The low-level reference transactions used to update references are currently completely opaque to the user. While certainly desirable in most usecases, there are some which might want to hook into the transaction to observe all queued reference updates as well as observing the abortion or commit of a prepared transaction. One such usecase would be to have a set of replicas of a given Git repository, where we perform Git operations on all of the repositories at once and expect the outcome to be the same in all of them. While there exist hooks already for a certain subset of Git commands that could be used to implement a voting mechanism for this, many others currently don't have any mechanism for this. The above scenario is the motivation for the new "reference-transaction" hook that reaches directly into Git's reference transaction mechanism. The hook receives as parameter the current state the transaction was moved to ("prepared", "committed" or "aborted") and gets via its standard input all queued reference updates. While the exit code gets ignored in the "committed" and "aborted" states, a non-zero exit code in the "prepared" state will cause the transaction to be aborted prematurely. Given the usecase described above, a voting mechanism can now be implemented via this hook: as soon as it gets called, it will take all of stdin and use it to cast a vote to a central service. When all replicas of the repository agree, the hook will exit with zero, otherwise it will abort the transaction by returning non-zero. The most important upside is that this will catch _all_ commands writing references at once, allowing to implement strong consistency for reference updates via a single mechanism. In order to test the impact on the case where we don't have any "reference-transaction" hook installed in the repository, this commit introduce two new performance tests for git-update-refs(1). Run against an empty repository, it produces the following results: Test origin/master HEAD -------------------------------------------------------------------- 1400.2: update-ref 2.70(2.10+0.71) 2.71(2.10+0.73) +0.4% 1400.3: update-ref --stdin 0.21(0.09+0.11) 0.21(0.07+0.14) +0.0% The performance test p1400.2 creates, updates and deletes a branch a thousand times, thus averaging runtime of git-update-refs over 3000 invocations. p1400.3 instead calls `git-update-refs --stdin` three times and queues a thousand creations, updates and deletes respectively. As expected, p1400.3 consistently shows no noticeable impact, as for each batch of updates there's a single call to access(3P) for the negative hook lookup. On the other hand, for p1400.2, one can see an impact caused by this patchset. But doing five runs of the performance tests where each one was run with GIT_PERF_REPEAT_COUNT=10, the overhead ranged from -1.5% to +1.1%. These inconsistent performance numbers can be explained by the overhead of spawning 3000 processes. This shows that the overhead of assembling the hook path and executing access(3P) once to check if it's there is mostly outweighed by the operating system's overhead. Signed-off-by: Patrick Steinhardt --- The only change was simplified error handling as proposed by Junio. Thanks for your feedback, the result definitely looks easier to read to me. Range-diff against v3: 1: 1de96b96e3 ! 1: 55c58e9b09 refs: implement reference transaction hook @@ refs.c: int ref_update_reject_duplicates(struct string_list *refnames, +{ + struct child_process proc = CHILD_PROCESS_INIT; + struct strbuf buf = STRBUF_INIT; -+ int saved_errno = 0, ret, i; ++ int ret = 0, i; + + if (hook == &hook_not_found) -+ return 0; ++ return ret; + if (!hook) + hook = find_hook("reference-transaction"); + if (!hook) { + hook = &hook_not_found; -+ return 0; ++ return ret; + } + + argv_array_pushl(&proc.args, hook, state, NULL); @@ refs.c: int ref_update_reject_duplicates(struct string_list *refnames, + + if (write_in_full(proc.in, buf.buf, buf.len) < 0) { + if (errno != EPIPE) -+ saved_errno = errno; ++ ret = -1; + break; + } + } @@ refs.c: int ref_update_reject_duplicates(struct string_list *refnames, + sigchain_pop(SIGPIPE); + strbuf_release(&buf); + -+ ret = finish_command(&proc); -+ if (ret) -+ return ret; -+ -+ return saved_errno; ++ ret |= finish_command(&proc); ++ return ret; +} + int ref_transaction_prepare(struct ref_transaction *transaction, Documentation/githooks.txt | 29 ++++++++ refs.c | 76 ++++++++++++++++++++- t/perf/p1400-update-ref.sh | 32 +++++++++ t/t1416-ref-transaction-hooks.sh | 109 +++++++++++++++++++++++++++++++ 4 files changed, 244 insertions(+), 2 deletions(-) create mode 100755 t/perf/p1400-update-ref.sh create mode 100755 t/t1416-ref-transaction-hooks.sh diff --git a/Documentation/githooks.txt b/Documentation/githooks.txt index 81f2a87e88..642471109f 100644 --- a/Documentation/githooks.txt +++ b/Documentation/githooks.txt @@ -404,6 +404,35 @@ Both standard output and standard error output are forwarded to `git send-pack` on the other end, so you can simply `echo` messages for the user. +ref-transaction +~~~~~~~~~~~~~~~ + +This hook is invoked by any Git command that performs reference +updates. It executes whenever a reference transaction is prepared, +committed or aborted and may thus get called multiple times. + +The hook takes exactly one argument, which is the current state the +given reference transaction is in: + + - "prepared": All reference updates have been queued to the + transaction and references were locked on disk. + + - "committed": The reference transaction was committed and all + references now have their respective new value. + + - "aborted": The reference transaction was aborted, no changes + were performed and the locks have been released. + +For each reference update that was added to the transaction, the hook +receives on standard input a line of the format: + + SP SP LF + +The exit status of the hook is ignored for any state except for the +"prepared" state. In the "prepared" state, a non-zero exit status will +cause the transaction to be aborted. The hook will not be called with +"aborted" state in that case. + push-to-checkout ~~~~~~~~~~~~~~~~ diff --git a/refs.c b/refs.c index 224ff66c7b..f05295f503 100644 --- a/refs.c +++ b/refs.c @@ -9,6 +9,7 @@ #include "iterator.h" #include "refs.h" #include "refs/refs-internal.h" +#include "run-command.h" #include "object-store.h" #include "object.h" #include "tag.h" @@ -16,6 +17,7 @@ #include "worktree.h" #include "argv-array.h" #include "repository.h" +#include "sigchain.h" /* * List of all available backends @@ -1986,10 +1988,65 @@ int ref_update_reject_duplicates(struct string_list *refnames, return 0; } +static const char hook_not_found; +static const char *hook; + +static int run_transaction_hook(struct ref_transaction *transaction, + const char *state) +{ + struct child_process proc = CHILD_PROCESS_INIT; + struct strbuf buf = STRBUF_INIT; + int ret = 0, i; + + if (hook == &hook_not_found) + return ret; + if (!hook) + hook = find_hook("reference-transaction"); + if (!hook) { + hook = &hook_not_found; + return ret; + } + + argv_array_pushl(&proc.args, hook, state, NULL); + proc.in = -1; + proc.stdout_to_stderr = 1; + proc.trace2_hook_name = "reference-transaction"; + + ret = start_command(&proc); + if (ret) + return ret; + + sigchain_push(SIGPIPE, SIG_IGN); + + for (i = 0; i < transaction->nr; i++) { + struct ref_update *update = transaction->updates[i]; + + strbuf_reset(&buf); + strbuf_addf(&buf, "%s %s %s\n", + oid_to_hex(&update->old_oid), + oid_to_hex(&update->new_oid), + update->refname); + + if (write_in_full(proc.in, buf.buf, buf.len) < 0) { + if (errno != EPIPE) + ret = -1; + break; + } + } + + close(proc.in); + sigchain_pop(SIGPIPE); + strbuf_release(&buf); + + ret |= finish_command(&proc); + return ret; +} + int ref_transaction_prepare(struct ref_transaction *transaction, struct strbuf *err) { struct ref_store *refs = transaction->ref_store; + int ret; switch (transaction->state) { case REF_TRANSACTION_OPEN: @@ -2012,7 +2069,17 @@ int