Dear Maintainer,

When boot with changing kernel version, force starting FSCK for F2FS partitions.
So, spend a lot of time (some minutes or longer) at boot time.

This is upstream's feature issue still not fixed.
See https://bbs.archlinux.org/viewtopic.php?id=245702 .

Regards,
Ohta

Hi Chao, Jaeguk,

Could you take a look at this complaint and let me know if I should
close the bug as Working As Intended or not?

The concern seems to be that for desktop distros (which I know was not
f2fs's original target), some users update their kernels frequently,
and for them, they are finding the overhead running fsck.f2fs on every
kernel upgrade to be overly burdensome.

Thanks,

					- Ted

We can bypass kernel check by adding an option "--no-kernel-check".
Like this?

From 0fd26684d7e6d279a6a5553c7b82856cb6068ec3 Mon Sep 17 00:00:00 2001
From: Jaegeuk Kim <jaegeuk@kernel.org>
Date: Mon, 14 Oct 2019 10:10:31 -0700
Subject: [PATCH] fsck.f2fs: add --no-kernel-check to bypass kernel version
 diff

Given this option, fsck.f2fs does not run fsck forcefully, even if kernel
is updated.

Signed-off-by: Jaegeuk Kim <jaegeuk@kernel.org>
---
 fsck/main.c       | 6 ++++++
 fsck/mount.c      | 3 ++-
 include/f2fs_fs.h | 1 +
 3 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/fsck/main.c b/fsck/main.c
index 8c62a14..a402ffb 100644
--- a/fsck/main.c
+++ b/fsck/main.c
@@ -66,6 +66,7 @@ void fsck_usage()
 	MSG(0, "  -y fix all the time\n");
 	MSG(0, "  -V print the version number and exit\n");
 	MSG(0, "  --dry-run do not really fix corruptions\n");
+	MSG(0, "  --no-kernel-check skips detecting kernel change\n");
 	exit(1);
 }

@@ -192,6 +193,7 @@ void f2fs_parse_options(int argc, char *argv[])
 		char *token;
 		struct option long_opt[] = {
 			{"dry-run", no_argument, 0, 1},
+			{"no-kernel-check", no_argument, 0, 2},
 			{0, 0, 0, 0}
 		};

@@ -203,6 +205,10 @@ void f2fs_parse_options(int argc, char *argv[])
 				c.dry_run = 1;
 				MSG(0, "Info: Dry run\n");
 				break;
+			case 2:
+				c.no_kernel_check = 1;
+				MSG(0, "Info: No Kernel Check\n");
+				break;
 			case 'a':
 				c.auto_fix = 1;
 				MSG(0, "Info: Fix the reported corruption.\n");
diff --git a/fsck/mount.c b/fsck/mount.c
index 2d01c2f..98227e7 100644
--- a/fsck/mount.c
+++ b/fsck/mount.c
@@ -883,7 +883,8 @@ int validate_super_block(struct f2fs_sb_info *sbi, enum SB_ADDR sb_addr)
 		MSG(0, "Info: MKFS version\n  \"%s\"\n", c.init_version);
 		MSG(0, "Info: FSCK version\n  from \"%s\"\n    to \"%s\"\n",
 					c.sb_version, c.version);
-		if (memcmp(c.sb_version, c.version, VERSION_LEN)) {
+		if (!c.no_kernel_check &&
+				memcmp(c.sb_version, c.version, VERSION_LEN)) {
 			memcpy(sbi->raw_super->version,
 						c.version, VERSION_LEN);
 			update_superblock(sbi->raw_super, SB_MASK(sb_addr));
diff --git a/include/f2fs_fs.h b/include/f2fs_fs.h
index a722be4..eb178ea 100644
--- a/include/f2fs_fs.h
+++ b/include/f2fs_fs.h
@@ -372,6 +372,7 @@ struct f2fs_configuration {
 	int func;
 	void *private;
 	int dry_run;
+	int no_kernel_check;
 	int fix_on;
 	int force;
 	int defset;

The challenge is that for desktop and server installations of Linux,
in general some program like /sbin/fsck parses /etc/fstab, and then
runs the appropriate file-system specific fsck driver, e.g.,
/sbin/fsck.ext4, or /sbin/fsck.f2fs, etc. for each particular file
system.  On more modern-day systems systemd will run the
/sbin/fsck.<FSTYPE> for each file system, but the issue remains the
same: there in general isn't a good way to configure the OS to pass in
file system-specific options, such as --no-kernel-check, to the
/sbin/fsck.<FSTYPE> program.

The solution that I have for this is to create a config file.  See
"man e2fsck.conf" for the documentation for it.  Code to parse this
WIN.INI-style "profile" is quite small; it was originally written for
Kerberos, and then imported into e2fsprogs.  There is a single C
file[1] and a single header file[2], licensed under a MIT-style
permissive free software license, so feel free to take and use it if
it's helpful.

[1] https://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git/tree/lib/support/profile.c
[2] https://git.kernel.org/pub/scm/fs/ext2/e2fsprogs.git/tree/lib/support/profile.h

Cheers,

						- Ted

P.S.  In case it isn't obvious, this is the same library code used to
parse /etc/mke2fs.conf.  See the man page for mke2fs.conf to see why
we find it very useful to have a configuration file for mkfs.ext4.

P.P.S.  I also notice that f2fstools seem to bump the major version
number of its shared libraries at every single release.  There are
ways this can be avoided; in fact, I haven't needed to bump the shared
libraries for e2fsprogs in over a decade.

Because new shared libraries require new debian packages, my upload of
Debian package for f2fstools version 1.12.0 has been hung up in the
NEW queue[1] for manual Debian ftpmaster review for over two months.
So there some advantage in trying to avoid needing to bump the shared
library version unnecessarily (although it does require more care to
consider API/ABI backwards compatibility in your design and
development).

[1] https://ftp-master.debian.org/new.html