Sort the linked list of packs directly using llist_mergesort() instead
of building an array, calling qsort(3) and fixing up the list pointers.
This is shorter and less complicated.
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The "git diff --submodule={short,log}" mechanism has been enhanced
to allow "--submodule=diff" to show the patch between the submodule
commits bound to the superproject.
* jk/diff-submodule-diff-inline:
diff: teach diff to display submodule difference with an inline diff
submodule: refactor show_submodule_summary with helper function
submodule: convert show_submodule_summary to use struct object_id *
allow do_submodule_path to work even if submodule isn't checked out
diff: prepare for additional submodule formats
graph: add support for --line-prefix on all graph-aware output
diff.c: remove output_prefix_length field
cache: add empty_tree_oid object and helper function
We may remove elements from the list while we are iterating,
which requires using a second temporary pointer. Otherwise
stepping to the next element of the list might involve
looking at freed memory (which generally works in practice,
as we _just_ freed it, but of course is wrong to rely on;
valgrind notices it).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git clone --resurse-submodules --reference $path $URL" is a way to
reduce network transfer cost by borrowing objects in an existing
$path repository when cloning the superproject from $URL; it
learned to also peek into $path for presense of corresponding
repositories of submodules and borrow objects from there when able.
* sb/submodule-clone-rr:
clone: recursive and reference option triggers submodule alternates
clone: implement optional references
clone: clarify option_reference as required
clone: factor out checking for an alternate path
submodule--helper update-clone: allow multiple references
submodule--helper module-clone: allow multiple references
t7408: merge short tests, factor out testing method
t7408: modernize style
Similar to is_null_oid(), and is_empty_blob_sha1() add an
empty_tree_oid along with helper function is_empty_tree_oid(). For
completeness, also add an "is_empty_tree_sha1()",
"is_empty_blob_sha1()", "is_empty_tree_oid()" and "is_empty_blob_oid()"
helpers.
To ensure we only get one singleton, implement EMPTY_BLOB_SHA1_BIN as
simply getting the hash of empty_blob_oid structure.
Signed-off-by: Jacob Keller <jacob.keller@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The fundamental data structure of the delta base cache is a
hash table mapping pairs of "(packfile, offset)" into
structs containing the actual object data. The hash table
implementation dates back to e5e0161 (Implement a simple
delta_base cache, 2007-03-17), and uses a fixed-size table.
The current size is a hard-coded 256 entries.
Because we need to be able to remove objects from the hash
table, entry lookup does not do any kind of probing to
handle collisions. Colliding items simply replace whatever
is in their slot. As a result, we have fewer usable slots
than even the 256 we allocate. At half full, each new item
has a 50% chance of displacing another one. Or another way
to think about it: every item has a 1/256 chance of being
ejected due to hash collision, without regard to our LRU
strategy.
So it would be interesting to see the effect of increasing
the cache size on the runtime for some common operations. As
with the previous patch, we'll measure "git log --raw" for
tree-only operations, and "git log -Sfoo --raw" for
operations that touch trees and blobs. All times are
wall-clock best-of-3, done against fully packed repos with
--depth=50, and the default core.deltaBaseCacheLimit of
96MB.
Here are timings for various values of MAX_DELTA_CACHE
against git.git (the asterisk marks the minimum time for
each operation):
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ---------
256 0m02.227s 0m12.821s
512 0m02.143s 0m10.602s
1024 0m02.127s 0m08.642s
2048 0m02.148s 0m07.123s
4096 0m02.194s 0m06.448s*
8192 0m02.239s 0m06.504s
16384 0m02.144s* 0m06.502s
32768 0m02.202s 0m06.622s
65536 0m02.230s 0m06.677s
The log-raw case isn't changed much at all here (probably
because our trees just aren't that big in the first place,
or possibly because we have so _few_ trees in git.git that
the 256-entry cache is enough). But once we start putting
blobs in the cache, too, we see a big improvement (almost
50%). The curve levels off around 4096, which means that we
can hold about that many entries before hitting the 96MB
memory limit (or possibly that the workload is small enough
that there is simply no more work to be optimized out by
caching more).
(As a side note, I initially timed my existing git.git pack,
which was a base of --aggressive combined with some pulls on
top. So it had quite a few deeper delta chains. The
256-cache case was more like 15s, and it still dropped to
~6.5s in the same way).
Here are the timings for linux.git:
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ---------
256 0m41.661s 5m12.410s
512 0m39.547s 5m07.920s
1024 0m37.054s 4m54.666s
2048 0m35.871s 4m41.194s*
4096 0m34.646s 4m51.648s
8192 0m33.881s 4m55.342s
16384 0m35.190s 5m00.122s
32768 0m35.060s 4m58.851s
65536 0m33.311s* 4m51.420s
As we grow we see a nice 20% speedup in the tree traversal,
and more modest 10% in the log-S. This is probably an
indication that we are bound less by the number of entries,
and more by the memory limit (more on that below). What is
interesting is that the numbers bounce around a bit;
increasing the number of entries isn't always a strict
improvement.
Partially this is due to noise in the measurement. But it
may also be an indication that our LRU ejection scheme is
not optimal. The smaller cache sizes introduce some
randomness into the ejection (due to collisions), which may
sometimes work in our favor (and sometimes not!).
So what is the optimal setting of MAX_DELTA_CACHE? The
"bouncing" in the linux.git log-S numbers notwithstanding,
it mostly seems like bigger is better. And even if we were
to try to find a "sweet spot", these are just two
repositories, that are not necessarily representative. The
shape of history, the size of trees and blobs, the memory
limit configuration, etc, all will affect the outcome.
Rather than trying to find the "right" number, another
strategy is to just switch to a hash table that can actually
store collisions: namely our hashmap.h implementation.
Here are numbers for that compared to the "best" we saw from
adjusting MAX_DELTA_CACHE:
| log-raw | log-S
| best hashmap | best hashmap
| --------- --------- | --------- ---------
git | 0m02.144s 0m02.144s | 0m06.448s 0m06.688s
linux | 0m33.311s 0m33.092s | 4m41.194s 4m57.172s
We can see the results are similar in most cases, which is
what we'd expect. We're not ejecting due to collisions at
all, so this is purely representing the LRU. So really, we'd
expect this to model most closely the larger values of the
static MAX_DELTA_CACHE limit. And that does seem to be
what's happening, including the "bounce" in the linux log-S
case.
So while the value for that case _isn't_ as good as the
optimal one measured above (which was 2048 entries), given
the bouncing I'm hesitant to suggest that 2048 is any kind
of optimum (not even for linux.git, let alone as a general
rule). The generic hashmap has the appeal that it drops the
number of tweakable numbers by one, which means we can focus
on tuning other elements, like the LRU strategy or the
core.deltaBaseCacheLimit setting.
And indeed, if we bump the cache limit to 1G (which is
probably silly for general use, but maybe something people
with big workstations would want to do), the linux.git log-S
time drops to 3m32s. That's something you really _can't_ do
easily with the static hash table, because the number of
entries needs to grow in proportion to the memory limit (so
2048 is almost certainly not going to be the right value
there).
This patch takes that direction, and drops the static hash
table entirely in favor of using the hashmap.h API.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the delta base cache runs out of allowed memory, it has
to drop entries. It does so by walking an LRU list, dropping
objects until we are under the memory limit. But we actually
walk the list twice: once to drop blobs, and then again to
drop other objects (which are generally trees). This comes
from 18bdec1 (Limit the size of the new delta_base_cache,
2007-03-19).
This performs poorly as the number of entries grows, because
any time dropping blobs does not satisfy the limit, we have
to walk the _entire_ list, trees included, looking for blobs
to drop, before starting to drop any trees.
It's not generally a problem now, as the cache is limited to
only 256 entries. But as we could benefit from increasing
that in a future patch, it's worth looking at how it
performs as the cache size grows. And the answer is "not
well".
The table below shows times for various operations with
different values of MAX_DELTA_CACHE (which is not a run-time
knob; I recompiled with -DMAX_DELTA_CACHE=$n for each).
I chose "git log --raw" ("log-raw" in the table) because it
will access all of the trees, but no blobs at all (so in a
sense it is a worst case for this problem, because we will
always walk over the entire list of trees once before
realizing there are no blobs to drop). This is also
representative of other tree-only operations like "rev-list
--objects" and "git log -- <path>".
I also timed "git log -Sfoo --raw" ("log-S" in the table).
It similarly accesses all of the trees, but also the blobs
for each commit. It's representative of "git log -p", though
it emphasizes the cost of blob access more, as "-S" is
cheaper than computing an actual blob diff.
All timings are best-of-3 wall-clock times (though they all
were CPU bound, so the user CPU times are similar). The
repositories were fully packed with --depth=50, and the
default core.deltaBaseCacheLimit of 96M was in effect. The
current value of MAX_DELTA_CACHE is 256, so I started there
and worked up by factors of 2.
First, here are values for git.git (the asterisk signals the
fastest run for each operation):
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ---------
256 0m02.212s 0m12.634s
512 0m02.136s* 0m10.614s
1024 0m02.156s 0m08.614s
2048 0m02.208s 0m07.062s
4096 0m02.190s 0m06.484s*
8192 0m02.176s 0m07.635s
16384 0m02.913s 0m19.845s
32768 0m03.617s 1m05.507s
65536 0m04.031s 1m18.488s
You can see that for the tree-only log-raw case, we don't
actually benefit that much as the cache grows (all the
differences up through 8192 are basically just noise; this
is probably because we don't actually have that many
distinct trees in git.git). But for log-S, we get a definite
speed improvement as the cache grows, but the improvements
are lost as cache size grows and the linear LRU management
starts to dominate.
Here's the same thing run against linux.git:
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ----------
256 0m40.987s 5m13.216s
512 0m37.949s 5m03.243s
1024 0m35.977s 4m50.580s
2048 0m33.855s 4m39.818s
4096 0m32.913s 4m47.299s*
8192 0m32.176s* 5m14.650s
16384 0m32.185s 6m31.625s
32768 0m38.056s 9m31.136s
65536 1m30.518s 17m38.549s
The pattern is similar, though the effect in log-raw is more
pronounced here. The times dip down in the middle, and then
go back up as we keep growing.
So we know there's a problem. What's the solution?
The obvious one is to improve the data structure to avoid
walking over tree entries during the looking-for-blobs
traversal. We can do this by keeping _two_ LRU lists: one
for blobs, and one for other objects. We drop items from the
blob LRU first, and then from the tree LRU (if necessary).
Here's git.git using that strategy:
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ----------
256 0m02.264s 0m12.830s
512 0m02.201s 0m10.771s
1024 0m02.181s 0m08.593s
2048 0m02.205s 0m07.116s
4096 0m02.158s 0m06.537s*
8192 0m02.213s 0m07.246s
16384 0m02.155s* 0m10.975s
32768 0m02.159s 0m16.047s
65536 0m02.181s 0m16.992s
The upswing on log-raw is gone completely. But log-S still
has it (albeit much better than without this strategy).
Let's see what linux.git shows:
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ---------
256 0m42.519s 5m14.654s
512 0m39.106s 5m04.708s
1024 0m36.802s 4m51.454s
2048 0m34.685s 4m39.378s*
4096 0m33.663s 4m44.047s
8192 0m33.157s 4m50.644s
16384 0m33.090s* 4m49.648s
32768 0m33.458s 4m53.371s
65536 0m33.563s 5m04.580s
The results are similar. The tree-only case again performs
well (not surprising; we're literally just dropping the one
useless walk, and not otherwise changing the cache eviction
strategy at all). But the log-S case again does a bit worse
as the cache grows (though possibly that's within the noise,
which is much larger for this case).
Perhaps this is an indication that the "remove blobs first"
strategy is not actually optimal. The intent of it is to
avoid blowing out the tree cache when we see large blobs,
but it also means we'll throw away useful, recent blobs in
favor of older trees.
Let's run the same numbers without caring about object type
at all (i.e., one LRU list, and always evicting whatever is
at the head, regardless of type).
Here's git.git:
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ---------
256 0m02.227s 0m12.821s
512 0m02.143s 0m10.602s
1024 0m02.127s 0m08.642s
2048 0m02.148s 0m07.123s
4096 0m02.194s 0m06.448s*
8192 0m02.239s 0m06.504s
16384 0m02.144s* 0m06.502s
32768 0m02.202s 0m06.622s
65536 0m02.230s 0m06.677s
Much smoother; there's no dramatic upswing as we increase
the cache size (some remains, though it's small enough that
it's mostly run-to-run noise. E.g., in the log-raw case,
note how 8192 is 50-100ms higher than its neighbors). Note
also that we stop getting any real benefit for log-S after
about 4096 entries; that number will depend on the size of
the repository, the size of the blob entries, and the memory
limit of the cache.
Let's see what linux.git shows for the same strategy:
MAX_DELTA_CACHE log-raw log-S
--------------- --------- ---------
256 0m41.661s 5m12.410s
512 0m39.547s 5m07.920s
1024 0m37.054s 4m54.666s
2048 0m35.871s 4m41.194s*
4096 0m34.646s 4m51.648s
8192 0m33.881s 4m55.342s
16384 0m35.190s 5m00.122s
32768 0m35.060s 4m58.851s
65536 0m33.311s* 4m51.420s
It's similarly good. As with the "separate blob LRU"
strategy, there's a lot of noise on the log-S run here. But
it's certainly not any worse, is possibly a bit better, and
the improvement over "separate blob LRU" on the git.git case
is dramatic.
So it seems like a clear winner, and that's what this patch
implements.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We keep an LRU list of entries for when we need to drop
something from an over-full cache. The list is implemented
as a circular doubly-linked list, which is exactly what
list.h provides. We can save a few lines by using the list.h
macros and functions. More importantly, this makes the code
easier to follow, as the reader sees explicit concepts like
"list_add_tail()" instead of pointer manipulation.
As a bonus, the list_entry() macro lets us place the lru
pointers anywhere inside the delta_base_cache_entry struct
(as opposed to just casting the pointer, which requires it
at the front of the struct). This will be useful in later
patches when we need to place other items at the front of
the struct (e.g., our hashmap implementation requires this).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function drops an entry entirely from the cache,
meaning that aside from the freeing of the buffer, it is
exactly equivalent to detach_delta_base_cache_entry(). Let's
build on top of the detach function, which shortens the code
and will make it simpler when we change out the underlying
storage in future patches.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The delta base cache entries are stored in a fixed-length
hash table. So the way to remove an entry is to "clear" the
slot in the table, and that is what this function does.
However, the name is a leaky abstraction. If we were to
change the hash table implementation, it would no longer be
about "clearing". We should name it after _what_ it does,
not _how_ it does it. I.e., something like "remove" instead
of "clear".
But that does not tell the whole story, either. The subtle
thing about this function is that it removes the entry, but
does not free the entry data. So a more descriptive name is
"detach"; we give ownership of the data buffer to the
caller, and remove any other resources.
This patch uses the name detach_delta_base_cache_entry().
We could further model this after functions like
strbuf_detach(), which pass back all of the detached
information. However, since there are so many bits of
information in the struct (the data, the size, the type),
and so few callers (only one), it's not worth that
awkwardness. The name change and a comment can make the
intent clear.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There is only one caller of cache_or_unpack_entry() and it
always passes 1 for the keep_cache parameter. We can
simplify it by dropping the "!keep_cache" case.
