git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v6 01/12] bulk-checkin: rename 'state' variable and separate 'plugged' boolean

From
Nnksingh85@gmail.com <nksingh85@gmail.com>
Date
Apr 5, 2022, 05:20 UTC
Message-ID
<20220405052018.11247-2-neerajsi@microsoft.com>
In-Reply-To
<pull.1134.v5.git.1648616734.gitgitgadget@gmail.com>
From: Neeraj Singh <neerajsi@microsoft.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_packfile', 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.
* Rename finish_bulk_checkin to flush_bulk_checkin_packfile and call it
  unconditionally from unplug_bulk_checkin. Internally it will
  conditionally do a flush if there's any work to do.
* 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.

The net effect of these changes is to make a clear separation between the portion of the bulk-checkin infrastructure that is related to the packfile (nearly all of it at present) and the part that is related to other future optimizations of the ODB.

Signed-off-by: Neeraj Singh <neerajsi@microsoft.com>
---
 bulk-checkin.c | 33 +++++++++++++++++----------------
 1 file changed, 17 insertions(+), 16 deletions(-)
diff --git a/bulk-checkin.c b/bulk-checkin.c
index 6d6c37171c9..88d72178b2c 100644
--- a/bulk-checkin.c
+++ b/bulk-checkin.c
@@ -10,9 +10,9 @@
 #include "packfile.h"
 #include "object-store.h"
 
-static struct bulk_checkin_state {
-	unsigned plugged:1;
+static int bulk_checkin_plugged;
 
+static struct bulk_checkin_packfile {
 	char *pack_tmp_name;
 	struct hashfile *f;
 	off_t offset;
@@ -21,7 +21,7 @@ static struct bulk_checkin_state {
 	struct pack_idx_entry **written;
 	uint32_t alloc_written;
 	uint32_t nr_written;
-} state;
+} bulk_checkin_packfile;
 
 static void finish_tmp_packfile(struct strbuf *basename,
 				const char *pack_tmp_name,
@@ -39,7 +39,7 @@ static void finish_tmp_packfile(struct strbuf *basename,
 	free(idx_tmp_name);
 }
 
-static void finish_bulk_checkin(struct bulk_checkin_state *state)
+static void flush_bulk_checkin_packfile(struct bulk_checkin_packfile *state)
 {
 	unsigned char hash[GIT_MAX_RAWSZ];
 	struct strbuf packname = STRBUF_INIT;
@@ -80,7 +80,7 @@ static void finish_bulk_checkin(struct bulk_checkin_state *state)
 	reprepare_packed_git(the_repository);
 }
 
-static int already_written(struct bulk_checkin_state *state, struct object_id *oid)
+static int already_written(struct bulk_checkin_packfile *state, struct object_id *oid)
 {
 	int i;
 
@@ -112,7 +112,7 @@ static int already_written(struct bulk_checkin_state *state, struct object_id *o
  * status before calling us just in case we ask it to call us again
  * with a new pack.
  */
-static int stream_to_pack(struct bulk_checkin_state *state,
+static int stream_to_pack(struct bulk_checkin_packfile *state,
 			  git_hash_ctx *ctx, off_t *already_hashed_to,
 			  int fd, size_t size, enum object_type type,
 			  const char *path, unsigned flags)
@@ -189,7 +189,7 @@ static int stream_to_pack(struct bulk_checkin_state *state,
 }
 
 /* Lazily create backing packfile for the state */
-static void prepare_to_stream(struct bulk_checkin_state *state,
+static void prepare_to_stream(struct bulk_checkin_packfile *state,
 			      unsigned flags)
 {
 	if (!(flags & HASH_WRITE_OBJECT) || state->f)
@@ -204,7 +204,7 @@ static void prepare_to_stream(struct bulk_checkin_state *state,
 		die_errno("unable to write pack header");
 }
 
