devtmpfs: actually reclaim some init memory
authorRasmus Villemoes <linux@rasmusvillemoes.dk>
Fri, 12 Mar 2021 10:30:27 +0000 (11:30 +0100)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Tue, 23 Mar 2021 13:57:35 +0000 (14:57 +0100)
commit01085e24ff0ae775e7407a6e40c2156a724ae884
treed00c705efc4360b2e54b23aa2c263f901247819f
parent38f087de8947700d3b06d3d1594490e0f611c5d1
devtmpfs: actually reclaim some init memory

Currently gcc seems to inline devtmpfs_setup() into devtmpfsd(), so
its memory footprint isn't reclaimed as intended. Mark it noinline to
make sure it gets put in .init.text.

While here, setup_done can also be put in .init.data: After complete()
releases the internal spinlock, the completion object is never touched
again by that thread, and the waiting thread doesn't proceed until it
observes ->done while holding that spinlock.

This is now the same pattern as for kthreadd_done in init/main.c:
complete() is done in a __ref function, while the corresponding
wait_for_completion() is in an __init function.

Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: Rasmus Villemoes <linux@rasmusvillemoes.dk>
Link: https://lore.kernel.org/r/20210312103027.2701413-2-linux@rasmusvillemoes.dk
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
drivers/base/devtmpfs.c