Another call, which did pass 0, was dropped in abe601b
(sha1_file: remove recursion in unpack_entry, 2013-03-27),
as unpack_entry() now does more complicated things than a
simple unpack when there is a cache miss.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In a later patch we want to determine if a path is suitable as an
alternate from other commands than builtin/clone. Move the checking
functionality of `add_one_reference` to `compute_alternate_path` that is
defined in cache.h.
Signed-off-by: Stefan Beller <sbeller@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Some code may have a pack/offset pair for an object, but
would like to look up more information. Using
sha1_object_info() is too heavy-weight; it starts from the
sha1 and has to find the pack again (so not only does it waste
time, it might not even find the same instance).
In some cases, this problem is solved by helpers like
get_size_from_delta(), which is used by pack-objects to take
a shortcut for objects whose packed representation has
already been found. But there's no similar function for
getting the object type, for instance. Rather than introduce
one, let's just make the whole packed_object_info() available.
It is smart enough to spend effort only on the items the
caller wants.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
An all-zero initializer is fine for this struct, but because
the first element is a pointer, call sites need to know to
use "NULL" instead of "0". Otherwise some static checkers
like "sparse" will complain; see d099b71 (Fix some sparse
warnings, 2013-07-18) for example. So let's provide an
initializer to make this easier to get right.
But let's also comment that memset() to zero is explicitly
OK[1]. One of the callers embeds object_info in another
struct which is initialized via memset (expand_data in
builtin/cat-file.c). Since our subset of C doesn't allow
assignment from a compound literal, handling this in any
other way is awkward, so we'd like to keep the ability to
initialize by memset(). By documenting this property, it
should make anybody who wants to change the initializer
think twice before doing so.
There's one other caller of interest. In parse_sha1_header(),
we did not initialize the struct fully in the first place.
This turned out not to be a bug because the sub-function it
calls does not look at any other fields except the ones we
did initialize. But that assumption might not hold in the
future, so it's a dangerous construct. This patch switches
it to initializing the whole struct, which protects us
against unexpected reads of the other fields.
[1] Obviously using memset() to initialize a pointer
violates the C standard, but we long ago decided that it
was an acceptable tradeoff in the real world.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git am -3" calls "git merge-recursive" when it needs to fall back
to a three-way merge; this call has been turned into an internal
subroutine call instead of spawning a separate subprocess.
* js/am-3-merge-recursive-direct:
merge-recursive: flush output buffer even when erroring out
merge_trees(): ensure that the callers release output buffer
merge-recursive: offer an option to retain the output in 'obuf'
merge-recursive: write the commit title in one go
merge-recursive: flush output buffer before printing error messages
am -3: use merge_recursive() directly again
merge-recursive: switch to returning errors instead of dying
merge-recursive: handle return values indicating errors
merge-recursive: allow write_tree_from_memory() to error out
merge-recursive: avoid returning a wholesale struct
merge_recursive: abort properly upon errors
prepare the builtins for a libified merge_recursive()
merge-recursive: clarify code in was_tracked()
die(_("BUG")): avoid translating bug messages
die("bug"): report bugs consistently
t5520: verify that `pull --rebase` shows the helpful advice when failing
"git pack-objects" has a few options that tell it not to pack
objects found in certain packfiles, which require it to scan .idx
files of all available packs. The codepaths involved in these
operations have been optimized for a common case of not having any
non-local pack and/or any .kept pack.
* jk/pack-objects-optim:
pack-objects: compute local/ignore_pack_keep early
pack-objects: break out of want_object loop early
find_pack_entry: replace last_found_pack with MRU cache
add generic most-recently-used list
sha1_file: drop free_pack_by_name
t/perf: add tests for many-pack scenarios
"git pack-objects" and "git index-pack" mostly operate with off_t
when talking about the offset of objects in a packfile, but there
were a handful of places that used "unsigned long" to hold that
value, leading to an unintended truncation.
* nd/pack-ofs-4gb-limit:
fsck: use streaming interface for large blobs in pack
pack-objects: do not truncate result in-pack object size on 32-bit systems
index-pack: correct "offset" type in unpack_entry_data()
index-pack: report correct bad object offsets even if they are large
index-pack: correct "len" type in unpack_data()
sha1_file.c: use type off_t* for object_info->disk_sizep
pack-objects: pass length to check_pack_crc() without truncation
Each pack has an index for looking up entries in O(log n)
time, but if we have multiple packs, we have to scan through
them linearly. This can produce a measurable overhead for
some operations.
We dealt with this long ago in f7c22cc (always start looking
up objects in the last used pack first, 2007-05-30), which
keeps what is essentially a 1-element most-recently-used
cache. In theory, we should be able to do better by keeping
a similar but longer cache, that is the same length as the
pack-list itself.
Since we now have a convenient generic MRU structure, we can
plug it in and measure. Here are the numbers for running
p5303 against linux.git:
Test HEAD^ HEAD
------------------------------------------------------------------------
5303.3: rev-list (1) 31.56(31.28+0.27) 31.30(31.08+0.20) -0.8%
5303.4: repack (1) 40.62(39.35+2.36) 40.60(39.27+2.44) -0.0%
5303.6: rev-list (50) 31.31(31.06+0.23) 31.23(31.00+0.22) -0.3%
5303.7: repack (50) 58.65(69.12+1.94) 58.27(68.64+2.05) -0.6%
5303.9: rev-list (1000) 38.74(38.40+0.33) 31.87(31.62+0.24) -17.7%
5303.10: repack (1000) 367.20(441.80+4.62) 342.00(414.04+3.72) -6.9%
The main numbers of interest here are the rev-list ones
(since that is exercising the normal object lookup code
path). The single-pack case shouldn't improve at all; the
260ms speedup there is just part of the run-to-run noise
(but it's important to note that we didn't make anything
worse with the overhead of maintaining our cache). In the
50-pack case, we see similar results. There may be a slight
improvement, but it's mostly within the noise.
The 1000-pack case does show a big improvement, though. That
carries over to the repack case, as well. Even though we
haven't touched its pack-search loop yet, it does still do a
lot of normal object lookups (e.g., for the internal
revision walk), and so improves.
As a point of reference, I also ran the 1000-pack test
against a version of HEAD^ with the last_found_pack
optimization disabled. It takes ~60s, so that gives an
indication of how much even the single-element cache is
helping.
For comparison, here's a smaller repository, git.git:
Test HEAD^ HEAD
---------------------------------------------------------------------
5303.3: rev-list (1) 1.56(1.54+0.01) 1.54(1.51+0.02) -1.3%
5303.4: repack (1) 1.84(1.80+0.10) 1.82(1.80+0.09) -1.1%
5303.6: rev-list (50) 1.58(1.55+0.02) 1.59(1.57+0.01) +0.6%
5303.7: repack (50) 2.50(3.18+0.04) 2.50(3.14+0.04) +0.0%
5303.9: rev-list (1000) 2.76(2.71+0.04) 2.24(2.21+0.02) -18.8%
5303.10: repack (1000) 13.21(19.56+0.25) 11.66(18.01+0.21) -11.7%
You can see that the percentage improvement is similar.
That's because the lookup we are optimizing is roughly
O(nr_objects * nr_packs). Since the number of packs is
constant in both tests, we'd expect the improvement to be
linear in the number of objects. But the whole process is
also linear in the number of objects, so the improvement
is a constant factor.
The exact improvement does also depend on the contents of
the packs. In p5303, the extra packs all have 5 first-parent
commits in them, which is a reasonable simulation of a
pushed-to repository. But it also means that only 250
first-parent commits are in those packs (compared to almost
50,000 total in linux.git), and the rest are in the huge
"base" pack. So once we start looking at history in taht big
pack, that's where we'll find most everything, and even the
1-element cache gets close to 100% cache hits. You could
almost certainly show better numbers with a more
pathological case (e.g., distributing the objects more
evenly across the packs). But that's simply not that
realistic a scenario, so it makes more sense to focus on
these numbers.
The implementation itself is a straightforward application
of the MRU code. We provide an MRU-ordered list of packs
that shadows the packed_git list. This is easy to do because
we only create and revise the pack list in one place. The
"reprepare" code path actually drops the whole MRU and
replaces it for simplicity. It would be more efficient to
just add new entries, but there's not much point in
optimizing here; repreparing happens rarely, and only after
doing a lot of other expensive work. The key things to keep
optimized are traversal (which is just a normal linked list,
albeit with one extra level of indirection over the regular
packed_git list), and marking (which is a constant number of
pointer assignments, though slightly more than the old
last_found_pack was; it doesn't seem to create a measurable
slowdown, though).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The point of this function is to drop an entry from the
"packed_git" cache that points to a file we might be
overwriting, because our contents may not be the same (and
hence the only caller was pack-objects as it moved a
temporary packfile into place).
In older versions of git, this could happen because the
names of packfiles were derived from the set of objects they
contained, not the actual bits on disk. But since 1190a1a
(pack-objects: name pack files after trailer hash,
2013-12-05), the name reflects the actual bits on disk, and
any two packfiles with the same name can be used
interchangeably.
Dropping this function not only saves a few lines of code,
it makes the lifetime of "struct packed_git" much easier to
reason about: namely, we now do not ever free these structs.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git pack-objects" and "git index-pack" mostly operate with off_t
when talking about the offset of objects in a packfile, but there
were a handful of places that used "unsigned long" to hold that
value, leading to an unintended truncation.
* nd/pack-ofs-4gb-limit:
fsck: use streaming interface for large blobs in pack
pack-objects: do not truncate result in-pack object size on 32-bit systems
index-pack: correct "offset" type in unpack_entry_data()
index-pack: report correct bad object offsets even if they are large
index-pack: correct "len" type in unpack_data()
sha1_file.c: use type off_t* for object_info->disk_sizep
pack-objects: pass length to check_pack_crc() without truncation
The vast majority of error messages in Git's source code which report a
bug use the convention to prefix the message with "BUG:".
As part of cleaning up merge-recursive to stop die()ing except in case of
detected bugs, let's just make the remainder of the bug reports consistent
with the de facto rule.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On 32 bit systems with large file support, unsigned long is 32-bit
while the two offsets in the subtraction expression (pack-objects has
the exact same expression as in sha1_file.c but not shown in diff) are
in 64-bit. If an in-pack object is larger than 2^32 len/datalen is
truncated and we get a misleading "error: bad packed object CRC for
..." as a result.
Use off_t for len and datalen. check_pack_crc() already accepts this
argument as off_t and can deal with 4+ GB.
Noticed-by: Christoph Michelbach <michelbach94@gmail.com>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The experimental "multiple worktree" feature gains more safety to
forbid operations on a branch that is checked out or being actively
worked on elsewhere, by noticing that e.g. it is being rebased.
* nd/worktree-various-heads:
branch: do not rename a branch under bisect or rebase
worktree.c: check whether branch is bisected in another worktree
wt-status.c: split bisect detection out of wt_status_get_state()
worktree.c: check whether branch is rebased in another worktree
worktree.c: avoid referencing to worktrees[i] multiple times
wt-status.c: make wt_status_check_rebase() work on any worktree
wt-status.c: split rebase detection out of wt_status_get_state()
path.c: refactor and add worktree_git_path()
worktree.c: mark current worktree
worktree.c: make find_shared_symref() return struct worktree *
worktree.c: store "id" instead of "git_dir"
path.c: add git_common_path() and strbuf_git_common_path()
dir.c: rename str(n)cmp_icase to fspath(n)cmp
These functions compare two paths that are taken from file system.
Depending on the running file system, paths may need to be compared
case-sensitively or not, and maybe even something else in future. The
current names do not convey that well.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code to read the pack data using the offsets stored in the pack
idx file has been made more carefully check the validity of the
data in the idx.
* jk/pack-idx-corruption-safety:
sha1_file.c: mark strings for translation
use_pack: handle signed off_t overflow
nth_packed_object_offset: bounds-check extended offset
t5313: test bounds-checks of corrupted/malicious pack/idx files
Update various codepaths to avoid manually-counted malloc().
* jk/tighten-alloc: (22 commits)
ewah: convert to REALLOC_ARRAY, etc
convert ewah/bitmap code to use xmalloc
diff_populate_gitlink: use a strbuf
transport_anonymize_url: use xstrfmt
git-compat-util: drop mempcpy compat code
sequencer: simplify memory allocation of get_message
test-path-utils: fix normalize_path_copy output buffer size
fetch-pack: simplify add_sought_entry
fast-import: simplify allocation in start_packfile
write_untracked_extension: use FLEX_ALLOC helper
prepare_{git,shell}_cmd: use argv_array
use st_add and st_mult for allocation size computation
convert trivial cases to FLEX_ARRAY macros
use xmallocz to avoid size arithmetic
convert trivial cases to ALLOC_ARRAY
convert manual allocations to argv_array
argv-array: add detach function
add helpers for allocating flex-array structs
harden REALLOC_ARRAY and xcalloc against size_t overflow
tree-diff: catch integer overflow in combine_diff_path allocation
...
A v2 pack index file can specify an offset within a packfile
of up to 2^64-1 bytes. On a system with a signed 64-bit
off_t, we can represent only up to 2^63-1. This means that a
corrupted .idx file can end up with a negative offset in the
pack code. Our bounds-checking use_pack function looks for
too-large offsets, but not for ones that have wrapped around
to negative. Let's do so, which fixes an out-of-bounds
access demonstrated in t5313.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If a pack .idx file has a corrupted offset for an object, we
may try to access an offset in the .idx or .pack file that
is larger than the file's size. For the .pack case, we have
use_pack() to protect us, which realizes the access is out
of bounds. But if the corrupted value asks us to look in the
.idx file's secondary 64-bit offset table, we blindly add it
to the mmap'd index data and access arbitrary memory.
We can fix this with a simple bounds-check compared to the
size we found when we opened the .idx file.
Note that there's similar code in index-pack that is
triggered only during "index-pack --verify". To support
both, we pull the bounds-check into a separate function,
which dies when it sees a corrupted file.
It would be nice if we could return an error, so that the
pack code could try to find a good copy of the object
elsewhere. Currently nth_packed_object_offset doesn't have
any way to return an error, but it could probably use "0" as
a sentinel value (since no object can start there). This is
the minimal fix, and we can improve the resilience later on
top.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If our size computation overflows size_t, we may allocate a
much smaller buffer than we expected and overflow it. It's
probably impossible to trigger an overflow in most of these
sites in practice, but it is easy enough convert their
additions and multiplications into overflow-checking
variants. This may be fixing real bugs, and it makes
auditing the code easier.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Each of these cases can be converted to use ALLOC_ARRAY or
REALLOC_ARRAY, which has two advantages:
1. It automatically checks the array-size multiplication
for overflow.
2. It always uses sizeof(*array) for the element-size,
so that it can never go out of sync with the declared
type of the array.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
$GIT_OBJECT_DIRECTORY/info/alternates is a text file that can be
edited with a DOS editor. We do not want to use the real path with
CR appended at the end.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The strbuf_getline() interface allows a byte other than LF or NUL as
the line terminator, but this is only because I wrote these
codepaths anticipating that there might be a value other than NUL
and LF that could be useful when I introduced line_termination long
time ago. No useful caller that uses other value has emerged.
