On Tue, Sep 03, 2024 at 07:13:42AM GMT, Christoph Hellwig wrote:

On Mon, Sep 02, 2024 at 02:52:52PM -0700, Andrew Morton wrote:

quoted

It would be helpful to summarize your concerns.

And that'd better be a really good argument for a change that was
pushed directly to Linus bypassing the maintainer after multiple
reviewers pointed out it was broken.  This series simply undoes the
damage done by that, while also keeping the code dependend on it
working.

Well, to be blunt, I thought the "we don't want the allocator to even
know if we're in a non-sleepable context" argument was too crazy to have
real support, and moving towards PF_MEMALLOC flags is something we've
been talking about quite a bit going back years.

Little did I know the minefield I was walking into...

But the disccussion seems to finally be cooling off and going in a more
productive direction.