The next commit will need to use the bitmap's MIDX (if one exists) to
translate bit positions into pack-relative positions in the source pack.
Ordinarily, we'd use the "midx" field of the bitmap_index struct. But
since that struct is defined within pack-bitmap.c, and our caller is in
a separate compilation unit, we do not have access to the MIDX field.
Instead, add a "from_midx" field to the bitmapped_pack structure so that
we can use that piece of data from outside of pack-bitmap.c. The caller
that uses this new piece of information will be added in the following
commit.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When loading a chained midx, we build up an array of hashes, one per
layer of the chain. But since the chain is also represented by the
linked list of multi_pack_index structs, nobody actually reads this
array. We pass it to add_midx_to_chain(), but the parameters are
completely ignored.
So we can drop those unused parameters. And then we can see that its
sole caller, load_midx_chain_fd_st(), only cares about one layer hash at a
time (for parsing each line and feeding it to the single-layer midx
code). So we can replace the array with a single object_id on the stack.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the rest of the MIDX subsystem and relevant callers have been
updated to learn about how to read and process incremental MIDX chains,
let's finally update the implementation in `write_midx_internal()` to be
able to write incremental MIDX chains.
This new feature is available behind the `--incremental` option for the
`multi-pack-index` builtin, like so:
$ git multi-pack-index write --incremental
The implementation for doing so is relatively straightforward, and boils
down to a handful of different kinds of changes implemented in this
patch:
- The `compute_sorted_entries()` function is taught to reject objects
which appear in any existing MIDX layer.
- Functions like `write_midx_revindex()` are adjusted to write
pack_order values which are offset by the number of objects in the
base MIDX layer.
- The end of `write_midx_internal()` is adjusted to move
non-incremental MIDX files when necessary (i.e. when creating an
incremental chain with an existing non-incremental MIDX in the
repository).
There are a handful of other changes that are introduced, like new
functions to clear incremental MIDX files that are unrelated to the
current chain (using the same "keep_hash" mechanism as in the
non-incremental case).
The tests explicitly exercising the new incremental MIDX feature are
relatively limited for two reasons:
1. Most of the "interesting" behavior is already thoroughly covered in
t5319-multi-pack-index.sh, which handles the core logic of reading
objects through a MIDX.
The new tests in t5334-incremental-multi-pack-index.sh are mostly
focused on creating and destroying incremental MIDXs, as well as
stitching their results together across layers.
2. A new GIT_TEST environment variable is added called
"GIT_TEST_MULTI_PACK_INDEX_WRITE_INCREMENTAL", which modifies the
entire test suite to write incremental MIDXs after repacking when
combined with the "GIT_TEST_MULTI_PACK_INDEX" variable.
This exercises the long tail of other interesting behavior that is
defined implicitly throughout the rest of the CI suite. It is
likewise added to the linux-TEST-vars job.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach the verification implementation used by `git multi-pack-index
verify` to perform verification for incremental MIDX chains by
independently validating each layer within the chain.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the MIDX machinery's internals have been taught to understand
incremental MIDXs over the previous handful of commits, the MIDX
machinery itself can begin reading incremental MIDXs.
(Note that while the on-disk format for incremental MIDXs has been
defined, the writing end has not been implemented. This will take place
in the commit after next.)
The core of this change involves following the order specified in the
MIDX chain in reverse and opening up MIDXs in the chain one-by-one,
adding them to the previous layer's `->base_midx` pointer at each step.
In order to implement this, the `load_multi_pack_index()` function is
taught to call a new `load_multi_pack_index_chain()` function if loading
a non-incremental MIDX failed via `load_multi_pack_index_one()`.
When loading a MIDX chain, `load_midx_chain_fd_st()` reads each line in
the file one-by-one and dispatches calls to
`load_multi_pack_index_one()` to read each layer of the MIDX chain. When
a layer was successfully read, it is added to the MIDX chain by calling
`add_midx_to_chain()` which validates the contents of the `BASE` chunk,
performs some bounds checks on the number of combined packs and objects,
and attaches the new MIDX by assigning its `base_midx` pointer to the
existing part of the chain.
As a supplement to this, introduce a new mode in the test-read-midx
test-tool which allows us to read the information for a specific MIDX in
the chain by specifying its trailing checksum via the command-line
arguments like so:
$ test-tool read-midx .git/objects [checksum]
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function `midx_preferred_pack()` is used to determine the identity
of the preferred pack, which is the identity of a unique pack within
the MIDX which is used as a tie-breaker when selecting from which pack
to represent an object that appears in multiple packs within the MIDX.
Historically we have said that the MIDX's preferred pack has the unique
property that all objects from that pack are represented in the MIDX.
But that isn't quite true: a more precise statement would be that all
objects from that pack *which appear in the MIDX* are selected from that
pack.