ref_transaction_prepare(struct ref_transaction *transaction, return -1; } - return refs->be->transaction_prepare(refs, transaction, err); + ret = refs->be->transaction_prepare(refs, transaction, err); + if (ret) + return ret; + + ret = run_transaction_hook(transaction, "prepared"); + if (ret) { + ref_transaction_abort(transaction, err); + die(_("ref updates aborted by hook")); + } + + return 0; } int ref_transaction_abort(struct ref_transaction *transaction, @@ -2036,6 +2103,8 @@ int ref_transaction_abort(struct ref_transaction *transaction, break; } + run_transaction_hook(transaction, "aborted"); + ref_transaction_free(transaction); return ret; } @@ -2064,7 +2133,10 @@ int ref_transaction_commit(struct ref_transaction *transaction, break; } - return refs->be->transaction_finish(refs, transaction, err); + ret = refs->be->transaction_finish(refs, transaction, err); + if (!ret) + run_transaction_hook(transaction, "committed"); + return ret; } int refs_verify_refname_available(struct ref_store *refs, diff --git a/t/perf/p1400-update-ref.sh b/t/perf/p1400-update-ref.sh new file mode 100755 index 0000000000..d275a81248 --- /dev/null +++ b/t/perf/p1400-update-ref.sh @@ -0,0 +1,32 @@ +#!/bin/sh + +test_description="Tests performance of update-ref" + +. ./perf-lib.sh + +test_perf_fresh_repo + +test_expect_success "setup" ' + test_commit PRE && + test_commit POST && + printf "create refs/heads/%d PRE\n" $(test_seq 1000) >create && + printf "update refs/heads/%d POST PRE\n" $(test_seq 1000) >update && + printf "delete refs/heads/%d POST\n" $(test_seq 1000) >delete +' + +test_perf "update-ref" ' + for i in $(test_seq 1000) + do + git update-ref refs/heads/branch PRE && + git update-ref refs/heads/branch POST PRE && + git update-ref -d refs/heads/branch + done +' + +test_perf "update-ref --stdin" ' + git update-ref --stdin >actual + EOF + cat >expect <<-EOF && + prepared + committed + EOF + git update-ref HEAD POST && + test_cmp expect actual +' + +test_expect_success 'hook aborts updating ref in prepared state' ' + test_when_finished "rm .git/hooks/reference-transaction" && + git reset --hard PRE && + write_script .git/hooks/reference-transaction <<-\EOF && + if test "$1" = prepared + then + exit 1 + fi + EOF + test_must_fail git update-ref HEAD POST 2>err && + test_i18ngrep "ref updates aborted by hook" err +' + +test_expect_success 'hook gets all queued updates in prepared state' ' + test_when_finished "rm .git/hooks/reference-transaction actual" && + git reset --hard PRE && + write_script .git/hooks/reference-transaction <<-\EOF && + if test "$1" = prepared + then + while read -r line + do + printf "%s\n" "$line" + done >actual + fi + EOF + cat >expect <<-EOF && + $ZERO_OID $POST_OID HEAD + $ZERO_OID $POST_OID refs/heads/master + EOF + git update-ref HEAD POST <<-EOF && + update HEAD $ZERO_OID $POST_OID + update refs/heads/master $ZERO_OID $POST_OID + EOF + test_cmp expect actual +' + +test_expect_success 'hook gets all queued updates in committed state' ' + test_when_finished "rm .git/hooks/reference-transaction actual" && + git reset --hard PRE && + write_script .git/hooks/reference-transaction <<-\EOF && + if test "$1" = committed + then + while read -r line + do + printf "%s\n" "$line" + done >actual + fi + EOF + cat >expect <<-EOF && + $ZERO_OID $POST_OID HEAD + $ZERO_OID $POST_OID refs/heads/master + EOF + git update-ref HEAD POST && + test_cmp expect actual +' + +test_expect_success 'hook gets all queued updates in aborted state' ' + test_when_finished "rm .git/hooks/reference-transaction actual" && + git reset --hard PRE && + write_script .git/hooks/reference-transaction <<-\EOF && + if test "$1" = aborted + then + while read -r line + do + printf "%s\n" "$line" + done >actual + fi + EOF + cat >expect <<-EOF && + $ZERO_OID $POST_OID HEAD + $ZERO_OID $POST_OID refs/heads/master + EOF + git update-ref --stdin <<-EOF && + start + update HEAD POST $ZERO_OID + update refs/heads/master POST $ZERO_OID + abort + EOF + test_cmp expect actual +' + +test_done -- 2.27.0