Re: [PATCH 3/3] mm: require ->set_page_dirty to be explicitly wire up

[Christoph Hellwig]

On Fri, Jun 18, 2021 at 10:23:55PM +0300, Dmitry Osipenko wrote:
> >  21 files changed, 29 insertions(+), 14 deletions(-)
> 
> The ecryptfs is now crashing with NULL deference, please fix.

Which means it crashed the same before on configs without CONFIG_BLOCK.

Tyler, can you look at what ecryptfs should do for ->set_page_dirty?
Currently it implicitly gets __set_page_dirty_buffers for kernels
with COFIG_BLOCK set, but given that this function looks at buffer_heads
that ecryptfs doesn't use it can't really be the right choice.
__set_page_dirty_nobuffers will probably work, but I'd love to see an
audit of the page dirtying and writeback for ecryptfs while we're at it.