re: patch to support >16 bit g/uids on ext2fs

[matthew green <mrg%eterna.com.au@localhost>]

   >    > i recently had need to recover some data from an ext3 that used
   >    > uids greater than 16 bits, and discovered they were incorrectly
   >    > represented as only 16 bits.
   >    > 
   >    > after a bit of searching, i noticed that linux has a per-mount
   >    > option to disable 32 bit but otherwise has it there by default.
   >    > 
   >    > since we don't have a similar mount option, i've made it non
   >    > optional in our kernel (simple ways to add it welcome.)
   >    
   >    I wonder if the features bits in struct ext2fs have
   >    some bits for the uid/gid extension or not, as LARGEFILE etc.
   > 
   > i could be wrong, but the code i read indicated it was a
   > mount option not a file system option.
   
   With a quick glance at Linux ext2_fs.h, you are right.
   Umm, what an awful design...
   (what happens if old FS has garbages in unused [ug]id_high!?)

the worst case is that each file will be owned by a "random" uid,
which is fixable (just mask out uids with find + chmod).  the
current lossage is not quite so fixable.
   
   At least should we check e2fs_rev for sanity?

for what values?  >REV0?


.mrg.