[GFS2] assertion failure after writing to journaled file, umount

This patch passes all my nasty tests that were causing the code to
fail under one circumstance or another.  Here is a complete summary
of all changes from today's git tree, in order of appearance:

1. There are now separate variables for metadata buffer accounting.
2. Variable sd_log_num_hdrs is no longer needed, since the header
   accounting is taken care of by the reserve/refund sequence.
3. Fixed a tiny grammatical problem in a comment.
4. Added a new function "calc_reserved" to calculate the reserved
   log space.  This isn't entirely necessary, but it has two benefits:
   First, it simplifies the gfs2_log_refund function greatly.
   Second, it allows for easier debugging because I could sprinkle the
   code with calls to this function to make sure the accounting is
   proper (by adding asserts and printks) at strategic point of the code.
5. In log_pull_tail there apparently was a kludge to fix up the
   accounting based on a "pull" parameter.  The buffer accounting is
   now done properly, so the kludge was removed.
6. File sync operations were making a call to gfs2_log_flush that
   writes another journal header.  Since that header was unplanned
   for (reserved) by the reserve/refund sequence, the free space had
   to be decremented so that when log_pull_tail gets called, the free
   space is be adjusted properly.  (Did I hear you call that a kludge?
   well, maybe, but a lot more justifiable than the one I removed).
7. In the gfs2_log_shutdown code, it optionally syncs the log by
   specifying the PULL parameter to log_write_header.  I'm not sure
   this is necessary anymore.  It just seems to me there could be
   cases where shutdown is called while there are outstanding log
   buffers.
8. In the (data)buf_lo_before_commit functions, I changed some offset
   values from being calculated on the fly to being constants.	That
   simplified some code and we might as well let the compiler do the
   calculation once rather than redoing those cycles at run time.
9. This version has my rewritten databuf_lo_add function.
   This version is much more like its predecessor, buf_lo_add, which
   makes it easier to understand.  Again, this might not be necessary,
   but it seems as if this one works as well as the previous one,
   maybe even better, so I decided to leave it in.
10. In databuf_lo_before_commit, a previous data corruption problem
   was caused by going off the end of the buffer.  The proper solution
   is to have the proper limit in place, rather than stopping earlier.
   (Thus my previous attempt to fix it is wrong).
   If you don't wrap the buffer, you're stopping too early and that
   causes more log buffer accounting problems.
11. In lops.h there are two new (previously mentioned) constants for
   figuring out the data offset for the journal buffers.
12. There are also two new functions, buf_limit and databuf_limit to
   calculate how many entries will fit in the buffer.
13. In function gfs2_meta_wipe, it needs to distinguish between pinned
   metadata buffers and journaled data buffers for proper journal buffer
   accounting.	It can't use the JDATA gfs2_inode flag because it's
   sometimes passed the "real" inode and sometimes the "metadata
   inode" and the inode flags will be random bits in a metadata
   gfs2_inode.	It needs to base its decision on which was passed in.

Signed-off-by: Bob Peterson <rpeterso@redhat.com>
Signed-off-by: Steven Whitehouse <swhiteho@redhat.com>

1 files changed, 76 insertions, 24 deletions

diff --git a/fs/gfs2/log.c b/fs/gfs2/log.c
index fbdc0dc9923e..8fcfb784f906 100644
--- a/fs/gfs2/log.c
+++ b/fs/gfs2/log.c
@@ -276,7 +276,7 @@ static void ail2_empty(struct gfs2_sbd *sdp, unsigned int new_tail)
  * @blks: The number of blocks to reserve
  *
  * Note that we never give out the last few blocks of the journal. Thats
- * due to the fact that there is are a small number of header blocks
+ * due to the fact that there is a small number of header blocks
  * associated with each log flush. The exact number can't be known until
  * flush time, so we ensure that we have just enough free blocks at all
  * times to avoid running out during a log flush.
@@ -371,6 +371,58 @@ static inline unsigned int log_distance(struct gfs2_sbd *sdp, unsigned int newer
 	return dist;
 }
 
