Received-SPF: pass (google.com: domain of linux-ext4-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 31 Jul 2023 08:41:33 -0700
From:   "Darrick J. Wong" <djwong@kernel.org>
To:     zhangshida <starzhangzsd@gmail.com>
Cc:     tytso@mit.edu, adilger.kernel@dilger.ca,
        linux-ext4@vger.kernel.org, linux-kernel@vger.kernel.org,
        zhangshida@kylinos.cn
Subject: Re: [PATCH v2] ext4: Fix rec_len verify error
Message-ID: <20230731154133.GA11332@frogsfrogsfrogs>
References: <20230731010104.1781335-1-zhangshida@kylinos.cn>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230731010104.1781335-1-zhangshida@kylinos.cn>
Precedence: bulk

On Mon, Jul 31, 2023 at 09:01:04AM +0800, zhangshida wrote:
> From: Shida Zhang <zhangshida@kylinos.cn>
> 
> With the configuration PAGE_SIZE 64k and filesystem blocksize 64k,
> a problem occurred when more than 13 million files were directly created
> under a directory:
> 
> EXT4-fs error (device xx): ext4_dx_csum_set:492: inode #xxxx: comm xxxxx: dir seems corrupt?  Run e2fsck -D.
> EXT4-fs error (device xx): ext4_dx_csum_verify:463: inode #xxxx: comm xxxxx: dir seems corrupt?  Run e2fsck -D.
> EXT4-fs error (device xx): dx_probe:856: inode #xxxx: block 8188: comm xxxxx: Directory index failed checksum
> 
> When enough files are created, the fake_dirent->reclen will be 0xffff.
> it doesn't equal to the blocksize 65536, i.e. 0x10000.
> 
> But it is not the same condition when blocksize equals to 4k.
> when enough file are created, the fake_dirent->reclen will be 0x1000.
> it equals to the blocksize 4k, i.e. 0x1000.
> 
> The problem seems to be related to the limitation of the 16-bit field
> when the blocksize is set to 64k. To address this, Modify the check so
> as to handle it properly.

urughghahrhrhr<shudder>

Sorry that I missed that rec_len is an encoded number, not a plain le16
integer...

> Signed-off-by: Shida Zhang <zhangshida@kylinos.cn>
> ---
> v1->v2:
>   Use a better way to check the condition, as suggested by Andreas.
> 
>  fs/ext4/namei.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
> index 0caf6c730ce3..fffed95f8531 100644
> --- a/fs/ext4/namei.c
> +++ b/fs/ext4/namei.c
> @@ -445,8 +445,9 @@ static struct dx_countlimit *get_dx_countlimit(struct inode *inode,
>  	struct ext4_dir_entry *dp;
>  	struct dx_root_info *root;
>  	int count_offset;
> +	int blocksize = EXT4_BLOCK_SIZE(inode->i_sb);
>  
> -	if (le16_to_cpu(dirent->rec_len) == EXT4_BLOCK_SIZE(inode->i_sb))
> +	if (ext4_rec_len_from_disk(dirent->rec_len, blocksize) == blocksize)
>  		count_offset = 8;
>  	else if (le16_to_cpu(dirent->rec_len) == 12) {

...but what about all the other le16_to_cpu(ext4_dir_entry{,_2}.rec_len)
accesses in this file?  Don't those also need to be converted to
ext4_rec_len_from_disk calls?

Also,
Fixes: dbe89444042ab ("ext4: Calculate and verify checksums for htree nodes")

--D

>  		dp = (struct ext4_dir_entry *)(((void *)dirent) + 12);
> -- 
> 2.27.0
>