The unpack-objects functionality is used by fetch, push, and fast-import
to turn the transfered data into object database entries when there are
fewer objects than the 'unpacklimit' setting.
By enabling an odb-transaction when unpacking objects, we can take advantage
of batched fsyncs.
Here are some performance numbers to justify batch mode for
unpack-objects, collected on a WSL2 Ubuntu VM.
Fsync Mode | Time for 90 objects (ms)
-------------------------------------
Off | 170
On,fsync | 760
On,batch | 230
Note that the default unpackLimit is 100 objects, so there's a 3x
benefit in the worst case. The non-batch mode fsync scales linearly
with the number of objects, so there are significant benefits even with
smaller numbers of objects.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The update-index functionality is used internally by 'git stash push' to
setup the internal stashed commit.
This change enables odb-transactions for update-index infrastructure to
speed up adding new objects to the object database by leveraging the
batch fsync functionality.
There is some risk with this change, since under batch fsync, the object
files will be in a tmp-objdir until update-index is complete, so callers
using the --stdin option will not see them until update-index is done.
This risk is mitigated by not keeping an ODB transaction open around
--stdin processing if in --verbose mode. Without --verbose mode,
a caller feeding update-index via --stdin wouldn't know when
update-index adds an object, event without an ODB transaction.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
The add_files_to_cache function is invoked internally by
builtin/commit.c and builtin/checkout.c for their flags that stage
modified files before doing the larger operation. These commands
can benefit from batched fsyncing.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Take advantage of the odb transaction infrastructure around writing the
cached tree to the object database.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
When adding many objects to a repo with `core.fsync=loose-object`,
the cost of fsync'ing each object file can become prohibitive.
One major source of the cost of fsync is the implied flush of the
hardware writeback cache within the disk drive. This commit introduces
a new `core.fsyncMethod=batch` option that batches up hardware flushes.
It hooks into the bulk-checkin odb-transaction functionality, takes
advantage of tmp-objdir, and uses the writeout-only support code.
When the new mode is enabled, we do the following for each new object:
1a. Create the object in a tmp-objdir.
2a. Issue a pagecache writeback request and wait for it to complete.
At the end of the entire transaction when unplugging bulk checkin:
1b. Issue an fsync against a dummy file to flush the log and hardware
writeback cache, which should by now have seen the tmp-objdir writes.
2b. Rename all of the tmp-objdir files to their final names.
3b. When updating the index and/or refs, we assume that Git will issue
another fsync internal to that operation. This is not the default
today, but the user now has the option of syncing the index and there
is a separate patch series to implement syncing of refs.
On a filesystem with a singular journal that is updated during name
operations (e.g. create, link, rename, etc), such as NTFS, HFS+, or XFS
we would expect the fsync to trigger a journal writeout so that this
sequence is enough to ensure that the user's data is durable by the time
the git command returns. This sequence also ensures that no object files
appear in the main object store unless they are fsync-durable.
Batch mode is only enabled if core.fsync includes loose-objects. If
the legacy core.fsyncObjectFiles setting is enabled, but core.fsync does
not include loose-objects, we will use file-by-file fsyncing.
In step (1a) of the sequence, the tmp-objdir is created lazily to avoid
work if no loose objects are ever added to the ODB. We use a tmp-objdir
to maintain the invariant that no loose-objects are visible in the main
ODB unless they are properly fsync-durable. This is important since
future ODB operations that try to create an object with specific
contents will silently drop the new data if an object with the target
hash exists without checking that the loose-object contents match the
hash. Only a full git-fsck would restore the ODB to a functional state
where dataloss doesn't occur.
In step (1b) of the sequence, we issue a fsync against a dummy file
created specifically for the purpose. This method has a little higher
cost than using one of the input object files, but makes adding new
callers of this mechanism easier, since we don't need to figure out
which object file is "last" or risk sharing violations by caching the fd
of the last object file.
_Performance numbers_:
Linux - Hyper-V VM running Kernel 5.11 (Ubuntu 20.04) on a fast SSD.
Mac - macOS 11.5.1 running on a Mac mini on a 1TB Apple SSD.
Windows - Same host as Linux, a preview version of Windows 11.
Adding 500 files to the repo with 'git add' Times reported in seconds.
object file syncing | Linux | Mac | Windows
--------------------|-------|-------|--------
disabled | 0.06 | 0.35 | 0.61
fsync | 1.88 | 11.18 | 2.47
batch | 0.15 | 0.41 | 1.53
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Make it clearer in the naming and documentation of the plug_bulk_checkin
and unplug_bulk_checkin APIs that they can be thought of as
a "transaction" to optimize operations on the object database. These
transactions may be nested so that subsystems like the cache-tree
writing code can optimize their operations without caring whether the
top-level code has a transaction active.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
This commit prepares for adding batch-fsync to the bulk-checkin
infrastructure.
