Re: Intentionally corrupted vfat fs causing BUG

[OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>]

Al Viro <viro@ZenIV.linux.org.uk> writes:

> On Thu, Oct 23, 2014 at 05:55:33PM +0100, Al Viro wrote:
>> On Fri, Oct 24, 2014 at 01:45:49AM +0900, OGAWA Hirofumi wrote:
>> 
>> > d_splice_alias() calls __d_find_alias() with want_discon==1, so
>> > __d_find_alias() doesn't return dentry, and d_splice_alias() doesn't use
>> > d_move() path, right?
>> 
>> Hmm...  Not in the current mainline (and not because of want_discon - that's
>> gone already).  However, with the fixes I've got in the local tree it
>> will both find and move it - same as d_materialise_unique() would in the
>> current mainline.
>
> Untested interim fix follows; as soon as d_splice_alias()/d_materialise_unique()
> merge happens, we'll be able to clean vfat_lookup() a bit more.
>
> a) don't bother with ->d_time for positives - we only check it for negatives
> anyway.
> b) make sure to set it at unlink and rmdir time - at *that* point soon-to-be
> negative dentry matches then-current directory contents
> c) don't go into renaming of old alias in vfat_lookup() unless it has
> the same parent (which it will, unless we are seeing corrupted image) *and*
> is a non-directory
> d) use (for now) d_materialise_unique() instead of d_splice_alias() - that one
> will do renames of old directory aliases just fine (and pretty soon so will
> d_splice_alias(), but this bug is -stable fodder)

Ah, this calls d_move() for directory? 

I recalled why it didn't change alias if directory. Changing result of
getcwd() may become the surprise of apps... If doesn't surprise, looks
good.

> Signed-off-by: Al Viro <viro@zeniv.linux.org.uk>
> ---
> diff --git a/fs/fat/namei_vfat.c b/fs/fat/namei_vfat.c
> index 6df8d3d..eed856f 100644
> --- a/fs/fat/namei_vfat.c
> +++ b/fs/fat/namei_vfat.c
> @@ -736,17 +736,17 @@ static struct dentry *vfat_lookup(struct inode *dir, struct dentry *dentry,
>  	}
>  
>  	alias = d_find_alias(inode);
> -	if (alias && !vfat_d_anon_disconn(alias)) {
> +	if (alias && alias->d_parent == dentry->d_parent &&
> +	    !S_ISDIR(inode->i_mode) && !vfat_d_anon_disconn(alias)) {
>  		/*
> -		 * This inode has non anonymous-DCACHE_DISCONNECTED
> +		 * This file has non anonymous-DCACHE_DISCONNECTED
>  		 * dentry. This means, the user did ->lookup() by an
>  		 * another name (longname vs 8.3 alias of it) in past.
>  		 *
>  		 * Switch to new one for reason of locality if possible.
>  		 */
>  		BUG_ON(d_unhashed(alias));
> -		if (!S_ISDIR(inode->i_mode))
> -			d_move(alias, dentry);
> +		d_move(alias, dentry);
>  		iput(inode);
>  		mutex_unlock(&MSDOS_SB(sb)->s_lock);
>  		return alias;
> @@ -755,12 +755,9 @@ static struct dentry *vfat_lookup(struct inode *dir, struct dentry *dentry,
>  
>  out:
>  	mutex_unlock(&MSDOS_SB(sb)->s_lock);
> -	dentry->d_time = dentry->d_parent->d_inode->i_version;
> -	dentry = d_splice_alias(inode, dentry);
> -	if (dentry)
> -		dentry->d_time = dentry->d_parent->d_inode->i_version;
> -	return dentry;
> -
> +	if (!inode)
> +		dentry->d_time = dir->i_version;
> +	return d_materialise_unique(dentry, inode);
>  error:
>  	mutex_unlock(&MSDOS_SB(sb)->s_lock);
>  	return ERR_PTR(err);
> @@ -793,7 +790,6 @@ static int vfat_create(struct inode *dir, struct dentry *dentry, umode_t mode,
>  	inode->i_mtime = inode->i_atime = inode->i_ctime = ts;
>  	/* timestamp is already written, so mark_inode_dirty() is unneeded. */
>  
> -	dentry->d_time = dentry->d_parent->d_inode->i_version;
>  	d_instantiate(dentry, inode);
>  out:
>  	mutex_unlock(&MSDOS_SB(sb)->s_lock);
> @@ -824,6 +820,7 @@ static int vfat_rmdir(struct inode *dir, struct dentry *dentry)
>  	clear_nlink(inode);
>  	inode->i_mtime = inode->i_atime = CURRENT_TIME_SEC;
>  	fat_detach(inode);
> +	dentry->d_time = dir->i_version;
>  out:
>  	mutex_unlock(&MSDOS_SB(sb)->s_lock);
>  
> @@ -849,6 +846,7 @@ static int vfat_unlink(struct inode *dir, struct dentry *dentry)
>  	clear_nlink(inode);
>  	inode->i_mtime = inode->i_atime = CURRENT_TIME_SEC;
>  	fat_detach(inode);
> +	dentry->d_time = dir->i_version;
>  out:
>  	mutex_unlock(&MSDOS_SB(sb)->s_lock);
>  
> @@ -889,7 +887,6 @@ static int vfat_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode)
>  	inode->i_mtime = inode->i_atime = inode->i_ctime = ts;
>  	/* timestamp is already written, so mark_inode_dirty() is unneeded. */
>  
> -	dentry->d_time = dentry->d_parent->d_inode->i_version;
>  	d_instantiate(dentry, inode);
>  
>  	mutex_unlock(&MSDOS_SB(sb)->s_lock);

-- 
OGAWA Hirofumi <hirofumi@mail.parknet.co.jp>