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

Re: [PATCH] Avoid running lstat(2) on the same cache entry.

From
Linus Torvalds <torvalds@linux-foundation.org>
Date
Jan 19, 2008, 19:47 UTC
Message-ID
<alpine.LFD.1.00.0801191133330.2957@woody.linux-foundation.org>
In-Reply-To
<7vfxwu9s2z.fsf@gitster.siamese.dyndns.org>
On Fri, 18 Jan 2008, Junio C Hamano wrote:
> 
> Signed-off-by: Junio C Hamano <gitster@pobox.com>
Acked-by: Linus Torvalds <torvalds@linux-foundation.org>
>  * Again, this is a rebase on top of your "Updated in-memory
>    index".  "echo >>Makefile && git commit -m test Makefile" in
>    the kernel repository does 23k lstat(2) with this patch, 46k
>    without.

Sadly, doing a simple "git commit" without anything else, will still apparently get us 69230 lstat() calls, with or without this patch. So we get 3 lstat() calls per path.

The call trace for the first one is
 - builtin-commit.c: line 786
	prepare_index
	  refresh_index
	    refresh_cache_ent
	      lstat
while the second one is
 - builtin-commit.c: line 793
	prepare_log_message
	  run_status
	    wt_status_print
	      wt_status_print_changed
	        run_diff_files
	          lstat
and the third one is
 - builtin-commit.c: line 795
	run_status
	  wt_status_print
	    wt_status_print_changed
	      run_diff_files
	        lstat
and the reason is that 
 - "run_diff_files()" doesn't honor ce_uptodate
 - ..but even if it did, wt_status_print_changed() does a wt_read_cache() 
   which will invalidate and re-read the index!

The first problem is trivially fixed (appended), the second one is notas trivial.

		Linus
---
 diff-lib.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)
diff --git a/diff-lib.c b/diff-lib.c
index 4a05b02..5d5a950 100644
--- a/diff-lib.c
+++ b/diff-lib.c
@@ -435,6 +435,8 @@ int run_diff_files(struct rev_info *revs, unsigned int option)
 				continue;
 		}
 
+		if (ce_uptodate(ce))
+			continue;
 		if (lstat(ce->name, &st) < 0) {
 			if (errno != ENOENT && errno != ENOTDIR) {
 				perror(ce->name);
Previous: Junio C HamanoNext: Linus Torvalds
Message 4 of 26 in “Updated in-memory index cleanup”
  1. Linus TorvaldsJan 19, 2008
  2. index: be careful when handling long namesJunio C Hamano, Jan 19, 2008
  3. Avoid running lstat(2) on the same cache entry.Junio C Hamano, Jan 19, 2008
  4. Linus TorvaldsJan 19, 2008
  5. Linus TorvaldsJan 19, 2008
  6. Linus TorvaldsJan 20, 2008
  7. Linus TorvaldsJan 20, 2008
  8. Johannes SchindelinJan 20, 2008
  9. Linus TorvaldsJan 20, 2008
  10. Steffen ProhaskaJan 20, 2008
  11. Johannes SchindelinJan 20, 2008
  12. Junio C HamanoJan 21, 2008
  13. Steffen ProhaskaJan 20, 2008
  14. Johannes SchindelinJan 20, 2008
  15. Also use unpack_trees() in do_diff_cache()Johannes Schindelin, Jan 20, 2008
  16. Linus TorvaldsJan 20, 2008
  17. Linus TorvaldsJan 20, 2008
  18. Johannes SchindelinJan 20, 2008
  19. Linus TorvaldsJan 20, 2008
  20. Johannes SchindelinJan 21, 2008
  21. Try to resurrect the handling for 'diff-index -m'Johannes Schindelin, Jan 20, 2008
  22. Linus TorvaldsJan 20, 2008
  23. Brian DowningJan 20, 2008
  24. Steffen ProhaskaJan 20, 2008
  25. Linus TorvaldsJan 20, 2008
  26. Johannes SchindelinJan 20, 2008

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.