The bulk-checkin infrastructure is currently used to batch up addition
of large blobs to a packfile. When a blob is larger than
big_file_threshold, we unconditionally add it to a pack. If bulk
checkins are 'plugged', we allow multiple large blobs to be added to a
single pack until we reach the packfile size limit; otherwise, we simply
make a new packfile for each large blob. The 'unplug' call tells us when
the series of blob additions is done so that we can finish the packfiles
and make their objects available to subsequent operations.
Stated another way, bulk-checkin allows callers to define a transaction
that adds multiple objects to the object database, where the object
database can optimize its internal operations within the transaction
boundary.
Batched fsync will fit into bulk-checkin by taking advantage of the
plug/unplug functionality to determine the appropriate time to fsync
and make newly-added objects available in the primary object database.
* Rename 'state' variable to 'bulk_checkin_state', since we will later
be adding 'bulk_fsync_objdir'. This also makes the variable easier to
find in the debugger, since the name is more unique.
* Move the 'plugged' data member of 'bulk_checkin_state' into a separate
static variable. Doing this avoids resetting the variable in
finish_bulk_checkin when zeroing the 'bulk_checkin_state'. As-is, we
seem to unintentionally disable the plugging functionality the first
time a new packfile must be created due to packfile size limits. While
disabling the plugging state only results in suboptimal behavior for
the current code, it would be fatal for the bulk-fsync functionality
later in this patch series.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git worktree list --porcelain" did not c-quote pathnames and lock
reasons with unsafe bytes correctly, which is worked around by
introducing NUL terminated output format with "-z".
* pw/worktree-list-with-z:
worktree: add -z option for list subcommand
Built-in fsmonitor (part 2).
* jh/builtin-fsmonitor-part2: (30 commits)
t7527: test status with untracked-cache and fsmonitor--daemon
fsmonitor: force update index after large responses
fsmonitor--daemon: use a cookie file to sync with file system
fsmonitor--daemon: periodically truncate list of modified files
t/perf/p7519: add fsmonitor--daemon test cases
t/perf/p7519: speed up test on Windows
t/perf/p7519: fix coding style
t/helper/test-chmtime: skip directories on Windows
t/perf: avoid copying builtin fsmonitor files into test repo
t7527: create test for fsmonitor--daemon
t/helper/fsmonitor-client: create IPC client to talk to FSMonitor Daemon
help: include fsmonitor--daemon feature flag in version info
fsmonitor--daemon: implement handle_client callback
compat/fsmonitor/fsm-listen-darwin: implement FSEvent listener on MacOS
compat/fsmonitor/fsm-listen-darwin: add MacOS header files for FSEvent
compat/fsmonitor/fsm-listen-win32: implement FSMonitor backend on Windows
fsmonitor--daemon: create token-based changed path cache
fsmonitor--daemon: define token-ids
fsmonitor--daemon: add pathname classification
fsmonitor--daemon: implement 'start' command
...
Give hint when branch tracking cannot be established because fetch
refspecs from multiple remote repositories overlap.
* tk/ambiguous-fetch-refspec:
tracking branches: add advice to ambiguous refspec error
"git fetch --refetch" learned to fetch everything without telling
the other side what we already have, which is useful when you
cannot trust what you have in the local object store.
* rc/fetch-refetch:
docs: mention --refetch fetch option
fetch: after refetch, encourage auto gc repacking
t5615-partial-clone: add test for fetch --refetch
fetch: add --refetch option
builtin/fetch-pack: add --refetch option
fetch-pack: add refetch
fetch-negotiator: add specific noop initializer
"git am" can read from the standard input when no mailbox is given
on the command line, but the end-user gets no indication when it
happens, making Git appear stuck.
* jc/mailsplit-warn-on-tty:
am/apply: warn if we end up reading patches from terminal
A handful of obvious clean-ups around a topic that is already in
'master'.
* gc/branch-recurse-submodules-fix:
branch.c: simplify advice-and-die sequence
branch: rework comments for future developers
branch: remove negative exit code
branch --set-upstream-to: be consistent when advising
branch: give submodule updating advice before exit
branch: support more tracking modes when recursing
When creating a loose object file, we didn't report the exact
filename of the file we failed to fsync, even though the
information was readily available, which has been corrected.
* ns/fsync-or-die-message-fix:
object-file: pass filename to fsync_or_die
A couple of fix-up to a topic that is now in 'master'.
