From: Nikolay Borisov <nborisov@suse.com>
To: Qu Wenruo <quwenruo.btrfs@gmx.com>, Qu Wenruo <wqu@suse.com>,
linux-btrfs@vger.kernel.org
Cc: dsterba@suse.cz
Subject: Re: [PATCH 1/2] btrfs: tree-checker: Add checker for variable length item
Date: Thu, 2 Nov 2017 14:52:10 +0200 [thread overview]
Message-ID: <40b9f60b-e71b-0893-29e7-750f9d714831@suse.com> (raw)
In-Reply-To: <dd9afa2c-cc6b-3f1a-d5f4-88e53908d444@gmx.com>
On 2.11.2017 14:37, Qu Wenruo wrote:
>
>
> On 2017年11月02日 20:06, Nikolay Borisov wrote:
>>
>>
>> On 1.11.2017 14:14, Qu Wenruo wrote:
>>> For the following types, we have items with variable length:
>>> (With BTRFS_ prefix and _KEY suffix snipped)
>>>
>>> DIR_ITEM
>>> DIR_INDEX
>>> XATTR_ITEM
>>> INODE_REF
>>> INODE_EXTREF
>>> ROOT_REF
>>> ROOT_BACKREF
>>>
>>> They all use @name_len to indicate their name length, and XATTR_ITEM has
>>> extra @data_len to indicate it data length.
>>>
>>> Despite their variable length, it's also possible to have several such
>>> structure inside one item.
>>>
>>> This patch will add checker to ensure:
>>>
>>> 1) No structure header and its data cross item boundary
>>> 2) Except XATTR_ITEM, no structure should have non-zero @data_len
>>>
>>> This checker is especially useful to avoid possible access beyond
>>> boundary for fuzzed image.
>>>> Signed-off-by: Qu Wenruo <wqu@suse.com>
>>> ---
>>> fs/btrfs/tree-checker.c | 123 ++++++++++++++++++++++++++++++++++++++++++++++++
>>> 1 file changed, 123 insertions(+)
>>>
>>> diff --git a/fs/btrfs/tree-checker.c b/fs/btrfs/tree-checker.c
>>> index 114fc5f0ecc5..f26e86fcbd74 100644
>>> --- a/fs/btrfs/tree-checker.c
>>> +++ b/fs/btrfs/tree-checker.c
>>> @@ -222,6 +222,120 @@ static int check_csum_item(struct btrfs_root *root, struct extent_buffer *leaf,
>>> return 0;
>>> }
>>>
>>> +static u32 get_header_size(u8 type)
>>> +{
>>> + switch (type) {
>>> + case BTRFS_DIR_ITEM_KEY:
>>> + case BTRFS_DIR_INDEX_KEY:
>>> + case BTRFS_XATTR_ITEM_KEY:
>>> + return sizeof(struct btrfs_dir_item);
>>> + case BTRFS_INODE_REF_KEY:
>>> + return sizeof(struct btrfs_inode_ref);
>>> + case BTRFS_INODE_EXTREF_KEY:
>>> + return sizeof(struct btrfs_inode_extref);
>>> + case BTRFS_ROOT_REF_KEY:
>>> + case BTRFS_ROOT_BACKREF_KEY:
>>> + return sizeof(struct btrfs_root_ref);
>>> + }
>>> + WARN_ON(1);
>>> + return 0;
>>> +}
>>> +
>>> +static u16 get_header_namelen(struct extent_buffer *leaf, u8 type,
>>> + u32 header_offset)
>>> +{
>>> + /*
>>> + * @header_offset is offset starts after leaf header, while the
>>> + * accessors expects offset starts from leaf header.
>>> + * Sowe need to adds LEAF_DATA_OFFSET here
>>> + */
>>
>> The comment is confusing (but so are the internals of accessing an
>> extent buffer and all the offset magic happening behind the scenes).
>
> The sentence inside the brackets is the point. :)
>
>>
>> You are essentially opencoding btrfs_item_ptr_offset but instead of
>> pointing at the first entry in the data portion, you really point to
>> 'cur' entry and in order to make it "accessor friendly" you do the +
>> BTRFS_LEAF_DATA_OFFSET. I'd say just put something like :
>>
>> "This is the same as btrfs_item_ptr_offset but starting at item data
>> pointed by header_offset". In fact I'd suggest renaming header_offset to
>> data_offset
>
> Only if we can ensure all offsets passed in are starting after header,
> then I'm OK to rename it to "data_offset"
That's one of the things which confused me. Which header:
1. There is the btrfs_leaf which is at the beginning of the of the leaf
item and in order to access a pointer inside the leaf extent_buffer via
the helpers you really want the offset which skips this header and
points to the data item hence your + BTRFS_LEAF_DATA_OFFSET code
2. (The one which I assume you meant) There is the actual struct in the
data portion of the item, which acts as a header for the variable length
data (either name or name + data).
Since this function is only ever called from check_variable_length_item
and it guarantees that the passed value always points inside the data
portion I believe this is fine, no ?
>
> The "header_offset" here is to info any caller that one should pass
> "normal" offset here, to avoid possible LEAF_DATA_OFFSET hassle.
