Re: Review of Thomas's >2GB ext2fs proposal

[Neal H. Walfield]

At Tue, 17 Aug 2004 12:50:25 +0300,
Ognyan Kulev wrote:
> 
> Neal H. Walfield wrote:
> > If that is the case, then I think we should drop the whole reference
> > counting system and have the accessor functions just do a vm_map and
> > vm_unmap as required.
> 
> Surely you don't want ext2_getblk to vm_map/vm_unmap 1024 times the same 
> single indirect block for a 4M file when this file is read. 
> (ext2_getblk is called for each block when file is read.)
> 
> And what about two threads that access the same block?  They should work 
> with the same page (we don't want to merge changes made in two pages, 
> right?), and whoever fininishes first must not vm_unmap the page, 
> because the other works with the same page too.

Nope.  You accidently cut one too many lines of context.  Namely,
Thomas's statement that

    Mappings carry no overhead, and are quick and easy to set up and
    tear down.

Neal