This helps us extend the concept of preferred packs across a MIDX chain,
where some object(s) in the preferred pack may appear in other packs
in an earlier MIDX layer, in which case those object(s) will not appear
in a subsequent MIDX layer from either the preferred pack or any other
pack.
Extend the concept of preferred packs by using the pack which represents
the object at the first position in MIDX pseudo-pack order belonging to
the current MIDX layer (i.e., at position 'm->num_objects_in_base').
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the `midx_contains_pack()` versus `midx_locate_pack()` debacle
has been cleaned up, teach the former about how to operate in an
incremental MIDX-aware world in a similar fashion as in previous
commits.
Instead of using either of the two `midx_for_object()` or
`midx_for_pack()` helpers, this function is split into two: one that
determines whether a pack is contained in a single MIDX, and another
which calls the former in a loop over all MIDXs.
This approach does not require that we change any of the implementation
in what is now `midx_contains_pack_1()` as it still operates over a
single MIDX.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 307d75bbe6 (midx: implement `midx_locate_pack()`, 2023-12-14)
introduced `midx_locate_pack()`, which was described at the time as a
complement to the function `midx_contains_pack()` which allowed
callers to determine where in the MIDX lexical order a pack appeared, as
opposed to whether or not it was simply contained.
307d75bbe6 suggests that future patches would be added which would
introduce callers for this new function, but none ever were, meaning the
function has gone unused since its introduction.
Clean this up by in effect reverting 307d75bbe6, which removes the
unused functions and inlines its definition back into
`midx_contains_pack()`.
(Looking back through the list archives when 307d75bbe6 was written,
this was in preparation for this[1] patch from back when we had the
concept of "disjoint" packs while developing multi-pack verbatim reuse.
That concept was abandoned before the series was merged, but I never
dropped what would become 307d75bbe6 from the series, leading to the
state prior to this commit).
[1]: https://lore.kernel.org/git/3019738b52ba8cd78ea696a3b800fa91e722eb66.1701198172.git.me@ttaylorr.com/
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a similar fashion as previous commits, teach the `fill_midx_entry()`
function to work in a incremental MIDX-aware fashion.
This function, unlike others which accept an index into either the
lexical order of objects or packs, takes in an object_id, and attempts
to fill a caller-provided 'struct pack_entry' with the remaining pieces
of information about that object from the MIDX.
The function uses `bsearch_midx()` which fills out the frame-local 'pos'
variable, recording the given object_id's lexical position within the
MIDX chain, if found (if no matching object ID was found, we'll return
immediately without filling out the `pack_entry` structure).
Once given that position, we jump back through the `->base_midx` pointer
to ensure that our `m` points at the MIDX layer which contains the given
object_id (and not an ancestor or descendant of it in the chain). Note
that we can drop the bounds check "if (pos >= m->num_objects)" because
`midx_for_object()` performs this check for us.
After that point, we only need to make two special considerations within
this function:
- First, the pack_int_id returned to us by `nth_midxed_pack_int_id()`
is a position in the concatenated lexical order of packs, so we must
ensure that we subtract `m->num_packs_in_base` before accessing the
MIDX-local `packs` array.
- Second, we must avoid translating the `pos` back to a MIDX-local
index, since we use it as an argument to `nth_midxed_offset()` which
expects a position relative to the concatenated lexical order of
objects.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a similar fashion as in previous commits, teach the function
`nth_midxed_offset()` about incremental MIDXs.
The given object `pos` is used to find the containing MIDX, and
translated back into a MIDX-local position by assigning the return value
of `midx_for_object()` to it.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that the special cases callers of `bsearch_midx()` have been dealt
with, teach `bsearch_midx()` to handle incremental MIDX chains.
The incremental MIDX-aware version of `bsearch_midx()` works by
repeatedly searching for a given OID in each layer along the
`->base_midx` pointer, stopping either when an exact match is found, or
the end of the chain is reached.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `bsearch_midx()` function will be extended in a following commit to
search for the location of a given object ID across all MIDXs in a chain
(or the single non-chain MIDX if no chain is available).
While most callers will naturally want to use the updated
`bsearch_midx()` function, there are a handful of special cases that
will want finer control and will only want to search through a single
MIDX.
For instance, the object abbreviation code, which cares about object IDs
near to where we'd expect to find a match in a MIDX. In that case, we
want to look at the nearby matches in each layer of the MIDX chain, not
just a single one).
Split the more fine-grained control out into a separate function called
`bsearch_one_midx()` which searches only a single MIDX.
At present both `bsearch_midx()` and `bsearch_one_midx()` have identical
behavior, but the following commit will rewrite the former to be aware
of incremental MIDXs for the remaining non-special case callers.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a similar fashion as in previous commits, teach the function
`nth_bitmapped_pack()` about incremental MIDXs by translating the given
`pack_int_id` from the concatenated lexical order to a MIDX-local
lexical position.