+/**
+ * calc_reserved - Calculate the number of blocks to reserve when
+ *                 refunding a transaction's unused buffers.
+ * @sdp: The GFS2 superblock
+ *
+ * This is complex.  We need to reserve room for all our currently used
+ * metadata buffers (e.g. normal file I/O rewriting file time stamps) and 
+ * all our journaled data buffers for journaled files (e.g. files in the 
+ * meta_fs like rindex, or files for which chattr +j was done.)
+ * If we don't reserve enough space, gfs2_log_refund and gfs2_log_flush
+ * will count it as free space (sd_log_blks_free) and corruption will follow.
+ *
+ * We can have metadata bufs and jdata bufs in the same journal.  So each
+ * type gets its own log header, for which we need to reserve a block.
+ * In fact, each type has the potential for needing more than one header 
+ * in cases where we have more buffers than will fit on a journal page.
+ * Metadata journal entries take up half the space of journaled buffer entries.
+ * Thus, metadata entries have buf_limit (502) and journaled buffers have
+ * databuf_limit (251) before they cause a wrap around.
+ *
+ * Also, we need to reserve blocks for revoke journal entries and one for an
+ * overall header for the lot.
+ *
+ * Returns: the number of blocks reserved
+ */
+static unsigned int calc_reserved(struct gfs2_sbd *sdp)
+{
+	unsigned int reserved = 0;
+	unsigned int mbuf_limit, metabufhdrs_needed;
+	unsigned int dbuf_limit, databufhdrs_needed;
+	unsigned int revokes = 0;
+
+	mbuf_limit = buf_limit(sdp);
+	metabufhdrs_needed = (sdp->sd_log_commited_buf +
+			      (mbuf_limit - 1)) / mbuf_limit;
+	dbuf_limit = databuf_limit(sdp);
+	databufhdrs_needed = (sdp->sd_log_commited_databuf +
+			      (dbuf_limit - 1)) / dbuf_limit;
+
+	if (sdp->sd_log_commited_revoke)
+		revokes = gfs2_struct2blk(sdp, sdp->sd_log_commited_revoke,
+					  sizeof(u64));
+
+	reserved = sdp->sd_log_commited_buf + metabufhdrs_needed +
+		sdp->sd_log_commited_databuf + databufhdrs_needed +
+		revokes;
+	/* One for the overall header */
+	if (reserved)
+		reserved++;
+	return reserved;
+}
+
 static unsigned int current_tail(struct gfs2_sbd *sdp)
 {
 	struct gfs2_ail *ai;
@@ -461,14 +513,14 @@ struct buffer_head *gfs2_log_fake_buf(struct gfs2_sbd *sdp,
 	return bh;
 }
 
-static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail, int pull)
+static void log_pull_tail(struct gfs2_sbd *sdp, unsigned int new_tail)
 {
 	unsigned int dist = log_distance(sdp, new_tail, sdp->sd_log_tail);
 
 	ail2_empty(sdp, new_tail);
 
 	gfs2_log_lock(sdp);
-	sdp->sd_log_blks_free += dist - (pull ? 1 : 0);
+	sdp->sd_log_blks_free += dist;
 	gfs2_assert_withdraw(sdp, sdp->sd_log_blks_free <= sdp->sd_jdesc->jd_blocks);
 	gfs2_log_unlock(sdp);
 
@@ -518,7 +570,7 @@ static void log_write_header(struct gfs2_sbd *sdp, u32 flags, int pull)
 	brelse(bh);
 
 	if (sdp->sd_log_tail != tail)
-		log_pull_tail(sdp, tail, pull);
+		log_pull_tail(sdp, tail);
 	else
 		gfs2_assert_withdraw(sdp, !pull);
 
@@ -579,7 +631,10 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl)
 	INIT_LIST_HEAD(&ai->ai_ail1_list);
 	INIT_LIST_HEAD(&ai->ai_ail2_list);
 
-	gfs2_assert_withdraw(sdp, sdp->sd_log_num_buf + sdp->sd_log_num_jdata == sdp->sd_log_commited_buf);
+	gfs2_assert_withdraw(sdp,
+			     sdp->sd_log_num_buf + sdp->sd_log_num_jdata ==
+			     sdp->sd_log_commited_buf +
+			     sdp->sd_log_commited_databuf);
 	gfs2_assert_withdraw(sdp,
 			sdp->sd_log_num_revoke == sdp->sd_log_commited_revoke);
 
