[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: ARE: Changing journaling mode on root / loop-aes

On Tue, Jul 09, 2002 at 09:04:25AM +0300, Ivan Ivanov wrote:
> This is the most complicated way to boot a linux box that I ever heared.
> Why dont't you simple recompile your kernel with ext3 support and boot
> your box using lilo.
Well, it may be "complicated," but it is extremely flexible, as it allows
things like MD, LVM, crypto, etc., policy decisions to be made out in userland,
not hardwired into the kernel.

Al Viro and H. Peter Anvin are working on replacing the initrd
mechanism with initramfs.  [You can find the draft proposal in the lkml
archives.] Basically, initramfs system will create a
(dynamically-sized) ramfs root filesystem, into which it extracts a
cpio archive, and then proceeds similarly to initrd.

Running the initial root filesystem out of RAM allows hardware probing,
handling of quirks, blacklists, etc., to be done in userland, e.g.,
dmidecode.  This has the very desirable property of partially
decoupling kernel upgrades from motherboard and peripheral upgrades.

Mr. Tener offers a good example of where this model is useful -- security.
Running entirely out of RAM makes it rather trivial to extract a
passphrase from the user in order to unlock disk partitions, make a
secure network connection, etc.  I, for one, hope to see this as
a standard install option, especially for laptops.


   Bill Rugolsky

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]