+ jbd-strictly-check-for-write-errors-on-data-buffers.patch added to -mm tree

+ jbd-strictly-check-for-write-errors-on-data-buffers.patch added to -mm tree

[akpm]

The patch titled
     jbd: strictly check for write errors on data buffers
has been added to the -mm tree.  Its filename is
     jbd-strictly-check-for-write-errors-on-data-buffers.patch

Before you just go and hit "reply", please:
   a) Consider who else should be cc'ed
   b) Prefer to cc a suitable mailing list as well
   c) Ideally: find the original patch on the mailing list and do a
      reply-to-all to that, adding suitable additional cc's

*** Remember to use Documentation/SubmitChecklist when testing your code ***

See http://www.zip.com.au/~akpm/linux/patches/stuff/added-to-mm.txt to find
out what to do about this

The current -mm tree may be found at http://userweb.kernel.org/~akpm/mmotm/

------------------------------------------------------
Subject: jbd: strictly check for write errors on data buffers
From: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>

In ordered mode, we should abort journaling when an I/O error has occurred
on a file data buffer in the committing transaction.  But there can be
data buffers which are not checked for error:

  (a) the buffer which has already been written out by pdflush
  (b) the buffer which has been unlocked before scanned in the
      t_locked_list loop

This patch adds missing error checks and aborts journaling
appropriately.

Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitacih.com>
Acked-by: Jan Kara <jack@suse.cz>
Cc: <linux-ext4@vger.kernel.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 fs/jbd/commit.c |   14 +++++++++-----
 1 file changed, 9 insertions(+), 5 deletions(-)

diff -puN fs/jbd/commit.c~jbd-strictly-check-for-write-errors-on-data-buffers fs/jbd/commit.c
--- a/fs/jbd/commit.c~jbd-strictly-check-for-write-errors-on-data-buffers
+++ a/fs/jbd/commit.c
@@ -172,7 +172,7 @@ static void journal_do_submit_data(struc
 /*
  *  Submit all the data buffers to disk
  */
-static void journal_submit_data_buffers(journal_t *journal,
+static int journal_submit_data_buffers(journal_t *journal,
 				transaction_t *commit_transaction)
 {
 	struct journal_head *jh;
@@ -180,6 +180,7 @@ static void journal_submit_data_buffers(
 	int locked;
 	int bufs = 0;
 	struct buffer_head **wbuf = journal->j_wbuf;
+	int err = 0;
 
 	/*
 	 * Whenever we unlock the journal and sleep, things can get added
@@ -253,6 +254,8 @@ write_out_data:
 			put_bh(bh);
 		} else {
 			BUFFER_TRACE(bh, "writeout complete: unfile");
+			if (unlikely(!buffer_uptodate(bh)))
+				err = -EIO;
 			__journal_unfile_buffer(jh);
 			jbd_unlock_bh_state(bh);
 			if (locked)
@@ -271,6 +274,8 @@ write_out_data:
 	}
 	spin_unlock(&journal->j_list_lock);
 	journal_do_submit_data(wbuf, bufs);
+
+	return err;
 }
 
 /*
@@ -410,8 +415,7 @@ void journal_commit_transaction(journal_
 	 * Now start flushing things to disk, in the order they appear
 	 * on the transaction lists.  Data blocks go first.
 	 */
-	err = 0;
-	journal_submit_data_buffers(journal, commit_transaction);
+	err = journal_submit_data_buffers(journal, commit_transaction);
 
 	/*
 	 * Wait for all previously submitted IO to complete.
@@ -426,10 +430,10 @@ void journal_commit_transaction(journal_
 		if (buffer_locked(bh)) {
 			spin_unlock(&journal->j_list_lock);
 			wait_on_buffer(bh);
-			if (unlikely(!buffer_uptodate(bh)))
-				err = -EIO;
 			spin_lock(&journal->j_list_lock);
 		}
+		if (unlikely(!buffer_uptodate(bh)))
+			err = -EIO;
 		if (!inverted_lock(journal, bh)) {
 			put_bh(bh);
 			spin_lock(&journal->j_list_lock);
_

Patches currently in -mm which might be from hidehiro.kawai.ez@hitachi.com are

jbd-strictly-check-for-write-errors-on-data-buffers.patch
jbd-ordered-data-integrity-fix.patch
jbd-abort-when-failed-to-log-metadata-buffers.patch
jbd-fix-error-handling-for-checkpoint-io.patch
ext3-abort-ext3-if-the-journal-has-aborted.patch

Re: + jbd-strictly-check-for-write-errors-on-data-buffers.patch added to -mm tree