When accessing the containing MIDX's array of packs, use the local pack
ID. Likewise, when reading the 'BTMP' chunk, use the MIDX-local offset
when accessing the data within that chunk.
(Note that the both the call to prepare_midx_pack() and the assignment
of bp->pack_int_id both care about the global pack_int_id, so avoid
shadowing the given 'pack_int_id' parameter).
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function `nth_midxed_object_oid()` returns the object ID for a given
object position in the MIDX lexicographic order.
Teach this function to instead operate over the concatenated
lexicographic order defined in an earlier step so that it is able to be
used with incremental MIDXs.
To do this, we need to both (a) adjust the bounds check for the given
'n', as well as record the MIDX-local position after chasing the
`->base_midx` pointer to find the MIDX which contains that object.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function `prepare_midx_pack()` is part of the midx.h API and
loads the pack identified by the MIDX-local 'pack_int_id'. This patch
prepares that function to be aware of an incremental MIDX world.
To do this, introduce the second of the two general purpose helpers
mentioned in the previous commit. This commit introduces
`midx_for_pack()`, which is the pack-specific analog of
`midx_for_object()`, and works in the same fashion.
Like `midx_for_object()`, this function chases down the '->base_midx'
field until it finds the MIDX layer within the chain that contains the
given pack.
Use this function within `prepare_midx_pack()` so that the `pack_int_id`
it expects is now relative to the entire MIDX chain, and that it
prepares the given pack in the appropriate MIDX.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The function `nth_midxed_pack_int_id()` takes in a object position in
MIDX lexicographic order and returns an identifier of the pack from
which that object was selected in the MIDX.
Currently, the given object position is an index into the lexicographic
order of objects in a single MIDX. Change this position to instead refer
into the concatenated lexicographic order of all MIDXs in a MIDX chain.
This has two visible effects within the implementation of
`prepare_midx_pack()`:
- First, the given position is now an index into the concatenated
lexicographic order of all MIDXs in the order in which they appear
in the MIDX chain.
- Second the pack ID returned from this function is now also in the
concatenated order of packs among all layers of the MIDX chain in
the same order that they appear in the MIDX chain.
To do this, introduce the first of two general purpose helpers, this one
being `midx_for_object()`. `midx_for_object()` takes a double pointer to
a `struct multi_pack_index` as well as an object `pos` in terms of the
entire MIDX chain[^1].
The function chases down the '->base_midx' field until it finds the MIDX
layer within the chain that contains the given object. It then:
- modifies the double pointer to point to the containing MIDX, instead
of the tip of the chain, and
- returns the MIDX-local position[^2] at which the given object can be
found.
Use this function within `nth_midxed_pack_int_id()` so that the `pos` it
expects is now relative to the entire MIDX chain, and that it returns
the appropriate pack position for that object.
[^1]: As a reminder, this means that the object is identified among the
objects contained in all layers of the incremental MIDX chain, not any
particular layer. For example, consider MIDX chain with two individual
MIDXs, one with 4 objects and another with 3 objects. If the MIDX with
4 objects appears earlier in the chain, then asking for object 6 would
return the second object in the MIDX with 3 objects.
[^2]: Building on the previous example, asking for object 6 in a MIDX
chain with (4, 3) objects, respectively, this would set the double
pointer to point at the MIDX containing three objects, and would
return an index to the second object within that MIDX.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Use of the `the_repository` variable is deprecated nowadays, and we
slowly but steadily convert the codebase to not use it anymore. Instead,
callers should be passing down the repository to work on via parameters.
It is hard though to prove that a given code unit does not use this
variable anymore. The most trivial case, merely demonstrating that there
is no direct use of `the_repository`, is already a bit of a pain during
code reviews as the reviewer needs to manually verify claims made by the
patch author. The bigger problem though is that we have many interfaces
that implicitly rely on `the_repository`.
Introduce a new `USE_THE_REPOSITORY_VARIABLE` macro that allows code
units to opt into usage of `the_repository`. The intent of this macro is
to demonstrate that a certain code unit does not use this variable
anymore, and to keep it from new dependencies on it in future changes,
be it explicit or implicit
For now, the macro only guards `the_repository` itself as well as
`the_hash_algo`. There are many more known interfaces where we have an
implicit dependency on `the_repository`, but those are not guarded at
the current point in time. Over time though, we should start to add
guards as required (or even better, just remove them).