-static int deflate_to_pack(struct bulk_checkin_state *state,
+static int deflate_to_pack(struct bulk_checkin_packfile *state,
 			   struct object_id *result_oid,
 			   int fd, size_t size,
 			   enum object_type type, const char *path,
@@ -251,7 +251,7 @@ static int deflate_to_pack(struct bulk_checkin_state *state,
 			BUG("should not happen");
 		hashfile_truncate(state->f, &checkpoint);
 		state->offset = checkpoint.offset;
-		finish_bulk_checkin(state);
+		flush_bulk_checkin_packfile(state);
 		if (lseek(fd, seekback, SEEK_SET) == (off_t) -1)
 			return error("cannot seek back");
 	}
@@ -278,21 +278,22 @@ int index_bulk_checkin(struct object_id *oid,
 		       int fd, size_t size, enum object_type type,
 		       const char *path, unsigned flags)
 {
-	int status = deflate_to_pack(&state, oid, fd, size, type,
+	int status = deflate_to_pack(&bulk_checkin_packfile, oid, fd, size, type,
 				     path, flags);
-	if (!state.plugged)
-		finish_bulk_checkin(&state);
+	if (!bulk_checkin_plugged)
+		flush_bulk_checkin_packfile(&bulk_checkin_packfile);
 	return status;
 }
 
 void plug_bulk_checkin(void)
 {
-	state.plugged = 1;
+	assert(!bulk_checkin_plugged);
+	bulk_checkin_plugged = 1;
 }
 
