2012-05-13 18:43:37

by Dan Carpenter

[permalink] [raw]
Subject: dereference before check in ext4_move_extents()

Hi, going through some static checker warnings and reporting bugs.
These were introduced a long time ago.

The patch 748de6736c1e: "ext4: online defrag -- Add EXT4_IOC_MOVE_EXT
ioctl" from Jun 17, 2009, leads to the following Smatch complaint:

fs/ext4/move_extent.c:1381 ext4_move_extents()
warn: variable dereferenced before check 'holecheck_path' (see line 1292)

fs/ext4/move_extent.c
1291 ext_prev = ext_cur;
1292 last_extent = mext_next_extent(orig_inode, holecheck_path,
^^^^^^^^^^^^^^
Dereferenced unconditionally inside the mext_next_extent() function.

1293 &ext_cur);
1294 if (last_extent < 0) {
1295 ret1 = last_extent;
1296 break;
1297 }

[snip]

1376 double_down_write_data_sem(orig_inode, donor_inode);
1377 if (ret1 < 0)
1378 break;
1379
1380 /* Decrease buffer counter */
1381 if (holecheck_path)
^^^^^^^^^^^^^^
Checked here.

1382 ext4_ext_drop_refs(holecheck_path);
1383 ret1 = get_ext_path(orig_inode, seq_start, &holecheck_path);

regards,
dan carpenter



2012-05-14 08:31:17

by Akira Fujita

[permalink] [raw]
Subject: Re: dereference before check in ext4_move_extents()

Hi Dan,

> fs/ext4/move_extent.c:1381 ext4_move_extents()
> warn: variable dereferenced before check 'holecheck_path' (see line 1292)

Thanks for reporting.
I tried to reproduce this with Smatch and kernel 3.4-rc7,
but another compile warning turned up.

# make CHECK="/home/KERN/smatch/smatch -p=kernel" C=1 bzImage modules

<snip>

CHECK fs/ext4/block_validity.c
CC fs/ext4/block_validity.o
CHECK fs/ext4/move_extent.c
fs/ext4/move_extent.c:696 mext_replace_branches() warn: variable dereferenced before check 'dext' (see line 686)
CC fs/ext4/move_extent.o
CHECK fs/ext4/mmp.c

I'll fix above compile warning surely, but it's not the original
you reported. Hmm, how can I reproduce yours?

Regards,
Akira Fujita


(2012/05/14 3:43), Dan Carpenter wrote:
> Hi, going through some static checker warnings and reporting bugs.
> These were introduced a long time ago.
>
> The patch 748de6736c1e: "ext4: online defrag -- Add EXT4_IOC_MOVE_EXT
> ioctl" from Jun 17, 2009, leads to the following Smatch complaint:
>
> fs/ext4/move_extent.c:1381 ext4_move_extents()
> warn: variable dereferenced before check 'holecheck_path' (see line 1292)
>
> fs/ext4/move_extent.c
> 1291 ext_prev = ext_cur;
> 1292 last_extent = mext_next_extent(orig_inode, holecheck_path,
> ^^^^^^^^^^^^^^
> Dereferenced unconditionally inside the mext_next_extent() function.
>
> 1293 &ext_cur);
> 1294 if (last_extent< 0) {
> 1295 ret1 = last_extent;
> 1296 break;
> 1297 }
>
> [snip]
>
> 1376 double_down_write_data_sem(orig_inode, donor_inode);
> 1377 if (ret1< 0)
> 1378 break;
> 1379
> 1380 /* Decrease buffer counter */
> 1381 if (holecheck_path)
> ^^^^^^^^^^^^^^
> Checked here.
>
> 1382 ext4_ext_drop_refs(holecheck_path);
> 1383 ret1 = get_ext_path(orig_inode, seq_start,&holecheck_path);
>
> regards,
> dan carpenter
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-ext4" in
> the body of a message to [email protected]
> More majordomo info at http://vger.kernel.org/majordomo-info.html
>


2012-05-14 08:48:16

by Dan Carpenter

[permalink] [raw]
Subject: Re: dereference before check in ext4_move_extents()

On Mon, May 14, 2012 at 05:30:23PM +0900, Akira Fujita wrote:
> I'll fix above compile warning surely, but it's not the original
> you reported. Hmm, how can I reproduce yours?
>

Sorry, I apologize for that. It's some new Smatch stuff I'm working
on. I haven't pushed it yet.

regards,
dan carpenter


2012-05-15 02:02:29

by Akira Fujita

[permalink] [raw]
Subject: Re: dereference before check in ext4_move_extents()

> Sorry, I apologize for that. It's some new Smatch stuff I'm working
> on. I haven't pushed it yet.

Ok, I'll make a patch when new Smatch appears.

Regards,
Akira Fujita

(2012/05/14 17:51), Dan Carpenter wrote:
> On Mon, May 14, 2012 at 05:30:23PM +0900, Akira Fujita wrote:
>> I'll fix above compile warning surely, but it's not the original
>> you reported. Hmm, how can I reproduce yours?
>>
>
> Sorry, I apologize for that. It's some new Smatch stuff I'm working
> on. I haven't pushed it yet.
>
> regards,
> dan carpenter
>
>

2012-05-15 10:47:21

by Dan Carpenter

[permalink] [raw]
Subject: Re: dereference before check in ext4_move_extents()

On Tue, May 15, 2012 at 11:01:55AM +0900, Akira Fujita wrote:
> > Sorry, I apologize for that. It's some new Smatch stuff I'm working
> > on. I haven't pushed it yet.
>
> Ok, I'll make a patch when new Smatch appears.
>

That works. The fix would be to remove the unnecessary NULL check
so there is no rush.

regards,
dan carpenter