By now, it is clear that the interface is overly broad without a
good reason. Many codepaths have hardcoded preference to read
either LF terminated or NUL terminated records from their input, and
then call strbuf_getline() with LF or NUL as the third parameter.
This step introduces two thin wrappers around strbuf_getline(),
namely, strbuf_getline_lf() and strbuf_getline_nul(), and
mechanically rewrites these call sites to call either one of
them. The changes contained in this patch are:
* introduction of these two functions in strbuf.[ch]
* mechanical conversion of all callers to strbuf_getline() with
either '\n' or '\0' as the third parameter to instead call the
respective thin wrapper.
After this step, output from "git grep 'strbuf_getline('" would
become a lot smaller. An interim goal of this series is to make
this an empty set, so that we can have strbuf_getline_crlf() take
over the shorter name strbuf_getline().
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"format-patch" has learned a new option to zero-out the commit
object name on the mbox "From " line.
* bc/format-patch-null-from-line:
format-patch: check that header line has expected format
format-patch: add an option to suppress commit hash
sha1_file.c: introduce a null_oid constant
The helper used to iterate over loose object directories to prune
stale objects did not closedir() immediately when it is done with a
directory--a callback such as the one used for "git prune" may want
to do rmdir(), but it would fail on open directory on platforms
such as WinXP.
* jk/prune-mtime:
prune: close directory earlier during loose-object directory traversal
null_oid is the struct object_id equivalent to null_sha1.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add has_object_file, which is a wrapper around has_sha1_file, but for
struct object_id.
Signed-off-by: brian m. carlson <sandals@crustytoothpaste.net>
Signed-off-by: Jeff King <peff@peff.net>
Having a leftover .idx file without corresponding .pack file in
the repository hurts performance; "git gc" learned to prune them.
* dk/gc-idx-wo-pack:
gc: remove garbage .idx files from pack dir
t5304: test cleaning pack garbage
prepare_packed_git(): refactor garbage reporting in pack directory
Various compilation fixes and squelching of warnings.
* js/misc-fixes:
Correct fscanf formatting string for I64u values
Silence GCC's "cast of pointer to integer of a different size" warning
Squelch warning about an integer overflow
When calculating hashes from pointers, it actually makes sense to cut
off the most significant bits. In that case, said warning does not make
a whole lot of sense.
So let's just work around it by casting the pointer first to intptr_t
and then casting up/down to the final integral type.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Many allocations that is manually counted (correctly) that are
followed by strcpy/sprintf have been replaced with a less error
prone constructs such as xstrfmt.
Macintosh-specific breakage was noticed and corrected in this
reroll.
* jk/war-on-sprintf: (70 commits)
name-rev: use strip_suffix to avoid magic numbers
use strbuf_complete to conditionally append slash
fsck: use for_each_loose_file_in_objdir
Makefile: drop D_INO_IN_DIRENT build knob
fsck: drop inode-sorting code
convert strncpy to memcpy
notes: document length of fanout path with a constant
color: add color_set helper for copying raw colors
prefer memcpy to strcpy
help: clean up kfmclient munging
receive-pack: simplify keep_arg computation
avoid sprintf and strcpy with flex arrays
use alloc_ref rather than hand-allocating "struct ref"
color: add overflow checks for parsing colors
drop strcpy in favor of raw sha1_to_hex
use sha1_to_hex_r() instead of strcpy
daemon: use cld->env_array when re-spawning
stat_tracking_info: convert to argv_array
http-push: use an argv_array for setup_revisions
fetch-pack: use argv_array for index-pack / unpack-objects
...
"git clone --dissociate" runs a big "git repack" process at the
end, and it helps to close file descriptors that are open on the
packs and their idx files before doing so on filesystems that
cannot remove a file that is still open.
* js/clone-dissociate:
clone --dissociate: avoid locking pack files
sha1_file.c: add a function to release all packs
sha1_file: consolidate code to close a pack's file descriptor
t5700: demonstrate a Windows file locking issue with `git clone --dissociate`
On Windows, files that are in use cannot be removed or renamed. That
means that we have to release pack files when we are about to, say,
repack them. Let's introduce a convenient function to close all the
pack files and their idx files.
While at it, we consolidate the close windows/close fd/close index
stanza in `free_pack_by_name()` into the `close_pack()` function that
is used by the new `close_all_packs()` function to avoid repeated code.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There was a lot of repeated code to close the file descriptor of
a given pack. Let's just refactor this code into a single function.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we are allocating a struct with a FLEX_ARRAY member, we
generally compute the size of the array and then sprintf or
strcpy into it. Normally we could improve a dynamic allocation
like this by using xstrfmt, but it doesn't work here; we
have to account for the size of the rest of the struct.
But we can improve things a bit by storing the length that
we use for the allocation, and then feeding it to xsnprintf
or memcpy, which makes it more obvious that we are not
writing more than the allocated number of bytes.
It would be nice if we had some kind of helper for
allocating generic flex arrays, but it doesn't work that
well:
- the call signature is a little bit unwieldy:
d = flex_struct(sizeof(*d), offsetof(d, path), fmt, ...);
You need offsetof here instead of just writing to the
end of the base size, because we don't know how the
struct is packed (partially this is because FLEX_ARRAY
might not be zero, though we can account for that; but
the size of the struct may actually be rounded up for
alignment, and we can't know that).
- some sites do clever things, like over-allocating because
they know they will write larger things into the buffer
later (e.g., struct packed_git here).
So we're better off to just write out each allocation (or
add type-specific helpers, though many of these are one-off
allocations anyway).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When creating a loose object tempfile, we use a fixed
PATH_MAX-sized buffer, and strcpy directly into it. This
isn't buggy, because we do a rough check of the size, but
there's no verification that our guesstimate of the required
space is enough (in fact, it's several bytes too big for the
current naming scheme).
Let's switch to a strbuf, which makes this much easier to
verify. The allocation overhead should be negligible, since
we are replacing a static buffer with a static strbuf, and
we'll only need to allocate on the first call.
While we're here, we can also document a subtle interaction
with mkstemp that would be easy to overlook.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We do some manual memory computation here, and there's no
check that our 60 is not overflowed by the raw sprintf (it
isn't, because the "which" parameter is never longer than
"pack"). We can simplify this greatly with a strbuf.
Technically the end result is not identical, as the original
took care not to rewrite the object directory on each call
for performance reasons. We could do that here, too (by
saving the baselen and resetting to it), but it's not worth
the complexity; this function is not called a lot (generally
once per packfile that we open).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we want to convert "foo.pack" to "foo.idx", we do it by
duplicating the original string and then munging the bytes
in place. Let's use strip_suffix and xstrfmt instead, which
has several advantages:
1. It's more clear what the intent is.
2. It does not implicitly rely on the fact that
strlen(".idx") <= strlen(".pack") to avoid an overflow.
3. We communicate the assumption that the input file ends
with ".pack" (and get a run-time check that this is so).
4. We drop calls to strcpy, which makes auditing the code
base easier.
Likewise, we can do this to convert ".pack" to ".bitmap",
avoiding some manual memory computation.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We have the path "foo.idx", and we create a buffer big
enough to hold "foo.pack" and "foo.keep", and then strcpy
straight into it. This isn't a bug (we have enough space),
but it's very hard to tell from the strcpy that this is so.
Let's instead use strip_suffix to take off the ".idx",
record the size of our allocation, and use xsnprintf to make
sure we don't violate our assumptions.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We generally use 32-byte buffers to format git's "type size"
header fields. These should not generally overflow unless
you can produce some truly gigantic objects (and our types
come from our internal array of constant strings). But it is
a good idea to use xsnprintf to make sure this is the case.
Note that we slightly modify the interface to
write_sha1_file_prepare, which nows uses "hdrlen" as an "in"
parameter as well as an "out" (on the way in it stores the
allocated size of the header, and on the way out it returns
the ultimate size of the header).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This function assumes that the relative_base path passed
into it is no larger than PATH_MAX, and writes into a
fixed-size buffer. However, this path may not have actually
come from the filesystem; for example, add_submodule_odb
generates a path using a strbuf and passes it in. This is
hard to trigger in practice, though, because the long
submodule directory would have to exist on disk before we
would try to open its info/alternates file.
We can easily avoid the bug, though, by simply creating the
filename on the heap.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When trying to see that an object does not exist, a state errno
leaked from our "first try to open a packfile with O_NOATIME and
then if it fails retry without it" logic on a system that refuses
O_NOATIME. This confused us and caused us to die, saying that the
packfile is unreadable, when we should have just reported that the
object does not exist in that packfile to the caller.
* cb/open-noatime-clear-errno:
git_open_noatime: return with errno=0 on success
When trying to see that an object does not exist, a state errno
leaked from our "first try to open a packfile with O_NOATIME and
then if it fails retry without it" logic on a system that refuses
O_NOATIME. This confused us and caused us to die, saying that the
packfile is unreadable, when we should have just reported that the
object does not exist in that packfile to the caller.
* cb/open-noatime-clear-errno:
git_open_noatime: return with errno=0 on success
git_path() and mkpath() are handy helper functions but it is easy
to misuse, as the callers need to be careful to keep the number of
active results below 4. Their uses have been reduced.
* jk/git-path:
memoize common git-path "constant" files
get_repo_path: refactor path-allocation
find_hook: keep our own static buffer
refs.c: remove_empty_directories can take a strbuf
refs.c: avoid git_path assignment in lock_ref_sha1_basic
refs.c: avoid repeated git_path calls in rename_tmp_log
refs.c: simplify strbufs in reflog setup and writing
path.c: drop git_path_submodule
refs.c: remove extra git_path calls from read_loose_refs
remote.c: drop extraneous local variable from migrate_file
prefer mkpathdup to mkpath in assignments
prefer git_pathdup to git_path in some possibly-dangerous cases
add_to_alternates_file: don't add duplicate entries
t5700: modernize style
cache.h: complete set of git_path_submodule helpers
cache.h: clarify documentation for git_path, et al
The hook to report "garbage" files in $GIT_OBJECT_DIRECTORY/pack/
could be generic but is too specific to count-object's needs.
Move the part to produce human-readable messages to count-objects,
and refine the interface to callback with the "bits" with values
defined in the cache.h header file, so that other callers (e.g.
prune) can later use the same mechanism to enumerate different
kinds of garbage files and do something intelligent about them,
other than reporting in textual messages.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In read_sha1_file_extended we die if read_object fails with a fatal
error. We detect a fatal error if errno is non-zero and is not
ENOENT. If the object could not be read because it does not exist,
this is not considered a fatal error and we want to return NULL.
Somewhere down the line, read_object calls git_open_noatime to open
a pack index file, for example. We first try open with O_NOATIME.
If O_NOATIME fails with EPERM, we retry without O_NOATIME. When the
second open succeeds, errno is however still set to EPERM from the
first attempt. When we finally determine that the object does not
exist, read_object returns NULL and read_sha1_file_extended dies
with a fatal error:
fatal: failed to read object <sha1>: Operation not permitted
Fix this by resetting errno to zero before we call open again.
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Signed-off-by: Clemens Buchacher <clemens.buchacher@intel.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
27e1e22d (prune: factor out loose-object directory traversal, 2014-10-16)
introduced a new function for_each_loose_file_in_objdir() with a helper
for_each_file_in_obj_subdir(). The latter calls callbacks for each file
found during a directory traversal and finally also a callback for the
directory itself.
git-prune uses the function to clean up the object directory. In
particular, in the directory callback it calls rmdir(). On Windows XP,
this rmdir call fails, because the directory is still open while the
callback is called. Close the directory before calling the callback.
Signed-off-by: Johannes Sixt <j6t@kdbg.org>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The add_to_alternates_file function blindly uses
hold_lock_file_for_append to copy the existing contents, and
then adds the new line to it. This has two minor problems:
1. We might add duplicate entries, which are ugly and
inefficient.
2. We do not check that the file ends with a newline, in
which case we would bogusly append to the final line.
This is quite unlikely in practice, though, as we call
this function only from git-clone, so presumably we are
the only writers of the file (and we always add a
newline).
Instead of using hold_lock_file_for_append, let's copy the
file line by line, which ensures all records are properly
terminated. If we see an extra line, we can simply abort the
update (there is no point in even copying the rest, as we
know that it would be identical to the original).
As a bonus, we also get rid of some calls to the
static-buffer mkpath and git_path functions.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 5a688fe4 ("core.sharedrepository = 0mode" should set, not
loosen, 2009-03-25), we kept reminding ourselves:
NEEDSWORK: this should be renamed to finalize_temp_file() as
"moving" is only a part of what it does, when no patch between
master to pu changes the call sites of this function.
without doing anything about it. Let's do so.
The purpose of this function was not to move but to finalize. The
detail of the primarily implementation of finalizing was to link the
temporary file to its final name and then to unlink, which wasn't
even "moving". The alternative implementation did "move" by calling
rename(2), which is a fun tangent.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Disable "have we lost a race with competing repack?" check while
receiving a huge object transfer that runs index-pack.
* jk/index-pack-reduce-recheck:
index-pack: avoid excessive re-reading of pack directory
The for_each_packed_object() API function did not iterate over
objects in a packfile that hasn't been used yet.
* jk/maint-for-each-packed-object:
for_each_packed_object: automatically open pack index
When we want to write out a loose object file, we have
always first made sure we don't already have the object
somewhere. Since 33d4221 (write_sha1_file: freshen existing
objects, 2014-10-15), we also update the timestamp on the
file, so that a simultaneous prune knows somebody is
likely to reference it soon.
If our utime() call fails, we treat this the same as not
having the object in the first place; the safe thing to do
is write out another copy. However, the loose-object check
accidentally inverts the utime() check; it returns failure
_only_ when the utime() call actually succeeded. Thus it was
failing to protect us there, and in the normal case where
utime() succeeds, it caused us to pointlessly write out and
link the object.
This passed our freshening tests, because writing out the
new object is certainly _one_ way of updating its utime. So
the normal case was inefficient, but not wrong.
While we're here, let's also drop a comment in front of the
check_and_freshen functions, making a note of their return
type (since it is not our usual "0 for success, -1 for
error").
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The configuration reader/writer uses mmap(2) interface to access
the files; when we find a directory, it barfed with "Out of memory?".
* jk/diagnose-config-mmap-failure:
xmmap(): drop "Out of memory?"
config.c: rewrite ENODEV into EISDIR when mmap fails
config.c: avoid xmmap error messages
config.c: fix mmap leak when writing config
read-cache.c: drop PROT_WRITE from mmap of index
Disable "have we lost a race with competing repack?" check while
receiving a huge object transfer that runs index-pack.
* jk/index-pack-reduce-recheck:
index-pack: avoid excessive re-reading of pack directory
When for_each_packed_object is called, we call
prepare_packed_git() to make sure we have the actual list of
packs. But the latter does not actually open the pack
indices, meaning that pack->nr_objects may simply be 0 if
the pack has not otherwise been used since the program
started.
In practice, this didn't come up for the current callers,
because they iterate the packed objects only after iterating
all reachable objects (so for it to matter you would have to
have a pack consisting only of unreachable objects). But it
is a dangerous and confusing interface that should be fixed
for future callers.
Note that we do not end the iteration when a pack cannot be
opened, but we do return an error. That lets you complete
the iteration even in actively-repacked repository where an
.idx file may racily go away, but it also lets callers know
that they may not have gotten the complete list (which the
current reachability-check caller does care about).