@@ -590,16 +645,19 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl)
 	lops_before_commit(sdp);
 	if (!list_empty(&sdp->sd_log_flush_list))
 		log_flush_commit(sdp);
-	else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle)
+	else if (sdp->sd_log_tail != current_tail(sdp) && !sdp->sd_log_idle){
+		gfs2_log_lock(sdp);
+		sdp->sd_log_blks_free--; /* Adjust for unreserved buffer */
+		gfs2_log_unlock(sdp);
 		log_write_header(sdp, 0, PULL);
+	}
 	lops_after_commit(sdp, ai);
 
 	gfs2_log_lock(sdp);
 	sdp->sd_log_head = sdp->sd_log_flush_head;
-	sdp->sd_log_blks_free -= sdp->sd_log_num_hdrs;
 	sdp->sd_log_blks_reserved = 0;
 	sdp->sd_log_commited_buf = 0;
-	sdp->sd_log_num_hdrs = 0;
+	sdp->sd_log_commited_databuf = 0;
 	sdp->sd_log_commited_revoke = 0;
 
 	if (!list_empty(&ai->ai_ail1_list)) {
@@ -616,32 +674,26 @@ void gfs2_log_flush(struct gfs2_sbd *sdp, struct gfs2_glock *gl)
 
 static void log_refund(struct gfs2_sbd *sdp, struct gfs2_trans *tr)
 {
-	unsigned int reserved = 0;
+	unsigned int reserved;
 	unsigned int old;
 
 	gfs2_log_lock(sdp);
 
 	sdp->sd_log_commited_buf += tr->tr_num_buf_new - tr->tr_num_buf_rm;
-	gfs2_assert_withdraw(sdp, ((int)sdp->sd_log_commited_buf) >= 0);
+	sdp->sd_log_commited_databuf += tr->tr_num_databuf_new -
+		tr->tr_num_databuf_rm;
+	gfs2_assert_withdraw(sdp, (((int)sdp->sd_log_commited_buf) >= 0) ||
+			     (((int)sdp->sd_log_commited_databuf) >= 0));
 	sdp->sd_log_commited_revoke += tr->tr_num_revoke - tr->tr_num_revoke_rm;
 	gfs2_assert_withdraw(sdp, ((int)sdp->sd_log_commited_revoke) >= 0);
-
-	if (sdp->sd_log_commited_buf)
-		reserved += sdp->sd_log_commited_buf;
-	if (sdp->sd_log_commited_revoke)
-		reserved += gfs2_struct2blk(sdp, sdp->sd_log_commited_revoke,
-					    sizeof(u64));
-	if (reserved)
-		reserved++;
-
+	reserved = calc_reserved(sdp);
 	old = sdp->sd_log_blks_free;
 	sdp->sd_log_blks_free += tr->tr_reserved -
 				 (reserved - sdp->sd_log_blks_reserved);
 
 	gfs2_assert_withdraw(sdp, sdp->sd_log_blks_free >= old);
-	gfs2_assert_withdraw(sdp,
-			     sdp->sd_log_blks_free <= sdp->sd_jdesc->jd_blocks +
-			     sdp->sd_log_num_hdrs);
+	gfs2_assert_withdraw(sdp, sdp->sd_log_blks_free <=
+			     sdp->sd_jdesc->jd_blocks);
 
 	sdp->sd_log_blks_reserved = reserved;
 
@@ -687,13 +739,13 @@ void gfs2_log_shutdown(struct gfs2_sbd *sdp)
 	gfs2_assert_withdraw(sdp, !sdp->sd_log_num_revoke);
 	gfs2_assert_withdraw(sdp, !sdp->sd_log_num_rg);
 	gfs2_assert_withdraw(sdp, !sdp->sd_log_num_databuf);
-	gfs2_assert_withdraw(sdp, !sdp->sd_log_num_hdrs);
 	gfs2_assert_withdraw(sdp, list_empty(&sdp->sd_ail1_list));
 
 	sdp->sd_log_flush_head = sdp->sd_log_head;
 	sdp->sd_log_flush_wrapped = 0;
 
-	log_write_header(sdp, GFS2_LOG_HEAD_UNMOUNT, 0);
+	log_write_header(sdp, GFS2_LOG_HEAD_UNMOUNT,
+			 (sdp->sd_log_tail == current_tail(sdp)) ? 0 : PULL);
 
 	gfs2_assert_warn(sdp, sdp->sd_log_blks_free == sdp->sd_jdesc->jd_blocks);
 	gfs2_assert_warn(sdp, sdp->sd_log_head == sdp->sd_log_tail);