[PATCH] e2freefrag: fix free blocks count during live scan

[PATCH] e2freefrag: fix free blocks count during live scan

[Darrick J. Wong]

From: Darrick J. Wong <darrick.wong@oracle.com>

In e2freefrag live scan mode, we take the free block count from the
ondisk superblock.  This leads to screwy histogram percentages:

Extent Size Range :  Free extents   Free Blocks  Percent
    4M...    8M-  :             5          8234    1.05%
   64M...  128M-  :             2         52279    6.64%
  512M... 1024M-  :             1        202752   25.74%

...because there could be superblock updates in the journal that haven't
yet been checkpointed.  The online scan is perfectly capable of tallying
the free blocks on its own, so teach it do that and make a more accurate
report.

Reported-by: Elana Hashman <Elana.Hashman@twosigma.com>
Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>
---
 misc/e2freefrag.c |   26 ++++++++++++++++----------
 1 file changed, 16 insertions(+), 10 deletions(-)

diff --git a/misc/e2freefrag.c b/misc/e2freefrag.c
index 6eb4c736d..f7fede54b 100644
--- a/misc/e2freefrag.c
+++ b/misc/e2freefrag.c
@@ -163,7 +163,8 @@ static void scan_block_bitmap(ext2_filsys fs, struct chunk_info *info)
 
 #if defined(HAVE_EXT2_IOCTLS) && !defined(DEBUGFS)
 # define FSMAP_EXTENTS	1024
-static int scan_online(ext2_filsys fs, struct chunk_info *info)
+static int scan_online(ext2_filsys fs, struct chunk_info *info,
+		       blk64_t *free_blks)
 {
 	struct fsmap_head *fsmap;
 	struct fsmap *extent;
@@ -204,6 +205,7 @@ static int scan_online(ext2_filsys fs, struct chunk_info *info)
 	fsmap->fmh_keys[1].fmr_offset = ULLONG_MAX;
 	fsmap->fmh_keys[1].fmr_flags = UINT_MAX;
 
+	*free_blks = 0;
 	/* Fill the extent histogram with live data */
 	while (1) {
 		ret = ioctl(fd, FS_IOC_GETFSMAP, fsmap);
@@ -225,6 +227,7 @@ static int scan_online(ext2_filsys fs, struct chunk_info *info)
 				continue;
 			update_chunk_stats(info,
 					   extent->fmr_length / fs->blocksize);
+			*free_blks += (extent->fmr_length / fs->blocksize);
 		}
 
 		p = &fsmap->fmh_recs[fsmap->fmh_entries - 1];
@@ -236,13 +239,15 @@ static int scan_online(ext2_filsys fs, struct chunk_info *info)
 	return 1;
 }
 #else
-# define scan_online(fs, info)	(0)
+# define scan_online(fs, info, free_blks)	(0)
 #endif /* HAVE_EXT2_IOCTLS */
 
-static errcode_t scan_offline(ext2_filsys fs, struct chunk_info *info)
+static errcode_t scan_offline(ext2_filsys fs, struct chunk_info *info,
+			      blk64_t *free_blks)
 {
 	errcode_t retval;
 
+	*free_blks = ext2fs_free_blocks_count(fs->super);
 	retval = ext2fs_read_block_bitmap(fs);
 	if (retval)
 		return retval;
@@ -251,7 +256,7 @@ static errcode_t scan_offline(ext2_filsys fs, struct chunk_info *info)
 }
 
 static errcode_t dump_chunk_info(ext2_filsys fs, struct chunk_info *info,
-				 FILE *f)
+				 FILE *f, blk64_t free_blks)
 {
 	unsigned long total_chunks;
 	const char *unitp = "KMGTPEZY";
@@ -261,8 +266,8 @@ static errcode_t dump_chunk_info(ext2_filsys fs, struct chunk_info *info,
 
 	fprintf(f, "Total blocks: %llu\nFree blocks: %llu (%0.1f%%)\n",
 		ext2fs_blocks_count(fs->super),
-		ext2fs_free_blocks_count(fs->super),
-		(double)ext2fs_free_blocks_count(fs->super) * 100 /
+		free_blks,
+		(double)free_blks * 100 /
 		ext2fs_blocks_count(fs->super));
 
 	if (info->chunkbytes) {
@@ -306,7 +311,7 @@ static errcode_t dump_chunk_info(ext2_filsys fs, struct chunk_info *info,
 				info->histogram.fc_chunks[i],
 				info->histogram.fc_blocks[i],
 				(double)info->histogram.fc_blocks[i] * 100 /
-				ext2fs_free_blocks_count(fs->super));
+				free_blks);
 		}
 		start = end;
 		if (start == 1<<10) {
@@ -330,14 +335,15 @@ static void close_device(char *device_name, ext2_filsys fs)
 static void collect_info(ext2_filsys fs, struct chunk_info *chunk_info, FILE *f)
 {
 	unsigned int retval = 0;
+	blk64_t free_blks = 0;
 
 	fprintf(f, "Device: %s\n", fs->device_name);
 	fprintf(f, "Blocksize: %u bytes\n", fs->blocksize);
 
 	init_chunk_info(fs, chunk_info);
-	if (!scan_online(fs, chunk_info)) {
+	if (!scan_online(fs, chunk_info, &free_blks)) {
 		init_chunk_info(fs, chunk_info);
-		retval = scan_offline(fs, chunk_info);
+		retval = scan_offline(fs, chunk_info, &free_blks);
 	}
 	if (retval) {
 		com_err(fs->device_name, retval, "while reading block bitmap");
@@ -345,7 +351,7 @@ static void collect_info(ext2_filsys fs, struct chunk_info *chunk_info, FILE *f)
 		exit(1);
 	}
 
-	retval = dump_chunk_info(fs, chunk_info, f);
+	retval = dump_chunk_info(fs, chunk_info, f, free_blks);
 	if (retval) {
 		com_err(fs->device_name, retval, "while dumping chunk info");
                 close_device(fs->device_name, fs);

Re: [PATCH] e2freefrag: fix free blocks count during live scan

[Theodore Y. Ts'o]

On Thu, Nov 08, 2018 at 10:44:31AM -0800, Darrick J. Wong wrote:
> From: Darrick J. Wong <darrick.wong@oracle.com>
> 
> In e2freefrag live scan mode, we take the free block count from the
> ondisk superblock.  This leads to screwy histogram percentages:
> 
> Extent Size Range :  Free extents   Free Blocks  Percent
>     4M...    8M-  :             5          8234    1.05%
>    64M...  128M-  :             2         52279    6.64%
>   512M... 1024M-  :             1        202752   25.74%
> 
> ...because there could be superblock updates in the journal that haven't
> yet been checkpointed.  The online scan is perfectly capable of tallying
> the free blocks on its own, so teach it do that and make a more accurate
> report.
> 
> Reported-by: Elana Hashman <Elana.Hashman@twosigma.com>
> Signed-off-by: Darrick J. Wong <darrick.wong@oracle.com>

Thanks, applied.

					- Ted