Re: [datapoint] Re: 2.5.7 rm -r in tmpfs problem

[Mike Galbraith]

On Sat, 23 Mar 2002, Alexander Viro wrote:

> There are known problems with rm(1) on ramfs/tmpfs/etc. - the thing assumes
> that offsets in directory remain stable after unlink(2), but locking changes
> didn't affect.

Ok, I finally found a wee bit of clue.  I understand the assumption,
and in virgin 2.5.4, that assumption is enforced... somefsckingwhere ;-)

However, the locking changes do affect it in some magical way, and do
cause (well, trigger anyway) the breakage.

In 2.5.4, I creats a directory with 170 (two more than will fit in one call)
entries, and rm -r it.  Over all three dcache_readdir() calls, (one for the
first 168 entries (+.+..), one for the last two entries + seek, and one to
determine that we're empty), the dentry count remains stable at 170.  If I
apply the locking changes, POOF, my dentries disappear exactly as they do in
kernels >= 2.5.5-pre1.

64 Kibble-$$ question:

Why are my dentries going awol with the locking changes applied?

2.5.4.virgin
enter pos=0 count=170
seek pos=0 count=170
exit pos=170 count=170
enter pos=170 count=170
seek pos=0 count=170
exit pos=172 count=170
enter pos=172 count=170
exit EOF

2.5.4.locking
enter pos=0 count=170
seek pos=0 count=170
exit pos=170 count=170
enter pos=170 count=2
exit EOF

	-Mike