Define the macro as required in our code units. As expected, most of our
code still relies on the global variable. Nearly all of our builtins
rely on the variable as there is no way yet to pass `the_repository` to
their entry point. For now, declare the macro in "biultin.h" to keep the
required changes at least a little bit more contained.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Both `oidread()` and `oidclr()` use `the_repository` to derive the hash
function that shall be used. Require callers to pass in the hash
algorithm to get rid of this implicit dependency.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit f894081dea (pack-revindex: read multi-pack reverse indexes,
2021-03-30) introduced the `get_midx_rev_filename()` helper (later
modified by commit 60980aed78 (midx.c: write MIDX filenames to
strbuf, 2021-10-26)).
This function returns the location of the classic ".rev" files we used
to write for MIDXs (prior to 95e8383bac (midx.c: make changing the
preferred pack safe, 2022-01-25)), which is always of the form:
$GIT_DIR/objects/pack/multi-pack-index-$HASH.rev
Replace this function with a generic helper that populates a strbuf with
the above form, replacing the ".rev" extension with a caller-provided
argument.
This will allow us to remove a similarly-defined function in the
pack-bitmap code (used to determine the location of a MIDX .bitmap file)
by reimplementing it in terms of `get_midx_filename_ext()`.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
GIt 2.44 introduced a regression that makes the updated code to
barf in repositories with multi-pack index written by older
versions of Git, which has been corrected.
* ps/missing-btmp-fix:
pack-bitmap: gracefully handle missing BTMP chunks
In 0fea6b73f1 (Merge branch 'tb/multi-pack-verbatim-reuse', 2024-01-12)
we have introduced multi-pack verbatim reuse of objects. This series has
introduced a new BTMP chunk, which encodes information about bitmapped
objects in the multi-pack index. Starting with dab60934e3 (pack-bitmap:
pass `bitmapped_pack` struct to pack-reuse functions, 2023-12-14) we use
this information to figure out objects which we can reuse from each of
the packfiles.
One thing that we glossed over though is backwards compatibility with
repositories that do not yet have BTMP chunks in their multi-pack index.
In that case, `nth_bitmapped_pack()` would return an error, which causes
us to emit a warning followed by another error message. These warnings
are visible to users that fetch from a repository:
```
$ git fetch
...
remote: error: MIDX does not contain the BTMP chunk
remote: warning: unable to load pack: 'pack-f6bb7bd71d345ea9fe604b60cab9ba9ece54ffbe.idx', disabling pack-reuse
remote: Enumerating objects: 40, done.
remote: Counting objects: 100% (40/40), done.
remote: Compressing objects: 100% (39/39), done.
remote: Total 40 (delta 5), reused 0 (delta 0), pack-reused 0 (from 0)
...
```
While the fetch succeeds the user is left wondering what they did wrong.
Furthermore, as visible both from the warning and from the reuse stats,
pack-reuse is completely disabled in such repositories.
What is quite interesting is that this issue can even be triggered in
case `pack.allowPackReuse=single` is set, which is the default value.
One could have expected that in this case we fall back to the old logic,
which is to use the preferred packfile without consulting BTMP chunks at
all. But either we fail with the above error in case they are missing,
or we use the first pack in the multi-pack-index. The former case
disables pack-reuse altogether, whereas the latter case may result in
reusing objects from a suboptimal packfile.
Fix this issue by partially reverting the logic back to what we had
before this patch series landed. Namely, in the case where we have no
BTMP chunks or when `pack.allowPackReuse=single` are set, we use the
preferred pack instead of consulting the BTMP chunks.
Helped-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Introduce a new midx-write.c source file, which holds all of the
functionality from the MIDX sub-system related to writing new MIDX files.
Similar to the relationship between "pack-bitmap.c" and
"pack-bitmap-write.c", this source file will hold code that is specific
to writing MIDX files as opposed to reading them (the latter will remain
in midx.c).
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Build the pack base name argument directly using strvec_pushf() instead
of with an intermediate strbuf. This is shorter, simpler and avoids the
need for explicit cleanup.
Signed-off-by: René Scharfe <l.s.r@web.de>
Reviewed-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When performing a binary search over the objects in a MIDX's bitmap
(i.e. in pseudo-pack order), the reader reconstructs the pseudo-pack
ordering using a combination of (a) the preferred pack, (b) the pack's
lexical position in the MIDX based on pack names, and (c) the object
offset within the pack.
In order to perform this binary search, the reader must know the
identity of the preferred pack. This could be stored in the MIDX, but
isn't for historical reasons, mostly because it can easily be inferred
at read-time by looking at the object in the first bit position and
finding out which pack it was selected from in the MIDX, like so:
nth_midxed_pack_int_id(m, pack_pos_to_midx(m, 0));
In midx_to_pack_pos() which performs this binary search, we look up the
identity of the preferred pack before each search. This is relatively
quick, since it involves two table-driven lookups (one in the MIDX's
revindex for `pack_pos_to_midx()`, and another in the MIDX's object
table for `nth_midxed_pack_int_id()`).