We have to tweak one of the prune tests due to the changed
return value; an earlier test creates bogus .idx files and
does not clean them up. Having to make this tweak is a good
thing; it means we will not prune in a broken repository,
and the test confirms that we do not negatively impact a
more lenient caller, count-objects.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The clean/smudge interface did not work well when filtering an
empty contents (failed and then passed the empty input through).
It can be argued that a filter that produces anything but empty for
an empty input is nonsense, but if the user wants to do strange
things, then why not?
* jh/filter-empty-contents:
sha1_file: pass empty buffer to index empty file
The configuration reader/writer uses mmap(2) interface to access
the files; when we find a directory, it barfed with "Out of memory?".
* jk/diagnose-config-mmap-failure:
xmmap(): drop "Out of memory?"
config.c: rewrite ENODEV into EISDIR when mmap fails
config.c: avoid xmmap error messages
config.c: fix mmap leak when writing config
read-cache.c: drop PROT_WRITE from mmap of index
Since 45e8a74 (has_sha1_file: re-check pack directory before
giving up, 2013-08-30), we spend extra effort for
has_sha1_file to give the right answer when somebody else is
repacking. Usually this effort does not matter, because
after finding that the object does not exist, the next step
is usually to die().
However, some code paths make a large number of
has_sha1_file checks which are _not_ expected to return 1.
The collision test in index-pack.c is such a case. On a
local system, this can cause a performance slowdown of
around 5%. But on a system with high-latency system calls
(like NFS), it can be much worse.
This patch introduces a "quick" flag to has_sha1_file which
callers can use when they would prefer high performance at
the cost of false negatives during repacks. There may be
other code paths that can use this, but the index-pack one
is the most obviously critical, so we'll start with
switching that one.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The clean/smudge interface did not work well when filtering an
empty contents (failed and then passed the empty input through).
It can be argued that a filter that produces anything but empty for
an empty input is nonsense, but if the user wants to do strange
things, then why not?
* jh/filter-empty-contents:
sha1_file: pass empty buffer to index empty file
We show that message with die_errno(), but the OS is ought to know
why mmap(2) failed much better than we do. There is no reason for
us to say "Out of memory?" here.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The config-writing code uses xmmap to map the existing
config file, which will die if the map fails. This has two
downsides:
1. The error message is not very helpful, as it lacks any
context about the file we are mapping:
$ mkdir foo
$ git config --file=foo some.key value
fatal: Out of memory? mmap failed: No such device
2. We normally do not die in this code path; instead, we'd
rather report the error and return an appropriate exit
status (which is part of the public interface
documented in git-config.1).
This patch introduces a "gentle" form of xmmap which lets us
produce our own error message. We do not want to use mmap
directly, because we would like to use the other
compatibility elements of xmmap (e.g., handling 0-length
maps portably).
The end result is:
$ git.compile config --file=foo some.key value
error: unable to mmap 'foo': No such device
$ echo $?
3
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"hash-object --literally" introduced in v2.2 was not prepared to
take a really long object type name.
* jc/hash-object:
write_sha1_file(): do not use a separate sha1[] array
t1007: add hash-object --literally tests
hash-object --literally: fix buffer overrun with extra-long object type
git-hash-object.txt: document --literally option
Add the "--allow-unknown-type" option to "cat-file" to allow
inspecting loose objects of an experimental or a broken type.
* kn/cat-file-literally:
t1006: add tests for git cat-file --allow-unknown-type
cat-file: teach cat-file a '--allow-unknown-type' option
cat-file: make the options mutually exclusive
sha1_file: support reading from a loose object of unknown type
`git add` of an empty file with a filter pops complaints from
`copy_fd` about a bad file descriptor.
This traces back to these lines in sha1_file.c:index_core:
if (!size) {
ret = index_mem(sha1, NULL, size, type, path, flags);
The problem here is that content to be added to the index can be
supplied from an fd, or from a memory buffer, or from a pathname. This
call is supplying a NULL buffer pointer and a zero size.
Downstream logic takes the complete absence of a buffer to mean the
data is to be found elsewhere -- for instance, these, from convert.c:
if (params->src) {
write_err = (write_in_full(child_process.in, params->src, params->size) < 0);
} else {
write_err = copy_fd(params->fd, child_process.in);
}
~If there's a buffer, write from that, otherwise the data must be coming
from an open fd.~
Perfectly reasonable logic in a routine that's going to write from
either a buffer or an fd.
So change `index_core` to supply an empty buffer when indexing an empty
file.
There's a patch out there that instead changes the logic quoted above to
take a `-1` fd to mean "use the buffer", but it seems to me that the
distinction between a missing buffer and an empty one carries intrinsic
semantics, where the logic change is adapting the code to handle
incorrect arguments.
Signed-off-by: Jim Hill <gjthill@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Access to objects in repositories that borrow from another one on a
slow NFS server unnecessarily got more expensive due to recent code
becoming more cautious in a naive way not to lose objects to pruning.
* jk/prune-mtime:
sha1_file: only freshen packs once per run
sha1_file: freshen pack objects before loose
reachable: only mark local objects as recent
"hash-object --literally" introduced in v2.2 was not prepared to
take a really long object type name.
* jc/hash-object:
write_sha1_file(): do not use a separate sha1[] array
t1007: add hash-object --literally tests
hash-object --literally: fix buffer overrun with extra-long object type
git-hash-object.txt: document --literally option
A replacement for contrib/workdir/git-new-workdir that does not
rely on symbolic links and make sharing of objects and refs safer
by making the borrowee and borrowers aware of each other.
* nd/multiple-work-trees: (41 commits)
prune --worktrees: fix expire vs worktree existence condition
t1501: fix test with split index
t2026: fix broken &&-chain
t2026 needs procondition SANITY
git-checkout.txt: a note about multiple checkout support for submodules
checkout: add --ignore-other-wortrees
checkout: pass whole struct to parse_branchname_arg instead of individual flags
git-common-dir: make "modules/" per-working-directory directory
checkout: do not fail if target is an empty directory
t2025: add a test to make sure grafts is working from a linked checkout
checkout: don't require a work tree when checking out into a new one
git_path(): keep "info/sparse-checkout" per work-tree
count-objects: report unused files in $GIT_DIR/worktrees/...
gc: support prune --worktrees
gc: factor out gc.pruneexpire parsing code
gc: style change -- no SP before closing parenthesis
checkout: clean up half-prepared directories in --to mode
checkout: reject if the branch is already checked out elsewhere
prune: strategies for linked checkouts
checkout: support checking out into a new working directory
...
Update sha1_loose_object_info() to optionally allow it to read
from a loose object file of unknown/bogus type; as the function
usually returns the type of the object it read in the form of enum
for known types, add an optional "typename" field to receive the
name of the type in textual form and a flag to indicate the reading
of a loose object file of unknown/bogus type.
Add parse_sha1_header_extended() which acts as a wrapper around
parse_sha1_header() allowing more information to be obtained.
Add unpack_sha1_header_to_strbuf() to unpack sha1 headers of
unknown/corrupt objects which have a unknown sha1 header size to
a strbuf structure. This was written by Junio C Hamano but tested
by me.
Helped-by: Junio C Hamano <gitster@pobox.com>
Helped-by: Eric Sunshine <sunshine@sunshineco.com>
Helped-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Hepled-by: Jeff King <peff@peff.net>
Signed-off-by: Karthik Nayak <karthik.188@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Access to objects in repositories that borrow from another one on a
slow NFS server unnecessarily got more expensive due to recent code
becoming more cautious in a naive way not to lose objects to pruning.
* jk/prune-mtime:
sha1_file: only freshen packs once per run
sha1_file: freshen pack objects before loose
reachable: only mark local objects as recent
In the beginning, write_sha1_file() did not have a way to tell the
caller the name of the object it wrote to the caller. This was
changed in d6d3f9d0 (This implements the new "recursive tree"
write-tree., 2005-04-09) by adding the "returnsha1" parameter to the
function so that the callers who are interested in the value can
optionally pass a pointer to receive it.
It turns out that all callers do want to know the name of the object
it just has written. Nobody passes a NULL to this parameter, hence
it is not necessary to use a separate sha1[] array to receive the
result from write_sha1_file_prepare(), and copy the result to the
returnsha1 supplied by the caller.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"hash-object" learned in 5ba9a93 (hash-object: add --literally
option, 2014-09-11) to allow crafting a corrupt/broken object of
unknown type.
When the user-provided type is particularly long, however, it can
overflow the relatively small stack-based character array handed to
write_sha1_file_prepare() by hash_sha1_file() and write_sha1_file(),
leading to stack corruption (and crash). Introduce a custom helper
to allow arbitrarily long typenames just for "hash-object --literally".
[jc: Eric's original used a strbuf in the more common codepaths, and
I rewrote it to avoid penalizing the non-literally code. Bugs are mine]
Signed-off-by: Eric Sunshine <sunshine@sunshineco.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 33d4221 (write_sha1_file: freshen existing objects,
2014-10-15), we update the mtime of existing objects that we
would have written out (had they not existed). For the
common case in which many objects are packed, we may update
the mtime on a single packfile repeatedly. This can result
in a noticeable performance problem if calling utime() is
expensive (e.g., because your storage is on NFS).
We can fix this by keeping a per-pack flag that lets us
freshen only once per program invocation.
An alternative would be to keep the packed_git.mtime flag up
to date as we freshen, and freshen only once every N
seconds. In practice, it's not worth the complexity. We are
racing against prune expiration times here, which inherently
must be set to accomodate reasonable program running times
(because they really care about the time between an object
being written and it becoming referenced, and the latter is
typically the last step a program takes).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When writing out an object file, we first check whether it
already exists and if so optimize out the write. Prior to
33d4221, we did this by calling has_sha1_file(), which will
check for packed objects followed by loose. Since that
commit, we check loose objects first.
For the common case of a repository whose objects are mostly
packed, this means we will make a lot of extra access()
system calls checking for loose objects. We should follow
the same packed-then-loose order that all of our other
lookups use.
Reported-by: Stefan Saasen <ssaasen@atlassian.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When pruning and repacking a repository that has an
alternate object store configured, we may traverse a large
number of objects in the alternate. This serves no purpose,
and may be expensive to do. A longer explanation is below.
Commits d3038d2 and abcb865 taught prune and pack-objects
(respectively) to treat "recent" objects as tips for
reachability, so that we keep whole chunks of history. They
built on the object traversal in 660c889 (sha1_file: add
for_each iterators for loose and packed objects,
2014-10-15), which covers both local and alternate objects.
In both cases, covering alternate objects is unnecessary, as
both commands can only drop objects from the local
repository. In the case of prune, we traverse only the local
object directory. And in the case of repacking, while we may
or may not include local objects in our pack, we will never
reach into the alternate with "repack -d". The "-l" option
is only a question of whether we are migrating objects from
the alternate into our repository, or leaving them
untouched.
It is possible that we may drop an object that is depended
upon by another object in the alternate. For example,
imagine two repositories, A and B, with A pointing to B as
an alternate. Now imagine a commit that is in B which
references a tree that is only in A. Traversing from recent
objects in B might prevent A from dropping that tree. But
this case isn't worth covering. Repo B should take
responsibility for its own objects. It would never have had
the commit in the first place if it did not also have the
tree, and assuming it is using the same "keep recent chunks
of history" scheme, then it would itself keep the tree, as
well.
So checking the alternate objects is not worth doing, and
come with a significant performance impact. In both cases,
we skip any recent objects that have already been marked
SEEN (i.e., that we know are already reachable for prune, or
included in the pack for a repack). So there is a slight
waste of time in opening the alternate packs at all, only to
notice that we have already considered each object. But much
worse, the alternate repository may have a large number of
objects that are not reachable from the local repository at
all, and we end up adding them to the traversal.
We can fix this by considering only local unseen objects.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we find an object in a packfile index, we make sure we
can still open the packfile itself (or that it is already
open), as it might have been deleted by a simultaneous
repack. If we can't access the packfile, we print a warning
for the user and tell the caller that we don't have the
object (we can then look in other packfiles, or find a loose
version, before giving up).
The warning we print to the user isn't really accomplishing
anything, and it is potentially confusing to users. In the
normal case, it is complete noise; we find the object
elsewhere, and the user does not have to care that we racily
saw a packfile index that became stale. It didn't affect the
operation at all.
A possibly more interesting case is when we later can't find
the object, and report failure to the user. In this case the
warning could be considered a clue toward that ultimate
failure. But it's not really a useful clue in practice. We
wouldn't even print it consistently (since we are racing
with another process, we might not even see the .idx file,
or we might win the race and open the packfile, completing
the operation).
This patch drops the warning entirely (not only from the
fill_pack_entry site, but also from an identical use in
pack-objects). If we did find the warning interesting in the
error case, we could stuff it away and reveal it to the user
when we later die() due to the broken object. But that
complexity just isn't worth it.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Clear the git_zstream variable at the start of git_deflate_init() etc.
so that callers don't have to do that.
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In v2.2.0, we broke "git prune" that runs in a repository that
borrows from an alternate object store.
* jk/prune-mtime:
sha1_file: fix iterating loose alternate objects
for_each_loose_file_in_objdir: take an optional strbuf path
In v2.2.0, we broke "git prune" that runs in a repository that
borrows from an alternate object store.
* jk/prune-mtime:
sha1_file: fix iterating loose alternate objects
for_each_loose_file_in_objdir: take an optional strbuf path
The string in 'base' contains a path suffix to a specific object;
when its value is used, the suffix must either be filled (as in
stat_sha1_file, open_sha1_file, check_and_freshen_nonlocal) or
cleared (as in prepare_packed_git) to avoid junk at the end.
660c889e (sha1_file: add for_each iterators for loose and packed
objects, 2014-10-15) introduced loose_from_alt_odb(), but this did
neither and treated 'base' as a complete path to the "base" object
directory, instead of a pointer to the "base" of the full path
string.
The trailing path after 'base' is still initialized to NUL, hiding
the bug in some common cases. Additionally the descendent
for_each_file_in_obj_subdir() function swallows ENOENT, so an error
only shows if the alternate's path was last filled with a valid
object (where statting /path/to/existing/00/0bjectfile/00 fails).
Signed-off-by: Jonathon Mah <me@JonathonMah.com>
Helped-by: Kyle J. McKay <mackyle@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We feed a root "objdir" path to this iterator function,
which then copies the result into a strbuf, so that it can
repeatedly append the object sub-directories to it. Let's
make it easy for callers to just pass us a strbuf in the
first place.
We leave the original interface as a convenience for callers
who want to just pass a const string like the result of
get_object_directory().
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Before the previous commit, get_pathname returns an array of PATH_MAX
length. Even if git_path() and similar functions does not use the
whole array, git_path() caller can, in theory.
After the commit, get_pathname() may return a buffer that has just
enough room for the returned string and git_path() caller should never
write beyond that.
Make git_path(), mkpath() and git_path_submodule() return a const
buffer to make sure callers do not write in it at all.
This could have been part of the previous commit, but the "const"
conversion is too much distraction from the core changes in path.c.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The new name is more consistent with the names of other
string_list-related functions.