>
> Thanks,
> Qu
>
>>
>>> + unsigned long leaf_offset = header_offset + BTRFS_LEAF_DATA_OFFSET;
>>> +
>>> + switch (type) {
>>> + case BTRFS_DIR_ITEM_KEY:
>>> + case BTRFS_DIR_INDEX_KEY:
>>> + case BTRFS_XATTR_ITEM_KEY:
>>> + return btrfs_dir_name_len(leaf, (void *)leaf_offset);
>>> + case BTRFS_INODE_REF_KEY:
>>> + return btrfs_inode_ref_name_len(leaf, (void *)leaf_offset);
>>> + case BTRFS_INODE_EXTREF_KEY:
>>> + return btrfs_inode_extref_name_len(leaf, (void *)leaf_offset);
>>> + case BTRFS_ROOT_REF_KEY:
>>> + case BTRFS_ROOT_BACKREF_KEY:
>>> + return btrfs_root_ref_name_len(leaf, (void *)leaf_offset);
>>> + }
>>> + WARN_ON(1);
>>> + return 0;
>>> +}
>>> +
>>> +static u16 get_header_datalen(struct extent_buffer *leaf, u8 type,
>>> + unsigned long header_offset)
>>> +{
>>> + /* Same as get_header_namelen */
>>> + unsigned long leaf_offset = header_offset + BTRFS_LEAF_DATA_OFFSET;
>>> +
>>> + switch (type) {
>>> + case BTRFS_DIR_ITEM_KEY:
>>> + case BTRFS_DIR_INDEX_KEY:
>>> + case BTRFS_XATTR_ITEM_KEY:
>>> + return btrfs_dir_data_len(leaf, (void *)leaf_offset);
>>> + }
>>> + return 0;
>>> +}
>>> +
>>> +/*
>>> + * For items with variable length, normally with namelen and tailing data.
>>> + * Like INODE_REF or XATTR
>>> + */
>>> +static int check_variable_length_item(struct btrfs_root *root,
>>> + struct extent_buffer *leaf,
>>> + struct btrfs_key *key, int slot)
>>> +{
>>> + u8 type = key->type;
>>> + u32 item_start = btrfs_item_offset_nr(leaf, slot);
>>> + u32 item_end = btrfs_item_end_nr(leaf, slot);
>>> + u32 header_size = get_header_size(type);
>>> + u32 total_size;
>>> + u32 cur = item_start;
>>
>> nit: Rename cur to cur_offset to make it clear this is an offset.
>>
>>> +
>>> + while (cur < item_end) {
>>> + u32 namelen;
>>> + u32 datalen;
>>> +
>>> + /* header itself should not cross item boundary */
>>> + if (cur + header_size > item_end) {
>>> + generic_err(root, leaf, slot,
>>> + "structure header crosses item boundary, have %u expect (%u, %u]",
>>> + cur + header_size, cur, item_end);
>>> + return -EUCLEAN;
>>> + }
>>> +
>>> + namelen = get_header_namelen(leaf, type, cur);
>>> + datalen = get_header_datalen(leaf, type, cur);
>>> +
>>> + /* Only XATTR can own data */
>>> + if (type != BTRFS_XATTR_ITEM_KEY && datalen) {
>>> + generic_err(root, leaf, slot,
>>> + "item has invalid data len, have %u expect 0",
>>> + datalen);
>>> + return -EUCLEAN;
>>> + }
>>> +
>>> + total_size = header_size + namelen + datalen;
>>> +
>>> + /* header and name/data should not cross item boundary */
>>> + if (cur + total_size > item_end) {
>>> + generic_err(root, leaf, slot,
>>> + "structure data crosses item boundary, have %u expect (%u, %u]",
>>> + cur + total_size, cur + header_size, item_end);
>>> + return -EUCLEAN;
>>> + }
>>> +
>>> + cur += total_size;
>>> + }
>>> + return 0;
>>> +}
>>> +
>>> /*
>>> * Common point to switch the item-specific validation.
>>> */
>>> @@ -238,6 +352,15 @@ static int check_leaf_item(struct btrfs_root *root,
>>> case BTRFS_EXTENT_CSUM_KEY:
>>> ret = check_csum_item(root, leaf, key, slot);
>>> break;
>>> + case BTRFS_DIR_ITEM_KEY:
>>> + case BTRFS_XATTR_ITEM_KEY:
>>> + case BTRFS_DIR_INDEX_KEY:
>>> + case BTRFS_INODE_REF_KEY:
>>> + case BTRFS_INODE_EXTREF_KEY:
>>> + case BTRFS_ROOT_REF_KEY:
>>> + case BTRFS_ROOT_BACKREF_KEY:
>>> + ret = check_variable_length_item(root, leaf, key, slot);
>>> + break;
>>> }
>>> return ret;
>>> }
>>>
>> --
>> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
next prev parent reply other threads:[~2017-11-02 12:52 UTC|newest]
Thread overview: 13+ messages / expand[flat|nested] mbox.gz Atom feed top
2017-11-01 12:14 [PATCH 1/2] btrfs: tree-checker: Add checker for variable length item Qu Wenruo
2017-11-01 12:14 ` [PATCH 2/2] btrfs: Cleanup existing name_len checks Qu Wenruo
2017-11-02 12:06 ` [PATCH 1/2] btrfs: tree-checker: Add checker for variable length item Nikolay Borisov
2017-11-02 12:37 ` Qu Wenruo
2017-11-02 12:52 ` Nikolay Borisov [this message]
2017-11-02 12:12 ` Nikolay Borisov
2017-11-02 12:48 ` Qu Wenruo
2017-11-02 12:56 ` Nikolay Borisov
2017-11-02 13:35 ` Qu Wenruo
2017-11-01 12:22 Qu Wenruo
2017-11-07 20:50 ` David Sterba
2017-11-08 0:13 ` Qu Wenruo
2017-11-08 0:51 ` Qu Wenruo
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=40b9f60b-e71b-0893-29e7-750f9d714831@suse.com \
--to=nborisov@suse.com \
--cc=dsterba@suse.cz \
--cc=linux-btrfs@vger.kernel.org \
--cc=quwenruo.btrfs@gmx.com \
--cc=wqu@suse.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).