 void unplug_bulk_checkin(void)
 {
-	state.plugged = 0;
-	if (state.f)
-		finish_bulk_checkin(&state);
+	assert(bulk_checkin_plugged);
+	bulk_checkin_plugged = 0;
+	flush_bulk_checkin_packfile(&bulk_checkin_packfile);
 }
-- 
2.34.1.78.g86e39b8f8d
Previous: Neeraj Singh via GitGitGadgetNext: nksingh85@gmail.com
Message 159 of 175 in “core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objects”
  1. 0/7 core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objectsNeeraj K. Singh via GitGitGadget, Mar 15, 2022
  2. 1/7 bulk-checkin: rename 'state' variable and separate 'plugged' booleanNeeraj Singh via GitGitGadget, Mar 15, 2022
  3. Junio C HamanoMar 16, 2022
  4. Neeraj SinghMar 16, 2022
  5. Junio C HamanoMar 16, 2022
  6. Neeraj SinghMar 16, 2022
  7. Junio C HamanoMar 16, 2022
  8. Neeraj SinghMar 16, 2022
  9. 2/7 core.fsyncmethod: batched disk flushes for loose-objectsNeeraj Singh via GitGitGadget, Mar 15, 2022
  10. Patrick SteinhardtMar 16, 2022
  11. Neeraj SinghMar 16, 2022
  12. Patrick SteinhardtMar 17, 2022
  13. Bagas SanjayaMar 16, 2022
  14. Neeraj SinghMar 16, 2022
  15. 4/7 unpack-objects: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 15, 2022
  16. 3/7 update-index: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 15, 2022
  17. 5/7 core.fsync: use batch mode and sync loose objects by default on WindowsNeeraj Singh via GitGitGadget, Mar 15, 2022
  18. 6/7 core.fsyncmethod: tests for batch modeNeeraj Singh via GitGitGadget, Mar 15, 2022
  19. 7/7 core.fsyncmethod: performance tests for add and stashNeeraj Singh via GitGitGadget, Mar 15, 2022
  20. 0/7 core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objectsNeeraj K. Singh via GitGitGadget, Mar 20, 2022
  21. 1/7 bulk-checkin: rename 'state' variable and separate 'plugged' booleanNeeraj Singh via GitGitGadget, Mar 20, 2022
  22. 2/7 core.fsyncmethod: batched disk flushes for loose-objectsNeeraj Singh via GitGitGadget, Mar 20, 2022
  23. Ævar Arnfjörð BjarmasonMar 21, 2022
  24. Neeraj SinghMar 21, 2022
  25. Ævar Arnfjörð BjarmasonMar 21, 2022
  26. Neeraj SinghMar 21, 2022
  27. Ævar Arnfjörð BjarmasonMar 21, 2022
  28. Neeraj SinghMar 22, 2022
  29. Ævar Arnfjörð BjarmasonMar 22, 2022
  30. Neeraj SinghMar 22, 2022
  31. 0/7 bottom-up ns/batched-fsync & "plugging" in object-file.cÆvar Arnfjörð Bjarmason, Mar 23, 2022
  32. 2/7 unpack-objects: add skeleton HASH_N_OBJECTS{,_{FIRST,LAST}} flagsÆvar Arnfjörð Bjarmason, Mar 23, 2022
  33. 1/7 write-or-die.c: remove unused fsync_component() functionÆvar Arnfjörð Bjarmason, Mar 23, 2022
  34. Neeraj SinghMar 23, 2022
  35. 4/7 update-index: use a utility function for stdin consumptionÆvar Arnfjörð Bjarmason, Mar 23, 2022
  36. 3/7 object-file: pass down unpack-objects.c flags for "bulk" checkinÆvar Arnfjörð Bjarmason, Mar 23, 2022
  37. 5/7 update-index: pass down an "oflags" argumentÆvar Arnfjörð Bjarmason, Mar 23, 2022
  38. 6/7 update-index: rename "buf" to "line"Ævar Arnfjörð Bjarmason, Mar 23, 2022
  39. 7/7 update-index: make use of HASH_N_OBJECTS{,_{FIRST,LAST}} flagsÆvar Arnfjörð Bjarmason, Mar 23, 2022
  40. Neeraj SinghMar 23, 2022
  41. Ævar Arnfjörð BjarmasonMar 23, 2022
  42. Neeraj SinghMar 23, 2022
  43. 0/7 bottom-up ns/batched-fsync & "plugging" in object-file.cÆvar Arnfjörð Bjarmason, Mar 23, 2022
  44. 1/7 unpack-objects: add skeleton HASH_N_OBJECTS{,_{FIRST,LAST}} flagsÆvar Arnfjörð Bjarmason, Mar 23, 2022
  45. Neeraj SinghMar 23, 2022
  46. 2/7 object-file: pass down unpack-objects.c flags for "bulk" checkinÆvar Arnfjörð Bjarmason, Mar 23, 2022
  47. Neeraj SinghMar 23, 2022
  48. 3/7 update-index: pass down skeleton "oflags" argumentÆvar Arnfjörð Bjarmason, Mar 23, 2022
  49. 4/7 update-index: have the index fsync() flush the loose objectsÆvar Arnfjörð Bjarmason, Mar 23, 2022
  50. Neeraj SinghMar 23, 2022
  51. 5/7 add: use WLI_NEED_LOOSE_FSYNC for new "only the index" bulk fsync()Ævar Arnfjörð Bjarmason, Mar 23, 2022
  52. 7/7 fsync docs: add new fsyncMethod.batch.quarantine, elaborate on oldÆvar Arnfjörð Bjarmason, Mar 23, 2022
  53. Neeraj SinghMar 23, 2022
  54. 6/7 fsync docs: update for new syncing semanticsÆvar Arnfjörð Bjarmason, Mar 23, 2022
  55. Junio C HamanoMar 21, 2022
