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

[PATCH v2 19/25] get_packed_ref_cache(): assume "packed-refs" won't change while locked

From
Michael Haggerty <mhagger@alum.mit.edu>
Date
May 22, 2017, 14:17 UTC
Message-ID
<6f28d8349797120fe344d6ed160df2e80a41c7a2.1495460199.git.mhagger@alum.mit.edu>
In-Reply-To
<cover.1495460199.git.mhagger@alum.mit.edu>

If we've got the "packed-refs" file locked, then it can't change; there's no need to keep calling `stat_validity_check()` on it.

Signed-off-by: Michael Haggerty <mhagger@alum.mit.edu>
---
 refs/files-backend.c | 16 +++++++++++-----
 1 file changed, 11 insertions(+), 5 deletions(-)
diff --git a/refs/files-backend.c b/refs/files-backend.c
index 29514392b0..c4bc9550d3 100644
--- a/refs/files-backend.c
+++ b/refs/files-backend.c
@@ -342,13 +342,18 @@ static void files_ref_path(struct files_ref_store *refs,
 
 /*
  * Get the packed_ref_cache for the specified files_ref_store,
- * creating it if necessary.
+ * creating and populating it if it hasn't been read before or if the
+ * file has been changed (according to its `validity` field) since it
+ * was last read. On the other hand, if we hold the lock, then assume
+ * that the file hasn't been changed out from under us, so skip the
+ * extra `stat()` call in `stat_validity_check()`.
  */
 static struct packed_ref_cache *get_packed_ref_cache(struct files_ref_store *refs)
 {
 	const char *packed_refs_file = files_packed_refs_path(refs);
 
 	if (refs->packed &&
+	    !is_lock_file_locked(&refs->packed_refs_lock) &&
 	    !stat_validity_check(&refs->packed->validity, packed_refs_file))
 		clear_packed_ref_cache(refs);
 
@@ -1288,10 +1293,11 @@ static int lock_packed_refs(struct files_ref_store *refs, int flags)
 			    flags, timeout_value) < 0)
 		return -1;
 	/*
-	 * Get the current packed-refs while holding the lock.  If the
-	 * packed-refs file has been modified since we last read it,
-	 * this will automatically invalidate the cache and re-read
-	 * the packed-refs file.
+	 * Get the current packed-refs while holding the lock. It is
+	 * important that we call `get_packed_ref_cache()` before
+	 * setting `packed_ref_cache->lock`, because otherwise the
+	 * former will see that the file is locked and assume that the
+	 * cache can't be stale.
 	 */
 	packed_ref_cache = get_packed_ref_cache(refs);
 	/* Increment the reference count to prevent it from being freed: */
-- 
2.11.0
Previous: Michael HaggertyNext: Michael Haggerty
Message 18 of 31 in “Prepare to separate out a packed_ref_store”
  1. 00/25 Prepare to separate out a packed_ref_storeMichael Haggerty, May 22, 2017
  2. 01/25 t3600: clean up permissions test properlyMichael Haggerty, May 22, 2017
  3. 02/25 refs.h: clarify docstring for the ref_transaction_update()-related fnsMichael Haggerty, May 22, 2017
  4. 03/25 ref_iterator_begin_fn(): fix docstringMichael Haggerty, May 22, 2017
  5. 04/25 files-backend: use `die("BUG: ...")`, not `die("internal error: ...")`Michael Haggerty, May 22, 2017
  6. 05/25 prefix_ref_iterator: don't trim too muchMichael Haggerty, May 22, 2017
  7. 06/25 refs_ref_iterator_begin(): don't check prefixes redundantlyMichael Haggerty, May 22, 2017
  8. 07/25 refs: use `size_t` indexes when iterating over ref transaction updatesMichael Haggerty, May 22, 2017
  9. 08/25 ref_store: take a `msg` parameter when deleting referencesMichael Haggerty, May 22, 2017
  10. 10/25 files-backend: move `lock` member to `files_ref_store`Michael Haggerty, May 22, 2017
  11. 12/25 files_transaction_cleanup(): new helper functionMichael Haggerty, May 22, 2017
  12. 11/25 files_ref_store: put the packed files lock directly in this structMichael Haggerty, May 22, 2017
  13. 09/25 lockfile: add a new method, is_lock_file_locked()Michael Haggerty, May 22, 2017
  14. 14/25 ref_transaction_prepare(): new optional step for reference updatesMichael Haggerty, May 22, 2017
  15. 16/25 ref_update_reject_duplicates(): use `size_t` rather than `int`Michael Haggerty, May 22, 2017
  16. 15/25 ref_update_reject_duplicates(): expose function to whole refs moduleMichael Haggerty, May 22, 2017
  17. 17/25 ref_update_reject_duplicates(): add a sanity checkMichael Haggerty, May 22, 2017
  18. 19/25 get_packed_ref_cache(): assume "packed-refs" won't change while lockedMichael Haggerty, May 22, 2017
  19. 20/25 read_packed_refs(): do more of the work of reading packed refsMichael Haggerty, May 22, 2017
  20. 22/25 refs_ref_iterator_begin(): handle `GIT_REF_PARANOIA`Michael Haggerty, May 22, 2017
  21. 23/25 create_ref_entry(): remove `check_name` optionMichael Haggerty, May 22, 2017
  22. 24/25 ref-filter: limit traversal to prefixMichael Haggerty, May 22, 2017
  23. 25/25 cache_ref_iterator_begin(): avoid priming unneeded directoriesMichael Haggerty, May 22, 2017
  24. Jeff KingMay 23, 2017
  25. Michael HaggertyMay 24, 2017
  26. Junio C HamanoMay 24, 2017
  27. 21/25 read_packed_refs(): report unexpected fopen() failuresMichael Haggerty, May 22, 2017
  28. 13/25 ref_transaction_commit(): check for valid `transaction->state`Michael Haggerty, May 22, 2017
  29. 18/25 should_pack_ref(): new function, extracted from `files_pack_refs()`Michael Haggerty, May 22, 2017
  30. Stefan BellerMay 22, 2017
  31. Junio C HamanoMay 24, 2017

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.