From 4f3d6d026176a48991c4e5ff95c8db583af38148 Mon Sep 17 00:00:00 2001 From: Nipunn Koorapati Date: Wed, 17 Mar 2021 21:22:21 +0000 Subject: [PATCH 1/3] fsmonitor: skip lstat deletion check during git diff-index Teach git to honor fsmonitor rather than issuing an lstat when checking for dirty local deletes. Eliminates O(files) lstats during `git diff HEAD` Signed-off-by: Nipunn Koorapati Signed-off-by: Junio C Hamano --- diff-lib.c | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) diff --git a/diff-lib.c b/diff-lib.c index b73cc1859a..3fb538ad18 100644 --- a/diff-lib.c +++ b/diff-lib.c @@ -30,7 +30,7 @@ */ static int check_removed(const struct cache_entry *ce, struct stat *st) { - if (lstat(ce->name, st) < 0) { + if (!(ce->ce_flags & CE_FSMONITOR_VALID) && lstat(ce->name, st) < 0) { if (!is_missing_file_error(errno)) return -1; return 1; @@ -574,6 +574,7 @@ int run_diff_index(struct rev_info *revs, unsigned int option) struct object_id oid; const char *name; char merge_base_hex[GIT_MAX_HEXSZ + 1]; + struct index_state *istate = revs->diffopt.repo->index; if (revs->pending.nr != 1) BUG("run_diff_index must be passed exactly one tree"); @@ -581,6 +582,8 @@ int run_diff_index(struct rev_info *revs, unsigned int option) trace_performance_enter(); ent = revs->pending.objects; + refresh_fsmonitor(istate); + if (merge_base) { diff_get_merge_base(revs, &oid); name = oid_to_hex_r(merge_base_hex, &oid); From 0ec9949f78e348250f55ba0343344ec8c606be11 Mon Sep 17 00:00:00 2001 From: Nipunn Koorapati Date: Wed, 17 Mar 2021 21:22:22 +0000 Subject: [PATCH 2/3] fsmonitor: add assertion that fsmonitor is valid to check_removed Validate that fsmonitor is valid to futureproof against bugs where check_removed might be called from places that haven't refreshed. Signed-off-by: Nipunn Koorapati Signed-off-by: Junio C Hamano --- diff-lib.c | 18 +++++++++++------- fsmonitor.h | 11 +++++++++++ 2 files changed, 22 insertions(+), 7 deletions(-) diff --git a/diff-lib.c b/diff-lib.c index 3fb538ad18..e5a58c9259 100644 --- a/diff-lib.c +++ b/diff-lib.c @@ -28,8 +28,9 @@ * exists for ce that is a submodule -- it is a submodule that is not * checked out). Return negative for an error. */ -static int check_removed(const struct cache_entry *ce, struct stat *st) +static int check_removed(const struct index_state *istate, const struct cache_entry *ce, struct stat *st) { + assert(is_fsmonitor_refreshed(istate)); if (!(ce->ce_flags & CE_FSMONITOR_VALID) && lstat(ce->name, st) < 0) { if (!is_missing_file_error(errno)) return -1; @@ -136,7 +137,7 @@ int run_diff_files(struct rev_info *revs, unsigned int option) memset(&(dpath->parent[0]), 0, sizeof(struct combine_diff_parent)*5); - changed = check_removed(ce, &st); + changed = check_removed(istate, ce, &st); if (!changed) wt_mode = ce_mode_from_stat(ce, st.st_mode); else { @@ -216,7 +217,7 @@ int run_diff_files(struct rev_info *revs, unsigned int option) } else { struct stat st; - changed = check_removed(ce, &st); + changed = check_removed(istate, ce, &st); if (changed) { if (changed < 0) { perror(ce->name); @@ -278,7 +279,8 @@ static void diff_index_show_file(struct rev_info *revs, oid, oid_valid, ce->name, dirty_submodule); } -static int get_stat_data(const struct cache_entry *ce, +static int get_stat_data(const struct index_state *istate, + const struct