Suggested-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Tighten the logic to decide that an unreachable cruft is
sufficiently old by covering corner cases such as an ancient object
becoming reachable and then going unreachable again, in which case
its retention period should be prolonged.
* jk/prune-mtime: (28 commits)
drop add_object_array_with_mode
revision: remove definition of unused 'add_object' function
pack-objects: double-check options before discarding objects
repack: pack objects mentioned by the index
pack-objects: use argv_array
reachable: use revision machinery's --indexed-objects code
rev-list: add --indexed-objects option
rev-list: document --reflog option
t5516: test pushing a tag of an otherwise unreferenced blob
traverse_commit_list: support pending blobs/trees with paths
make add_object_array_with_context interface more sane
write_sha1_file: freshen existing objects
pack-objects: match prune logic for discarding objects
pack-objects: refactor unpack-unreachable expiration check
prune: keep objects reachable from recent objects
sha1_file: add for_each iterators for loose and packed objects
count-objects: use for_each_loose_file_in_objdir
count-objects: do not use xsize_t when counting object size
prune-packed: use for_each_loose_file_in_objdir
reachable: mark index blobs as SEEN
...
When we try to write a loose object file, we first check
whether that object already exists. If so, we skip the
write as an optimization. However, this can interfere with
prune's strategy of using mtimes to mark files in progress.
For example, if a branch contains a particular tree object
and is deleted, that tree object may become unreachable, and
have an old mtime. If a new operation then tries to write
the same tree, this ends up as a noop; we notice we
already have the object and do nothing. A prune running
simultaneously with this operation will see the object as
old, and may delete it.
We can solve this by "freshening" objects that we avoid
writing by updating their mtime. The algorithm for doing so
is essentially the same as that of has_sha1_file. Therefore
we provide a new (static) interface "check_and_freshen",
which finds and optionally freshens the object. It's trivial
to implement freshening and simple checking by tweaking a
single parameter.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We typically iterate over the reachable objects in a
repository by starting at the tips and walking the graph.
There's no easy way to iterate over all of the objects,
including unreachable ones. Let's provide a way of doing so.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Prune has to walk $GIT_DIR/objects/?? in order to find the
set of loose objects to prune. Other parts of the code
(e.g., count-objects) want to do the same. Let's factor it
out into a reusable for_each-style function.
Note that this is not quite a straight code movement. The
original code had strange behavior when it found a file of
the form "[0-9a-f]{2}/.{38}" that did _not_ contain all hex
digits. It executed a "break" from the loop, meaning that we
stopped pruning in that directory (but still pruned other
directories!). This was probably a bug; we do not want to
process the file as an object, but we should keep going
otherwise (and that is how the new code handles it).
We are also a little more careful with loose object
directories which fail to open. The original code silently
ignored any failures, but the new code will complain about
any problems besides ENOENT.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We check the return value of the callback and stop iterating
if it is non-zero. However, we do not make the non-zero
return value available to the caller, so they have no way of
knowing whether the operation succeeded or not (technically
they can keep their own error flag in the callback data, but
that is unlike our other for_each functions).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The lockfile API and its users have been cleaned up.
* mh/lockfile: (38 commits)
lockfile.h: extract new header file for the functions in lockfile.c
hold_locked_index(): move from lockfile.c to read-cache.c
hold_lock_file_for_append(): restore errno before returning
get_locked_file_path(): new function
lockfile.c: rename static functions
lockfile: rename LOCK_NODEREF to LOCK_NO_DEREF
commit_lock_file_to(): refactor a helper out of commit_lock_file()
trim_last_path_component(): replace last_path_elm()
resolve_symlink(): take a strbuf parameter
resolve_symlink(): use a strbuf for internal scratch space
lockfile: change lock_file::filename into a strbuf
commit_lock_file(): use a strbuf to manage temporary space
try_merge_strategy(): use a statically-allocated lock_file object
try_merge_strategy(): remove redundant lock_file allocation
struct lock_file: declare some fields volatile
lockfile: avoid transitory invalid states
git_config_set_multivar_in_file(): avoid call to rollback_lock_file()
dump_marks(): remove a redundant call to rollback_lock_file()
api-lockfile: document edge cases
commit_lock_file(): rollback lock file on failure to rename
...
When running a required clean filter, we do not have to mmap the
original before feeding the filter. Instead, stream the file
contents directly to the filter and process its output.
* sp/stream-clean-filter:
sha1_file: don't convert off_t to size_t too early to avoid potential die()
convert: stream from fd to required clean filter to reduce used address space
copy_fd(): do not close the input file descriptor
mmap_limit: introduce GIT_MMAP_LIMIT to allow testing expected mmap size
memory_limit: use git_env_ulong() to parse GIT_ALLOC_LIMIT
config.c: add git_env_ulong() to parse environment variable
convert: drop arguments other than 'path' from would_convert_to_git()
Move the interface declaration for the functions in lockfile.c from
cache.h to a new file, lockfile.h. Add #includes where necessary (and
remove some redundant includes of cache.h by files that already
include builtin.h).
Move the documentation of the lock_file state diagram from lockfile.c
to the new header file.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
xsize_t() checks if an off_t argument can be safely converted to
a size_t return value. If the check is executed too early, it could
fail for large files on 32-bit architectures even if the size_t code
path is not taken. Other paths might be able to handle the large file.
Specifically, index_stream_convert_blob() is able to handle a large file
if a filter is configured that returns a small result.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach a few codepaths to punt (instead of dying) when large blobs
that would not fit in core are involved in the operation.
* nd/large-blobs:
diff: shortcut for diff'ing two binary SHA-1 objects
diff --stat: mark any file larger than core.bigfilethreshold binary
diff.c: allow to pass more flags to diff_populate_filespec
sha1_file.c: do not die failing to malloc in unpack_compressed_entry
wrapper.c: introduce gentle xmallocz that does not die()
Reduce the use of fixed sized buffer passed to getcwd() calls
by introducing xgetcwd() helper.
* rs/strbuf-getcwd:
use strbuf_add_absolute_path() to add absolute paths
abspath: convert absolute_path() to strbuf
use xgetcwd() to set $GIT_DIR
use xgetcwd() to get the current directory or die
wrapper: add xgetcwd()
abspath: convert real_path_internal() to strbuf
abspath: use strbuf_getcwd() to remember original working directory
setup: convert setup_git_directory_gently_1 et al. to strbuf
unix-sockets: use strbuf_getcwd()
strbuf: add strbuf_getcwd()
The data is streamed to the filter process anyway. Better avoid mapping
the file if possible. This is especially useful if a clean filter
reduces the size, for example if it computes a sha1 for binary data,
like git media. The file size that the previous implementation could
handle was limited by the available address space; large files for
example could not be handled with (32-bit) msysgit. The new
implementation can filter files of any size as long as the filter output
is small enough.
The new code path is only taken if the filter is required. The filter
consumes data directly from the fd. If it fails, the original data is
not immediately available. The condition can easily be handled as
a fatal error, which is expected for a required filter anyway.
If the filter was not required, the condition would need to be handled
in a different way, like seeking to 0 and reading the data. But this
would require more restructuring of the code and is probably not worth
it. The obvious approach of falling back to reading all data would not
help achieving the main purpose of this patch, which is to handle large
files with limited address space. If reading all data is an option, we
can simply take the old code path right away and mmap the entire file.
The environment variable GIT_MMAP_LIMIT, which has been introduced in
a previous commit is used to test that the expected code path is taken.
A related test that exercises required filters is modified to verify
that the data actually has been modified on its way from the file system
to the object store.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In order to test expectations about mmap in a way similar to testing
expectations about malloc with GIT_ALLOC_LIMIT introduced by
d41489a6 (Add more large blob test cases, 2012-03-07), introduce a
new environment variable GIT_MMAP_LIMIT to limit the largest allowed
mmap length.
xmmap() is modified to check the size of the requested region and
fail it if it is beyond the limit. Together with GIT_ALLOC_LIMIT
tests can now confirm expectations about memory consumption.
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It is only the path that matters in the decision whether to filter
or not. Clarify this by making path the only argument of
would_convert_to_git().
Signed-off-by: Steffen Prohaska <prohaska@zib.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fewer die() gives better control to the caller, provided that the
caller _can_ handle it. And in unpack_compressed_entry() case, it can,
because unpack_compressed_entry() already returns NULL if it fails to
inflate data.
A side effect from this is fsck continues to run when very large blobs
are present (and do not fit in memory).
Noticed-by: Dale R. Worley <worley@alum.mit.edu>
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* kb/perf-trace:
api-trace.txt: add trace API documentation
progress: simplify performance measurement by using getnanotime()
wt-status: simplify performance measurement by using getnanotime()
git: add performance tracing for git's main() function to debug scripts
trace: add trace_performance facility to debug performance issues
trace: add high resolution timer function to debug performance issues
trace: add 'file:line' to all trace output
trace: move code around, in preparation to file:line output
trace: add current timestamp to all trace output
trace: disable additional trace output for unit tests
trace: add infrastructure to augment trace output with additional info
sha1_file: change GIT_TRACE_PACK_ACCESS logging to use trace API
Documentation/git.txt: improve documentation of 'GIT_TRACE*' variables
trace: improve trace performance
trace: remove redundant printf format attribute
trace: consistently name the format parameter
trace: move trace declarations from cache.h to new trace.h
* jk/strip-suffix:
prepare_packed_git_one: refactor duplicate-pack check
verify-pack: use strbuf_strip_suffix
strbuf: implement strbuf_strip_suffix
index-pack: use strip_suffix to avoid magic numbers
use strip_suffix instead of ends_with in simple cases
replace has_extension with ends_with
implement ends_with via strip_suffix
add strip_suffix function
sha1_file: replace PATH_MAX buffer with strbuf in prepare_packed_git_one()
Code to avoid adding the same alternate object store twice was
subtly broken for a long time, but nobody seems to have noticed.
* rs/fix-alt-odb-path-comparison:
sha1_file: avoid overrunning alternate object base string
When adding alternate object directories, we try not to add the
directory of the current repository to avoid cycles. Unfortunately,
that test was broken, since it compared an absolute with a relative
path.
Signed-off-by: Ephrim Khong <dr.khong@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This changes GIT_TRACE_PACK_ACCESS functionality as follows:
* supports the same options as GIT_TRACE (e.g. printing to stderr)
* no longer supports relative paths
* appends to the trace file rather than overwriting
Signed-off-by: Karsten Blees <blees@dcon.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
While checking if a new alternate object database is a duplicate make
sure that old and new base paths have the same length before comparing
them with memcmp. This avoids overrunning the buffer of the existing
entry if the new one is longer and it stops rejecting foobar/ after
foo/ was already added.
Signed-off-by: Rene Scharfe <ls.r@web.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we are reloading the list of packs, we check whether a
particular pack has been loaded. This is slightly tricky,
because we load packs based on the presence of their ".idx"
files, but record the name of the matching ".pack" file.
Therefore we want to compare their bases.
The existing code stripped off ".idx" from a file we found,
then compared that whole base length to strings containing
the ".pack" version. This meant we could end up comparing
bytes past what the ".pack" string contained, if the ".idx"
file name was much longer.
In practice, it worked OK because memcmp would end up seeing
a difference in the two strings and would return early
before hitting the full length. However, memcmp may
sometimes read extra bytes past a difference (e.g., because
it is comparing 64-bit words), or is even free to compare in
reverse order.
Furthermore, our memcmp made no guarantees that we matched
the whole pack name, up to ".pack". So "foo.idx" would match
"foo-bar.pack", which is wrong (but does not typically
happen, because our pack names have a fixed size).
We can fix both issues, avoid magic numbers, and document
that we expect to compare against a string with ".pack" by
using strip_suffix.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
These two are almost the same function, with the exception
that has_extension only matches if there is content before
the suffix. So ends_with(".exe", ".exe") is true, but
has_extension would not be.
This distinction does not matter to any of the callers,
though, and we can just replace uses of has_extension with
ends_with. We prefer the "ends_with" name because it is more
generic, and there is nothing about the function that
requires it to be used for file extensions.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of using strbuf to create a message string in case a path is
too long for our fixed-size buffer, replace that buffer with a strbuf
and thus get rid of the limitation.
Helped-by: Duy Nguyen <pclouds@gmail.com>
Signed-off-by: Rene Scharfe <l.s.r@web.de>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Reworded the error message given upon a failure to open an existing
loose object file due to e.g. permission issues; it was reported as
the object being corrupt, but that is not quite true.
* jk/report-fail-to-read-objects-better:
open_sha1_file: report "most interesting" errno
When we try to open a loose object file, we first attempt to
open in the local object database, and then try any
alternates. This means that the errno value when we return
will be from the last place we looked (and due to the way
the code is structured, simply ENOENT if we do not have have
any alternates).
This can cause confusing error messages, as read_sha1_file
checks for ENOENT when reporting a missing object. If errno
is something else, we report that. If it is ENOENT, but
has_loose_object reports that we have it, then we claim the
object is corrupted. For example:
$ chmod 0 .git/objects/??/*
$ git rev-list --all
fatal: loose object b2d6fab18b92d49eac46dc3c5a0bcafabda20131 (stored in .git/objects/b2/d6fab18b92d49eac46dc3c5a0bcafabda20131) is corrupt
This patch instead keeps track of the "most interesting"
errno we receive during our search. We consider ENOENT to be
the least interesting of all, and otherwise report the first
error found (so problems in the object database take
precedence over ones in alternates). Here it is with this
patch:
$ git rev-list --all
fatal: failed to read object b2d6fab18b92d49eac46dc3c5a0bcafabda20131: Permission denied
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Eradicate mistaken use of "nor" (that is, essentially "nor" used
not in "neither A nor B" ;-)) from in-code comments, command output
strings, and documentations.
* jl/nor-or-nand-and:
code and test: fix misuses of "nor"
comments: fix misuses of "nor"
contrib: fix misuses of "nor"
Documentation: fix misuses of "nor"
Replace open-coded reallocation with ALLOC_GROW() macro.
* dd/use-alloc-grow:
sha1_file.c: use ALLOC_GROW() in pretend_sha1_file()
read-cache.c: use ALLOC_GROW() in add_index_entry()
builtin/mktree.c: use ALLOC_GROW() in append_to_tree()
attr.c: use ALLOC_GROW() in handle_attr_line()
dir.c: use ALLOC_GROW() in create_simplify()
reflog-walk.c: use ALLOC_GROW()
replace_object.c: use ALLOC_GROW() in register_replace_object()
patch-ids.c: use ALLOC_GROW() in add_commit()
diffcore-rename.c: use ALLOC_GROW()
diff.c: use ALLOC_GROW()
commit.c: use ALLOC_GROW() in register_commit_graft()
cache-tree.c: use ALLOC_GROW() in find_subtree()
bundle.c: use ALLOC_GROW() in add_to_ref_list()
builtin/pack-objects.c: use ALLOC_GROW() in check_pbase_path()
Fix a small leak in the delta stack used when resolving a long
delta chain at runtime.
* nd/sha1-file-delta-stack-leakage-fix:
sha1_file: fix delta_stack memory leak in unpack_entry
* mh/object-code-cleanup:
sha1_file.c: document a bunch of functions defined in the file
sha1_file_name(): declare to return a const string
find_pack_entry(): document last_found_pack
replace_object: use struct members instead of an array
Helped-by: He Sun <sunheehnus@gmail.com>
Signed-off-by: Dmitry S. Dolzhenko <dmitrys.dolzhenko@yandex.ru>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Borrow the bitmap index into packfiles from JGit to speed up
enumeration of objects involved in a commit range without having to
fully traverse the history.
