On Tue, Dec 08, 2020 at 12:25:55PM -0800, Linus Torvalds wrote:

On Tue, Dec 8, 2020 at 11:49 AM Al Viro [off-list ref] wrote:

quoted

Said that, it does appear to survive all beating, and it does fix
a regression introduced in this cycle, so, provided that amount of
comments in there is OK with you...

Ok, considering Greg's note, I've pulled it. It's early in the last
week, if something comes up we can still fix it.

That said, considering that I think the only use-case was that odd
/proc splice use, and the really special WSL2 thing, and both of those
are verified, it does sound safe to pull.

Famous last words...

Al, since you're around, would you mind looking at the two
DCACHE_DONTCACHE patches too? Honestly, since they seem to be an issue
only for DAX, and only for DAX policy changes, I don't consider them
critical for 5.10, but they've been around for a while now.

Umm...  I've got
fs: Handle I_DONTCACHE in iput_final() instead of generic_drop_inode()
and
fs: Kill DCACHE_DONTCACHE dentry even if DCACHE_REFERENCED is set
in "to apply" pile; if that's what you are talking about, I don't
think they are anywhere critical enough for 5.10-final, but I might
be missing something...

Al, still buried under piles of email ;-/