mirror of https://lore.kernel.org/lkml/
 help / color / mirror / Atom feed
From: Daeho Jeong <daeho43@gmail.com>
To: linux-kernel@vger.kernel.org,
	linux-f2fs-devel@lists.sourceforge.net, kernel-team@android.com
Cc: Daeho Jeong <daehojeong@google.com>, stable@vger.kernel.org
Subject: [PATCH] f2fs: don't run out of free sections in the dry run of resize_fs
Date: Mon, 16 Nov 2026 11:35:42 -0800	[thread overview]
Message-ID: <20261116193542.111825-1-daeho43@gmail.com> (raw)

From: Daeho Jeong <daehojeong@google.com>

f2fs_resize_fs() migrates the blocks in the range to be shrunk in a dry
run before freeze_super(). Checkpoint is blocked there to protect
MAIN_SECS, so the space freed in the meantime can't be reused, while GC
and other writers keep using free sections in the remaining area. On a
busy filesystem, free sections can run out and we hit the below.

 kernel BUG at fs/f2fs/segment.c:3020!
 Call trace:
  new_curseg+0x9dc/0xa50
  f2fs_allocate_data_block+0xa2c/0xcb8
  do_write_block+0xd4/0x2b4
  f2fs_outplace_write_data+0x78/0xcc
  f2fs_do_write_data_page+0x564/0x69c
  do_garbage_collect+0x1668/0x21ec
  f2fs_gc_range+0x274/0x41c
  free_segment_range+0x290/0x33c
  f2fs_resize_fs+0x200/0x530

Without CONFIG_F2FS_CHECK_FS, checkpoint is stopped by
STOP_CP_REASON_NO_SEGMENT instead, and writes fail with -EIO.

Stop the dry run before free sections run out. To migrate as much as
possible before freeze_super(), write a checkpoint before the dry run,
and repeat them as long as the valid blocks in the range decrease. The
rest, if any, is migrated after freeze_super() as before, where no one
else allocates blocks.

Shrinking a 4GB filesystem down to 4MB of free space with two writers
running hit the BUG in 3 of 4 runs on a 4KB page kernel, and in none of
4 runs with this patch.

Fixes: b4b10061ef98 ("f2fs: refactor resize_fs to avoid meta updates in progress")
Cc: stable@vger.kernel.org
Signed-off-by: Daeho Jeong <daehojeong@google.com>
---
This applies on top of "f2fs: reserve space to be shrunk during resize_fs"
(v3), which changes the same function.
https://lore.kernel.org/linux-f2fs-devel/20261114011807.3811774-1-daeho43@gmail.com/

 fs/f2fs/gc.c | 51 ++++++++++++++++++++++++++++++++++++++++++++++-----
 1 file changed, 46 insertions(+), 5 deletions(-)

diff --git a/fs/f2fs/gc.c b/fs/f2fs/gc.c
index 7609c23ec27b..2d2cc3417c36 100644
--- a/fs/f2fs/gc.c
+++ b/fs/f2fs/gc.c
@@ -2241,6 +2241,14 @@ int f2fs_gc_range(struct f2fs_sb_info *sbi,
 		if (is_cursec(sbi, GET_SEC_FROM_SEG(sbi, segno)))
 			goto next;
 
+		/*
+		 * Checkpoint is blocked in the dry run of resize, so the space
+		 * freed in the meantime can't be reused. Stop before free
+		 * sections run out, and let the caller migrate the rest later.
+		 */
+		if (dry_run && !lock && has_not_enough_free_secs(sbi, 0, 0))
+			break;
+
 		do_garbage_collect(sbi, segno, &gc_list, FG_GC, true, false);
 		put_gc_inode(&gc_list);
 
@@ -2411,15 +2419,30 @@ static void update_fs_metadata(struct f2fs_sb_info *sbi, int secs)
 	}
 }
 
+/* Count the valid blocks in the last @secs sections. */
+static block_t f2fs_valid_blocks_in_tail(struct f2fs_sb_info *sbi,
+					 unsigned int secs)
+{
+	unsigned int secno;
+	block_t valid = 0;
+
+	for (secno = MAIN_SECS(sbi) - secs; secno < MAIN_SECS(sbi); secno++)
+		valid += get_valid_blocks(sbi, GET_SEG_FROM_SEC(sbi, secno),
+					  true);
+	return valid;
+}
+
 int f2fs_resize_fs(struct file *filp, __u64 block_count)
 {
 	struct f2fs_sb_info *sbi = F2FS_I_SB(file_inode(filp));
 	__u64 old_block_count, shrunk_blocks;
 	struct cp_control cpc = { CP_RESIZE, 0, 0, 0 };
+	struct cp_control sync_cpc = { __get_cp_reason(sbi), 0, 0, 0 };
 	struct f2fs_lock_context lc;
 	struct f2fs_lock_context glc;
 	struct f2fs_lock_context clc;
 	unsigned int secs, old_tail;
+	block_t valid, prev_valid;
 	bool reserved = false;
 	int err = 0;
 	__u32 rem;
@@ -2468,9 +2491,6 @@ int f2fs_resize_fs(struct file *filp, __u64 block_count)
 		goto out_drop_write;
 	}
 
-	/* stop CP to protect MAIN_SEC in free_segment_range */
-	f2fs_lock_op(sbi, &lc);
-
 	spin_lock(&sbi->stat_lock);
 	if (shrunk_blocks + valid_user_blocks(sbi) >
 			__get_available_block_count(sbi, false)) {
@@ -2488,10 +2508,31 @@ int f2fs_resize_fs(struct file *filp, __u64 block_count)
 	if (err)
 		goto out_unlock;
 
-	err = free_segment_range(sbi, secs, true);
+	/*
+	 * Migrate the blocks in the range to be shrunk before freeze_super().
+	 * Checkpoint is blocked in the dry run, so the dry run stops before
+	 * free sections run out. A checkpoint makes the space freed so far
+	 * reusable: write one first, and repeat as long as the range gets
+	 * emptier. The rest is migrated after freeze_super().
+	 */
+	valid = f2fs_valid_blocks_in_tail(sbi, secs);
+	do {
+		prev_valid = valid;
+
+		stat_inc_cp_call_count(sbi, TOTAL_CALL);
+		err = f2fs_write_checkpoint(sbi, &sync_cpc);
+		if (err)
+			break;
+
+		/* stop CP to protect MAIN_SEC in free_segment_range */
+		f2fs_lock_op(sbi, &lc);
+		err = free_segment_range(sbi, secs, true);
+		f2fs_unlock_op(sbi, &lc);
+
+		valid = f2fs_valid_blocks_in_tail(sbi, secs);
+	} while (!err && valid && valid < prev_valid);
 
 out_unlock:
-	f2fs_unlock_op(sbi, &lc);
 	f2fs_up_write_trace(&sbi->gc_lock, &glc);
 out_drop_write:
 	mnt_drop_write_file(filp);
-- 
2.56.0.679.gc527c65306-goog


                 reply	other threads:[~2026-11-16 19:35 UTC|newest]

Thread overview: [no followups] expand[flat|nested]  mbox.gz  Atom feed

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20261116193542.111825-1-daeho43@gmail.com \
    --to=daeho43@gmail.com \
    --cc=daehojeong@google.com \
    --cc=kernel-team@android.com \
    --cc=linux-f2fs-devel@lists.sourceforge.net \
    --cc=linux-kernel@vger.kernel.org \
    --cc=stable@vger.kernel.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox

all inboxes | Powered by JetHome®