[Patch 7/13] Many inodes in filesystem.

[Patch 7/13] Many inodes in filesystem.

[Girish Shilamkar]

Handle filesystems with many hard links that have more than ~700M inodes.
The float variable lacks precision at this scale.

Signed-off-by: Andreas Dilger <adilger@clusterfs.com>
Signed-off-by: Girish Shilamkar <girish@clusterfs.com>
Signed-off-by: Kalpak Shah <kalpak@clusterfs.com>

Index: e2fsprogs-1.40.1/lib/ext2fs/icount.c
===================================================================
--- e2fsprogs-1.40.1.orig/lib/ext2fs/icount.c
+++ e2fsprogs-1.40.1/lib/ext2fs/icount.c
@@ -376,6 +376,10 @@ static struct ext2_icount_el *get_icount
 					range = 0.1;
 			}
 			mid = low + ((int) (range * (high-low)));
+			if (mid > high)
+				mid = high;
+			if (mid < low)
+				mid = low;
 		}
 #endif
 		if (ino == icount->list[mid].ino) {

Re: [Patch 7/13] Many inodes in filesystem.

[Theodore Tso]

On Tue, Jul 24, 2007 at 04:34:59PM +0530, Girish Shilamkar wrote:
> Handle filesystems with many hard links that have more than ~700M inodes.
> The float variable lacks precision at this scale.

This was fixed in e2fsprogs mainline a slightly different way.  Just
above it:

				range = ((float) (ino - lowval)) /
					(highval - lowval);
				if (range > 0.9)
					range = 0.9;
				if (range < 0.1)
					range = 0.1;

The problem with just correcting the midpoint calculation as you
proposed in your patch is that you could end up doing a linear search
through the whole array in some really abberrant cases.  With this fix
applied in e2fsprogs, the patch below isn't required.

Regards,

					- Ted

> Signed-off-by: Andreas Dilger <adilger@clusterfs.com>
> Signed-off-by: Girish Shilamkar <girish@clusterfs.com>
> Signed-off-by: Kalpak Shah <kalpak@clusterfs.com>
> 
> Index: e2fsprogs-1.40.1/lib/ext2fs/icount.c
> ===================================================================
> --- e2fsprogs-1.40.1.orig/lib/ext2fs/icount.c
> +++ e2fsprogs-1.40.1/lib/ext2fs/icount.c
> @@ -376,6 +376,10 @@ static struct ext2_icount_el *get_icount
>  					range = 0.1;
>  			}
>  			mid = low + ((int) (range * (high-low)));
> +			if (mid > high)
> +				mid = high;
> +			if (mid < low)
> +				mid = low;
>  		}
>  #endif
>  		if (ino == icount->list[mid].ino) {
> 
>