On Wed, Jun 15, 2022 at 03:38:45PM +0200, Jan Kara wrote:
On Wed 15-06-22 13:00:22, Jan Kara wrote:
On Tue 14-06-22 12:00:22, Linus Torvalds wrote:
On Tue, Jun 14, 2022 at 11:51 AM Linus Torvalds torvalds@linux-foundation.org wrote:
Or just make sure that noop_backing_dev_info is fully initialized before it's used.
I don't see any real reason why that
err = bdi_init(&noop_backing_dev_info);couldn't just be done very early. Maybe as the first call in driver_init(), before the whole devtmpfs_init() etc.
I've checked the dependencies and cgroups (which are the only non-trivial dependency besides per-CPU infrastructure) are initialized early enough so it should work fine. So let's try that.
Attached patch boots for me. Guys, who was able to reproduce the failure: Can you please confirm this patch fixes your problem?
It does for me.
Honza-- Jan Kara jack@suse.com SUSE Labs, CR
From 8f998b182be7563fc92aa8914cc7d21f75a3c20e Mon Sep 17 00:00:00 2001 From: Jan Kara jack@suse.cz Date: Wed, 15 Jun 2022 15:22:29 +0200 Subject: [PATCH] init: Initialized noop_backing_dev_info early
noop_backing_dev_info is used by superblocks of various pseudofilesystems such as kdevtmpfs. Initialize it before the filesystems get mounted.
Signed-off-by: Jan Kara jack@suse.cz
Tested-by: Guenter Roeck linux@roeck-us.net
Guenter