[PATCH] e2fsck: avoid overflow in pass5 check_block_end()

[PATCH] e2fsck: avoid overflow in pass5 check_block_end()

[Eric Sandeen]

When making a filesystem as in:

mke2fs -E lazy_itable_init=1 -O uninit_bg -b 4096 bigfile 4294967298

a subsequent fsck would result in:

Pass 5: Checking group summary information
Internal error: fudging end of bitmap (3)
e2fsck: aborted

This is because check_block_end() was overflowing in the calculation
for "end", and giving it a value of -1 (0xFFFF....) which eventually
ended up tripping up a test in ext2fs_fudge_generic_bmap_end,

        if (end > bitmap->real_end)
                return neq;

Fix another such error in read_bitmaps() as well.

lib/ext2fs/imager.c likely has similar problems but it looks like
it has no 64-bit treatment at all yet.

Signed-off-by: Eric Sandeen <sandeen@redhat.com>
---

Applies to pu branch.

Index: e2fsprogs/e2fsck/pass5.c
===================================================================
--- e2fsprogs.orig/e2fsck/pass5.c
+++ e2fsprogs/e2fsck/pass5.c
@@ -662,7 +662,7 @@ static void check_block_end(e2fsck_t ctx
 	clear_problem_context(&pctx);
 
 	end = ext2fs_get_block_bitmap_start2(fs->block_map) +
-		(EXT2_BLOCKS_PER_GROUP(fs->super) * fs->group_desc_count) - 1;
+		((blk64_t)EXT2_BLOCKS_PER_GROUP(fs->super) * fs->group_desc_count) - 1;
 	pctx.errcode = ext2fs_fudge_block_bitmap_end2(fs->block_map, end,
 						     &save_blocks_count);
 	if (pctx.errcode) {
Index: e2fsprogs/lib/ext2fs/rw_bitmaps.c
===================================================================
--- e2fsprogs.orig/lib/ext2fs/rw_bitmaps.c
+++ e2fsprogs/lib/ext2fs/rw_bitmaps.c
@@ -212,7 +212,7 @@ static errcode_t read_bitmaps(ext2_filsy
 		}
 		blk = (fs->image_header->offset_blockmap /
 		       fs->blocksize);
-		blk_cnt = EXT2_BLOCKS_PER_GROUP(fs->super) *
+		blk_cnt = (blk64_t)EXT2_BLOCKS_PER_GROUP(fs->super) *
 			fs->group_desc_count;
 		while (block_nbytes > 0) {
 			retval = io_channel_read_blk64(fs->image_io, blk++,

Re: [PATCH] e2fsck: avoid overflow in pass5 check_block_end()

[Theodore Tso]

On Tue, Sep 01, 2009 at 05:40:28PM -0500, Eric Sandeen wrote:
> When making a filesystem as in:
> 
> mke2fs -E lazy_itable_init=1 -O uninit_bg -b 4096 bigfile 4294967298
> 
> a subsequent fsck would result in:
> 
> Pass 5: Checking group summary information
> Internal error: fudging end of bitmap (3)
> e2fsck: aborted
> 
> This is because check_block_end() was overflowing in the calculation
> for "end", and giving it a value of -1 (0xFFFF....) which eventually
> ended up tripping up a test in ext2fs_fudge_generic_bmap_end,
> 
>         if (end > bitmap->real_end)
>                 return neq;
> 
> Fix another such error in read_bitmaps() as well.
> 
> lib/ext2fs/imager.c likely has similar problems but it looks like
> it has no 64-bit treatment at all yet.

Applied to the e2fsprogs 64-bit patch queue / pu.

	       		 	      	      - Ted