On Wed, Mar 20, 2019 at 6:09 PM Dave Young dyoung@redhat.com wrote:
I remember we have a report that kdump service runs earlier than some udev modprobe, so that this can cause rebuild initrd for each time in that case. And we talked about revert the module checking code.
But extra_modules is different with the check_kmodules case, am I missing something?
Yes, I put the extra_modules checking code in check_kmodules because it fits well with the function name. (check kernel modules) And we already have a old_modules list in initramfs now, so it should be good to check if there is any change of any module that got loaded by the previous kernel. This may cover some case user updated the modules but initramfs is not rebuilt. Just a bold suggestion, I can remove this part if there is any risk.
And as currently I still haven't reproduced the initramfs being rebuilt unexpectedly problem, I'm not sure if there is any better way to solve the problem, I think revert the commit is the button line only if there is no other way to fix it...