On Sun, Jul 26, 2020 at 05:21:13PM +0100, Al Viro wrote:

On Sun, Jul 26, 2020 at 05:52:04PM +0200, Christoph Hellwig wrote:

quoted

On Sun, Jul 26, 2020 at 08:49:28AM -0700, Linus Torvalds wrote:

quoted

On Sun, Jul 26, 2020 at 12:14 AM Christoph Hellwig [off-list ref] wrote:

quoted

Hi Al and Linus,

currently a lot of the file system calls in the early in code (and the
devtmpfs kthread) rely on the implicit set_fs(KERNEL_DS) during boot.
This is one of the few last remaining places we need to deal with to kill
off set_fs entirely, so this series adds new helpers that take kernel
pointers.  These helpers are in init/ and marked __init and thus will
be discarded after bootup.  A few also need to be duplicated in devtmpfs,
though unfortunately.

I see nothing objectionable here.

The only bikeshed comment I have is that I think the "for_init.c" name
is ugly and pointless - I think you could just call it "fs/init.c" and
it's both simpler and more straightforward. It _is_ init code, it's
not "for" init.

That was Al's suggestion.  I personally don't care, so if between the
two of you, you can come up with a preferred choice I'll switch to it.

I can live with either variant; the only problem with fs/init.c is that
such name would imply the init code _of_ VFS, rather than VFS helpers for
init.

Anyway, the series looks generally sane; if no other objections are raised,
I'm adding it to vfs.git#for-next

Thanks!

Note that this is based on top of my init-user-pointers branch, so you'll
need to pull that in as.

Btw, care to take a look at 

http://git.infradead.org/users/hch/misc.git/shortlog/refs/heads/kernel_readwrite

it has been in linux-next for 2 1/2 weeks, and the only interesting
thing found was that btrfs didn't wire up iter_splice_write, which has
already been fixed in mainline.