* jk/pack-bitmap: (26 commits)
ewah: unconditionally ntohll ewah data
ewah: support platforms that require aligned reads
read-cache: use get_be32 instead of hand-rolled ntoh_l
block-sha1: factor out get_be and put_be wrappers
do not discard revindex when re-preparing packfiles
pack-bitmap: implement optional name_hash cache
t/perf: add tests for pack bitmaps
t: add basic bitmap functionality tests
count-objects: recognize .bitmap in garbage-checking
repack: consider bitmaps when performing repacks
repack: handle optional files created by pack-objects
repack: turn exts array into array-of-struct
repack: stop using magic number for ARRAY_SIZE(exts)
pack-objects: implement bitmap writing
rev-list: add bitmap mode to speed up object lists
pack-objects: use bitmaps when packing objects
pack-objects: split add_object_entry
pack-bitmap: add support for bitmap indexes
documentation: add documentation for the bitmap format
ewah: compressed bitmap implementation
...
Change the return value of sha1_file_name() to (const char *).
(Callers have no business mucking about here.) Change callers
accordingly, deleting a few superfluous temporary variables along the
way.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add a comment at the declaration of last_found_pack and where it is
used in find_pack_entry(). In the latter, separate the cases (1) to
make a place for the new comment and (2) to turn the success case into
affirmative logic.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Reviewed-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This delta_stack array can grow to any length depending on the actual
delta chain, but we forget to free it. Normally it does not matter
because we use small_delta_stack[] from stack and small_delta_stack
can hold 64-delta chains, more than standard --depth=50 in pack-objects.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git clone $origin foo\bar\baz" on Windows failed to create the
leading directories (i.e. a moral-equivalent of "mkdir -p").
* ss/safe-create-leading-dir-with-slash:
safe_create_leading_directories(): on Windows, \ can separate path components
Code clean-up and protection against concurrent write access to the
ref namespace.
* mh/safe-create-leading-directories:
rename_tmp_log(): on SCLD_VANISHED, retry
rename_tmp_log(): limit the number of remote_empty_directories() attempts
rename_tmp_log(): handle a possible mkdir/rmdir race
rename_ref(): extract function rename_tmp_log()
remove_dir_recurse(): handle disappearing files and directories
remove_dir_recurse(): tighten condition for removing unreadable dir
lock_ref_sha1_basic(): if locking fails with ENOENT, retry
lock_ref_sha1_basic(): on SCLD_VANISHED, retry
safe_create_leading_directories(): add new error value SCLD_VANISHED
cmd_init_db(): when creating directories, handle errors conservatively
safe_create_leading_directories(): introduce enum for return values
safe_create_leading_directories(): always restore slash at end of loop
safe_create_leading_directories(): split on first of multiple slashes
safe_create_leading_directories(): rename local variable
safe_create_leading_directories(): add explicit "slash" pointer
safe_create_leading_directories(): reduce scope of local variable
safe_create_leading_directories(): fix format of "if" chaining
When cloning to a directory "C:\foo\bar" from Windows' cmd.exe where
"foo" does not exist yet, Git would throw an error like
fatal: could not create work tree dir 'c:\foo\bar'.: No such file or directory
Fix this by not hard-coding a platform specific directory separator
into safe_create_leading_directories().
This patch, including its entire commit message, is derived from a
patch by Sebastian Schuberth.
Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Sebastian Schuberth <sschuberth@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When an object lookup fails, we re-read the objects/pack
directory to pick up any new packfiles that may have been
created since our last read. We also discard any pack
revindex structs we've allocated.
The discarding is a problem for the pack-bitmap code, which keeps
a pointer to the revindex for the bitmapped pack. After the
discard, the pointer is invalid, and we may read free()d
memory.
Other revindex users do not keep a bare pointer to the
revindex; instead, they always access it through
revindex_for_pack(), which lazily builds the revindex. So
one solution is to teach the pack-bitmap code a similar
trick. It would be slightly less efficient, but probably not
all that noticeable.
However, it turns out this discarding is not actually
necessary. When we call reprepare_packed_git, we do not
throw away our old pack list. We keep the existing entries,
and only add in new ones. So there is no safety problem; we
will still have the pack struct that matches each revindex.
The packfile itself may go away, of course, but we are
already prepared to handle that, and it may happen outside
of reprepare_packed_git anyway.
Throwing away the revindex may save some RAM if the pack
never gets reused (about 12 bytes per object). But it also
wastes some CPU time (to regenerate the index) if the pack
does get reused. It's hard to say which is more valuable,
but in either case, it happens very rarely (only when we
race with a simultaneous repack). Just leaving the revindex
in place is simple and safe both for current and future
code.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Teach "cat-file --batch" to show delta-base object name for a
packed object that is represented as a delta.
* jk/oi-delta-base:
cat-file: provide %(deltabase) batch format
sha1_object_info_extended: provide delta base sha1s
When we figure out how many file descriptors to allocate for
keeping packfiles open, a system with non-working getrlimit() could
cause us to die(), but because we make this call only to get a
rough estimate of how many is available and we do not even attempt
to use up all file descriptors available ourselves, it is nicer to
fall back to a reasonable low value rather than dying.
* jh/rlimit-nofile-fallback:
get_max_fd_limit(): fall back to OPEN_MAX upon getrlimit/sysconf failure
read_sha1_file() that is the workhorse to read the contents given
an object name honoured object replacements, but there is no
corresponding mechanism to sha1_object_info() that is used to
obtain the metainfo (e.g. type & size) about the object, leading
callers to weird inconsistencies.
* cc/replace-object-info:
replace info: rename 'full' to 'long' and clarify in-code symbols
Documentation/git-replace: describe --format option
builtin/replace: unset read_replace_refs
t6050: add tests for listing with --format
builtin/replace: teach listing using short, medium or full formats
sha1_file: perform object replacement in sha1_object_info_extended()
t6050: show that git cat-file --batch fails with replace objects
sha1_object_info_extended(): add an "unsigned flags" parameter
sha1_file.c: add lookup_replace_object_extended() to pass flags
replace_object: don't check read_replace_refs twice
rename READ_SHA1_FILE_REPLACE flag to LOOKUP_REPLACE_OBJECT
Add a new possible error result that can be returned by
safe_create_leading_directories() and
safe_create_leading_directories_const(): SCLD_VANISHED. This value
indicates that a file or directory on the path existed at one point
(either it already existed or the function created it), but then it
disappeared. This probably indicates that another process deleted the
directory while we were working. If SCLD_VANISHED is returned, the
caller might want to retry the function call, as there is a chance
that a new attempt will succeed.
Why doesn't safe_create_leading_directories() do the retrying
internally? Because an empty directory isn't really ever safe until
it holds a file. So even if safe_create_leading_directories() were
absolutely sure that the directory existed before it returned, there
would be no guarantee that the directory still existed when the caller
tried to write something in it.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Instead of returning magic integer values (which a couple of callers
go to the trouble of distinguishing), return values from an enum. Add
a docstring.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Always restore the slash that we scribbled over at the end of the
loop, rather than also fixing it up at each premature exit from the
loop. This makes it harder to forget to do the cleanup as new paths
are added to the code.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If the input path has multiple slashes between path components (e.g.,
"foo//bar"), then the old code was breaking the path at the last
slash, not the first one. So in the above example, the second slash
was overwritten with NUL, resulting in the parent directory being
sought as "foo/".
When stat() is called on "foo/", it fails with ENOTDIR if "foo" exists
but is not a directory. This caused the wrong path to be taken in the
subsequent logic.
So instead, split path components at the first intercomponent slash
rather than the last one.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Rename "pos" to "next_component", because now it always points at the
next component of the path name that has to be processed.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Keep track of the position of the slash character independently of
"pos", thereby making the purpose of each variable clearer and
working towards other upcoming changes.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This makes it more obvious that values of "st" don't persist across
loop iterations.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Count-objects will report any "garbage" files in the packs
directory, including files whose extensions it does not
know (case 1), and files whose matching ".pack" file is
missing (case 2). Without having learned about ".bitmap"
files, the current code reports all such files as garbage
(case 1), even if their pack exists. Instead, they should be
treated as case 2.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
A caller of sha1_object_info_extended technically has enough
information to determine the base sha1 from the results of
the call. It knows the pack, offset, and delta type of the
object, which is sufficient to find the base.
However, the functions to do so are not publicly available,
and the code itself is intimate enough with the pack details
that it should be abstracted away. We could add a public
helper to allow callers to query the delta base separately,
but it is simpler and slightly more efficient to optionally
grab it along with the rest of the object_info data.
For cases where the object is not stored as a delta, we
write the null sha1 into the query field. A careful caller
could check "oi.whence == OI_PACKED && oi.u.packed.is_delta"
before looking at the base sha1, but using the null sha1
provides a simple alternative (and gives a better sanity
check for a non-careful caller than simply returning random
bytes).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On broken systems where RLIMIT_NOFILE is visible by the compliers
but underlying getrlimit() system call does not behave, we used to
simply die() when we are trying to decide how many file descriptors
to allocate for keeping packfiles open. Instead, allow the fallback
codepath to take over when we get such a failure from getrlimit().
The same issue exists with _SC_OPEN_MAX and sysconf(); restructure
the code in a similar way to prepare for a broken sysconf() as well.
Noticed-by: Joey Hess <joey@kitenet.net>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Two processes creating loose objects at the same time could have
failed unnecessarily when the name of their new objects started
with the same byte value, due to a race condition.
* jh/loose-object-dirs-creation-race:
sha1_file.c:create_tmpfile(): Fix race when creating loose object dirs
"git cat-file --batch-check=ok" did not check the existence of the
named object.
* sb/sha1-loose-object-info-check-existence:
sha1_loose_object_info(): do not return success on missing object
sha1_object_info_extended() should perform object replacement
if it is needed.
The simplest way to do that is to make it call
lookup_replace_object_extended().
And now its "unsigned flags" parameter is used as it is passed
to lookup_replace_object_extended().
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This parameter is not used yet, but it will be used to tell
sha1_object_info_extended() if it should perform object
replacement or not.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, there is only one caller to lookup_replace_object()
that can benefit from passing it some flags, but we expect
that there could be more.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The READ_SHA1_FILE_REPLACE flag is more related to using the
lookup_replace_object() function rather than the
read_sha1_file() function.
We also need such a flag to be used with sha1_object_info()
instead of read_sha1_file().
The name LOOKUP_REPLACE_OBJECT is therefore better for this
flag.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git cat-file --batch-check=ok" did not check the existence of the
named object.
* sb/sha1-loose-object-info-check-existence:
sha1_loose_object_info(): do not return success on missing object
When two processes created one loose object file each, which fell
into the same fan-out bucket that previously did not have any
objects, they both tried to do an equivalent of
mkdir .git/objects/$fanout &&
chmod $shared_perm .git/objects/$fanout
before writing into their file .git/objects/$fanout/$remainder,
one of which could have failed unnecessarily when the second
invocation of mkdir found that the directory already has been
created by the first one.
* jh/loose-object-dirs-creation-race:
sha1_file.c:create_tmpfile(): Fix race when creating loose object dirs
In git v1.4.3, we introduced a new loose object format that
encoded some object information outside of the zlib stream.
Ultimately the format was dropped in v1.5.3, but we kept the
reading side around to help people migrate objects. Each
time we open a loose object, we use a heuristic to check
whether it is in the normal loose format, or the
experimental one.
This heuristic is robust in the face of valid data, but it
tends to treat corrupted or garbage data as an experimental
object. With the regular format, we would notice quickly
that zlib's crc does not check out and complain. With the
experimental object, we are likely to extract a nonsensical
object size and try to allocate a huge buffer, resulting in
xmalloc calling "die".
This latter behavior is much worse, for two reasons. One,
git reports an allocation error when the real error is
corruption. And two, the program dies unconditionally, so
you cannot even run fsck (which would otherwise ignore the
broken object and keep going).
We could try to improve the heuristic to err on the side of
normal objects in the face of corruption, but there is
really little point. The experimental format is long-dead,
and was never enabled by default to begin with. We can
instead simply remove it. The only affected repository would
be one that explicitly set core.legacyheaders in 2007, and
then never repacked in the intervening 6 years.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Since 052fe5ea (sha1_loose_object_info: make type lookup optional,
2013-07-12), sha1_loose_object_info() returns happily without
checking if the object in question exists, which is not what the the
caller sha1_object_info_extended() expects; the caller does not even
bother checking the existence of the object itself.
Noticed-by: Sven Brauch <svenbrauch@googlemail.com>
Helped-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
There are cases (e.g. when running concurrent fetches in a repo) where
multiple Git processes concurrently attempt to create loose objects
within the same objects/XX/ dir. The creation of the loose object files
is (AFAICS) safe from races, but the creation of the objects/XX/ dir in
which the loose objects reside is unsafe, for example:
Two concurrent fetches - A and B. As part of its fetch, A needs to store
12aaaaa as a loose object. B, on the other hand, needs to store 12bbbbb
as a loose object. The objects/12 directory does not already exist.
Concurrently, both A and B determine that they need to create the
objects/12 directory (because their first call to git_mkstemp_mode()
within create_tmpfile() fails witn ENOENT). One of them - let's say A -
executes the following mkdir() call before the other. This first call
returns success, and A moves on. When B gets around to calling mkdir(),
it fails with EEXIST, because A won the race. The mkdir() error causes B
to return -1 from create_tmpfile(), which propagates all the way,
resulting in the fetch failing with:
error: unable to create temporary file: File exists
fatal: failed to write object
fatal: unpack-objects failed
Although it's hard to add a testcase reproducing this issue, it's easy
to provoke if we insert a sleep after the
if (mkdir(buffer, 0777) || adjust_shared_perm(buffer))
return -1;
block, and then run two concurrent "git fetch"es against the same repo.
The fix is to simply handle mkdir() failing with EEXIST as a success.
If EEXIST is somehow returned for the wrong reasons (because the relevant
objects/XX is not a directory, or is otherwise unsuitable for object
storage), the following call to adjust_shared_perm(), or ultimately the
retried call to git_mkstemp_mode() will fail, and we end up returning
error from create_tmpfile() in any case.
Note that there are still cases where two users with unsuitable umasks
in a shared repo can end up in two races where one user first wins the
mkdir() race to create an objects/XX/ directory, and then the other user
wins the adjust_shared_perms() race to chmod() that directory, but fails
because it is (transiently, until the first users completes its chmod())
unwriteable to the other user. However, (an equivalent of) this race also
exists before this patch, and is made no worse by this patch.
Signed-off-by: Johan Herland <johan@herland.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit 5b0864070 (sha1_object_info_extended: make type calculation
optional, Jul 12 2013) changed the return value of the
sha1_object_info_extended function to 0/-1 for success/error.
Previously this function returned the object type for success or
-1 for error. But unfortunately the above commit forgot to change
or move the comment above this function that says "returns enum
object_type or negative".
To fix this inconsistency, let's move the comment above the
sha1_object_info function where it is still true.