  56. Neeraj SinghMar 21, 2022
  57. Ævar Arnfjörð BjarmasonMar 23, 2022
  58. Neeraj SinghMar 24, 2022
  59. 3/7 update-index: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 20, 2022
  60. Ævar Arnfjörð BjarmasonMar 21, 2022
  61. Neeraj SinghMar 21, 2022
  62. Ævar Arnfjörð BjarmasonMar 21, 2022
  63. Junio C HamanoMar 21, 2022
  64. Neeraj SinghMar 21, 2022
  65. 4/7 unpack-objects: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 20, 2022
  66. Junio C HamanoMar 21, 2022
  67. Neeraj SinghMar 21, 2022
  68. Neeraj SinghMar 22, 2022
  69. 7/7 core.fsyncmethod: performance tests for add and stashNeeraj Singh via GitGitGadget, Mar 20, 2022
  70. 5/7 core.fsync: use batch mode and sync loose objects by default on WindowsNeeraj Singh via GitGitGadget, Mar 20, 2022
  71. 6/7 core.fsyncmethod: tests for batch modeNeeraj Singh via GitGitGadget, Mar 20, 2022
  72. Junio C HamanoMar 21, 2022
  73. Neeraj SinghMar 22, 2022
  74. Junio C HamanoMar 21, 2022
  75. Neeraj SinghMar 21, 2022
  76. Junio C HamanoMar 21, 2022
  77. 00/11 core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objectsNeeraj K. Singh via GitGitGadget, Mar 24, 2022
  78. 01/11 bulk-checkin: rebrand plug/unplug APIs as 'odb transactions'Neeraj Singh via GitGitGadget, Mar 24, 2022
  79. Ævar Arnfjörð BjarmasonMar 24, 2022
  80. Neeraj SinghMar 24, 2022
  81. 02/11 bulk-checkin: rename 'state' variable and separate 'plugged' booleanNeeraj Singh via GitGitGadget, Mar 24, 2022
  82. 03/11 object-file: pass filename to fsync_or_dieNeeraj Singh via GitGitGadget, Mar 24, 2022
  83. 04/11 core.fsyncmethod: batched disk flushes for loose-objectsNeeraj Singh via GitGitGadget, Mar 24, 2022
  84. 05/11 update-index: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 24, 2022
  85. Junio C HamanoMar 24, 2022
  86. Neeraj SinghMar 24, 2022
  87. Junio C HamanoMar 24, 2022
  88. Neeraj SinghMar 24, 2022
  89. 06/11 unpack-objects: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 24, 2022
  90. 07/11 core.fsync: use batch mode and sync loose objects by default on WindowsNeeraj Singh via GitGitGadget, Mar 24, 2022
  91. 08/11 test-lib-functions: add parsing helpers for ls-files and ls-treeNeeraj Singh via GitGitGadget, Mar 24, 2022
  92. 10/11 core.fsyncmethod: performance tests for add and stashNeeraj Singh via GitGitGadget, Mar 24, 2022
  93. 11/11 core.fsyncmethod: correctly camel-case warning messageNeeraj Singh via GitGitGadget, Mar 24, 2022
  94. 09/11 core.fsyncmethod: tests for batch modeNeeraj Singh via GitGitGadget, Mar 24, 2022
  95. Ævar Arnfjörð BjarmasonMar 24, 2022
  96. Neeraj SinghMar 24, 2022
  97. Ævar Arnfjörð BjarmasonMar 26, 2022
  98. Junio C HamanoMar 24, 2022
  99. Neeraj SinghMar 24, 2022
  100. 00/13 core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objectsNeeraj K. Singh via GitGitGadget, Mar 29, 2022
  101. 01/13 bulk-checkin: rename 'state' variable and separate 'plugged' booleanNeeraj Singh via GitGitGadget, Mar 29, 2022
  102. 02/13 bulk-checkin: rebrand plug/unplug APIs as 'odb transactions'Neeraj Singh via GitGitGadget, Mar 29, 2022
  103. 03/13 object-file: pass filename to fsync_or_dieNeeraj Singh via GitGitGadget, Mar 29, 2022
  104. 04/13 core.fsyncmethod: batched disk flushes for loose-objectsNeeraj Singh via GitGitGadget, Mar 29, 2022
  105. 05/13 cache-tree: use ODB transaction around writing a treeNeeraj Singh via GitGitGadget, Mar 29, 2022
  106. 06/13 update-index: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 29, 2022
  107. 07/13 unpack-objects: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 29, 2022
  108. 08/13 core.fsync: use batch mode and sync loose objects by default on WindowsNeeraj Singh via GitGitGadget, Mar 29, 2022
  109. 12/13 core.fsyncmethod: performance tests for add and stashNeeraj Singh via GitGitGadget, Mar 29, 2022
  110. Neeraj SinghMar 29, 2022
  111. 10/13 core.fsyncmethod: tests for batch modeNeeraj Singh via GitGitGadget, Mar 29, 2022
  112. 13/13 core.fsyncmethod: correctly camel-case warning messageNeeraj Singh via GitGitGadget, Mar 29, 2022
  113. 11/13 t/perf: add iteration setup mechanism to perf-libNeeraj Singh via GitGitGadget, Mar 29, 2022
  114. Neeraj SinghMar 29, 2022
  115. Junio C HamanoMar 29, 2022
