2023-08-10 11:21:20

by Dan Carpenter

[permalink] [raw]
Subject: [bug report] ext4: convert symlink external data block mapping to bdev

Hello Zhang Yi,

The patch 6493792d3299: "ext4: convert symlink external data block
mapping to bdev" from Apr 24, 2022 (linux-next), leads to the
following Smatch static checker warning:

fs/ext4/namei.c:3353 ext4_init_symlink_block()
error: potential NULL/IS_ERR bug 'bh'

fs/ext4/namei.c
3337 static int ext4_init_symlink_block(handle_t *handle, struct inode *inode,
3338 struct fscrypt_str *disk_link)
3339 {
3340 struct buffer_head *bh;
3341 char *kaddr;
3342 int err = 0;
3343
3344 bh = ext4_bread(handle, inode, 0, EXT4_GET_BLOCKS_CREATE);
3345 if (IS_ERR(bh))
3346 return PTR_ERR(bh);

From reading the code, it looks like ext4_bread() can return both error
pointers and NULL. (Second return statement).

3347
3348 BUFFER_TRACE(bh, "get_write_access");
3349 err = ext4_journal_get_write_access(handle, inode->i_sb, bh, EXT4_JTR_NONE);
3350 if (err)
3351 goto out;
3352
--> 3353 kaddr = (char *)bh->b_data;
^^^^
Unchecked dereference

3354 memcpy(kaddr, disk_link->name, disk_link->len);
3355 inode->i_size = disk_link->len - 1;
3356 EXT4_I(inode)->i_disksize = inode->i_size;
3357 err = ext4_handle_dirty_metadata(handle, inode, bh);
3358 out:
3359 brelse(bh);
3360 return err;
3361 }

regards,
dan carpenter


2023-08-10 11:36:38

by Zhang Yi

[permalink] [raw]
Subject: Re: [bug report] ext4: convert symlink external data block mapping to bdev

On 2023/8/10 18:31, Dan Carpenter wrote:
> Hello Zhang Yi,
>
> The patch 6493792d3299: "ext4: convert symlink external data block
> mapping to bdev" from Apr 24, 2022 (linux-next), leads to the
> following Smatch static checker warning:
>
> fs/ext4/namei.c:3353 ext4_init_symlink_block()
> error: potential NULL/IS_ERR bug 'bh'
>
> fs/ext4/namei.c
> 3337 static int ext4_init_symlink_block(handle_t *handle, struct inode *inode,
> 3338 struct fscrypt_str *disk_link)
> 3339 {
> 3340 struct buffer_head *bh;
> 3341 char *kaddr;
> 3342 int err = 0;
> 3343
> 3344 bh = ext4_bread(handle, inode, 0, EXT4_GET_BLOCKS_CREATE);
> 3345 if (IS_ERR(bh))
> 3346 return PTR_ERR(bh);
>
>>From reading the code, it looks like ext4_bread() can return both error
> pointers and NULL. (Second return statement).

Hello, Dan,

After checking the code, we have passed in EXT4_GET_BLOCKS_CREATE to
ext4_bread(), the return value must be an error code or a valid
buffer_head, it's impossible to return NULL. So I think the warning
is a false positive.

Thanks,
Yi.

>
> 3347
> 3348 BUFFER_TRACE(bh, "get_write_access");
> 3349 err = ext4_journal_get_write_access(handle, inode->i_sb, bh, EXT4_JTR_NONE);
> 3350 if (err)
> 3351 goto out;
> 3352
> --> 3353 kaddr = (char *)bh->b_data;
> ^^^^
> Unchecked dereference
>
> 3354 memcpy(kaddr, disk_link->name, disk_link->len);
> 3355 inode->i_size = disk_link->len - 1;
> 3356 EXT4_I(inode)->i_disksize = inode->i_size;
> 3357 err = ext4_handle_dirty_metadata(handle, inode, bh);
> 3358 out:
> 3359 brelse(bh);
> 3360 return err;
> 3361 }
>
> regards,
> dan carpenter
> .
>

2023-08-10 12:20:27

by Dan Carpenter

[permalink] [raw]
Subject: Re: [bug report] ext4: convert symlink external data block mapping to bdev

On Thu, Aug 10, 2023 at 07:25:02PM +0800, Zhang Yi wrote:
> On 2023/8/10 18:31, Dan Carpenter wrote:
> > Hello Zhang Yi,
> >
> > The patch 6493792d3299: "ext4: convert symlink external data block
> > mapping to bdev" from Apr 24, 2022 (linux-next), leads to the
> > following Smatch static checker warning:
> >
> > fs/ext4/namei.c:3353 ext4_init_symlink_block()
> > error: potential NULL/IS_ERR bug 'bh'
> >
> > fs/ext4/namei.c
> > 3337 static int ext4_init_symlink_block(handle_t *handle, struct inode *inode,
> > 3338 struct fscrypt_str *disk_link)
> > 3339 {
> > 3340 struct buffer_head *bh;
> > 3341 char *kaddr;
> > 3342 int err = 0;
> > 3343
> > 3344 bh = ext4_bread(handle, inode, 0, EXT4_GET_BLOCKS_CREATE);
> > 3345 if (IS_ERR(bh))
> > 3346 return PTR_ERR(bh);
> >
> >>From reading the code, it looks like ext4_bread() can return both error
> > pointers and NULL. (Second return statement).
>
> Hello, Dan,
>
> After checking the code, we have passed in EXT4_GET_BLOCKS_CREATE to
> ext4_bread(), the return value must be an error code or a valid
> buffer_head, it's impossible to return NULL. So I think the warning
> is a false positive.
>

Yep. You're right. Thanks for taking a look at this.

Eventually, I will get around to tracking bits set across function
boundaries and that should silence this warning.

regards,
dan carpenter