Signed-off-by: Christian Couder <chriscool@tuxfamily.org>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The `git_open_noatime` helper can be of general interest for other
consumers of git's different on-disk formats.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When an object is not found after checking the packfiles and then
loose object directory, read_sha1_file() re-checks the packfiles to
prevent racing with a concurrent repacker; teach the same logic to
has_sha1_file().
* jk/has-sha1-file-retry-packed:
has_sha1_file: re-check pack directory before giving up
Current code makes pack-objects always do check_pack_crc() in
unpack_entry() even if right after that we find out there's a cached
version and pack access is not needed. Swap two code blocks, search
for cached version first, then check crc.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Handle memory pressure and file descriptor pressure separately when
deciding to release pack windows to honor resource limits.
* bc/unuse-packfile:
Don't close pack fd when free'ing pack windows
sha1_file: introduce close_one_pack() to close packs on fd pressure
When we read a sha1 file, we first look for a packed
version, then a loose version, and then re-check the pack
directory again before concluding that we cannot find it.
This lets us handle a process that is writing to the
repository simultaneously (e.g., receive-pack writing a new
pack followed by a ref update, or git-repack packing
existing loose objects into a new pack).
However, we do not do the same trick with has_sha1_file; we
only check the packed objects once, followed by loose
objects. This means that we might incorrectly report that we
do not have an object, even though we could find it if we
simply re-checked the pack directory.
By itself, this is usually not a big deal. The other process
is running simultaneously, so we may run has_sha1_file
before it writes, anyway. It is a race whether we see the
object or not. However, we may also see other things
the writing process has done (like updating refs); and in
that case, we must be able to also see the new objects.
For example, imagine we are doing a for_each_ref iteration,
and somebody simultaneously pushes. Receive-pack may write
the pack and update a ref after we have examined the
objects/pack directory, but before the iteration gets to the
updated ref. When we do finally see the updated ref,
for_each_ref will call has_sha1_file to check whether the
ref is broken. If has_sha1_file returns the wrong answer, we
erroneously will think that the ref is broken.
For a normal iteration without DO_FOR_EACH_INCLUDE_BROKEN,
this means that the caller does not see the ref at all
(neither the old nor the new value). So not only will we
fail to see the new value of the ref (which is acceptable,
since we are running simultaneously with the writer, and we
might well read the ref before the writer commits its
write), but we will not see the old value either. For
programs that act on reachability like pack-objects or
prune, this can cause data loss, as we may see the objects
referenced by the original ref value as dangling (and either
omit them from the pack, or delete them via prune).
There's no test included here, because the success case is
two processes running simultaneously forever. But you can
replicate the issue with:
# base.sh
# run this in one terminal; it creates and pushes
# repeatedly to a repository
git init parent &&
(cd parent &&
# create a base commit that will trigger us looking at
# the objects/pack directory before we hit the updated ref
echo content >file &&
git add file &&
git commit -m base &&
# set the unpack limit abnormally low, which
# lets us simulate full-size pushes using tiny ones
git config receive.unpackLimit 1
) &&
git clone parent child &&
cd child &&
n=0 &&
while true; do
echo $n >file && git add file && git commit -m $n &&
git push origin HEAD:refs/remotes/child/master &&
n=$(($n + 1))
done
# fsck.sh
# now run this simultaneously in another terminal; it
# repeatedly fscks, looking for us to consider the
# newly-pushed ref broken. We cannot use for-each-ref
# here, as it uses DO_FOR_EACH_INCLUDE_BROKEN, which
# skips the has_sha1_file check (and if it wants
# more information on the object, it will actually read
# the object, which does the proper two-step lookup)
cd parent &&
while true; do
broken=`git fsck 2>&1 | grep remotes/child`
if test -n "$broken"; then
echo $broken
exit 1
fi
done
Without this patch, the fsck loop fails within a few seconds
(and almost instantly if the test repository actually has a
large number of refs). With it, the two can run
indefinitely.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Now that close_one_pack() has been introduced to handle file
descriptor pressure, it is not strictly necessary to close the
pack file descriptor in unuse_one_window() when we're under memory
pressure.
Jeff King provided a justification for leaving the pack file open:
If you close packfile descriptors, you can run into racy situations
where somebody else is repacking and deleting packs, and they go away
while you are trying to access them. If you keep a descriptor open,
you're fine; they last to the end of the process. If you don't, then
they disappear from under you.
For normal object access, this isn't that big a deal; we just rescan
the packs and retry. But if you are packing yourself (e.g., because
you are a pack-objects started by upload-pack for a clone or fetch),
it's much harder to recover (and we print some warnings).
Let's do so (or uh, not do so).
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When the number of open packs exceeds pack_max_fds, unuse_one_window()
is called repeatedly to attempt to release the least-recently-used
pack windows, which, as a side-effect, will also close a pack file
after closing its last open window. If a pack file has been opened,
but no windows have been allocated into it, it will never be selected
by unuse_one_window() and hence its file descriptor will not be
closed. When this happens, git may exceed the number of file
descriptors permitted by the system.
This latter situation can occur in show-ref or receive-pack during ref
advertisement. During ref advertisement, receive-pack will iterate
over every ref in the repository and advertise it to the client after
ensuring that the ref exists in the local repository. If the ref is
located inside a pack, then the pack is opened to ensure that it
exists, but since the object is not actually read from the pack, no
mmap windows are allocated. When the number of open packs exceeds
pack_max_fds, unuse_one_window() will not be able to find any windows to
free and will not be able to close any packs. Once the per-process
file descriptor limit is exceeded, receive-pack will produce a warning,
not an error, for each pack it cannot open, and will then most likely
fail with an error to spawn rev-list or index-pack like:
error: cannot create standard input pipe for rev-list: Too many open files
error: Could not run 'git rev-list'
This may also occur during upload-pack when refs are packed (in the
packed-refs file) and the number of packs that must be opened to
verify that these packed refs exist exceeds the file descriptor
limit. If the refs are loose, then upload-pack will read each ref
from the object database (if the object is in a pack, allocating one
or more mmap windows for it) in order to peel tags and advertise the
underlying object. But when the refs are packed and peeled,
upload-pack will use the peeled sha1 in the packed-refs file and
will not need to read from the pack files, so no mmap windows will
be allocated and just like with receive-pack, unuse_one_window()
will never select these opened packs to close.
When we have file descriptor pressure, we just need to find an open
pack to close. We can leave the existing mmap windows open. If
additional windows need to be mapped into the pack file, it will be
reopened when necessary. If the pack file has been rewritten in the
mean time, open_packed_git_1() should notice when it compares the file
size or the pack's sha1 checksum to what was previously read from the
pack index, and reject it.
Let's introduce a new function close_one_pack() designed specifically
for this purpose to search for and close the least-recently-used pack,
where LRU is defined as (in order of preference):
* pack with oldest mtime and no allocated mmap windows
* pack with the least-recently-used windows, i.e. the pack
with the oldest most-recently-used window, where none of
the windows are in use
* pack with the least-recently-used windows
Signed-off-by: Brandon Casey <drafnel@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If somebody wants to only know on-disk footprint of an object
without having to know its type or payload size, we can bypass a
lot of code to cheaply learn it.
* jk/cat-file-batch-optim:
Fix some sparse warnings
sha1_object_info_extended: pass object_info to helpers
sha1_object_info_extended: make type calculation optional
packed_object_info: make type lookup optional
packed_object_info: hoist delta type resolution to helper
sha1_loose_object_info: make type lookup optional
sha1_object_info_extended: rename "status" to "type"
cat-file: disable object/refname ambiguity check for batch mode
Sparse issues some "Using plain integer as NULL pointer" warnings.
Each warning relates to the use of an '{0}' initialiser expression
in the declaration of an 'struct object_info'. The first field of
this structure has pointer type. Thus, in order to suppress these
warnings, we replace the initialiser expression with '{NULL}'.
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
Acked-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git cat-file --batch-check=<format>" is added, primarily to allow
on-disk footprint of objects in packfiles (often they are a lot
smaller than their true size, when expressed as deltas) to be
reported.
* jk/in-pack-size-measurement:
pack-revindex: radix-sort the revindex
pack-revindex: use unsigned to store number of objects
cat-file: split --batch input lines on whitespace
cat-file: add %(objectsize:disk) format atom
cat-file: add --batch-check=<format>
cat-file: refactor --batch option parsing
cat-file: teach --batch to stream blob objects
t1006: modernize output comparisons
teach sha1_object_info_extended a "disk_size" query
zero-initialize object_info structs
We take in a "struct object_info" which contains pointers to
storage for items the caller cares about. But then rather
than pass the whole object to the low-level loose/packed
helper functions, we pass the individual pointers.
Let's pass the whole struct instead, which will make adding
more items later easier.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Each caller of sha1_object_info_extended sets up an
object_info struct to tell the function which elements of
the object it wants to get. Until now, getting the type of
the object has always been required (and it is returned via
the return type rather than a pointer in object_info).
This can involve actually opening a loose object file to
determine its type, or following delta chains to determine a
packed file's base type. These effects produce a measurable
slow-down when doing a "cat-file --batch-check" that does
not include %(objecttype).
This patch adds a "typep" query to struct object_info, so
that it can be optionally queried just like size and
disk_size. As a result, the return type of the function is
no longer the object type, but rather 0/-1 for success/error.
As there are only three callers total, we just fix up each
caller rather than keep a compatibility wrapper:
1. The simpler sha1_object_info wrapper continues to
always ask for and return the type field.
2. The istream_source function wants to know the type, and
so always asks for it.
3. The cat-file batch code asks for the type only when
%(objecttype) is part of the format string.
On linux.git, the best-of-five for running:
$ git rev-list --objects --all >objects
$ time git cat-file --batch-check='%(objectsize:disk)'
on a fully packed repository goes from:
real 0m8.680s
user 0m8.160s
sys 0m0.512s
to:
real 0m7.205s
user 0m6.580s
sys 0m0.608s
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Currently, packed_object_info can save some work by not
calculating the size or disk_size of the object if the
caller is not interested. However, it always calculates the
true object type, whether the caller cares or not, and only
optionally returns the easy-to-get "representation type".
Let's swap these types. The function will now return the
representation type (or OBJ_BAD on failure), and will only
optionally fill in the true type.
There should be no behavior change yet, as the only caller,
sha1_object_info_extended, will always feed it a type
pointer.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
To calculate the type of a packed object, we must walk down
its delta chain until we hit a true base object with a real
type. Most of the code in packed_object_info is for handling
this case.
Let's hoist it out into a separate helper function, which
will make it easier to make the type-lookup optional in the
future (and keep our indentation level sane).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Until recently, the only items to request from
sha1_object_info_extended were type and size. This meant
that we always had to open a loose object file to determine
one or the other. But with the addition of the disk_size
query, it's possible that we can fulfill the query without
even opening the object file at all. However, since the
function interface always returns the type, we have no way
of knowing whether the caller cares about it or not.
This patch only modified sha1_loose_object_info to make type
lookup optional using an out-parameter, similar to the way
the size is handled (and the return value is "0" or "-1" for
success or error, respectively).
There should be no functional change yet, though, as
sha1_object_info_extended, the only caller, will always ask
for a type.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The value we get from each low-level object_info function
(e.g., loose, packed) is actually the object type (or -1 for
error). Let's explicitly call it "type", which will make
further refactorings easier to read.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Using sha1_object_info_extended, a caller can find out the
type of an object, its size, and information about where it
is stored. In addition to the object's "true" size, it can
also be useful to know the size that the object takes on
disk (e.g., to generate statistics about which refs consume
space).
This patch adds a "disk_sizep" field to "struct object_info",
and fills it in during sha1_object_info_extended if it is
non-NULL.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The sha1_object_info_extended function expects the caller to
provide a "struct object_info" which contains pointers to
"query" items that will be filled in. The purpose of
providing pointers rather than storing the response directly
in the struct is so that callers can choose not to incur the
expense in finding particular fields that they do not care
about.
Right now the only query item is "sizep", and all callers
set it explicitly to choose whether or not to query it; they
can then leave the rest of the struct uninitialized.
However, as we add new query items, each caller will have to
be updated to explicitly turn off the new ones (by setting
them to NULL). Instead, let's teach each caller to
zero-initialize the struct, so that they do not have to
learn about each new query item added.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When we try to load an object from disk and fail, our
general strategy is to see if we can get it from somewhere
else (e.g., a loose object). That lets users fix corruption
problems by copying known-good versions of objects into the
object database.
We already handle the case where we were not able to read
the delta from disk. However, when we find that the delta we
read does not apply, we simply die. This case is harder to
trigger, as corruption in the delta data itself would
trigger a crc error from zlib. However, a corruption that
pointed us at the wrong delta base might cause it.
We can do the same "fail and try to find the object
elsewhere" trick instead of dying. This not only gives us a
chance to recover, but also puts us on code paths that will
alert the user to the problem (with the current message,
they do not even know which sha1 caused the problem).
Note that unlike some other pack corruptions, we do not
recover automatically from this case when doing a repack.
There is nothing apparently wrong with the delta, as it
points to a valid, accessible object, and we realize the
error only when the resulting size does not match up. And in
theory, one could even have a case where the corrupted size
is the same, and the problem would only be noticed by
recomputing the sha1.
We can get around this by recomputing the deltas with
--no-reuse-delta, which our test does (and this is probably
good advice for anyone recovering from pack corruption).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
5f44324 (core: log offset pack data accesses happened - 2011-07-06)
provides a way to observe pack access patterns via a config
switch. Setting an environment variable looks more obvious than a
config var, especially when you just need to _observe_, and more
inline with other tracing knobs we have.
Document it as it may be useful for remote troubleshooting.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
sha1_object_info() returns -1 (OBJ_BAD) if it cannot find the object
for some reason, which suggests that it wants the _caller_ to report
this error. However, part of its work happens in
sha1_loose_object_info, which _does_ report errors itself. This is
doubly strange because:
* packed_object_info(), which is the other half of the duo, does _not_
report this.
* In the event that an object is packed and pruned while
sha1_object_info_extended() goes looking for it, we would
erroneously show the error -- even though the code of the latter
function purports to handle this case gracefully.
* A caller might invoke sha1_object_info() to find the type of an
object even if that object is not known to exist.
Silence this error. The others remain untouched as a corrupt object
is a much more grave error than it merely being absent.
Signed-off-by: Thomas Rast <trast@inf.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the !delta_data error path of unpack_entry(), we run free(base).
This became a window for use-after-free() in abe601b (sha1_file:
remove recursion in unpack_entry, 2013-03-27), as follows:
Before abe601b, we got the 'base' from cache_or_unpack_entry(..., 0);
keep_cache=0 tells it to also remove that entry. So the 'base' is at
this point not cached, and freeing it in the error path is the right
thing.
After abe601b, the structure changed: we use a three-phase approach
where phase 1 finds the innermost base or a base that is already in
the cache. In phase 3 we therefore know that all bases we unpack are
not part of the delta cache yet. (Observe that we pop from the cache
in phase 1, so this is also true for the very first base.) So we make
no further attempts to look up the bases in the cache, and just call
add_delta_base_cache() on every base object we have assembled.
But the !delta_data error path remained unchanged, and now calls
free() on a base that has already been entered in the cache. This
means that there is a use-after-free if we later use the same base
again.
So remove that free(); we are still going to use that data.
