git/list[1] front-page[2] threads[3] people[4] search[5] about
 

Re: [PATCH] read-cache: avoid misaligned reads in index v4

From
PWPhillip Wood <phillip.wood123@gmail.com>
Date
Sep 25, 2022, 08:25 UTC
Message-ID
<bb3a2470-7ff5-e4a6-040a-96e0e3833978@gmail.com>
In-Reply-To
<pull.1366.git.1663962236069.gitgitgadget@gmail.com>
Hi Victoria
On 23/09/2022 20:43, Victoria Dye via GitGitGadget wrote:
Show 22 quoted lines
> From: Victoria Dye <vdye@github.com>
> 
> The process for reading the index into memory from disk is to first read its
> contents into a single memory-mapped file buffer (type 'char *'), then
> sequentially convert each on-disk index entry into a corresponding incore
> 'cache_entry'. To access the contents of the on-disk entry for processing, a
> moving pointer within the memory-mapped file is cast to type 'struct
> ondisk_cache_entry *'.
> 
> In index v4, the entries in the on-disk index file are written *without*
> aligning their first byte to a 4-byte boundary; entries are a variable
> length (depending on the entry name and whether or not extended flags are
> used). As a result, casting the 'char *' buffer pointer to 'struct
> ondisk_cache_entry *' then accessing its contents in a 'SANITIZE=undefined'
> build can trigger the following error:
> 
>    read-cache.c:1886:46: runtime error: member access within misaligned
>    address <address> for type 'struct ondisk_cache_entry', which requires 4
>    byte alignment
> 
> Avoid this error by reading fields directly from the 'char *' buffer, using
> the 'offsetof' individual fields in 'struct ondisk_cache_entry'.

I was confused as to why this was safe as it means we're still reading the individual fields from unaligned addresses. The reason it is safe is that we use get_bexx() to read the fields and those functions handle unaligned access. I wonder if it is worth adding a note to clarify that if you re-roll.

