> Am 17.02.2016 um 12:01 schrieb Guillaume Gardet <[email protected]>: > > > > Le 17/02/2016 11:15, Alexander Graf a écrit : >> >>> Am 17.02.2016 um 11:00 schrieb Guillaume Gardet <[email protected]>: >>> >>> Hi, >>> >>> ARMv7 images are broken. >>> >>> On boot, with latest build (kernel 4.4.0-3-default) I get the following >>> error: "Initramfs unpacking failed: write error". >>> Full kernel log (for beaglebone black): >>> ******************************************************** >>> Starting kernel ... >>> >>> [ 0.000530] WARNING: Your 'console=ttyO0' has been replaced by 'ttyS0' >>> [ 0.000538] This ensures that you still see kernel messages. Please >>> [ 0.000545] update your kernel commandline. >>> [ 5.154972] Initramfs unpacking failed: write error >>> [ 5.351054] omap_voltage_late_init: Voltage driver support not added >>> [ 5.357567] sr_dev_init: No voltage domain specified for smartreflex0. >>> Cannot initialize >>> [ 5.365720] sr_dev_init: No voltage domain specified for smartreflex1. >>> Cannot initialize >>> [ 5.563828] sr_init: platform driver register failed for SR >>> [ 5.575720] Unable to find PPMU node >>> [ 5.586412] Kernel panic - not syncing: Attempted to kill init! >>> exitcode=0x00007f00 >>> [ 5.586412] >>> [ 5.595610] CPU: 0 PID: 1 Comm: init Not tainted 4.4.0-3-default #1 >>> [ 5.601901] Hardware name: Generic AM33XX (Flattened Device Tree) >>> [ 5.608068] [<c022876c>] (unwind_backtrace) from [<c0221904>] >>> (show_stack+0x20/0x28) >>> [ 5.615856] [<c0221904>] (show_stack) from [<c0575ee8>] >>> (dump_stack+0x9c/0xdc) >>> [ 5.623125] [<c0575ee8>] (dump_stack) from [<c03803c0>] >>> (panic+0xb0/0x248) >>> [ 5.630039] [<c03803c0>] (panic) from [<c0274e54>] >>> (complete_and_exit+0x0/0x2c) >>> [ 5.637383] [<c0274e54>] (complete_and_exit) from [<c0274eec>] >>> (do_group_exit+0x4c/0xcc) >>> [ 5.645510] [<c0274eec>] (do_group_exit) from [<c0274f8c>] >>> (__wake_up_parent+0x0/0x34) >>> [ 5.653464] [<c0274f8c>] (__wake_up_parent) from [<c021ce60>] >>> (ret_fast_syscall+0x0/0x34) >>> [ 5.661690] Rebooting in 90 seconds.. >>> ******************************************************** >>> >>> Note that there is no early message (except warning on console, for which I >>> will send a JeOS SR). >>> >>> Any idea what is wrong? >> Looks like the initrd is too big (again). Please try to set the rootfstype >> to ramfs in the kernel command line to verify. > > It did the trick indeed! > >> >> I guess we really need to come up with a good way around this whole mess. I >> like the squashfs-in-initramfs approach the best so far... > > Why kernel does not handle it right without 'rootfstype=ramfs'?
The kernel defaults to tmpfs here, limiting the available space to half (or third? don't remember) of your system ram. > > Could we tweak kernel config so that it works with large initramfs? We could create a patch that defaults to ramfs instead. The real problem imho is that the initrd is too big though. > > squashfs-in-initramfs approach would need some kiwi work, I guess? Yes, it would. But then we would also get around the 64k page size problems. On 64k page size systems, every file in tmpfs occupies at least 64k. So if you have a few thousand like we do in the kiwi initrd, you waste a few 100 mbs just because of that. Alex > > > Guillaume > > >> >> >> Alex >> >>> Any kernel config update related to initramfs? >>> >>> >>> Guillaume >>> >>> -- >>> To unsubscribe, e-mail: [email protected] >>> To contact the owner, e-mail: [email protected] > > -- > To unsubscribe, e-mail: [email protected] > To contact the owner, e-mail: [email protected] > -- To unsubscribe, e-mail: [email protected] To contact the owner, e-mail: [email protected]