cache_entry *ce, const struct object_id **oidp, unsigned int *modep, int cached, int match_missing, @@ -290,7 +292,7 @@ static int get_stat_data(const struct cache_entry *ce, if (!cached && !ce_uptodate(ce)) { int changed; struct stat st; - changed = check_removed(ce, &st); + changed = check_removed(istate, ce, &st); if (changed < 0) return -1; else if (changed) { @@ -321,12 +323,13 @@ static void show_new_file(struct rev_info *revs, const struct object_id *oid; unsigned int mode; unsigned dirty_submodule = 0; + struct index_state *istate = revs->diffopt.repo->index; /* * New file in the index: it might actually be different in * the working tree. */ - if (get_stat_data(new_file, &oid, &mode, cached, match_missing, + if (get_stat_data(istate, new_file, &oid, &mode, cached, match_missing, &dirty_submodule, &revs->diffopt) < 0) return; @@ -342,8 +345,9 @@ static int show_modified(struct rev_info *revs, unsigned int mode, oldmode; const struct object_id *oid; unsigned dirty_submodule = 0; + struct index_state *istate = revs->diffopt.repo->index; - if (get_stat_data(new_entry, &oid, &mode, cached, match_missing, + if (get_stat_data(istate, new_entry, &oid, &mode, cached, match_missing, &dirty_submodule, &revs->diffopt) < 0) { if (report_missing) diff_index_show_file(revs, "-", old_entry, diff --git a/fsmonitor.h b/fsmonitor.h index 7f1794b90b..f20d72631d 100644 --- a/fsmonitor.h +++ b/fsmonitor.h @@ -49,6 +49,17 @@ void refresh_fsmonitor(struct index_state *istate); */ int fsmonitor_is_trivial_response(const struct strbuf *query_result); +/* + * Check if refresh_fsmonitor has been called at least once. + * refresh_fsmonitor is idempotent. Returns true if fsmonitor is + * not enabled (since the state will be "fresh" w/ CE_FSMONITOR_VALID unset) + * This version is useful for assertions + */ +static inline int is_fsmonitor_refreshed(const struct index_state *istate) +{ + return !core_fsmonitor || istate->fsmonitor_has_run_once; +} + /* * Set the given cache entries CE_FSMONITOR_VALID bit. This should be * called any time the cache entry has been updated to reflect the From 7e5aa13d2cb590f7f21f195312ae1cf7de1cdd75 Mon Sep 17 00:00:00 2001 From: Nipunn Koorapati Date: Wed, 17 Mar 2021 21:22:23 +0000 Subject: [PATCH 3/3] fsmonitor: add perf test for git diff HEAD Update the xargs call so that if your large repo contains symlinks, test-tool chmtime failure does not end the script. On Linux Test this tree upstream/master --------------------------------------------------------------------------------------------------------- 7519.4: status (fsmonitor=fsmonitor-watchman) 0.52(0.43+0.10) 0.53(0.49+0.05) +1.9% 7519.5: status -uno (fsmonitor=fsmonitor-watchman) 0.21(0.15+0.07) 0.22(0.13+0.09) +4.8% 7519.6: status -uall (fsmonitor=fsmonitor-watchman) 1.65(0.93+0.71) 1.69(1.03+0.65) +2.4% 7519.7: status (dirty) (fsmonitor=fsmonitor-watchman) 11.99(11.34+1.58) 11.95(11.02+1.79) -0.3% 7519.8: diff (fsmonitor=fsmonitor-watchman) 0.25(0.17+0.26) 0.25(0.18+0.26) +0.0% 7519.9: diff HEAD (fsmonitor=fsmonitor-watchman) 0.39(0.25+0.34) 0.89(0.35+0.74) +128.2% 7519.10: diff -- 0_files (fsmonitor=fsmonitor-watchman) 0.16(0.13+0.04) 0.16(0.12+0.05) +0.0% 7519.11: diff -- 10_files (fsmonitor=fsmonitor-watchman) 0.16(0.12+0.05) 0.16(0.12+0.05) +0.0% 7519.12: diff -- 100_files (fsmonitor=fsmonitor-watchman) 