Received: from out30-111.freemail.mail.aliyun.com (out30-111.freemail.mail.aliyun.com [115.124.30.111]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.subspace.kernel.org (Postfix) with ESMTPS id EEED5384 for ; Sat, 22 Jun 2024 00:31:58 +0000 (UTC) Authentication-Results: smtp.subspace.kernel.org; arc=none smtp.client-ip=115.124.30.111 ARC-Seal:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1719016322; cv=none; b=HRZZ9c6jtfUM9uDMfdPL96juPNYBHWmzwsY8WzuQt712zOGjhHOEJXjbJfbvMANC4BJO8eodo1X361H7dVjsre+LeBybhKKS1QBC9V7M7oWF81F/N4jcQnDlV4XF20Io8ZCtTIYKhgw20aJh0qEjRvYC8Ep8ENXMOU2+Y3cxYxM= ARC-Message-Signature:i=1; a=rsa-sha256; d=subspace.kernel.org; s=arc-20240116; t=1719016322; c=relaxed/simple; bh=ODnbx8AZLGkfoj4rtwitG40WMKWXT6wrkyohCRqMANI=; h=Message-ID:Date:MIME-Version:Subject:To:Cc:References:From: In-Reply-To:Content-Type; b=VqG5FD+mYkzv5tzDf891d+5exxPqtmNx5EWenoHvc7z09yxUhPBEXhMO7X5uRTPHPBd6pnhgPlfBuKmk5IEYynZ8e5A+BcXS7PcdoN3u28SImLywnJaYoOlPjZ73NEJsYAjJAYIjv3qgQXlU3hdmxHk42JlNduuejPeoKw2tv3o= ARC-Authentication-Results:i=1; smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=linux.alibaba.com; spf=pass smtp.mailfrom=linux.alibaba.com; dkim=pass (1024-bit key) header.d=linux.alibaba.com header.i=@linux.alibaba.com header.b=Y226snLF; arc=none smtp.client-ip=115.124.30.111 Authentication-Results: smtp.subspace.kernel.org; dmarc=pass (p=none dis=none) header.from=linux.alibaba.com Authentication-Results: smtp.subspace.kernel.org; spf=pass smtp.mailfrom=linux.alibaba.com Authentication-Results: smtp.subspace.kernel.org; dkim=pass (1024-bit key) header.d=linux.alibaba.com header.i=@linux.alibaba.com header.b="Y226snLF" DKIM-Signature:v=1; a=rsa-sha256; c=relaxed/relaxed; d=linux.alibaba.com; s=default; t=1719016309; h=Message-ID:Date:MIME-Version:Subject:To:From:Content-Type; bh=YahOcd0GT7H1WMW76zlE3sMvGTtmpJ69A1gJsg09h8s=; b=Y226snLF0GTh03n+uXHPKJvHIhxKVUfAktQZbVvrhs+uveeFZwSbldj7ZL3gWhlr4n1d1/A1k4KVPowUgsa4VlFWPWozMBexV0/4OcPuE9xcPhy/mrrmBW9JYbbbVUiFzGtmW0S+et8RIgeUtvik4F0kKv9Dq5AXBtcxV/VqP8I= X-Alimail-AntiSpam:AC=PASS;BC=-1|-1;BR=01201311R801e4;CH=green;DM=||false|;DS=||;FP=0|-1|-1|-1|0|-1|-1|-1;HT=maildocker-contentspam033037067111;MF=joseph.qi@linux.alibaba.com;NM=1;PH=DS;RN=3;SR=0;TI=SMTPD_---0W8xZ2Xb_1719016308; Received: from 192.168.31.113(mailfrom:joseph.qi@linux.alibaba.com fp:SMTPD_---0W8xZ2Xb_1719016308) by smtp.aliyun-inc.com; Sat, 22 Jun 2024 08:31:49 +0800 Message-ID: <6f97e658-6dcd-47e9-9cf4-3b413cdb5b57@linux.alibaba.com> Date: Sat, 22 Jun 2024 08:31:45 +0800 Precedence: bulk X-Mailing-List: ocfs2-devel@lists.linux.dev List-Id: List-Subscribe: List-Unsubscribe: MIME-Version: 1.0 User-Agent: Mozilla Thunderbird Subject: Re: [PATCH v5] ocfs2: add bounds checking to ocfs2_check_dir_entry() To: lei lu , Heming Zhao Cc: ocfs2-devel@lists.linux.dev References: <20240618065427.143612-1-llfamsec@gmail.com> <1af4ff4a-e963-4969-86cc-27419d326614@suse.com> Content-Language: en-US From: Joseph Qi In-Reply-To: Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit Hi Heming, Any more comments on this version? Thanks, Joseph On 6/18/24 9:09 PM, lei lu wrote: > On Tue, Jun 18, 2024 at 8:04 PM Heming Zhao wrote: >> >> On 6/18/24 14:54, lei lu wrote: >>> This adds sanity checks for ocfs2_dir_entry to make sure all members of >>> ocfs2_dir_entry don't stray beyond valid memory region. >>> >>> Signed-off-by: lei lu >>> --- >>> fs/ocfs2/dir.c | 39 +++++++++++++++++++++++++-------------- >>> 1 file changed, 25 insertions(+), 14 deletions(-) >>> >>> diff --git a/fs/ocfs2/dir.c b/fs/ocfs2/dir.c >>> index d620d4c53c6f..cdd92b4f9bf1 100644 >>> --- a/fs/ocfs2/dir.c >>> +++ b/fs/ocfs2/dir.c >>> @@ -294,13 +294,16 @@ static void ocfs2_dx_dir_name_hash(struct inode *dir, const