But since the preferred pack does not change after the MIDX is written,
it is safe to cache this value on the MIDX itself.
Write a helper to do just that, and rewrite all of the existing
call-sites that care about the identity of the preferred pack in terms
of this new helper.
This will prepare us for a subsequent patch where we will need to binary
search through the MIDX's pseudo-pack order multiple times.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The multi-pack index API exposes a `midx_contains_pack()` function that
takes in a string ending in either ".idx" or ".pack" and returns whether
or not the MIDX contains a given pack corresponding to that string.
There is no corresponding function to locate the position of a pack
within the MIDX's pack order (sorted lexically by pack filename).
We could add an optional out parameter to `midx_contains_pack()` that is
filled out with the pack's position when the parameter is non-NULL. To
minimize the amount of fallout from this change, instead introduce a new
function by renaming `midx_contains_pack()` to `midx_locate_pack()`,
adding that output parameter, and then reimplementing
`midx_contains_pack()` in terms of it.
Future patches will make use of this new function.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When a multi-pack bitmap is used to implement verbatim pack reuse (that
is, when verbatim chunks from an on-disk packfile are copied
directly[^1]), it does so by using its "preferred pack" as the source
for pack-reuse.
This allows repositories to pack the majority of their objects into a
single (often large) pack, and then use it as the single source for
verbatim pack reuse. This increases the amount of objects that are
reused verbatim (and consequently, decrease the amount of time it takes
to generate many packs). But this performance comes at a cost, which is
that the preferred packfile must pace its growth with that of the entire
repository in order to maintain the utility of verbatim pack reuse.
As repositories grow beyond what we can reasonably store in a single
packfile, the utility of verbatim pack reuse diminishes. Or, at the very
least, it becomes increasingly more expensive to maintain as the pack
grows larger and larger.
It would be beneficial to be able to perform this same optimization over
multiple packs, provided some modest constraints (most importantly, that
the set of packs eligible for verbatim reuse are disjoint with respect
to the subset of their objects being sent).
If we assume that the packs which we treat as candidates for verbatim
reuse are disjoint with respect to any of their objects we may output,
we need to make only modest modifications to the verbatim pack-reuse
code itself. Most notably, we need to remove the assumption that the
bits in the reachability bitmap corresponding to objects from the single
reuse pack begin at the first bit position.
Future patches will unwind these assumptions and reimplement their
existing functionality as special cases of the more general assumptions
(e.g. that reuse bits can start anywhere within the bitset, but happen
to start at 0 for all existing cases).
This patch does not yet relax any of those assumptions. Instead, it
implements a foundational data-structure, the "Bitampped Packs" (`BTMP`)
chunk of the multi-pack index. The `BTMP` chunk's contents are described
in detail here. Importantly, the `BTMP` chunk contains information to
map regions of a multi-pack index's reachability bitmap to the packs
whose objects they represent.
For now, this chunk is only written, not read (outside of the test-tool
used in this patch to test the new chunk's behavior). Future patches
will begin to make use of this new chunk.
[^1]: Modulo patching any `OFS_DELTA`'s that cross over a region of the
pack that wasn't used verbatim.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When selecting which packfiles will be written while generating a MIDX,
the MIDX internals fill out a 'struct pack_info' with various pieces of
book-keeping.
Instead of filling out each field of the `pack_info` structure
individually in each of the two spots that modify the array of such
structures (`ctx->info`), extract a common routine that does this for
us.
This reduces the code duplication by a modest amount. But more
importantly, it zero-initializes the structure before assigning values
into it. This hardens us for a future change which will add additional
fields to this structure which (until this patch) was not
zero-initialized.
As a result, any new fields added to the `pack_info` structure need only
be updated in a single location, instead of at each spot within midx.c.
There are no functional changes in this patch.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The pack-objects internals use a packing_data struct to track what
objects are part of the pack(s) being formed.
Since these structures contain allocated fields, failing to
appropriately free() them results in a leak. Plug that leak by
introducing a clear_packing_data() function, and call it in the
appropriate spots.
This is a fairly straightforward leak to plug, since none of the callers
expect to read any values or have any references to parts of the address
space being freed.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The commit-graph, midx, and pack idx on-disk formats all have oid fanout
tables which are fed to bsearch_hash(). If these tables do not increase
monotonically, then the binary search may not only produce bogus values,
it may cause out of bounds reads.
We fixed this for commit graphs in 4169d89645 (commit-graph: check
consistency of fanout table, 2023-10-09). That commit argued that we did
not need to do the same for midx and pack idx files, because they
already did this check. However, that is wrong. We _do_ check the fanout
table for pack idx files when we load them, but we only do so for midx
files when running "git multi-pack-index verify". So it is possible to
get an out-of-bounds read by running a normal command with a specially
crafted midx file.
