linux-ext4 - Re: [PATCH v4 03/10] f2fs: Simplify the handling of cached insensitive names

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite for Android: free password hash cracker in your pocket

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <YnyR6Z/IdLtTkejE@sol.localdomain>
Date:   Wed, 11 May 2022 21:49:45 -0700
From:   Eric Biggers <ebiggers@...nel.org>
To:     Gabriel Krisman Bertazi <krisman@...labora.com>
Cc:     tytso@....edu, adilger.kernel@...ger.ca, jaegeuk@...nel.org,
        linux-ext4@...r.kernel.org, linux-f2fs-devel@...ts.sourceforge.net,
        kernel@...labora.com
Subject: Re: [PATCH v4 03/10] f2fs: Simplify the handling of cached
 insensitive names

On Wed, May 11, 2022 at 03:31:39PM -0400, Gabriel Krisman Bertazi wrote:
> Keeping it as qstr avoids the unnecessary conversion in f2fs_match
> 
> Signed-off-by: Gabriel Krisman Bertazi <krisman@...labora.com>
> ---
>  fs/f2fs/dir.c      | 52 ++++++++++++++++++++++++++++------------------
>  fs/f2fs/f2fs.h     |  3 ++-
>  fs/f2fs/recovery.c |  5 +----
>  3 files changed, 35 insertions(+), 25 deletions(-)
> 
> diff --git a/fs/f2fs/dir.c b/fs/f2fs/dir.c
> index 166f08623362..c2a02003c5b9 100644
> --- a/fs/f2fs/dir.c
> +++ b/fs/f2fs/dir.c
> @@ -81,28 +81,47 @@ int f2fs_init_casefolded_name(const struct inode *dir,
>  {
>  #if IS_ENABLED(CONFIG_UNICODE)
>  	struct super_block *sb = dir->i_sb;
> +	unsigned char *buf;
> +	int len;
>  
>  	if (IS_CASEFOLDED(dir)) {
> -		fname->cf_name.name = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
> +		buf = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
>  					GFP_NOFS, false, F2FS_SB(sb));
> -		if (!fname->cf_name.name)
> -			return -ENOMEM;
> -		fname->cf_name.len = utf8_casefold(sb->s_encoding,
> -						   fname->usr_fname,
> -						   fname->cf_name.name,
> -						   F2FS_NAME_LEN);
> -		if ((int)fname->cf_name.len <= 0) {
> -			kmem_cache_free(f2fs_cf_name_slab, fname->cf_name.name);
> +		if (!buf) {
>  			fname->cf_name.name = NULL;
> +			return -ENOMEM;
> +		}
> +
> +		len = utf8_casefold(sb->s_encoding, fname->usr_fname,
> +				    buf, F2FS_NAME_LEN);
> +
> +		if (len <= 0) {
> +			kmem_cache_free(f2fs_cf_name_slab, buf);
> +			buf = NULL;
>  			if (sb_has_strict_encoding(sb))
>  				return -EINVAL;
>  			/* fall back to treating name as opaque byte sequence */
>  		}
> +		fname->cf_name.name = buf;
> +		fname->cf_name.len = (unsigned int) len;
>  	}
>  #endif

There's some inconsistent behavior above; now sometimes fname->cf_name.name is
set to NULL on failure and sometime it's not.  Also now fname->cf_name.len can
be set to a negative value.

Since struct f2fs_filename is always zero-initialized, how about only setting
the fname->cf_name fields if we actually have a valid value to assign?  I.e.

		buf = f2fs_kmem_cache_alloc(f2fs_cf_name_slab,
					GFP_NOFS, false, F2FS_SB(sb));
		if (!buf)
			return -ENOMEM;

		len = utf8_casefold(sb->s_encoding, fname->usr_fname,
				    buf, F2FS_NAME_LEN);
		if (len <= 0) {
			kmem_cache_free(f2fs_cf_name_slab, buf);
			if (sb_has_strict_encoding(sb))
				return -EINVAL;
			/* fall back to treating name as opaque byte sequence */
			return 0;
		}
		fname->cf_name.name = buf;
		fname->cf_name.len = len;

> +void f2fs_free_casefolded_name(struct f2fs_filename *fname)
> +{
> +#if IS_ENABLED(CONFIG_UNICODE)
> +	unsigned char *buf = (unsigned char *)fname->cf_name.name;
> +
> +	if (buf) {
> +		kmem_cache_free(f2fs_cf_name_slab, buf);
> +		fname->cf_name.name = NULL;
> +	}
> +
> +#endif
> +}

Kernel code usually uses static inline stubs for the !CONFIG_$FOO case in cases
like this, as that causes the function calls to be compiled away to nothing when
they're unneeded.

- Eric