char *name, int len, >>> * bh passed here can be an inode block or a dir data block, depending >>> * on the inode inline data flag. >>> */ >>> -static int ocfs2_check_dir_entry(struct inode * dir, >>> - struct ocfs2_dir_entry * de, >>> - struct buffer_head * bh, >>> +static int ocfs2_check_dir_entry(struct inode *dir, >>> + struct ocfs2_dir_entry *de, >>> + struct buffer_head *bh, >>> + char *buf, >>> + unsigned int size, >>> unsigned long offset) >>> { >>> const char *error_msg = NULL; >>> const int rlen = le16_to_cpu(de->rec_len); >>> + const unsigned long next_offset = ((char *) de - buf) + rlen; >> >> Hi Lei & Joseph, >> >> I withdraw my previous saying: the rest codes look good to me. >> >> I checked some code. In my view, if the code runs as expected, >> next_offset should be equal to: "((char *) de - bh->b_data) + rlen". > > Hi, > > As you said below, In inline mode, dir entry list starts at > bh->b_data->id2.i_data, > In the dir block mode, dir entry list starts at bh->b_data directly. > > So the prerequisite for checking here is to make sure that buf is the first_de > and size is the de list size. We make sure that de don't stray beyond the de > list regardless of its mode. > > The above patch code refers to the __ext4_check_dir_entry. > >> >> You could add codes here to verify, e.g: >> printk("..."); >> BUG_ON(next_offset != (((char *)de - bh->b_data) + rlen)); >> >> This BUG_ON() will never be triggered. >> >> Please take some time to build and run the test, If my thinking is right, >> you need to send v6 patch. >> >> btw, please don't forget to check my below comments. >> It seems your code also has other bug. >> >> @Joseph, I also need your opinion. >> >>> >>> if (unlikely(rlen < OCFS2_DIR_REC_LEN(1))) >>> error_msg = "rec_len is smaller than minimal"; >>> @@ -308,6 +311,11 @@ static int ocfs2_check_dir_entry(struct inode * dir, >>> error_msg = "rec_len % 4 != 0"; >>> else if (unlikely(rlen < OCFS2_DIR_REC_LEN(de->name_len))) >>> error_msg = "rec_len is too small for name_len"; >>> + else if (unlikely(next_offset > size)) >>> + error_msg = "directory entry overrun"; >>> + else if (unlikely(next_offset > size - OCFS2_DIR_REC_LEN(1)) && >>> + next_offset != size) >>> + error_msg = "directory entry too close to end"; >>> else if (unlikely( >>> ((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)) >>> error_msg = "directory entry across blocks"; >>> @@ -352,16 +360,16 @@ static inline int ocfs2_search_dirblock(struct buffer_head *bh, >>> de_buf = first_de; >>> dlimit = de_buf + bytes; >>> >>> - while (de_buf < dlimit) { >>> + while (de_buf < dlimit - OCFS2_DIR_MEMBER_LEN) { >>> /* this code is executed quadratically often */ >>> /* do minimal checking `by hand' */ >>> >>> de = (struct ocfs2_dir_entry *) de_buf; >>> >>> - if (de_buf + namelen <= dlimit && >>> + if (de->name + namelen <= dlimit && >>> ocfs2_match(namelen, name, de)) { >>> /* found a match - just to be sure, do a full check */ >>> - if (!ocfs2_check_dir_entry(dir, de, bh, offset)) { >>> + if (!ocfs2_check_dir_entry(dir, de, bh, first_de, bytes, offset)) { >> >> >> ocfs2 uses two ways to store dir entry: >> inline mode (OCFS2_INLINE_DATA_FL) & dir block mode >> >> Above function (ocfs2_search_dirblock) handles dir block case, the code >> can work under expectation. > > I think ocfs2_search_dirblock handles both two modes? Because In the caller > ocfs2_find_entry_id, the first_de is from di_bh->b_data->id2.i_data->id_data. > >> >>> ret = -1; >>> goto bail; >>> } >>> @@ -1138,7 +1146,7 @@ static int __ocfs2_delete_entry(handle_t *handle, struct inode *dir, >>> pde = NULL; >>> de = (struct ocfs2_dir_entry *) first_de; >>> while (i < bytes) { >>> - if (!ocfs2_check_dir_entry(dir, de, bh, i)) { >>> + if (!ocfs2_check_dir_entry(dir, de, bh, first_de, bytes, i)) { >>> status = -EIO; >>> mlog_errno(status); >>> goto bail; >>> @@ -1635,7 +1643,7 @@ int __ocfs2_add_entry(handle_t *handle, >>> /* These checks should've already been passed by the >>> * prepare function, but I guess we can leave them >>> * here anyway. */ >>> - if (!ocfs2_check_dir_entry(dir, de, insert_bh, offset)) { >>> + if (!ocfs2_check_dir_entry(dir, de, insert_bh, data_start, size, offset)) { >>> retval = -ENOENT; >>> goto bail; >>> } >>> @@ -1774,7 +1782,8 @@ static int ocfs2_dir_foreach_blk_id(struct inode *inode, >>> } >>> >>> de = (struct ocfs2_dir_entry *) (data->id_data + ctx->pos); >>> - if (!ocfs2_check_dir_entry(inode, de, di_bh, ctx->pos)) { >>> + if (!ocfs2_check_dir_entry(inode, de, di_bh, (char *)data->id_data, >>> + i_size_read(inode), ctx->pos)) { >> >> This function (ocfs2_dir_foreach_blk_id) handles inline mode. >> >> The de & data->id_data are both located in the inode struct, starting at >> offset 0xc8 in ((struct ocfs2_dinode)->id2.i_data). however di_bh->b_data >> points to the start address of the inode. >> >> if ocfs2_check_dir_entry() uses de & data->id_data to compare the offset, it >> misses the 0xc8 space. >> >> struct ocfs2_dinode >> +----------------------------------+ >> | ... | id2.i_data | >> | |---------------------| >> | | |id_data[]| >> ^------------------------^-------^-+ >> | | ... | >> 0x0 0xc8 0xc8+xxx: de >> di_bh->b_data data->id_da >> >> >> (This patch) in ocfs2_check_dir_entry(), it uses "de - data->id_da" to get >> the offset. However, from the above pic, the algorithm is wrong; it forgets >> the [0,0xc8]. >> the correct way: (de - bh->b_data) >> >> the same issue also occurs in ocfs2_find_dir_space_id(). >> >> >> Thanks, >> Heming >> >>> /* On error, skip the f_pos to the end. */ >>> ctx->pos = i_size_read(inode); >>> break; >>> @@ -1867,7 +1876,8 @@ static int ocfs2_dir_foreach_blk_el(struct inode *inode, >>> while (ctx->pos < i_size_read(inode) >>> && offset < sb->s_blocksize) { >>> de = (struct ocfs2_dir_entry *) (bh->b_data + offset); >>> - if (!ocfs2_check_dir_entry(inode, de, bh, offset)) { >>> + if (!ocfs2_check_dir_entry(inode, de, bh, bh->b_data, >>> + sb->s_blocksize, offset)) { >>> /* On error, skip the f_pos to the >>> next block. */ >>> ctx->pos = (ctx->pos | (sb->s_blocksize - 1)) + 1; >>> @@ -3339,7 +3349,7 @@ static int ocfs2_find_dir_space_id(struct inode *dir, struct buffer_head *di_bh, >>> struct super_block *sb = dir->i_sb; >>> struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; >>> struct ocfs2_dir_entry *de, *last_de = NULL; >>> - char *de_buf, *limit; >>> + char *first_de, *de_buf, *limit; >>> unsigned long offset = 0; >>> unsigned int rec_len, new_rec_len, free_space; >>> >>> @@ -3352,14 +3362,15 @@ static int ocfs2_find_dir_space_id(struct inode *dir, struct buffer_head *di_bh, >>> else >>> free_space = dir->i_sb->s_blocksize - i_size_read(dir); >>> >>> - de_buf = di->id2.i_data.id_data; >>> + first_de = di->id2.i_data.id_data; >>> + de_buf = first_de; >>> limit = de_buf + i_size_read(dir); >>> rec_len = OCFS2_DIR_REC_LEN(namelen); >>> >>> while (de_buf < limit) { >>> de = (struct ocfs2_dir_entry *)de_buf; >>> >>> - if (!ocfs2_check_dir_entry(dir, de, di_bh, offset)) { >>> + if (!ocfs2_check_dir_entry(dir, de, di_bh, first_de, i_size_read(dir), offset)) { >>> ret = -ENOENT; >>> goto out; >>> } >>> @@ -3441,7 +3452,7 @@ static int ocfs2_find_dir_space_el(struct inode *dir, const char *name, >>> /* move to next block */ >>> de = (struct ocfs2_dir_entry *) bh->b_data; >>> } >>> - if (!ocfs2_check_dir_entry(dir, de, bh, offset)) { >>> + if (!ocfs2_check_dir_entry(dir, de, bh, bh->b_data, blocksize, offset)) { >>> status = -ENOENT; >>> goto bail; >>> } >>