[11/28] ocfs2: extend enough credits for freeing one truncate record while replaying truncate records
diff mbox

Message ID 55de39a4.mm24W9YHh8D6r9FS%akpm@linux-foundation.org
State New
Headers show

Commit Message

Andrew Morton Aug. 26, 2015, 10:11 p.m. UTC
From: Xue jiufei <xuejiufei@huawei.com>
Subject: ocfs2: extend enough credits for freeing one truncate record while replaying truncate records

Now function ocfs2_replay_truncate_records() first modifies tl_used, then
calls ocfs2_extend_trans() to extend transactions for gd and alloc inode
used for freeing clusters.  jbd2_journal_restart() may be called and it
may happen that tl_used in truncate log is decreased but the clusters are
not freed, which means these clusters are lost.  So we should avoid
extending transactions in these two operations.

Signed-off-by: joyce.xue <xuejiufei@huawei.com>
Cc: Mark Fasheh <mfasheh@suse.com>
Cc: Joel Becker <jlbec@evilplan.org>
Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
---

 fs/ocfs2/alloc.c |   19 ++++++++-----------
 1 file changed, 8 insertions(+), 11 deletions(-)

Comments

Mark Fasheh Aug. 28, 2015, 11:55 p.m. UTC | #1
On Wed, Aug 26, 2015 at 03:11:48PM -0700, Andrew Morton wrote:
> From: Xue jiufei <xuejiufei@huawei.com>
> Subject: ocfs2: extend enough credits for freeing one truncate record while replaying truncate records
> 
> Now function ocfs2_replay_truncate_records() first modifies tl_used, then
> calls ocfs2_extend_trans() to extend transactions for gd and alloc inode
> used for freeing clusters.  jbd2_journal_restart() may be called and it
> may happen that tl_used in truncate log is decreased but the clusters are
> not freed, which means these clusters are lost.  So we should avoid
> extending transactions in these two operations.
> 
> Signed-off-by: joyce.xue <xuejiufei@huawei.com>
> Cc: Mark Fasheh <mfasheh@suse.com>
> Cc: Joel Becker <jlbec@evilplan.org>
> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> ---
> 
>  fs/ocfs2/alloc.c |   19 ++++++++-----------
>  1 file changed, 8 insertions(+), 11 deletions(-)
> 
> diff -puN fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records fs/ocfs2/alloc.c
> --- a/fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records
> +++ a/fs/ocfs2/alloc.c

<looks good>

> @@ -6063,7 +6060,7 @@ int __ocfs2_flush_truncate_log(struct oc
>  		goto out_mutex;
>  	}
>  
> -	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
> +	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
>  	if (IS_ERR(handle)) {
>  		status = PTR_ERR(handle);
>  		mlog_errno(status);

Why is this particular change here?
	--Mark

--
Mark Fasheh
Joseph Qi Aug. 29, 2015, 2:38 a.m. UTC | #2
Hi Mark,

On 2015/8/29 7:55, Mark Fasheh wrote:
> On Wed, Aug 26, 2015 at 03:11:48PM -0700, Andrew Morton wrote:
>> From: Xue jiufei <xuejiufei@huawei.com>
>> Subject: ocfs2: extend enough credits for freeing one truncate record while replaying truncate records
>>
>> Now function ocfs2_replay_truncate_records() first modifies tl_used, then
>> calls ocfs2_extend_trans() to extend transactions for gd and alloc inode
>> used for freeing clusters.  jbd2_journal_restart() may be called and it
>> may happen that tl_used in truncate log is decreased but the clusters are
>> not freed, which means these clusters are lost.  So we should avoid
>> extending transactions in these two operations.
>>
>> Signed-off-by: joyce.xue <xuejiufei@huawei.com>
>> Cc: Mark Fasheh <mfasheh@suse.com>
>> Cc: Joel Becker <jlbec@evilplan.org>
>> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
>> ---
>>
>>  fs/ocfs2/alloc.c |   19 ++++++++-----------
>>  1 file changed, 8 insertions(+), 11 deletions(-)
>>
>> diff -puN fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records fs/ocfs2/alloc.c
>> --- a/fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records
>> +++ a/fs/ocfs2/alloc.c
> 
> <looks good>
> 
>> @@ -6063,7 +6060,7 @@ int __ocfs2_flush_truncate_log(struct oc
>>  		goto out_mutex;
>>  	}
>>  
>> -	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
>> +	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
>>  	if (IS_ERR(handle)) {
>>  		status = PTR_ERR(handle);
>>  		mlog_errno(status);
> 
> Why is this particular change here?
> 	--Mark
> 
I think Joyce wants to get enough credits at first and then call
ocfs2_replay_truncate_records.