0.16(0.12+0.05) 0.16(0.12+0.05) +0.0% 7519.13: diff -- 1000_files (fsmonitor=fsmonitor-watchman) 0.16(0.11+0.06) 0.16(0.12+0.05) +0.0% 7519.14: diff -- 10000_files (fsmonitor=fsmonitor-watchman) 0.18(0.13+0.06) 0.17(0.10+0.08) -5.6% 7519.15: add (fsmonitor=fsmonitor-watchman) 2.25(1.53+0.68) 2.25(1.47+0.74) +0.0% 7519.18: status (fsmonitor=disabled) 0.88(0.73+1.03) 0.89(0.67+1.08) +1.1% 7519.19: status -uno (fsmonitor=disabled) 0.45(0.43+0.89) 0.45(0.34+0.98) +0.0% 7519.20: status -uall (fsmonitor=disabled) 1.88(1.16+1.58) 1.88(1.22+1.51) +0.0% 7519.21: status (dirty) (fsmonitor=disabled) 7.53(7.05+2.11) 7.53(6.98+2.04) +0.0% 7519.22: diff (fsmonitor=disabled) 0.42(0.37+0.92) 0.42(0.38+0.91) +0.0% 7519.23: diff HEAD (fsmonitor=disabled) 0.44(0.41+0.90) 0.44(0.40+0.91) +0.0% 7519.24: diff -- 0_files (fsmonitor=disabled) 0.13(0.09+0.05) 0.13(0.09+0.05) +0.0% 7519.25: diff -- 10_files (fsmonitor=disabled) 0.13(0.10+0.04) 0.13(0.10+0.04) +0.0% 7519.26: diff -- 100_files (fsmonitor=disabled) 0.13(0.09+0.05) 0.13(0.10+0.04) +0.0% 7519.27: diff -- 1000_files (fsmonitor=disabled) 0.13(0.09+0.06) 0.13(0.09+0.05) +0.0% 7519.28: diff -- 10000_files (fsmonitor=disabled) 0.14(0.11+0.05) 0.14(0.10+0.05) +0.0% 7519.29: add (fsmonitor=disabled) 2.43(1.61+1.64) 2.43(1.69+1.57) +0.0% On linux (2.29.2 vs w/ this patch): nipunn@nipunn-dbx:~/src/server3$ strace -f -c git diff 2>&1 | grep lstat 0.04 0.000063 3 20 6 lstat nipunn@nipunn-dbx:~/src/server3$ strace -f -c git diff HEAD 2>&1 | grep lstat 94.98 5.242262 10 523783 13 lstat nipunn@nipunn-dbx:~/src/server3$ strace -f -c ../git/bin-wrappers/git diff 2>&1 | grep lstat 0.38 0.000032 5 7 3 lstat nipunn@nipunn-dbx:~/src/server3$ strace -f -c ../git/bin-wrappers/git diff HEAD 2>&1 | grep lstat 99.44 0.741892 9 81634 10 lstat On mac (2.29.2 vs w/ this patch): nipunn-mbp:server nipunn$ sudo dtruss -L -f -c git diff 2>&1 | grep "^lstat64 " lstat64 8 nipunn-mbp:server nipunn$ sudo dtruss -L -f -c git diff HEAD 2>&1 | grep "^lstat64 " lstat64 120242 nipunn-mbp:server nipunn$ sudo dtruss -L -f -c ../git/bin-wrappers/git diff 2>&1 | grep "^lstat64 " lstat64 4 nipunn-mbp:server nipunn$ sudo dtruss -L -f -c ../git/bin-wrappers/git diff HEAD 2>&1 | grep "^lstat64 " lstat64 4497 There are still a bunch of lstats - on directories, but not every file. Progress! Signed-off-by: Nipunn Koorapati Signed-off-by: Junio C Hamano --- t/helper/test-chmtime.c | 4 ++-- t/perf/p7519-fsmonitor.sh | 4 ++++ 2 files changed, 6 insertions(+), 2 deletions(-) diff --git a/t/helper/test-chmtime.c b/t/helper/test-chmtime.c index aa22af48c2..524b55ca49 100644 --- a/t/helper/test-chmtime.c +++ b/t/helper/test-chmtime.c @@ -109,9 +109,9 @@ int cmd__chmtime(int argc, const char **argv) uintmax_t mtime; if (stat(argv[i], &sb) < 0) { - fprintf(stderr, "Failed to stat %s: %s\n", + fprintf(stderr, "Failed to stat %s: %s. Skipping\n", argv[i], strerror(errno)); - return 1; + continue; } #ifdef GIT_WINDOWS_NATIVE diff --git a/t/perf/p7519-fsmonitor.sh b/t/perf/p7519-fsmonitor.sh index b657564aed..5eb5044a10 100755 --- a/t/perf/p7519-fsmonitor.sh +++ b/t/perf/p7519-fsmonitor.sh @@ -216,6 +216,10 @@ test_fsmonitor_suite() { git diff ' + test_perf_w_drop_caches "diff HEAD ($DESC)" ' + git diff HEAD + ' + test_perf_w_drop_caches "diff -- 0_files ($DESC)" ' git diff -- 1_file '