Let's fix this using the same solution (and roughly the same test) we
did for the commit-graph in 4169d89645. This replaces the same check
from "multi-pack-index verify", because verify uses the same read
routines, we'd bail on reading the midx much sooner now. So let's make
sure to copy its verbose error message.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we load a revindex from disk, we check the size of the file
compared to the number of objects we expect it to have. But when we use
a RIDX chunk stored directly in the midx, we just access the memory
directly. This can lead to out-of-bounds memory access for a corrupted
or malicious multi-pack-index file.
We can catch this by recording the RIDX chunk size, and then checking it
against the expected size when we "load" the revindex. Note that this
check is much simpler than the one that load_revindex_from_disk() does,
because we just have the data array with no header (so we do not need
to account for the header size, and nor do we need to bother validating
the header values).
The test confirms both that we catch this case, and that we continue the
process (the revindex is required to use the midx bitmaps, but we
fallback to a non-bitmap traversal).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we see a large offset bit in the regular midx offset table, we
use the entry as an index into a separate large offset table (just like
a pack idx does). But we don't bounds-check the access to that large
offset table (nor even record its size when we parse the chunk!).
The equivalent code for a regular pack idx is in check_pack_index_ptr().
But things are a bit simpler here because of the chunked format: we can
just check our array index directly.
As a bonus, we can get rid of the st_mult() here. If our array
bounds-check is successful, then we know that the result will fit in a
size_t (and the bounds check uses a division to avoid overflow
entirely).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The object offset chunk has one fixed-size entry for each object in the
midx. But since we don't check its size, we may access out-of-bounds
memory if we see a corrupt or malicious midx file.
Sine the entries are fixed-size, the total length can be known up-front,
and we can just check it while parsing the chunk (this is similar to
what we do when opening pack idx files, which contain a similar offset
table).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The midx reader assumes chunks are aligned to a 4-byte boundary: we
treat the fanout chunk as an array of uint32_t, indexing it to feed the
results to ntohl(). Without aligning the chunks, we may violate the
CPU's alignment constraints. Though many platforms allow this, some do
not. And certanily UBSan will complain, since it is undefined behavior.
Even though most chunks are naturally 4-byte-aligned (because they are
storing uint32_t or larger types), PNAM is not. It stores NUL-terminated
pack names, so you can have a valid chunk with any length. The writing
side handles this by 4-byte-aligning the chunk, introducing a few extra
NULs as necessary. But since we don't check this on the reading side, we
may end up with a misaligned fanout and trigger the undefined behavior.
We have two options here:
1. Swap out ntohl(fanout[i]) for get_be32(fanout+i) everywhere. The
latter handles alignment itself. It's possible that it's slightly
slower (though in practice I'm not sure how true that is,
especially for these code paths which then go on to do a binary
search).
2. Enforce the alignment when reading the chunks. This is easy to do,
since the table-of-contents reader can check it in one spot.
I went with the second option here, just because it places less burden
on maintenance going forward (it is OK to continue using ntohl), and we
know it can't have any performance impact on the actual reads.
The commit-graph code uses the same chunk API. It's usually also 4-byte
aligned, but some chunks are not (like Bloom filter BDAT chunks). So
we'll pass "1" here to allow any alignment. It doesn't suffer from the
same problem as midx with its fanout because the fanout chunk is always
the first (and the rest of the format dictates that the first chunk will
start aligned).
The new test shows the effect on a midx with a misaligned PNAM chunk.
Note that the midx-reading code treats chunk-toc errors as soft, falling
back to the non-midx path rather than calling die(), as we do for other
parsing errors. Arguably we should make all of these behave the same,
but that's out of scope for this patch. For now the test just expects
the fallback behavior.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We parse the pack-name chunk as a series of NUL-terminated strings. But
since we don't look at the chunk size, there's nothing to guarantee that
we don't parse off the end of the chunk (or even off the end of the
mapped file).
We can record the length, and then as we parse make sure that we never
walk past it.
The new test exercises the case, though note that it does not actually
segfault before this patch. It hits a NUL byte somewhere in one of the
other chunks, and comes up with a garbage pack name. You could construct
one that reads out-of-bounds (e.g., a PNAM chunk at the end of file),
but this case is simple and sufficient to check that we detect the
problem.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When reading an on-disk multi-pack-index, we take the number of objects
in the midx from the final value of the fanout table. But we just
blindly assume that the chunk containing the actual oid entries is the
correct size. This can lead to us reading out-of-bounds memory if the
lookup chunk is too small (or if the fanout is corrupted; when they
don't agree we cannot tell which one is wrong).