Thanks
Joseph

> --
> Mark Fasheh
> 
> _______________________________________________
> Ocfs2-devel mailing list
> Ocfs2-devel@oss.oracle.com
> https://oss.oracle.com/mailman/listinfo/ocfs2-devel
> 
>
Mark Fasheh Sept. 1, 2015, 5:54 p.m. UTC | #3
On Sat, Aug 29, 2015 at 10:38:09AM +0800, Joseph Qi wrote:
> Hi Mark,
> 
> On 2015/8/29 7:55, Mark Fasheh wrote:
> > On Wed, Aug 26, 2015 at 03:11:48PM -0700, Andrew Morton wrote:
> >> From: Xue jiufei <xuejiufei@huawei.com>
> >> Subject: ocfs2: extend enough credits for freeing one truncate record while replaying truncate records
> >>
> >> Now function ocfs2_replay_truncate_records() first modifies tl_used, then
> >> calls ocfs2_extend_trans() to extend transactions for gd and alloc inode
> >> used for freeing clusters.  jbd2_journal_restart() may be called and it
> >> may happen that tl_used in truncate log is decreased but the clusters are
> >> not freed, which means these clusters are lost.  So we should avoid
> >> extending transactions in these two operations.
> >>
> >> Signed-off-by: joyce.xue <xuejiufei@huawei.com>
> >> Cc: Mark Fasheh <mfasheh@suse.com>
> >> Cc: Joel Becker <jlbec@evilplan.org>
> >> Signed-off-by: Andrew Morton <akpm@linux-foundation.org>
> >> ---
> >>
> >>  fs/ocfs2/alloc.c |   19 ++++++++-----------
> >>  1 file changed, 8 insertions(+), 11 deletions(-)
> >>
> >> diff -puN fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records fs/ocfs2/alloc.c
> >> --- a/fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records
> >> +++ a/fs/ocfs2/alloc.c
> > 
> > <looks good>
> > 
> >> @@ -6063,7 +6060,7 @@ int __ocfs2_flush_truncate_log(struct oc
> >>  		goto out_mutex;
> >>  	}
> >>  
> >> -	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
> >> +	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
> >>  	if (IS_ERR(handle)) {
> >>  		status = PTR_ERR(handle);
> >>  		mlog_errno(status);
> > 
> > Why is this particular change here?
> > 	--Mark
> > 
> I think Joyce wants to get enough credits at first and then call
> ocfs2_replay_truncate_records.

Oh ok I see. That's fine then, the patch otherwise looked good to me,
thanks.

Reviewed-by: Mark Fasheh <mfasheh@suse.de>
	--Mark

--
Mark Fasheh
Andrew Morton Sept. 1, 2015, 10:12 p.m. UTC | #4
On Tue, 1 Sep 2015 10:54:03 -0700 Mark Fasheh <mfasheh@suse.de> wrote:

> > > Why is this particular change here?
> > > 	--Mark
> > > 
> > I think Joyce wants to get enough credits at first and then call
> > ocfs2_replay_truncate_records.
> 
> Oh ok I see. That's fine then, the patch otherwise looked good to me,
> thanks.
> 
> Reviewed-by: Mark Fasheh <mfasheh@suse.de>

Cool.  And where are we at with
http://ozlabs.org/~akpm/mmots/broken-out/ocfs2-extend-transaction-for-ocfs2_remove_rightmost_path-and-ocfs2_update_edge_lengths-before-to-avoid-inconsistency-between-inode-and-et.patch

?
Mark Fasheh Sept. 1, 2015, 10:36 p.m. UTC | #5
On Tue, Sep 01, 2015 at 03:12:07PM -0700, Andrew Morton wrote:
> On Tue, 1 Sep 2015 10:54:03 -0700 Mark Fasheh <mfasheh@suse.de> wrote:
> 
> > > > Why is this particular change here?
> > > > 	--Mark
> > > > 
> > > I think Joyce wants to get enough credits at first and then call
> > > ocfs2_replay_truncate_records.
> > 
> > Oh ok I see. That's fine then, the patch otherwise looked good to me,
> > thanks.
> > 
> > Reviewed-by: Mark Fasheh <mfasheh@suse.de>
> 
> Cool.  And where are we at with
> http://ozlabs.org/~akpm/mmots/broken-out/ocfs2-extend-transaction-for-ocfs2_remove_rightmost_path-and-ocfs2_update_edge_lengths-before-to-avoid-inconsistency-between-inode-and-et.patch
> 
> ?

I had a couple questions, we're waiting for a response:

https://www.mail-archive.com/ocfs2-devel@oss.oracle.com/msg10167.html
	--Mark

--
Mark Fasheh

Patch
diff mbox

diff -puN fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records fs/ocfs2/alloc.c
--- a/fs/ocfs2/alloc.c~extend-enough-credits-for-freeing-one-truncate-record-while-replaying-truncate-records
+++ a/fs/ocfs2/alloc.c
@@ -5975,16 +5975,6 @@  static int ocfs2_replay_truncate_records
 
 		ocfs2_journal_dirty(handle, tl_bh);
 
-		/* TODO: Perhaps we can calculate the bulk of the
-		 * credits up front rather than extending like
-		 * this. */
-		status = ocfs2_extend_trans(handle,
-					    OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
-		if (status < 0) {
-			mlog_errno(status);
-			goto bail;
-		}
-
 		rec = tl->tl_recs[i];
 		start_blk = ocfs2_clusters_to_blocks(data_alloc_inode->i_sb,
 						    le32_to_cpu(rec.t_start));
@@ -6005,6 +5995,13 @@  static int ocfs2_replay_truncate_records
 				goto bail;
 			}
 		}
+
+		status = ocfs2_extend_trans(handle,
+				OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
+		if (status < 0) {
+			mlog_errno(status);
+			goto bail;
+		}
 		i--;
 	}
 
@@ -6063,7 +6060,7 @@  int __ocfs2_flush_truncate_log(struct oc
 		goto out_mutex;
 	}
 
-	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_UPDATE);
+	handle = ocfs2_start_trans(osb, OCFS2_TRUNCATE_LOG_FLUSH_ONE_REC);
 	if (IS_ERR(handle)) {
 		status = PTR_ERR(handle);
 		mlog_errno(status);