounts the partition and caanot execute a file on ot.
Sry. Cannot execute a file on it.
We really need the kernel log from the failing case then. Either find a serial cable, or explore netconsole. I don't know how workable the console-over-firewire is these days, but console-over-USB-serial should work.
Could you give me a link about Console -over-USB (or firewire)? I found a desciption on the internet but I make out that I need (kermit - minicom) runnig to get console over USB. Will it run if the kernel couldn't execute anything on the drive? Or put in an initrd? Will it run before init?
Rebuild the kernel, compiling into the kernel the following: - USB host controller - CONFIG_USB_SERIAL - The right USB-serial hardware support. After you have the above selected as built-in, there is a new option: CONFIG_USB_SERIAL_CONSOLE Enabling that new option, recompile and install the new kernel. Boot with the following added to your kernel parameters: console=tty0 console=ttyUSB0,115200n8 Does your system not have a serial port or something (and you've got no PCI cards that give you a serial port)? I forgot that serial-console-over-firewire never made it to the final Linux kernel from an external module (it is in freebsd however). Alternatively, do you have a digital camera? You could try to snap each page after the output stops (using the pageup to start from the top). If you take them at high resolution and fully aligned, the OCR stuff in the tree is usually able to convert it back to text (and you can do the manual corrections). This is probably more time-consuming and error-prone than finding some serial rig.
the initramfs is also maybe an idea, if you built a custom busybox with stuff to send the dmesg capture somewhere else over the network.
First of all, I'm not in the IT. I have lot of work and two children so I will have time for that at later this week. (Now I'm using 2.6.25 kernel with other config) > "Boot with the following added to your kernel parameters: console=tty0 console=ttyUSB0,115200n8" Thanks. I will try. I have an old laptop with Xubuntu. It will be a good console I think. > "Alternatively, do you have a digital camera? You could try to snap each page after the output stops (using the pageup to start from the top). If you take them at high resolution and fully aligned, the OCR stuff" Yes, I have. I think I've OCR too. But the batteries are depleted today so again I can try it later. > "the initramfs is also maybe an idea, if you built a custom busybox with stuff to send the dmesg capture somewhere else over the network." Probably I could solve it (again I'm not in the IT), but I haven't enough time for it. I will write again, if I have the log. (And yes I will have the log) Thanks.
Hello, Don't know if it helps you. I solved the problem by creating a new Kernel. mv /usr/src/linux/.config /usr/src/linux/.config_bak mv /usr/src/linux/.config.old /usr/src/linux/.config.old_bak Then I created with make menuconfig the new .config file. Problem: You have to configure the complete kernel from the scratch. I think between 2.6.25 and 2.6.29 too much changes happen in .config file. Cheers Roland
roland: You can use make oldconfig. That's not the issue here at all, you can diff Lorant's .config files he provided, they are nearly identical. The ONLY difference is: 2512c2512,2513 < # CONFIG_SOUND_PRIME is not set --- > CONFIG_SOUND_PRIME=y > # CONFIG_SOUND_OSS is not set (unless he uploaded the wrong .configs again).
I have similar error. Tried 'root=UUID=xxx-yyy-zzz-...' instead of /dev/sd... - did not help.
I have exactly the same problems. I didn't write them here before cause I thought, it's a stupid mistake in the .config by me cause current Live-Gentoos work. I still believe it's my mistake, but after the previous messages... Perhabs this information helps. gentoo-sources-2.6.26-r4 works fine, I can configure it any serious way and it boots up. After this release, every newer kernel I tried (2.6.27-r8, 2.6.28-r5 and finally 2.6.29-r5) produced this error message, no matter what I gave to the init=-parameter. I have only a single HardDisk in my notebook(Thinkpad T61, amd64 / nocona, DualCore). sda1 is /boot, sda2 is /, sda3 is hibernate (not yet working :-( ) and sda4 is swap. /sbin/init is 35k-0755-file and works fine with older kernels. sda2 is detected as a reiserfs3.6 like it really is, it's mounted read-only. My kernel-parameters are only root=/dev/sda2. Adding a init=<anything> brings an additional line of "cannot execute init, returning to defaults". After this line, the cannot execute of init follows and the system hangs. Using similar .config files on my older notebook(T23, x86) as well as on my desktop (amd64 / athlon64, DualCore), works fine.
Stephan, please file a separate bug for your issue unless the ideas below solve your problem. Lóránt, in your config you have this: CONFIG_INITRAMFS_SOURCE="/usr/share/v86d/initramfs" this means that your root= arguments are being ignored by the kernel and it is building an initramfs from that directory, which is supposed to be able to find the root partition and mount it. My guess is that directory is empty or incomplete. As a next step, please blank out that setting so that you boot without an initramfs. If that works then we can reassign to the v86d maintainer to figure out the initramfs problem.
I have in my .config CONFIG_INITRAMFS_SOURCE="", but the same problem with one motherboard and no problems with another (like enforce4). I think the problem is in sata driver for nvidia chip: > 00:0e.0 IDE interface: nVidia Corporation MCP51 Serial ATA Controller (rev a1) > 00:0f.0 IDE interface: nVidia Corporation MCP51 Serial ATA Controller (rev a1)
Created attachment 195850 [details] working config for 2.6.29-gentoo-r5 There were nothing relevant in the log of non-working kernel. However at later the same kernel didn't give the failed init message but simply hanged up at the same place. BUT, now I have a working new kernel. I have made this by make mrproper then make defconfig then I selected (with make menuconfig) the proper Processor type IDE/SATA chipset etc... Now everything is working. I didn't find any serious difference between the working and non-working config. Maybe it is a problem with the import of the old config (as I see most of the people here has tried to import their old config). I suggest everyone to try config his/her new kernel as above.
I am still reasonably confident that it is a problem with your initramfs. If you rebuilt your working kernel using the same config right now, your newly rebuilt version would encounter the same problem. This is because once upon a time you had a usable initramfs in /usr/share/v86d/initramfs which was built into your kernel, but now for some reason you do not, so any kernel built at this point includes an unusable initramfs. This is only an educated guess though, but I imagine your new kernel .config does not have CONFIG_INITRAMFS_SOURCE="/usr/share/v86d/initramfs" (and if it did then you would not be able to boot). Either way, if it's working now, I'll close this bug. I also do not recommend copying a .config between a big kernel change e.g. 2.6.29 to 2.6.30 because of the huge amount of change in each kernel release.
Roman, you should file a separate bug.
It has nothing to do with CONFIG_INITRAMFS_SOURCE="/usr/share/v86d/initramfs". It is only needed for uvesafb. See http://dev.gentoo.org/~spock/projects/uvesafb/ My first kernel which started to work was a 2.6.30 kernel, without CONFIG_INITRAMFS_SOURCE="/usr/share/v86d/initramfs". An it worked (though without uvesafb). The problem was with the import of an old config. I imported the config-2.6.25-gentoo-r5 manually (In the menuconfig of 2.6.29). I think this was a mistake. If you encounter this problem try configure as above.
Having that setting set to *anything* drastically changes the way the kernel boots, to the point where the kernel itself doesn't even look at your hard drives in order to try and find init. So it still seems very likely to me that this was the cause of your problem. If there is one small problem inside that directory then you'll see exactly this problem.