Note that we bump the assignment of m->num_objects into the fanout
parser callback, so that it's set when we parse the lookup table
(otherwise we'd have to manually record the lookup table size and check
it later).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we load the oid-fanout chunk, our callback checks that its size is
reasonable and returns an error if not. However, the caller only checks
our return value against CHUNK_NOT_FOUND, so we end up ignoring the
error completely! Using a too-small fanout table means we end up
accessing random memory for the fanout and segfault.
We can fix this by checking for any non-zero return value, rather than
just CHUNK_NOT_FOUND, and adjusting our error message to cover both
cases. We could handle each error code individually, but there's not
much point for such a rare case. The extra message produced in the
callback makes it clear what is going on.
The same pattern is used in the adjacent code. Those cases are actually
OK for now because they do not use a custom callback, so the only error
they can get is CHUNK_NOT_FOUND. But let's convert them, as this is an
accident waiting to happen (especially as we convert some of them away
from pair_chunk). The error messages are more verbose, but it should be
rare for a user to see these anyway.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The pair_chunk() function is provided as an easy helper for parsing
chunks that just want a pointer to a set of bytes. But every caller has
a hidden bug: because we return only the pointer without the matching
chunk size, the callers have no clue how many bytes they are allowed to
look at. And as a result, they may read off the end of the mmap'd data
when the on-disk file does not match their expectations.
Since chunk files are typically used for local-repository data like
commit-graph files and midx's, the security implications here are pretty
mild. The worst that can happen is that you hand somebody a corrupted
repository tarball, and running Git on it does an out-of-bounds read and
crashes. So it's worth being more defensive, but we don't need to drop
everything and fix every caller immediately.
I noticed the problem because the pair_chunk_fn() callback does not look
at its chunk_size argument, and wanted to annotate it to silence
-Wunused-parameter. We could do that now, but we'd lose the hint that
this code should be audited and fixed.
So instead, let's set ourselves up for going down that path:
1. Provide a pair_chunk() function that does return the size, which
prepares us for fixing these cases.
2. Rename the existing function to pair_chunk_unsafe(). That gives us
an easy way to grep for cases which still need to be fixed, and the
name should cause anybody adding new calls to think twice before
using it.
There are no callers of the "safe" version yet, but we'll add some in
subsequent patches.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Various offset computation in the code that accesses the packfiles
and other data in the object layer has been hardened against
arithmetic overflow, especially on 32-bit systems.
* tb/object-access-overflow-protection:
commit-graph.c: prevent overflow in `verify_commit_graph()`
commit-graph.c: prevent overflow in `write_commit_graph()`
commit-graph.c: prevent overflow in `merge_commit_graph()`
commit-graph.c: prevent overflow in `split_graph_merge_strategy()`
commit-graph.c: prevent overflow in `load_tree_for_commit()`
commit-graph.c: prevent overflow in `fill_commit_in_graph()`
commit-graph.c: prevent overflow in `fill_commit_graph_info()`
commit-graph.c: prevent overflow in `load_oid_from_graph()`
commit-graph.c: prevent overflow in add_graph_to_chain()
commit-graph.c: prevent overflow in `write_commit_graph_file()`
pack-bitmap.c: ensure that eindex lookups don't overflow
midx.c: prevent overflow in `fill_included_packs_batch()`
midx.c: prevent overflow in `write_midx_internal()`
midx.c: store `nr`, `alloc` variables as `size_t`'s
midx.c: prevent overflow in `nth_midxed_offset()`
midx.c: prevent overflow in `nth_midxed_object_oid()`
midx.c: use `size_t`'s for fanout nr and alloc
packfile.c: use checked arithmetic in `nth_packed_object_offset()`
packfile.c: prevent overflow in `load_idx()`
packfile.c: prevent overflow in `nth_packed_object_id()`
In a similar spirit as in previous commits, avoid an integer overflow
when computing the expected size of a MIDX.
(Note that this is also OK as-is, since `p->pack_size` is an `off_t`, so
this computation should already be done as 64-bit integers. But again,
let's use `st_mult()` to make this fact clear).
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When writing a MIDX, we use the chunk-format API to write out each
individual chunk of the MIDX. Each chunk of the MIDX is tracked via a
call to `add_chunk()`, along with the expected size of that chunk.
Guard against overflow when dealing with a MIDX with a large number of
entries (and consequently, large chunks within the MIDX file itself) to
avoid corrupting the contents of the MIDX itself.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the `write_midx_context` structure, we use two `uint32_t`'s to track
the length and allocated size of the packs, and one `uint32_t` to track
the number of objects in the MIDX.
In practice, having these be 32-bit unsigned values shouldn't cause any
problems since we are unlikely to have that many objects or packs in any
real-world repository. But these values should be `size_t`'s, so change
their type to reflect that.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a similar spirit as previous patches, avoid an overflow when looking
up object offsets in the MIDX's large offset table by guarding the
computation via `st_mult()`.