* ns/core-fsyncmethod:
core.fsyncmethod: correctly camel-case warning message
core.fsync: fix incorrect expression for default configuration
Work around AIX C compiler that does not seem to grok
initialization of a union member of a struct.
* ab/reftable-aix-xlc-12:
reftable: make assignments portable to AIX xlc v12.01
Move more "git submodule update" to C.
* gc/submodule-update-part2:
submodule--helper: remove forward declaration
submodule: move core cmd_update() logic to C
submodule--helper: reduce logic in run_update_procedure()
submodule--helper: teach update_data more options
builtin/submodule--helper.c: rename option struct to "opt"
submodule update: use die_message()
submodule--helper: run update using child process struct
Code clean-up.
* ds/partial-bundle-more:
pack-objects: lazily set up "struct rev_info", don't leak
bundle: output hash information in 'verify'
bundle: move capabilities to end of 'verify'
pack-objects: parse --filter directly into revs.filter
pack-objects: move revs out of get_object_list()
list-objects-filter: remove CL_ARG__FILTER
"git ls-tree" learns "--oid-only" option, similar to "--name-only",
and more generalized "--format" option.
* tl/ls-tree-oid-only:
ls-tree: split up "fast path" callbacks
ls-tree: detect and error on --name-only --name-status
ls-tree: support --object-only option for "git-ls-tree"
ls-tree: introduce "--format" option
cocci: allow padding with `strbuf_addf()`
ls-tree: introduce struct "show_tree_data"
ls-tree: slightly refactor `show_tree()`
ls-tree: fix "--name-only" and "--long" combined use bug
ls-tree: simplify nesting if/else logic in "show_tree()"
ls-tree: rename "retval" to "recurse" in "show_tree()"
ls-tree: use "size_t", not "int" for "struct strbuf"'s "len"
ls-tree: use "enum object_type", not {blob,tree,commit}_type
ls-tree: add missing braces to "else" arms
ls-tree: remove commented-out code
ls-tree tests: add tests for --name-status
"git reflog" command now uses parse-options API to parse its
command line options.
* ab/reflog-parse-options:
reflog: fix 'show' subcommand's argv
reflog [show]: display sensible -h output
reflog: convert to parse_options() API
reflog exists: use parse_options() API
git reflog [expire|delete]: make -h output consistent with SYNOPSIS
reflog: move "usage" variables and use macros
reflog tests: add missing "git reflog exists" tests
reflog: refactor cmd_reflog() to "if" branches
reflog.c: indent argument lists
The error "not tracking: ambiguous information for ref" is raised
when we are evaluating what tracking information to set on a branch,
and find that the ref to be added as tracking branch is mapped
under multiple remotes' fetch refspecs.
This can easily happen when a user copy-pastes a remote definition
in their git config, and forgets to change the tracking path.
Add advice in this situation, explicitly highlighting which remotes
are involved and suggesting how to correct the situation. Also
update a test to explicitly expect that advice.
Signed-off-by: Tao Klerks <tao@klerks.biz>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
In the dwim_branch_start(), when we cannot find an appropriate
upstream, we will die with the same message anyway, whether we
issue an advice message.
Flip the code around a bit and simplify the flow using
advise_if_enabled() function.
Helped-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
For two cases in which we do not explicitly pass --track=<choice>
option down to the submodule--helper subprocess, we have comments
that say "we do not have to pass --track", but in fact we not just
do not have to, but it would be incorrect to pass any --track option
to the subprocess (instead, the correct behaviour is to let the
subprocess figure out what is the appropriate tracking mode to use).
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
* ab/usage-die-message:
config API: use get_error_routine(), not vreportf()
usage.c + gc: add and use a die_message_errno()
gc: return from cmd_gc(), don't call exit()
usage.c API users: use die_message() for error() + exit 128
usage.c API users: use die_message() for "fatal :" + exit 128
usage.c: add a die_message() routine
Add a -z option to be used in conjunction with --porcelain that gives
NUL-terminated output. As 'worktree list --porcelain' does not quote
worktree paths this enables it to handle worktree paths that contain
newlines.
Signed-off-by: Phillip Wood <phillip.wood@dunelm.org.uk>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
We raised the weather balloon to see if we can allow the construct
in 44ba10d6 (revision: use C99 declaration of variable in for()
loop, 2021-11-14), which was shipped as a part of Git v2.35.
Document that fact in the coding guidelines, and more importantly,
give ourselves a deadline to revisit and update.
