Re: [git pull] vfs.git spurious ENOTDIR fix

[Hugh Dickins <hughd@google.com>]

On Sat, 1 Aug 2015, Linus Torvalds wrote:
> 
> Well, I'd not be against continuing cleanups for 4.3... Well, as long
> as we can make sure 4.2 is solid first, of course. I'd still like to
> have Hugh verify that the current -git tree works for his load, but
> obviously that wasn't easily reproducible, so that will presumably
> take a few days. Dominique seems to at least not see it any more with
> that patch.

I've had Al's patch under load since this morning, and it's going fine.
But I've not actually managed to reproduce the issue for several days
now, however bogus a "fix" I've been testing.  So certainly don't wait
to hear from me: but of course I'll keep on with the loads, and shout
if it does go wrong again.

(I don't actually understand why the clearing of DCACHE_ENTRY_TYPE in
dentry_iput() is not of continuing concern; but don't worry, there's
plenty I don't understand - so long as you're both satisfied that
it's not a concern, no need to persuade me.)

Do we have any idea why a bug introduced in v3.13 should only now
stand out, both for Dominique and for me?  Has the RCU lookup somehow
become much more effective recently?  (I don't think symlinks were a
big deal for either of us, though the kernel build does use some).

Hugh