[PATCH 2/7] xfs: fix min bufsize bugs in two places

[PATCH 2/7] xfs: fix min bufsize bugs in two places

[Alex Elder]

This fixes a bug in two places that I found by inspection.  In
xlog_find_verify_cycle() and xlog_write_log_records(), the code
attempts to allocate a buffer to hold as many blocks as possible.
It gives up if the number of blocks to be allocated gets too small.
Right now it uses log->l_sectbb_log as that lower bound, but I'm
sure it's supposed to be the actual log sector size instead.  That
is, the lower bound should be (1 << log->l_sectbb_log).

Also define a simple macro xlog_sectbb(log) to represent the number
of basic blocks in a sector for the given log.

Signed-off-by: Alex Elder <aelder@sgi.com>

---
 fs/xfs/xfs_log_recover.c |   21 +++++++++++++++++----
 1 file changed, 17 insertions(+), 4 deletions(-)

Index: b/fs/xfs/xfs_log_recover.c
===================================================================
--- a/fs/xfs/xfs_log_recover.c
+++ b/fs/xfs/xfs_log_recover.c
@@ -66,6 +66,9 @@ STATIC void	xlog_recover_check_summary(x
 	((bbs + (log)->l_sectbb_mask + 1) & ~(log)->l_sectbb_mask) : (bbs) )
 #define XLOG_SECTOR_ROUNDDOWN_BLKNO(log, bno)	((bno) & ~(log)->l_sectbb_mask)
 
+/* Number of basic blocks in a log sector */
+#define xlog_sectbb(log) (1 << (log)->l_sectbb_log)
+
 STATIC xfs_buf_t *
 xlog_get_bp(
 	xlog_t		*log,
@@ -376,12 +379,16 @@ xlog_find_verify_cycle(
 	xfs_caddr_t	buf = NULL;
 	int		error = 0;
 
+	/*
+	 * Greedily allocate a buffer big enough to handle the full
+	 * range of basic blocks we'll be examining.  If that fails,
+	 * try a smaller size.  We need to be able to read at least
+	 * a log sector, or we're out of luck.
+	 */
 	bufblks = 1 << ffs(nbblks);
-
 	while (!(bp = xlog_get_bp(log, bufblks))) {
-		/* can't get enough memory to do everything in one big buffer */
 		bufblks >>= 1;
-		if (bufblks <= log->l_sectbb_log)
+		if (bufblks < xlog_sectbb(log))
 			return ENOMEM;
 	}
 
@@ -1158,10 +1165,16 @@ xlog_write_log_records(
 	int		error = 0;
 	int		i, j = 0;
 
+	/*
+	 * Greedily allocate a buffer big enough to handle the full
+	 * range of basic blocks to be written.  If that fails, try
+	 * a smaller size.  We need to be able to write at least a
+	 * log sector, or we're out of luck.
+	 */
 	bufblks = 1 << ffs(blocks);
 	while (!(bp = xlog_get_bp(log, bufblks))) {
 		bufblks >>= 1;
-		if (bufblks <= log->l_sectbb_log)
+		if (bufblks < xlog_sectbb(log))
 			return ENOMEM;
 	}
 

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH 2/7] xfs: fix min bufsize bugs in two places

[Christoph Hellwig]

> +/* Number of basic blocks in a log sector */
> +#define xlog_sectbb(log) (1 << (log)->l_sectbb_log)

Looking at all uses of (log)->l_sectbb_log I wonder if we should
bother storing this in the log structure in this form, or rather
as the multipler of the basic block size, ala l_sectsize.  All the

	if (log->l_sectbb_log) {

checks would just become

	if (log->l_sectsize > 1) {

and the xlog_find_verify_cycle/xlog_write_log_records checks
would also be a natural

	if (bufblks < log->l_sectsize)

The comments added are defintively useful, btw.

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs

Re: [PATCH 2/7] xfs: fix min bufsize bugs in two places

[Alex Elder]

On Sat, 2010-03-20 at 12:42 -0400, Christoph Hellwig wrote:
> > +/* Number of basic blocks in a log sector */
> > +#define xlog_sectbb(log) (1 << (log)->l_sectbb_log)
> 
> Looking at all uses of (log)->l_sectbb_log I wonder if we should
> bother storing this in the log structure in this form, or rather
> as the multipler of the basic block size, ala l_sectsize.  All the

I agree, and I was sort of headed in that direction.  I have
more work in this file that eventually will lead to some larger
scale (algorithmic) simplification.  But for now I'm starting
small.

> 	if (log->l_sectbb_log) {
> 
> checks would just become
> 
> 	if (log->l_sectsize > 1) {
> 
> and the xlog_find_verify_cycle/xlog_write_log_records checks
> would also be a natural
> 
> 	if (bufblks < log->l_sectsize)

I'll attack this in a later patch.

> The comments added are defintively useful, btw.

Thanks, I felt they were necessary.  More to come.

I got "looks good" from you on all but this patch,
and numbers 5 and 7 in the series.  Is this one OK,
and do you plan to review those other two?

					-Alex

_______________________________________________
xfs mailing list
xfs@oss.sgi.com
http://oss.sgi.com/mailman/listinfo/xfs