Let's declare that we will officially adopt the variable declaration
in the initializaiton part of "for ()" statement this winter, unless
we find that a platform we care about does not grok it.
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Update tests around the use of hook scripts.
* ab/hook-tests-updates:
http tests: use "test_hook" for "smart" and "dumb" http tests
proc-receive hook tests: use "test_hook" instead of "write_script"
tests: extend "test_hook" for "rm" and "chmod -x", convert "$HOOK"
tests: use "test_hook" for misc "mkdir -p" and "chmod" cases
tests: change "mkdir -p && write_script" to use "test_hook"
tests: change "cat && chmod +x" to use "test_hook"
gc + p4 tests: use "test_hook", remove sub-shells
fetch+push tests: use "test_hook" and "test_when_finished" pattern
bugreport tests: tighten up "git bugreport -s hooks" test
tests: assume the hooks are disabled by default
http tests: don't rely on "hook/post-update.sample"
hook tests: turn exit code assertions into a loop
test-lib-functions: add and use a "test_hook" wrapper
Tweaks in the command line prompt (in contrib/) code around its
GIT_PS1_SHOWUPSTREAM feature.
* jd/prompt-upstream-mark:
git-prompt: put upstream comments together
git-prompt: make long upstream state indicator consistent
git-prompt: make upstream state indicator location consistent
git-prompt: rename `upstream` to `upstream_type`
Finishing touches to C rewrite of "git add -i" in single-key
interactive mode.
* pw/add-p-single-key:
terminal: restore settings on SIGTSTP
terminal: work around macos poll() bug
terminal: don't assume stdin is /dev/tty
terminal: use flags for save_term()
"git stash" does not allow subcommands it internally runs as its
implementation detail, except for "git reset", to emit messages;
now "git reset" part has also been squelched.
* vd/stash-silence-reset:
reset: show --no-refresh in the short-help
reset: remove 'reset.refresh' config option
reset: remove 'reset.quiet' config option
reset: do not make '--quiet' disable index refresh
stash: make internal resets quiet and refresh index
reset: suppress '--no-refresh' advice if logging is silenced
reset: replace '--quiet' with '--no-refresh' in performance advice
reset: introduce --[no-]refresh option to --mixed
reset: revise index refresh advice
Replace an instance of "exit(-1)" with "exit(1)". We don't use negative
exit codes - they are misleading because Unix machines will coerce them
to 8-bit unsigned values, losing the sign.
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
If we die while trying to fsync a loose object file, pass the actual
filename we're trying to sync. This is likely to be more helpful for a
user trying to diagnose the cause of the failure than the former
'loose object file' string. It also sidesteps any concerns about
translating the die message differently for loose objects versus
something else that has a real path.
Reported-by: Ævar Arnfjörð Bjarmason <avarab@gmail.com>
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git branch --set-upstream-to" behaves differently when advice is
enabled/disabled:
| | error prefix | exit code |
|-----------------+--------------+-----------|
| advice enabled | error: | 1 |
| advice disabled | fatal: | 128 |
Make both cases consistent by using die_message() when advice is
enabled (this was first proposed in [1]).
[1] https://lore.kernel.org/git/211210.86ee6ldwlc.gmgdl@evledraar.gmail.com
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Fix a bug where "hint:" was printed _before_ "fatal:" (instead of the
other way around).
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
"git branch --recurse-submodules" does not propagate "--track=inherit"
or "--no-track" to submodules, which causes submodule branches to use
the wrong tracking mode [1]. To fix this, pass the correct options to
the "submodule--helper create-branch" child process and test for it.
While we are refactoring the same code, replace "--track" with the
synonymous, but more consistent-looking "--track=direct" option
(introduced at the same time as "--track=inherit", d3115660b4 (branch:
add flags and config to inherit tracking, 2021-12-20)).
[1] This bug is partially a timing issue: "branch --recurse-submodules"
was introduced around the same time as "--track=inherit", and even
though I rebased "branch --recurse-submodules" on top of that, I had
neglected to support the new tracking mode. Omitting "--no-track"
was just a plain old mistake, though.
Signed-off-by: Glen Choo <chooglen@google.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Add some global trace2 statistics for the number of fsyncs performed
during the lifetime of a Git process.
These stats are printed as part of trace2_cmd_exit_fl, which is
presumably where we might want to print any other cross-cutting
statistics.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
Commit b9f5d035 (core.fsync: documentation and user-friendly
aggregate options, 2022-03-15) introduced an incorrect value for
FSYNC_COMPONENTS_DEFAULT. We need an AND-NOT rather than OR-NOT.
Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
Signed-off-by: Junio C Hamano <gitster@pobox.com>