Best Wishes
Phillip
Show 129 quoted lines
> 
> Reported-by: Jeff King <peff@peff.net>
> Signed-off-by: Victoria Dye <vdye@github.com>
> ---
>      read-cache: avoid misaligned reads in index v4
>      
>      This fixes the bug reported in [1], where unaligned index entries in the
>      memory-mapped index file triggered a 'SANITIZE=undefined' error due to
>      casting to & accessing unaligned data a 4 byte-aligned 'struct
>      ondisk_cache_entry *' type.
>      
>      In addition to the originally-reported 't9210-scalar.sh' now passing in
>      a 'SANITIZE=undefined' build, I did some light testing by first writing
>      a v4 index with a released version of Git (v2.37), then running some
>      index-modifying operations ('git status', 'git add') with this patch's
>      changes, then again running 'git status' with the stable version. I
>      didn't see anything out of the ordinary but, considering how critical
>      "reading the index" is, I'd very much appreciate some extra-thorough
>      reviews on this patch. :)
>      
>      Thanks!
>      
>       * Victoria
>      
>      [1]
>      https://lore.kernel.org/git/YywzNTzd72tox8Z+@coredump.intra.peff.net/
> 
> Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-1366%2Fvdye%2Fbugfix%2Findex-v4-v1
> Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-1366/vdye/bugfix/index-v4-v1
> Pull-Request: https://github.com/gitgitgadget/git/pull/1366
> 
>   read-cache.c | 42 +++++++++++++++++++++++-------------------
>   1 file changed, 23 insertions(+), 19 deletions(-)
> 
> diff --git a/read-cache.c b/read-cache.c
> index b09128b1884..d16eb979060 100644
> --- a/read-cache.c
> +++ b/read-cache.c
> @@ -1875,7 +1875,7 @@ static int read_index_extension(struct index_state *istate,
>   
>   static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
>   					    unsigned int version,
> -					    struct ondisk_cache_entry *ondisk,
> +					    const char *ondisk,
>   					    unsigned long *ent_size,
>   					    const struct cache_entry *previous_ce)
>   {
> @@ -1883,7 +1883,7 @@ static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
>   	size_t len;
>   	const char *name;
>   	const unsigned hashsz = the_hash_algo->rawsz;
> -	const uint16_t *flagsp = (const uint16_t *)(ondisk->data + hashsz);
> +	const char *flagsp = ondisk + offsetof(struct ondisk_cache_entry, data) + hashsz;
>   	unsigned int flags;
>   	size_t copy_len = 0;
>   	/*
> @@ -1901,15 +1901,15 @@ static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
>   
>   	if (flags & CE_EXTENDED) {
>   		int extended_flags;
> -		extended_flags = get_be16(flagsp + 1) << 16;
> +		extended_flags = get_be16(flagsp + sizeof(uint16_t)) << 16;
>   		/* We do not yet understand any bit out of CE_EXTENDED_FLAGS */
>   		if (extended_flags & ~CE_EXTENDED_FLAGS)
>   			die(_("unknown index entry format 0x%08x"), extended_flags);
>   		flags |= extended_flags;
> -		name = (const char *)(flagsp + 2);
> +		name = (const char *)(flagsp + 2 * sizeof(uint16_t));
>   	}
>   	else
> -		name = (const char *)(flagsp + 1);
> +		name = (const char *)(flagsp + sizeof(uint16_t));
>   
>   	if (expand_name_field) {
>   		const unsigned char *cp = (const unsigned char *)name;
> @@ -1935,20 +1935,24 @@ static struct cache_entry *create_from_disk(struct mem_pool *ce_mem_pool,
>   
>   	ce = mem_pool__ce_alloc(ce_mem_pool, len);
>   
> -	ce->ce_stat_data.sd_ctime.sec = get_be32(&ondisk->ctime.sec);
> -	ce->ce_stat_data.sd_mtime.sec = get_be32(&ondisk->mtime.sec);
> -	ce->ce_stat_data.sd_ctime.nsec = get_be32(&ondisk->ctime.nsec);
> -	ce->ce_stat_data.sd_mtime.nsec = get_be32(&ondisk->mtime.nsec);
> -	ce->ce_stat_data.sd_dev   = get_be32(&ondisk->dev);
> -	ce->ce_stat_data.sd_ino   = get_be32(&ondisk->ino);
> -	ce->ce_mode  = get_be32(&ondisk->mode);
> -	ce->ce_stat_data.sd_uid   = get_be32(&ondisk->uid);
> -	ce->ce_stat_data.sd_gid   = get_be32(&ondisk->gid);
> -	ce->ce_stat_data.sd_size  = get_be32(&ondisk->size);
> +	ce->ce_stat_data.sd_ctime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime)
> +							+ offsetof(struct cache_time, sec));
> +	ce->ce_stat_data.sd_mtime.sec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime)
> +							+ offsetof(struct cache_time, sec));
> +	ce->ce_stat_data.sd_ctime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ctime)
> +							 + offsetof(struct cache_time, nsec));
> +	ce->ce_stat_data.sd_mtime.nsec = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mtime)
> +							 + offsetof(struct cache_time, nsec));
> +	ce->ce_stat_data.sd_dev   = get_be32(ondisk + offsetof(struct ondisk_cache_entry, dev));
> +	ce->ce_stat_data.sd_ino   = get_be32(ondisk + offsetof(struct ondisk_cache_entry, ino));
> +	ce->ce_mode  = get_be32(ondisk + offsetof(struct ondisk_cache_entry, mode));
> +	ce->ce_stat_data.sd_uid   = get_be32(ondisk + offsetof(struct ondisk_cache_entry, uid));
> +	ce->ce_stat_data.sd_gid   = get_be32(ondisk + offsetof(struct ondisk_cache_entry, gid));
> +	ce->ce_stat_data.sd_size  = get_be32(ondisk + offsetof(struct ondisk_cache_entry, size));
>   	ce->ce_flags = flags & ~CE_NAMEMASK;
>   	ce->ce_namelen = len;
>   	ce->index = 0;
> -	oidread(&ce->oid, ondisk->data);
> +	oidread(&ce->oid, (const unsigned char *)ondisk + offsetof(struct ondisk_cache_entry, data));
>   
>   	if (expand_name_field) {
>   		if (copy_len)
> @@ -2117,12 +2121,12 @@ static unsigned long load_cache_entry_block(struct index_state *istate,
>   	unsigned long src_offset = start_offset;
>   
>   	for (i = offset; i < offset + nr; i++) {
> -		struct ondisk_cache_entry *disk_ce;
>   		struct cache_entry *ce;
>   		unsigned long consumed;
>   
> -		disk_ce = (struct ondisk_cache_entry *)(mmap + src_offset);
> -		ce = create_from_disk(ce_mem_pool, istate->version, disk_ce, &consumed, previous_ce);
> +		ce = create_from_disk(ce_mem_pool, istate->version,
> +				      mmap + src_offset,
> +				      &consumed, previous_ce);
>   		set_index_entry(istate, i, ce);
>   
>   		src_offset += consumed;
> 
> base-commit: 1b3d6e17fe83eb6f79ffbac2f2c61bbf1eaef5f8
Previous: Junio C HamanoNext: Junio C Hamano
Message 9 of 12 in “read-cache: avoid misaligned reads in index v4”
  1. read-cache: avoid misaligned reads in index v4Victoria Dye via GitGitGadget, Sep 23, 2022
  2. Jeff KingSep 23, 2022
  3. Junio C HamanoSep 23, 2022
  4. Victoria DyeSep 26, 2022
  5. Jeff KingSep 26, 2022
  6. Jeff KingSep 26, 2022
  7. Jeff KingSep 26, 2022
  8. Junio C HamanoSep 26, 2022
  9. Phillip WoodSep 25, 2022
  10. Junio C HamanoSep 26, 2022
  11. read-cache: avoid misaligned reads in index v4Victoria Dye via GitGitGadget, Sep 28, 2022
  12. Junio C HamanoSep 28, 2022

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.