  116. 09/13 test-lib-functions: add parsing helpers for ls-files and ls-treeNeeraj Singh via GitGitGadget, Mar 29, 2022
  117. Ævar Arnfjörð BjarmasonMar 29, 2022
  118. Neeraj SinghMar 29, 2022
  119. Ævar Arnfjörð BjarmasonMar 29, 2022
  120. Neeraj SinghMar 29, 2022
  121. 00/14 core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objectsNeeraj K. Singh via GitGitGadget, Mar 30, 2022
  122. 02/14 bulk-checkin: rebrand plug/unplug APIs as 'odb transactions'Neeraj Singh via GitGitGadget, Mar 30, 2022
  123. Junio C HamanoMar 30, 2022
  124. Neeraj SinghMar 31, 2022
  125. 01/14 bulk-checkin: rename 'state' variable and separate 'plugged' booleanNeeraj Singh via GitGitGadget, Mar 30, 2022
  126. Junio C HamanoMar 30, 2022
  127. Neeraj SinghMar 30, 2022
  128. Junio C HamanoMar 30, 2022
  129. Neeraj SinghMar 31, 2022
  130. Junio C HamanoMar 31, 2022
  131. Neeraj SinghMar 31, 2022
  132. 03/14 object-file: pass filename to fsync_or_dieNeeraj Singh via GitGitGadget, Mar 30, 2022
  133. Junio C HamanoMar 30, 2022
  134. Neeraj SinghMar 30, 2022
  135. 04/14 core.fsyncmethod: batched disk flushes for loose-objectsNeeraj Singh via GitGitGadget, Mar 30, 2022
  136. Junio C HamanoMar 30, 2022
  137. Neeraj SinghMar 31, 2022
  138. Junio C HamanoMar 31, 2022
  139. Neeraj SinghMar 31, 2022
  140. Junio C HamanoApr 1, 2022
  141. 05/14 cache-tree: use ODB transaction around writing a treeNeeraj Singh via GitGitGadget, Mar 30, 2022
  142. Junio C HamanoMar 30, 2022
  143. Neeraj SinghMar 30, 2022
  144. 07/14 update-index: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 30, 2022
  145. Junio C HamanoMar 30, 2022
  146. Neeraj SinghMar 30, 2022
  147. 09/14 core.fsync: use batch mode and sync loose objects by default on WindowsNeeraj Singh via GitGitGadget, Mar 30, 2022
  148. 10/14 test-lib-functions: add parsing helpers for ls-files and ls-treeNeeraj Singh via GitGitGadget, Mar 30, 2022
  149. 08/14 unpack-objects: use the bulk-checkin infrastructureNeeraj Singh via GitGitGadget, Mar 30, 2022
  150. 06/14 builtin/add: add ODB transaction around add_files_to_cacheNeeraj Singh via GitGitGadget, Mar 30, 2022
  151. Junio C HamanoMar 30, 2022
  152. 11/14 core.fsyncmethod: tests for batch modeNeeraj Singh via GitGitGadget, Mar 30, 2022
  153. Junio C HamanoMar 30, 2022
  154. Neeraj SinghMar 31, 2022
  155. 12/14 t/perf: add iteration setup mechanism to perf-libNeeraj Singh via GitGitGadget, Mar 30, 2022
  156. 13/14 core.fsyncmethod: performance tests for batch modeNeeraj Singh via GitGitGadget, Mar 30, 2022
  157. Neeraj SinghMar 31, 2022
  158. 14/14 core.fsyncmethod: correctly camel-case warning messageNeeraj Singh via GitGitGadget, Mar 30, 2022
  159. 01/12 bulk-checkin: rename 'state' variable and separate 'plugged' booleannksingh85@gmail.com, Apr 5, 2022
  160. 00/12 core.fsyncmethod: add 'batch' mode for faster fsyncing of multiple objectsnksingh85@gmail.com, Apr 5, 2022
  161. Junio C HamanoApr 6, 2022
  162. Junio C HamanoMay 19, 2022
  163. Neeraj SinghMay 19, 2022
  164. Johannes SchindelinMay 24, 2022
  165. 04/12 cache-tree: use ODB transaction around writing a treenksingh85@gmail.com, Apr 5, 2022
  166. 10/12 core.fsyncmethod: tests for batch modenksingh85@gmail.com, Apr 5, 2022
  167. 12/12 core.fsyncmethod: performance tests for batch modenksingh85@gmail.com, Apr 5, 2022
  168. 11/12 t/perf: add iteration setup mechanism to perf-libnksingh85@gmail.com, Apr 5, 2022
  169. 09/12 test-lib-functions: add parsing helpers for ls-files and ls-treenksingh85@gmail.com, Apr 5, 2022
  170. 07/12 unpack-objects: use the bulk-checkin infrastructurenksingh85@gmail.com, Apr 5, 2022
  171. 05/12 builtin/add: add ODB transaction around add_files_to_cachenksingh85@gmail.com, Apr 5, 2022
  172. 08/12 core.fsync: use batch mode and sync loose objects by default on Windowsnksingh85@gmail.com, Apr 5, 2022
  173. 02/12 bulk-checkin: rebrand plug/unplug APIs as 'odb transactions'nksingh85@gmail.com, Apr 5, 2022
  174. 03/12 core.fsyncmethod: batched disk flushes for loose-objectsnksingh85@gmail.com, Apr 5, 2022
  175. 06/12 update-index: use the bulk-checkin infrastructurenksingh85@gmail.com, Apr 5, 2022

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.