[Andreas Dilger]

On Jun 03, 2008  15:36 -0700, Andrew Morton wrote:
> ------------------------------------------------------
> Subject: jbd: strictly check for write errors on data buffers
> From: Hidehiro Kawai <hidehiro.kawai.ez@hitachi.com>
> 
> In ordered mode, we should abort journaling when an I/O error has occurred
> on a file data buffer in the committing transaction.  But there can be
> data buffers which are not checked for error:
> 
>   (a) the buffer which has already been written out by pdflush
>   (b) the buffer which has been unlocked before scanned in the
>       t_locked_list loop
> 
> This patch adds missing error checks and aborts journaling
> appropriately.

Should we actually abort the journal for file data write errors?

It appears journal_submit_data_buffers() is only for file data
(BJ_SyncData) and not any metadata (BJ_Metadata).  In data=ordered
mode there is no risk to the filesystem integrity.

> Signed-off-by: Hidehiro Kawai <hidehiro.kawai.ez@hitacih.com>
> Acked-by: Jan Kara <jack@suse.cz>
> Cc: <linux-ext4@vger.kernel.org>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
>  fs/jbd/commit.c |   14 +++++++++-----
>  1 file changed, 9 insertions(+), 5 deletions(-)
> 
> diff -puN fs/jbd/commit.c~jbd-strictly-check-for-write-errors-on-data-buffers fs/jbd/commit.c
> --- a/fs/jbd/commit.c~jbd-strictly-check-for-write-errors-on-data-buffers
> +++ a/fs/jbd/commit.c
> @@ -172,7 +172,7 @@ static void journal_do_submit_data(struc
>  /*
>   *  Submit all the data buffers to disk
>   */
> -static void journal_submit_data_buffers(journal_t *journal,
> +static int journal_submit_data_buffers(journal_t *journal,
>  				transaction_t *commit_transaction)
>  {
>  	struct journal_head *jh;
> @@ -180,6 +180,7 @@ static void journal_submit_data_buffers(
>  	int locked;
>  	int bufs = 0;
>  	struct buffer_head **wbuf = journal->j_wbuf;
> +	int err = 0;
>  
>  	/*
>  	 * Whenever we unlock the journal and sleep, things can get added
> @@ -253,6 +254,8 @@ write_out_data:
>  			put_bh(bh);
>  		} else {
>  			BUFFER_TRACE(bh, "writeout complete: unfile");
> +			if (unlikely(!buffer_uptodate(bh)))
> +				err = -EIO;
>  			__journal_unfile_buffer(jh);
>  			jbd_unlock_bh_state(bh);
>  			if (locked)
> @@ -271,6 +274,8 @@ write_out_data:
>  	}
>  	spin_unlock(&journal->j_list_lock);
>  	journal_do_submit_data(wbuf, bufs);
> +
> +	return err;
>  }
>  
>  /*
> @@ -410,8 +415,7 @@ void journal_commit_transaction(journal_
>  	 * Now start flushing things to disk, in the order they appear
>  	 * on the transaction lists.  Data blocks go first.
>  	 */
> -	err = 0;
> -	journal_submit_data_buffers(journal, commit_transaction);
> +	err = journal_submit_data_buffers(journal, commit_transaction);
>  
>  	/*
>  	 * Wait for all previously submitted IO to complete.
> @@ -426,10 +430,10 @@ void journal_commit_transaction(journal_
>  		if (buffer_locked(bh)) {
>  			spin_unlock(&journal->j_list_lock);
>  			wait_on_buffer(bh);
> -			if (unlikely(!buffer_uptodate(bh)))
> -				err = -EIO;
>  			spin_lock(&journal->j_list_lock);
>  		}
> +		if (unlikely(!buffer_uptodate(bh)))
> +			err = -EIO;
>  		if (!inverted_lock(journal, bh)) {
>  			put_bh(bh);
>  			spin_lock(&journal->j_list_lock);
> _
> 
> Patches currently in -mm which might be from hidehiro.kawai.ez@hitachi.com are
> 
> jbd-strictly-check-for-write-errors-on-data-buffers.patch
> jbd-ordered-data-integrity-fix.patch
> jbd-abort-when-failed-to-log-metadata-buffers.patch
> jbd-fix-error-handling-for-checkpoint-io.patch
> ext3-abort-ext3-if-the-journal-has-aborted.patch
> 
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Cheers, Andreas
--
Andreas Dilger
Sr. Staff Engineer, Lustre Group
Sun Microsystems of Canada, Inc.