Richard wrote:
>       I can only say that, as an occasional producer of Live CDs for my own
> consumption, that unionfs has never caused me any trouble whatsoever,
> and nowadays seems to be quite mature.

As I said, this is quite possible. How many CPU cores does your computer 
have? (intention: figure out whether you tested for SMP-related bugs, too)

> As a matter of fact, the
> linux-live script from the 6.x series now uses 'aufs' (another union
> filesystem (...)), a rewrite of unionfs with same concepts but different
> interface. I am not adocating anything here; just stating the facts.
>   

BTW, Knoppix does the same, for the same reasons as the LFS LiveCD (bugs 
in the original unionfs). I have not hit any bugs in aufs in Knoppix, 
thus it may indeed be a working solution. Also, the "filesystem union" 
semantics provided by unionfs and aufs (as opposed to "copy-on-write 
block device" provided by device mapper) has the following benefits:

 * the amount of free space in the unioned filesystem is directly 
related to the amount of free RAM for tmpfs (as opposed to a fixed value 
determined by the contents of root.ext2 file)

 * if one produces and then deletes a large file, the net result is "no 
change", i.e. the memory available to applications is the same as before 
(in contrast, this operation on a device-mapper based CD marks a lot of 
sectors as dirty, and, since there is no "dirty -> clean" transition, 
the RAM that stores their modified contents can only be reused for 
storing another file, not for memory allocation by applications)

Anyway, unionfs has been submitted for merging (and actually merged into 
the -mm kernel), thus, it is unclear which one of unionfs and aufs will 
survive in the long term.

-- 
Alexander E. Patrakov
Please remove [EMAIL PROTECTED] from your address book - this address will 
become invalid soon, after I quit my job at USU.

-- 
http://linuxfromscratch.org/mailman/listinfo/livecd
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page

Reply via email to