All of lore.kernel.org
 help / color / mirror / Atom feed
From: Jan Kara <jack@suse.cz>
To: Ted Tso <tytso@mit.edu>
Cc: <linux-ext4@vger.kernel.org>, Jan Kara <jack@suse.cz>
Subject: [PATCH 11/12] ext4: Save error info to sb through journal if available
Date: Fri, 27 Nov 2020 12:34:04 +0100	[thread overview]
Message-ID: <20201127113405.26867-12-jack@suse.cz> (raw)
In-Reply-To: <20201127113405.26867-1-jack@suse.cz>

If journalling is still working at the moment we get to writing error
information to the superblock we cannot write directly to the superblock
as such write could race with journalled update of the superblock and
cause journal checksum failures, writing inconsistent information to the
journal or other problems. We cannot journal the superblock directly
from the error handling functions as we are running in uncertain context
and could deadlock so just punt journalled superblock update to a
workqueue.

Signed-off-by: Jan Kara <jack@suse.cz>
---
 fs/ext4/super.c | 97 ++++++++++++++++++++++++++++++++++++++++++---------------
 1 file changed, 72 insertions(+), 25 deletions(-)

diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index fb1d61ed7dff..51670258ef84 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -65,6 +65,7 @@ static struct ratelimit_state ext4_mount_msg_ratelimit;
 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
 			     unsigned long journal_devnum);
 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
+static void ext4_update_super(struct super_block *sb);
 static int ext4_commit_super(struct super_block *sb);
 static int ext4_mark_recovery_complete(struct super_block *sb,
 					struct ext4_super_block *es);
@@ -586,9 +587,9 @@ static int ext4_errno_to_code(int errno)
 	return EXT4_ERR_UNKNOWN;
 }
 
-static void __save_error_info(struct super_block *sb, int error,
-			      __u32 ino, __u64 block,
-			      const char *func, unsigned int line)
+static void save_error_info(struct super_block *sb, int error,
+			    __u32 ino, __u64 block,
+			    const char *func, unsigned int line)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
 
@@ -615,15 +616,6 @@ static void __save_error_info(struct super_block *sb, int error,
 	spin_unlock(&sbi->s_error_lock);
 }
 
-static void save_error_info(struct super_block *sb, int error,
-			    __u32 ino, __u64 block,
-			    const char *func, unsigned int line)
-{
-	__save_error_info(sb, error, ino, block, func, line);
-	if (!bdev_read_only(sb->s_bdev))
-		ext4_commit_super(sb);
-}
-
 /* Deal with the reporting of failure conditions on a filesystem such as
  * inconsistencies detected or read IO failures.
  *
@@ -649,20 +641,35 @@ static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
 			      const char *func, unsigned int line)
 {
 	journal_t *journal = EXT4_SB(sb)->s_journal;
+	bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
 
 	EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
 	if (test_opt(sb, WARN_ON_ERROR))
 		WARN_ON_ONCE(1);
 
-	if (!bdev_read_only(sb->s_bdev))
+	if (!continue_fs && !sb_rdonly(sb)) {
+		ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
+		if (journal)
+			jbd2_journal_abort(journal, -EIO);
+	}
+
+	if (!bdev_read_only(sb->s_bdev)) {
 		save_error_info(sb, error, ino, block, func, line);
+		/*
+		 * In case the fs should keep running, we need to writeout
+		 * superblock through the journal. Due to lock ordering
+		 * constraints, it may not be safe to do it right here so we
+		 * defer superblock flushing to a workqueue.
+		 */
+		if (continue_fs)
+			schedule_work(&EXT4_SB(sb)->s_error_work);
+		else
+			ext4_commit_super(sb);
+	}
 
-	if (sb_rdonly(sb) || (!force_ro && test_opt(sb, ERRORS_CONT)))
+	if (sb_rdonly(sb) || continue_fs)
 		return;
 
-	ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
-	if (journal)
-		jbd2_journal_abort(journal, -EIO);
 	/*
 	 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
 	 * could panic during 'reboot -f' as the underlying device got already
@@ -685,7 +692,38 @@ static void flush_stashed_error_work(struct work_struct *work)
 {
 	struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
 						s_error_work);
+	journal_t *journal = sbi->s_journal;
+	handle_t *handle;
 
+	/*
+	 * If the journal is still running, we have to write out superblock
+	 * through the journal to avoid collisions of other journalled sb
+	 * updates.
+	 *
+	 * We use directly jbd2 functions here to avoid recursing back into
+	 * ext4 error handling code during handling of previous errors.
+	 */
+	if (!sb_rdonly(sbi->s_sb) && journal) {
+		handle = jbd2_journal_start(journal, 1);
+		if (IS_ERR(handle))
+			goto write_directly;
+		if (jbd2_journal_get_write_access(handle, sbi->s_sbh)) {
+			jbd2_journal_stop(handle);
+			goto write_directly;
+		}
+		ext4_update_super(sbi->s_sb);
+		if (jbd2_journal_dirty_metadata(handle, sbi->s_sbh)) {
+			jbd2_journal_stop(handle);
+			goto write_directly;
+		}
+		jbd2_journal_stop(handle);
+		return;
+	}
+write_directly:
+	/*
+	 * Write through journal failed. Write sb directly to get error info
+	 * out and hope for the best.
+	 */
 	ext4_commit_super(sbi->s_sb);
 }
 
