fsck shortcomings

[Scott Long]

Michael Nottebrock wrote:
> I recently had a filesystem go bad on me in such a way that it was 
> recognized way bigger than it actually was, causing fsck to fail while 
> trying to allocate and equally astronomic amount of memory (and my 
> machine already had 1 Gig of mem + 2 Gig swap available).
> I just newfs'd and I'm now in the process of restoring data, however, I 
> googled a bit on this and it seems that this kind of fs corruption is 
> occurring quite often, in particular due to power failures.

Yes, very troubling.  You said that the alternate superblocks didn't
help?

> 
> Is there really no way that fsck could be made smarter about dealing 
> with seemingly huge filesystems? Also, what kind of memory would be 
> required to fsck a _real_ 11TB filesystem?
> 


More than you can address in 32 bits.  Reducing the RAM footprint of
fsck_ufs is something that desperately needs to be done, especially
since it's now easy to trash crashdumps that are saved in swap because
fsck is consuming so much memory.

Scott