Reported-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Thomas Rast <trast@inf.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Attempts to reduce the stack footprint of sha1_object_info()
and unpack_entry() codepaths.
* tr/packed-object-info-wo-recursion:
sha1_file: remove recursion in unpack_entry
Refactor parts of in_delta_base_cache/cache_or_unpack_entry
sha1_file: remove recursion in packed_object_info
Have the streaming interface and other codepaths more carefully
examine for corrupt objects.
* jk/check-corrupt-objects-carefully:
clone: leave repo in place after checkout errors
clone: run check_everything_connected
clone: die on errors from unpack_trees
add tests for cloning corrupted repositories
streaming_write_entry: propagate streaming errors
add test for streaming corrupt blobs
avoid infinite loop in read_istream_loose
read_istream_filtered: propagate read error from upstream
check_sha1_signature: check return value from read_istream
stream_blob_to_fd: detect errors reading from stream
It's possible for read_istream to return an error, in which
case we just end up in an infinite loop (aside from EOF, we
do not even look at the result, but just feed it straight
into our running hash).
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Similar to the recursion in packed_object_info(), this leads to
problems on stack-space-constrained systems in the presence of long
delta chains.
We proceed in three phases:
1. Dig through the delta chain, saving each delta object's offsets and
size on an ad-hoc stack.
2. Unpack the base object at the bottom.
3. Unpack and apply the deltas from the stack.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The delta base cache lookup and test were shared. Refactor them;
we'll need both parts again. Also, we'll use the clearing routine
later.
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If two processes are racing to create the same directory tree, they
will both see that the directory doesn't exist, both try to mkdir(),
and one of them will fail. This is okay, as we only care that the
directory gets created. So, we add a check for EEXIST from mkdir,
and continue when the directory exists, taking the same codepath as
the case where the earlier stat() succeeds and finds a directory.
Signed-off-by: Steven Walter <stevenrwalter@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
packed_object_info() and packed_delta_info() were mutually recursive.
The former would handle ordinary types and defer deltas to the latter;
the latter would use the former to resolve the delta base.
This arrangement, however, leads to trouble with threaded index-pack
and long delta chains on platforms where thread stacks are small, as
happened on OS X (512kB thread stacks by default) with the chromium
repo.
The task of the two functions is not all that hard to describe without
any recursion, however. It proceeds in three steps:
- determine the representation type and size, based on the outermost
object (delta or not)
- follow through the delta chain, if any
- determine the object type from what is found at the end of the delta
chain
The only complication stems from the error recovery. If parsing fails
at any step, we want to mark that object (within the pack) as bad and
try getting the corresponding SHA1 from elsewhere. If that also
fails, we want to repeat this process back up the delta chain until we
find a reasonable solution or conclude that there is no way to
reconstruct the object. (This is conveniently checked by t5303.)
To achieve that within the pack, we keep track of the entire delta
chain in a stack. When things go sour, we process that stack from the
top, marking entries as bad and attempting to re-resolve by sha1. To
avoid excessive malloc(), the stack starts out with a small
stack-allocated array. The choice of 64 is based on the default of
pack.depth, which is 50, in the hope that it covers "most" delta
chains without any need for malloc().
It's much harder to make the actual re-resolving by sha1 nonrecursive,
so we skip that. If you can't afford *that* recursion, your
corruption problems are more serious than your stack size problems.
Reported-by: Stefan Zager <szager@google.com>
Signed-off-by: Thomas Rast <trast@student.ethz.ch>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
prepare_packed_git_one() is modified to allow count-objects to hook a
report function to so we don't need to duplicate the pack searching
logic in count-objects.c. When report_pack_garbage is NULL, the
overhead is insignificant.
The garbage is reported with warning() instead of error() in packed
garbage case because it's not an error to have garbage. Loose garbage
is still reported as errors and will be converted to warnings later.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The current loop does
while (...) {
if (it is not an .idx file)
continue;
process .idx file;
}
and is reordered to
while (...) {
if (it is an .idx file) {
process .idx file;
}
}
This makes it easier to add new extension file processing.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Change link_alt_odb_entries() to take the length of the "alt"
parameter rather than a pointer to the end of the "alt" string. This
is the more common calling convention and simplifies the code a tiny
bit.
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
Change link_alt_odb_entry() to take a NUL-terminated string instead of
(char *, len). Use string_list_split_in_place() rather than inline
code in link_alt_odb_entries().
This approach saves some code and also avoids the (probably harmless)
error of passing a non-NUL-terminated string to is_absolute_path().
Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: Jeff King <peff@peff.net>
Not all platforms have getrlimit(), but there are other ways to see
the maximum number of files that a process can have open. If
getrlimit() is unavailable, fall back to sysconf(_SC_OPEN_MAX) if
available, and use OPEN_MAX from <limits.h>.
Signed-off-by: Joachim Schmitz <jojo@schmitz-digital.de>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The code to avoid mistaken attempt to add the object directory
itself as its own alternate could read beyond end of a string while
comparison.
* hv/link-alt-odb-entry:
link_alt_odb_entry: fix read over array bounds reported by valgrind
pfxlen can be longer than the path in objdir when relative_base
contains the path to gits object directory. Here we are interested
in checking if ent->base[] (the part that corresponds to .git/objects)
is the same string as objdir, and the code NUL-terminated ent->base[]
to
LEADING PATH\0XX/XXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXXX\0
in preparation for these "duplicate check" step (before we return
from the function, the first NUL is turned into '/' so that we can
fill XX when probing for loose objects). All we need to do is to
compare the string with the path to our object directory.
Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When peeking into object stores of submodules, the code forgot that they
might borrow objects from alternate object stores on their own.
By Heiko Voigt
* hv/submodule-alt-odb:
teach add_submodule_odb() to look for alternates
Since we allow to link other object databases when loading a submodules
database we should also load possible alternates.
Signed-off-by: Heiko Voigt <hvoigt@hvoigt.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When write_loose_object() finds that it is unable to
create a temporary file, it complains, for instance:
unable to create temporary sha1 filename : Too many open files
That extra space was supposed to be the name of the file,
and will be an empty string if the git_mkstemps_mode() fails.
The name of the temporary file is unimportant; delete it.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The error handling routines add a newline. Remove
the duplicate ones in error messages.
Signed-off-by: Pete Wyckoff <pw@padd.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Traditionally, all the callers of check_sha1_signature() first
called read_sha1_file() to prepare the whole object data in core,
and called this function. The function is used to revalidate what
we read from the object database actually matches the object name we
used to ask for the data from the object database.
Update the API to allow callers to pass NULL as the object data, and
have the function read and hash the object data using streaming API
to recompute the object name, without having to hold everything in
core at the same time. This is most useful in parse_object() that
parses a blob object, because this caller does not have to keep the
actual blob data around in memory after a "struct blob" is returned.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jk/maint-avoid-streaming-filtered-contents:
do not stream large files to pack when filters are in use
teach dry-run convert_to_git not to require a src buffer
teach convert_to_git a "dry run" mode
* jk/maint-avoid-streaming-filtered-contents:
do not stream large files to pack when filters are in use
teach dry-run convert_to_git not to require a src buffer
teach convert_to_git a "dry run" mode
Because git's object format requires us to specify the
number of bytes in the object in its header, we must know
the size before streaming a blob into the object database.
This is not a problem when adding a regular file, as we can
get the size from stat(). However, when filters are in use
(such as autocrlf, or the ident, filter, or eol
gitattributes), we have no idea what the ultimate size will
be.
The current code just punts on the whole issue and ignores
filter configuration entirely for files larger than
core.bigfilethreshold. This can generate confusing results
if you use filters for large binary files, as the filter
will suddenly stop working as the file goes over a certain
size. Rather than try to handle unknown input sizes with
streaming, this patch just turns off the streaming
optimization when filters are in use.
This has a slight performance regression in a very specific
case: if you have autocrlf on, but no gitattributes, a large
binary file will avoid the streaming code path because we
don't know beforehand whether it will need conversion or
not. But if you are handling large binary files, you should
be marking them as such via attributes (or at least not
using autocrlf, and instead marking your text files as
such). And the flip side is that if you have a large
_non_-binary file, there is a correctness improvement;
before we did not apply the conversion at all.
The first half of the new t1051 script covers these failures
on input. The second half tests the matching output code
paths. These already work correctly, and do not need any
adjustment.
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* nd/find-pack-entry-recent-cache-invalidation:
find_pack_entry(): do not keep packed_git pointer locally
sha1_file.c: move the core logic of find_pack_entry() into fill_pack_entry()
* nd/find-pack-entry-recent-cache-invalidation:
find_pack_entry(): do not keep packed_git pointer locally
sha1_file.c: move the core logic of find_pack_entry() into fill_pack_entry()
Since 3ba7a06552 (A loose object is not corrupt if it
cannot be read due to EMFILE), "git fsck" on a repository with an empty
loose object file complains with the error message
fatal: failed to read object <sha1>: Invalid argument
This comes from a failure of mmap on this empty file, which sets errno to
EINVAL. Instead of calling xmmap on empty file, we display a clean error
message ourselves, and return a NULL pointer. The new message is
error: object file .git/objects/09/<rest-of-sha1> is empty
fatal: loose object <sha1> (stored in .git/objects/09/<rest-of-sha1>) is corrupt
The second line was already there before the regression in 3ba7a06552,
and the first is an additional message, that should help diagnosing the
problem for the user.
Signed-off-by: Matthieu Moy <Matthieu.Moy@imag.fr>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit f7c22cc (always start looking up objects in the last used pack
first - 2007-05-30) introduce a static packed_git* pointer as an
optimization. The kept pointer however may become invalid if
free_pack_by_name() happens to free that particular pack.
Current code base does not access packs after calling
free_pack_by_name() so it should not be a problem. Anyway, move the
pointer out so that free_pack_by_name() can reset it to avoid running
into troubles in future.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The new helper function implements the logic to find the offset for the
object in one pack and fill a pack_entry structure. The next patch will
restructure the loop and will call the helper from two places.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
On Solaris the system headers define the "tmpfile" name, which'll
cause Git compiled with Sun Studio 12 Update 1 to whine about us
redefining the name:
"pack-write.c", line 76: warning: name redefined by pragma redefine_extname declared static: tmpfile (E_PRAGMA_REDEFINE_STATIC)
"sha1_file.c", line 2455: warning: name redefined by pragma redefine_extname declared static: tmpfile (E_PRAGMA_REDEFINE_STATIC)
"fast-import.c", line 858: warning: name redefined by pragma redefine_extname declared static: tmpfile (E_PRAGMA_REDEFINE_STATIC)
"builtin/index-pack.c", line 175: warning: name redefined by pragma redefine_extname declared static: tmpfile (E_PRAGMA_REDEFINE_STATIC)
Just renaming the "tmpfile" variable to "tmp_file" in the relevant
places is the easiest way to fix this.
Signed-off-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jc/stream-to-pack:
bulk-checkin: replace fast-import based implementation
csum-file: introduce sha1file_checkpoint
finish_tmp_packfile(): a helper function
create_tmp_packfile(): a helper function
write_pack_header(): a helper function
Conflicts:
pack.h
* nd/misc-cleanups:
unpack_object_header_buffer(): clear the size field upon error
tree_entry_interesting: make use of local pointer "item"
tree_entry_interesting(): give meaningful names to return values
read_directory_recursive: reduce one indentation level
get_tree_entry(): do not call find_tree_entry() on an empty tree
tree-walk.c: do not leak internal structure in tree_entry_len()
* nd/misc-cleanups:
unpack_object_header_buffer(): clear the size field upon error
tree_entry_interesting: make use of local pointer "item"
tree_entry_interesting(): give meaningful names to return values
read_directory_recursive: reduce one indentation level
get_tree_entry(): do not call find_tree_entry() on an empty tree
tree-walk.c: do not leak internal structure in tree_entry_len()
This extends the earlier approach to stream a large file directly from the
filesystem to its own packfile, and allows "git add" to send large files
directly into a single pack. Older code used to spawn fast-import, but the
new bulk-checkin API replaces it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The callers do not use the returned size when the function says
it did not use any bytes and sets the type to OBJ_BAD, so this
should not matter in practice, but it is a good code hygiene
anyway.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* jk/maint-pack-objects-compete-with-delete:
downgrade "packfile cannot be accessed" errors to warnings
pack-objects: protect against disappearing packs
These can happen if another process simultaneously prunes a
pack. But that is not usually an error condition, because a
properly-running prune should have repacked the object into
a new pack. So we will notice that the pack has disappeared
unexpectedly, print a message, try other packs (possibly
after re-scanning the list of packs), and find it in the new
pack.
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
It's possible that while pack-objects is running, a
simultaneously running prune process might delete a pack
that we are interested in. Because we load the pack indices
early on, we know that the pack contains our item, but by
the time we try to open and map it, it is gone.
Since c715f78, we already protect against this in the normal
object access code path, but pack-objects accesses the packs
at a lower level. In the normal access path, we call
find_pack_entry, which will call find_pack_entry_one on each
pack index, which does the actual lookup. If it gets a hit,
we will actually open and verify the validity of the
matching packfile (using c715f78's is_pack_valid). If we
can't open it, we'll issue a warning and pretend that we
didn't find it, causing us to go on to the next pack (or on
to loose objects).
Furthermore, we will cache the descriptor to the opened
packfile. Which means that later, when we actually try to
access the object, we are likely to still have that packfile
opened, and won't care if it has been unlinked from the
filesystem.
Notice the "likely" above. If there is another pack access
in the interim, and we run out of descriptors, we could
close the pack. And then a later attempt to access the
closed pack could fail (we'll try to re-open it, of course,
but it may have been deleted). In practice, this doesn't
happen because we tend to look up items and then access them
immediately.
Pack-objects does not follow this code path. Instead, it
accesses the packs at a much lower level, using
find_pack_entry_one directly. This means we skip the
is_pack_valid check, and may end up with the name of a
packfile, but no open descriptor.
We can add the same is_pack_valid check here. Unfortunately,
the access patterns of pack-objects are not quite as nice
for keeping lookup and object access together. We look up
each object as we find out about it, and the only later when
writing the packfile do we necessarily access it. Which
means that the opened packfile may be closed in the interim.
In practice, however, adding this check still has value, for
three reasons.
1. If you have a reasonable number of packs and/or a
reasonable file descriptor limit, you can keep all of
your packs open simultaneously. If this is the case,
then the race is impossible to trigger.
2. Even if you can't keep all packs open at once, you
may end up keeping the deleted one open (i.e., you may
get lucky).
3. The race window is shortened. You may notice early that
the pack is gone, and not try to access it. Triggering
the problem without this check means deleting the pack
any time after we read the list of index files, but
before we access the looked-up objects. Triggering it
with this check means deleting the pack means deleting
the pack after we do a lookup (and successfully access
the packfile), but before we access the object. Which
is a smaller window.
Acked-by: Nicolas Pitre <nico@fluxnic.net>
Signed-off-by: Jeff King <peff@peff.net>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When it needs to compare and add an alt object path to the
alt_odb_list, we normalize this path first since comparing normalized
path is easy to get correct result.
Use strbuf to replace some string operations, since it is cleaner and
safer.
Helped-by: Junio C Hamano <gitster@pobox.com>
Signed-off-by: Hui Wang <Hui.Wang@windriver.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>