@@ -944,7 +982,7 @@ __acquires(bitlock)
 		if (test_opt(sb, WARN_ON_ERROR))
 			WARN_ON_ONCE(1);
 		EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
-		__save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
+		save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
 		schedule_work(&EXT4_SB(sb)->s_error_work);
 		return;
 	}
@@ -5425,15 +5463,12 @@ static int ext4_load_journal(struct super_block *sb,
 	return err;
 }
 
-static int ext4_commit_super(struct super_block *sb)
+/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
+static void ext4_update_super(struct super_block *sb)
 {
 	struct ext4_sb_info *sbi = EXT4_SB(sb);
 	struct ext4_super_block *es = EXT4_SB(sb)->s_es;
 	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
-	int error = 0;
-
-	if (!sbh || block_device_ejected(sb))
-		return error;
 
 	lock_buffer(sbh);
 	/*
@@ -5501,8 +5536,20 @@ static int ext4_commit_super(struct super_block *sb)
 	}
 	spin_unlock(&sbi->s_error_lock);
 
-	BUFFER_TRACE(sbh, "marking dirty");
 	ext4_superblock_csum_set(sb);
+	unlock_buffer(sbh);
+}
+
+static int ext4_commit_super(struct super_block *sb)
+{
+	struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
+	int error = 0;
+
+	if (!sbh || block_device_ejected(sb))
+		return error;
+
+	ext4_update_super(sb);
+
 	if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
 		/*
 		 * Oh, dear.  A previous attempt to write the
@@ -5517,8 +5564,8 @@ static int ext4_commit_super(struct super_block *sb)
 		clear_buffer_write_io_error(sbh);
 		set_buffer_uptodate(sbh);
 	}
+	BUFFER_TRACE(sbh, "marking dirty");
 	mark_buffer_dirty(sbh);
-	unlock_buffer(sbh);
 	error = __sync_dirty_buffer(sbh,
 		REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
 	if (buffer_write_io_error(sbh)) {
-- 
2.16.4


  parent reply	other threads:[~2020-11-27 11:34 UTC|newest]

Thread overview: 34+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-11-27 11:33 [PATCH 00/12] ext4: Various fixes of ext4 handling of fs errors Jan Kara
2020-11-27 11:33 ` [PATCH 01/12] ext4: Don't remount read-only with errors=continue on reboot Jan Kara
2020-11-29 21:33   ` Andreas Dilger
2020-12-16  4:59     ` Theodore Y. Ts'o
2020-11-27 11:33 ` [PATCH 02/12] ext4: Remove redundant sb checksum recomputation Jan Kara
2020-11-29 22:11   ` Andreas Dilger
2020-11-30 10:59     ` Jan Kara
2020-12-16  5:01   ` Theodore Y. Ts'o
2020-11-27 11:33 ` [PATCH 03/12] ext4: Standardize error message in ext4_protect_reserved_inode() Jan Kara
2020-11-29 21:39   ` Andreas Dilger
2020-12-16  5:04   ` Theodore Y. Ts'o
2020-11-27 11:33 ` [PATCH 04/12] ext4: Make ext4_abort() use __ext4_error() Jan Kara
2020-11-29 22:12   ` Andreas Dilger
2020-12-16  5:07   ` Theodore Y. Ts'o
2020-11-27 11:33 ` [PATCH 05/12] ext4: Move functions in super.c Jan Kara
2020-11-29 22:13   ` Andreas Dilger
2020-12-16  5:09   ` Theodore Y. Ts'o
2020-11-27 11:33 ` [PATCH 06/12] ext4: Simplify ext4 error translation Jan Kara
2020-11-29 21:57   ` Andreas Dilger
2020-12-16  5:11   ` Theodore Y. Ts'o
2020-11-27 11:34 ` [PATCH 07/12] ext4: Defer saving error info from atomic context Jan Kara
2020-12-16  5:31   ` Theodore Y. Ts'o
2020-12-16  5:40     ` Theodore Y. Ts'o
2020-12-16  9:56       ` Jan Kara
2020-11-27 11:34 ` [PATCH 08/12] ext4: Combine ext4_handle_error() and save_error_info() Jan Kara
2020-12-14 19:23   ` harshad shirwadkar
2020-12-16 10:11     ` Jan Kara
2020-12-16 10:24       ` Jan Kara
2020-11-27 11:34 ` [PATCH 09/12] ext4: Drop sync argument of ext4_commit_super() Jan Kara
2020-12-14 19:25   ` harshad shirwadkar
2020-11-27 11:34 ` [PATCH 10/12] ext4: Protect superblock modifications with a buffer lock Jan Kara
2020-11-27 11:34 ` Jan Kara [this message]
2020-11-27 11:34 ` [PATCH 12/12] ext4: Use sbi instead of EXT4_SB(sb) in ext4_update_super() Jan Kara
2020-12-14 19:07 ` [PATCH 00/12] ext4: Various fixes of ext4 handling of fs errors harshad shirwadkar

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=20201127113405.26867-12-jack@suse.cz \
    --to=jack@suse.cz \
    --cc=linux-ext4@vger.kernel.org \
    --cc=tytso@mit.edu \
    --subject='Re: [PATCH 11/12] ext4: Save error info to sb through journal if available' \
    /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

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.