This instance is also OK as-is, since the left operand is the result of
`sizeof(...)`, which is already a `size_t`. But use `st_mult()` instead
here to make it explicit that this computation is to be performed using
64-bit unsigned integers.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a similar spirit as previous commits, avoid overflow when looking up
an object's OID in a MIDX when its position is greater than
`2^32-1/m->hash_len`.
As usual, it is perfectly OK for a MIDX to have as many as 2^32-1
objects (since we use 32-bit fields to count the number of objects at
each fanout layer). But if we have more than `2^32-1/m->hash_len` number
of objects, we will incorrectly perform the computation using 32-bit
integers, overflowing the result.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `midx_fanout` struct is used to keep track of a set of OIDs
corresponding to each layer of the MIDX's fanout table. It stores an
array of entries, along with the number of entries in the table, and the
allocated size of the array.
Both `nr` and `alloc` are stored as 32-bit unsigned integers. In
practice, this should never cause any problems, since most packs have
far fewer than 2^32-1 objects.
But storing these as `size_t`'s is more appropriate, and prevents us
from accidentally overflowing some result when multiplying or adding to
either of these values. Update these struct members to be `size_t`'s as
appropriate.
Signed-off-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
alloc_nr, ALLOC_GROW, and ALLOC_GROW_BY are commonly used macros for
dynamic array allocation. Moving these macros to git-compat-util.h with
the other alloc macros focuses alloc.[ch] to allocation for Git objects
and additionally allows us to remove inclusions to alloc.h from files
that solely used the above macros.
Signed-off-by: Calvin Wan <calvinwan@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The vast majority of files including object-store.h did not need dir.h
nor khash.h. Split the header into two files, and let most just depend
upon object-store-ll.h, while letting the two callers that need it
depend on the full object-store.h.
After this patch:
$ git grep -h include..object-store | sort | uniq -c
2 #include "object-store.h"
129 #include "object-store-ll.h"
Diff best viewed with `--color-moved`.
Signed-off-by: Elijah Newren <newren@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
More header clean-up.
* en/header-split-cache-h-part-2: (22 commits)
reftable: ensure git-compat-util.h is the first (indirect) include
diff.h: reduce unnecessary includes
object-store.h: reduce unnecessary includes
commit.h: reduce unnecessary includes
fsmonitor: reduce includes of cache.h
cache.h: remove unnecessary headers
treewide: remove cache.h inclusion due to previous changes
cache,tree: move basic name compare functions from read-cache to tree
cache,tree: move cmp_cache_name_compare from tree.[ch] to read-cache.c
hash-ll.h: split out of hash.h to remove dependency on repository.h
tree-diff.c: move S_DIFFTREE_IFXMIN_NEQ define from cache.h
dir.h: move DTYPE defines from cache.h
versioncmp.h: move declarations for versioncmp.c functions from cache.h
ws.h: move declarations for ws.c functions from cache.h
match-trees.h: move declarations for match-trees.c functions from cache.h
pkt-line.h: move declarations for pkt-line.c functions from cache.h
base85.h: move declarations for base85.c functions from cache.h
copy.h: move declarations for copy.c functions from cache.h
server-info.h: move declarations for server-info.c functions from cache.h
packfile.h: move pack_window and pack_entry from cache.h
...
Geometric repacking ("git repack --geometric=<n>") in a repository
that borrows from an alternate object database had various corner
case bugs, which have been corrected.
* ps/fix-geom-repack-with-alternates:
repack: disable writing bitmaps when doing a local repack
repack: honor `-l` when calculating pack geometry
t/helper: allow chmtime to print verbosely without modifying mtime
pack-objects: extend test coverage of `--stdin-packs` with alternates
pack-objects: fix error when same packfile is included and excluded
pack-objects: fix error when packing same pack twice
pack-objects: split out `--stdin-packs` tests into separate file
repack: fix generating multi-pack-index with only non-local packs
repack: fix trying to use preferred pack in alternates
midx: fix segfault with no packs and invalid preferred pack
When asked to write a multi-pack-index the user can specify a preferred
pack that is used as a tie breaker when multiple packs contain the same
objects. When this packfile cannot be found, we just pick the first pack
that is getting tracked by the newly written multi-pack-index as a
fallback.
Picking the fallback can fail in the case where we're asked to write a
multi-pack-index with no packfiles at all: picking the fallback value
will cause a segfault as we blindly index into the array of packfiles,
which would be empty.
Fix this bug by resetting the preferred packfile index to `-1` before
searching for the preferred pack. This fixes the segfault as we already
check for whether the index is `> - 1`. If it is not, we simply don't
pick a preferred packfile at all.
Helped-by: Taylor Blau <me@ttaylorr.com>
Signed-off-by: Patrick Steinhardt <ps@pks.im>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
